diff --git a/parse/dev/-NOQJw5z_KY/images/03a12ae800ef2fbde40a9e6f2edf25ce039ef43dec426231e17d61df9352749c.jpg b/parse/dev/-NOQJw5z_KY/images/03a12ae800ef2fbde40a9e6f2edf25ce039ef43dec426231e17d61df9352749c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7154142c08fec6de510363fa2c9625341ca16cc3 --- /dev/null +++ b/parse/dev/-NOQJw5z_KY/images/03a12ae800ef2fbde40a9e6f2edf25ce039ef43dec426231e17d61df9352749c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ebc41c42a71508478e351bee4d55a2f2ac28210a7ebb8f1aa3180f71f8f12aa +size 75950 diff --git a/parse/dev/-NOQJw5z_KY/images/1453e03317c00f1e9bc6f5afe49742e1dba6439e41cc30c786a7bf00a139aed5.jpg b/parse/dev/-NOQJw5z_KY/images/1453e03317c00f1e9bc6f5afe49742e1dba6439e41cc30c786a7bf00a139aed5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6dc317da9560f040d754a90abc9e932a2eaa6f34 --- /dev/null +++ b/parse/dev/-NOQJw5z_KY/images/1453e03317c00f1e9bc6f5afe49742e1dba6439e41cc30c786a7bf00a139aed5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:92c54ad8ec5973b33ec6968bba5aa773c04c8ae827d6af36da90401d244a82c6 +size 20411 diff --git a/parse/dev/-NOQJw5z_KY/images/3af81a44807e4bd3399bcdb6165f4176b5c2807cbdc33ab478fdffea53c11575.jpg b/parse/dev/-NOQJw5z_KY/images/3af81a44807e4bd3399bcdb6165f4176b5c2807cbdc33ab478fdffea53c11575.jpg new file mode 100644 index 0000000000000000000000000000000000000000..db975d31cd0dab3c9861b22e08805fd3fa2b6003 --- /dev/null +++ b/parse/dev/-NOQJw5z_KY/images/3af81a44807e4bd3399bcdb6165f4176b5c2807cbdc33ab478fdffea53c11575.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d5620c01e17aecd0c1145074e0dd251d687baa0a4d5e05c0f1de9d4419613d2 +size 65788 diff --git a/parse/dev/-NOQJw5z_KY/images/4e331f124119c8758a744a811647c826b4c1af228b5af625adf124b50420c54b.jpg b/parse/dev/-NOQJw5z_KY/images/4e331f124119c8758a744a811647c826b4c1af228b5af625adf124b50420c54b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..701ca241245328379f0abb7f9413caf99dd92fb5 --- /dev/null +++ b/parse/dev/-NOQJw5z_KY/images/4e331f124119c8758a744a811647c826b4c1af228b5af625adf124b50420c54b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b063bfc08638a7ff531217e98a9c51ea25b32944b3793737819267c92327cde +size 21783 diff --git a/parse/dev/-NOQJw5z_KY/images/53710fea8c99c91556580251a05b8374eb54910ded17dc54d98c4bda41723211.jpg b/parse/dev/-NOQJw5z_KY/images/53710fea8c99c91556580251a05b8374eb54910ded17dc54d98c4bda41723211.jpg new file mode 100644 index 0000000000000000000000000000000000000000..88499b42cc1152d33a91495d70cb72b87d7ed9ab --- /dev/null +++ b/parse/dev/-NOQJw5z_KY/images/53710fea8c99c91556580251a05b8374eb54910ded17dc54d98c4bda41723211.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bfceccf1027bef0b56f29da07763ba933d78e4b41f93b2d2f9f1d699091470e9 +size 134730 diff --git a/parse/dev/-NOQJw5z_KY/images/6123e09bf595a2d3cb6b5ebf16f1253af7de97be3c9d91a1f5a1a8dd867ec37a.jpg b/parse/dev/-NOQJw5z_KY/images/6123e09bf595a2d3cb6b5ebf16f1253af7de97be3c9d91a1f5a1a8dd867ec37a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ac6b86c3a68d3b4aab0a0ba0d9d53048b26f0275 --- /dev/null +++ b/parse/dev/-NOQJw5z_KY/images/6123e09bf595a2d3cb6b5ebf16f1253af7de97be3c9d91a1f5a1a8dd867ec37a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4eeb465cd39746e48d07f3fddaa83804e86b5fda7a8ca05076e9e6f6aad1a85d +size 24247 diff --git a/parse/dev/-NOQJw5z_KY/images/68b5a5d5a2ff8ba9bc909d7cb8de829c43193dd1ad18d2dd977b30d8cfb87bac.jpg b/parse/dev/-NOQJw5z_KY/images/68b5a5d5a2ff8ba9bc909d7cb8de829c43193dd1ad18d2dd977b30d8cfb87bac.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ed2f5a95e5485da448ef2cc5e5bae9e2959dad14 --- /dev/null +++ b/parse/dev/-NOQJw5z_KY/images/68b5a5d5a2ff8ba9bc909d7cb8de829c43193dd1ad18d2dd977b30d8cfb87bac.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:655f86864f26093169ea47ba02d9fdb58c87331374ae7e3f98cea47612b2f5df +size 30067 diff --git a/parse/dev/-NOQJw5z_KY/images/885a1efefd0fd983dcbbc79e2507cb34d9c2ec38adcba084a69fc31b5c336be5.jpg b/parse/dev/-NOQJw5z_KY/images/885a1efefd0fd983dcbbc79e2507cb34d9c2ec38adcba084a69fc31b5c336be5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d11858f5732bd5c16df140644318e881535f2bf8 --- /dev/null +++ b/parse/dev/-NOQJw5z_KY/images/885a1efefd0fd983dcbbc79e2507cb34d9c2ec38adcba084a69fc31b5c336be5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eee20bd0c6f4854c7f6d6c925bec6676d674b8dcce58550a1a634ea41c03804d +size 26637 diff --git a/parse/dev/-NOQJw5z_KY/images/92626be4d32e49ea3cb2d5fd609d0e4efd4bace02c6c7dd642547205afd2be97.jpg b/parse/dev/-NOQJw5z_KY/images/92626be4d32e49ea3cb2d5fd609d0e4efd4bace02c6c7dd642547205afd2be97.jpg new file mode 100644 index 0000000000000000000000000000000000000000..acaa066813a5155582b237043d129bd68ce59e0b --- /dev/null +++ b/parse/dev/-NOQJw5z_KY/images/92626be4d32e49ea3cb2d5fd609d0e4efd4bace02c6c7dd642547205afd2be97.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d2c55e7484961e1418229a76c4fbfe31cc5bdb4223d1f95423b56813641c5ad4 +size 29962 diff --git a/parse/dev/0IywQ8uxJx/images/0d7e8be689840459aa1ffe8642b06bbea77c3f7ad6f1af46e1567c0e0253e831.jpg b/parse/dev/0IywQ8uxJx/images/0d7e8be689840459aa1ffe8642b06bbea77c3f7ad6f1af46e1567c0e0253e831.jpg new file mode 100644 index 0000000000000000000000000000000000000000..322130aa551b176ac6f190efaa7cf2a664b93e95 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/0d7e8be689840459aa1ffe8642b06bbea77c3f7ad6f1af46e1567c0e0253e831.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b80a42c998ce09347b024c1bf43a9297e1d220c658451cbef86cd52efd46ba07 +size 28216 diff --git a/parse/dev/0IywQ8uxJx/images/18bee69209d99b387d7f595c3f1ccea11cf1602ea5b0705c9c6f474fdf124c4a.jpg b/parse/dev/0IywQ8uxJx/images/18bee69209d99b387d7f595c3f1ccea11cf1602ea5b0705c9c6f474fdf124c4a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a6f652639246aeb3ded849ba22e86a55e98cb0ba --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/18bee69209d99b387d7f595c3f1ccea11cf1602ea5b0705c9c6f474fdf124c4a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6afac73593d112aef2d1452524c6b1cb99f503c2cc76a9563a3db8a0935c7a71 +size 4985 diff --git a/parse/dev/0IywQ8uxJx/images/1b33f9de7819308f2db0d2637e21957ff23731887d18d39d43f6f214a994fe51.jpg b/parse/dev/0IywQ8uxJx/images/1b33f9de7819308f2db0d2637e21957ff23731887d18d39d43f6f214a994fe51.jpg new file mode 100644 index 0000000000000000000000000000000000000000..119c11981f586315ab36473b688f4dbd6daae672 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/1b33f9de7819308f2db0d2637e21957ff23731887d18d39d43f6f214a994fe51.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f1145349566cc900e427db9970f89a56372f478ed2d60efef6067b7048c3c8e +size 4947 diff --git a/parse/dev/0IywQ8uxJx/images/1d5af9037f1cddaa07ba89180595c72ad99320e923035dc07c3cbd9acfcfd9a4.jpg b/parse/dev/0IywQ8uxJx/images/1d5af9037f1cddaa07ba89180595c72ad99320e923035dc07c3cbd9acfcfd9a4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..85aa417b4072c4662e32c42bb9f679b7f59b43d6 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/1d5af9037f1cddaa07ba89180595c72ad99320e923035dc07c3cbd9acfcfd9a4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b32ef680da874040f052d6496da8751d9e24f921662ecddeebc1a0654e4cc2d +size 10842 diff --git a/parse/dev/0IywQ8uxJx/images/1dfe2211c43f6bd57d2f04ef4d531f6cc7e248648a7e9ad5ace4cee301eb712f.jpg b/parse/dev/0IywQ8uxJx/images/1dfe2211c43f6bd57d2f04ef4d531f6cc7e248648a7e9ad5ace4cee301eb712f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..47c36faf51abaa376b6530033db7027a5cd30ca2 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/1dfe2211c43f6bd57d2f04ef4d531f6cc7e248648a7e9ad5ace4cee301eb712f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8368860c3bb589ccee6e5b5e1f60809ce05993a08c9a8fdd02f41a2d8f959563 +size 2943 diff --git a/parse/dev/0IywQ8uxJx/images/2322b75518b1be37fc38d1cc5a4ad06c8ac11d7d9e54132cd51c0bc2bfd0a064.jpg b/parse/dev/0IywQ8uxJx/images/2322b75518b1be37fc38d1cc5a4ad06c8ac11d7d9e54132cd51c0bc2bfd0a064.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bf046d073e12e8a314ce0db0142daaeb8b874bf5 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/2322b75518b1be37fc38d1cc5a4ad06c8ac11d7d9e54132cd51c0bc2bfd0a064.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e4fd6b635efeab398948bfddd2112ee246069da02ccbb5438f03a5a5cadbbcf +size 7996 diff --git a/parse/dev/0IywQ8uxJx/images/3c708d5cb179be13684860a8c4a9bf890cefc966715d9fa13ddba06d31898cbd.jpg b/parse/dev/0IywQ8uxJx/images/3c708d5cb179be13684860a8c4a9bf890cefc966715d9fa13ddba06d31898cbd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..871205df2f328b98b58457f0b48af85f7807cf05 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/3c708d5cb179be13684860a8c4a9bf890cefc966715d9fa13ddba06d31898cbd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcef64383d55106c3002faa9b40e6218d7ec3e0683ceb54c4fbf60dc673dbe31 +size 7922 diff --git a/parse/dev/0IywQ8uxJx/images/5dab6a26b3d663f87b14576a7c986d2a48e21b018e4ec15ae7db5f5dd45501d5.jpg b/parse/dev/0IywQ8uxJx/images/5dab6a26b3d663f87b14576a7c986d2a48e21b018e4ec15ae7db5f5dd45501d5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..db832526d5d9d67c13aa67bf72de083b1ccc949f --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/5dab6a26b3d663f87b14576a7c986d2a48e21b018e4ec15ae7db5f5dd45501d5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4db18e80f07f371ddf071dcf97bacf9cf04b58a78162e1a109884b14953a22c1 +size 9237 diff --git a/parse/dev/0IywQ8uxJx/images/6d7bd5d53e6a6296ce7cc8e1dbc9f1baa3643924c5ed706ca4adc890421b6550.jpg b/parse/dev/0IywQ8uxJx/images/6d7bd5d53e6a6296ce7cc8e1dbc9f1baa3643924c5ed706ca4adc890421b6550.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5e4524502181676a32b228408328b6c777d54aee --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/6d7bd5d53e6a6296ce7cc8e1dbc9f1baa3643924c5ed706ca4adc890421b6550.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb155fbf76dc53aeeb75d3f830bdffe872c803fa4e582f318a12aef9e99a5193 +size 9802 diff --git a/parse/dev/0IywQ8uxJx/images/8f3f6fc69b739a5dcdc95e3d3b71849bfd8e813f051e84fe3c7e60d6235fb45c.jpg b/parse/dev/0IywQ8uxJx/images/8f3f6fc69b739a5dcdc95e3d3b71849bfd8e813f051e84fe3c7e60d6235fb45c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..576c68c9d2d016068eb1a6908c8a98cc66635c61 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/8f3f6fc69b739a5dcdc95e3d3b71849bfd8e813f051e84fe3c7e60d6235fb45c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4807a4153ad3aa40a608dc35f84eb53ed9ec88ad4b4e5c13223b20b448e8d7ec +size 8862 diff --git a/parse/dev/0IywQ8uxJx/images/9021cfad3ad02d3eda490bcda65d4332392f50a1266fb88b44a6910ead172db1.jpg b/parse/dev/0IywQ8uxJx/images/9021cfad3ad02d3eda490bcda65d4332392f50a1266fb88b44a6910ead172db1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f5d3379c87a2644226ff95ec4d6a9d57467c6a1b --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/9021cfad3ad02d3eda490bcda65d4332392f50a1266fb88b44a6910ead172db1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a6af5aac342347c473a49fab4292ffd6c9edf769c6b1f8018671b49f4eaff76 +size 109195 diff --git a/parse/dev/0IywQ8uxJx/images/95c9c069d4facaa33c48db84921de26d64884ce17c2f540687ce59ecdf154b34.jpg b/parse/dev/0IywQ8uxJx/images/95c9c069d4facaa33c48db84921de26d64884ce17c2f540687ce59ecdf154b34.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e5be55dd2b4816baa5adf8364eb4448ddb7e8d56 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/95c9c069d4facaa33c48db84921de26d64884ce17c2f540687ce59ecdf154b34.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b5f0c2bb4d258bc7aa0c0b213872be7af3316f78097c32f78f2cf20dae8b558 +size 8758 diff --git a/parse/dev/0IywQ8uxJx/images/b8a1bae3225f2c79fd9a45b1324eaeb367605880a18a518af633abc6aad48181.jpg b/parse/dev/0IywQ8uxJx/images/b8a1bae3225f2c79fd9a45b1324eaeb367605880a18a518af633abc6aad48181.jpg new file mode 100644 index 0000000000000000000000000000000000000000..52b963580395c236b5fafd79ce7759b158b9ad7e --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/b8a1bae3225f2c79fd9a45b1324eaeb367605880a18a518af633abc6aad48181.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b0f0ef21146c13385667708f31bcb656d8f5816c0eb275af02ce6dfcfec0e300 +size 4853 diff --git a/parse/dev/0IywQ8uxJx/images/b8bc3e3e6d92424ce12c4d1f1d578bd54939bb81d114a96b4359856fbeaf2585.jpg b/parse/dev/0IywQ8uxJx/images/b8bc3e3e6d92424ce12c4d1f1d578bd54939bb81d114a96b4359856fbeaf2585.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ba877a45abdb3054febb0428e7659961e3beb974 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/b8bc3e3e6d92424ce12c4d1f1d578bd54939bb81d114a96b4359856fbeaf2585.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f9c25b2af34517440af36baded3b48449827e95da536be9a6519959a362ce384 +size 11916 diff --git a/parse/dev/0IywQ8uxJx/images/bbd3047d2cbd0b9bc92189a22e8836ee7d60b4c122913cbc88bde0574145b67a.jpg b/parse/dev/0IywQ8uxJx/images/bbd3047d2cbd0b9bc92189a22e8836ee7d60b4c122913cbc88bde0574145b67a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..171d5c4f38beb7fe8b5753747c3f450406e400fb --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/bbd3047d2cbd0b9bc92189a22e8836ee7d60b4c122913cbc88bde0574145b67a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9edf014b6e82325661c37cb891ec3ddf8504302ccde053d82098edf8eb028cbe +size 15145 diff --git a/parse/dev/0IywQ8uxJx/images/caf7071b5e44eb992bdac1865787a450a18a6822fbbb17b131512c0b3bd91df2.jpg b/parse/dev/0IywQ8uxJx/images/caf7071b5e44eb992bdac1865787a450a18a6822fbbb17b131512c0b3bd91df2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a348338d07ff54639681918d137604bdd3c66e65 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/caf7071b5e44eb992bdac1865787a450a18a6822fbbb17b131512c0b3bd91df2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7891287356d24c0ac499a38ad21e8a0d1c43d0df19cfe65ac02cd5227c904ec6 +size 5668 diff --git a/parse/dev/0IywQ8uxJx/images/cb34dad6e7a575313485e28a3c895f9e4982683f7e64518e4756c21f51f25781.jpg b/parse/dev/0IywQ8uxJx/images/cb34dad6e7a575313485e28a3c895f9e4982683f7e64518e4756c21f51f25781.jpg new file mode 100644 index 0000000000000000000000000000000000000000..60e33d9d1a2accfbcee406d896858490d769b1d8 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/cb34dad6e7a575313485e28a3c895f9e4982683f7e64518e4756c21f51f25781.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e330ebb877dfb5b3e95fd4701f468f846e035a8dd39dfd2d59509626e0984e9f +size 4889 diff --git a/parse/dev/0IywQ8uxJx/images/e05e27c2cfe54fa4828d9675759b7302ebf31a15528936ec50858043bda97eb0.jpg b/parse/dev/0IywQ8uxJx/images/e05e27c2cfe54fa4828d9675759b7302ebf31a15528936ec50858043bda97eb0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bcbe4151637cb04b99fbeea74076da5a2467eb65 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/e05e27c2cfe54fa4828d9675759b7302ebf31a15528936ec50858043bda97eb0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:200fb0765c56c8fd4602bca2949e2e4a2ea3547299ba82d66f4253756776ce9c +size 5593 diff --git a/parse/dev/0IywQ8uxJx/images/ea428720bada9ad2302dbe5eaef898cc58ff51b1b353db0aa69c812e991684c9.jpg b/parse/dev/0IywQ8uxJx/images/ea428720bada9ad2302dbe5eaef898cc58ff51b1b353db0aa69c812e991684c9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dea6ce6fe9e102f77035fa8a89dd0a1a8d8dc48b --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/ea428720bada9ad2302dbe5eaef898cc58ff51b1b353db0aa69c812e991684c9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:682588086923c2520b07ca9b290a9a003ba96fa9178e4de0f91f7f0a4a8f9b09 +size 5973 diff --git a/parse/dev/0IywQ8uxJx/images/faaace426bfe6b0a392878a365a28bf7cf87a44a8b17f69cbf9f2d60f041c1ce.jpg b/parse/dev/0IywQ8uxJx/images/faaace426bfe6b0a392878a365a28bf7cf87a44a8b17f69cbf9f2d60f041c1ce.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7d0375943ae97d94317c9a77ad75ba0bded8f0b9 --- /dev/null +++ b/parse/dev/0IywQ8uxJx/images/faaace426bfe6b0a392878a365a28bf7cf87a44a8b17f69cbf9f2d60f041c1ce.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5a34b351f67ab4734cc5b91570b8594f124d325d0d6103c1755e9356179da4b +size 11778 diff --git a/parse/dev/0LXEvcD3dB/images/18e296fb0af1200b4362fce127e53612728fab4197c4d19c9740c246b9e519ce.jpg b/parse/dev/0LXEvcD3dB/images/18e296fb0af1200b4362fce127e53612728fab4197c4d19c9740c246b9e519ce.jpg new file mode 100644 index 0000000000000000000000000000000000000000..56bfffe719f6b19bfc8ab88cdd9477e938db8aa9 --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/18e296fb0af1200b4362fce127e53612728fab4197c4d19c9740c246b9e519ce.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c1bcf75eaaf9272d212a446ce901245318bbc3d84838c2a3a94a8b94d19d50b +size 79385 diff --git a/parse/dev/0LXEvcD3dB/images/1d61b01892386d5a98c656e490472f5748ac72f5299b244e6050df23e1219fca.jpg b/parse/dev/0LXEvcD3dB/images/1d61b01892386d5a98c656e490472f5748ac72f5299b244e6050df23e1219fca.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9d64e1b97c1674c990db22dd29d811e9b9157bef --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/1d61b01892386d5a98c656e490472f5748ac72f5299b244e6050df23e1219fca.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7236c2d844e1c3723b3327538db454ba05d120f864973d903f24492dd943301c +size 21251 diff --git a/parse/dev/0LXEvcD3dB/images/2914927effd25a1062b348dbbc7abea459d3b55a988bd889ae878e885b349122.jpg b/parse/dev/0LXEvcD3dB/images/2914927effd25a1062b348dbbc7abea459d3b55a988bd889ae878e885b349122.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6c5b5635830856d777c2fed89dbd39eb239b8255 --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/2914927effd25a1062b348dbbc7abea459d3b55a988bd889ae878e885b349122.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a61d177f10e9e47f94af8b8de26efe5a6ab8a02fdfac32b59e18a7c2a5f2b082 +size 31343 diff --git a/parse/dev/0LXEvcD3dB/images/29ebe76d95a7e8df91127675821b59b7dfd208841d7b69c05c18697836b3f716.jpg b/parse/dev/0LXEvcD3dB/images/29ebe76d95a7e8df91127675821b59b7dfd208841d7b69c05c18697836b3f716.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7aed0dd55c858b8966a16999b5684b9769f82895 --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/29ebe76d95a7e8df91127675821b59b7dfd208841d7b69c05c18697836b3f716.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43cecf87f993eaaea0c36e39e04a74421da95c63523a8259947b914b52b838d0 +size 44065 diff --git a/parse/dev/0LXEvcD3dB/images/2a5ff459674a5e6a9098d06007bb122a5347290ea0bb0eb37dcdfe681b740ee4.jpg b/parse/dev/0LXEvcD3dB/images/2a5ff459674a5e6a9098d06007bb122a5347290ea0bb0eb37dcdfe681b740ee4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e260869cf6f054c4e67a1ebb59a561f04c62e412 --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/2a5ff459674a5e6a9098d06007bb122a5347290ea0bb0eb37dcdfe681b740ee4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b996ea647adefebe20b7795b4b5d6b5197d5388fef83c273430f8d3454f8b722 +size 2723 diff --git a/parse/dev/0LXEvcD3dB/images/5c24d3200a6c1c1a8ab469f2b58bd7a939e61229813d75742509df2d4030eccd.jpg b/parse/dev/0LXEvcD3dB/images/5c24d3200a6c1c1a8ab469f2b58bd7a939e61229813d75742509df2d4030eccd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..982ed525b67740d0d3140df22c5a02a115153c84 --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/5c24d3200a6c1c1a8ab469f2b58bd7a939e61229813d75742509df2d4030eccd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c3022582ec494e57d76b1c12425c181d60f30b20c62e14dbbdac28a8fce43a8 +size 7224 diff --git a/parse/dev/0LXEvcD3dB/images/736c4a0fae6a175438d4a57d6bf4d60b0599e47380be66ebaf270706a95705b4.jpg b/parse/dev/0LXEvcD3dB/images/736c4a0fae6a175438d4a57d6bf4d60b0599e47380be66ebaf270706a95705b4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d7ef04161d4e96ccb533aeb46b572d2929101eb5 --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/736c4a0fae6a175438d4a57d6bf4d60b0599e47380be66ebaf270706a95705b4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc7b7fad8e255c02ac80bc06243ede836197fbab4ae207fe7266f9b1186feb52 +size 7855 diff --git a/parse/dev/0LXEvcD3dB/images/9a7fcd22dd35afec89ad794b756a0b89d8a71276f648f06dff3208200eb58e3d.jpg b/parse/dev/0LXEvcD3dB/images/9a7fcd22dd35afec89ad794b756a0b89d8a71276f648f06dff3208200eb58e3d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9b5c7d47bb979a22b2550d965bfe787ca172adb4 --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/9a7fcd22dd35afec89ad794b756a0b89d8a71276f648f06dff3208200eb58e3d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7fdc5f527b02a581f0ad1113fe99c2f63295846db42acefb5e926f3671067c6c +size 68370 diff --git a/parse/dev/0LXEvcD3dB/images/da7e7b4aab6aa78969c08f258ac74e2b540f2e61c6360fd974bc580739c6c728.jpg b/parse/dev/0LXEvcD3dB/images/da7e7b4aab6aa78969c08f258ac74e2b540f2e61c6360fd974bc580739c6c728.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ad500379bbb26030395e7ebfacdccda30c3383e4 --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/da7e7b4aab6aa78969c08f258ac74e2b540f2e61c6360fd974bc580739c6c728.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:94104ab039f5e4916370ac9ffe521cb312bcf8e6aee76413389437e2acfb0abb +size 29531 diff --git a/parse/dev/0LXEvcD3dB/images/e46aa44b83753245531eae6af65862d3d2b33291238053589833d6d978b8c55e.jpg b/parse/dev/0LXEvcD3dB/images/e46aa44b83753245531eae6af65862d3d2b33291238053589833d6d978b8c55e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8efc0bdeed1ab72779def28c881097437062685f --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/e46aa44b83753245531eae6af65862d3d2b33291238053589833d6d978b8c55e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:71238ad747aac31ad3da2e31080a5c520961f75221079131371586f91aa44c4b +size 31013 diff --git a/parse/dev/0LXEvcD3dB/images/ecf3e6ed2da4cca7d61a363613f9a1609345ce3c1d02ce37899c0cd76172bc7f.jpg b/parse/dev/0LXEvcD3dB/images/ecf3e6ed2da4cca7d61a363613f9a1609345ce3c1d02ce37899c0cd76172bc7f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7f812b533b14f6ce12edb0248a17bb3176f105ed --- /dev/null +++ b/parse/dev/0LXEvcD3dB/images/ecf3e6ed2da4cca7d61a363613f9a1609345ce3c1d02ce37899c0cd76172bc7f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57c24f7bfd51d5c7ce4a415c4447146da7a77c3fcf0e09b32fd9d77b704e8f46 +size 1940 diff --git a/parse/dev/3R3Pz5i0tye/images/042f8fad029b3eccb9b2940378efa3272789deb428476874bd69fbe6d8a34faf.jpg b/parse/dev/3R3Pz5i0tye/images/042f8fad029b3eccb9b2940378efa3272789deb428476874bd69fbe6d8a34faf.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b9f4163bb23f53fd9ef1ad2a85ffbb54259468f8 --- /dev/null +++ b/parse/dev/3R3Pz5i0tye/images/042f8fad029b3eccb9b2940378efa3272789deb428476874bd69fbe6d8a34faf.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a70f3263d7ec7feff3e178c1714bbccd3700d7298932f0eabdf102d3207254a2 +size 60589 diff --git a/parse/dev/3R3Pz5i0tye/images/23f63105548f5a25f6be3e43c2c0827df5c8ada1ffc79fc8b275dd8d4dc37c46.jpg b/parse/dev/3R3Pz5i0tye/images/23f63105548f5a25f6be3e43c2c0827df5c8ada1ffc79fc8b275dd8d4dc37c46.jpg new file mode 100644 index 0000000000000000000000000000000000000000..21465ec88dd580617ddfff1cb2bcbdedea6b8a21 --- /dev/null +++ b/parse/dev/3R3Pz5i0tye/images/23f63105548f5a25f6be3e43c2c0827df5c8ada1ffc79fc8b275dd8d4dc37c46.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c7595ea9dc6316e515c33cdecd66addc342736650b97eaf467a6f96ea755c539 +size 69441 diff --git a/parse/dev/3R3Pz5i0tye/images/6e4d56d21ac1d79c836010878e48898181199d009ab6fd9d5a4387fba16cf9ab.jpg b/parse/dev/3R3Pz5i0tye/images/6e4d56d21ac1d79c836010878e48898181199d009ab6fd9d5a4387fba16cf9ab.jpg new file mode 100644 index 0000000000000000000000000000000000000000..04f088ddbfc60d94819057b3f51502ba082adfbd --- /dev/null +++ b/parse/dev/3R3Pz5i0tye/images/6e4d56d21ac1d79c836010878e48898181199d009ab6fd9d5a4387fba16cf9ab.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:46857f901958ddd5b5b617af9480d12010697fc385270712bd5ed63c325c1202 +size 161521 diff --git a/parse/dev/3R3Pz5i0tye/images/7ed3f15bf2a139e412449021d4a4d3a56239547a742e782baae455d0285b4c5e.jpg b/parse/dev/3R3Pz5i0tye/images/7ed3f15bf2a139e412449021d4a4d3a56239547a742e782baae455d0285b4c5e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4f7c7c886dbe5ece49ede6fc43c6d9b14e4fe85a --- /dev/null +++ b/parse/dev/3R3Pz5i0tye/images/7ed3f15bf2a139e412449021d4a4d3a56239547a742e782baae455d0285b4c5e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f4ed10e281df18548795e69a5f39034920f935456b24a29f4bf047c5a231031 +size 43097 diff --git a/parse/dev/3R3Pz5i0tye/images/b69f1e6523a1bf91127c8424584ad2b72679a7d90508907785b1eab73cac9df5.jpg b/parse/dev/3R3Pz5i0tye/images/b69f1e6523a1bf91127c8424584ad2b72679a7d90508907785b1eab73cac9df5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b14879e7d7c2f8645eaa66903aa54c85c5819668 --- /dev/null +++ b/parse/dev/3R3Pz5i0tye/images/b69f1e6523a1bf91127c8424584ad2b72679a7d90508907785b1eab73cac9df5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e827c4b2a80f713f0edad8407322768de9a05f18a08175095ff379bbf6afb18 +size 23297 diff --git a/parse/dev/3R3Pz5i0tye/images/b751ebf5a9eeb4eacad63238d86487141852922195f38985cb0dd571d16bf2ea.jpg b/parse/dev/3R3Pz5i0tye/images/b751ebf5a9eeb4eacad63238d86487141852922195f38985cb0dd571d16bf2ea.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8764e70f3216da92c0689664ea93e110b52852a8 --- /dev/null +++ b/parse/dev/3R3Pz5i0tye/images/b751ebf5a9eeb4eacad63238d86487141852922195f38985cb0dd571d16bf2ea.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a5fb8d4fe619e529be73917ee017f6e602fdc5874ffe2ee3581d9ef3f18de62 +size 139973 diff --git a/parse/dev/3R3Pz5i0tye/images/bbccad41cc5e20aea1abb341ed8f534bf72ee44935a7f38a47d65f5334398156.jpg b/parse/dev/3R3Pz5i0tye/images/bbccad41cc5e20aea1abb341ed8f534bf72ee44935a7f38a47d65f5334398156.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7044369026168e87c2c5ec66aa0e840a1117a19a --- /dev/null +++ b/parse/dev/3R3Pz5i0tye/images/bbccad41cc5e20aea1abb341ed8f534bf72ee44935a7f38a47d65f5334398156.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d76b6bf0890b1b289a28f5eb3764a727d7f8b33eb66adc907824ab3183cb177 +size 73039 diff --git a/parse/dev/3R3Pz5i0tye/images/db55033730673a7d5c48e55e77fc8eed317a3f44e4db10126e2a17ac9ad10a76.jpg b/parse/dev/3R3Pz5i0tye/images/db55033730673a7d5c48e55e77fc8eed317a3f44e4db10126e2a17ac9ad10a76.jpg new file mode 100644 index 0000000000000000000000000000000000000000..30f86d13a7b88693af177b587eb8d94a3fa6d19d --- /dev/null +++ b/parse/dev/3R3Pz5i0tye/images/db55033730673a7d5c48e55e77fc8eed317a3f44e4db10126e2a17ac9ad10a76.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5826043989465c932765a389187306588b7a78a812d5c8f2b82173b9e4f9d9c5 +size 26569 diff --git a/parse/dev/3uwj8QZROL/images/0add43e1b5e7daf184ac991319dfa3f26c966b50cb187a76cf9a0f2f3edf5792.jpg b/parse/dev/3uwj8QZROL/images/0add43e1b5e7daf184ac991319dfa3f26c966b50cb187a76cf9a0f2f3edf5792.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7ecd19f3d24459f62954dcb454489bb1b7d28abe --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/0add43e1b5e7daf184ac991319dfa3f26c966b50cb187a76cf9a0f2f3edf5792.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:21a3afac61b34835c5eb8e2eec39ee23ba6336269fca639aaaf8040f25bf0ae9 +size 28162 diff --git a/parse/dev/3uwj8QZROL/images/22c7522a90c85b5b94f32cb97df5872220f10ae2c946c31d19d6e3eb392b4902.jpg b/parse/dev/3uwj8QZROL/images/22c7522a90c85b5b94f32cb97df5872220f10ae2c946c31d19d6e3eb392b4902.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f3db3b767afc9bd5c66d72d2d58ee93ade119628 --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/22c7522a90c85b5b94f32cb97df5872220f10ae2c946c31d19d6e3eb392b4902.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65d42036e66a05be532f792732498b345c09cd14d667b873ce6f02a9c2617b88 +size 73639 diff --git a/parse/dev/3uwj8QZROL/images/2a01ee3c09f480b88ac8375b319968f15d3d2a0a747b2937e52ed97ed42cfdf5.jpg b/parse/dev/3uwj8QZROL/images/2a01ee3c09f480b88ac8375b319968f15d3d2a0a747b2937e52ed97ed42cfdf5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1a4c675050844886d241e509b55670ae3467da29 --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/2a01ee3c09f480b88ac8375b319968f15d3d2a0a747b2937e52ed97ed42cfdf5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77720aab2ecf30315bee4fc2c4629d8221a1d59eb9e55926d9ada12f476f1f14 +size 47876 diff --git a/parse/dev/3uwj8QZROL/images/30b00f83487263cbecc9212528670563a1ad3f986a7739c3026049373ee3b1f0.jpg b/parse/dev/3uwj8QZROL/images/30b00f83487263cbecc9212528670563a1ad3f986a7739c3026049373ee3b1f0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..14df041d411afb3c617074bd6fb95abf1a1e5a46 --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/30b00f83487263cbecc9212528670563a1ad3f986a7739c3026049373ee3b1f0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f732b056a73a6e1c91b798d925aaa3555aa522fdd88aac4f4cb90939b06b89ca +size 76686 diff --git a/parse/dev/3uwj8QZROL/images/4744b196017d0467fd75e21b1770f0e42c8b6ad74eed9c78004a816824d44cc1.jpg b/parse/dev/3uwj8QZROL/images/4744b196017d0467fd75e21b1770f0e42c8b6ad74eed9c78004a816824d44cc1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3ff9467a50cd0154bf5b4054c695f3dd43c3eccc --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/4744b196017d0467fd75e21b1770f0e42c8b6ad74eed9c78004a816824d44cc1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07bc35c9976b394dcac864602d714fc690150f01d24b6e0770dbb6fb50ffb76b +size 33300 diff --git a/parse/dev/3uwj8QZROL/images/78a03ced9845e5e36be93f05d891ac4d3e7a6d2c26453a50febf301bf9995670.jpg b/parse/dev/3uwj8QZROL/images/78a03ced9845e5e36be93f05d891ac4d3e7a6d2c26453a50febf301bf9995670.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3fa77fe1ed49592c8063d7f56b9b51c47d1334e8 --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/78a03ced9845e5e36be93f05d891ac4d3e7a6d2c26453a50febf301bf9995670.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bbcfcbddc6082327ba003c9e50be1b4c2161485a3501641f9599c84525b6eed5 +size 13255 diff --git a/parse/dev/3uwj8QZROL/images/8b84c8bcafb3c5523a3a5292076f5a24e44c545a54376ab9b4b74b865087bd27.jpg b/parse/dev/3uwj8QZROL/images/8b84c8bcafb3c5523a3a5292076f5a24e44c545a54376ab9b4b74b865087bd27.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2da58776074120b231ad6937d44e4e7cff1ba1a5 --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/8b84c8bcafb3c5523a3a5292076f5a24e44c545a54376ab9b4b74b865087bd27.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:608780d09c71886a1a00e06d52d0eaa22998a9310c0a9af960de84db3013c608 +size 77052 diff --git a/parse/dev/3uwj8QZROL/images/9ae4afdd3193c47d2f2b4a04951d6281d21f9cc3a160399a0b424263980e3765.jpg b/parse/dev/3uwj8QZROL/images/9ae4afdd3193c47d2f2b4a04951d6281d21f9cc3a160399a0b424263980e3765.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8e5cf815360c05cfa557ca893c672baa6c776504 --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/9ae4afdd3193c47d2f2b4a04951d6281d21f9cc3a160399a0b424263980e3765.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f84000a8355846e4ef1ed42de9bf38c183860ed34dc6aa2bc5db7ac1cedca7bc +size 29890 diff --git a/parse/dev/3uwj8QZROL/images/aac2af01ab14e71b284d50ba38a37b40f4d7e2ba7bae0cef76d32b765dba60cb.jpg b/parse/dev/3uwj8QZROL/images/aac2af01ab14e71b284d50ba38a37b40f4d7e2ba7bae0cef76d32b765dba60cb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8bf97caa3b6693b3d00a1adec96cb7a8c3760254 --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/aac2af01ab14e71b284d50ba38a37b40f4d7e2ba7bae0cef76d32b765dba60cb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a37b22e37eb56157bd6ac933378b8aea55a991a896fb9ad2e815c651a27e74d2 +size 16551 diff --git a/parse/dev/3uwj8QZROL/images/af501840695a83365f589251c0fbc7f8aabfd9958eafb1d1c3b46a1d762c10b3.jpg b/parse/dev/3uwj8QZROL/images/af501840695a83365f589251c0fbc7f8aabfd9958eafb1d1c3b46a1d762c10b3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e577d0a5e8e37466753c1cb9f94cc99ba7af3f49 --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/af501840695a83365f589251c0fbc7f8aabfd9958eafb1d1c3b46a1d762c10b3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e5935c9f714cc5b9341b953d1424bce15cf6a8a4009383908a64744f62fcd7c +size 10515 diff --git a/parse/dev/3uwj8QZROL/images/f63bc4948ea74ccafeed52fb6956ad8708828d0027d13f9da9fa68a447de7324.jpg b/parse/dev/3uwj8QZROL/images/f63bc4948ea74ccafeed52fb6956ad8708828d0027d13f9da9fa68a447de7324.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7e6bc8afd82022c799b7da11624c1f5caa16a3b5 --- /dev/null +++ b/parse/dev/3uwj8QZROL/images/f63bc4948ea74ccafeed52fb6956ad8708828d0027d13f9da9fa68a447de7324.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c0e6fbd652846753ae49a11755f77fe853225623557b59a1c418f32a084f139c +size 26881 diff --git a/parse/dev/5HLoTvVGDe/images/07785b29428f7b754ea962d1d0e9223b7335974dc2a9dcfcc173504ec79fe461.jpg b/parse/dev/5HLoTvVGDe/images/07785b29428f7b754ea962d1d0e9223b7335974dc2a9dcfcc173504ec79fe461.jpg new file mode 100644 index 0000000000000000000000000000000000000000..498d3d7631a94976f590ad9056500e40efd054ea --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/07785b29428f7b754ea962d1d0e9223b7335974dc2a9dcfcc173504ec79fe461.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:10e84cdba72788b7991be34aac1cb0f4ffb16be740aaa3f2b1f6b1564801ac9f +size 8024 diff --git a/parse/dev/5HLoTvVGDe/images/0afb2f71971abc39ddad7e47bc616a93c5b9a03f912363401ccca81d7c56f32d.jpg b/parse/dev/5HLoTvVGDe/images/0afb2f71971abc39ddad7e47bc616a93c5b9a03f912363401ccca81d7c56f32d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9f0ac851d7347ef7d062d12f7f044ac4f80a6ce0 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/0afb2f71971abc39ddad7e47bc616a93c5b9a03f912363401ccca81d7c56f32d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8e242e50f30899af02cffaba6c1e10ddd1c64892d660d4e54ad7477447f3c32 +size 9958 diff --git a/parse/dev/5HLoTvVGDe/images/2a7c4e1aadce5a2456cd8cabf97f1b26371bab2b5b5ca1053de92b69e8f118e7.jpg b/parse/dev/5HLoTvVGDe/images/2a7c4e1aadce5a2456cd8cabf97f1b26371bab2b5b5ca1053de92b69e8f118e7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4e3406c19307d6f05f39f2f6b8d579db9e4bd3c4 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/2a7c4e1aadce5a2456cd8cabf97f1b26371bab2b5b5ca1053de92b69e8f118e7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:95e9bc1ed7888c6dcb48f9e37b8cf19194d2e12cd715807bd01ec34051da3dba +size 9613 diff --git a/parse/dev/5HLoTvVGDe/images/2b61eb9f42e2b1f466db582a6a11ff82b05efaaa0dfc5885b8484b5d963b81d1.jpg b/parse/dev/5HLoTvVGDe/images/2b61eb9f42e2b1f466db582a6a11ff82b05efaaa0dfc5885b8484b5d963b81d1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..707ea133046e76bfca91bb2551166eec6da42e73 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/2b61eb9f42e2b1f466db582a6a11ff82b05efaaa0dfc5885b8484b5d963b81d1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a72b8542608fa2e76b63b5a6a4354f71a9d7f2f54dc740693c339e60537df22a +size 52877 diff --git a/parse/dev/5HLoTvVGDe/images/3258de5665a5c2a16d56cf4d2c018f21fecf14a9b389df28f63e27fe67c77b76.jpg b/parse/dev/5HLoTvVGDe/images/3258de5665a5c2a16d56cf4d2c018f21fecf14a9b389df28f63e27fe67c77b76.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bc89e4d5afb7fc92bfddbcbe40b60271162b1b8f --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/3258de5665a5c2a16d56cf4d2c018f21fecf14a9b389df28f63e27fe67c77b76.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:30cded79ead9cb5939cb31da65ba0106ee4b955550555ffee29d289ca88a2a8a +size 51792 diff --git a/parse/dev/5HLoTvVGDe/images/468d8cf66c2feae617666c379545a5d342da54656551c4b468384e7872b006df.jpg b/parse/dev/5HLoTvVGDe/images/468d8cf66c2feae617666c379545a5d342da54656551c4b468384e7872b006df.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e833ea5fed6683065d0ed88331fba945943e5c0d --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/468d8cf66c2feae617666c379545a5d342da54656551c4b468384e7872b006df.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be6f219370339ee410692852ef76ce6bd0d81b71675c77306559559ef2709b01 +size 19974 diff --git a/parse/dev/5HLoTvVGDe/images/649b746d19b2ec60e11404bf7366e6b99da18f6224f71705c869e6e0c897b10e.jpg b/parse/dev/5HLoTvVGDe/images/649b746d19b2ec60e11404bf7366e6b99da18f6224f71705c869e6e0c897b10e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..44d19f367d547d27ae551a2ee37c6e9758e667bc --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/649b746d19b2ec60e11404bf7366e6b99da18f6224f71705c869e6e0c897b10e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df235e44bda7787af3e9fcb407309504973219a2d1c05e011bd9d9d780ab77b5 +size 125481 diff --git a/parse/dev/5HLoTvVGDe/images/6ce5f7422a6cd137398786eafd2f96d43e6d3216c82dc7136f40c0f7dfd17c78.jpg b/parse/dev/5HLoTvVGDe/images/6ce5f7422a6cd137398786eafd2f96d43e6d3216c82dc7136f40c0f7dfd17c78.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c1d18e7853580b3414bc2d09740b8fdcfbbeb802 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/6ce5f7422a6cd137398786eafd2f96d43e6d3216c82dc7136f40c0f7dfd17c78.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c897f586c478bd033ce7f07a65d0709ed214fb17a5788db5b9e738ba0a72700 +size 5472 diff --git a/parse/dev/5HLoTvVGDe/images/7a3d55f81dee5d6bb207a44ff8dab4e5c1c8a460b491f0381fe66b18e8e9e4ef.jpg b/parse/dev/5HLoTvVGDe/images/7a3d55f81dee5d6bb207a44ff8dab4e5c1c8a460b491f0381fe66b18e8e9e4ef.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6c4e7c778d631d2cc56efd50ed2dd182d4f3390b --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/7a3d55f81dee5d6bb207a44ff8dab4e5c1c8a460b491f0381fe66b18e8e9e4ef.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c22e19c1cffccd004724b5ab6c2497778bde800424cd16ff9c9e247e38d8c52c +size 10712 diff --git a/parse/dev/5HLoTvVGDe/images/96924226f650ec72a1b0f4bffc85669d3bb53e018354455fa300015d4759a89f.jpg b/parse/dev/5HLoTvVGDe/images/96924226f650ec72a1b0f4bffc85669d3bb53e018354455fa300015d4759a89f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1dd0bc3f45907a47157989f6db8dc2eab211d34f --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/96924226f650ec72a1b0f4bffc85669d3bb53e018354455fa300015d4759a89f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0cd7da6b625a3d6f6ddac818089609a06284e45f16b66c17b4590a98e373f4b9 +size 5869 diff --git a/parse/dev/5HLoTvVGDe/images/9a11302c72e2c4164abfc78bddbf2ac251badbafda7dcee77d3f1b0337d52e44.jpg b/parse/dev/5HLoTvVGDe/images/9a11302c72e2c4164abfc78bddbf2ac251badbafda7dcee77d3f1b0337d52e44.jpg new file mode 100644 index 0000000000000000000000000000000000000000..15acefdaee5c0765cbab0fb5018c7ef214e314f1 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/9a11302c72e2c4164abfc78bddbf2ac251badbafda7dcee77d3f1b0337d52e44.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3c2f9430858fd6954b45863a4b753a75d7bab06a26a1078ed1c7c607a868b07b +size 79592 diff --git a/parse/dev/5HLoTvVGDe/images/9a9cd802cd7efc288712eac010a63baa13a07222fa8452fa01361d7e72ad5ea0.jpg b/parse/dev/5HLoTvVGDe/images/9a9cd802cd7efc288712eac010a63baa13a07222fa8452fa01361d7e72ad5ea0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..386f7b5fbff05ed67a457cf6e7f5ee2aacef4488 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/9a9cd802cd7efc288712eac010a63baa13a07222fa8452fa01361d7e72ad5ea0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c0ec42a97b5e453e924e6ea1db3f83c53c681caadf7353603e16456abfeca2e2 +size 16501 diff --git a/parse/dev/5HLoTvVGDe/images/9c49e5f0ab830b6b1f359c6d17497bc2ac880460d824f5395fdeb30197f6f7c9.jpg b/parse/dev/5HLoTvVGDe/images/9c49e5f0ab830b6b1f359c6d17497bc2ac880460d824f5395fdeb30197f6f7c9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..112cab3ddce358cb18fb13a6e5d9210c8fa22810 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/9c49e5f0ab830b6b1f359c6d17497bc2ac880460d824f5395fdeb30197f6f7c9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:17f52c4112bbc88197dd0d4315f696139b4f8c33602d3e67c544bda3422f790a +size 25535 diff --git a/parse/dev/5HLoTvVGDe/images/ae78672377be50dd79010dbe570b9745d13641e689747b703ee356dcbc6d14b0.jpg b/parse/dev/5HLoTvVGDe/images/ae78672377be50dd79010dbe570b9745d13641e689747b703ee356dcbc6d14b0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e0d7d7520fe6fab4fc51a52c8de24b3f1272b77e --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/ae78672377be50dd79010dbe570b9745d13641e689747b703ee356dcbc6d14b0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8855d26fb573cf5ada4caac9d004d85d0974018c18edae9951ca52ac994ec077 +size 49588 diff --git a/parse/dev/5HLoTvVGDe/images/c1f26c0a6bef908cc1ead156af8e57cbcc448fa17dda125fa0fb83aa3933213b.jpg b/parse/dev/5HLoTvVGDe/images/c1f26c0a6bef908cc1ead156af8e57cbcc448fa17dda125fa0fb83aa3933213b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4da142c35e7814a52c0c2bfb0cf0abbae45288ae --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/c1f26c0a6bef908cc1ead156af8e57cbcc448fa17dda125fa0fb83aa3933213b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce357072157d06b295ecc3c8d98c3d8b13f55d19ca8931e8639ca29e2a518a90 +size 65834 diff --git a/parse/dev/5HLoTvVGDe/images/ccb8103213bf4508747916601fb8ec38828642f5c8052f50af2b5c931d859c79.jpg b/parse/dev/5HLoTvVGDe/images/ccb8103213bf4508747916601fb8ec38828642f5c8052f50af2b5c931d859c79.jpg new file mode 100644 index 0000000000000000000000000000000000000000..503152b51b163b2e492ec7b1880dc3c3e016a0cc --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/ccb8103213bf4508747916601fb8ec38828642f5c8052f50af2b5c931d859c79.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:47328508d0421541a165230a1d9ce555afb921107d01dd287b7582c8a04ff508 +size 15153 diff --git a/parse/dev/5HLoTvVGDe/images/cd8510744b3893bbc51c9621e5cf8cb1efd83273b7882e05ecc0a0c2741629ca.jpg b/parse/dev/5HLoTvVGDe/images/cd8510744b3893bbc51c9621e5cf8cb1efd83273b7882e05ecc0a0c2741629ca.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8db6cb58cb1931831a4d873389c77b6ffda9ac38 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/cd8510744b3893bbc51c9621e5cf8cb1efd83273b7882e05ecc0a0c2741629ca.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d381fb36c2391390a32b4587423a94ca3d58c5140f0fa4edb6331cdd8cd36926 +size 9974 diff --git a/parse/dev/5HLoTvVGDe/images/e35ff7df5f979ad6ab5875bfd8f6d3c0b1d97b83ba06eec8df3b28086906e91d.jpg b/parse/dev/5HLoTvVGDe/images/e35ff7df5f979ad6ab5875bfd8f6d3c0b1d97b83ba06eec8df3b28086906e91d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..618acdb8e80791a33574434d61f7045c6a90fe4d --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/e35ff7df5f979ad6ab5875bfd8f6d3c0b1d97b83ba06eec8df3b28086906e91d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5ed24758601f22bf6a67e8ba4506ab9fd6d51d4867d27aa1aeace2e62696165b +size 104854 diff --git a/parse/dev/5HLoTvVGDe/images/e3c00ea5c7b96551cd12cc1ae7e09967d617e72a91b85c18aca8ce87095df1fe.jpg b/parse/dev/5HLoTvVGDe/images/e3c00ea5c7b96551cd12cc1ae7e09967d617e72a91b85c18aca8ce87095df1fe.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0f4e451eb75c37a0787b83b3defd0a18d52d2e2a --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/e3c00ea5c7b96551cd12cc1ae7e09967d617e72a91b85c18aca8ce87095df1fe.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b93bc4392022413c56e8fd6587fdfee6d8caa5cb24470ca3d1422c9c3859853 +size 4381 diff --git a/parse/dev/5HLoTvVGDe/images/eb9720f73471a98b4cb6e1dc549e94a1be65a245fff317111ee10826050b4b06.jpg b/parse/dev/5HLoTvVGDe/images/eb9720f73471a98b4cb6e1dc549e94a1be65a245fff317111ee10826050b4b06.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9e2fb1dfd8545233fe4ce2a90082f8f471793cff --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/eb9720f73471a98b4cb6e1dc549e94a1be65a245fff317111ee10826050b4b06.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f2f9e8e2aa35bad47d31d773967799f4db810e2172c552d398b9779e8de6eeff +size 6543 diff --git a/parse/dev/5HLoTvVGDe/images/ed3cd99dadfa47191b869e324f869f2c593d640de753c52387bb79a0c6017a93.jpg b/parse/dev/5HLoTvVGDe/images/ed3cd99dadfa47191b869e324f869f2c593d640de753c52387bb79a0c6017a93.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5cc0ccb52cc3f18005653da7110a3f8cb5fe4ddd --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/ed3cd99dadfa47191b869e324f869f2c593d640de753c52387bb79a0c6017a93.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e03927ae5a3a75ce849761c7760990b52dc8cb44725ba809763326f2e49915a +size 266062 diff --git a/parse/dev/5HLoTvVGDe/images/edcce6da43511e77d3a1621dfbad2dbefec21e522b8df91230ac5eeff27bf73c.jpg b/parse/dev/5HLoTvVGDe/images/edcce6da43511e77d3a1621dfbad2dbefec21e522b8df91230ac5eeff27bf73c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..840554f404a76cf1efa3bb09a180e4d4e8569396 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/edcce6da43511e77d3a1621dfbad2dbefec21e522b8df91230ac5eeff27bf73c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce9b84bf49eee86d0b5be1a3c96dcc7a7d38518e7f2a4053a7d7e2118e1bc13b +size 32973 diff --git a/parse/dev/5HLoTvVGDe/images/ee647e9b50d95ff7634b2c1a2544b7ce20dcfb49996f55b37b905c2d91de733f.jpg b/parse/dev/5HLoTvVGDe/images/ee647e9b50d95ff7634b2c1a2544b7ce20dcfb49996f55b37b905c2d91de733f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..364de097237045c0b88aaf29b29cd024b15aa8c8 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/ee647e9b50d95ff7634b2c1a2544b7ce20dcfb49996f55b37b905c2d91de733f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:920bc2cda69710a9705a2a70f4744b75b85d9c39c84494fe64b34859651ca160 +size 5960 diff --git a/parse/dev/5HLoTvVGDe/images/f2c70f2d78578b1bbd43b1f9367fdd8af8134aa81254cde01a6590a8d7a7cd41.jpg b/parse/dev/5HLoTvVGDe/images/f2c70f2d78578b1bbd43b1f9367fdd8af8134aa81254cde01a6590a8d7a7cd41.jpg new file mode 100644 index 0000000000000000000000000000000000000000..00e5b611ad28a67d656972067aab155319378de6 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/f2c70f2d78578b1bbd43b1f9367fdd8af8134aa81254cde01a6590a8d7a7cd41.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ec32bb9e4b9c5e53fc707b8d9bbbc04502398e60ecdf267026401e67c0f5920 +size 10017 diff --git a/parse/dev/5HLoTvVGDe/images/f825e98d00916d4d037bbc77ab52f7cbad1259edd0566c2dc402140eb276e554.jpg b/parse/dev/5HLoTvVGDe/images/f825e98d00916d4d037bbc77ab52f7cbad1259edd0566c2dc402140eb276e554.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e69ce709c36fa4513365df825e6d3681cfaf5484 --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/f825e98d00916d4d037bbc77ab52f7cbad1259edd0566c2dc402140eb276e554.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0875a668f003cea5139b53b3bbf3c0ed37a66dfb2988b08b1a270f43578389be +size 8312 diff --git a/parse/dev/5HLoTvVGDe/images/f8eee46971c0bbc18b331790f27547b8f4719e98acab4f51a354b7d1fee65edb.jpg b/parse/dev/5HLoTvVGDe/images/f8eee46971c0bbc18b331790f27547b8f4719e98acab4f51a354b7d1fee65edb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b2f2cd089a0f76f43ce949db8e551a1520637e3d --- /dev/null +++ b/parse/dev/5HLoTvVGDe/images/f8eee46971c0bbc18b331790f27547b8f4719e98acab4f51a354b7d1fee65edb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e7bd05967d8b4d3ea3c842471d76f0f290032bc1810c758228ae451e821fd06 +size 9764 diff --git a/parse/dev/6u6N8WWwYSM/images/01730836a383cf595ab1cb93774d8e67234f48693672101e2fdcb2c8cc7761b2.jpg b/parse/dev/6u6N8WWwYSM/images/01730836a383cf595ab1cb93774d8e67234f48693672101e2fdcb2c8cc7761b2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ae848000a44d1e8e9d843d170513da1a99aa57a1 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/01730836a383cf595ab1cb93774d8e67234f48693672101e2fdcb2c8cc7761b2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c67d6630fe1d36f5a4452c297d2cabd5729fd7dc1358da1b99cc512f05b37b0 +size 19726 diff --git a/parse/dev/6u6N8WWwYSM/images/0739cc72fbaee2cab769c477c1d34bcdd520f1abbf4dca31741349e102bb6193.jpg b/parse/dev/6u6N8WWwYSM/images/0739cc72fbaee2cab769c477c1d34bcdd520f1abbf4dca31741349e102bb6193.jpg new file mode 100644 index 0000000000000000000000000000000000000000..aed68a9531ef2ec9eb1ad003d342d15f78cc4822 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/0739cc72fbaee2cab769c477c1d34bcdd520f1abbf4dca31741349e102bb6193.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5f5541a8cc047f49687ab34a33ed6051b83de49af599fbef9bb27e63a8a5f3b2 +size 105496 diff --git a/parse/dev/6u6N8WWwYSM/images/080f1e165fde866182ca85eb2ad2769d22d31503a8937a7b391377094cf6925e.jpg b/parse/dev/6u6N8WWwYSM/images/080f1e165fde866182ca85eb2ad2769d22d31503a8937a7b391377094cf6925e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3e9c2aa323c78be000e4dee42caf59c39e038197 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/080f1e165fde866182ca85eb2ad2769d22d31503a8937a7b391377094cf6925e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82a3ba4f3641f78aca5885031746afd2efcdc7586cc7c6f89b6c586e9e1cb8ce +size 230280 diff --git a/parse/dev/6u6N8WWwYSM/images/1175caa05b585655e133c94e3b0f6dc9ae581471e69922fae7fb62f0686910cb.jpg b/parse/dev/6u6N8WWwYSM/images/1175caa05b585655e133c94e3b0f6dc9ae581471e69922fae7fb62f0686910cb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2db66cd8fc9f6e75871b93496bcc6e132ef08449 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/1175caa05b585655e133c94e3b0f6dc9ae581471e69922fae7fb62f0686910cb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:52a64ba189a7d6ed8658d11e452fc19d1c2d0a5ca94bde4015d8131ba94cf094 +size 50008 diff --git a/parse/dev/6u6N8WWwYSM/images/1d1faee2aebe0d662445de16e655943bbf0508b697b27d5ffb3f0dd6bbe939cf.jpg b/parse/dev/6u6N8WWwYSM/images/1d1faee2aebe0d662445de16e655943bbf0508b697b27d5ffb3f0dd6bbe939cf.jpg new file mode 100644 index 0000000000000000000000000000000000000000..34e7766a7fa0547b5426d3cdc7755b6e95946beb --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/1d1faee2aebe0d662445de16e655943bbf0508b697b27d5ffb3f0dd6bbe939cf.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d65302dab185a1ec6db48eb2f1438783f9844b5b5eda0ab718df4b34eb3dff7d +size 14857 diff --git a/parse/dev/6u6N8WWwYSM/images/2de69cd8fe5affc54c5eea18f83dffe6076343a4a726d39127761a9440d728d2.jpg b/parse/dev/6u6N8WWwYSM/images/2de69cd8fe5affc54c5eea18f83dffe6076343a4a726d39127761a9440d728d2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2c69a8c87b2048d477e881db07015a9bc8b57b5d --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/2de69cd8fe5affc54c5eea18f83dffe6076343a4a726d39127761a9440d728d2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d84f994edc2f62985fda7f951308a43ff79c56cde96d2a9d349681c05e8f0b4 +size 5573 diff --git a/parse/dev/6u6N8WWwYSM/images/4baebf3db414acb944b4740be2ecea8e45ce61ea41a16fc75f5505a8fb4982f1.jpg b/parse/dev/6u6N8WWwYSM/images/4baebf3db414acb944b4740be2ecea8e45ce61ea41a16fc75f5505a8fb4982f1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..504f36e3dddd98d465d4bdac8653128ed52ed18b --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/4baebf3db414acb944b4740be2ecea8e45ce61ea41a16fc75f5505a8fb4982f1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:263b9aeee707708b8c4ac4ebeb2a758be69df802bc68d99b3605968d88a9b230 +size 16088 diff --git a/parse/dev/6u6N8WWwYSM/images/603366c355d5ea2d62cecf1e19c5c243f2f550ba816503caab91d72d647512c7.jpg b/parse/dev/6u6N8WWwYSM/images/603366c355d5ea2d62cecf1e19c5c243f2f550ba816503caab91d72d647512c7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e3cc9a4598c32607eb1776ad60532dcc52e440df --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/603366c355d5ea2d62cecf1e19c5c243f2f550ba816503caab91d72d647512c7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be85a88e12eaf557d76c54de277d77396b13672d1827bbe469e4b162e9870a61 +size 66320 diff --git a/parse/dev/6u6N8WWwYSM/images/68b0547fe0941acb03fda9253e486c4915d291ecb3bfee98130687549e23e0e2.jpg b/parse/dev/6u6N8WWwYSM/images/68b0547fe0941acb03fda9253e486c4915d291ecb3bfee98130687549e23e0e2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8cb01d31a2a868f9f417995df09d91fea5445d71 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/68b0547fe0941acb03fda9253e486c4915d291ecb3bfee98130687549e23e0e2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:00f22bb35aa4410b4c01c22fc7d157143c534f6d5011fa7e64cd15c652737eb4 +size 12203 diff --git a/parse/dev/6u6N8WWwYSM/images/6a0eedd3a7abbcf3b91e4d92c859b2e3c046c295a8645eb107a586a2f42f5e73.jpg b/parse/dev/6u6N8WWwYSM/images/6a0eedd3a7abbcf3b91e4d92c859b2e3c046c295a8645eb107a586a2f42f5e73.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a37d565e3ae88776a3b182842df7db60a87fc2d5 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/6a0eedd3a7abbcf3b91e4d92c859b2e3c046c295a8645eb107a586a2f42f5e73.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ec54d7cc3958b305e6cb013b0964ab8841de5b4d223b27c03dcd442c2332cce4 +size 102693 diff --git a/parse/dev/6u6N8WWwYSM/images/771442127b8a75cea7cd11e5ae86b9f0e36cdfc473466a89f17776aa3768d645.jpg b/parse/dev/6u6N8WWwYSM/images/771442127b8a75cea7cd11e5ae86b9f0e36cdfc473466a89f17776aa3768d645.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b7eb09bc8dbb49a47c313ed9048a6c58fcefc254 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/771442127b8a75cea7cd11e5ae86b9f0e36cdfc473466a89f17776aa3768d645.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:47f81987fc7abec1aa19b576b5197c2117e3ebb898b11fdcb3134197bbce73cd +size 22029 diff --git a/parse/dev/6u6N8WWwYSM/images/7aa3fd00b4020c4d45e31341769e04cda9730f7f908bd1bec7abfe00ae47168a.jpg b/parse/dev/6u6N8WWwYSM/images/7aa3fd00b4020c4d45e31341769e04cda9730f7f908bd1bec7abfe00ae47168a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a2dba81b8c2677db3bec79405b038e1699a574e0 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/7aa3fd00b4020c4d45e31341769e04cda9730f7f908bd1bec7abfe00ae47168a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:80bc676c52376d17c7b9d65a5dbb3d4c219a209cb0b9dad65bc47d145771ef20 +size 23090 diff --git a/parse/dev/6u6N8WWwYSM/images/809972c8991576f173e37564b9603381a78cd20d01cc70b40de50090bdd9171b.jpg b/parse/dev/6u6N8WWwYSM/images/809972c8991576f173e37564b9603381a78cd20d01cc70b40de50090bdd9171b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ddeb35f821b2d95d386299f060e28fc729c0ab8f --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/809972c8991576f173e37564b9603381a78cd20d01cc70b40de50090bdd9171b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cbdb4c8a72b73d9f02f948cf82079d9394bf9edb64d32a716c3a0b828e5071c2 +size 62227 diff --git a/parse/dev/6u6N8WWwYSM/images/87a4765d9dd11e7bc7baebbec9ccec6dba349253e537720cdd9a860d8bfadc80.jpg b/parse/dev/6u6N8WWwYSM/images/87a4765d9dd11e7bc7baebbec9ccec6dba349253e537720cdd9a860d8bfadc80.jpg new file mode 100644 index 0000000000000000000000000000000000000000..464dc9a9ab6aa9109850bcf0ab6953bd376a8d83 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/87a4765d9dd11e7bc7baebbec9ccec6dba349253e537720cdd9a860d8bfadc80.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a001eff7501f5a392040e32087175337530d30ee9942d59231515201bbc27a0b +size 22157 diff --git a/parse/dev/6u6N8WWwYSM/images/af38e8ba44a2c1b5d9a0e115e404f68144368852fbabe582024d6a7fc0f8fde5.jpg b/parse/dev/6u6N8WWwYSM/images/af38e8ba44a2c1b5d9a0e115e404f68144368852fbabe582024d6a7fc0f8fde5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..45006a01d2b10b36ab9dd35dec18172cb77115d3 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/af38e8ba44a2c1b5d9a0e115e404f68144368852fbabe582024d6a7fc0f8fde5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b7e81d64bd469881f0b26b141c5e9e1d7f137931069880b09613d4870a06ee71 +size 11831 diff --git a/parse/dev/6u6N8WWwYSM/images/b5fe9e69bd192e091d00e294308b8c1874870f5345a5f8c9859125ce40a6ca8c.jpg b/parse/dev/6u6N8WWwYSM/images/b5fe9e69bd192e091d00e294308b8c1874870f5345a5f8c9859125ce40a6ca8c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e16bd4a7d40e9f32cc5d2c73b47598cae0088b8a --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/b5fe9e69bd192e091d00e294308b8c1874870f5345a5f8c9859125ce40a6ca8c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5314bfbd2370f25c32455cba241c2d69b53b244ec3c144fef4161ee3309d31bb +size 182882 diff --git a/parse/dev/6u6N8WWwYSM/images/b85969c2bd36442199a9700f045f97743d090da7c225d4735d013e2dd9a957c8.jpg b/parse/dev/6u6N8WWwYSM/images/b85969c2bd36442199a9700f045f97743d090da7c225d4735d013e2dd9a957c8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..61a7d1e80cdb3eb9cb251c7468f3a9543e0704c0 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/b85969c2bd36442199a9700f045f97743d090da7c225d4735d013e2dd9a957c8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4681841b7bb8299962eef795b53d2d0f6ee06aa6130b399606bbfe1a9a93997 +size 24212 diff --git a/parse/dev/6u6N8WWwYSM/images/b8a41e6882e35596dfde824f3e6ffcf816d7c94c2b55f9e11bfc357a683da7f7.jpg b/parse/dev/6u6N8WWwYSM/images/b8a41e6882e35596dfde824f3e6ffcf816d7c94c2b55f9e11bfc357a683da7f7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1a8434dad93530608161794ae5f012f27c9c6922 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/b8a41e6882e35596dfde824f3e6ffcf816d7c94c2b55f9e11bfc357a683da7f7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f9171105ba5aa2edaabae72e74272c9c2b3d8148e43fda0f2a3b2667043f6cd1 +size 132472 diff --git a/parse/dev/6u6N8WWwYSM/images/b9475f7cb7f0241ddf150f447da1afbbb875d847ca5e16caf9d24fe874152f74.jpg b/parse/dev/6u6N8WWwYSM/images/b9475f7cb7f0241ddf150f447da1afbbb875d847ca5e16caf9d24fe874152f74.jpg new file mode 100644 index 0000000000000000000000000000000000000000..af34ae09086341bcd7ab0334cb34466a8a70102f --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/b9475f7cb7f0241ddf150f447da1afbbb875d847ca5e16caf9d24fe874152f74.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:38819ee36f6b249c0c58387ced12c04e79135343e030a6bbb6146e33b65cb7ca +size 137438 diff --git a/parse/dev/6u6N8WWwYSM/images/c392360a05db92b63807900471b937e8fe659774ca8eca14c432b20af0fd9b6c.jpg b/parse/dev/6u6N8WWwYSM/images/c392360a05db92b63807900471b937e8fe659774ca8eca14c432b20af0fd9b6c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..aff4d643972cb543248eb1353212ca958eb045bf --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/c392360a05db92b63807900471b937e8fe659774ca8eca14c432b20af0fd9b6c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3338a88dcb9e131d55cffdbb2c3df3a8c7120a2dbdfc89549b7d4edce616ec99 +size 31098 diff --git a/parse/dev/6u6N8WWwYSM/images/ccedc51e47d33c2e43fe410c17e67c255fd12c5f5f87bb340a2f90fbc047f49c.jpg b/parse/dev/6u6N8WWwYSM/images/ccedc51e47d33c2e43fe410c17e67c255fd12c5f5f87bb340a2f90fbc047f49c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7b2165e8a093d695f4e582d80cb2abb85f737faf --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/ccedc51e47d33c2e43fe410c17e67c255fd12c5f5f87bb340a2f90fbc047f49c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f04920449f94364c219d3464307c523246392f2ad555146ea83a1b3c282ec1d +size 4888 diff --git a/parse/dev/6u6N8WWwYSM/images/d1528c733884e8502f3353f61a0c9b71cc1164ca318641422551372d3bc937f2.jpg b/parse/dev/6u6N8WWwYSM/images/d1528c733884e8502f3353f61a0c9b71cc1164ca318641422551372d3bc937f2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..881033416540a7530ec360c34d24aac0f652a32f --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/d1528c733884e8502f3353f61a0c9b71cc1164ca318641422551372d3bc937f2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e98fbcd663d680dbaaf61b6aafdb8e8f2f3068fbd8d957e6c5715e6657280b6 +size 40432 diff --git a/parse/dev/6u6N8WWwYSM/images/d7a0d4015766f4796dde34cff6b6e3cc473bbe0377ad9e6741668a650b58f0e5.jpg b/parse/dev/6u6N8WWwYSM/images/d7a0d4015766f4796dde34cff6b6e3cc473bbe0377ad9e6741668a650b58f0e5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..190a1279f7bdf44b125adaf0be17fcc7c4f68fe1 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/d7a0d4015766f4796dde34cff6b6e3cc473bbe0377ad9e6741668a650b58f0e5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc1565e8c14fdc77386eaf9e8bb991fa870b959a6205e31c06bee52dac904178 +size 216265 diff --git a/parse/dev/6u6N8WWwYSM/images/e86a8ea56e0e97c7c3c88b45f168d27a4fb399e303820baa1abb18992e4612a5.jpg b/parse/dev/6u6N8WWwYSM/images/e86a8ea56e0e97c7c3c88b45f168d27a4fb399e303820baa1abb18992e4612a5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e42ebbf3e9524c541391f812fa0a56a0427ed2a9 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/e86a8ea56e0e97c7c3c88b45f168d27a4fb399e303820baa1abb18992e4612a5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be30c32b6c1302c72efc2d4e5c0da6332e7d9b8293dc2863f4f125d78d968e37 +size 41667 diff --git a/parse/dev/6u6N8WWwYSM/images/f6a481df57ef8250e889be80e66df9374a91bf090642c75370927ce0b0489792.jpg b/parse/dev/6u6N8WWwYSM/images/f6a481df57ef8250e889be80e66df9374a91bf090642c75370927ce0b0489792.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7e9025130a3e6f1280621db5be6dcdda7a186b88 --- /dev/null +++ b/parse/dev/6u6N8WWwYSM/images/f6a481df57ef8250e889be80e66df9374a91bf090642c75370927ce0b0489792.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a5b5f2047660348bef0c6f551898985a79235199f55f0531650d960d3e4e8eb +size 8655 diff --git a/parse/dev/716PvHoDct/716PvHoDct_model.json b/parse/dev/716PvHoDct/716PvHoDct_model.json new file mode 100644 index 0000000000000000000000000000000000000000..251552c863dfccc7066b67334ddcdb9a124bc2b7 --- /dev/null +++ b/parse/dev/716PvHoDct/716PvHoDct_model.json @@ -0,0 +1,42467 @@ +[ + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 397, + 764, + 1306, + 764, + 1306, + 1312, + 397, + 1312 + ], + "score": 0.983 + }, + { + "category_id": 3, + "poly": [ + 303, + 1361, + 1387, + 1361, + 1387, + 1639, + 303, + 1639 + ], + "score": 0.968 + }, + { + "category_id": 4, + "poly": [ + 297, + 1660, + 1405, + 1660, + 1405, + 1782, + 297, + 1782 + ], + "score": 0.957 + }, + { + "category_id": 0, + "poly": [ + 368, + 270, + 1330, + 270, + 1330, + 380, + 368, + 380 + ], + "score": 0.953 + }, + { + "category_id": 1, + "poly": [ + 317, + 485, + 1436, + 485, + 1436, + 617, + 317, + 617 + ], + "score": 0.939 + }, + { + "category_id": 1, + "poly": [ + 299, + 1871, + 1403, + 1871, + 1403, + 1934, + 299, + 1934 + ], + "score": 0.934 + }, + { + "category_id": 0, + "poly": [ + 299, + 1818, + 531, + 1818, + 531, + 1855, + 299, + 1855 + ], + "score": 0.896 + }, + { + "category_id": 2, + "poly": [ + 297, + 2033, + 1071, + 2033, + 1071, + 2062, + 297, + 2062 + ], + "score": 0.887 + }, + { + "category_id": 2, + "poly": [ + 332, + 1954, + 777, + 1954, + 777, + 1983, + 332, + 1983 + ], + "score": 0.886 + }, + { + "category_id": 0, + "poly": [ + 787, + 695, + 913, + 695, + 913, + 733, + 787, + 733 + ], + "score": 0.883 + }, + { + "category_id": 1, + "poly": [ + 566, + 638, + 1129, + 638, + 1129, + 675, + 566, + 675 + ], + "score": 0.869 + }, + { + "category_id": 0, + "poly": [ + 314, + 489, + 449, + 489, + 449, + 522, + 314, + 522 + ], + "score": 0.238 + }, + { + "category_id": 13, + "poly": [ + 526, + 1160, + 602, + 1160, + 602, + 1189, + 526, + 1189 + ], + "score": 0.87, + "latex": "1 0 . 7 \\%" + }, + { + "category_id": 13, + "poly": [ + 1080, + 1130, + 1135, + 1130, + 1135, + 1159, + 1080, + 1159 + ], + "score": 0.86, + "latex": "10 \\%" + }, + { + "category_id": 13, + "poly": [ + 1205, + 1160, + 1301, + 1160, + 1301, + 1191, + 1205, + 1191 + ], + "score": 0.81, + "latex": "\\mathrm { O P T } _ { 6 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 814, + 1130, + 909, + 1130, + 909, + 1162, + 814, + 1162 + ], + "score": 0.77, + "latex": "\\mathrm { O P T } _ { 6 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 591, + 1130, + 687, + 1130, + 687, + 1161, + 591, + 1161 + ], + "score": 0.76, + "latex": "\\mathrm { O P T } _ { 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 486, + 522, + 596, + 522, + 596, + 551, + 486, + 551 + ], + "score": 0.56, + "latex": "\\mathbf { N E x T + + }" + }, + { + "category_id": 13, + "poly": [ + 696, + 487, + 773, + 487, + 773, + 520, + 696, + 520 + ], + "score": 0.26, + "latex": "\\mathbf { Y a o } ^ { 2 * }" + }, + { + "category_id": 15, + "poly": [ + 866.0, + 1361.0, + 952.0, + 1361.0, + 952.0, + 1388.0, + 866.0, + 1388.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 1385.0, + 892.0, + 1385.0, + 892.0, + 1394.0, + 882.0, + 1394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 926.0, + 1383.0, + 936.0, + 1383.0, + 936.0, + 1393.0, + 926.0, + 1393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 819.0, + 1391.0, + 868.0, + 1391.0, + 868.0, + 1421.0, + 819.0, + 1421.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 921.0, + 1394.0, + 936.0, + 1394.0, + 936.0, + 1402.0, + 921.0, + 1402.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 947.0, + 1391.0, + 1000.0, + 1391.0, + 1000.0, + 1420.0, + 947.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1009.0, + 1395.0, + 1026.0, + 1395.0, + 1026.0, + 1420.0, + 1009.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 1395.0, + 1143.0, + 1395.0, + 1143.0, + 1422.0, + 1028.0, + 1422.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 723.0, + 1422.0, + 755.0, + 1422.0, + 755.0, + 1452.0, + 723.0, + 1452.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 924.0, + 1422.0, + 930.0, + 1422.0, + 930.0, + 1428.0, + 924.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1434.0, + 401.0, + 1434.0, + 401.0, + 1459.0, + 354.0, + 1459.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 497.0, + 1435.0, + 574.0, + 1435.0, + 574.0, + 1462.0, + 497.0, + 1462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 576.0, + 1440.0, + 591.0, + 1440.0, + 591.0, + 1456.0, + 576.0, + 1456.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 669.0, + 1431.0, + 719.0, + 1431.0, + 719.0, + 1460.0, + 669.0, + 1460.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 1441.0, + 849.0, + 1441.0, + 849.0, + 1450.0, + 838.0, + 1450.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 923.0, + 1433.0, + 931.0, + 1433.0, + 931.0, + 1442.0, + 923.0, + 1442.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 967.0, + 1442.0, + 981.0, + 1442.0, + 981.0, + 1450.0, + 967.0, + 1450.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1454.0, + 405.0, + 1454.0, + 405.0, + 1483.0, + 348.0, + 1483.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 511.0, + 1454.0, + 561.0, + 1454.0, + 561.0, + 1483.0, + 511.0, + 1483.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 666.0, + 1449.0, + 722.0, + 1449.0, + 722.0, + 1482.0, + 666.0, + 1482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 804.0, + 1455.0, + 893.0, + 1455.0, + 893.0, + 1488.0, + 804.0, + 1488.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 921.0, + 1459.0, + 931.0, + 1459.0, + 931.0, + 1467.0, + 921.0, + 1467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 933.0, + 1452.0, + 1015.0, + 1452.0, + 1015.0, + 1489.0, + 933.0, + 1489.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1052.0, + 1452.0, + 1127.0, + 1452.0, + 1127.0, + 1479.0, + 1052.0, + 1479.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 883.0, + 1473.0, + 893.0, + 1473.0, + 893.0, + 1482.0, + 883.0, + 1482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 301.0, + 1489.0, + 447.0, + 1489.0, + 447.0, + 1537.0, + 301.0, + 1537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 491.0, + 1493.0, + 577.0, + 1493.0, + 577.0, + 1524.0, + 491.0, + 1524.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 663.0, + 1493.0, + 727.0, + 1493.0, + 727.0, + 1527.0, + 663.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1053.0, + 1504.0, + 1134.0, + 1504.0, + 1134.0, + 1541.0, + 1053.0, + 1541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 928.0, + 1550.0, + 934.0, + 1550.0, + 934.0, + 1556.0, + 928.0, + 1556.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.0, + 1565.0, + 894.0, + 1565.0, + 894.0, + 1596.0, + 793.0, + 1596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 923.0, + 1565.0, + 1025.0, + 1565.0, + 1025.0, + 1596.0, + 923.0, + 1596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 1587.0, + 665.0, + 1587.0, + 665.0, + 1617.0, + 395.0, + 1617.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 966.0, + 1599.0, + 1225.0, + 1599.0, + 1225.0, + 1626.0, + 966.0, + 1626.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 382.0, + 1612.0, + 676.0, + 1612.0, + 676.0, + 1639.0, + 382.0, + 1639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1154.0, + 1387.0, + 1234.0, + 1387.0, + 1234.0, + 1420.0, + 1154.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1284.0, + 1384.5, + 1363.0, + 1384.5, + 1363.0, + 1421.5, + 1284.0, + 1421.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 880.0, + 1407.5, + 896.0, + 1407.5, + 896.0, + 1419.5, + 880.0, + 1419.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1150.0, + 1445.5, + 1240.0, + 1445.5, + 1240.0, + 1481.0, + 1150.0, + 1481.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1281.0, + 1445.5, + 1373.0, + 1445.5, + 1373.0, + 1481.0, + 1281.0, + 1481.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 1488.0, + 893.0, + 1488.0, + 893.0, + 1492.0, + 882.0, + 1492.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1150.0, + 1503.0, + 1240.0, + 1503.0, + 1240.0, + 1539.5, + 1150.0, + 1539.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1282.0, + 1502.5, + 1371.0, + 1502.5, + 1371.0, + 1539.0, + 1282.0, + 1539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 804.0, + 1515.5, + 882.0, + 1515.5, + 882.0, + 1547.5, + 804.0, + 1547.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 932.25, + 1514.0, + 1014.25, + 1514.0, + 1014.25, + 1550.0, + 932.25, + 1550.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1657.0, + 1406.0, + 1657.0, + 1406.0, + 1698.0, + 292.0, + 1698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1690.0, + 1406.0, + 1690.0, + 1406.0, + 1726.0, + 291.0, + 1726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1721.0, + 1406.0, + 1721.0, + 1406.0, + 1757.0, + 294.0, + 1757.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1746.0, + 1382.0, + 1746.0, + 1382.0, + 1789.0, + 291.0, + 1789.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 361.0, + 265.0, + 1339.0, + 265.0, + 1339.0, + 332.0, + 361.0, + 332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 708.0, + 331.0, + 989.0, + 331.0, + 989.0, + 378.0, + 708.0, + 378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1814.0, + 535.0, + 1814.0, + 535.0, + 1861.0, + 292.0, + 1861.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2029.0, + 1074.0, + 2029.0, + 1074.0, + 2067.0, + 294.0, + 2067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 1951.0, + 777.0, + 1951.0, + 777.0, + 1987.0, + 333.0, + 1987.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 782.0, + 692.0, + 920.0, + 692.0, + 920.0, + 738.0, + 782.0, + 738.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 309.0, + 485.0, + 456.0, + 485.0, + 456.0, + 528.0, + 309.0, + 528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 764.0, + 1310.0, + 764.0, + 1310.0, + 801.0, + 394.0, + 801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 794.0, + 1308.0, + 794.0, + 1308.0, + 833.0, + 394.0, + 833.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 825.0, + 1310.0, + 825.0, + 1310.0, + 863.0, + 392.0, + 863.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 856.0, + 1308.0, + 856.0, + 1308.0, + 892.0, + 393.0, + 892.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 888.0, + 1310.0, + 888.0, + 1310.0, + 921.0, + 394.0, + 921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 918.0, + 1307.0, + 918.0, + 1307.0, + 951.0, + 394.0, + 951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 948.0, + 1308.0, + 948.0, + 1308.0, + 980.0, + 393.0, + 980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 976.0, + 1307.0, + 976.0, + 1307.0, + 1014.0, + 392.0, + 1014.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1007.0, + 1306.0, + 1007.0, + 1306.0, + 1043.0, + 393.0, + 1043.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1038.0, + 1306.0, + 1038.0, + 1306.0, + 1074.0, + 392.0, + 1074.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1068.0, + 1305.0, + 1068.0, + 1305.0, + 1103.0, + 393.0, + 1103.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1098.0, + 1308.0, + 1098.0, + 1308.0, + 1134.0, + 393.0, + 1134.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1128.0, + 590.0, + 1128.0, + 590.0, + 1165.0, + 392.0, + 1165.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 688.0, + 1128.0, + 813.0, + 1128.0, + 813.0, + 1165.0, + 688.0, + 1165.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 910.0, + 1128.0, + 1079.0, + 1128.0, + 1079.0, + 1165.0, + 910.0, + 1165.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1136.0, + 1128.0, + 1306.0, + 1128.0, + 1306.0, + 1165.0, + 1136.0, + 1165.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1156.0, + 525.0, + 1156.0, + 525.0, + 1197.0, + 392.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 603.0, + 1156.0, + 1204.0, + 1156.0, + 1204.0, + 1197.0, + 603.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1189.0, + 1306.0, + 1189.0, + 1306.0, + 1226.0, + 393.0, + 1226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 391.0, + 1219.0, + 1305.0, + 1219.0, + 1305.0, + 1254.0, + 391.0, + 1254.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1247.0, + 1306.0, + 1247.0, + 1306.0, + 1287.0, + 392.0, + 1287.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1281.0, + 1185.0, + 1281.0, + 1185.0, + 1314.0, + 394.0, + 1314.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 311.0, + 487.0, + 455.0, + 487.0, + 455.0, + 527.0, + 311.0, + 527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 478.0, + 487.0, + 695.0, + 487.0, + 695.0, + 522.0, + 478.0, + 522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 774.0, + 487.0, + 793.0, + 487.0, + 793.0, + 522.0, + 774.0, + 522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 798.0, + 487.0, + 1221.0, + 487.0, + 1221.0, + 523.0, + 798.0, + 523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1233.0, + 487.0, + 1439.0, + 487.0, + 1439.0, + 526.0, + 1233.0, + 526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 469.0, + 515.0, + 485.0, + 515.0, + 485.0, + 562.0, + 469.0, + 562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 597.0, + 515.0, + 1285.0, + 515.0, + 1285.0, + 562.0, + 597.0, + 562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 468.0, + 547.0, + 1284.0, + 547.0, + 1284.0, + 593.0, + 468.0, + 593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 430.0, + 582.0, + 1322.0, + 582.0, + 1322.0, + 622.0, + 430.0, + 622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1871.0, + 1404.0, + 1871.0, + 1404.0, + 1906.0, + 295.0, + 1906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1901.0, + 1404.0, + 1901.0, + 1404.0, + 1936.0, + 295.0, + 1936.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 566.0, + 635.0, + 1133.0, + 635.0, + 1133.0, + 679.0, + 566.0, + 679.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 0, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1097, + 1405, + 1097, + 1405, + 1374, + 297, + 1374 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 297, + 202, + 1405, + 202, + 1405, + 477, + 297, + 477 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 297, + 495, + 1404, + 495, + 1404, + 710, + 297, + 710 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 331, + 1467, + 1406, + 1467, + 1406, + 1758, + 331, + 1758 + ], + "score": 0.96 + }, + { + "category_id": 1, + "poly": [ + 330, + 834, + 1405, + 834, + 1405, + 1085, + 330, + 1085 + ], + "score": 0.955 + }, + { + "category_id": 1, + "poly": [ + 303, + 729, + 1399, + 729, + 1399, + 822, + 303, + 822 + ], + "score": 0.954 + }, + { + "category_id": 1, + "poly": [ + 300, + 1392, + 1403, + 1392, + 1403, + 1454, + 300, + 1454 + ], + "score": 0.942 + }, + { + "category_id": 2, + "poly": [ + 296, + 1831, + 1406, + 1831, + 1406, + 2008, + 296, + 2008 + ], + "score": 0.898 + }, + { + "category_id": 1, + "poly": [ + 302, + 1772, + 1403, + 1772, + 1403, + 1806, + 302, + 1806 + ], + "score": 0.858 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.726 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.204 + }, + { + "category_id": 13, + "poly": [ + 715, + 1282, + 804, + 1282, + 804, + 1311, + 715, + 1311 + ], + "score": 0.91, + "latex": "( 5 \\times l r )" + }, + { + "category_id": 13, + "poly": [ + 634, + 1661, + 712, + 1661, + 712, + 1691, + 634, + 1691 + ], + "score": 0.89, + "latex": "1 0 . 7 \\%" + }, + { + "category_id": 13, + "poly": [ + 992, + 1629, + 1148, + 1629, + 1148, + 1662, + 992, + 1662 + ], + "score": 0.88, + "latex": "\\mathbf { O P T _ { 2 . 7 B 6 . 7 B } }" + }, + { + "category_id": 13, + "poly": [ + 644, + 1727, + 700, + 1727, + 700, + 1756, + 644, + 1756 + ], + "score": 0.84, + "latex": "\\mathbf { + 2 . 9 }" + }, + { + "category_id": 13, + "poly": [ + 353, + 1660, + 411, + 1660, + 411, + 1691, + 353, + 1691 + ], + "score": 0.83, + "latex": "10 \\%" + }, + { + "category_id": 13, + "poly": [ + 1048, + 1727, + 1104, + 1727, + 1104, + 1756, + 1048, + 1756 + ], + "score": 0.82, + "latex": "\\mathbf { + 3 . 4 }" + }, + { + "category_id": 13, + "poly": [ + 1249, + 866, + 1346, + 866, + 1346, + 897, + 1249, + 897 + ], + "score": 0.64, + "latex": "\\mathrm { O P T } _ { 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 1139, + 1696, + 1234, + 1696, + 1234, + 1728, + 1139, + 1728 + ], + "score": 0.64, + "latex": "\\mathrm { O P T } _ { 6 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 719, + 897, + 814, + 897, + 814, + 928, + 719, + 928 + ], + "score": 0.61, + "latex": "\\mathrm { O P T } _ { 6 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 332, + 1917, + 362, + 1917, + 362, + 1946, + 332, + 1946 + ], + "score": 0.56, + "latex": "^ 4 \\mathrm { I t }" + }, + { + "category_id": 13, + "poly": [ + 297, + 1423, + 328, + 1423, + 328, + 1455, + 297, + 1455 + ], + "score": 0.46, + "latex": "\\ S 4" + }, + { + "category_id": 13, + "poly": [ + 818, + 1099, + 872, + 1099, + 872, + 1131, + 818, + 1131 + ], + "score": 0.4, + "latex": "\\ S 3 . 1 \\ r ," + }, + { + "category_id": 13, + "poly": [ + 1205, + 1129, + 1304, + 1129, + 1304, + 1162, + 1205, + 1162 + ], + "score": 0.39, + "latex": "( c f . \\ \\ S 3 . 2 )" + }, + { + "category_id": 13, + "poly": [ + 1041, + 527, + 1102, + 527, + 1102, + 559, + 1041, + 559 + ], + "score": 0.32, + "latex": "{ \\mathrm { 7 } } 5 _ { \\mathrm { X X L } }" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 1827.0, + 923.0, + 1827.0, + 923.0, + 1865.0, + 329.0, + 1865.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 328.0, + 1854.0, + 1406.0, + 1854.0, + 1406.0, + 1897.0, + 328.0, + 1897.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1889.0, + 710.0, + 1889.0, + 710.0, + 1919.0, + 296.0, + 1919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 327.0, + 1913.0, + 331.0, + 1913.0, + 331.0, + 1954.0, + 327.0, + 1954.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1913.0, + 1407.0, + 1913.0, + 1407.0, + 1954.0, + 363.0, + 1954.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1945.0, + 573.0, + 1945.0, + 573.0, + 1981.0, + 295.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 1971.0, + 1041.0, + 1971.0, + 1041.0, + 2011.0, + 329.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2093.0, + 838.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2093.0, + 838.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1098.0, + 817.0, + 1098.0, + 817.0, + 1135.0, + 295.0, + 1135.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 873.0, + 1098.0, + 1408.0, + 1098.0, + 1408.0, + 1135.0, + 873.0, + 1135.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1129.0, + 1204.0, + 1129.0, + 1204.0, + 1166.0, + 294.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1305.0, + 1129.0, + 1405.0, + 1129.0, + 1405.0, + 1166.0, + 1305.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1160.0, + 1408.0, + 1160.0, + 1408.0, + 1196.0, + 295.0, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1184.0, + 1405.0, + 1184.0, + 1405.0, + 1229.0, + 292.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1217.0, + 1405.0, + 1217.0, + 1405.0, + 1255.0, + 292.0, + 1255.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1251.0, + 1405.0, + 1251.0, + 1405.0, + 1284.0, + 296.0, + 1284.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1280.0, + 714.0, + 1280.0, + 714.0, + 1317.0, + 294.0, + 1317.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 805.0, + 1280.0, + 1405.0, + 1280.0, + 1405.0, + 1317.0, + 805.0, + 1317.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1307.0, + 1406.0, + 1307.0, + 1406.0, + 1347.0, + 294.0, + 1347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1343.0, + 773.0, + 1343.0, + 773.0, + 1378.0, + 292.0, + 1378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 203.0, + 1408.0, + 203.0, + 1408.0, + 239.0, + 295.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 233.0, + 1407.0, + 233.0, + 1407.0, + 269.0, + 294.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 261.0, + 1406.0, + 261.0, + 1406.0, + 300.0, + 294.0, + 300.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 294.0, + 1403.0, + 294.0, + 1403.0, + 330.0, + 295.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 325.0, + 1407.0, + 325.0, + 1407.0, + 362.0, + 294.0, + 362.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 354.0, + 1406.0, + 354.0, + 1406.0, + 390.0, + 294.0, + 390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 385.0, + 1403.0, + 385.0, + 1403.0, + 418.0, + 295.0, + 418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 415.0, + 1405.0, + 415.0, + 1405.0, + 451.0, + 294.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 444.0, + 410.0, + 444.0, + 410.0, + 478.0, + 293.0, + 478.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 493.0, + 1404.0, + 493.0, + 1404.0, + 534.0, + 294.0, + 534.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 525.0, + 1040.0, + 525.0, + 1040.0, + 563.0, + 294.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1103.0, + 525.0, + 1405.0, + 525.0, + 1405.0, + 563.0, + 1103.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 556.0, + 1407.0, + 556.0, + 1407.0, + 593.0, + 292.0, + 593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 587.0, + 1406.0, + 587.0, + 1406.0, + 625.0, + 294.0, + 625.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 615.0, + 1405.0, + 615.0, + 1405.0, + 653.0, + 292.0, + 653.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 648.0, + 1406.0, + 648.0, + 1406.0, + 682.0, + 295.0, + 682.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 679.0, + 955.0, + 679.0, + 955.0, + 713.0, + 295.0, + 713.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 1465.0, + 1406.0, + 1465.0, + 1406.0, + 1504.0, + 329.0, + 1504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1499.0, + 1406.0, + 1499.0, + 1406.0, + 1533.0, + 353.0, + 1533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1532.0, + 418.0, + 1532.0, + 418.0, + 1566.0, + 352.0, + 1566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 1561.0, + 1403.0, + 1561.0, + 1403.0, + 1600.0, + 329.0, + 1600.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1592.0, + 1294.0, + 1592.0, + 1294.0, + 1632.0, + 350.0, + 1632.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 330.0, + 1630.0, + 991.0, + 1630.0, + 991.0, + 1667.0, + 330.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1149.0, + 1630.0, + 1406.0, + 1630.0, + 1406.0, + 1667.0, + 1149.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1656.0, + 352.0, + 1656.0, + 352.0, + 1699.0, + 348.0, + 1699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 412.0, + 1656.0, + 633.0, + 1656.0, + 633.0, + 1699.0, + 412.0, + 1699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 713.0, + 1656.0, + 1354.0, + 1656.0, + 1354.0, + 1699.0, + 713.0, + 1699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 1692.0, + 1138.0, + 1692.0, + 1138.0, + 1733.0, + 329.0, + 1733.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1235.0, + 1692.0, + 1407.0, + 1692.0, + 1407.0, + 1733.0, + 1235.0, + 1733.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1727.0, + 643.0, + 1727.0, + 643.0, + 1762.0, + 354.0, + 1762.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 701.0, + 1727.0, + 1047.0, + 1727.0, + 1047.0, + 1762.0, + 701.0, + 1762.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1105.0, + 1727.0, + 1406.0, + 1727.0, + 1406.0, + 1762.0, + 1105.0, + 1762.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 330.0, + 834.0, + 1406.0, + 834.0, + 1406.0, + 872.0, + 330.0, + 872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 864.0, + 1248.0, + 864.0, + 1248.0, + 902.0, + 353.0, + 902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1347.0, + 864.0, + 1407.0, + 864.0, + 1407.0, + 902.0, + 1347.0, + 902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 897.0, + 718.0, + 897.0, + 718.0, + 931.0, + 355.0, + 931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 815.0, + 897.0, + 1404.0, + 897.0, + 1404.0, + 931.0, + 815.0, + 931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 927.0, + 1406.0, + 927.0, + 1406.0, + 961.0, + 351.0, + 961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 956.0, + 892.0, + 956.0, + 892.0, + 994.0, + 351.0, + 994.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 330.0, + 989.0, + 1406.0, + 989.0, + 1406.0, + 1030.0, + 330.0, + 1030.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1018.0, + 1410.0, + 1018.0, + 1410.0, + 1061.0, + 351.0, + 1061.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1055.0, + 1397.0, + 1055.0, + 1397.0, + 1085.0, + 356.0, + 1085.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 727.0, + 1404.0, + 727.0, + 1404.0, + 763.0, + 295.0, + 763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 756.0, + 1404.0, + 756.0, + 1404.0, + 795.0, + 295.0, + 795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 790.0, + 892.0, + 790.0, + 892.0, + 824.0, + 298.0, + 824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1388.0, + 1408.0, + 1388.0, + 1408.0, + 1429.0, + 294.0, + 1429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 1422.0, + 857.0, + 1422.0, + 857.0, + 1457.0, + 329.0, + 1457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1767.0, + 1408.0, + 1767.0, + 1408.0, + 1811.0, + 296.0, + 1811.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 1, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 300, + 1593, + 1404, + 1593, + 1404, + 1772, + 300, + 1772 + ], + "score": 0.978, + "html": "
MLLMsVPGVPG TrainableLLMLLMTrainable
KOSMOS-1[19]CLIP [43]AllRand.Init.LMAll
Frozen [54]NF-ResNet-50 [6]NF-ResNet-50GPT-2-like† [42]No
Flamingo [2]NFNet-F6 [6]+Resampler [21]ResamplerChinchilla [18]Xattn-Dense
PaLM-E[16]ViT[15]/OSRT[49]AllPaLM[11]No
BLIP-2 [30]EVA-CLIP [52] + Q-Former [30]Q-FormerOPT[60]/Flan-T5[12]No
" + }, + { + "category_id": 1, + "poly": [ + 297, + 1310, + 1406, + 1310, + 1406, + 1492, + 297, + 1492 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 297, + 1792, + 1404, + 1792, + 1404, + 2008, + 297, + 2008 + ], + "score": 0.968 + }, + { + "category_id": 1, + "poly": [ + 331, + 837, + 1405, + 837, + 1405, + 1095, + 331, + 1095 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 300, + 702, + 1405, + 702, + 1405, + 824, + 300, + 824 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 302, + 1187, + 1402, + 1187, + 1402, + 1248, + 302, + 1248 + ], + "score": 0.953 + }, + { + "category_id": 1, + "poly": [ + 330, + 526, + 1405, + 526, + 1405, + 687, + 330, + 687 + ], + "score": 0.94 + }, + { + "category_id": 3, + "poly": [ + 306, + 199, + 1396, + 199, + 1396, + 420, + 306, + 420 + ], + "score": 0.933 + }, + { + "category_id": 4, + "poly": [ + 300, + 443, + 1406, + 443, + 1406, + 506, + 300, + 506 + ], + "score": 0.929 + }, + { + "category_id": 6, + "poly": [ + 298, + 1514, + 1397, + 1514, + 1397, + 1578, + 298, + 1578 + ], + "score": 0.919 + }, + { + "category_id": 0, + "poly": [ + 297, + 1134, + 522, + 1134, + 522, + 1172, + 297, + 1172 + ], + "score": 0.915 + }, + { + "category_id": 0, + "poly": [ + 297, + 1269, + 454, + 1269, + 454, + 1299, + 297, + 1299 + ], + "score": 0.887 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.655 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.424 + }, + { + "category_id": 13, + "poly": [ + 1270, + 558, + 1351, + 558, + 1351, + 589, + 1270, + 589 + ], + "score": 0.89, + "latex": "L L M _ { s r c }" + }, + { + "category_id": 13, + "poly": [ + 1015, + 528, + 1102, + 528, + 1102, + 558, + 1015, + 558 + ], + "score": 0.88, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 794, + 528, + 880, + 528, + 880, + 562, + 794, + 562 + ], + "score": 0.88, + "latex": "\\mathrm { L L M _ { \\mathrm { t g t } } }" + }, + { + "category_id": 13, + "poly": [ + 675, + 528, + 762, + 528, + 762, + 559, + 675, + 559 + ], + "score": 0.87, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 1102, + 475, + 1252, + 475, + 1252, + 506, + 1102, + 506 + ], + "score": 0.86, + "latex": "\\mathrm { O P T } _ { 2 . 7 \\mathrm { B } 6 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 957, + 476, + 1049, + 476, + 1049, + 505, + 957, + 505 + ], + "score": 0.72, + "latex": "\\bf { \\Phi } _ { \\mathrm { X L \\to X X L } }" + }, + { + "category_id": 13, + "poly": [ + 1304, + 655, + 1363, + 655, + 1363, + 687, + 1304, + 687 + ], + "score": 0.35, + "latex": "\\ S 5 . 2 )" + }, + { + "category_id": 13, + "poly": [ + 531, + 212, + 553, + 212, + 553, + 233, + 531, + 233 + ], + "score": 0.34, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 991, + 211, + 1013, + 211, + 1013, + 233, + 991, + 233 + ], + "score": 0.33, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 850, + 589, + 949, + 589, + 949, + 621, + 850, + 621 + ], + "score": 0.32, + "latex": "( c f . \\ \\ S 4 . 2 )" + }, + { + "category_id": 13, + "poly": [ + 399, + 1033, + 433, + 1033, + 433, + 1064, + 399, + 1064 + ], + "score": 0.31, + "latex": "\\ S 6" + }, + { + "category_id": 15, + "poly": [ + 362.0, + 198.0, + 530.0, + 198.0, + 530.0, + 242.0, + 362.0, + 242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 554.0, + 198.0, + 689.0, + 198.0, + 689.0, + 242.0, + 554.0, + 242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 823.0, + 196.0, + 990.0, + 196.0, + 990.0, + 242.0, + 823.0, + 242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1014.0, + 196.0, + 1124.0, + 196.0, + 1124.0, + 242.0, + 1014.0, + 242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 247.0, + 386.0, + 247.0, + 386.0, + 269.0, + 344.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 447.0, + 246.0, + 499.0, + 246.0, + 499.0, + 271.0, + 447.0, + 271.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 558.0, + 247.0, + 599.0, + 247.0, + 599.0, + 269.0, + 558.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.0, + 244.0, + 836.0, + 244.0, + 836.0, + 267.0, + 793.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 893.0, + 241.0, + 945.0, + 241.0, + 945.0, + 267.0, + 893.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 247.0, + 1047.0, + 247.0, + 1047.0, + 269.0, + 1006.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1162.0, + 235.0, + 1392.0, + 235.0, + 1392.0, + 281.0, + 1162.0, + 281.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 642.0, + 271.0, + 703.0, + 271.0, + 703.0, + 292.0, + 642.0, + 292.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1166.0, + 314.0, + 1397.0, + 314.0, + 1397.0, + 350.0, + 1166.0, + 350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 350.0, + 359.0, + 350.0, + 359.0, + 373.0, + 324.0, + 373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 427.0, + 346.0, + 468.0, + 346.0, + 468.0, + 372.0, + 427.0, + 372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 542.0, + 347.0, + 572.0, + 347.0, + 572.0, + 371.0, + 542.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 772.0, + 343.0, + 808.0, + 343.0, + 808.0, + 366.0, + 772.0, + 366.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 872.0, + 343.0, + 917.0, + 343.0, + 917.0, + 370.0, + 872.0, + 370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 984.0, + 345.0, + 1020.0, + 345.0, + 1020.0, + 368.0, + 984.0, + 368.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 301.0, + 373.0, + 407.0, + 373.0, + 407.0, + 402.0, + 301.0, + 402.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 435.0, + 376.0, + 487.0, + 376.0, + 487.0, + 402.0, + 435.0, + 402.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 518.0, + 376.0, + 620.0, + 376.0, + 620.0, + 401.0, + 518.0, + 401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 637.0, + 375.0, + 712.0, + 375.0, + 712.0, + 404.0, + 637.0, + 404.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 751.0, + 373.0, + 858.0, + 373.0, + 858.0, + 402.0, + 751.0, + 402.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 884.0, + 376.0, + 938.0, + 376.0, + 938.0, + 401.0, + 884.0, + 401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 969.0, + 376.0, + 1070.0, + 376.0, + 1070.0, + 401.0, + 969.0, + 401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1089.0, + 376.0, + 1162.0, + 376.0, + 1162.0, + 401.0, + 1089.0, + 401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 387.0, + 393.0, + 387.0, + 393.0, + 424.0, + 315.0, + 424.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 427.0, + 391.0, + 494.0, + 391.0, + 494.0, + 421.0, + 427.0, + 421.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 545.0, + 391.0, + 592.0, + 391.0, + 592.0, + 422.0, + 545.0, + 422.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 640.0, + 390.0, + 708.0, + 390.0, + 708.0, + 424.0, + 640.0, + 424.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 767.0, + 391.0, + 841.0, + 391.0, + 841.0, + 421.0, + 767.0, + 421.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 877.0, + 391.0, + 945.0, + 391.0, + 945.0, + 421.0, + 877.0, + 421.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 995.0, + 391.0, + 1044.0, + 391.0, + 1044.0, + 422.0, + 995.0, + 422.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1091.0, + 390.0, + 1160.0, + 390.0, + 1160.0, + 424.0, + 1091.0, + 424.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1090.0, + 274.0, + 1152.0, + 274.0, + 1152.0, + 304.5, + 1090.0, + 304.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 441.0, + 1404.0, + 441.0, + 1404.0, + 477.0, + 295.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 468.0, + 956.0, + 468.0, + 956.0, + 511.0, + 293.0, + 511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1050.0, + 468.0, + 1101.0, + 468.0, + 1101.0, + 511.0, + 1050.0, + 511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1253.0, + 468.0, + 1410.0, + 468.0, + 1410.0, + 511.0, + 1253.0, + 511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1513.0, + 1403.0, + 1513.0, + 1403.0, + 1549.0, + 294.0, + 1549.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1545.0, + 542.0, + 1545.0, + 542.0, + 1582.0, + 294.0, + 1582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 1128.0, + 527.0, + 1128.0, + 527.0, + 1185.0, + 288.0, + 1185.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1267.0, + 458.0, + 1267.0, + 458.0, + 1306.0, + 293.0, + 1306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1305.0, + 1408.0, + 1305.0, + 1408.0, + 1346.0, + 295.0, + 1346.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1339.0, + 1406.0, + 1339.0, + 1406.0, + 1375.0, + 295.0, + 1375.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1370.0, + 1409.0, + 1370.0, + 1409.0, + 1405.0, + 294.0, + 1405.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1400.0, + 1404.0, + 1400.0, + 1404.0, + 1435.0, + 294.0, + 1435.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1429.0, + 1408.0, + 1429.0, + 1408.0, + 1465.0, + 294.0, + 1465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1458.0, + 977.0, + 1458.0, + 977.0, + 1494.0, + 292.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1794.0, + 1404.0, + 1794.0, + 1404.0, + 1828.0, + 296.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1821.0, + 1409.0, + 1821.0, + 1409.0, + 1862.0, + 292.0, + 1862.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1851.0, + 1408.0, + 1851.0, + 1408.0, + 1891.0, + 294.0, + 1891.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1884.0, + 1405.0, + 1884.0, + 1405.0, + 1920.0, + 294.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1915.0, + 1406.0, + 1915.0, + 1406.0, + 1949.0, + 295.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1946.0, + 1405.0, + 1946.0, + 1405.0, + 1980.0, + 295.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1975.0, + 1405.0, + 1975.0, + 1405.0, + 2010.0, + 296.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 330.0, + 837.0, + 1407.0, + 837.0, + 1407.0, + 875.0, + 330.0, + 875.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 868.0, + 1404.0, + 868.0, + 1404.0, + 903.0, + 353.0, + 903.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 898.0, + 1177.0, + 898.0, + 1177.0, + 935.0, + 352.0, + 935.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 931.0, + 1406.0, + 931.0, + 1406.0, + 973.0, + 331.0, + 973.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 962.0, + 1276.0, + 962.0, + 1276.0, + 1002.0, + 350.0, + 1002.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 334.0, + 1002.0, + 1404.0, + 1002.0, + 1404.0, + 1033.0, + 334.0, + 1033.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1032.0, + 398.0, + 1032.0, + 398.0, + 1067.0, + 354.0, + 1067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 434.0, + 1032.0, + 1405.0, + 1032.0, + 1405.0, + 1067.0, + 434.0, + 1067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1060.0, + 1179.0, + 1060.0, + 1179.0, + 1097.0, + 353.0, + 1097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 702.0, + 1405.0, + 702.0, + 1405.0, + 734.0, + 296.0, + 734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 731.0, + 1406.0, + 731.0, + 1406.0, + 766.0, + 294.0, + 766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 762.0, + 1405.0, + 762.0, + 1405.0, + 799.0, + 293.0, + 799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 794.0, + 456.0, + 794.0, + 456.0, + 824.0, + 295.0, + 824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 1185.0, + 1403.0, + 1185.0, + 1403.0, + 1220.0, + 298.0, + 1220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1218.0, + 736.0, + 1218.0, + 736.0, + 1250.0, + 297.0, + 1250.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 525.0, + 674.0, + 525.0, + 674.0, + 563.0, + 329.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 763.0, + 525.0, + 793.0, + 525.0, + 793.0, + 563.0, + 763.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 881.0, + 525.0, + 1014.0, + 525.0, + 1014.0, + 563.0, + 881.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1103.0, + 525.0, + 1410.0, + 525.0, + 1410.0, + 563.0, + 1103.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 555.0, + 1269.0, + 555.0, + 1269.0, + 594.0, + 351.0, + 594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1352.0, + 555.0, + 1410.0, + 555.0, + 1410.0, + 594.0, + 1352.0, + 594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 590.0, + 849.0, + 590.0, + 849.0, + 623.0, + 353.0, + 623.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 950.0, + 590.0, + 961.0, + 590.0, + 961.0, + 623.0, + 950.0, + 623.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 624.0, + 1404.0, + 624.0, + 1404.0, + 657.0, + 329.0, + 657.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 654.0, + 1303.0, + 654.0, + 1303.0, + 690.0, + 352.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1364.0, + 654.0, + 1374.0, + 654.0, + 1374.0, + 690.0, + 1364.0, + 690.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 2, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1362, + 1405, + 1362, + 1405, + 1666, + 297, + 1666 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 298, + 884, + 1405, + 884, + 1405, + 1068, + 298, + 1068 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 203, + 1405, + 203, + 1405, + 355, + 298, + 355 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 298, + 745, + 1405, + 745, + 1405, + 866, + 298, + 866 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 298, + 605, + 1405, + 605, + 1405, + 727, + 298, + 727 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 299, + 417, + 1404, + 417, + 1404, + 509, + 299, + 509 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 296, + 1164, + 1401, + 1164, + 1401, + 1225, + 296, + 1225 + ], + "score": 0.946 + }, + { + "category_id": 1, + "poly": [ + 297, + 1686, + 1406, + 1686, + 1406, + 1954, + 297, + 1954 + ], + "score": 0.945 + }, + { + "category_id": 1, + "poly": [ + 299, + 1287, + 1404, + 1287, + 1404, + 1349, + 299, + 1349 + ], + "score": 0.938 + }, + { + "category_id": 0, + "poly": [ + 299, + 1110, + 1359, + 1110, + 1359, + 1148, + 299, + 1148 + ], + "score": 0.938 + }, + { + "category_id": 1, + "poly": [ + 294, + 526, + 1403, + 526, + 1403, + 588, + 294, + 588 + ], + "score": 0.936 + }, + { + "category_id": 2, + "poly": [ + 333, + 1979, + 801, + 1979, + 801, + 2007, + 333, + 2007 + ], + "score": 0.917 + }, + { + "category_id": 0, + "poly": [ + 299, + 376, + 604, + 376, + 604, + 408, + 299, + 408 + ], + "score": 0.916 + }, + { + "category_id": 0, + "poly": [ + 302, + 1246, + 1132, + 1246, + 1132, + 1279, + 302, + 1279 + ], + "score": 0.9 + }, + { + "category_id": 2, + "poly": [ + 841, + 2062, + 858, + 2062, + 858, + 2084, + 841, + 2084 + ], + "score": 0.771 + }, + { + "category_id": 0, + "poly": [ + 300, + 1681, + 1400, + 1681, + 1400, + 1710, + 300, + 1710 + ], + "score": 0.1 + }, + { + "category_id": 13, + "poly": [ + 1098, + 1892, + 1150, + 1892, + 1150, + 1921, + 1098, + 1921 + ], + "score": 0.86, + "latex": "10 \\%" + }, + { + "category_id": 13, + "poly": [ + 887, + 1923, + 950, + 1923, + 950, + 1952, + 887, + 1952 + ], + "score": 0.86, + "latex": "5 . 4 \\%" + }, + { + "category_id": 13, + "poly": [ + 299, + 294, + 382, + 294, + 382, + 324, + 299, + 324 + ], + "score": 0.85, + "latex": "( 6 5 . 0 \\%" + }, + { + "category_id": 13, + "poly": [ + 1173, + 1423, + 1277, + 1423, + 1277, + 1454, + 1173, + 1454 + ], + "score": 0.84, + "latex": "\\mathrm { O P T } _ { 1 2 5 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 1296, + 1394, + 1398, + 1394, + 1398, + 1424, + 1296, + 1424 + ], + "score": 0.83, + "latex": "\\mathrm { O P T } _ { 1 2 5 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 1225, + 295, + 1301, + 295, + 1301, + 325, + 1225, + 325 + ], + "score": 0.82, + "latex": "5 6 . 3 \\%" + }, + { + "category_id": 13, + "poly": [ + 815, + 295, + 893, + 295, + 893, + 325, + 815, + 325 + ], + "score": 0.82, + "latex": "5 1 . 0 \\%" + }, + { + "category_id": 13, + "poly": [ + 745, + 976, + 836, + 976, + 836, + 1005, + 745, + 1005 + ], + "score": 0.76, + "latex": "l r { = } 1 \\mathrm { e } { - } 5" + }, + { + "category_id": 13, + "poly": [ + 709, + 1393, + 812, + 1393, + 812, + 1424, + 709, + 1424 + ], + "score": 0.55, + "latex": "\\mathrm { O P T } _ { 3 5 0 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 695, + 266, + 742, + 266, + 742, + 295, + 695, + 295 + ], + "score": 0.44, + "latex": "\\mathfrak { j } _ { \\mathrm { X X L } }" + }, + { + "category_id": 15, + "poly": [ + 285.0, + 1103.0, + 1366.0, + 1103.0, + 1366.0, + 1160.0, + 285.0, + 1160.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 1974.0, + 806.0, + 1974.0, + 806.0, + 2009.0, + 333.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 368.0, + 607.0, + 368.0, + 607.0, + 419.0, + 291.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1244.0, + 1135.0, + 1244.0, + 1135.0, + 1290.0, + 293.0, + 1290.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2059.0, + 862.0, + 2059.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1673.0, + 1408.0, + 1673.0, + 1408.0, + 1717.0, + 295.0, + 1717.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1360.0, + 959.0, + 1360.0, + 959.0, + 1397.0, + 294.0, + 1397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 967.0, + 1361.0, + 1403.0, + 1361.0, + 1403.0, + 1395.0, + 967.0, + 1395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1388.0, + 708.0, + 1388.0, + 708.0, + 1433.0, + 290.0, + 1433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 813.0, + 1388.0, + 1295.0, + 1388.0, + 1295.0, + 1433.0, + 813.0, + 1433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1399.0, + 1388.0, + 1411.0, + 1388.0, + 1411.0, + 1433.0, + 1399.0, + 1433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1422.0, + 1172.0, + 1422.0, + 1172.0, + 1457.0, + 295.0, + 1457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1278.0, + 1422.0, + 1405.0, + 1422.0, + 1405.0, + 1457.0, + 1278.0, + 1457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1455.0, + 1403.0, + 1455.0, + 1403.0, + 1486.0, + 295.0, + 1486.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1484.0, + 1405.0, + 1484.0, + 1405.0, + 1516.0, + 292.0, + 1516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1513.0, + 1405.0, + 1513.0, + 1405.0, + 1547.0, + 292.0, + 1547.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1543.0, + 1406.0, + 1543.0, + 1406.0, + 1578.0, + 295.0, + 1578.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1572.0, + 1405.0, + 1572.0, + 1405.0, + 1609.0, + 292.0, + 1609.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1605.0, + 1406.0, + 1605.0, + 1406.0, + 1638.0, + 294.0, + 1638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1634.0, + 490.0, + 1634.0, + 490.0, + 1671.0, + 294.0, + 1671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 881.0, + 1405.0, + 881.0, + 1405.0, + 920.0, + 293.0, + 920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 909.0, + 1409.0, + 909.0, + 1409.0, + 953.0, + 292.0, + 953.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 943.0, + 1407.0, + 943.0, + 1407.0, + 981.0, + 292.0, + 981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 975.0, + 744.0, + 975.0, + 744.0, + 1010.0, + 293.0, + 1010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 837.0, + 975.0, + 1405.0, + 975.0, + 1405.0, + 1010.0, + 837.0, + 1010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1008.0, + 1405.0, + 1008.0, + 1405.0, + 1039.0, + 296.0, + 1039.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1033.0, + 1250.0, + 1033.0, + 1250.0, + 1072.0, + 292.0, + 1072.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 203.0, + 1405.0, + 203.0, + 1405.0, + 240.0, + 295.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 234.0, + 1403.0, + 234.0, + 1403.0, + 267.0, + 295.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 261.0, + 694.0, + 261.0, + 694.0, + 300.0, + 292.0, + 300.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 743.0, + 261.0, + 1406.0, + 261.0, + 1406.0, + 300.0, + 743.0, + 300.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 293.0, + 298.0, + 293.0, + 298.0, + 330.0, + 295.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 383.0, + 293.0, + 814.0, + 293.0, + 814.0, + 330.0, + 383.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 894.0, + 293.0, + 1224.0, + 293.0, + 1224.0, + 330.0, + 894.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1302.0, + 293.0, + 1408.0, + 293.0, + 1408.0, + 330.0, + 1302.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 324.0, + 1129.0, + 324.0, + 1129.0, + 360.0, + 295.0, + 360.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 742.0, + 1405.0, + 742.0, + 1405.0, + 783.0, + 293.0, + 783.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 774.0, + 1408.0, + 774.0, + 1408.0, + 811.0, + 293.0, + 811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 801.0, + 1405.0, + 801.0, + 1405.0, + 843.0, + 293.0, + 843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 836.0, + 882.0, + 836.0, + 882.0, + 870.0, + 295.0, + 870.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 604.0, + 1405.0, + 604.0, + 1405.0, + 638.0, + 295.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 636.0, + 1406.0, + 636.0, + 1406.0, + 668.0, + 295.0, + 668.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 664.0, + 1407.0, + 664.0, + 1407.0, + 701.0, + 292.0, + 701.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 696.0, + 1049.0, + 696.0, + 1049.0, + 730.0, + 292.0, + 730.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 414.0, + 1409.0, + 414.0, + 1409.0, + 452.0, + 293.0, + 452.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 447.0, + 1405.0, + 447.0, + 1405.0, + 481.0, + 294.0, + 481.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 479.0, + 1346.0, + 479.0, + 1346.0, + 512.0, + 295.0, + 512.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1161.0, + 1405.0, + 1161.0, + 1405.0, + 1200.0, + 294.0, + 1200.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1193.0, + 800.0, + 1193.0, + 800.0, + 1227.0, + 296.0, + 1227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1674.0, + 1411.0, + 1674.0, + 1411.0, + 1717.0, + 291.0, + 1717.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1709.0, + 1404.0, + 1709.0, + 1404.0, + 1745.0, + 294.0, + 1745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1737.0, + 1404.0, + 1737.0, + 1404.0, + 1775.0, + 294.0, + 1775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1769.0, + 1406.0, + 1769.0, + 1406.0, + 1803.0, + 292.0, + 1803.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1800.0, + 1406.0, + 1800.0, + 1406.0, + 1835.0, + 295.0, + 1835.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1826.0, + 1409.0, + 1826.0, + 1409.0, + 1870.0, + 292.0, + 1870.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1857.0, + 1407.0, + 1857.0, + 1407.0, + 1899.0, + 291.0, + 1899.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1892.0, + 1097.0, + 1892.0, + 1097.0, + 1924.0, + 295.0, + 1924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1151.0, + 1892.0, + 1407.0, + 1892.0, + 1407.0, + 1924.0, + 1151.0, + 1924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1921.0, + 886.0, + 1921.0, + 886.0, + 1954.0, + 294.0, + 1954.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 951.0, + 1921.0, + 1406.0, + 1921.0, + 1406.0, + 1954.0, + 951.0, + 1954.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1283.0, + 1404.0, + 1283.0, + 1404.0, + 1322.0, + 295.0, + 1322.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1315.0, + 1317.0, + 1315.0, + 1317.0, + 1351.0, + 295.0, + 1351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 524.0, + 1404.0, + 524.0, + 1404.0, + 561.0, + 294.0, + 561.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 554.0, + 1408.0, + 554.0, + 1408.0, + 592.0, + 294.0, + 592.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 3, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 299, + 629, + 881, + 629, + 881, + 993, + 299, + 993 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 297, + 1008, + 1406, + 1008, + 1406, + 1130, + 297, + 1130 + ], + "score": 0.974 + }, + { + "category_id": 5, + "poly": [ + 1028, + 1312, + 1379, + 1312, + 1379, + 1469, + 1028, + 1469 + ], + "score": 0.971, + "html": "
Epochw/ init.w/o init.
1130.2126.1
2132.7131.6
3133.4132.8
" + }, + { + "category_id": 1, + "poly": [ + 298, + 1152, + 990, + 1152, + 990, + 1455, + 298, + 1455 + ], + "score": 0.971 + }, + { + "category_id": 3, + "poly": [ + 298, + 196, + 1399, + 196, + 1399, + 411, + 298, + 411 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 297, + 1515, + 1404, + 1515, + 1404, + 1667, + 297, + 1667 + ], + "score": 0.964 + }, + { + "category_id": 3, + "poly": [ + 904, + 628, + 1394, + 628, + 1394, + 869, + 904, + 869 + ], + "score": 0.961 + }, + { + "category_id": 4, + "poly": [ + 901, + 882, + 1403, + 882, + 1403, + 973, + 901, + 973 + ], + "score": 0.959 + }, + { + "category_id": 1, + "poly": [ + 298, + 1884, + 1403, + 1884, + 1403, + 2008, + 298, + 2008 + ], + "score": 0.959 + }, + { + "category_id": 4, + "poly": [ + 298, + 422, + 1402, + 422, + 1402, + 514, + 298, + 514 + ], + "score": 0.95 + }, + { + "category_id": 6, + "poly": [ + 1009, + 1189, + 1405, + 1189, + 1405, + 1311, + 1009, + 1311 + ], + "score": 0.95 + }, + { + "category_id": 1, + "poly": [ + 296, + 545, + 1404, + 545, + 1404, + 608, + 296, + 608 + ], + "score": 0.941 + }, + { + "category_id": 0, + "poly": [ + 299, + 1480, + 832, + 1480, + 832, + 1512, + 299, + 1512 + ], + "score": 0.922 + }, + { + "category_id": 1, + "poly": [ + 299, + 1749, + 1403, + 1749, + 1403, + 1872, + 299, + 1872 + ], + "score": 0.805 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 858, + 2061, + 858, + 2085, + 841, + 2085 + ], + "score": 0.721 + }, + { + "category_id": 1, + "poly": [ + 335, + 1711, + 1335, + 1711, + 1335, + 1744, + 335, + 1744 + ], + "score": 0.572 + }, + { + "category_id": 0, + "poly": [ + 302, + 1681, + 676, + 1681, + 676, + 1711, + 302, + 1711 + ], + "score": 0.552 + }, + { + "category_id": 1, + "poly": [ + 302, + 1681, + 676, + 1681, + 676, + 1711, + 302, + 1711 + ], + "score": 0.249 + }, + { + "category_id": 0, + "poly": [ + 297, + 1153, + 988, + 1153, + 988, + 1181, + 297, + 1181 + ], + "score": 0.236 + }, + { + "category_id": 13, + "poly": [ + 926, + 1808, + 1134, + 1808, + 1134, + 1842, + 926, + 1842 + ], + "score": 0.93, + "latex": "{ \\dot { \\boldsymbol { f } } } _ { t } ( { \\boldsymbol { x } } ) = W _ { t } { \\boldsymbol { x } } + b _ { t }" + }, + { + "category_id": 13, + "poly": [ + 467, + 1915, + 553, + 1915, + 553, + 1949, + 467, + 1949 + ], + "score": 0.92, + "latex": "\\mathrm { L L M _ { \\mathrm { t g t } } }" + }, + { + "category_id": 13, + "poly": [ + 413, + 811, + 498, + 811, + 498, + 845, + 413, + 845 + ], + "score": 0.91, + "latex": "\\mathrm { L L M _ { \\mathrm { t g t } } }" + }, + { + "category_id": 13, + "poly": [ + 878, + 1711, + 963, + 1711, + 963, + 1745, + 878, + 1745 + ], + "score": 0.91, + "latex": "\\mathrm { L L M _ { \\mathrm { t g t } } }" + }, + { + "category_id": 13, + "poly": [ + 540, + 751, + 628, + 751, + 628, + 782, + 540, + 782 + ], + "score": 0.9, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 519, + 843, + 612, + 843, + 612, + 875, + 519, + 875 + ], + "score": 0.9, + "latex": "\\mathrm { L L M } _ { t g t }" + }, + { + "category_id": 13, + "poly": [ + 416, + 722, + 503, + 722, + 503, + 752, + 416, + 752 + ], + "score": 0.9, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 328, + 1841, + 539, + 1841, + 539, + 1873, + 328, + 1873 + ], + "score": 0.89, + "latex": "g _ { c } ( x ) = W _ { c } x + b _ { c }" + }, + { + "category_id": 13, + "poly": [ + 328, + 1808, + 542, + 1808, + 542, + 1840, + 328, + 1840 + ], + "score": 0.89, + "latex": "{ \\bar { f } } _ { s } ( x ) = W _ { s } x + b _ { s }" + }, + { + "category_id": 13, + "poly": [ + 1234, + 1885, + 1319, + 1885, + 1319, + 1916, + 1234, + 1916 + ], + "score": 0.88, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 434, + 1779, + 520, + 1779, + 520, + 1809, + 434, + 1809 + ], + "score": 0.88, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 719, + 781, + 806, + 781, + 806, + 811, + 719, + 811 + ], + "score": 0.88, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 1111, + 1748, + 1197, + 1748, + 1197, + 1782, + 1111, + 1782 + ], + "score": 0.88, + "latex": "\\mathrm { L L M _ { \\mathrm { t g t } } }" + }, + { + "category_id": 13, + "poly": [ + 1163, + 1009, + 1268, + 1009, + 1268, + 1040, + 1163, + 1040 + ], + "score": 0.88, + "latex": "\\mathrm { O P T } _ { 1 2 5 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 1314, + 1779, + 1401, + 1779, + 1401, + 1810, + 1314, + 1810 + ], + "score": 0.87, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 1038, + 1976, + 1123, + 1976, + 1123, + 2007, + 1038, + 2007 + ], + "score": 0.87, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 949, + 423, + 1055, + 423, + 1055, + 454, + 949, + 454 + ], + "score": 0.87, + "latex": "\\mathrm { O P T } _ { 1 2 5 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 808, + 1809, + 893, + 1809, + 893, + 1843, + 808, + 1843 + ], + "score": 0.86, + "latex": "\\mathrm { L L M _ { \\mathrm { t g t } } }" + }, + { + "category_id": 13, + "poly": [ + 1173, + 1976, + 1257, + 1976, + 1257, + 2010, + 1173, + 2010 + ], + "score": 0.86, + "latex": "\\mathrm { L L M _ { \\mathrm { t g t } } }" + }, + { + "category_id": 13, + "poly": [ + 1243, + 1711, + 1329, + 1711, + 1329, + 1742, + 1243, + 1742 + ], + "score": 0.85, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 890, + 1071, + 995, + 1071, + 995, + 1100, + 890, + 1100 + ], + "score": 0.83, + "latex": "\\mathrm { O P T } _ { 1 2 5 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 1303, + 1009, + 1398, + 1009, + 1398, + 1040, + 1303, + 1040 + ], + "score": 0.77, + "latex": "\\mathrm { O P T } _ { 1 . 3 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 866, + 1515, + 917, + 1515, + 917, + 1547, + 866, + 1547 + ], + "score": 0.76, + "latex": "\\ S 3 . 1" + }, + { + "category_id": 13, + "poly": [ + 1307, + 1040, + 1399, + 1040, + 1399, + 1070, + 1307, + 1070 + ], + "score": 0.69, + "latex": "\\mathrm { O P T } _ { 1 . 3 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 455, + 629, + 554, + 629, + 554, + 663, + 455, + 663 + ], + "score": 0.66, + "latex": "\\mathbf { L L M _ { \\mathrm { t g t } } } ^ { \\prime }" + }, + { + "category_id": 13, + "poly": [ + 554, + 453, + 656, + 453, + 656, + 484, + 554, + 484 + ], + "score": 0.42, + "latex": "\\mathrm { O P T } _ { 3 5 0 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 346, + 1748, + 382, + 1748, + 382, + 1780, + 346, + 1780 + ], + "score": 0.41, + "latex": "( b )" + }, + { + "category_id": 13, + "poly": [ + 882, + 1039, + 970, + 1039, + 970, + 1070, + 882, + 1070 + ], + "score": 0.32, + "latex": "\\ S 3 . 2 ( \\mathsf { b } ) )" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 195.0, + 489.0, + 195.0, + 489.0, + 222.0, + 372.0, + 222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 616.0, + 195.0, + 684.0, + 195.0, + 684.0, + 225.0, + 616.0, + 225.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 837.0, + 195.0, + 902.0, + 195.0, + 902.0, + 224.0, + 837.0, + 224.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1066.0, + 196.0, + 1114.0, + 196.0, + 1114.0, + 222.0, + 1066.0, + 222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1276.0, + 196.0, + 1344.0, + 196.0, + 1344.0, + 221.0, + 1276.0, + 221.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 534.0, + 219.0, + 568.0, + 219.0, + 568.0, + 242.0, + 534.0, + 242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 762.0, + 225.0, + 787.0, + 225.0, + 787.0, + 244.0, + 762.0, + 244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 981.0, + 225.0, + 1007.0, + 225.0, + 1007.0, + 244.0, + 981.0, + 244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1203.0, + 226.0, + 1226.0, + 226.0, + 1226.0, + 246.0, + 1203.0, + 246.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 243.0, + 349.0, + 243.0, + 349.0, + 267.0, + 316.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 265.0, + 347.0, + 265.0, + 347.0, + 326.0, + 297.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 516.0, + 257.0, + 568.0, + 257.0, + 568.0, + 324.0, + 516.0, + 324.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 741.0, + 256.0, + 797.0, + 256.0, + 797.0, + 323.0, + 741.0, + 323.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 958.0, + 254.0, + 1012.0, + 254.0, + 1012.0, + 321.0, + 958.0, + 321.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1181.0, + 260.0, + 1234.0, + 260.0, + 1234.0, + 323.0, + 1181.0, + 323.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 761.0, + 317.0, + 788.0, + 317.0, + 788.0, + 340.0, + 761.0, + 340.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 980.0, + 316.0, + 1009.0, + 316.0, + 1009.0, + 340.0, + 980.0, + 340.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 334.0, + 349.0, + 334.0, + 349.0, + 358.0, + 316.0, + 358.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 383.0, + 332.0, + 466.0, + 332.0, + 466.0, + 351.0, + 383.0, + 351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 541.0, + 335.0, + 570.0, + 335.0, + 570.0, + 359.0, + 541.0, + 359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 605.0, + 332.0, + 686.0, + 332.0, + 686.0, + 351.0, + 605.0, + 351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 825.0, + 332.0, + 906.0, + 332.0, + 906.0, + 351.0, + 825.0, + 351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1044.0, + 332.0, + 1125.0, + 332.0, + 1125.0, + 351.0, + 1044.0, + 351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1204.0, + 331.0, + 1226.0, + 331.0, + 1226.0, + 351.0, + 1204.0, + 351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1264.0, + 332.0, + 1345.0, + 332.0, + 1345.0, + 351.0, + 1264.0, + 351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 358.0, + 347.0, + 501.0, + 347.0, + 501.0, + 370.0, + 358.0, + 370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 578.0, + 347.0, + 720.0, + 347.0, + 720.0, + 370.0, + 578.0, + 370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 760.0, + 348.0, + 788.0, + 348.0, + 788.0, + 371.0, + 760.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 797.0, + 347.0, + 940.0, + 347.0, + 940.0, + 370.0, + 797.0, + 370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 980.0, + 348.0, + 1009.0, + 348.0, + 1009.0, + 371.0, + 980.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1017.0, + 349.0, + 1044.0, + 349.0, + 1044.0, + 366.0, + 1017.0, + 366.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1045.0, + 349.0, + 1159.0, + 349.0, + 1159.0, + 369.0, + 1045.0, + 369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1239.0, + 349.0, + 1378.0, + 349.0, + 1378.0, + 369.0, + 1239.0, + 369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 337.0, + 372.0, + 523.0, + 372.0, + 523.0, + 396.0, + 337.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 557.0, + 374.0, + 743.0, + 374.0, + 743.0, + 397.0, + 557.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 776.0, + 374.0, + 964.0, + 374.0, + 964.0, + 397.0, + 776.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 995.0, + 372.0, + 1183.0, + 372.0, + 1183.0, + 396.0, + 995.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1234.0, + 374.0, + 1403.0, + 374.0, + 1403.0, + 397.0, + 1234.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 390.0, + 463.0, + 390.0, + 463.0, + 414.0, + 398.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 618.0, + 390.0, + 682.0, + 390.0, + 682.0, + 414.0, + 618.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 390.0, + 902.0, + 390.0, + 902.0, + 414.0, + 838.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 390.0, + 1122.0, + 390.0, + 1122.0, + 411.0, + 1057.0, + 411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1279.0, + 390.0, + 1341.0, + 390.0, + 1341.0, + 414.0, + 1279.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1208.0, + 364.0, + 1234.0, + 364.0, + 1234.0, + 391.0, + 1208.0, + 391.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 997.0, + 627.0, + 1101.0, + 627.0, + 1101.0, + 652.0, + 997.0, + 652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1265.0, + 628.0, + 1320.0, + 628.0, + 1320.0, + 652.0, + 1265.0, + 652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 920.0, + 656.0, + 950.0, + 656.0, + 950.0, + 678.0, + 920.0, + 678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1098.0, + 666.0, + 1111.0, + 666.0, + 1111.0, + 677.0, + 1098.0, + 677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1171.0, + 656.0, + 1193.0, + 656.0, + 1193.0, + 674.0, + 1171.0, + 674.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 920.0, + 684.0, + 950.0, + 684.0, + 950.0, + 705.0, + 920.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1171.0, + 686.0, + 1193.0, + 686.0, + 1193.0, + 705.0, + 1171.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 902.0, + 712.0, + 949.0, + 712.0, + 949.0, + 767.0, + 902.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1171.0, + 717.0, + 1193.0, + 717.0, + 1193.0, + 736.0, + 1171.0, + 736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1210.0, + 716.0, + 1218.0, + 716.0, + 1218.0, + 723.0, + 1210.0, + 723.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 921.0, + 737.0, + 950.0, + 737.0, + 950.0, + 758.0, + 921.0, + 758.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1156.0, + 727.0, + 1178.0, + 727.0, + 1178.0, + 761.0, + 1156.0, + 761.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1171.0, + 748.0, + 1193.0, + 748.0, + 1193.0, + 767.0, + 1171.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 920.0, + 764.0, + 950.0, + 764.0, + 950.0, + 785.0, + 920.0, + 785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 919.0, + 790.0, + 950.0, + 790.0, + 950.0, + 812.0, + 919.0, + 812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 992.0, + 784.0, + 1115.0, + 784.0, + 1115.0, + 807.0, + 992.0, + 807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1171.0, + 779.0, + 1193.0, + 779.0, + 1193.0, + 798.0, + 1171.0, + 798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1217.0, + 792.0, + 1225.0, + 792.0, + 1225.0, + 799.0, + 1217.0, + 799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1238.0, + 784.0, + 1360.0, + 784.0, + 1360.0, + 807.0, + 1238.0, + 807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 920.0, + 816.0, + 950.0, + 816.0, + 950.0, + 838.0, + 920.0, + 838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 962.0, + 807.0, + 994.0, + 807.0, + 994.0, + 823.0, + 962.0, + 823.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 995.0, + 804.0, + 1115.0, + 804.0, + 1115.0, + 826.0, + 995.0, + 826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1171.0, + 811.0, + 1193.0, + 811.0, + 1193.0, + 830.0, + 1171.0, + 830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1214.0, + 810.0, + 1236.0, + 810.0, + 1236.0, + 819.0, + 1214.0, + 819.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1238.0, + 803.0, + 1361.0, + 803.0, + 1361.0, + 828.0, + 1238.0, + 828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 942.0, + 838.0, + 958.0, + 838.0, + 958.0, + 853.0, + 942.0, + 853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 965.0, + 837.0, + 1011.0, + 837.0, + 1011.0, + 855.0, + 965.0, + 855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1013.0, + 831.0, + 1158.0, + 831.0, + 1158.0, + 858.0, + 1013.0, + 858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1186.0, + 838.0, + 1202.0, + 838.0, + 1202.0, + 853.0, + 1186.0, + 853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1210.0, + 838.0, + 1228.0, + 838.0, + 1228.0, + 854.0, + 1210.0, + 854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1234.0, + 838.0, + 1252.0, + 838.0, + 1252.0, + 854.0, + 1234.0, + 854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1259.0, + 838.0, + 1277.0, + 838.0, + 1277.0, + 853.0, + 1259.0, + 853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1281.0, + 836.0, + 1352.0, + 836.0, + 1352.0, + 855.0, + 1281.0, + 855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1357.0, + 837.0, + 1376.0, + 837.0, + 1376.0, + 855.0, + 1357.0, + 855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1377.0, + 838.0, + 1399.0, + 838.0, + 1399.0, + 853.0, + 1377.0, + 853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1016.0, + 849.0, + 1077.0, + 849.0, + 1077.0, + 871.0, + 1016.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1260.0, + 849.0, + 1322.0, + 849.0, + 1322.0, + 871.0, + 1260.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 877.0, + 1405.0, + 877.0, + 1405.0, + 919.0, + 899.0, + 919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 912.0, + 1404.0, + 912.0, + 1404.0, + 944.0, + 899.0, + 944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 901.0, + 942.0, + 1382.0, + 942.0, + 1382.0, + 977.0, + 901.0, + 977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 421.0, + 948.0, + 421.0, + 948.0, + 458.0, + 294.0, + 458.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1056.0, + 421.0, + 1406.0, + 421.0, + 1406.0, + 458.0, + 1056.0, + 458.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 449.0, + 553.0, + 449.0, + 553.0, + 491.0, + 292.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 657.0, + 449.0, + 1406.0, + 449.0, + 1406.0, + 491.0, + 657.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 482.0, + 451.0, + 482.0, + 451.0, + 514.0, + 296.0, + 514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1012.0, + 1190.0, + 1406.0, + 1190.0, + 1406.0, + 1221.0, + 1012.0, + 1221.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1009.0, + 1220.0, + 1405.0, + 1220.0, + 1405.0, + 1253.0, + 1009.0, + 1253.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1010.0, + 1251.0, + 1403.0, + 1251.0, + 1403.0, + 1282.0, + 1010.0, + 1282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1011.0, + 1282.0, + 1389.0, + 1282.0, + 1389.0, + 1313.0, + 1011.0, + 1313.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1477.0, + 833.0, + 1477.0, + 833.0, + 1516.0, + 293.0, + 1516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2058.0, + 861.0, + 2058.0, + 861.0, + 2093.0, + 839.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1675.0, + 680.0, + 1675.0, + 680.0, + 1723.0, + 294.0, + 1723.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 1151.0, + 994.0, + 1151.0, + 994.0, + 1185.0, + 298.0, + 1185.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 628.0, + 454.0, + 628.0, + 454.0, + 662.0, + 296.0, + 662.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 555.0, + 628.0, + 883.0, + 628.0, + 883.0, + 662.0, + 555.0, + 662.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 658.0, + 883.0, + 658.0, + 883.0, + 693.0, + 295.0, + 693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 690.0, + 882.0, + 690.0, + 882.0, + 722.0, + 295.0, + 722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 719.0, + 415.0, + 719.0, + 415.0, + 754.0, + 295.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 504.0, + 719.0, + 882.0, + 719.0, + 882.0, + 754.0, + 504.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 749.0, + 539.0, + 749.0, + 539.0, + 785.0, + 295.0, + 785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 629.0, + 749.0, + 884.0, + 749.0, + 884.0, + 785.0, + 629.0, + 785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 780.0, + 718.0, + 780.0, + 718.0, + 813.0, + 293.0, + 813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 807.0, + 780.0, + 884.0, + 780.0, + 884.0, + 813.0, + 807.0, + 813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 810.0, + 412.0, + 810.0, + 412.0, + 844.0, + 292.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 499.0, + 810.0, + 884.0, + 810.0, + 884.0, + 844.0, + 499.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 839.0, + 518.0, + 839.0, + 518.0, + 876.0, + 292.0, + 876.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 613.0, + 839.0, + 883.0, + 839.0, + 883.0, + 876.0, + 613.0, + 876.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 872.0, + 883.0, + 872.0, + 883.0, + 904.0, + 295.0, + 904.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 902.0, + 882.0, + 902.0, + 882.0, + 933.0, + 294.0, + 933.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 934.0, + 881.0, + 934.0, + 881.0, + 963.0, + 297.0, + 963.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 964.0, + 574.0, + 964.0, + 574.0, + 993.0, + 296.0, + 993.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1002.0, + 1162.0, + 1002.0, + 1162.0, + 1047.0, + 290.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1269.0, + 1002.0, + 1302.0, + 1002.0, + 1302.0, + 1047.0, + 1269.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1399.0, + 1002.0, + 1412.0, + 1002.0, + 1412.0, + 1047.0, + 1399.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1033.0, + 881.0, + 1033.0, + 881.0, + 1077.0, + 291.0, + 1077.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 971.0, + 1033.0, + 1306.0, + 1033.0, + 1306.0, + 1077.0, + 971.0, + 1077.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1400.0, + 1033.0, + 1407.0, + 1033.0, + 1407.0, + 1077.0, + 1400.0, + 1077.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1068.0, + 889.0, + 1068.0, + 889.0, + 1104.0, + 294.0, + 1104.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 996.0, + 1068.0, + 1411.0, + 1068.0, + 1411.0, + 1104.0, + 996.0, + 1104.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1098.0, + 1189.0, + 1098.0, + 1189.0, + 1133.0, + 294.0, + 1133.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 1149.0, + 994.0, + 1149.0, + 994.0, + 1186.0, + 298.0, + 1186.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1182.0, + 991.0, + 1182.0, + 991.0, + 1212.0, + 297.0, + 1212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1212.0, + 990.0, + 1212.0, + 990.0, + 1245.0, + 296.0, + 1245.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1243.0, + 991.0, + 1243.0, + 991.0, + 1274.0, + 294.0, + 1274.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1271.0, + 993.0, + 1271.0, + 993.0, + 1306.0, + 294.0, + 1306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1302.0, + 993.0, + 1302.0, + 993.0, + 1337.0, + 294.0, + 1337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1333.0, + 991.0, + 1333.0, + 991.0, + 1365.0, + 296.0, + 1365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1361.0, + 995.0, + 1361.0, + 995.0, + 1399.0, + 293.0, + 1399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1393.0, + 992.0, + 1393.0, + 992.0, + 1426.0, + 295.0, + 1426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1420.0, + 422.0, + 1420.0, + 422.0, + 1457.0, + 294.0, + 1457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1514.0, + 865.0, + 1514.0, + 865.0, + 1546.0, + 296.0, + 1546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 918.0, + 1514.0, + 1402.0, + 1514.0, + 1402.0, + 1546.0, + 918.0, + 1546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1541.0, + 1405.0, + 1541.0, + 1405.0, + 1581.0, + 294.0, + 1581.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1574.0, + 1406.0, + 1574.0, + 1406.0, + 1608.0, + 292.0, + 1608.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1601.0, + 1404.0, + 1601.0, + 1404.0, + 1644.0, + 294.0, + 1644.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1634.0, + 1104.0, + 1634.0, + 1104.0, + 1671.0, + 294.0, + 1671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1885.0, + 1233.0, + 1885.0, + 1233.0, + 1918.0, + 296.0, + 1918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1320.0, + 1885.0, + 1404.0, + 1885.0, + 1404.0, + 1918.0, + 1320.0, + 1918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1913.0, + 466.0, + 1913.0, + 466.0, + 1950.0, + 294.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 554.0, + 1913.0, + 1405.0, + 1913.0, + 1405.0, + 1950.0, + 554.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1947.0, + 1403.0, + 1947.0, + 1403.0, + 1980.0, + 296.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1978.0, + 1037.0, + 1978.0, + 1037.0, + 2010.0, + 294.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1124.0, + 1978.0, + 1172.0, + 1978.0, + 1172.0, + 2010.0, + 1124.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1258.0, + 1978.0, + 1405.0, + 1978.0, + 1405.0, + 2010.0, + 1258.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 542.0, + 1404.0, + 542.0, + 1404.0, + 582.0, + 293.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 572.0, + 401.0, + 572.0, + 401.0, + 612.0, + 290.0, + 612.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 383.0, + 1747.0, + 1110.0, + 1747.0, + 1110.0, + 1783.0, + 383.0, + 1783.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 1747.0, + 1404.0, + 1747.0, + 1404.0, + 1783.0, + 1198.0, + 1783.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1778.0, + 433.0, + 1778.0, + 433.0, + 1814.0, + 294.0, + 1814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 521.0, + 1778.0, + 1313.0, + 1778.0, + 1313.0, + 1814.0, + 521.0, + 1814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1806.0, + 327.0, + 1806.0, + 327.0, + 1845.0, + 293.0, + 1845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 543.0, + 1806.0, + 807.0, + 1806.0, + 807.0, + 1845.0, + 543.0, + 1845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 894.0, + 1806.0, + 925.0, + 1806.0, + 925.0, + 1845.0, + 894.0, + 1845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1135.0, + 1806.0, + 1408.0, + 1806.0, + 1408.0, + 1845.0, + 1135.0, + 1845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1838.0, + 327.0, + 1838.0, + 327.0, + 1876.0, + 294.0, + 1876.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 540.0, + 1838.0, + 551.0, + 1838.0, + 551.0, + 1876.0, + 540.0, + 1876.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 341.0, + 1707.0, + 877.0, + 1707.0, + 877.0, + 1748.0, + 341.0, + 1748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 1707.0, + 1242.0, + 1707.0, + 1242.0, + 1748.0, + 964.0, + 1748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1330.0, + 1707.0, + 1339.0, + 1707.0, + 1339.0, + 1748.0, + 1330.0, + 1748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1675.0, + 680.0, + 1675.0, + 680.0, + 1723.0, + 294.0, + 1723.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 4, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1530, + 1405, + 1530, + 1405, + 1716, + 298, + 1716 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 299, + 1730, + 1405, + 1730, + 1405, + 1854, + 299, + 1854 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 299, + 1914, + 1404, + 1914, + 1404, + 2007, + 299, + 2007 + ], + "score": 0.971 + }, + { + "category_id": 3, + "poly": [ + 304, + 195, + 1394, + 195, + 1394, + 582, + 304, + 582 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 298, + 1299, + 1405, + 1299, + 1405, + 1392, + 298, + 1392 + ], + "score": 0.966 + }, + { + "category_id": 4, + "poly": [ + 295, + 608, + 1404, + 608, + 1404, + 733, + 295, + 733 + ], + "score": 0.954 + }, + { + "category_id": 1, + "poly": [ + 298, + 1452, + 1399, + 1452, + 1399, + 1514, + 298, + 1514 + ], + "score": 0.951 + }, + { + "category_id": 1, + "poly": [ + 292, + 757, + 1402, + 757, + 1402, + 821, + 292, + 821 + ], + "score": 0.951 + }, + { + "category_id": 1, + "poly": [ + 298, + 1127, + 1405, + 1127, + 1405, + 1220, + 298, + 1220 + ], + "score": 0.951 + }, + { + "category_id": 0, + "poly": [ + 297, + 1245, + 985, + 1245, + 985, + 1284, + 297, + 1284 + ], + "score": 0.944 + }, + { + "category_id": 8, + "poly": [ + 699, + 833, + 999, + 833, + 999, + 874, + 699, + 874 + ], + "score": 0.934 + }, + { + "category_id": 1, + "poly": [ + 303, + 1048, + 1403, + 1048, + 1403, + 1111, + 303, + 1111 + ], + "score": 0.932 + }, + { + "category_id": 0, + "poly": [ + 293, + 1874, + 1303, + 1874, + 1303, + 1907, + 293, + 1907 + ], + "score": 0.927 + }, + { + "category_id": 1, + "poly": [ + 290, + 902, + 1372, + 902, + 1372, + 936, + 290, + 936 + ], + "score": 0.925 + }, + { + "category_id": 0, + "poly": [ + 299, + 1411, + 627, + 1411, + 627, + 1445, + 299, + 1445 + ], + "score": 0.924 + }, + { + "category_id": 1, + "poly": [ + 294, + 1000, + 1084, + 1000, + 1084, + 1036, + 294, + 1036 + ], + "score": 0.917 + }, + { + "category_id": 9, + "poly": [ + 1368, + 953, + 1402, + 953, + 1402, + 983, + 1368, + 983 + ], + "score": 0.879 + }, + { + "category_id": 8, + "poly": [ + 608, + 948, + 1092, + 948, + 1092, + 989, + 608, + 989 + ], + "score": 0.873 + }, + { + "category_id": 9, + "poly": [ + 1368, + 839, + 1401, + 839, + 1401, + 869, + 1368, + 869 + ], + "score": 0.852 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 859, + 2062, + 859, + 2085, + 840, + 2085 + ], + "score": 0.795 + }, + { + "category_id": 8, + "poly": [ + 609, + 948, + 1092, + 948, + 1092, + 989, + 609, + 989 + ], + "score": 0.199 + }, + { + "category_id": 13, + "poly": [ + 690, + 1002, + 840, + 1002, + 840, + 1034, + 690, + 1034 + ], + "score": 0.92, + "latex": "W _ { t } = W _ { s } W _ { c }" + }, + { + "category_id": 13, + "poly": [ + 692, + 903, + 748, + 903, + 748, + 937, + 692, + 937 + ], + "score": 0.91, + "latex": "g _ { c } ( \\cdot )" + }, + { + "category_id": 13, + "poly": [ + 891, + 1002, + 1070, + 1002, + 1070, + 1034, + 891, + 1034 + ], + "score": 0.91, + "latex": "b _ { t } = W _ { s } b _ { c } + b _ { s }" + }, + { + "category_id": 14, + "poly": [ + 697, + 834, + 1001, + 834, + 1001, + 873, + 697, + 873 + ], + "score": 0.91, + "latex": "\\mathcal { L } = 1 - s i m ( g _ { c } ( x _ { s } ) , x _ { t } ) ." + }, + { + "category_id": 14, + "poly": [ + 605, + 949, + 1093, + 949, + 1093, + 988, + 605, + 988 + ], + "score": 0.91, + "latex": "f _ { t } ( x ) = f _ { s } ( g _ { c } ( x ) ) = W _ { s } ( W _ { c } x + b _ { c } ) + b _ { s } ," + }, + { + "category_id": 13, + "poly": [ + 945, + 640, + 1031, + 640, + 1031, + 671, + 945, + 671 + ], + "score": 0.89, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 1247, + 904, + 1333, + 904, + 1333, + 935, + 1247, + 935 + ], + "score": 0.88, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 430, + 1004, + 456, + 1004, + 456, + 1035, + 430, + 1035 + ], + "score": 0.87, + "latex": "f _ { t }" + }, + { + "category_id": 13, + "poly": [ + 848, + 789, + 933, + 789, + 933, + 823, + 848, + 823 + ], + "score": 0.87, + "latex": "\\mathrm { L L M _ { \\mathrm { t g t } } }" + }, + { + "category_id": 13, + "poly": [ + 1313, + 610, + 1400, + 610, + 1400, + 640, + 1313, + 640 + ], + "score": 0.86, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 966, + 793, + 997, + 793, + 997, + 819, + 966, + 819 + ], + "score": 0.85, + "latex": "x _ { s }" + }, + { + "category_id": 13, + "poly": [ + 1047, + 793, + 1076, + 793, + 1076, + 819, + 1047, + 819 + ], + "score": 0.85, + "latex": "x _ { t }" + }, + { + "category_id": 13, + "poly": [ + 713, + 789, + 798, + 789, + 798, + 820, + 713, + 820 + ], + "score": 0.84, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 298, + 790, + 316, + 790, + 316, + 816, + 298, + 816 + ], + "score": 0.78, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 761, + 707, + 781, + 707, + 781, + 728, + 761, + 728 + ], + "score": 0.72, + "latex": "n" + }, + { + "category_id": 13, + "poly": [ + 1381, + 1165, + 1401, + 1165, + 1401, + 1185, + 1381, + 1185 + ], + "score": 0.69, + "latex": "n" + }, + { + "category_id": 13, + "poly": [ + 969, + 1824, + 1012, + 1824, + 1012, + 1853, + 969, + 1853 + ], + "score": 0.41, + "latex": "\\bar { \\mathsf { \\Sigma } } _ { \\mathrm { b a s e } }" + }, + { + "category_id": 13, + "poly": [ + 1128, + 1824, + 1172, + 1824, + 1172, + 1854, + 1128, + 1854 + ], + "score": 0.37, + "latex": "{ \\mathrm { 5 } } _ { \\mathrm { X L } }" + }, + { + "category_id": 13, + "poly": [ + 296, + 1130, + 322, + 1130, + 322, + 1157, + 296, + 1157 + ], + "score": 0.3, + "latex": "\\blacktriangleright" + }, + { + "category_id": 13, + "poly": [ + 345, + 1159, + 379, + 1159, + 379, + 1189, + 345, + 1189 + ], + "score": 0.27, + "latex": "( d )" + }, + { + "category_id": 15, + "poly": [ + 843.0, + 197.0, + 943.0, + 197.0, + 943.0, + 222.0, + 843.0, + 222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 701.0, + 243.0, + 725.0, + 243.0, + 725.0, + 273.0, + 701.0, + 273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1158.0, + 236.0, + 1190.0, + 236.0, + 1190.0, + 267.0, + 1158.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 640.0, + 260.0, + 687.0, + 260.0, + 687.0, + 287.0, + 640.0, + 287.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 851.0, + 262.0, + 935.0, + 262.0, + 935.0, + 289.0, + 851.0, + 289.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 302.0, + 442.0, + 302.0, + 442.0, + 337.0, + 346.0, + 337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 303.0, + 325.0, + 485.0, + 325.0, + 485.0, + 361.0, + 303.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 540.0, + 321.0, + 656.0, + 321.0, + 656.0, + 360.0, + 540.0, + 360.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 325.0, + 895.0, + 325.0, + 895.0, + 358.0, + 787.0, + 358.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1060.0, + 333.0, + 1186.0, + 333.0, + 1186.0, + 357.0, + 1060.0, + 357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 574.0, + 344.0, + 623.0, + 344.0, + 623.0, + 369.0, + 574.0, + 369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 799.0, + 343.0, + 882.0, + 343.0, + 882.0, + 371.0, + 799.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 637.0, + 399.0, + 688.0, + 399.0, + 688.0, + 429.0, + 637.0, + 429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 695.0, + 373.0, + 731.0, + 373.0, + 731.0, + 410.0, + 695.0, + 410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 851.0, + 399.0, + 934.0, + 399.0, + 934.0, + 425.0, + 851.0, + 425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 937.0, + 395.0, + 958.0, + 395.0, + 958.0, + 419.0, + 937.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1158.0, + 380.0, + 1190.0, + 380.0, + 1190.0, + 411.0, + 1158.0, + 411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1215.0, + 395.0, + 1378.0, + 395.0, + 1378.0, + 423.0, + 1215.0, + 423.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1217.0, + 411.0, + 1376.0, + 411.0, + 1376.0, + 439.0, + 1217.0, + 439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 475.0, + 479.0, + 475.0, + 479.0, + 532.0, + 307.0, + 532.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 637.0, + 491.0, + 688.0, + 491.0, + 688.0, + 522.0, + 637.0, + 522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 700.0, + 472.0, + 725.0, + 472.0, + 725.0, + 503.0, + 700.0, + 503.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 849.0, + 486.0, + 958.0, + 486.0, + 958.0, + 519.0, + 849.0, + 519.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1159.0, + 470.0, + 1191.0, + 470.0, + 1191.0, + 503.0, + 1159.0, + 503.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1211.0, + 487.0, + 1400.0, + 487.0, + 1400.0, + 543.0, + 1211.0, + 543.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 843.0, + 557.0, + 943.0, + 557.0, + 943.0, + 583.0, + 843.0, + 583.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1082.0, + 256.0, + 1164.0, + 256.0, + 1164.0, + 286.5, + 1082.0, + 286.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1081.0, + 398.0, + 1164.0, + 398.0, + 1164.0, + 432.0, + 1081.0, + 432.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1082.0, + 490.5, + 1164.0, + 490.5, + 1164.0, + 525.0, + 1082.0, + 525.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 606.0, + 1312.0, + 606.0, + 1312.0, + 647.0, + 292.0, + 647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1401.0, + 606.0, + 1405.0, + 606.0, + 1405.0, + 647.0, + 1401.0, + 647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 639.0, + 944.0, + 639.0, + 944.0, + 675.0, + 294.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1032.0, + 639.0, + 1405.0, + 639.0, + 1405.0, + 675.0, + 1032.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 671.0, + 1405.0, + 671.0, + 1405.0, + 704.0, + 293.0, + 704.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 701.0, + 760.0, + 701.0, + 760.0, + 734.0, + 297.0, + 734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 782.0, + 701.0, + 872.0, + 701.0, + 872.0, + 734.0, + 782.0, + 734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1244.0, + 989.0, + 1244.0, + 989.0, + 1291.0, + 291.0, + 1291.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1870.0, + 1306.0, + 1870.0, + 1306.0, + 1914.0, + 290.0, + 1914.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1406.0, + 629.0, + 1406.0, + 629.0, + 1453.0, + 293.0, + 1453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2060.0, + 862.0, + 2060.0, + 862.0, + 2091.0, + 840.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1530.0, + 1408.0, + 1530.0, + 1408.0, + 1566.0, + 295.0, + 1566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1560.0, + 1407.0, + 1560.0, + 1407.0, + 1596.0, + 295.0, + 1596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1592.0, + 1407.0, + 1592.0, + 1407.0, + 1628.0, + 295.0, + 1628.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1622.0, + 1405.0, + 1622.0, + 1405.0, + 1657.0, + 295.0, + 1657.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1650.0, + 1405.0, + 1650.0, + 1405.0, + 1688.0, + 293.0, + 1688.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1682.0, + 1374.0, + 1682.0, + 1374.0, + 1720.0, + 292.0, + 1720.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1730.0, + 1404.0, + 1730.0, + 1404.0, + 1765.0, + 294.0, + 1765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1760.0, + 1407.0, + 1760.0, + 1407.0, + 1797.0, + 294.0, + 1797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1792.0, + 1405.0, + 1792.0, + 1405.0, + 1827.0, + 294.0, + 1827.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1820.0, + 968.0, + 1820.0, + 968.0, + 1860.0, + 293.0, + 1860.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1013.0, + 1820.0, + 1127.0, + 1820.0, + 1127.0, + 1860.0, + 1013.0, + 1860.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1173.0, + 1820.0, + 1184.0, + 1820.0, + 1184.0, + 1860.0, + 1173.0, + 1860.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1915.0, + 1406.0, + 1915.0, + 1406.0, + 1948.0, + 295.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1945.0, + 1406.0, + 1945.0, + 1406.0, + 1981.0, + 294.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1976.0, + 1404.0, + 1976.0, + 1404.0, + 2010.0, + 295.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1300.0, + 1403.0, + 1300.0, + 1403.0, + 1333.0, + 295.0, + 1333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1331.0, + 1403.0, + 1331.0, + 1403.0, + 1365.0, + 296.0, + 1365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1361.0, + 889.0, + 1361.0, + 889.0, + 1395.0, + 296.0, + 1395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1451.0, + 1401.0, + 1451.0, + 1401.0, + 1487.0, + 294.0, + 1487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1482.0, + 1235.0, + 1482.0, + 1235.0, + 1517.0, + 294.0, + 1517.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 754.0, + 1405.0, + 754.0, + 1405.0, + 793.0, + 294.0, + 793.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 317.0, + 788.0, + 712.0, + 788.0, + 712.0, + 824.0, + 317.0, + 824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 799.0, + 788.0, + 847.0, + 788.0, + 847.0, + 824.0, + 799.0, + 824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 934.0, + 788.0, + 965.0, + 788.0, + 965.0, + 824.0, + 934.0, + 824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 998.0, + 788.0, + 1046.0, + 788.0, + 1046.0, + 824.0, + 998.0, + 824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1077.0, + 788.0, + 1409.0, + 788.0, + 1409.0, + 824.0, + 1077.0, + 824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1125.0, + 672.0, + 1125.0, + 672.0, + 1161.0, + 323.0, + 1161.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 341.0, + 1155.0, + 344.0, + 1155.0, + 344.0, + 1193.0, + 341.0, + 1193.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 380.0, + 1155.0, + 1380.0, + 1155.0, + 1380.0, + 1193.0, + 380.0, + 1193.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 1155.0, + 1406.0, + 1155.0, + 1406.0, + 1193.0, + 1402.0, + 1193.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1187.0, + 694.0, + 1187.0, + 694.0, + 1223.0, + 294.0, + 1223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 338.0, + 1046.0, + 1405.0, + 1046.0, + 1405.0, + 1083.0, + 338.0, + 1083.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1077.0, + 1343.0, + 1077.0, + 1343.0, + 1116.0, + 297.0, + 1116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 901.0, + 691.0, + 901.0, + 691.0, + 940.0, + 297.0, + 940.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 749.0, + 901.0, + 1246.0, + 901.0, + 1246.0, + 940.0, + 749.0, + 940.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1334.0, + 901.0, + 1374.0, + 901.0, + 1374.0, + 940.0, + 1334.0, + 940.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 998.0, + 429.0, + 998.0, + 429.0, + 1041.0, + 295.0, + 1041.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 457.0, + 998.0, + 689.0, + 998.0, + 689.0, + 1041.0, + 457.0, + 1041.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 841.0, + 998.0, + 890.0, + 998.0, + 890.0, + 1041.0, + 841.0, + 1041.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1071.0, + 998.0, + 1081.0, + 998.0, + 1081.0, + 1041.0, + 1071.0, + 1041.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 5, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 407, + 1063, + 1290, + 1063, + 1290, + 1393, + 407, + 1393 + ], + "score": 0.982, + "html": "
TransferCOCO CaptionNoCapsVQAv2GQAOKVQA
OPT125M→350M1.73.01.05.05.0
OPT125M→1.3B9.010.09.0=2.0
OPT350M→1.3B4.55.09.02.02.0
OPT125M→2.7B10.010.02.02.03.0
OPT350M→2.7B10.010.02.013.0
OPT1.3B-→2.7B3.33.32.0-1.5
FlanT5base→XL1.01.13.04.02.0
FlanT5xL →OPT2.7B5.05.02.02.03.0
OPT2.7B→FlanT5xL1.72.012.01
" + }, + { + "category_id": 1, + "poly": [ + 297, + 1421, + 1406, + 1421, + 1406, + 1641, + 297, + 1641 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 299, + 1820, + 1402, + 1820, + 1402, + 2006, + 299, + 2006 + ], + "score": 0.975 + }, + { + "category_id": 3, + "poly": [ + 295, + 192, + 1393, + 192, + 1393, + 848, + 295, + 848 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 299, + 1653, + 1404, + 1653, + 1404, + 1746, + 299, + 1746 + ], + "score": 0.971 + }, + { + "category_id": 4, + "poly": [ + 298, + 859, + 1402, + 859, + 1402, + 950, + 298, + 950 + ], + "score": 0.949 + }, + { + "category_id": 1, + "poly": [ + 294, + 1758, + 1400, + 1758, + 1400, + 1792, + 294, + 1792 + ], + "score": 0.914 + }, + { + "category_id": 6, + "poly": [ + 296, + 984, + 1402, + 984, + 1402, + 1046, + 296, + 1046 + ], + "score": 0.911 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 858, + 2061, + 858, + 2084, + 841, + 2084 + ], + "score": 0.721 + }, + { + "category_id": 13, + "poly": [ + 416, + 1715, + 567, + 1715, + 567, + 1746, + 416, + 1746 + ], + "score": 0.91, + "latex": "\\mathrm { O P T } _ { 1 . 3 \\mathrm { B } 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 1001, + 1883, + 1088, + 1883, + 1088, + 1913, + 1001, + 1913 + ], + "score": 0.89, + "latex": "\\mathbf { L L M } _ { \\mathrm { s r c } }" + }, + { + "category_id": 13, + "poly": [ + 1214, + 1914, + 1365, + 1914, + 1365, + 1944, + 1214, + 1944 + ], + "score": 0.88, + "latex": "\\mathrm { O P T } _ { 1 . 3 \\mathrm { B } 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 1364, + 1513, + 1397, + 1513, + 1397, + 1552, + 1364, + 1552 + ], + "score": 0.88, + "latex": "\\frac { e _ { \\mathrm { t f s } } } { e _ { \\mathrm { v t } } }" + }, + { + "category_id": 13, + "poly": [ + 430, + 1914, + 591, + 1914, + 591, + 1945, + 430, + 1945 + ], + "score": 0.88, + "latex": "\\mathrm { O P T } _ { 1 2 5 \\mathrm { M } 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 888, + 1517, + 922, + 1517, + 922, + 1545, + 888, + 1545 + ], + "score": 0.87, + "latex": "e _ { \\mathrm { v t } }" + }, + { + "category_id": 13, + "poly": [ + 642, + 1915, + 802, + 1915, + 802, + 1945, + 642, + 1945 + ], + "score": 0.87, + "latex": "\\mathrm { O P T } _ { 3 5 0 \\mathbf { M } 2 . 7 \\mathbf { B } }" + }, + { + "category_id": 13, + "poly": [ + 1123, + 1487, + 1161, + 1487, + 1161, + 1514, + 1123, + 1514 + ], + "score": 0.87, + "latex": "e _ { \\mathrm { { t f s } } }" + }, + { + "category_id": 13, + "poly": [ + 1158, + 1547, + 1212, + 1547, + 1212, + 1577, + 1158, + 1577 + ], + "score": 0.86, + "latex": "40 \\%" + }, + { + "category_id": 13, + "poly": [ + 298, + 1608, + 459, + 1608, + 459, + 1640, + 298, + 1640 + ], + "score": 0.86, + "latex": "\\mathrm { O P T } _ { 1 2 5 \\mathrm { M } 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 888, + 1853, + 971, + 1853, + 971, + 1886, + 888, + 1886 + ], + "score": 0.85, + "latex": "\\mathrm { L L M _ { \\mathrm { t g t } } }" + }, + { + "category_id": 13, + "poly": [ + 537, + 1822, + 635, + 1822, + 635, + 1853, + 537, + 1853 + ], + "score": 0.85, + "latex": "\\mathbf { L L M _ { s r c } } ," + }, + { + "category_id": 13, + "poly": [ + 447, + 1974, + 600, + 1974, + 600, + 2006, + 447, + 2006 + ], + "score": 0.83, + "latex": "\\mathrm { O P T } _ { 1 . 3 \\mathrm { B } 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 948, + 1714, + 1003, + 1714, + 1003, + 1746, + 948, + 1746 + ], + "score": 0.83, + "latex": "\\mathrm { \\ : \\ : \\xi { 8 C . 3 } }" + }, + { + "category_id": 13, + "poly": [ + 1194, + 1515, + 1221, + 1515, + 1221, + 1541, + 1194, + 1541 + ], + "score": 0.81, + "latex": "D" + }, + { + "category_id": 13, + "poly": [ + 921, + 1488, + 940, + 1488, + 940, + 1515, + 921, + 1515 + ], + "score": 0.8, + "latex": "p" + }, + { + "category_id": 13, + "poly": [ + 984, + 1484, + 1011, + 1484, + 1011, + 1511, + 984, + 1511 + ], + "score": 0.79, + "latex": "D" + }, + { + "category_id": 13, + "poly": [ + 759, + 1519, + 779, + 1519, + 779, + 1546, + 759, + 1546 + ], + "score": 0.77, + "latex": "p" + }, + { + "category_id": 13, + "poly": [ + 479, + 1484, + 506, + 1484, + 506, + 1510, + 479, + 1510 + ], + "score": 0.74, + "latex": "D" + }, + { + "category_id": 13, + "poly": [ + 1240, + 1947, + 1399, + 1947, + 1399, + 1976, + 1240, + 1976 + ], + "score": 0.74, + "latex": "\\mathrm { O P T } _ { 1 2 5 \\mathrm { M } 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 15, + "poly": [ + 421.0, + 195.0, + 512.0, + 195.0, + 512.0, + 218.0, + 421.0, + 218.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 659.0, + 198.0, + 713.0, + 198.0, + 713.0, + 218.0, + 659.0, + 218.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 881.0, + 199.0, + 928.0, + 199.0, + 928.0, + 215.0, + 881.0, + 215.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1107.0, + 199.0, + 1141.0, + 199.0, + 1141.0, + 215.0, + 1107.0, + 215.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1317.0, + 199.0, + 1367.0, + 199.0, + 1367.0, + 215.0, + 1317.0, + 215.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 207.0, + 340.0, + 207.0, + 340.0, + 224.0, + 315.0, + 224.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 210.0, + 559.0, + 210.0, + 559.0, + 229.0, + 532.0, + 229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 229.0, + 340.0, + 229.0, + 340.0, + 248.0, + 315.0, + 248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 227.0, + 778.0, + 227.0, + 778.0, + 245.0, + 757.0, + 245.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 975.0, + 225.0, + 996.0, + 225.0, + 996.0, + 244.0, + 975.0, + 244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 224.0, + 1213.0, + 224.0, + 1213.0, + 244.0, + 1193.0, + 244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 251.0, + 341.0, + 251.0, + 341.0, + 270.0, + 315.0, + 270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 246.0, + 559.0, + 246.0, + 559.0, + 265.0, + 532.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 255.0, + 778.0, + 255.0, + 778.0, + 270.0, + 757.0, + 270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 253.0, + 996.0, + 253.0, + 996.0, + 271.0, + 974.0, + 271.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 249.0, + 1215.0, + 249.0, + 1215.0, + 269.0, + 1193.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 300.0, + 274.0, + 341.0, + 274.0, + 341.0, + 321.0, + 300.0, + 321.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 518.0, + 273.0, + 559.0, + 273.0, + 559.0, + 323.0, + 518.0, + 323.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 274.0, + 1215.0, + 274.0, + 1215.0, + 294.0, + 1193.0, + 294.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 296.0, + 340.0, + 296.0, + 340.0, + 316.0, + 315.0, + 316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 307.0, + 779.0, + 307.0, + 779.0, + 326.0, + 757.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 975.0, + 306.0, + 996.0, + 306.0, + 996.0, + 325.0, + 975.0, + 325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 319.0, + 341.0, + 319.0, + 341.0, + 338.0, + 315.0, + 338.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 330.0, + 461.0, + 330.0, + 461.0, + 346.0, + 351.0, + 346.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 319.0, + 559.0, + 319.0, + 559.0, + 337.0, + 537.0, + 337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 591.0, + 331.0, + 678.0, + 331.0, + 678.0, + 347.0, + 591.0, + 347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 755.0, + 333.0, + 778.0, + 333.0, + 778.0, + 352.0, + 755.0, + 352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 810.0, + 331.0, + 896.0, + 331.0, + 896.0, + 347.0, + 810.0, + 347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 333.0, + 996.0, + 333.0, + 996.0, + 352.0, + 974.0, + 352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1027.0, + 330.0, + 1115.0, + 330.0, + 1115.0, + 346.0, + 1027.0, + 346.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 324.0, + 1216.0, + 324.0, + 1216.0, + 343.0, + 1193.0, + 343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 331.0, + 1334.0, + 331.0, + 1334.0, + 347.0, + 1248.0, + 347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 342.0, + 341.0, + 342.0, + 341.0, + 383.0, + 314.0, + 383.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 341.0, + 436.0, + 341.0, + 436.0, + 378.0, + 354.0, + 378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 355.0, + 559.0, + 355.0, + 559.0, + 373.0, + 537.0, + 373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 572.0, + 363.0, + 586.0, + 363.0, + 586.0, + 372.0, + 572.0, + 372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 591.0, + 343.0, + 654.0, + 343.0, + 654.0, + 378.0, + 591.0, + 378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 755.0, + 361.0, + 778.0, + 361.0, + 778.0, + 379.0, + 755.0, + 379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 792.0, + 348.0, + 807.0, + 348.0, + 807.0, + 372.0, + 792.0, + 372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 809.0, + 343.0, + 873.0, + 343.0, + 873.0, + 378.0, + 809.0, + 378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 361.0, + 997.0, + 361.0, + 997.0, + 379.0, + 974.0, + 379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1011.0, + 363.0, + 1021.0, + 363.0, + 1021.0, + 372.0, + 1011.0, + 372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1027.0, + 343.0, + 1091.0, + 343.0, + 1091.0, + 378.0, + 1027.0, + 378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 348.0, + 1216.0, + 348.0, + 1216.0, + 367.0, + 1193.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1226.0, + 363.0, + 1240.0, + 363.0, + 1240.0, + 373.0, + 1226.0, + 373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1247.0, + 341.0, + 1310.0, + 341.0, + 1310.0, + 378.0, + 1247.0, + 378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 373.0, + 1216.0, + 373.0, + 1216.0, + 392.0, + 1193.0, + 392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 336.0, + 384.0, + 346.0, + 384.0, + 346.0, + 397.0, + 336.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 384.0, + 403.0, + 384.0, + 403.0, + 399.0, + 351.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 383.0, + 423.0, + 383.0, + 423.0, + 398.0, + 405.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 425.0, + 384.0, + 442.0, + 384.0, + 442.0, + 398.0, + 425.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 445.0, + 382.0, + 520.0, + 382.0, + 520.0, + 402.0, + 445.0, + 402.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 554.0, + 386.0, + 566.0, + 386.0, + 566.0, + 397.0, + 554.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 568.0, + 383.0, + 606.0, + 383.0, + 606.0, + 399.0, + 568.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 608.0, + 382.0, + 737.0, + 382.0, + 737.0, + 399.0, + 608.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 772.0, + 384.0, + 783.0, + 384.0, + 783.0, + 397.0, + 772.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 386.0, + 802.0, + 386.0, + 802.0, + 397.0, + 789.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 807.0, + 384.0, + 823.0, + 384.0, + 823.0, + 399.0, + 807.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 825.0, + 384.0, + 842.0, + 384.0, + 842.0, + 398.0, + 825.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 843.0, + 384.0, + 861.0, + 384.0, + 861.0, + 398.0, + 843.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 863.0, + 384.0, + 880.0, + 384.0, + 880.0, + 398.0, + 863.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 881.0, + 382.0, + 956.0, + 382.0, + 956.0, + 402.0, + 881.0, + 402.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 991.0, + 384.0, + 1001.0, + 384.0, + 1001.0, + 397.0, + 991.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1007.0, + 384.0, + 1041.0, + 384.0, + 1041.0, + 399.0, + 1007.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1042.0, + 384.0, + 1060.0, + 384.0, + 1060.0, + 398.0, + 1042.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1061.0, + 384.0, + 1080.0, + 384.0, + 1080.0, + 398.0, + 1061.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1081.0, + 384.0, + 1099.0, + 384.0, + 1099.0, + 398.0, + 1081.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 384.0, + 1116.0, + 384.0, + 1116.0, + 398.0, + 1100.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1117.0, + 383.0, + 1173.0, + 383.0, + 1173.0, + 399.0, + 1117.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1211.0, + 386.0, + 1220.0, + 386.0, + 1220.0, + 397.0, + 1211.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1226.0, + 383.0, + 1261.0, + 383.0, + 1261.0, + 399.0, + 1226.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1263.0, + 386.0, + 1276.0, + 386.0, + 1276.0, + 397.0, + 1263.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1277.0, + 382.0, + 1392.0, + 382.0, + 1392.0, + 399.0, + 1277.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 401.0, + 396.0, + 450.0, + 396.0, + 450.0, + 414.0, + 401.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 616.0, + 394.0, + 666.0, + 394.0, + 666.0, + 411.0, + 616.0, + 411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 396.0, + 884.0, + 396.0, + 884.0, + 413.0, + 838.0, + 413.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 396.0, + 1104.0, + 396.0, + 1104.0, + 413.0, + 1057.0, + 413.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1275.0, + 396.0, + 1321.0, + 396.0, + 1321.0, + 413.0, + 1275.0, + 413.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 421.0, + 416.0, + 511.0, + 416.0, + 511.0, + 435.0, + 421.0, + 435.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 574.0, + 414.0, + 711.0, + 414.0, + 711.0, + 438.0, + 574.0, + 438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 880.0, + 417.0, + 926.0, + 417.0, + 926.0, + 434.0, + 880.0, + 434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1105.0, + 417.0, + 1139.0, + 417.0, + 1139.0, + 434.0, + 1105.0, + 434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1313.0, + 416.0, + 1366.0, + 416.0, + 1366.0, + 435.0, + 1313.0, + 435.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 430.0, + 341.0, + 430.0, + 341.0, + 449.0, + 315.0, + 449.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 533.0, + 445.0, + 559.0, + 445.0, + 559.0, + 464.0, + 533.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 438.0, + 778.0, + 438.0, + 778.0, + 457.0, + 757.0, + 457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 975.0, + 443.0, + 996.0, + 443.0, + 996.0, + 462.0, + 975.0, + 462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 440.0, + 1216.0, + 440.0, + 1216.0, + 459.0, + 1193.0, + 459.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 455.0, + 340.0, + 455.0, + 340.0, + 475.0, + 315.0, + 475.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 758.0, + 474.0, + 778.0, + 474.0, + 778.0, + 489.0, + 758.0, + 489.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 470.0, + 1216.0, + 470.0, + 1216.0, + 489.0, + 1193.0, + 489.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 300.0, + 481.0, + 341.0, + 481.0, + 341.0, + 540.0, + 300.0, + 540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 520.0, + 488.0, + 559.0, + 488.0, + 559.0, + 537.0, + 520.0, + 537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 740.0, + 499.0, + 783.0, + 499.0, + 783.0, + 539.0, + 740.0, + 539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 484.0, + 996.0, + 484.0, + 996.0, + 532.0, + 964.0, + 532.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 532.0, + 341.0, + 532.0, + 341.0, + 551.0, + 315.0, + 551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 529.0, + 559.0, + 529.0, + 559.0, + 547.0, + 538.0, + 547.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 975.0, + 525.0, + 996.0, + 525.0, + 996.0, + 544.0, + 975.0, + 544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 531.0, + 1215.0, + 531.0, + 1215.0, + 550.0, + 1193.0, + 550.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 550.0, + 461.0, + 550.0, + 461.0, + 566.0, + 372.0, + 566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 550.0, + 678.0, + 550.0, + 678.0, + 566.0, + 592.0, + 566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 755.0, + 541.0, + 778.0, + 541.0, + 778.0, + 560.0, + 755.0, + 560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 810.0, + 550.0, + 896.0, + 550.0, + 896.0, + 566.0, + 810.0, + 566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1029.0, + 550.0, + 1116.0, + 550.0, + 1116.0, + 566.0, + 1029.0, + 566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 550.0, + 1337.0, + 550.0, + 1337.0, + 566.0, + 1248.0, + 566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 557.0, + 341.0, + 557.0, + 341.0, + 576.0, + 314.0, + 576.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 357.0, + 567.0, + 370.0, + 567.0, + 370.0, + 576.0, + 357.0, + 576.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 562.0, + 436.0, + 562.0, + 436.0, + 582.0, + 372.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 591.0, + 562.0, + 654.0, + 562.0, + 654.0, + 582.0, + 591.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 792.0, + 567.0, + 807.0, + 567.0, + 807.0, + 576.0, + 792.0, + 576.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 810.0, + 562.0, + 873.0, + 562.0, + 873.0, + 582.0, + 810.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 566.0, + 997.0, + 566.0, + 997.0, + 585.0, + 974.0, + 585.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1029.0, + 562.0, + 1091.0, + 562.0, + 1091.0, + 582.0, + 1029.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 561.0, + 1216.0, + 561.0, + 1216.0, + 580.0, + 1193.0, + 580.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1247.0, + 562.0, + 1309.0, + 562.0, + 1309.0, + 582.0, + 1247.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 583.0, + 341.0, + 583.0, + 341.0, + 601.0, + 315.0, + 601.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 581.0, + 366.0, + 581.0, + 366.0, + 591.0, + 352.0, + 591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 573.0, + 431.0, + 573.0, + 431.0, + 597.0, + 372.0, + 597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 571.0, + 559.0, + 571.0, + 559.0, + 590.0, + 537.0, + 590.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 572.0, + 581.0, + 587.0, + 581.0, + 587.0, + 591.0, + 572.0, + 591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 590.0, + 573.0, + 650.0, + 573.0, + 650.0, + 597.0, + 590.0, + 597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 755.0, + 576.0, + 778.0, + 576.0, + 778.0, + 595.0, + 755.0, + 595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 792.0, + 581.0, + 804.0, + 581.0, + 804.0, + 591.0, + 792.0, + 591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 810.0, + 573.0, + 868.0, + 573.0, + 868.0, + 597.0, + 810.0, + 597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1011.0, + 582.0, + 1021.0, + 582.0, + 1021.0, + 591.0, + 1011.0, + 591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1027.0, + 573.0, + 1086.0, + 573.0, + 1086.0, + 597.0, + 1027.0, + 597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1226.0, + 582.0, + 1241.0, + 582.0, + 1241.0, + 591.0, + 1226.0, + 591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1247.0, + 573.0, + 1306.0, + 573.0, + 1306.0, + 597.0, + 1247.0, + 597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 336.0, + 603.0, + 347.0, + 603.0, + 347.0, + 614.0, + 336.0, + 614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 600.0, + 520.0, + 600.0, + 520.0, + 619.0, + 349.0, + 619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 554.0, + 600.0, + 738.0, + 600.0, + 738.0, + 619.0, + 554.0, + 619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 770.0, + 602.0, + 787.0, + 602.0, + 787.0, + 616.0, + 770.0, + 616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 788.0, + 598.0, + 956.0, + 598.0, + 956.0, + 619.0, + 788.0, + 619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 991.0, + 603.0, + 1002.0, + 603.0, + 1002.0, + 614.0, + 991.0, + 614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1007.0, + 600.0, + 1175.0, + 600.0, + 1175.0, + 619.0, + 1007.0, + 619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1240.0, + 598.0, + 1392.0, + 598.0, + 1392.0, + 618.0, + 1240.0, + 618.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 401.0, + 613.0, + 450.0, + 613.0, + 450.0, + 632.0, + 401.0, + 632.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 619.0, + 613.0, + 666.0, + 613.0, + 666.0, + 631.0, + 619.0, + 631.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 613.0, + 884.0, + 613.0, + 884.0, + 631.0, + 838.0, + 631.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 613.0, + 1102.0, + 613.0, + 1102.0, + 631.0, + 1057.0, + 631.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1275.0, + 613.0, + 1323.0, + 613.0, + 1323.0, + 632.0, + 1275.0, + 632.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 312.0, + 649.0, + 342.0, + 649.0, + 342.0, + 672.0, + 312.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 421.0, + 632.0, + 511.0, + 632.0, + 511.0, + 655.0, + 421.0, + 655.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 649.0, + 559.0, + 649.0, + 559.0, + 668.0, + 532.0, + 668.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 657.0, + 634.0, + 712.0, + 634.0, + 712.0, + 654.0, + 657.0, + 654.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 745.0, + 641.0, + 780.0, + 641.0, + 780.0, + 662.0, + 745.0, + 662.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 795.0, + 630.0, + 926.0, + 630.0, + 926.0, + 658.0, + 795.0, + 658.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 652.0, + 996.0, + 652.0, + 996.0, + 670.0, + 974.0, + 670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1105.0, + 634.0, + 1139.0, + 634.0, + 1139.0, + 652.0, + 1105.0, + 652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1316.0, + 636.0, + 1366.0, + 636.0, + 1366.0, + 652.0, + 1316.0, + 652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 745.0, + 662.0, + 780.0, + 662.0, + 780.0, + 684.0, + 745.0, + 684.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 664.0, + 1215.0, + 664.0, + 1215.0, + 683.0, + 1193.0, + 683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 312.0, + 674.0, + 342.0, + 674.0, + 342.0, + 697.0, + 312.0, + 697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 674.0, + 559.0, + 674.0, + 559.0, + 693.0, + 537.0, + 693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 675.0, + 996.0, + 675.0, + 996.0, + 695.0, + 974.0, + 695.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 300.0, + 702.0, + 341.0, + 702.0, + 341.0, + 757.0, + 300.0, + 757.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 518.0, + 699.0, + 559.0, + 699.0, + 559.0, + 757.0, + 518.0, + 757.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 732.0, + 684.0, + 783.0, + 684.0, + 783.0, + 755.0, + 732.0, + 755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 959.0, + 700.0, + 1001.0, + 700.0, + 1001.0, + 756.0, + 959.0, + 756.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1177.0, + 693.0, + 1220.0, + 693.0, + 1220.0, + 756.0, + 1177.0, + 756.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 319.0, + 753.0, + 341.0, + 753.0, + 341.0, + 771.0, + 319.0, + 771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 749.0, + 559.0, + 749.0, + 559.0, + 767.0, + 538.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 745.0, + 750.0, + 780.0, + 750.0, + 780.0, + 771.0, + 745.0, + 771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 749.0, + 996.0, + 749.0, + 996.0, + 767.0, + 974.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 750.0, + 1216.0, + 750.0, + 1216.0, + 769.0, + 1193.0, + 769.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 319.0, + 777.0, + 341.0, + 777.0, + 341.0, + 796.0, + 319.0, + 796.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 767.0, + 461.0, + 767.0, + 461.0, + 801.0, + 372.0, + 801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 774.0, + 559.0, + 774.0, + 559.0, + 792.0, + 537.0, + 792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 591.0, + 767.0, + 679.0, + 767.0, + 679.0, + 800.0, + 591.0, + 800.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 745.0, + 771.0, + 780.0, + 771.0, + 780.0, + 792.0, + 745.0, + 792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 810.0, + 767.0, + 896.0, + 767.0, + 896.0, + 801.0, + 810.0, + 801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 772.0, + 997.0, + 772.0, + 997.0, + 792.0, + 974.0, + 792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1027.0, + 767.0, + 1114.0, + 767.0, + 1114.0, + 801.0, + 1027.0, + 801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 777.0, + 1216.0, + 777.0, + 1216.0, + 796.0, + 1193.0, + 796.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1247.0, + 767.0, + 1333.0, + 767.0, + 1333.0, + 800.0, + 1247.0, + 800.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 800.0, + 366.0, + 800.0, + 366.0, + 808.0, + 354.0, + 808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 795.0, + 431.0, + 795.0, + 431.0, + 815.0, + 372.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 799.0, + 559.0, + 799.0, + 559.0, + 817.0, + 537.0, + 817.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 572.0, + 800.0, + 587.0, + 800.0, + 587.0, + 810.0, + 572.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 591.0, + 795.0, + 649.0, + 795.0, + 649.0, + 815.0, + 591.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 745.0, + 794.0, + 780.0, + 794.0, + 780.0, + 815.0, + 745.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 792.0, + 800.0, + 803.0, + 800.0, + 803.0, + 810.0, + 792.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 810.0, + 792.0, + 868.0, + 792.0, + 868.0, + 816.0, + 810.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 797.0, + 997.0, + 797.0, + 997.0, + 816.0, + 974.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1011.0, + 800.0, + 1021.0, + 800.0, + 1021.0, + 808.0, + 1011.0, + 808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1027.0, + 795.0, + 1086.0, + 795.0, + 1086.0, + 815.0, + 1027.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1226.0, + 800.0, + 1241.0, + 800.0, + 1241.0, + 810.0, + 1226.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1247.0, + 795.0, + 1306.0, + 795.0, + 1306.0, + 815.0, + 1247.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 805.0, + 339.0, + 805.0, + 339.0, + 820.0, + 321.0, + 820.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 806.0, + 1217.0, + 806.0, + 1217.0, + 825.0, + 1193.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 336.0, + 821.0, + 346.0, + 821.0, + 346.0, + 832.0, + 336.0, + 832.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 817.0, + 520.0, + 817.0, + 520.0, + 837.0, + 347.0, + 837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 554.0, + 817.0, + 738.0, + 817.0, + 738.0, + 837.0, + 554.0, + 837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 769.0, + 817.0, + 956.0, + 817.0, + 956.0, + 837.0, + 769.0, + 837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 988.0, + 817.0, + 1175.0, + 817.0, + 1175.0, + 837.0, + 988.0, + 837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1211.0, + 821.0, + 1221.0, + 821.0, + 1221.0, + 833.0, + 1211.0, + 833.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1222.0, + 816.0, + 1393.0, + 816.0, + 1393.0, + 837.0, + 1222.0, + 837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 401.0, + 832.0, + 450.0, + 832.0, + 450.0, + 851.0, + 401.0, + 851.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 619.0, + 832.0, + 668.0, + 832.0, + 668.0, + 851.0, + 619.0, + 851.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 832.0, + 884.0, + 832.0, + 884.0, + 850.0, + 838.0, + 850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 832.0, + 1102.0, + 832.0, + 1102.0, + 850.0, + 1057.0, + 850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1275.0, + 832.0, + 1323.0, + 832.0, + 1323.0, + 851.0, + 1275.0, + 851.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 334.0, + 197.5, + 423.0, + 197.5, + 423.0, + 218.5, + 334.0, + 218.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 573.0, + 198.5, + 659.0, + 198.5, + 659.0, + 218.5, + 573.0, + 218.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.0, + 198.5, + 880.0, + 198.5, + 880.0, + 218.5, + 793.0, + 218.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1017.0, + 198.5, + 1103.0, + 198.5, + 1103.0, + 218.5, + 1017.0, + 218.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1228.0, + 198.5, + 1314.0, + 198.5, + 1314.0, + 218.5, + 1228.0, + 218.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1190.0, + 291.5, + 1201.0, + 291.5, + 1201.0, + 316.0, + 1190.0, + 316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 728.0, + 283.0, + 789.0, + 283.0, + 789.0, + 307.0, + 728.0, + 307.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 946.25, + 283.0, + 1006.25, + 283.0, + 1006.25, + 306.0, + 946.25, + 306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 337.0, + 416.0, + 420.0, + 416.0, + 420.0, + 437.5, + 337.0, + 437.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 795.0, + 416.0, + 878.0, + 416.0, + 878.0, + 437.5, + 795.0, + 437.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1019.0, + 416.0, + 1102.0, + 416.0, + 1102.0, + 437.5, + 1019.0, + 437.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 416.0, + 1311.0, + 416.0, + 1311.0, + 437.0, + 1229.0, + 437.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1164.0, + 500.0, + 1225.0, + 500.0, + 1225.0, + 527.5, + 1164.0, + 527.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1200.0, + 594.0, + 1218.0, + 594.0, + 1218.0, + 613.0, + 1200.0, + 613.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1226.75, + 600.5, + 1238.75, + 600.5, + 1238.75, + 617.0, + 1226.75, + 617.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 337.25, + 634.0, + 421.25, + 634.0, + 421.25, + 655.5, + 337.25, + 655.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 572.75, + 634.0, + 659.75, + 634.0, + 659.75, + 655.5, + 572.75, + 655.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1018.25, + 634.0, + 1101.25, + 634.0, + 1101.25, + 655.5, + 1018.25, + 655.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 634.0, + 1314.0, + 634.0, + 1314.0, + 654.0, + 1229.0, + 654.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 858.0, + 1406.0, + 858.0, + 1406.0, + 893.0, + 293.0, + 893.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 888.0, + 1404.0, + 888.0, + 1404.0, + 924.0, + 293.0, + 924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 918.0, + 592.0, + 918.0, + 592.0, + 951.0, + 295.0, + 951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 982.0, + 1406.0, + 982.0, + 1406.0, + 1021.0, + 292.0, + 1021.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1006.0, + 1050.0, + 1006.0, + 1050.0, + 1052.0, + 291.0, + 1052.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2059.0, + 860.0, + 2059.0, + 860.0, + 2091.0, + 839.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1421.0, + 1406.0, + 1421.0, + 1406.0, + 1460.0, + 294.0, + 1460.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1452.0, + 1406.0, + 1452.0, + 1406.0, + 1487.0, + 295.0, + 1487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1481.0, + 478.0, + 1481.0, + 478.0, + 1518.0, + 292.0, + 1518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 507.0, + 1481.0, + 920.0, + 1481.0, + 920.0, + 1518.0, + 507.0, + 1518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1481.0, + 983.0, + 1481.0, + 983.0, + 1518.0, + 941.0, + 1518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1012.0, + 1481.0, + 1122.0, + 1481.0, + 1122.0, + 1518.0, + 1012.0, + 1518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1162.0, + 1481.0, + 1406.0, + 1481.0, + 1406.0, + 1518.0, + 1162.0, + 1518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1510.0, + 758.0, + 1510.0, + 758.0, + 1552.0, + 290.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 780.0, + 1510.0, + 887.0, + 1510.0, + 887.0, + 1552.0, + 780.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 923.0, + 1510.0, + 1193.0, + 1510.0, + 1193.0, + 1552.0, + 923.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1222.0, + 1510.0, + 1363.0, + 1510.0, + 1363.0, + 1552.0, + 1222.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1398.0, + 1510.0, + 1408.0, + 1510.0, + 1408.0, + 1552.0, + 1398.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1545.0, + 1157.0, + 1545.0, + 1157.0, + 1582.0, + 294.0, + 1582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1213.0, + 1545.0, + 1406.0, + 1545.0, + 1406.0, + 1582.0, + 1213.0, + 1582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1576.0, + 1404.0, + 1576.0, + 1404.0, + 1614.0, + 292.0, + 1614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 460.0, + 1603.0, + 790.0, + 1603.0, + 790.0, + 1646.0, + 460.0, + 1646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1820.0, + 536.0, + 1820.0, + 536.0, + 1854.0, + 295.0, + 1854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 636.0, + 1820.0, + 1404.0, + 1820.0, + 1404.0, + 1854.0, + 636.0, + 1854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1852.0, + 887.0, + 1852.0, + 887.0, + 1886.0, + 294.0, + 1886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 972.0, + 1852.0, + 1404.0, + 1852.0, + 1404.0, + 1886.0, + 972.0, + 1886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1884.0, + 1000.0, + 1884.0, + 1000.0, + 1916.0, + 295.0, + 1916.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1089.0, + 1884.0, + 1403.0, + 1884.0, + 1403.0, + 1916.0, + 1089.0, + 1916.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1908.0, + 429.0, + 1908.0, + 429.0, + 1951.0, + 292.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 1908.0, + 641.0, + 1908.0, + 641.0, + 1951.0, + 592.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 803.0, + 1908.0, + 1213.0, + 1908.0, + 1213.0, + 1951.0, + 803.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1366.0, + 1908.0, + 1407.0, + 1908.0, + 1407.0, + 1951.0, + 1366.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1937.0, + 1239.0, + 1937.0, + 1239.0, + 1983.0, + 292.0, + 1983.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1400.0, + 1937.0, + 1406.0, + 1937.0, + 1406.0, + 1983.0, + 1400.0, + 1983.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1970.0, + 446.0, + 1970.0, + 446.0, + 2009.0, + 293.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 601.0, + 1970.0, + 795.0, + 1970.0, + 795.0, + 2009.0, + 601.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1651.0, + 1408.0, + 1651.0, + 1408.0, + 1689.0, + 295.0, + 1689.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1683.0, + 1404.0, + 1683.0, + 1404.0, + 1718.0, + 295.0, + 1718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1713.0, + 415.0, + 1713.0, + 415.0, + 1749.0, + 294.0, + 1749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 568.0, + 1713.0, + 947.0, + 1713.0, + 947.0, + 1749.0, + 568.0, + 1749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1004.0, + 1713.0, + 1263.0, + 1713.0, + 1263.0, + 1749.0, + 1004.0, + 1749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1753.0, + 1405.0, + 1753.0, + 1405.0, + 1800.0, + 293.0, + 1800.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 6, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1180, + 1405, + 1180, + 1405, + 1486, + 297, + 1486 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 298, + 1567, + 1047, + 1567, + 1047, + 1690, + 298, + 1690 + ], + "score": 0.968 + }, + { + "category_id": 5, + "poly": [ + 301, + 271, + 1404, + 271, + 1404, + 524, + 301, + 524 + ], + "score": 0.964, + "html": "
ModelsVQAv2 valGQA test-devOKVQA testGPU hourstraining data
BLIP-2 ViT-G OPT6.7B54.336.436.4631.5129M
BLIP-2 ViT-G OPT2.7B-→6.7B (0urs)57.236.239.859.013.8M
VL-LLaMA7B (ours)58.137.537.467.113.8M
BLIP-2 ViT-G FlanT5xXL65.244.745.9684.0121.6M
BLIP-2 ViT-G FlanT5xL-→XXL (ours)65.245.045.032.45.3M
" + }, + { + "category_id": 3, + "poly": [ + 297, + 573, + 1392, + 573, + 1392, + 1035, + 297, + 1035 + ], + "score": 0.963 + }, + { + "category_id": 4, + "poly": [ + 299, + 1050, + 1401, + 1050, + 1401, + 1142, + 299, + 1142 + ], + "score": 0.954 + }, + { + "category_id": 3, + "poly": [ + 1068, + 1527, + 1402, + 1527, + 1402, + 1803, + 1068, + 1803 + ], + "score": 0.95 + }, + { + "category_id": 6, + "poly": [ + 296, + 193, + 1399, + 193, + 1399, + 257, + 296, + 257 + ], + "score": 0.944 + }, + { + "category_id": 1, + "poly": [ + 298, + 1720, + 1047, + 1720, + 1047, + 2026, + 298, + 2026 + ], + "score": 0.939 + }, + { + "category_id": 4, + "poly": [ + 1066, + 1810, + 1404, + 1810, + 1404, + 2023, + 1066, + 2023 + ], + "score": 0.933 + }, + { + "category_id": 0, + "poly": [ + 300, + 1519, + 622, + 1519, + 622, + 1552, + 300, + 1552 + ], + "score": 0.924 + }, + { + "category_id": 2, + "poly": [ + 840, + 2061, + 858, + 2061, + 858, + 2085, + 840, + 2085 + ], + "score": 0.807 + }, + { + "category_id": 13, + "poly": [ + 1071, + 1334, + 1167, + 1334, + 1167, + 1365, + 1071, + 1365 + ], + "score": 0.9, + "latex": "\\mathbf { L L M } _ { t a r }" + }, + { + "category_id": 13, + "poly": [ + 296, + 1629, + 390, + 1629, + 390, + 1660, + 296, + 1660 + ], + "score": 0.89, + "latex": "\\mathrm { O P T } _ { 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 564, + 1364, + 659, + 1364, + 659, + 1395, + 564, + 1395 + ], + "score": 0.89, + "latex": "\\mathbf { L L M } _ { s r c }" + }, + { + "category_id": 13, + "poly": [ + 970, + 1394, + 1065, + 1394, + 1065, + 1425, + 970, + 1425 + ], + "score": 0.88, + "latex": "\\mathbf { L L M } _ { s r c }" + }, + { + "category_id": 13, + "poly": [ + 433, + 1274, + 528, + 1274, + 528, + 1304, + 433, + 1304 + ], + "score": 0.87, + "latex": "\\mathbf { L L M } _ { s r c }" + }, + { + "category_id": 13, + "poly": [ + 420, + 1629, + 512, + 1629, + 512, + 1660, + 420, + 1660 + ], + "score": 0.86, + "latex": "\\mathrm { O P T } _ { 6 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 633, + 1783, + 706, + 1783, + 706, + 1811, + 633, + 1811 + ], + "score": 0.86, + "latex": "1 0 . 7 \\%" + }, + { + "category_id": 13, + "poly": [ + 581, + 1751, + 744, + 1751, + 744, + 1782, + 581, + 1782 + ], + "score": 0.85, + "latex": "\\mathbf { O P T } _ { 2 . 7 \\mathbf { B } 6 . 7 \\mathbf { B } } ;" + }, + { + "category_id": 13, + "poly": [ + 403, + 1874, + 446, + 1874, + 446, + 1900, + 403, + 1900 + ], + "score": 0.84, + "latex": "5 \\%" + }, + { + "category_id": 13, + "poly": [ + 1169, + 1993, + 1262, + 1993, + 1262, + 2024, + 1169, + 2024 + ], + "score": 0.83, + "latex": "\\mathrm { L L M } _ { s r c }" + }, + { + "category_id": 13, + "poly": [ + 573, + 1659, + 627, + 1659, + 627, + 1691, + 573, + 1691 + ], + "score": 0.82, + "latex": "\\mathrm { \\ : \\ S C . 4 }" + }, + { + "category_id": 13, + "poly": [ + 370, + 1843, + 477, + 1843, + 477, + 1873, + 370, + 1873 + ], + "score": 0.65, + "latex": "\\bf \\Pi ^ { \\prime } 5 _ { X L X X L }" + }, + { + "category_id": 13, + "poly": [ + 948, + 1934, + 1042, + 1934, + 1042, + 1964, + 948, + 1964 + ], + "score": 0.62, + "latex": "\\mathsf { \\Pi } _ { \\mathrm { X L X X L } }" + }, + { + "category_id": 13, + "poly": [ + 297, + 225, + 391, + 225, + 391, + 257, + 297, + 257 + ], + "score": 0.58, + "latex": "\\mathrm { O P T } _ { 6 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 847, + 1630, + 889, + 1630, + 889, + 1660, + 847, + 1660 + ], + "score": 0.49, + "latex": "5 _ { \\mathrm { X L } }" + }, + { + "category_id": 13, + "poly": [ + 677, + 226, + 732, + 226, + 732, + 256, + 677, + 256 + ], + "score": 0.34, + "latex": "5 _ { \\mathrm { X X L } }" + }, + { + "category_id": 13, + "poly": [ + 944, + 1051, + 1018, + 1051, + 1018, + 1080, + 944, + 1080 + ], + "score": 0.3, + "latex": "2 \\ \\mathrm { T a T }" + }, + { + "category_id": 15, + "poly": [ + 377.0, + 572.0, + 474.0, + 572.0, + 474.0, + 596.0, + 377.0, + 596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 615.0, + 572.0, + 672.0, + 572.0, + 672.0, + 596.0, + 615.0, + 596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 836.0, + 573.0, + 887.0, + 573.0, + 887.0, + 595.0, + 836.0, + 595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1061.0, + 573.0, + 1100.0, + 573.0, + 1100.0, + 596.0, + 1061.0, + 596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1270.0, + 573.0, + 1326.0, + 573.0, + 1326.0, + 595.0, + 1270.0, + 595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 607.0, + 342.0, + 607.0, + 342.0, + 626.0, + 315.0, + 626.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 583.0, + 500.0, + 583.0, + 500.0, + 614.0, + 349.0, + 614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 610.0, + 561.0, + 610.0, + 561.0, + 628.0, + 538.0, + 628.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 567.0, + 583.0, + 719.0, + 583.0, + 719.0, + 614.0, + 567.0, + 614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 605.0, + 780.0, + 605.0, + 780.0, + 623.0, + 756.0, + 623.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 786.0, + 583.0, + 937.0, + 583.0, + 937.0, + 614.0, + 786.0, + 614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 606.0, + 999.0, + 606.0, + 999.0, + 627.0, + 964.0, + 627.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1004.0, + 583.0, + 1155.0, + 583.0, + 1155.0, + 614.0, + 1004.0, + 614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1182.0, + 600.0, + 1217.0, + 600.0, + 1217.0, + 623.0, + 1182.0, + 623.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1222.0, + 583.0, + 1372.0, + 583.0, + 1372.0, + 615.0, + 1222.0, + 615.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 631.0, + 342.0, + 631.0, + 342.0, + 651.0, + 315.0, + 651.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 630.0, + 999.0, + 630.0, + 999.0, + 652.0, + 964.0, + 652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1182.0, + 625.0, + 1217.0, + 625.0, + 1217.0, + 648.0, + 1182.0, + 648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 646.0, + 561.0, + 646.0, + 561.0, + 666.0, + 538.0, + 666.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 643.0, + 780.0, + 643.0, + 780.0, + 663.0, + 756.0, + 663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 827.0, + 651.0, + 838.0, + 651.0, + 838.0, + 661.0, + 827.0, + 661.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1182.0, + 650.0, + 1217.0, + 650.0, + 1217.0, + 672.0, + 1182.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 301.0, + 656.0, + 342.0, + 656.0, + 342.0, + 712.0, + 301.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 516.0, + 657.0, + 569.0, + 657.0, + 569.0, + 721.0, + 516.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 740.0, + 669.0, + 784.0, + 669.0, + 784.0, + 711.0, + 740.0, + 711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 949.0, + 655.0, + 1004.0, + 655.0, + 1004.0, + 711.0, + 949.0, + 711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1167.0, + 669.0, + 1222.0, + 669.0, + 1222.0, + 712.0, + 1167.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 706.0, + 342.0, + 706.0, + 342.0, + 726.0, + 320.0, + 726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 705.0, + 999.0, + 705.0, + 999.0, + 727.0, + 964.0, + 727.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1184.0, + 702.0, + 1215.0, + 702.0, + 1215.0, + 721.0, + 1184.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 319.0, + 731.0, + 342.0, + 731.0, + 342.0, + 751.0, + 319.0, + 751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 383.0, + 726.0, + 490.0, + 726.0, + 490.0, + 746.0, + 383.0, + 746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 720.0, + 561.0, + 720.0, + 561.0, + 740.0, + 538.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 601.0, + 726.0, + 709.0, + 726.0, + 709.0, + 746.0, + 601.0, + 746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 722.0, + 780.0, + 722.0, + 780.0, + 743.0, + 756.0, + 743.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 818.0, + 726.0, + 928.0, + 726.0, + 928.0, + 746.0, + 818.0, + 746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 721.0, + 1005.0, + 721.0, + 1005.0, + 753.0, + 964.0, + 753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1037.0, + 726.0, + 1146.0, + 726.0, + 1146.0, + 746.0, + 1037.0, + 746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1183.0, + 726.0, + 1217.0, + 726.0, + 1217.0, + 749.0, + 1183.0, + 749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1254.0, + 726.0, + 1365.0, + 726.0, + 1365.0, + 746.0, + 1254.0, + 746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 327.0, + 758.0, + 340.0, + 758.0, + 340.0, + 774.0, + 327.0, + 774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 382.0, + 741.0, + 453.0, + 741.0, + 453.0, + 765.0, + 382.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 545.0, + 756.0, + 567.0, + 756.0, + 567.0, + 775.0, + 545.0, + 775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 601.0, + 741.0, + 672.0, + 741.0, + 672.0, + 765.0, + 601.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 819.0, + 741.0, + 890.0, + 741.0, + 890.0, + 765.0, + 819.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 755.0, + 999.0, + 755.0, + 999.0, + 778.0, + 964.0, + 778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1037.0, + 744.0, + 1107.0, + 744.0, + 1107.0, + 764.0, + 1037.0, + 764.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1183.0, + 751.0, + 1218.0, + 751.0, + 1218.0, + 773.0, + 1183.0, + 773.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1256.0, + 741.0, + 1327.0, + 741.0, + 1327.0, + 765.0, + 1256.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 337.0, + 775.0, + 348.0, + 775.0, + 348.0, + 786.0, + 337.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 773.0, + 388.0, + 773.0, + 388.0, + 789.0, + 349.0, + 789.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 391.0, + 770.0, + 521.0, + 770.0, + 521.0, + 791.0, + 391.0, + 791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 554.0, + 775.0, + 567.0, + 775.0, + 567.0, + 786.0, + 554.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 568.0, + 770.0, + 739.0, + 770.0, + 739.0, + 790.0, + 568.0, + 790.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 786.0, + 767.0, + 958.0, + 767.0, + 958.0, + 792.0, + 786.0, + 792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 988.0, + 770.0, + 1040.0, + 770.0, + 1040.0, + 791.0, + 988.0, + 791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1042.0, + 770.0, + 1176.0, + 770.0, + 1176.0, + 791.0, + 1042.0, + 791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1209.0, + 769.0, + 1395.0, + 769.0, + 1395.0, + 793.0, + 1209.0, + 793.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 399.0, + 784.0, + 451.0, + 784.0, + 451.0, + 804.0, + 399.0, + 804.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 617.0, + 784.0, + 670.0, + 784.0, + 670.0, + 805.0, + 617.0, + 805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 836.0, + 784.0, + 888.0, + 784.0, + 888.0, + 805.0, + 836.0, + 805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1054.0, + 784.0, + 1106.0, + 784.0, + 1106.0, + 805.0, + 1054.0, + 805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1271.0, + 784.0, + 1325.0, + 784.0, + 1325.0, + 805.0, + 1271.0, + 805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 377.0, + 804.0, + 473.0, + 804.0, + 473.0, + 828.0, + 377.0, + 828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 615.0, + 807.0, + 673.0, + 807.0, + 673.0, + 828.0, + 615.0, + 828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 836.0, + 807.0, + 887.0, + 807.0, + 887.0, + 828.0, + 836.0, + 828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1060.0, + 807.0, + 1100.0, + 807.0, + 1100.0, + 828.0, + 1060.0, + 828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1270.0, + 807.0, + 1326.0, + 807.0, + 1326.0, + 828.0, + 1270.0, + 828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 834.0, + 342.0, + 834.0, + 342.0, + 853.0, + 314.0, + 853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 358.0, + 820.0, + 491.0, + 820.0, + 491.0, + 844.0, + 358.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 531.0, + 828.0, + 562.0, + 828.0, + 562.0, + 850.0, + 531.0, + 850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 576.0, + 818.0, + 710.0, + 818.0, + 710.0, + 846.0, + 576.0, + 846.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 834.0, + 779.0, + 834.0, + 779.0, + 853.0, + 756.0, + 853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 795.0, + 818.0, + 928.0, + 818.0, + 928.0, + 846.0, + 795.0, + 846.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1013.0, + 820.0, + 1147.0, + 820.0, + 1147.0, + 844.0, + 1013.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1231.0, + 818.0, + 1365.0, + 818.0, + 1365.0, + 846.0, + 1231.0, + 846.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 854.0, + 352.0, + 854.0, + 352.0, + 955.0, + 291.0, + 955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 517.0, + 864.0, + 563.0, + 864.0, + 563.0, + 949.0, + 517.0, + 949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 740.0, + 859.0, + 784.0, + 859.0, + 784.0, + 945.0, + 740.0, + 945.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 957.0, + 847.0, + 1004.0, + 847.0, + 1004.0, + 946.0, + 957.0, + 946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1172.0, + 850.0, + 1225.0, + 850.0, + 1225.0, + 951.0, + 1172.0, + 951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 942.0, + 344.0, + 942.0, + 344.0, + 965.0, + 313.0, + 965.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 939.0, + 561.0, + 939.0, + 561.0, + 960.0, + 538.0, + 960.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 934.0, + 780.0, + 934.0, + 780.0, + 953.0, + 756.0, + 953.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 934.0, + 998.0, + 934.0, + 998.0, + 955.0, + 974.0, + 955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1192.0, + 947.0, + 1217.0, + 947.0, + 1217.0, + 967.0, + 1192.0, + 967.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 961.0, + 345.0, + 961.0, + 345.0, + 1007.0, + 313.0, + 1007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 381.0, + 960.0, + 490.0, + 960.0, + 490.0, + 1000.0, + 381.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 976.0, + 561.0, + 976.0, + 561.0, + 997.0, + 537.0, + 997.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 578.0, + 982.0, + 591.0, + 982.0, + 591.0, + 991.0, + 578.0, + 991.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 960.0, + 709.0, + 960.0, + 709.0, + 1000.0, + 600.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 958.0, + 779.0, + 958.0, + 779.0, + 1003.0, + 756.0, + 1003.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 817.0, + 960.0, + 927.0, + 960.0, + 927.0, + 1000.0, + 817.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 965.0, + 999.0, + 965.0, + 999.0, + 1013.0, + 974.0, + 1013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1036.0, + 960.0, + 1146.0, + 960.0, + 1146.0, + 1000.0, + 1036.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 971.0, + 1217.0, + 971.0, + 1217.0, + 991.0, + 1193.0, + 991.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1234.0, + 982.0, + 1245.0, + 982.0, + 1245.0, + 991.0, + 1234.0, + 991.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1253.0, + 960.0, + 1365.0, + 960.0, + 1365.0, + 1000.0, + 1253.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 335.0, + 1003.0, + 521.0, + 1003.0, + 521.0, + 1024.0, + 335.0, + 1024.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 551.0, + 1003.0, + 739.0, + 1003.0, + 739.0, + 1024.0, + 551.0, + 1024.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 769.0, + 1003.0, + 957.0, + 1003.0, + 957.0, + 1024.0, + 769.0, + 1024.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 991.0, + 1007.0, + 1001.0, + 1007.0, + 1001.0, + 1020.0, + 991.0, + 1020.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1003.0, + 1003.0, + 1174.0, + 1003.0, + 1174.0, + 1024.0, + 1003.0, + 1024.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1222.0, + 1003.0, + 1393.0, + 1003.0, + 1393.0, + 1024.0, + 1222.0, + 1024.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 399.0, + 1017.0, + 452.0, + 1017.0, + 452.0, + 1037.0, + 399.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 618.0, + 1017.0, + 670.0, + 1017.0, + 670.0, + 1037.0, + 618.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 835.0, + 1017.0, + 888.0, + 1017.0, + 888.0, + 1037.0, + 835.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1054.0, + 1017.0, + 1106.0, + 1017.0, + 1106.0, + 1037.0, + 1054.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1271.0, + 1017.0, + 1325.0, + 1017.0, + 1325.0, + 1037.0, + 1271.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 760.0, + 762.0, + 782.0, + 762.0, + 782.0, + 786.0, + 760.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 996.5, + 1218.0, + 996.5, + 1218.0, + 1019.5, + 1198.0, + 1019.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1048.0, + 943.0, + 1048.0, + 943.0, + 1084.0, + 294.0, + 1084.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1019.0, + 1048.0, + 1405.0, + 1048.0, + 1405.0, + 1084.0, + 1019.0, + 1084.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1079.0, + 1405.0, + 1079.0, + 1405.0, + 1116.0, + 294.0, + 1116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1110.0, + 592.0, + 1110.0, + 592.0, + 1144.0, + 296.0, + 1144.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1087.0, + 1542.0, + 1157.0, + 1542.0, + 1157.0, + 1610.0, + 1087.0, + 1610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1168.0, + 1540.0, + 1221.0, + 1540.0, + 1221.0, + 1575.0, + 1168.0, + 1575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 1542.0, + 1281.0, + 1542.0, + 1281.0, + 1573.0, + 1229.0, + 1573.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1288.0, + 1542.0, + 1340.0, + 1542.0, + 1340.0, + 1574.0, + 1288.0, + 1574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1374.0, + 1558.0, + 1406.0, + 1558.0, + 1406.0, + 1585.0, + 1374.0, + 1585.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1086.0, + 1598.0, + 1219.0, + 1598.0, + 1219.0, + 1672.0, + 1086.0, + 1672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1227.0, + 1600.0, + 1282.0, + 1600.0, + 1282.0, + 1635.0, + 1227.0, + 1635.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1287.0, + 1600.0, + 1342.0, + 1600.0, + 1342.0, + 1635.0, + 1287.0, + 1635.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1372.0, + 1608.0, + 1406.0, + 1608.0, + 1406.0, + 1635.0, + 1372.0, + 1635.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1066.0, + 1619.0, + 1093.0, + 1619.0, + 1093.0, + 1677.0, + 1066.0, + 1677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1087.0, + 1655.0, + 1278.0, + 1655.0, + 1278.0, + 1726.0, + 1087.0, + 1726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1285.0, + 1659.0, + 1342.0, + 1659.0, + 1342.0, + 1694.0, + 1285.0, + 1694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1373.0, + 1656.0, + 1406.0, + 1656.0, + 1406.0, + 1683.0, + 1373.0, + 1683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1374.0, + 1709.0, + 1397.0, + 1709.0, + 1397.0, + 1729.0, + 1374.0, + 1729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1087.0, + 1724.0, + 1109.0, + 1724.0, + 1109.0, + 1780.0, + 1087.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1110.0, + 1721.0, + 1163.0, + 1721.0, + 1163.0, + 1752.0, + 1110.0, + 1752.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1169.0, + 1718.0, + 1221.0, + 1718.0, + 1221.0, + 1754.0, + 1169.0, + 1754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1228.0, + 1718.0, + 1282.0, + 1718.0, + 1282.0, + 1753.0, + 1228.0, + 1753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1293.0, + 1720.0, + 1336.0, + 1720.0, + 1336.0, + 1753.0, + 1293.0, + 1753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1108.0, + 1760.0, + 1283.0, + 1760.0, + 1283.0, + 1795.0, + 1108.0, + 1795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1284.0, + 1757.0, + 1344.0, + 1757.0, + 1344.0, + 1796.0, + 1284.0, + 1796.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1378.0, + 1760.0, + 1396.0, + 1760.0, + 1396.0, + 1776.0, + 1378.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1196.0, + 1781.0, + 1253.0, + 1781.0, + 1253.0, + 1811.0, + 1196.0, + 1811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 193.0, + 1404.0, + 193.0, + 1404.0, + 229.0, + 294.0, + 229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 217.0, + 676.0, + 217.0, + 676.0, + 264.0, + 392.0, + 264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 733.0, + 217.0, + 745.0, + 217.0, + 745.0, + 264.0, + 733.0, + 264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1065.0, + 1809.0, + 1408.0, + 1809.0, + 1408.0, + 1843.0, + 1065.0, + 1843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1063.0, + 1837.0, + 1405.0, + 1837.0, + 1405.0, + 1875.0, + 1063.0, + 1875.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1066.0, + 1871.0, + 1408.0, + 1871.0, + 1408.0, + 1902.0, + 1066.0, + 1902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1065.0, + 1901.0, + 1404.0, + 1901.0, + 1404.0, + 1932.0, + 1065.0, + 1932.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1064.0, + 1931.0, + 1405.0, + 1931.0, + 1405.0, + 1962.0, + 1064.0, + 1962.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1063.0, + 1962.0, + 1405.0, + 1962.0, + 1405.0, + 1994.0, + 1063.0, + 1994.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1064.0, + 1985.0, + 1168.0, + 1985.0, + 1168.0, + 2030.0, + 1064.0, + 2030.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1263.0, + 1985.0, + 1277.0, + 1985.0, + 1277.0, + 2030.0, + 1263.0, + 2030.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1517.0, + 625.0, + 1517.0, + 625.0, + 1556.0, + 295.0, + 1556.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 836.0, + 2058.0, + 859.0, + 2058.0, + 859.0, + 2089.0, + 836.0, + 2089.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1176.0, + 1406.0, + 1176.0, + 1406.0, + 1220.0, + 292.0, + 1220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1210.0, + 1406.0, + 1210.0, + 1406.0, + 1247.0, + 294.0, + 1247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1244.0, + 1403.0, + 1244.0, + 1403.0, + 1276.0, + 296.0, + 1276.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1272.0, + 432.0, + 1272.0, + 432.0, + 1308.0, + 295.0, + 1308.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 529.0, + 1272.0, + 1406.0, + 1272.0, + 1406.0, + 1308.0, + 529.0, + 1308.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1303.0, + 1406.0, + 1303.0, + 1406.0, + 1338.0, + 294.0, + 1338.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1334.0, + 1070.0, + 1334.0, + 1070.0, + 1370.0, + 294.0, + 1370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1168.0, + 1334.0, + 1407.0, + 1334.0, + 1407.0, + 1370.0, + 1168.0, + 1370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1363.0, + 563.0, + 1363.0, + 563.0, + 1399.0, + 295.0, + 1399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 660.0, + 1363.0, + 1406.0, + 1363.0, + 1406.0, + 1399.0, + 660.0, + 1399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1394.0, + 969.0, + 1394.0, + 969.0, + 1429.0, + 294.0, + 1429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1066.0, + 1394.0, + 1405.0, + 1394.0, + 1405.0, + 1429.0, + 1066.0, + 1429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1423.0, + 1405.0, + 1423.0, + 1405.0, + 1460.0, + 292.0, + 1460.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1452.0, + 1036.0, + 1452.0, + 1036.0, + 1490.0, + 292.0, + 1490.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1568.0, + 1045.0, + 1568.0, + 1045.0, + 1599.0, + 296.0, + 1599.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1599.0, + 1046.0, + 1599.0, + 1046.0, + 1630.0, + 295.0, + 1630.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 391.0, + 1622.0, + 419.0, + 1622.0, + 419.0, + 1667.0, + 391.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 513.0, + 1622.0, + 846.0, + 1622.0, + 846.0, + 1667.0, + 513.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 890.0, + 1622.0, + 1053.0, + 1622.0, + 1053.0, + 1667.0, + 890.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1656.0, + 572.0, + 1656.0, + 572.0, + 1692.0, + 294.0, + 1692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 628.0, + 1656.0, + 936.0, + 1656.0, + 936.0, + 1692.0, + 628.0, + 1692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1719.0, + 1051.0, + 1719.0, + 1051.0, + 1755.0, + 294.0, + 1755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1748.0, + 580.0, + 1748.0, + 580.0, + 1786.0, + 293.0, + 1786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 745.0, + 1748.0, + 1049.0, + 1748.0, + 1049.0, + 1786.0, + 745.0, + 1786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1781.0, + 632.0, + 1781.0, + 632.0, + 1815.0, + 295.0, + 1815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 707.0, + 1781.0, + 1048.0, + 1781.0, + 1048.0, + 1815.0, + 707.0, + 1815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1810.0, + 1049.0, + 1810.0, + 1049.0, + 1848.0, + 293.0, + 1848.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1839.0, + 369.0, + 1839.0, + 369.0, + 1878.0, + 293.0, + 1878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 478.0, + 1839.0, + 1048.0, + 1839.0, + 1048.0, + 1878.0, + 478.0, + 1878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1870.0, + 402.0, + 1870.0, + 402.0, + 1906.0, + 294.0, + 1906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 447.0, + 1870.0, + 1047.0, + 1870.0, + 1047.0, + 1906.0, + 447.0, + 1906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1901.0, + 1050.0, + 1901.0, + 1050.0, + 1938.0, + 294.0, + 1938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1929.0, + 947.0, + 1929.0, + 947.0, + 1972.0, + 292.0, + 1972.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1043.0, + 1929.0, + 1048.0, + 1929.0, + 1048.0, + 1972.0, + 1043.0, + 1972.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1963.0, + 1047.0, + 1963.0, + 1047.0, + 1997.0, + 296.0, + 1997.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1990.0, + 792.0, + 1990.0, + 792.0, + 2031.0, + 294.0, + 2031.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 7, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 575, + 273, + 1126, + 273, + 1126, + 577, + 575, + 577 + ], + "score": 0.979, + "html": "
RankModel Elo Rating
1LLaMA-Adapter v2 [17] 1023.0
2 LLaVA [35]1019.9
3VL-Vicuna (ours) 1012.1
4MiniGPT-4 [62] 1011.9
5InstructBLIP[13] 999.5
6mPLUG-Owl[59] 996.3
7 Otter [26]981.5
8 BLIP-2 [30]955.8
" + }, + { + "category_id": 1, + "poly": [ + 297, + 1236, + 1405, + 1236, + 1405, + 1541, + 297, + 1541 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 298, + 1853, + 1404, + 1853, + 1404, + 2007, + 298, + 2007 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 297, + 1103, + 1404, + 1103, + 1404, + 1226, + 297, + 1226 + ], + "score": 0.972 + }, + { + "category_id": 1, + "poly": [ + 298, + 866, + 1405, + 866, + 1405, + 960, + 298, + 960 + ], + "score": 0.972 + }, + { + "category_id": 1, + "poly": [ + 299, + 1631, + 1401, + 1631, + 1401, + 1724, + 299, + 1724 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 296, + 1742, + 1401, + 1742, + 1401, + 1836, + 296, + 1836 + ], + "score": 0.959 + }, + { + "category_id": 1, + "poly": [ + 296, + 678, + 1402, + 678, + 1402, + 739, + 296, + 739 + ], + "score": 0.958 + }, + { + "category_id": 1, + "poly": [ + 299, + 977, + 1402, + 977, + 1402, + 1042, + 299, + 1042 + ], + "score": 0.95 + }, + { + "category_id": 1, + "poly": [ + 297, + 796, + 1400, + 796, + 1400, + 857, + 297, + 857 + ], + "score": 0.949 + }, + { + "category_id": 0, + "poly": [ + 296, + 631, + 1012, + 631, + 1012, + 669, + 296, + 669 + ], + "score": 0.947 + }, + { + "category_id": 6, + "poly": [ + 298, + 192, + 1400, + 192, + 1400, + 258, + 298, + 258 + ], + "score": 0.938 + }, + { + "category_id": 0, + "poly": [ + 298, + 1568, + 965, + 1568, + 965, + 1608, + 298, + 1608 + ], + "score": 0.936 + }, + { + "category_id": 0, + "poly": [ + 298, + 754, + 627, + 754, + 627, + 787, + 298, + 787 + ], + "score": 0.923 + }, + { + "category_id": 0, + "poly": [ + 297, + 1063, + 1231, + 1063, + 1231, + 1095, + 297, + 1095 + ], + "score": 0.921 + }, + { + "category_id": 2, + "poly": [ + 840, + 2061, + 859, + 2061, + 859, + 2084, + 840, + 2084 + ], + "score": 0.797 + }, + { + "category_id": 13, + "poly": [ + 1085, + 1165, + 1190, + 1165, + 1190, + 1196, + 1085, + 1196 + ], + "score": 0.87, + "latex": "\\mathrm { O P T } _ { 3 5 0 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 1110, + 1420, + 1342, + 1420, + 1342, + 1451, + 1110, + 1451 + ], + "score": 0.85, + "latex": "\\mathrm { F l a n T } 5 _ { \\mathrm { X L } } { } \\mathrm { O P T } _ { 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 816, + 1299, + 911, + 1299, + 911, + 1329, + 816, + 1329 + ], + "score": 0.84, + "latex": "\\mathrm { O P T } _ { 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 503, + 1010, + 597, + 1010, + 597, + 1042, + 503, + 1042 + ], + "score": 0.84, + "latex": "\\mathrm { O P T } _ { 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 297, + 1268, + 390, + 1268, + 390, + 1300, + 297, + 1300 + ], + "score": 0.82, + "latex": "\\mathrm { O P T } _ { 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 762, + 1359, + 864, + 1359, + 864, + 1390, + 762, + 1390 + ], + "score": 0.82, + "latex": "\\mathrm { O P T } _ { 3 5 0 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 1021, + 1804, + 1124, + 1804, + 1124, + 1836, + 1021, + 1836 + ], + "score": 0.8, + "latex": "\\mathrm { O P T } _ { 6 . 7 B }" + }, + { + "category_id": 13, + "poly": [ + 640, + 1916, + 733, + 1916, + 733, + 1947, + 640, + 1947 + ], + "score": 0.8, + "latex": "\\mathrm { O P T } _ { 6 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 1156, + 1389, + 1402, + 1389, + 1402, + 1419, + 1156, + 1419 + ], + "score": 0.79, + "latex": "\\mathrm { O P T } _ { 3 5 0 \\mathrm { M } } { } \\mathrm { F l a n T } 5 _ { \\mathrm { b a s e } }" + }, + { + "category_id": 13, + "poly": [ + 1245, + 978, + 1350, + 978, + 1350, + 1010, + 1245, + 1010 + ], + "score": 0.74, + "latex": "\\mathrm { O P T } _ { 3 5 0 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 640, + 1774, + 735, + 1774, + 735, + 1805, + 640, + 1805 + ], + "score": 0.71, + "latex": "\\mathrm { O P T } _ { 6 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 1164, + 1976, + 1197, + 1976, + 1197, + 2007, + 1164, + 2007 + ], + "score": 0.7, + "latex": "\\mathrm { \\Omega ^ { \\ S E } }" + }, + { + "category_id": 13, + "poly": [ + 672, + 1299, + 784, + 1299, + 784, + 1329, + 672, + 1329 + ], + "score": 0.61, + "latex": "\\mathrm { F l a n T 5 } _ { \\mathrm { X L } }" + }, + { + "category_id": 13, + "poly": [ + 504, + 1268, + 547, + 1268, + 547, + 1299, + 504, + 1299 + ], + "score": 0.48, + "latex": "{ \\mathrm { 5 } } _ { \\mathrm { X L } }" + }, + { + "category_id": 13, + "poly": [ + 1297, + 1166, + 1348, + 1166, + 1348, + 1196, + 1297, + 1196 + ], + "score": 0.34, + "latex": "5 _ { \\mathrm { b a s e } }" + }, + { + "category_id": 13, + "poly": [ + 594, + 1359, + 712, + 1359, + 712, + 1390, + 594, + 1390 + ], + "score": 0.33, + "latex": "{ \\mathrm { F l a n T } } 5 _ { \\mathrm { b a s e } }" + }, + { + "category_id": 13, + "poly": [ + 362, + 1010, + 415, + 1010, + 415, + 1041, + 362, + 1041 + ], + "score": 0.3, + "latex": "\\mathrm { \\Delta } 5 _ { \\mathrm { b a s e } }" + }, + { + "category_id": 13, + "poly": [ + 661, + 1360, + 712, + 1360, + 712, + 1389, + 661, + 1389 + ], + "score": 0.29, + "latex": "5 _ { \\mathrm { b a s e } }" + }, + { + "category_id": 13, + "poly": [ + 717, + 1011, + 758, + 1011, + 758, + 1041, + 717, + 1041 + ], + "score": 0.27, + "latex": "5 _ { \\mathrm { X L } }" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 625.0, + 1015.0, + 625.0, + 1015.0, + 681.0, + 290.0, + 681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 195.0, + 1402.0, + 195.0, + 1402.0, + 228.0, + 296.0, + 228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 222.0, + 1153.0, + 222.0, + 1153.0, + 261.0, + 293.0, + 261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1563.0, + 969.0, + 1563.0, + 969.0, + 1618.0, + 290.0, + 1618.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 746.0, + 630.0, + 746.0, + 630.0, + 797.0, + 292.0, + 797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1058.0, + 1235.0, + 1058.0, + 1235.0, + 1102.0, + 291.0, + 1102.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2060.0, + 861.0, + 2060.0, + 861.0, + 2090.0, + 839.0, + 2090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1238.0, + 1403.0, + 1238.0, + 1403.0, + 1270.0, + 297.0, + 1270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 391.0, + 1267.0, + 503.0, + 1267.0, + 503.0, + 1302.0, + 391.0, + 1302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 548.0, + 1267.0, + 1405.0, + 1267.0, + 1405.0, + 1302.0, + 548.0, + 1302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1297.0, + 671.0, + 1297.0, + 671.0, + 1333.0, + 295.0, + 1333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 785.0, + 1297.0, + 815.0, + 1297.0, + 815.0, + 1333.0, + 785.0, + 1333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 912.0, + 1297.0, + 1403.0, + 1297.0, + 1403.0, + 1333.0, + 912.0, + 1333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1328.0, + 1407.0, + 1328.0, + 1407.0, + 1362.0, + 295.0, + 1362.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1358.0, + 593.0, + 1358.0, + 593.0, + 1394.0, + 295.0, + 1394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 713.0, + 1358.0, + 761.0, + 1358.0, + 761.0, + 1394.0, + 713.0, + 1394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 865.0, + 1358.0, + 1406.0, + 1358.0, + 1406.0, + 1394.0, + 865.0, + 1394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1382.0, + 1155.0, + 1382.0, + 1155.0, + 1428.0, + 292.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1403.0, + 1382.0, + 1406.0, + 1382.0, + 1406.0, + 1428.0, + 1403.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1415.0, + 1109.0, + 1415.0, + 1109.0, + 1456.0, + 292.0, + 1456.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1343.0, + 1415.0, + 1406.0, + 1415.0, + 1406.0, + 1456.0, + 1343.0, + 1456.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1451.0, + 1402.0, + 1451.0, + 1402.0, + 1482.0, + 296.0, + 1482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1480.0, + 1405.0, + 1480.0, + 1405.0, + 1513.0, + 292.0, + 1513.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1510.0, + 651.0, + 1510.0, + 651.0, + 1542.0, + 296.0, + 1542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1855.0, + 1404.0, + 1855.0, + 1404.0, + 1888.0, + 297.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1884.0, + 1406.0, + 1884.0, + 1406.0, + 1921.0, + 294.0, + 1921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1913.0, + 639.0, + 1913.0, + 639.0, + 1950.0, + 294.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 734.0, + 1913.0, + 1406.0, + 1913.0, + 1406.0, + 1950.0, + 734.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1945.0, + 1408.0, + 1945.0, + 1408.0, + 1982.0, + 294.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1974.0, + 1163.0, + 1974.0, + 1163.0, + 2012.0, + 293.0, + 2012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 1974.0, + 1405.0, + 1974.0, + 1405.0, + 2012.0, + 1198.0, + 2012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 299.0, + 1103.0, + 1404.0, + 1103.0, + 1404.0, + 1135.0, + 299.0, + 1135.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1133.0, + 1404.0, + 1133.0, + 1404.0, + 1168.0, + 294.0, + 1168.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1163.0, + 1084.0, + 1163.0, + 1084.0, + 1198.0, + 294.0, + 1198.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1191.0, + 1163.0, + 1296.0, + 1163.0, + 1296.0, + 1198.0, + 1191.0, + 1198.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1349.0, + 1163.0, + 1404.0, + 1163.0, + 1404.0, + 1198.0, + 1349.0, + 1198.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1194.0, + 1254.0, + 1194.0, + 1254.0, + 1228.0, + 292.0, + 1228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 867.0, + 1403.0, + 867.0, + 1403.0, + 901.0, + 296.0, + 901.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 898.0, + 1403.0, + 898.0, + 1403.0, + 931.0, + 293.0, + 931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 930.0, + 447.0, + 930.0, + 447.0, + 965.0, + 293.0, + 965.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1629.0, + 1406.0, + 1629.0, + 1406.0, + 1668.0, + 293.0, + 1668.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1661.0, + 1405.0, + 1661.0, + 1405.0, + 1697.0, + 294.0, + 1697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1694.0, + 782.0, + 1694.0, + 782.0, + 1725.0, + 297.0, + 1725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1742.0, + 1406.0, + 1742.0, + 1406.0, + 1779.0, + 295.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1771.0, + 639.0, + 1771.0, + 639.0, + 1810.0, + 292.0, + 1810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 736.0, + 1771.0, + 1407.0, + 1771.0, + 1407.0, + 1810.0, + 736.0, + 1810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1804.0, + 1020.0, + 1804.0, + 1020.0, + 1838.0, + 295.0, + 1838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1125.0, + 1804.0, + 1298.0, + 1804.0, + 1298.0, + 1838.0, + 1125.0, + 1838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 671.0, + 1406.0, + 671.0, + 1406.0, + 715.0, + 291.0, + 715.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 709.0, + 674.0, + 709.0, + 674.0, + 740.0, + 295.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 973.0, + 1244.0, + 973.0, + 1244.0, + 1015.0, + 294.0, + 1015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1351.0, + 973.0, + 1404.0, + 973.0, + 1404.0, + 1015.0, + 1351.0, + 1015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1006.0, + 361.0, + 1006.0, + 361.0, + 1047.0, + 294.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 416.0, + 1006.0, + 502.0, + 1006.0, + 502.0, + 1047.0, + 416.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 598.0, + 1006.0, + 716.0, + 1006.0, + 716.0, + 1047.0, + 598.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 759.0, + 1006.0, + 769.0, + 1006.0, + 769.0, + 1047.0, + 759.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 795.0, + 1402.0, + 795.0, + 1402.0, + 830.0, + 295.0, + 830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 825.0, + 1236.0, + 825.0, + 1236.0, + 860.0, + 295.0, + 860.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 8, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1422, + 1404, + 1422, + 1404, + 1635, + 298, + 1635 + ], + "score": 0.978 + }, + { + "category_id": 3, + "poly": [ + 302, + 196, + 1393, + 196, + 1393, + 1132, + 302, + 1132 + ], + "score": 0.971 + }, + { + "category_id": 1, + "poly": [ + 298, + 1250, + 1404, + 1250, + 1404, + 1371, + 298, + 1371 + ], + "score": 0.968 + }, + { + "category_id": 1, + "poly": [ + 299, + 1745, + 1403, + 1745, + 1403, + 1808, + 299, + 1808 + ], + "score": 0.942 + }, + { + "category_id": 4, + "poly": [ + 490, + 1162, + 1206, + 1162, + 1206, + 1195, + 490, + 1195 + ], + "score": 0.917 + }, + { + "category_id": 0, + "poly": [ + 301, + 1676, + 561, + 1676, + 561, + 1714, + 301, + 1714 + ], + "score": 0.911 + }, + { + "category_id": 0, + "poly": [ + 299, + 1386, + 510, + 1386, + 510, + 1419, + 299, + 1419 + ], + "score": 0.909 + }, + { + "category_id": 0, + "poly": [ + 300, + 1849, + 455, + 1849, + 455, + 1883, + 300, + 1883 + ], + "score": 0.899 + }, + { + "category_id": 2, + "poly": [ + 335, + 1979, + 712, + 1979, + 712, + 2007, + 335, + 2007 + ], + "score": 0.893 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.848 + }, + { + "category_id": 1, + "poly": [ + 319, + 1899, + 1402, + 1899, + 1402, + 1956, + 319, + 1956 + ], + "score": 0.582 + }, + { + "category_id": 2, + "poly": [ + 319, + 1899, + 1402, + 1899, + 1402, + 1956, + 319, + 1956 + ], + "score": 0.328 + }, + { + "category_id": 13, + "poly": [ + 631, + 1746, + 738, + 1746, + 738, + 1775, + 631, + 1775 + ], + "score": 0.44, + "latex": "{ \\mathrm { N E x T } } + +" + }, + { + "category_id": 13, + "poly": [ + 702, + 1748, + 739, + 1748, + 739, + 1774, + 702, + 1774 + ], + "score": 0.26, + "latex": "^ { + + }" + }, + { + "category_id": 15, + "poly": [ + 508.0, + 198.0, + 623.0, + 198.0, + 623.0, + 227.0, + 508.0, + 227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1074.0, + 194.0, + 1191.0, + 194.0, + 1191.0, + 231.0, + 1074.0, + 231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 490.0, + 380.0, + 804.0, + 380.0, + 804.0, + 419.0, + 490.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1058.0, + 380.0, + 1375.0, + 380.0, + 1375.0, + 419.0, + 1058.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 317.0, + 439.0, + 352.0, + 439.0, + 352.0, + 469.0, + 317.0, + 469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 436.0, + 637.0, + 436.0, + 637.0, + 471.0, + 363.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 890.0, + 446.0, + 910.0, + 446.0, + 910.0, + 461.0, + 890.0, + 461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 936.0, + 431.0, + 1326.0, + 431.0, + 1326.0, + 480.0, + 936.0, + 480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 571.0, + 491.0, + 747.0, + 491.0, + 747.0, + 523.0, + 571.0, + 523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 767.0, + 482.0, + 812.0, + 482.0, + 812.0, + 531.0, + 767.0, + 531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1139.0, + 488.0, + 1315.0, + 488.0, + 1315.0, + 523.0, + 1139.0, + 523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1336.0, + 482.0, + 1381.0, + 482.0, + 1381.0, + 530.0, + 1336.0, + 530.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 312.0, + 530.0, + 357.0, + 530.0, + 357.0, + 569.0, + 312.0, + 569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 371.0, + 539.0, + 725.0, + 539.0, + 725.0, + 573.0, + 371.0, + 573.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 885.0, + 533.0, + 918.0, + 533.0, + 918.0, + 562.0, + 885.0, + 562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 938.0, + 539.0, + 1321.0, + 539.0, + 1321.0, + 573.0, + 938.0, + 573.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 562.0, + 731.0, + 562.0, + 731.0, + 593.0, + 372.0, + 593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 939.0, + 564.0, + 1323.0, + 564.0, + 1323.0, + 589.0, + 939.0, + 589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 583.0, + 728.0, + 583.0, + 728.0, + 610.0, + 374.0, + 610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 939.0, + 579.0, + 1318.0, + 579.0, + 1318.0, + 613.0, + 939.0, + 613.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 371.0, + 603.0, + 737.0, + 603.0, + 737.0, + 629.0, + 371.0, + 629.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 939.0, + 602.0, + 1305.0, + 602.0, + 1305.0, + 629.0, + 939.0, + 629.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 624.0, + 761.0, + 624.0, + 761.0, + 652.0, + 374.0, + 652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 939.0, + 622.0, + 1307.0, + 622.0, + 1307.0, + 653.0, + 939.0, + 653.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 644.0, + 732.0, + 644.0, + 732.0, + 671.0, + 374.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 938.0, + 640.0, + 1326.0, + 640.0, + 1326.0, + 673.0, + 938.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 664.0, + 690.0, + 664.0, + 690.0, + 691.0, + 372.0, + 691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 939.0, + 664.0, + 1240.0, + 664.0, + 1240.0, + 691.0, + 939.0, + 691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 447.0, + 889.0, + 735.0, + 889.0, + 735.0, + 939.0, + 447.0, + 939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 771.0, + 891.0, + 809.0, + 891.0, + 809.0, + 937.0, + 771.0, + 937.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1015.0, + 889.0, + 1302.0, + 889.0, + 1302.0, + 939.0, + 1015.0, + 939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1335.0, + 889.0, + 1380.0, + 889.0, + 1380.0, + 939.0, + 1335.0, + 939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 312.0, + 954.0, + 356.0, + 954.0, + 356.0, + 993.0, + 312.0, + 993.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 375.0, + 963.0, + 738.0, + 963.0, + 738.0, + 987.0, + 375.0, + 987.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 881.0, + 955.0, + 921.0, + 955.0, + 921.0, + 991.0, + 881.0, + 991.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 959.0, + 1328.0, + 959.0, + 1328.0, + 992.0, + 941.0, + 992.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 983.0, + 752.0, + 983.0, + 752.0, + 1006.0, + 372.0, + 1006.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 939.0, + 979.0, + 1321.0, + 979.0, + 1321.0, + 1011.0, + 939.0, + 1011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 1003.0, + 743.0, + 1003.0, + 743.0, + 1026.0, + 374.0, + 1026.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1003.0, + 1319.0, + 1003.0, + 1319.0, + 1030.0, + 941.0, + 1030.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 371.0, + 1019.0, + 737.0, + 1019.0, + 737.0, + 1048.0, + 371.0, + 1048.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1023.0, + 1311.0, + 1023.0, + 1311.0, + 1050.0, + 941.0, + 1050.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 370.0, + 1038.0, + 736.0, + 1038.0, + 736.0, + 1070.0, + 370.0, + 1070.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 939.0, + 1041.0, + 1290.0, + 1041.0, + 1290.0, + 1070.0, + 939.0, + 1070.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 370.0, + 1060.0, + 682.0, + 1060.0, + 682.0, + 1087.0, + 370.0, + 1087.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1062.0, + 1283.0, + 1062.0, + 1283.0, + 1090.0, + 941.0, + 1090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1082.0, + 1320.0, + 1082.0, + 1320.0, + 1110.0, + 941.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 491.0, + 1160.0, + 1207.0, + 1160.0, + 1207.0, + 1198.0, + 491.0, + 1198.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1672.0, + 565.0, + 1672.0, + 565.0, + 1721.0, + 295.0, + 1721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1375.0, + 516.0, + 1375.0, + 516.0, + 1430.0, + 290.0, + 1430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1845.0, + 460.0, + 1845.0, + 460.0, + 1889.0, + 295.0, + 1889.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 1973.0, + 716.0, + 1973.0, + 716.0, + 2013.0, + 331.0, + 2013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 317.0, + 1899.0, + 1403.0, + 1899.0, + 1403.0, + 1930.0, + 317.0, + 1930.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1924.0, + 1404.0, + 1924.0, + 1404.0, + 1961.0, + 348.0, + 1961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1422.0, + 1405.0, + 1422.0, + 1405.0, + 1456.0, + 293.0, + 1456.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1451.0, + 1405.0, + 1451.0, + 1405.0, + 1486.0, + 294.0, + 1486.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1478.0, + 1405.0, + 1478.0, + 1405.0, + 1520.0, + 292.0, + 1520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1511.0, + 1404.0, + 1511.0, + 1404.0, + 1548.0, + 293.0, + 1548.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1543.0, + 1404.0, + 1543.0, + 1404.0, + 1577.0, + 294.0, + 1577.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1574.0, + 1404.0, + 1574.0, + 1404.0, + 1604.0, + 294.0, + 1604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1604.0, + 1215.0, + 1604.0, + 1215.0, + 1639.0, + 296.0, + 1639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1248.0, + 1406.0, + 1248.0, + 1406.0, + 1287.0, + 293.0, + 1287.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1279.0, + 1405.0, + 1279.0, + 1405.0, + 1316.0, + 292.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1309.0, + 1404.0, + 1309.0, + 1404.0, + 1348.0, + 293.0, + 1348.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1342.0, + 390.0, + 1342.0, + 390.0, + 1374.0, + 293.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1745.0, + 630.0, + 1745.0, + 630.0, + 1780.0, + 295.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 740.0, + 1745.0, + 1404.0, + 1745.0, + 1404.0, + 1780.0, + 740.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1775.0, + 1358.0, + 1775.0, + 1358.0, + 1811.0, + 295.0, + 1811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 317.0, + 1899.0, + 1403.0, + 1899.0, + 1403.0, + 1930.0, + 317.0, + 1930.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1924.0, + 1404.0, + 1924.0, + 1404.0, + 1961.0, + 348.0, + 1961.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 9, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 863, + 2061, + 863, + 2086, + 836, + 2086 + ], + "score": 0.75 + }, + { + "category_id": 1, + "poly": [ + 293, + 138, + 1409, + 138, + 1409, + 2013, + 293, + 2013 + ], + "score": 0.673 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 863, + 2061, + 863, + 2086, + 836, + 2086 + ], + "score": 0.224 + }, + { + "category_id": 13, + "poly": [ + 657, + 1144, + 720, + 1144, + 720, + 1171, + 657, + 1171 + ], + "score": 0.83, + "latex": "9 0 \\% *" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2098.0, + 832.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2098.0, + 832.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 203.0, + 1172.0, + 203.0, + 1172.0, + 242.0, + 349.0, + 242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 255.0, + 1408.0, + 255.0, + 1408.0, + 288.0, + 306.0, + 288.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 277.0, + 1410.0, + 277.0, + 1410.0, + 318.0, + 344.0, + 318.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 310.0, + 1320.0, + 310.0, + 1320.0, + 343.0, + 349.0, + 343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 353.0, + 1404.0, + 353.0, + 1404.0, + 392.0, + 304.0, + 392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 386.0, + 1406.0, + 386.0, + 1406.0, + 419.0, + 349.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 409.0, + 1273.0, + 409.0, + 1273.0, + 450.0, + 344.0, + 450.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 462.0, + 1406.0, + 462.0, + 1406.0, + 495.0, + 308.0, + 495.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 491.0, + 1406.0, + 491.0, + 1406.0, + 524.0, + 349.0, + 524.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 516.0, + 779.0, + 516.0, + 779.0, + 553.0, + 344.0, + 553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 567.0, + 1406.0, + 567.0, + 1406.0, + 600.0, + 306.0, + 600.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 592.0, + 1408.0, + 592.0, + 1408.0, + 631.0, + 349.0, + 631.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 623.0, + 1225.0, + 623.0, + 1225.0, + 656.0, + 349.0, + 656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 666.0, + 1406.0, + 666.0, + 1406.0, + 709.0, + 302.0, + 709.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 697.0, + 1408.0, + 697.0, + 1408.0, + 736.0, + 346.0, + 736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 724.0, + 501.0, + 724.0, + 501.0, + 757.0, + 351.0, + 757.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 773.0, + 1408.0, + 773.0, + 1408.0, + 810.0, + 302.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 802.0, + 1410.0, + 802.0, + 1410.0, + 841.0, + 346.0, + 841.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 342.0, + 829.0, + 1098.0, + 829.0, + 1098.0, + 866.0, + 342.0, + 866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 880.0, + 1406.0, + 880.0, + 1406.0, + 913.0, + 306.0, + 913.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 903.0, + 1406.0, + 903.0, + 1406.0, + 944.0, + 344.0, + 944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 936.0, + 1005.0, + 936.0, + 1005.0, + 969.0, + 351.0, + 969.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 981.0, + 1408.0, + 981.0, + 1408.0, + 1020.0, + 304.0, + 1020.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1012.0, + 1406.0, + 1012.0, + 1406.0, + 1045.0, + 349.0, + 1045.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1041.0, + 1402.0, + 1041.0, + 1402.0, + 1073.0, + 349.0, + 1073.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1084.0, + 1406.0, + 1084.0, + 1406.0, + 1125.0, + 289.0, + 1125.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1117.0, + 1406.0, + 1117.0, + 1406.0, + 1150.0, + 351.0, + 1150.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1139.0, + 656.0, + 1139.0, + 656.0, + 1183.0, + 346.0, + 1183.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 1139.0, + 1402.0, + 1139.0, + 1402.0, + 1183.0, + 721.0, + 1183.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1193.0, + 1406.0, + 1193.0, + 1406.0, + 1226.0, + 294.0, + 1226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1215.0, + 1406.0, + 1215.0, + 1406.0, + 1259.0, + 347.0, + 1259.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1248.0, + 1020.0, + 1248.0, + 1020.0, + 1281.0, + 351.0, + 1281.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1292.0, + 1404.0, + 1292.0, + 1404.0, + 1331.0, + 294.0, + 1331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1320.0, + 1406.0, + 1320.0, + 1406.0, + 1361.0, + 346.0, + 1361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 1349.0, + 701.0, + 1349.0, + 701.0, + 1386.0, + 344.0, + 1386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1396.0, + 1406.0, + 1396.0, + 1406.0, + 1438.0, + 292.0, + 1438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1427.0, + 1408.0, + 1427.0, + 1408.0, + 1466.0, + 346.0, + 1466.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 1454.0, + 608.0, + 1454.0, + 608.0, + 1491.0, + 344.0, + 1491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1503.0, + 1408.0, + 1503.0, + 1408.0, + 1543.0, + 292.0, + 1543.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1530.0, + 1408.0, + 1530.0, + 1408.0, + 1569.0, + 346.0, + 1569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 1559.0, + 699.0, + 1559.0, + 699.0, + 1594.0, + 344.0, + 1594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1610.0, + 1404.0, + 1610.0, + 1404.0, + 1643.0, + 296.0, + 1643.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1637.0, + 1406.0, + 1637.0, + 1406.0, + 1670.0, + 351.0, + 1670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1662.0, + 1353.0, + 1662.0, + 1353.0, + 1699.0, + 346.0, + 1699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1711.0, + 1406.0, + 1711.0, + 1406.0, + 1750.0, + 292.0, + 1750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1736.0, + 1406.0, + 1736.0, + 1406.0, + 1779.0, + 347.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1769.0, + 836.0, + 1769.0, + 836.0, + 1802.0, + 351.0, + 1802.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1818.0, + 1404.0, + 1818.0, + 1404.0, + 1851.0, + 296.0, + 1851.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1843.0, + 1406.0, + 1843.0, + 1406.0, + 1882.0, + 346.0, + 1882.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1874.0, + 836.0, + 1874.0, + 836.0, + 1907.0, + 351.0, + 1907.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1919.0, + 1408.0, + 1919.0, + 1408.0, + 1958.0, + 294.0, + 1958.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1950.0, + 1402.0, + 1950.0, + 1402.0, + 1983.0, + 351.0, + 1983.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1975.0, + 999.0, + 1975.0, + 999.0, + 2012.0, + 346.0, + 2012.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 10, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.738 + }, + { + "category_id": 1, + "poly": [ + 287, + 195, + 1410, + 195, + 1410, + 2033, + 287, + 2033 + ], + "score": 0.617 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.23 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 204.0, + 1406.0, + 204.0, + 1406.0, + 240.0, + 291.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 232.0, + 1408.0, + 232.0, + 1408.0, + 270.0, + 348.0, + 270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 260.0, + 945.0, + 260.0, + 945.0, + 294.0, + 346.0, + 294.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 313.0, + 1404.0, + 313.0, + 1404.0, + 345.0, + 295.0, + 345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 341.0, + 1406.0, + 341.0, + 1406.0, + 373.0, + 350.0, + 373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 365.0, + 808.0, + 365.0, + 808.0, + 401.0, + 344.0, + 401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 416.0, + 1410.0, + 416.0, + 1410.0, + 454.0, + 291.0, + 454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 442.0, + 1408.0, + 442.0, + 1408.0, + 482.0, + 348.0, + 482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 474.0, + 561.0, + 474.0, + 561.0, + 506.0, + 350.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 523.0, + 1408.0, + 523.0, + 1408.0, + 559.0, + 293.0, + 559.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 547.0, + 1410.0, + 547.0, + 1410.0, + 589.0, + 344.0, + 589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 577.0, + 1291.0, + 577.0, + 1291.0, + 615.0, + 346.0, + 615.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 625.0, + 1408.0, + 625.0, + 1408.0, + 666.0, + 291.0, + 666.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 656.0, + 1333.0, + 656.0, + 1333.0, + 692.0, + 346.0, + 692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 702.0, + 1412.0, + 702.0, + 1412.0, + 749.0, + 289.0, + 749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 732.0, + 764.0, + 732.0, + 764.0, + 771.0, + 348.0, + 771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 781.0, + 1408.0, + 781.0, + 1408.0, + 825.0, + 293.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 815.0, + 1050.0, + 815.0, + 1050.0, + 849.0, + 346.0, + 849.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 864.0, + 1406.0, + 864.0, + 1406.0, + 900.0, + 291.0, + 900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 892.0, + 1178.0, + 892.0, + 1178.0, + 930.0, + 348.0, + 930.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 942.0, + 1408.0, + 942.0, + 1408.0, + 981.0, + 295.0, + 981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 971.0, + 1408.0, + 971.0, + 1408.0, + 1009.0, + 350.0, + 1009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1001.0, + 891.0, + 1001.0, + 891.0, + 1033.0, + 350.0, + 1033.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1045.0, + 1408.0, + 1045.0, + 1408.0, + 1086.0, + 291.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1076.0, + 1406.0, + 1076.0, + 1406.0, + 1114.0, + 348.0, + 1114.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1104.0, + 997.0, + 1104.0, + 997.0, + 1140.0, + 346.0, + 1140.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1154.0, + 1408.0, + 1154.0, + 1408.0, + 1193.0, + 293.0, + 1193.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1185.0, + 1406.0, + 1185.0, + 1406.0, + 1217.0, + 350.0, + 1217.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1211.0, + 812.0, + 1211.0, + 812.0, + 1249.0, + 348.0, + 1249.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1259.0, + 1410.0, + 1259.0, + 1410.0, + 1300.0, + 291.0, + 1300.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1292.0, + 1406.0, + 1292.0, + 1406.0, + 1324.0, + 348.0, + 1324.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1338.0, + 1408.0, + 1338.0, + 1408.0, + 1376.0, + 293.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1366.0, + 1408.0, + 1366.0, + 1408.0, + 1405.0, + 348.0, + 1405.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1397.0, + 742.0, + 1397.0, + 742.0, + 1429.0, + 350.0, + 1429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1445.0, + 1408.0, + 1445.0, + 1408.0, + 1483.0, + 293.0, + 1483.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1471.0, + 1406.0, + 1471.0, + 1406.0, + 1510.0, + 348.0, + 1510.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1502.0, + 1184.0, + 1502.0, + 1184.0, + 1540.0, + 348.0, + 1540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1552.0, + 1408.0, + 1552.0, + 1408.0, + 1590.0, + 293.0, + 1590.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1580.0, + 1120.0, + 1580.0, + 1120.0, + 1619.0, + 348.0, + 1619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1633.0, + 1406.0, + 1633.0, + 1406.0, + 1665.0, + 295.0, + 1665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1661.0, + 1408.0, + 1661.0, + 1408.0, + 1693.0, + 350.0, + 1693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1683.0, + 1408.0, + 1683.0, + 1408.0, + 1724.0, + 346.0, + 1724.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1716.0, + 613.0, + 1716.0, + 613.0, + 1748.0, + 352.0, + 1748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1766.0, + 1289.0, + 1766.0, + 1289.0, + 1798.0, + 295.0, + 1798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1812.0, + 1408.0, + 1812.0, + 1408.0, + 1853.0, + 291.0, + 1853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1843.0, + 1408.0, + 1843.0, + 1408.0, + 1881.0, + 348.0, + 1881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1871.0, + 420.0, + 1871.0, + 420.0, + 1905.0, + 348.0, + 1905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1924.0, + 1404.0, + 1924.0, + 1404.0, + 1956.0, + 295.0, + 1956.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1948.0, + 1408.0, + 1948.0, + 1408.0, + 1986.0, + 348.0, + 1986.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1976.0, + 1017.0, + 1976.0, + 1017.0, + 2014.0, + 348.0, + 2014.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 11, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.806 + }, + { + "category_id": 1, + "poly": [ + 291, + 130, + 1409, + 130, + 1409, + 2021, + 291, + 2021 + ], + "score": 0.558 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 202.0, + 1408.0, + 202.0, + 1408.0, + 242.0, + 294.0, + 242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 233.0, + 1406.0, + 233.0, + 1406.0, + 267.0, + 351.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 258.0, + 1142.0, + 258.0, + 1142.0, + 298.0, + 347.0, + 298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 308.0, + 1408.0, + 308.0, + 1408.0, + 350.0, + 292.0, + 350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 341.0, + 611.0, + 341.0, + 611.0, + 370.0, + 347.0, + 370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 393.0, + 1404.0, + 393.0, + 1404.0, + 426.0, + 296.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 420.0, + 1118.0, + 420.0, + 1118.0, + 455.0, + 345.0, + 455.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 466.0, + 1410.0, + 466.0, + 1410.0, + 513.0, + 290.0, + 513.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 499.0, + 1408.0, + 499.0, + 1408.0, + 538.0, + 349.0, + 538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 528.0, + 1307.0, + 528.0, + 1307.0, + 561.0, + 349.0, + 561.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 576.0, + 1408.0, + 576.0, + 1408.0, + 617.0, + 292.0, + 617.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 605.0, + 1074.0, + 605.0, + 1074.0, + 642.0, + 347.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 657.0, + 1406.0, + 657.0, + 1406.0, + 696.0, + 294.0, + 696.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 686.0, + 1402.0, + 686.0, + 1402.0, + 719.0, + 351.0, + 719.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 715.0, + 1408.0, + 715.0, + 1408.0, + 754.0, + 347.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 737.0, + 422.0, + 737.0, + 422.0, + 780.0, + 349.0, + 780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 796.0, + 1406.0, + 796.0, + 1406.0, + 829.0, + 296.0, + 829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 822.0, + 1408.0, + 822.0, + 1408.0, + 856.0, + 351.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 852.0, + 1049.0, + 852.0, + 1049.0, + 885.0, + 349.0, + 885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 903.0, + 1406.0, + 903.0, + 1406.0, + 937.0, + 296.0, + 937.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 926.0, + 1410.0, + 926.0, + 1410.0, + 970.0, + 347.0, + 970.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 957.0, + 689.0, + 957.0, + 689.0, + 992.0, + 344.0, + 992.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1009.0, + 1404.0, + 1009.0, + 1404.0, + 1042.0, + 296.0, + 1042.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 1036.0, + 1112.0, + 1036.0, + 1112.0, + 1071.0, + 345.0, + 1071.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1090.0, + 1406.0, + 1090.0, + 1406.0, + 1123.0, + 296.0, + 1123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1119.0, + 1294.0, + 1119.0, + 1294.0, + 1152.0, + 349.0, + 1152.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1169.0, + 1406.0, + 1169.0, + 1406.0, + 1202.0, + 296.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1198.0, + 1406.0, + 1198.0, + 1406.0, + 1231.0, + 351.0, + 1231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1221.0, + 1404.0, + 1221.0, + 1404.0, + 1260.0, + 347.0, + 1260.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1252.0, + 689.0, + 1252.0, + 689.0, + 1285.0, + 349.0, + 1285.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1304.0, + 1406.0, + 1304.0, + 1406.0, + 1337.0, + 296.0, + 1337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 1328.0, + 1410.0, + 1328.0, + 1410.0, + 1370.0, + 345.0, + 1370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 1358.0, + 1106.0, + 1358.0, + 1106.0, + 1395.0, + 345.0, + 1395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1414.0, + 1406.0, + 1414.0, + 1406.0, + 1447.0, + 296.0, + 1447.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 1436.0, + 1173.0, + 1436.0, + 1173.0, + 1476.0, + 344.0, + 1476.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1490.0, + 1408.0, + 1490.0, + 1408.0, + 1528.0, + 292.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1515.0, + 1406.0, + 1515.0, + 1406.0, + 1555.0, + 347.0, + 1555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1546.0, + 1404.0, + 1546.0, + 1404.0, + 1579.0, + 349.0, + 1579.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 1573.0, + 1055.0, + 1573.0, + 1055.0, + 1611.0, + 345.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1625.0, + 1406.0, + 1625.0, + 1406.0, + 1665.0, + 294.0, + 1665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 1652.0, + 930.0, + 1652.0, + 930.0, + 1689.0, + 345.0, + 1689.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1708.0, + 1406.0, + 1708.0, + 1406.0, + 1741.0, + 296.0, + 1741.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1735.0, + 1404.0, + 1735.0, + 1404.0, + 1768.0, + 349.0, + 1768.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 1760.0, + 943.0, + 1760.0, + 943.0, + 1797.0, + 345.0, + 1797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1814.0, + 1404.0, + 1814.0, + 1404.0, + 1847.0, + 296.0, + 1847.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1841.0, + 1408.0, + 1841.0, + 1408.0, + 1880.0, + 347.0, + 1880.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1863.0, + 518.0, + 1863.0, + 518.0, + 1907.0, + 347.0, + 1907.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1920.0, + 1406.0, + 1920.0, + 1406.0, + 1959.0, + 294.0, + 1959.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1944.0, + 1406.0, + 1944.0, + 1406.0, + 1988.0, + 347.0, + 1988.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1978.0, + 541.0, + 1978.0, + 541.0, + 2011.0, + 349.0, + 2011.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 12, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 293, + 201, + 1409, + 201, + 1409, + 898, + 293, + 898 + ], + "score": 0.884 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.833 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2095.0, + 832.0, + 2095.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 206.0, + 1407.0, + 206.0, + 1407.0, + 241.0, + 295.0, + 241.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 231.0, + 1406.0, + 231.0, + 1406.0, + 268.0, + 348.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 262.0, + 1279.0, + 262.0, + 1279.0, + 294.0, + 348.0, + 294.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 308.0, + 1407.0, + 308.0, + 1407.0, + 347.0, + 294.0, + 347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 339.0, + 1405.0, + 339.0, + 1405.0, + 374.0, + 348.0, + 374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 365.0, + 696.0, + 365.0, + 696.0, + 400.0, + 347.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 415.0, + 1405.0, + 415.0, + 1405.0, + 450.0, + 295.0, + 450.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 444.0, + 1407.0, + 444.0, + 1407.0, + 479.0, + 349.0, + 479.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 492.0, + 1407.0, + 492.0, + 1407.0, + 527.0, + 295.0, + 527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 521.0, + 1407.0, + 521.0, + 1407.0, + 556.0, + 349.0, + 556.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 544.0, + 1410.0, + 544.0, + 1410.0, + 587.0, + 346.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 578.0, + 415.0, + 578.0, + 415.0, + 608.0, + 349.0, + 608.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 625.0, + 1406.0, + 625.0, + 1406.0, + 660.0, + 295.0, + 660.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 654.0, + 1408.0, + 654.0, + 1408.0, + 689.0, + 349.0, + 689.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 683.0, + 756.0, + 683.0, + 756.0, + 714.0, + 351.0, + 714.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 729.0, + 1405.0, + 729.0, + 1405.0, + 767.0, + 294.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 759.0, + 1127.0, + 759.0, + 1127.0, + 792.0, + 348.0, + 792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 804.0, + 1406.0, + 804.0, + 1406.0, + 846.0, + 292.0, + 846.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 833.0, + 1408.0, + 833.0, + 1408.0, + 873.0, + 347.0, + 873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 862.0, + 417.0, + 862.0, + 417.0, + 896.0, + 348.0, + 896.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 13, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1068, + 1405, + 1068, + 1405, + 1615, + 298, + 1615 + ], + "score": 0.985 + }, + { + "category_id": 1, + "poly": [ + 298, + 623, + 1405, + 623, + 1405, + 958, + 298, + 958 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 334, + 1405, + 334, + 1405, + 609, + 297, + 609 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 1823, + 1402, + 1823, + 1402, + 2007, + 298, + 2007 + ], + "score": 0.974 + }, + { + "category_id": 0, + "poly": [ + 297, + 1670, + 970, + 1670, + 970, + 1709, + 297, + 1709 + ], + "score": 0.931 + }, + { + "category_id": 1, + "poly": [ + 299, + 1746, + 1142, + 1746, + 1142, + 1778, + 299, + 1778 + ], + "score": 0.921 + }, + { + "category_id": 0, + "poly": [ + 299, + 274, + 711, + 274, + 711, + 307, + 299, + 307 + ], + "score": 0.911 + }, + { + "category_id": 0, + "poly": [ + 299, + 1006, + 568, + 1006, + 568, + 1039, + 299, + 1039 + ], + "score": 0.905 + }, + { + "category_id": 0, + "poly": [ + 299, + 198, + 556, + 198, + 556, + 235, + 299, + 235 + ], + "score": 0.895 + }, + { + "category_id": 2, + "poly": [ + 836, + 2062, + 864, + 2062, + 864, + 2085, + 836, + 2085 + ], + "score": 0.84 + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1666.0, + 973.0, + 1666.0, + 973.0, + 1717.0, + 290.0, + 1717.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 273.0, + 715.0, + 273.0, + 715.0, + 312.0, + 295.0, + 312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1001.0, + 574.0, + 1001.0, + 574.0, + 1047.0, + 293.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 195.0, + 560.0, + 195.0, + 560.0, + 240.0, + 293.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 869.0, + 2058.0, + 869.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1070.0, + 1405.0, + 1070.0, + 1405.0, + 1101.0, + 295.0, + 1101.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1098.0, + 1407.0, + 1098.0, + 1407.0, + 1134.0, + 292.0, + 1134.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1127.0, + 1406.0, + 1127.0, + 1406.0, + 1165.0, + 291.0, + 1165.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1158.0, + 1407.0, + 1158.0, + 1407.0, + 1195.0, + 292.0, + 1195.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1190.0, + 1407.0, + 1190.0, + 1407.0, + 1225.0, + 293.0, + 1225.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1218.0, + 1406.0, + 1218.0, + 1406.0, + 1256.0, + 293.0, + 1256.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1247.0, + 1409.0, + 1247.0, + 1409.0, + 1286.0, + 292.0, + 1286.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1281.0, + 1408.0, + 1281.0, + 1408.0, + 1316.0, + 293.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1310.0, + 1406.0, + 1310.0, + 1406.0, + 1344.0, + 295.0, + 1344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1342.0, + 1406.0, + 1342.0, + 1406.0, + 1373.0, + 296.0, + 1373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1372.0, + 1407.0, + 1372.0, + 1407.0, + 1406.0, + 293.0, + 1406.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1401.0, + 1410.0, + 1401.0, + 1410.0, + 1439.0, + 291.0, + 1439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1431.0, + 1405.0, + 1431.0, + 1405.0, + 1467.0, + 293.0, + 1467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1464.0, + 1403.0, + 1464.0, + 1403.0, + 1496.0, + 296.0, + 1496.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1493.0, + 1405.0, + 1493.0, + 1405.0, + 1529.0, + 295.0, + 1529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1522.0, + 1405.0, + 1522.0, + 1405.0, + 1558.0, + 295.0, + 1558.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1553.0, + 1406.0, + 1553.0, + 1406.0, + 1588.0, + 295.0, + 1588.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1584.0, + 1020.0, + 1584.0, + 1020.0, + 1620.0, + 295.0, + 1620.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 624.0, + 1405.0, + 624.0, + 1405.0, + 658.0, + 295.0, + 658.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 652.0, + 1403.0, + 652.0, + 1403.0, + 687.0, + 295.0, + 687.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 683.0, + 1405.0, + 683.0, + 1405.0, + 721.0, + 293.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 713.0, + 1405.0, + 713.0, + 1405.0, + 750.0, + 293.0, + 750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 740.0, + 1406.0, + 740.0, + 1406.0, + 782.0, + 291.0, + 782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 776.0, + 1406.0, + 776.0, + 1406.0, + 810.0, + 295.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 804.0, + 1408.0, + 804.0, + 1408.0, + 842.0, + 293.0, + 842.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 836.0, + 1406.0, + 836.0, + 1406.0, + 871.0, + 295.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 865.0, + 1405.0, + 865.0, + 1405.0, + 902.0, + 293.0, + 902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 896.0, + 1405.0, + 896.0, + 1405.0, + 933.0, + 293.0, + 933.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 925.0, + 738.0, + 925.0, + 738.0, + 962.0, + 295.0, + 962.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 335.0, + 1402.0, + 335.0, + 1402.0, + 367.0, + 296.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 366.0, + 1406.0, + 366.0, + 1406.0, + 401.0, + 294.0, + 401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 397.0, + 1408.0, + 397.0, + 1408.0, + 434.0, + 294.0, + 434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 422.0, + 1405.0, + 422.0, + 1405.0, + 464.0, + 292.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 455.0, + 1405.0, + 455.0, + 1405.0, + 494.0, + 292.0, + 494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 487.0, + 1406.0, + 487.0, + 1406.0, + 524.0, + 295.0, + 524.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 516.0, + 1408.0, + 516.0, + 1408.0, + 554.0, + 294.0, + 554.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 544.0, + 1408.0, + 544.0, + 1408.0, + 585.0, + 294.0, + 585.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 580.0, + 664.0, + 580.0, + 664.0, + 612.0, + 296.0, + 612.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1820.0, + 1127.0, + 1820.0, + 1127.0, + 1858.0, + 294.0, + 1858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1145.0, + 1824.0, + 1406.0, + 1824.0, + 1406.0, + 1855.0, + 1145.0, + 1855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1855.0, + 1404.0, + 1855.0, + 1404.0, + 1887.0, + 296.0, + 1887.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1886.0, + 1404.0, + 1886.0, + 1404.0, + 1918.0, + 294.0, + 1918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1914.0, + 1404.0, + 1914.0, + 1404.0, + 1949.0, + 293.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1943.0, + 1406.0, + 1943.0, + 1406.0, + 1981.0, + 293.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1975.0, + 913.0, + 1975.0, + 913.0, + 2010.0, + 293.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1742.0, + 1146.0, + 1742.0, + 1146.0, + 1784.0, + 294.0, + 1784.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 14, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 934, + 1405, + 934, + 1405, + 1179, + 297, + 1179 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 1666, + 1404, + 1666, + 1404, + 1821, + 298, + 1821 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 297, + 1212, + 1404, + 1212, + 1404, + 1366, + 297, + 1366 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 1479, + 1404, + 1479, + 1404, + 1572, + 298, + 1572 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 299, + 1914, + 1403, + 1914, + 1403, + 2007, + 299, + 2007 + ], + "score": 0.974 + }, + { + "category_id": 3, + "poly": [ + 573, + 542, + 1116, + 542, + 1116, + 815, + 573, + 815 + ], + "score": 0.971 + }, + { + "category_id": 3, + "poly": [ + 299, + 195, + 1398, + 195, + 1398, + 412, + 299, + 412 + ], + "score": 0.964 + }, + { + "category_id": 4, + "poly": [ + 297, + 437, + 1405, + 437, + 1405, + 531, + 297, + 531 + ], + "score": 0.949 + }, + { + "category_id": 0, + "poly": [ + 297, + 1857, + 1034, + 1857, + 1034, + 1891, + 297, + 1891 + ], + "score": 0.94 + }, + { + "category_id": 0, + "poly": [ + 298, + 1610, + 932, + 1610, + 932, + 1644, + 298, + 1644 + ], + "score": 0.937 + }, + { + "category_id": 4, + "poly": [ + 402, + 843, + 1291, + 843, + 1291, + 877, + 402, + 877 + ], + "score": 0.931 + }, + { + "category_id": 0, + "poly": [ + 300, + 1408, + 746, + 1408, + 746, + 1447, + 300, + 1447 + ], + "score": 0.917 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.824 + }, + { + "category_id": 13, + "poly": [ + 954, + 439, + 1060, + 439, + 1060, + 470, + 954, + 470 + ], + "score": 0.89, + "latex": "\\mathrm { O P T } _ { 1 2 5 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 966, + 844, + 1070, + 844, + 1070, + 875, + 966, + 875 + ], + "score": 0.88, + "latex": "\\mathrm { O P T } _ { 1 2 5 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 554, + 470, + 656, + 470, + 656, + 501, + 554, + 501 + ], + "score": 0.57, + "latex": "\\mathrm { O P T } _ { 3 5 0 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 586, + 1668, + 712, + 1668, + 712, + 1701, + 586, + 1701 + ], + "score": 0.5, + "latex": "\\mathrm { F l a n T } 5 _ { \\mathrm { l a r g e } }" + }, + { + "category_id": 13, + "poly": [ + 975, + 1760, + 1016, + 1760, + 1016, + 1793, + 975, + 1793 + ], + "score": 0.48, + "latex": "_ \\mathrm { l a r g e }" + }, + { + "category_id": 13, + "poly": [ + 1132, + 1481, + 1166, + 1481, + 1166, + 1513, + 1132, + 1513 + ], + "score": 0.4, + "latex": "\\ S 4 )" + }, + { + "category_id": 13, + "poly": [ + 1298, + 470, + 1401, + 470, + 1401, + 502, + 1298, + 502 + ], + "score": 0.3, + "latex": "\\mathrm { O P T } _ { 3 5 0 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 895, + 1759, + 1016, + 1759, + 1016, + 1793, + 895, + 1793 + ], + "score": 0.28, + "latex": "\\mathrm { F l a n T } 5 _ { \\mathrm { l a r g e } }" + }, + { + "category_id": 15, + "poly": [ + 1007.0, + 585.0, + 1122.0, + 585.0, + 1122.0, + 622.0, + 1007.0, + 622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1004.0, + 629.0, + 1085.0, + 629.0, + 1085.0, + 673.0, + 1004.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 647.0, + 662.0, + 697.0, + 662.0, + 697.0, + 722.0, + 647.0, + 722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1004.0, + 678.0, + 1104.0, + 678.0, + 1104.0, + 717.0, + 1004.0, + 717.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 785.0, + 1060.0, + 785.0, + 1060.0, + 815.0, + 1006.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 379.0, + 196.0, + 473.0, + 196.0, + 473.0, + 220.0, + 379.0, + 220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 617.0, + 198.0, + 674.0, + 198.0, + 674.0, + 219.0, + 617.0, + 219.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 760.0, + 208.0, + 777.0, + 208.0, + 777.0, + 224.0, + 760.0, + 224.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 196.0, + 892.0, + 196.0, + 892.0, + 221.0, + 840.0, + 221.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1067.0, + 197.0, + 1105.0, + 197.0, + 1105.0, + 219.0, + 1067.0, + 219.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1201.0, + 208.0, + 1217.0, + 208.0, + 1217.0, + 225.0, + 1201.0, + 225.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1279.0, + 197.0, + 1332.0, + 197.0, + 1332.0, + 218.0, + 1279.0, + 218.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 534.0, + 218.0, + 559.0, + 218.0, + 559.0, + 237.0, + 534.0, + 237.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 980.0, + 225.0, + 997.0, + 225.0, + 997.0, + 240.0, + 980.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 238.0, + 339.0, + 238.0, + 339.0, + 256.0, + 314.0, + 256.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 760.0, + 245.0, + 777.0, + 245.0, + 777.0, + 260.0, + 760.0, + 260.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1201.0, + 232.0, + 1218.0, + 232.0, + 1218.0, + 248.0, + 1201.0, + 248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 251.0, + 560.0, + 251.0, + 560.0, + 270.0, + 538.0, + 270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 978.0, + 256.0, + 1000.0, + 256.0, + 1000.0, + 275.0, + 978.0, + 275.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1201.0, + 255.0, + 1217.0, + 255.0, + 1217.0, + 271.0, + 1201.0, + 271.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 301.0, + 274.0, + 338.0, + 274.0, + 338.0, + 315.0, + 301.0, + 315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 515.0, + 277.0, + 565.0, + 277.0, + 565.0, + 321.0, + 515.0, + 321.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 740.0, + 274.0, + 785.0, + 274.0, + 785.0, + 321.0, + 740.0, + 321.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 961.0, + 279.0, + 1003.0, + 279.0, + 1003.0, + 320.0, + 961.0, + 320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 299.0, + 1219.0, + 299.0, + 1219.0, + 320.0, + 1198.0, + 320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 310.0, + 339.0, + 310.0, + 339.0, + 328.0, + 314.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 380.0, + 318.0, + 454.0, + 318.0, + 454.0, + 338.0, + 380.0, + 338.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 540.0, + 318.0, + 557.0, + 318.0, + 557.0, + 334.0, + 540.0, + 334.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 318.0, + 674.0, + 318.0, + 674.0, + 338.0, + 600.0, + 338.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 759.0, + 317.0, + 780.0, + 317.0, + 780.0, + 336.0, + 759.0, + 336.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 820.0, + 318.0, + 893.0, + 318.0, + 893.0, + 338.0, + 820.0, + 338.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 977.0, + 322.0, + 1000.0, + 322.0, + 1000.0, + 341.0, + 977.0, + 341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1040.0, + 318.0, + 1113.0, + 318.0, + 1113.0, + 338.0, + 1040.0, + 338.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1199.0, + 326.0, + 1217.0, + 326.0, + 1217.0, + 342.0, + 1199.0, + 342.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1259.0, + 318.0, + 1333.0, + 318.0, + 1333.0, + 338.0, + 1259.0, + 338.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 380.0, + 336.0, + 485.0, + 336.0, + 485.0, + 356.0, + 380.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 336.0, + 705.0, + 336.0, + 705.0, + 356.0, + 600.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 820.0, + 336.0, + 925.0, + 336.0, + 925.0, + 356.0, + 820.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1040.0, + 336.0, + 1145.0, + 336.0, + 1145.0, + 356.0, + 1040.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1259.0, + 336.0, + 1364.0, + 336.0, + 1364.0, + 356.0, + 1259.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 346.0, + 339.0, + 346.0, + 339.0, + 364.0, + 314.0, + 364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 379.0, + 353.0, + 446.0, + 353.0, + 446.0, + 374.0, + 379.0, + 374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 539.0, + 352.0, + 559.0, + 352.0, + 559.0, + 367.0, + 539.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 576.0, + 358.0, + 589.0, + 358.0, + 589.0, + 367.0, + 576.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 599.0, + 353.0, + 666.0, + 353.0, + 666.0, + 374.0, + 599.0, + 374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 760.0, + 354.0, + 777.0, + 354.0, + 777.0, + 370.0, + 760.0, + 370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 796.0, + 358.0, + 807.0, + 358.0, + 807.0, + 367.0, + 796.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 819.0, + 353.0, + 886.0, + 353.0, + 886.0, + 374.0, + 819.0, + 374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 977.0, + 354.0, + 1001.0, + 354.0, + 1001.0, + 373.0, + 977.0, + 373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1016.0, + 358.0, + 1027.0, + 358.0, + 1027.0, + 367.0, + 1016.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1038.0, + 353.0, + 1107.0, + 353.0, + 1107.0, + 374.0, + 1038.0, + 374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 348.0, + 1219.0, + 348.0, + 1219.0, + 367.0, + 1198.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1236.0, + 358.0, + 1249.0, + 358.0, + 1249.0, + 367.0, + 1236.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1258.0, + 353.0, + 1327.0, + 353.0, + 1327.0, + 374.0, + 1258.0, + 374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1201.0, + 373.0, + 1218.0, + 373.0, + 1218.0, + 389.0, + 1201.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 383.0, + 370.0, + 383.0, + 370.0, + 396.0, + 354.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 383.0, + 390.0, + 383.0, + 390.0, + 396.0, + 374.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 399.0, + 384.0, + 410.0, + 384.0, + 410.0, + 395.0, + 399.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 420.0, + 384.0, + 433.0, + 384.0, + 433.0, + 395.0, + 420.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 439.0, + 383.0, + 455.0, + 383.0, + 455.0, + 398.0, + 439.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 463.0, + 384.0, + 475.0, + 384.0, + 475.0, + 395.0, + 463.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 484.0, + 384.0, + 496.0, + 384.0, + 496.0, + 395.0, + 484.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 505.0, + 384.0, + 518.0, + 384.0, + 518.0, + 395.0, + 505.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 555.0, + 384.0, + 564.0, + 384.0, + 564.0, + 395.0, + 555.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 576.0, + 384.0, + 589.0, + 384.0, + 589.0, + 395.0, + 576.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 596.0, + 384.0, + 609.0, + 384.0, + 609.0, + 395.0, + 596.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 619.0, + 384.0, + 631.0, + 384.0, + 631.0, + 395.0, + 619.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 637.0, + 383.0, + 655.0, + 383.0, + 655.0, + 396.0, + 637.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 659.0, + 383.0, + 675.0, + 383.0, + 675.0, + 398.0, + 659.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 684.0, + 384.0, + 695.0, + 384.0, + 695.0, + 395.0, + 684.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 704.0, + 384.0, + 716.0, + 384.0, + 716.0, + 395.0, + 704.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 725.0, + 384.0, + 737.0, + 384.0, + 737.0, + 395.0, + 725.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 775.0, + 384.0, + 784.0, + 384.0, + 784.0, + 395.0, + 775.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 794.0, + 383.0, + 811.0, + 383.0, + 811.0, + 396.0, + 794.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 814.0, + 383.0, + 830.0, + 383.0, + 830.0, + 396.0, + 814.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 384.0, + 850.0, + 384.0, + 850.0, + 395.0, + 839.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 858.0, + 383.0, + 875.0, + 383.0, + 875.0, + 396.0, + 858.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 878.0, + 383.0, + 895.0, + 383.0, + 895.0, + 398.0, + 878.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 903.0, + 384.0, + 915.0, + 384.0, + 915.0, + 395.0, + 903.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 923.0, + 384.0, + 936.0, + 384.0, + 936.0, + 395.0, + 923.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 943.0, + 383.0, + 960.0, + 383.0, + 960.0, + 396.0, + 943.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 995.0, + 384.0, + 1005.0, + 384.0, + 1005.0, + 395.0, + 995.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1016.0, + 384.0, + 1028.0, + 384.0, + 1028.0, + 395.0, + 1016.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1036.0, + 384.0, + 1048.0, + 384.0, + 1048.0, + 395.0, + 1036.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1058.0, + 384.0, + 1071.0, + 384.0, + 1071.0, + 395.0, + 1058.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1078.0, + 383.0, + 1095.0, + 383.0, + 1095.0, + 396.0, + 1078.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1098.0, + 383.0, + 1115.0, + 383.0, + 1115.0, + 398.0, + 1098.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1123.0, + 384.0, + 1135.0, + 384.0, + 1135.0, + 395.0, + 1123.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1145.0, + 384.0, + 1157.0, + 384.0, + 1157.0, + 395.0, + 1145.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1164.0, + 384.0, + 1177.0, + 384.0, + 1177.0, + 395.0, + 1164.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1216.0, + 384.0, + 1224.0, + 384.0, + 1224.0, + 395.0, + 1216.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1236.0, + 384.0, + 1248.0, + 384.0, + 1248.0, + 395.0, + 1236.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1253.0, + 383.0, + 1271.0, + 383.0, + 1271.0, + 396.0, + 1253.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1278.0, + 384.0, + 1291.0, + 384.0, + 1291.0, + 395.0, + 1278.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1298.0, + 383.0, + 1314.0, + 383.0, + 1314.0, + 396.0, + 1298.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1318.0, + 383.0, + 1336.0, + 383.0, + 1336.0, + 398.0, + 1318.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1341.0, + 383.0, + 1357.0, + 383.0, + 1357.0, + 396.0, + 1341.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1364.0, + 384.0, + 1377.0, + 384.0, + 1377.0, + 395.0, + 1364.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1386.0, + 384.0, + 1397.0, + 384.0, + 1397.0, + 395.0, + 1386.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 394.0, + 452.0, + 394.0, + 452.0, + 411.0, + 398.0, + 411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 618.0, + 394.0, + 672.0, + 394.0, + 672.0, + 411.0, + 618.0, + 411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 394.0, + 891.0, + 394.0, + 891.0, + 411.0, + 838.0, + 411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1058.0, + 394.0, + 1111.0, + 394.0, + 1111.0, + 411.0, + 1058.0, + 411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1278.0, + 393.0, + 1333.0, + 393.0, + 1333.0, + 412.0, + 1278.0, + 412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1168.0, + 282.0, + 1231.0, + 282.0, + 1231.0, + 305.5, + 1168.0, + 305.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 437.0, + 953.0, + 437.0, + 953.0, + 474.0, + 295.0, + 474.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1061.0, + 437.0, + 1405.0, + 437.0, + 1405.0, + 474.0, + 1061.0, + 474.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 466.0, + 553.0, + 466.0, + 553.0, + 508.0, + 292.0, + 508.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 657.0, + 466.0, + 1297.0, + 466.0, + 1297.0, + 508.0, + 657.0, + 508.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 466.0, + 1405.0, + 466.0, + 1405.0, + 508.0, + 1402.0, + 508.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 497.0, + 793.0, + 497.0, + 793.0, + 535.0, + 294.0, + 535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1855.0, + 1036.0, + 1855.0, + 1036.0, + 1895.0, + 296.0, + 1895.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1608.0, + 934.0, + 1608.0, + 934.0, + 1648.0, + 296.0, + 1648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 841.0, + 965.0, + 841.0, + 965.0, + 881.0, + 404.0, + 881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1071.0, + 841.0, + 1296.0, + 841.0, + 1296.0, + 881.0, + 1071.0, + 881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1403.0, + 749.0, + 1403.0, + 749.0, + 1455.0, + 293.0, + 1455.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 933.0, + 1406.0, + 933.0, + 1406.0, + 967.0, + 296.0, + 967.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 964.0, + 1405.0, + 964.0, + 1405.0, + 998.0, + 295.0, + 998.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 993.0, + 1403.0, + 993.0, + 1403.0, + 1029.0, + 294.0, + 1029.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1025.0, + 1405.0, + 1025.0, + 1405.0, + 1058.0, + 294.0, + 1058.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1055.0, + 1406.0, + 1055.0, + 1406.0, + 1090.0, + 292.0, + 1090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1086.0, + 1405.0, + 1086.0, + 1405.0, + 1120.0, + 295.0, + 1120.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1116.0, + 1406.0, + 1116.0, + 1406.0, + 1150.0, + 294.0, + 1150.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1143.0, + 1045.0, + 1143.0, + 1045.0, + 1184.0, + 292.0, + 1184.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1668.0, + 585.0, + 1668.0, + 585.0, + 1701.0, + 296.0, + 1701.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 713.0, + 1668.0, + 1404.0, + 1668.0, + 1404.0, + 1701.0, + 713.0, + 1701.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1696.0, + 1404.0, + 1696.0, + 1404.0, + 1732.0, + 293.0, + 1732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1726.0, + 1405.0, + 1726.0, + 1405.0, + 1763.0, + 294.0, + 1763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1754.0, + 894.0, + 1754.0, + 894.0, + 1798.0, + 292.0, + 1798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1017.0, + 1754.0, + 1408.0, + 1754.0, + 1408.0, + 1798.0, + 1017.0, + 1798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1787.0, + 631.0, + 1787.0, + 631.0, + 1825.0, + 294.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1210.0, + 1406.0, + 1210.0, + 1406.0, + 1247.0, + 296.0, + 1247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1241.0, + 1404.0, + 1241.0, + 1404.0, + 1274.0, + 296.0, + 1274.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1269.0, + 1406.0, + 1269.0, + 1406.0, + 1309.0, + 295.0, + 1309.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1301.0, + 1404.0, + 1301.0, + 1404.0, + 1341.0, + 292.0, + 1341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1330.0, + 476.0, + 1330.0, + 476.0, + 1373.0, + 294.0, + 1373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1478.0, + 1131.0, + 1478.0, + 1131.0, + 1516.0, + 293.0, + 1516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1167.0, + 1478.0, + 1405.0, + 1478.0, + 1405.0, + 1516.0, + 1167.0, + 1516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1507.0, + 1405.0, + 1507.0, + 1405.0, + 1546.0, + 293.0, + 1546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1538.0, + 420.0, + 1538.0, + 420.0, + 1577.0, + 292.0, + 1577.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1911.0, + 1407.0, + 1911.0, + 1407.0, + 1952.0, + 293.0, + 1952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1945.0, + 1404.0, + 1945.0, + 1404.0, + 1979.0, + 295.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1974.0, + 1405.0, + 1974.0, + 1405.0, + 2010.0, + 294.0, + 2010.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 15, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1614, + 1405, + 1614, + 1405, + 1920, + 297, + 1920 + ], + "score": 0.981 + }, + { + "category_id": 5, + "poly": [ + 408, + 808, + 1290, + 808, + 1290, + 1089, + 408, + 1089 + ], + "score": 0.979, + "html": "
TransferCOCO CaptionNoCapsVQAv2GQAOKVQA
OPT125M→350M1.12.710.06.06.0
OPT125M→1.3B-2.71-1
OPT350M→1.3B11.71.02.01.0
OPT125M→2.7B3.03.03.01.03.0
OPT350M→2.7B3.34.51.01.01.0
OPT1.3B→2.7B2.33.0111
FlanT5base-→XL11.01.89.00.4
" + }, + { + "category_id": 1, + "poly": [ + 298, + 1342, + 1405, + 1342, + 1405, + 1497, + 298, + 1497 + ], + "score": 0.979 + }, + { + "category_id": 3, + "poly": [ + 634, + 201, + 1064, + 201, + 1064, + 620, + 634, + 620 + ], + "score": 0.963 + }, + { + "category_id": 1, + "poly": [ + 295, + 1162, + 1402, + 1162, + 1402, + 1224, + 295, + 1224 + ], + "score": 0.953 + }, + { + "category_id": 0, + "poly": [ + 297, + 1278, + 961, + 1278, + 961, + 1312, + 297, + 1312 + ], + "score": 0.934 + }, + { + "category_id": 0, + "poly": [ + 299, + 1551, + 904, + 1551, + 904, + 1585, + 299, + 1585 + ], + "score": 0.927 + }, + { + "category_id": 6, + "poly": [ + 294, + 729, + 1402, + 729, + 1402, + 793, + 294, + 793 + ], + "score": 0.92 + }, + { + "category_id": 4, + "poly": [ + 453, + 651, + 1242, + 651, + 1242, + 687, + 453, + 687 + ], + "score": 0.916 + }, + { + "category_id": 2, + "poly": [ + 332, + 1978, + 765, + 1978, + 765, + 2007, + 332, + 2007 + ], + "score": 0.911 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 864, + 2061, + 864, + 2085, + 836, + 2085 + ], + "score": 0.856 + }, + { + "category_id": 13, + "poly": [ + 967, + 1466, + 1042, + 1466, + 1042, + 1495, + 967, + 1495 + ], + "score": 0.85, + "latex": "6 9 . 2 \\%" + }, + { + "category_id": 13, + "poly": [ + 296, + 1435, + 352, + 1435, + 352, + 1464, + 296, + 1464 + ], + "score": 0.83, + "latex": "74 \\%" + }, + { + "category_id": 13, + "poly": [ + 924, + 654, + 967, + 654, + 967, + 687, + 924, + 687 + ], + "score": 0.46, + "latex": "\\mathrm { l a r g e }" + }, + { + "category_id": 15, + "poly": [ + 672.0, + 195.0, + 1072.0, + 195.0, + 1072.0, + 245.0, + 672.0, + 245.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 658.0, + 237.0, + 709.0, + 237.0, + 709.0, + 276.0, + 658.0, + 276.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 657.0, + 280.0, + 711.0, + 280.0, + 711.0, + 321.0, + 657.0, + 321.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 777.0, + 317.0, + 831.0, + 317.0, + 831.0, + 330.0, + 777.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 858.0, + 313.0, + 884.0, + 313.0, + 884.0, + 326.0, + 858.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 658.0, + 325.0, + 709.0, + 325.0, + 709.0, + 363.0, + 658.0, + 363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 634.0, + 367.0, + 668.0, + 367.0, + 668.0, + 430.0, + 634.0, + 430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 666.0, + 369.0, + 708.0, + 369.0, + 708.0, + 404.0, + 666.0, + 404.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 665.0, + 409.0, + 711.0, + 409.0, + 711.0, + 454.0, + 665.0, + 454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 792.0, + 435.0, + 1000.0, + 435.0, + 1000.0, + 473.0, + 792.0, + 473.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 667.0, + 456.0, + 709.0, + 456.0, + 709.0, + 498.0, + 667.0, + 498.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 792.0, + 470.0, + 939.0, + 470.0, + 939.0, + 510.0, + 792.0, + 510.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 668.0, + 500.0, + 708.0, + 500.0, + 708.0, + 536.0, + 668.0, + 536.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 791.0, + 504.0, + 927.0, + 504.0, + 927.0, + 546.0, + 791.0, + 546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 681.0, + 547.0, + 704.0, + 547.0, + 704.0, + 575.0, + 681.0, + 575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 701.0, + 569.0, + 716.0, + 569.0, + 716.0, + 588.0, + 701.0, + 588.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 779.0, + 566.0, + 802.0, + 566.0, + 802.0, + 594.0, + 779.0, + 594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 859.0, + 564.0, + 885.0, + 564.0, + 885.0, + 596.0, + 859.0, + 596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 943.0, + 569.0, + 962.0, + 569.0, + 962.0, + 592.0, + 943.0, + 592.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1022.0, + 567.0, + 1046.0, + 567.0, + 1046.0, + 594.0, + 1022.0, + 594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 823.0, + 584.0, + 922.0, + 584.0, + 922.0, + 627.0, + 823.0, + 627.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1275.0, + 966.0, + 1275.0, + 966.0, + 1318.0, + 293.0, + 1318.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1549.0, + 906.0, + 1549.0, + 906.0, + 1592.0, + 295.0, + 1592.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 725.0, + 1405.0, + 725.0, + 1405.0, + 767.0, + 292.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 761.0, + 1017.0, + 761.0, + 1017.0, + 793.0, + 297.0, + 793.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 453.0, + 648.0, + 923.0, + 648.0, + 923.0, + 691.0, + 453.0, + 691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 968.0, + 648.0, + 1246.0, + 648.0, + 1246.0, + 691.0, + 968.0, + 691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 330.0, + 1970.0, + 770.0, + 1970.0, + 770.0, + 2015.0, + 330.0, + 2015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 869.0, + 2058.0, + 869.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1616.0, + 1405.0, + 1616.0, + 1405.0, + 1648.0, + 296.0, + 1648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1645.0, + 1408.0, + 1645.0, + 1408.0, + 1679.0, + 294.0, + 1679.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1677.0, + 1403.0, + 1677.0, + 1403.0, + 1709.0, + 295.0, + 1709.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1704.0, + 1407.0, + 1704.0, + 1407.0, + 1742.0, + 294.0, + 1742.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1734.0, + 1406.0, + 1734.0, + 1406.0, + 1772.0, + 292.0, + 1772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1766.0, + 1408.0, + 1766.0, + 1408.0, + 1804.0, + 294.0, + 1804.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1796.0, + 1407.0, + 1796.0, + 1407.0, + 1832.0, + 292.0, + 1832.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1827.0, + 1405.0, + 1827.0, + 1405.0, + 1862.0, + 294.0, + 1862.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1860.0, + 1403.0, + 1860.0, + 1403.0, + 1891.0, + 296.0, + 1891.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1886.0, + 1241.0, + 1886.0, + 1241.0, + 1924.0, + 294.0, + 1924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1344.0, + 1405.0, + 1344.0, + 1405.0, + 1377.0, + 296.0, + 1377.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1372.0, + 1405.0, + 1372.0, + 1405.0, + 1410.0, + 292.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1401.0, + 1408.0, + 1401.0, + 1408.0, + 1442.0, + 292.0, + 1442.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1431.0, + 295.0, + 1431.0, + 295.0, + 1470.0, + 292.0, + 1470.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1431.0, + 1407.0, + 1431.0, + 1407.0, + 1470.0, + 353.0, + 1470.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1465.0, + 966.0, + 1465.0, + 966.0, + 1502.0, + 295.0, + 1502.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1043.0, + 1465.0, + 1175.0, + 1465.0, + 1175.0, + 1502.0, + 1043.0, + 1502.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1161.0, + 1404.0, + 1161.0, + 1404.0, + 1197.0, + 294.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1194.0, + 1057.0, + 1194.0, + 1057.0, + 1226.0, + 297.0, + 1226.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 16, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 3, + "poly": [ + 297, + 294, + 1393, + 294, + 1393, + 1825, + 297, + 1825 + ], + "score": 0.968 + }, + { + "category_id": 4, + "poly": [ + 294, + 1836, + 1405, + 1836, + 1405, + 1901, + 294, + 1901 + ], + "score": 0.926 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.825 + }, + { + "category_id": 15, + "poly": [ + 425.0, + 302.0, + 512.0, + 302.0, + 512.0, + 319.0, + 425.0, + 319.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 662.0, + 300.0, + 716.0, + 300.0, + 716.0, + 322.0, + 662.0, + 322.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 298.0, + 932.0, + 298.0, + 932.0, + 320.0, + 882.0, + 320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1108.0, + 300.0, + 1142.0, + 300.0, + 1142.0, + 319.0, + 1108.0, + 319.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1317.0, + 298.0, + 1370.0, + 298.0, + 1370.0, + 320.0, + 1317.0, + 320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 758.0, + 314.0, + 780.0, + 314.0, + 780.0, + 332.0, + 758.0, + 332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 326.0, + 340.0, + 326.0, + 340.0, + 346.0, + 315.0, + 346.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 534.0, + 327.0, + 560.0, + 327.0, + 560.0, + 348.0, + 534.0, + 348.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 343.0, + 779.0, + 343.0, + 779.0, + 361.0, + 757.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 343.0, + 995.0, + 343.0, + 995.0, + 356.0, + 979.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 341.0, + 1215.0, + 341.0, + 1215.0, + 356.0, + 1198.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 349.0, + 340.0, + 349.0, + 340.0, + 370.0, + 313.0, + 370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 373.0, + 340.0, + 373.0, + 340.0, + 394.0, + 313.0, + 394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 541.0, + 370.0, + 556.0, + 370.0, + 556.0, + 385.0, + 541.0, + 385.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 758.0, + 371.0, + 779.0, + 371.0, + 779.0, + 392.0, + 758.0, + 392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 375.0, + 998.0, + 375.0, + 998.0, + 395.0, + 976.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 377.0, + 1215.0, + 377.0, + 1215.0, + 392.0, + 1198.0, + 392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 383.0, + 346.0, + 383.0, + 346.0, + 422.0, + 298.0, + 422.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 522.0, + 388.0, + 539.0, + 388.0, + 539.0, + 417.0, + 522.0, + 417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 966.0, + 388.0, + 983.0, + 388.0, + 983.0, + 417.0, + 966.0, + 417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1185.0, + 388.0, + 1202.0, + 388.0, + 1202.0, + 417.0, + 1185.0, + 417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 407.0, + 560.0, + 407.0, + 560.0, + 426.0, + 538.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 412.0, + 998.0, + 412.0, + 998.0, + 433.0, + 976.0, + 433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 412.0, + 1215.0, + 412.0, + 1215.0, + 427.0, + 1198.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 421.0, + 340.0, + 421.0, + 340.0, + 441.0, + 315.0, + 441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 431.0, + 461.0, + 431.0, + 461.0, + 453.0, + 373.0, + 453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 594.0, + 434.0, + 679.0, + 434.0, + 679.0, + 451.0, + 594.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 429.0, + 780.0, + 429.0, + 780.0, + 448.0, + 757.0, + 448.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 811.0, + 434.0, + 898.0, + 434.0, + 898.0, + 451.0, + 811.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1030.0, + 434.0, + 1117.0, + 434.0, + 1117.0, + 451.0, + 1030.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1251.0, + 434.0, + 1336.0, + 434.0, + 1336.0, + 451.0, + 1251.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 444.0, + 340.0, + 444.0, + 340.0, + 465.0, + 313.0, + 465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 448.0, + 434.0, + 448.0, + 434.0, + 465.0, + 374.0, + 465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 541.0, + 450.0, + 558.0, + 450.0, + 558.0, + 465.0, + 541.0, + 465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 594.0, + 448.0, + 653.0, + 448.0, + 653.0, + 465.0, + 594.0, + 465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 813.0, + 448.0, + 870.0, + 448.0, + 870.0, + 465.0, + 813.0, + 465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 448.0, + 998.0, + 448.0, + 998.0, + 467.0, + 976.0, + 467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1032.0, + 448.0, + 1091.0, + 448.0, + 1091.0, + 465.0, + 1032.0, + 465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 448.0, + 1215.0, + 448.0, + 1215.0, + 463.0, + 1198.0, + 463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1249.0, + 448.0, + 1310.0, + 448.0, + 1310.0, + 465.0, + 1249.0, + 465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 468.0, + 340.0, + 468.0, + 340.0, + 487.0, + 315.0, + 487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 463.0, + 369.0, + 463.0, + 369.0, + 478.0, + 352.0, + 478.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 460.0, + 430.0, + 460.0, + 430.0, + 482.0, + 373.0, + 482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 571.0, + 463.0, + 587.0, + 463.0, + 587.0, + 477.0, + 571.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 460.0, + 650.0, + 460.0, + 650.0, + 482.0, + 592.0, + 482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 458.0, + 780.0, + 458.0, + 780.0, + 477.0, + 757.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 463.0, + 806.0, + 463.0, + 806.0, + 477.0, + 789.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 811.0, + 461.0, + 867.0, + 461.0, + 867.0, + 478.0, + 811.0, + 478.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1008.0, + 463.0, + 1023.0, + 463.0, + 1023.0, + 477.0, + 1008.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 460.0, + 1088.0, + 460.0, + 1088.0, + 482.0, + 1028.0, + 482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 463.0, + 1246.0, + 463.0, + 1246.0, + 477.0, + 1229.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 460.0, + 1307.0, + 460.0, + 1307.0, + 482.0, + 1248.0, + 482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 487.0, + 366.0, + 487.0, + 366.0, + 501.0, + 352.0, + 501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 487.0, + 385.0, + 487.0, + 385.0, + 502.0, + 368.0, + 502.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 386.0, + 487.0, + 403.0, + 487.0, + 403.0, + 501.0, + 386.0, + 501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 485.0, + 442.0, + 485.0, + 442.0, + 502.0, + 405.0, + 502.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 459.0, + 484.0, + 519.0, + 484.0, + 519.0, + 506.0, + 459.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 487.0, + 604.0, + 487.0, + 604.0, + 502.0, + 570.0, + 502.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 607.0, + 489.0, + 621.0, + 489.0, + 621.0, + 501.0, + 607.0, + 501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 622.0, + 485.0, + 660.0, + 485.0, + 660.0, + 502.0, + 622.0, + 502.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 680.0, + 484.0, + 738.0, + 484.0, + 738.0, + 506.0, + 680.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 487.0, + 804.0, + 487.0, + 804.0, + 502.0, + 789.0, + 502.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 808.0, + 487.0, + 879.0, + 487.0, + 879.0, + 501.0, + 808.0, + 501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 896.0, + 484.0, + 955.0, + 484.0, + 955.0, + 506.0, + 896.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 989.0, + 484.0, + 1059.0, + 484.0, + 1059.0, + 506.0, + 989.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1061.0, + 487.0, + 1098.0, + 487.0, + 1098.0, + 501.0, + 1061.0, + 501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 485.0, + 1173.0, + 485.0, + 1173.0, + 502.0, + 1100.0, + 502.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1207.0, + 484.0, + 1394.0, + 484.0, + 1394.0, + 506.0, + 1207.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 495.0, + 447.0, + 495.0, + 447.0, + 518.0, + 398.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 617.0, + 495.0, + 667.0, + 495.0, + 667.0, + 518.0, + 617.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 499.0, + 884.0, + 499.0, + 884.0, + 516.0, + 840.0, + 516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 499.0, + 1103.0, + 499.0, + 1103.0, + 516.0, + 1057.0, + 516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1277.0, + 495.0, + 1324.0, + 495.0, + 1324.0, + 518.0, + 1277.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 419.0, + 514.0, + 514.0, + 514.0, + 514.0, + 540.0, + 419.0, + 540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 519.0, + 928.0, + 519.0, + 928.0, + 536.0, + 882.0, + 536.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1107.0, + 519.0, + 1141.0, + 519.0, + 1141.0, + 536.0, + 1107.0, + 536.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1319.0, + 519.0, + 1368.0, + 519.0, + 1368.0, + 536.0, + 1319.0, + 536.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 529.0, + 340.0, + 529.0, + 340.0, + 548.0, + 313.0, + 548.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 528.0, + 780.0, + 528.0, + 780.0, + 546.0, + 757.0, + 546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 545.0, + 996.0, + 545.0, + 996.0, + 560.0, + 979.0, + 560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 541.0, + 1217.0, + 541.0, + 1217.0, + 562.0, + 1195.0, + 562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 552.0, + 340.0, + 552.0, + 340.0, + 572.0, + 313.0, + 572.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 534.0, + 558.0, + 560.0, + 558.0, + 560.0, + 579.0, + 534.0, + 579.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 553.0, + 779.0, + 553.0, + 779.0, + 572.0, + 757.0, + 572.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 574.0, + 340.0, + 574.0, + 340.0, + 594.0, + 313.0, + 594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 574.0, + 998.0, + 574.0, + 998.0, + 594.0, + 976.0, + 594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1197.0, + 574.0, + 1214.0, + 574.0, + 1214.0, + 589.0, + 1197.0, + 589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 760.0, + 582.0, + 777.0, + 582.0, + 777.0, + 597.0, + 760.0, + 597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 596.0, + 346.0, + 596.0, + 346.0, + 642.0, + 298.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 522.0, + 596.0, + 560.0, + 596.0, + 560.0, + 636.0, + 522.0, + 636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 741.0, + 601.0, + 784.0, + 601.0, + 784.0, + 642.0, + 741.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 961.0, + 599.0, + 1001.0, + 599.0, + 1001.0, + 642.0, + 961.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1180.0, + 597.0, + 1220.0, + 597.0, + 1220.0, + 640.0, + 1180.0, + 640.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 640.0, + 340.0, + 640.0, + 340.0, + 659.0, + 315.0, + 659.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 630.0, + 560.0, + 630.0, + 560.0, + 650.0, + 538.0, + 650.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 760.0, + 635.0, + 777.0, + 635.0, + 777.0, + 648.0, + 760.0, + 648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 633.0, + 998.0, + 633.0, + 998.0, + 653.0, + 976.0, + 653.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 630.0, + 1215.0, + 630.0, + 1215.0, + 650.0, + 1195.0, + 650.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 652.0, + 459.0, + 652.0, + 459.0, + 669.0, + 352.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 655.0, + 590.0, + 655.0, + 590.0, + 667.0, + 570.0, + 667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 594.0, + 652.0, + 679.0, + 652.0, + 679.0, + 669.0, + 594.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 791.0, + 655.0, + 808.0, + 655.0, + 808.0, + 667.0, + 791.0, + 667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 811.0, + 652.0, + 896.0, + 652.0, + 896.0, + 669.0, + 811.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1030.0, + 652.0, + 1115.0, + 652.0, + 1115.0, + 669.0, + 1030.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 653.0, + 1248.0, + 653.0, + 1248.0, + 665.0, + 1229.0, + 665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1249.0, + 652.0, + 1334.0, + 652.0, + 1334.0, + 669.0, + 1249.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 662.0, + 340.0, + 662.0, + 340.0, + 682.0, + 313.0, + 682.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 669.0, + 368.0, + 669.0, + 368.0, + 681.0, + 352.0, + 681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 664.0, + 436.0, + 664.0, + 436.0, + 686.0, + 373.0, + 686.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 667.0, + 560.0, + 667.0, + 560.0, + 686.0, + 538.0, + 686.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 664.0, + 655.0, + 664.0, + 655.0, + 686.0, + 592.0, + 686.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 760.0, + 662.0, + 777.0, + 662.0, + 777.0, + 677.0, + 760.0, + 677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 809.0, + 664.0, + 874.0, + 664.0, + 874.0, + 686.0, + 809.0, + 686.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 664.0, + 998.0, + 664.0, + 998.0, + 684.0, + 976.0, + 684.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1030.0, + 664.0, + 1091.0, + 664.0, + 1091.0, + 686.0, + 1030.0, + 686.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 662.0, + 1217.0, + 662.0, + 1217.0, + 681.0, + 1195.0, + 681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 664.0, + 1310.0, + 664.0, + 1310.0, + 686.0, + 1248.0, + 686.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 684.0, + 340.0, + 684.0, + 340.0, + 704.0, + 315.0, + 704.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 682.0, + 368.0, + 682.0, + 368.0, + 696.0, + 352.0, + 696.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 681.0, + 430.0, + 681.0, + 430.0, + 698.0, + 374.0, + 698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 571.0, + 682.0, + 585.0, + 682.0, + 585.0, + 694.0, + 571.0, + 694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 594.0, + 681.0, + 648.0, + 681.0, + 648.0, + 698.0, + 594.0, + 698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 758.0, + 686.0, + 780.0, + 686.0, + 780.0, + 704.0, + 758.0, + 704.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 791.0, + 682.0, + 806.0, + 682.0, + 806.0, + 694.0, + 791.0, + 694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 811.0, + 681.0, + 867.0, + 681.0, + 867.0, + 698.0, + 811.0, + 698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1030.0, + 681.0, + 1086.0, + 681.0, + 1086.0, + 698.0, + 1030.0, + 698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1197.0, + 691.0, + 1217.0, + 691.0, + 1217.0, + 711.0, + 1197.0, + 711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 677.0, + 1307.0, + 677.0, + 1307.0, + 699.0, + 1248.0, + 699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 335.0, + 706.0, + 349.0, + 706.0, + 349.0, + 718.0, + 335.0, + 718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 706.0, + 403.0, + 706.0, + 403.0, + 720.0, + 352.0, + 720.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 704.0, + 442.0, + 704.0, + 442.0, + 721.0, + 405.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 447.0, + 708.0, + 456.0, + 708.0, + 456.0, + 716.0, + 447.0, + 716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 459.0, + 704.0, + 517.0, + 704.0, + 517.0, + 721.0, + 459.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 704.0, + 643.0, + 704.0, + 643.0, + 721.0, + 570.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 706.0, + 662.0, + 706.0, + 662.0, + 718.0, + 645.0, + 718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 679.0, + 704.0, + 736.0, + 704.0, + 736.0, + 721.0, + 679.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 775.0, + 708.0, + 784.0, + 708.0, + 784.0, + 716.0, + 775.0, + 716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 704.0, + 860.0, + 704.0, + 860.0, + 721.0, + 789.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 862.0, + 706.0, + 879.0, + 706.0, + 879.0, + 718.0, + 862.0, + 718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 894.0, + 701.0, + 955.0, + 701.0, + 955.0, + 723.0, + 894.0, + 723.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 989.0, + 704.0, + 1096.0, + 704.0, + 1096.0, + 721.0, + 989.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1098.0, + 704.0, + 1173.0, + 704.0, + 1173.0, + 721.0, + 1098.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1209.0, + 704.0, + 1261.0, + 704.0, + 1261.0, + 721.0, + 1209.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1263.0, + 706.0, + 1280.0, + 706.0, + 1280.0, + 718.0, + 1263.0, + 718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1317.0, + 706.0, + 1334.0, + 706.0, + 1334.0, + 718.0, + 1317.0, + 718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1336.0, + 704.0, + 1392.0, + 704.0, + 1392.0, + 721.0, + 1336.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 400.0, + 715.0, + 447.0, + 715.0, + 447.0, + 737.0, + 400.0, + 737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 837.0, + 715.0, + 886.0, + 715.0, + 886.0, + 737.0, + 837.0, + 737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1272.0, + 715.0, + 1321.0, + 715.0, + 1321.0, + 733.0, + 1272.0, + 733.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 335.0, + 733.0, + 512.0, + 733.0, + 512.0, + 759.0, + 335.0, + 759.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 733.0, + 715.0, + 733.0, + 715.0, + 761.0, + 570.0, + 761.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.0, + 735.0, + 881.0, + 735.0, + 881.0, + 761.0, + 793.0, + 761.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 738.0, + 928.0, + 738.0, + 928.0, + 755.0, + 882.0, + 755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1017.0, + 735.0, + 1141.0, + 735.0, + 1141.0, + 761.0, + 1017.0, + 761.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1228.0, + 735.0, + 1316.0, + 735.0, + 1316.0, + 761.0, + 1228.0, + 761.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1317.0, + 738.0, + 1367.0, + 738.0, + 1367.0, + 755.0, + 1317.0, + 755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 747.0, + 340.0, + 747.0, + 340.0, + 766.0, + 313.0, + 766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 769.0, + 340.0, + 769.0, + 340.0, + 810.0, + 313.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 534.0, + 776.0, + 563.0, + 776.0, + 563.0, + 794.0, + 534.0, + 794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 759.0, + 780.0, + 759.0, + 780.0, + 803.0, + 757.0, + 803.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 759.0, + 998.0, + 759.0, + 998.0, + 810.0, + 976.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 764.0, + 1217.0, + 764.0, + 1217.0, + 810.0, + 1195.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 813.0, + 340.0, + 813.0, + 340.0, + 834.0, + 313.0, + 834.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 522.0, + 813.0, + 561.0, + 813.0, + 561.0, + 854.0, + 522.0, + 854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 741.0, + 806.0, + 784.0, + 806.0, + 784.0, + 859.0, + 741.0, + 859.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1180.0, + 815.0, + 1220.0, + 815.0, + 1220.0, + 859.0, + 1180.0, + 859.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 857.0, + 340.0, + 857.0, + 340.0, + 878.0, + 313.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 539.0, + 847.0, + 560.0, + 847.0, + 560.0, + 868.0, + 539.0, + 868.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 758.0, + 859.0, + 777.0, + 859.0, + 777.0, + 873.0, + 758.0, + 873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 851.0, + 998.0, + 851.0, + 998.0, + 871.0, + 976.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 852.0, + 1215.0, + 852.0, + 1215.0, + 868.0, + 1198.0, + 868.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 868.0, + 461.0, + 868.0, + 461.0, + 890.0, + 373.0, + 890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 594.0, + 871.0, + 679.0, + 871.0, + 679.0, + 888.0, + 594.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 794.0, + 874.0, + 803.0, + 874.0, + 803.0, + 883.0, + 794.0, + 883.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 811.0, + 871.0, + 896.0, + 871.0, + 896.0, + 888.0, + 811.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1032.0, + 869.0, + 1117.0, + 869.0, + 1117.0, + 886.0, + 1032.0, + 886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 869.0, + 1334.0, + 869.0, + 1334.0, + 886.0, + 1248.0, + 886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 881.0, + 340.0, + 881.0, + 340.0, + 900.0, + 313.0, + 900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 886.0, + 369.0, + 886.0, + 369.0, + 900.0, + 352.0, + 900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 883.0, + 436.0, + 883.0, + 436.0, + 905.0, + 373.0, + 905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 541.0, + 886.0, + 558.0, + 886.0, + 558.0, + 901.0, + 541.0, + 901.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 883.0, + 655.0, + 883.0, + 655.0, + 905.0, + 592.0, + 905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 881.0, + 780.0, + 881.0, + 780.0, + 900.0, + 757.0, + 900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 888.0, + 806.0, + 888.0, + 806.0, + 900.0, + 789.0, + 900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 809.0, + 883.0, + 874.0, + 883.0, + 874.0, + 905.0, + 809.0, + 905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 883.0, + 998.0, + 883.0, + 998.0, + 901.0, + 976.0, + 901.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1030.0, + 883.0, + 1093.0, + 883.0, + 1093.0, + 905.0, + 1030.0, + 905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 881.0, + 1217.0, + 881.0, + 1217.0, + 901.0, + 1195.0, + 901.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 886.0, + 1246.0, + 886.0, + 1246.0, + 900.0, + 1229.0, + 900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 883.0, + 1312.0, + 883.0, + 1312.0, + 905.0, + 1248.0, + 905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 903.0, + 340.0, + 903.0, + 340.0, + 922.0, + 315.0, + 922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 901.0, + 368.0, + 901.0, + 368.0, + 913.0, + 354.0, + 913.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 896.0, + 432.0, + 896.0, + 432.0, + 918.0, + 373.0, + 918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 571.0, + 901.0, + 587.0, + 901.0, + 587.0, + 913.0, + 571.0, + 913.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 896.0, + 650.0, + 896.0, + 650.0, + 918.0, + 592.0, + 918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 758.0, + 905.0, + 780.0, + 905.0, + 780.0, + 924.0, + 758.0, + 924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 791.0, + 901.0, + 804.0, + 901.0, + 804.0, + 913.0, + 791.0, + 913.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 813.0, + 900.0, + 867.0, + 900.0, + 867.0, + 917.0, + 813.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 912.0, + 1215.0, + 912.0, + 1215.0, + 927.0, + 1198.0, + 927.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 901.0, + 1244.0, + 901.0, + 1244.0, + 913.0, + 1229.0, + 913.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 920.0, + 441.0, + 920.0, + 441.0, + 942.0, + 349.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 446.0, + 920.0, + 519.0, + 920.0, + 519.0, + 942.0, + 446.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 568.0, + 922.0, + 662.0, + 922.0, + 662.0, + 941.0, + 568.0, + 941.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 677.0, + 920.0, + 738.0, + 920.0, + 738.0, + 942.0, + 677.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 924.0, + 860.0, + 924.0, + 860.0, + 941.0, + 789.0, + 941.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 862.0, + 920.0, + 955.0, + 920.0, + 955.0, + 942.0, + 862.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 988.0, + 920.0, + 1175.0, + 920.0, + 1175.0, + 942.0, + 988.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1226.0, + 924.0, + 1280.0, + 924.0, + 1280.0, + 941.0, + 1226.0, + 941.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1283.0, + 920.0, + 1394.0, + 920.0, + 1394.0, + 942.0, + 1283.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 400.0, + 935.0, + 447.0, + 935.0, + 447.0, + 952.0, + 400.0, + 952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 619.0, + 935.0, + 667.0, + 935.0, + 667.0, + 952.0, + 619.0, + 952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 935.0, + 884.0, + 935.0, + 884.0, + 952.0, + 838.0, + 952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 935.0, + 1103.0, + 935.0, + 1103.0, + 952.0, + 1057.0, + 952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1277.0, + 935.0, + 1322.0, + 935.0, + 1322.0, + 952.0, + 1277.0, + 952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 963.0, + 340.0, + 963.0, + 340.0, + 983.0, + 315.0, + 983.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 424.0, + 958.0, + 510.0, + 958.0, + 510.0, + 975.0, + 424.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 969.0, + 561.0, + 969.0, + 561.0, + 988.0, + 532.0, + 988.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 658.0, + 952.0, + 715.0, + 952.0, + 715.0, + 979.0, + 658.0, + 979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1107.0, + 956.0, + 1142.0, + 956.0, + 1142.0, + 975.0, + 1107.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1316.0, + 954.0, + 1368.0, + 954.0, + 1368.0, + 976.0, + 1316.0, + 976.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 986.0, + 340.0, + 986.0, + 340.0, + 1007.0, + 313.0, + 1007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 985.0, + 779.0, + 985.0, + 779.0, + 1003.0, + 757.0, + 1003.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 983.0, + 998.0, + 983.0, + 998.0, + 1003.0, + 976.0, + 1003.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1197.0, + 983.0, + 1217.0, + 983.0, + 1217.0, + 1003.0, + 1197.0, + 1003.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1009.0, + 340.0, + 1009.0, + 340.0, + 1029.0, + 313.0, + 1029.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 1003.0, + 560.0, + 1003.0, + 560.0, + 1024.0, + 532.0, + 1024.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1010.0, + 779.0, + 1010.0, + 779.0, + 1031.0, + 757.0, + 1031.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1010.0, + 998.0, + 1010.0, + 998.0, + 1029.0, + 976.0, + 1029.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1007.0, + 1217.0, + 1007.0, + 1217.0, + 1026.0, + 1195.0, + 1026.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 1031.0, + 346.0, + 1031.0, + 346.0, + 1078.0, + 298.0, + 1078.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 967.0, + 1037.0, + 998.0, + 1037.0, + 998.0, + 1070.0, + 967.0, + 1070.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1197.0, + 1034.0, + 1214.0, + 1034.0, + 1214.0, + 1049.0, + 1197.0, + 1049.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 758.0, + 1065.0, + 780.0, + 1065.0, + 780.0, + 1085.0, + 758.0, + 1085.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1065.0, + 998.0, + 1065.0, + 998.0, + 1085.0, + 976.0, + 1085.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1076.0, + 340.0, + 1076.0, + 340.0, + 1141.0, + 313.0, + 1141.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1121.0, + 369.0, + 1121.0, + 369.0, + 1133.0, + 352.0, + 1133.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 1087.0, + 461.0, + 1087.0, + 461.0, + 1134.0, + 373.0, + 1134.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1076.0, + 560.0, + 1076.0, + 560.0, + 1133.0, + 538.0, + 1133.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 1119.0, + 587.0, + 1119.0, + 587.0, + 1133.0, + 570.0, + 1133.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 591.0, + 1087.0, + 680.0, + 1087.0, + 680.0, + 1138.0, + 591.0, + 1138.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1092.0, + 780.0, + 1092.0, + 780.0, + 1138.0, + 757.0, + 1138.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 1121.0, + 808.0, + 1121.0, + 808.0, + 1133.0, + 787.0, + 1133.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 809.0, + 1087.0, + 899.0, + 1087.0, + 899.0, + 1138.0, + 809.0, + 1138.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1092.0, + 1000.0, + 1092.0, + 1000.0, + 1138.0, + 976.0, + 1138.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 1090.0, + 1117.0, + 1090.0, + 1117.0, + 1138.0, + 1028.0, + 1138.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1082.0, + 1217.0, + 1082.0, + 1217.0, + 1126.0, + 1195.0, + 1126.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1245.0, + 1090.0, + 1334.0, + 1090.0, + 1334.0, + 1138.0, + 1245.0, + 1138.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1143.0, + 366.0, + 1143.0, + 366.0, + 1156.0, + 352.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1143.0, + 385.0, + 1143.0, + 385.0, + 1156.0, + 369.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 386.0, + 1143.0, + 403.0, + 1143.0, + 403.0, + 1156.0, + 386.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 407.0, + 1143.0, + 441.0, + 1143.0, + 441.0, + 1156.0, + 407.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 446.0, + 1143.0, + 459.0, + 1143.0, + 459.0, + 1155.0, + 446.0, + 1155.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 461.0, + 1143.0, + 478.0, + 1143.0, + 478.0, + 1158.0, + 461.0, + 1158.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 480.0, + 1141.0, + 517.0, + 1141.0, + 517.0, + 1158.0, + 480.0, + 1158.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 554.0, + 1143.0, + 566.0, + 1143.0, + 566.0, + 1156.0, + 554.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 1143.0, + 587.0, + 1143.0, + 587.0, + 1156.0, + 570.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 590.0, + 1143.0, + 622.0, + 1143.0, + 622.0, + 1156.0, + 590.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 624.0, + 1143.0, + 643.0, + 1143.0, + 643.0, + 1156.0, + 624.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 1144.0, + 660.0, + 1144.0, + 660.0, + 1156.0, + 645.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 680.0, + 1143.0, + 695.0, + 1143.0, + 695.0, + 1156.0, + 680.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 697.0, + 1141.0, + 736.0, + 1141.0, + 736.0, + 1158.0, + 697.0, + 1158.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 1143.0, + 804.0, + 1143.0, + 804.0, + 1158.0, + 789.0, + 1158.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 808.0, + 1143.0, + 825.0, + 1143.0, + 825.0, + 1156.0, + 808.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 826.0, + 1143.0, + 842.0, + 1143.0, + 842.0, + 1156.0, + 826.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 843.0, + 1143.0, + 860.0, + 1143.0, + 860.0, + 1156.0, + 843.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 862.0, + 1143.0, + 879.0, + 1143.0, + 879.0, + 1156.0, + 862.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 1143.0, + 915.0, + 1143.0, + 915.0, + 1156.0, + 882.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 916.0, + 1141.0, + 954.0, + 1141.0, + 954.0, + 1158.0, + 916.0, + 1158.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 991.0, + 1143.0, + 1005.0, + 1143.0, + 1005.0, + 1155.0, + 991.0, + 1155.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 1143.0, + 1023.0, + 1143.0, + 1023.0, + 1156.0, + 1006.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1025.0, + 1143.0, + 1042.0, + 1143.0, + 1042.0, + 1156.0, + 1025.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1045.0, + 1143.0, + 1059.0, + 1143.0, + 1059.0, + 1156.0, + 1045.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1081.0, + 1143.0, + 1096.0, + 1143.0, + 1096.0, + 1156.0, + 1081.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 1143.0, + 1119.0, + 1143.0, + 1119.0, + 1155.0, + 1100.0, + 1155.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1120.0, + 1141.0, + 1173.0, + 1141.0, + 1173.0, + 1160.0, + 1120.0, + 1160.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 1134.0, + 1215.0, + 1134.0, + 1215.0, + 1148.0, + 1198.0, + 1148.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1227.0, + 1143.0, + 1260.0, + 1143.0, + 1260.0, + 1156.0, + 1227.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1263.0, + 1143.0, + 1278.0, + 1143.0, + 1278.0, + 1156.0, + 1263.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1282.0, + 1143.0, + 1297.0, + 1143.0, + 1297.0, + 1156.0, + 1282.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1300.0, + 1143.0, + 1316.0, + 1143.0, + 1316.0, + 1156.0, + 1300.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1319.0, + 1139.0, + 1394.0, + 1139.0, + 1394.0, + 1161.0, + 1319.0, + 1161.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 1151.0, + 449.0, + 1151.0, + 449.0, + 1173.0, + 398.0, + 1173.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 617.0, + 1151.0, + 667.0, + 1151.0, + 667.0, + 1173.0, + 617.0, + 1173.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 837.0, + 1151.0, + 886.0, + 1151.0, + 886.0, + 1173.0, + 837.0, + 1173.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1056.0, + 1151.0, + 1105.0, + 1151.0, + 1105.0, + 1173.0, + 1056.0, + 1173.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1277.0, + 1151.0, + 1324.0, + 1151.0, + 1324.0, + 1173.0, + 1277.0, + 1173.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 421.0, + 1170.0, + 512.0, + 1170.0, + 512.0, + 1196.0, + 421.0, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 1175.0, + 928.0, + 1175.0, + 928.0, + 1192.0, + 882.0, + 1192.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1107.0, + 1175.0, + 1141.0, + 1175.0, + 1141.0, + 1192.0, + 1107.0, + 1192.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1316.0, + 1173.0, + 1370.0, + 1173.0, + 1370.0, + 1195.0, + 1316.0, + 1195.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 1182.0, + 340.0, + 1182.0, + 340.0, + 1202.0, + 315.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 1184.0, + 560.0, + 1184.0, + 560.0, + 1202.0, + 532.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1184.0, + 780.0, + 1184.0, + 780.0, + 1204.0, + 757.0, + 1204.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1187.0, + 1000.0, + 1187.0, + 1000.0, + 1207.0, + 976.0, + 1207.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1206.0, + 340.0, + 1206.0, + 340.0, + 1248.0, + 313.0, + 1248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 534.0, + 1221.0, + 560.0, + 1221.0, + 560.0, + 1240.0, + 534.0, + 1240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1214.0, + 779.0, + 1214.0, + 779.0, + 1234.0, + 757.0, + 1234.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1216.0, + 998.0, + 1216.0, + 998.0, + 1234.0, + 976.0, + 1234.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1209.0, + 1217.0, + 1209.0, + 1217.0, + 1228.0, + 1195.0, + 1228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 1250.0, + 340.0, + 1250.0, + 340.0, + 1270.0, + 315.0, + 1270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 522.0, + 1258.0, + 560.0, + 1258.0, + 560.0, + 1291.0, + 522.0, + 1291.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 741.0, + 1245.0, + 784.0, + 1245.0, + 784.0, + 1299.0, + 741.0, + 1299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 963.0, + 1243.0, + 1005.0, + 1243.0, + 1005.0, + 1297.0, + 963.0, + 1297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1185.0, + 1240.0, + 1217.0, + 1240.0, + 1217.0, + 1291.0, + 1185.0, + 1291.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1296.0, + 340.0, + 1296.0, + 340.0, + 1316.0, + 313.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1311.0, + 373.0, + 1311.0, + 373.0, + 1323.0, + 352.0, + 1323.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 1308.0, + 459.0, + 1308.0, + 459.0, + 1325.0, + 374.0, + 1325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1294.0, + 560.0, + 1294.0, + 560.0, + 1313.0, + 538.0, + 1313.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 594.0, + 1308.0, + 679.0, + 1308.0, + 679.0, + 1325.0, + 594.0, + 1325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1306.0, + 779.0, + 1306.0, + 779.0, + 1325.0, + 757.0, + 1325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 794.0, + 1308.0, + 894.0, + 1308.0, + 894.0, + 1325.0, + 794.0, + 1325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1301.0, + 1000.0, + 1301.0, + 1000.0, + 1321.0, + 976.0, + 1321.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1030.0, + 1308.0, + 1117.0, + 1308.0, + 1117.0, + 1325.0, + 1030.0, + 1325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1197.0, + 1296.0, + 1214.0, + 1296.0, + 1214.0, + 1311.0, + 1197.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 1308.0, + 1336.0, + 1308.0, + 1336.0, + 1325.0, + 1248.0, + 1325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1319.0, + 342.0, + 1319.0, + 342.0, + 1360.0, + 313.0, + 1360.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 1319.0, + 436.0, + 1319.0, + 436.0, + 1353.0, + 373.0, + 1353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 541.0, + 1333.0, + 558.0, + 1333.0, + 558.0, + 1348.0, + 541.0, + 1348.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 1319.0, + 655.0, + 1319.0, + 655.0, + 1353.0, + 592.0, + 1353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1336.0, + 780.0, + 1336.0, + 780.0, + 1355.0, + 757.0, + 1355.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 809.0, + 1319.0, + 874.0, + 1319.0, + 874.0, + 1353.0, + 809.0, + 1353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1330.0, + 998.0, + 1330.0, + 998.0, + 1350.0, + 976.0, + 1350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 1319.0, + 1093.0, + 1319.0, + 1093.0, + 1357.0, + 1028.0, + 1357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1323.0, + 1217.0, + 1323.0, + 1217.0, + 1342.0, + 1195.0, + 1342.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1246.0, + 1319.0, + 1312.0, + 1319.0, + 1312.0, + 1357.0, + 1246.0, + 1357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 335.0, + 1362.0, + 347.0, + 1362.0, + 347.0, + 1376.0, + 335.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1360.0, + 366.0, + 1360.0, + 366.0, + 1374.0, + 352.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1362.0, + 385.0, + 1362.0, + 385.0, + 1376.0, + 369.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 388.0, + 1362.0, + 403.0, + 1362.0, + 403.0, + 1374.0, + 388.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 407.0, + 1362.0, + 422.0, + 1362.0, + 422.0, + 1374.0, + 407.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 425.0, + 1362.0, + 441.0, + 1362.0, + 441.0, + 1374.0, + 425.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 444.0, + 1360.0, + 459.0, + 1360.0, + 459.0, + 1374.0, + 444.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 461.0, + 1360.0, + 478.0, + 1360.0, + 478.0, + 1376.0, + 461.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 480.0, + 1359.0, + 517.0, + 1359.0, + 517.0, + 1377.0, + 480.0, + 1377.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 556.0, + 1364.0, + 565.0, + 1364.0, + 565.0, + 1372.0, + 556.0, + 1372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 1360.0, + 587.0, + 1360.0, + 587.0, + 1376.0, + 570.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 590.0, + 1362.0, + 621.0, + 1362.0, + 621.0, + 1374.0, + 590.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 624.0, + 1362.0, + 641.0, + 1362.0, + 641.0, + 1374.0, + 624.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 1362.0, + 660.0, + 1362.0, + 660.0, + 1374.0, + 645.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 682.0, + 1360.0, + 695.0, + 1360.0, + 695.0, + 1376.0, + 682.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 697.0, + 1359.0, + 735.0, + 1359.0, + 735.0, + 1377.0, + 697.0, + 1377.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 1360.0, + 804.0, + 1360.0, + 804.0, + 1376.0, + 789.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 808.0, + 1362.0, + 825.0, + 1362.0, + 825.0, + 1376.0, + 808.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 826.0, + 1362.0, + 840.0, + 1362.0, + 840.0, + 1374.0, + 826.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 845.0, + 1360.0, + 859.0, + 1360.0, + 859.0, + 1374.0, + 845.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 864.0, + 1362.0, + 879.0, + 1362.0, + 879.0, + 1376.0, + 864.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 1360.0, + 896.0, + 1360.0, + 896.0, + 1374.0, + 882.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 1362.0, + 915.0, + 1362.0, + 915.0, + 1376.0, + 899.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 916.0, + 1359.0, + 954.0, + 1359.0, + 954.0, + 1377.0, + 916.0, + 1377.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 993.0, + 1364.0, + 1001.0, + 1364.0, + 1001.0, + 1372.0, + 993.0, + 1372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1008.0, + 1360.0, + 1022.0, + 1360.0, + 1022.0, + 1374.0, + 1008.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1027.0, + 1362.0, + 1042.0, + 1362.0, + 1042.0, + 1376.0, + 1027.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1045.0, + 1362.0, + 1059.0, + 1362.0, + 1059.0, + 1374.0, + 1045.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1062.0, + 1362.0, + 1076.0, + 1362.0, + 1076.0, + 1374.0, + 1062.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1083.0, + 1360.0, + 1096.0, + 1360.0, + 1096.0, + 1374.0, + 1083.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 1360.0, + 1117.0, + 1360.0, + 1117.0, + 1374.0, + 1100.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1119.0, + 1359.0, + 1173.0, + 1359.0, + 1173.0, + 1377.0, + 1119.0, + 1377.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 1352.0, + 1215.0, + 1352.0, + 1215.0, + 1367.0, + 1198.0, + 1367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1227.0, + 1360.0, + 1243.0, + 1360.0, + 1243.0, + 1374.0, + 1227.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1263.0, + 1362.0, + 1278.0, + 1362.0, + 1278.0, + 1374.0, + 1263.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1283.0, + 1362.0, + 1297.0, + 1362.0, + 1297.0, + 1374.0, + 1283.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1300.0, + 1362.0, + 1316.0, + 1362.0, + 1316.0, + 1374.0, + 1300.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1319.0, + 1357.0, + 1394.0, + 1357.0, + 1394.0, + 1379.0, + 1319.0, + 1379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 400.0, + 1370.0, + 447.0, + 1370.0, + 447.0, + 1392.0, + 400.0, + 1392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 619.0, + 1370.0, + 667.0, + 1370.0, + 667.0, + 1392.0, + 619.0, + 1392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 1370.0, + 886.0, + 1370.0, + 886.0, + 1392.0, + 838.0, + 1392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1056.0, + 1370.0, + 1105.0, + 1370.0, + 1105.0, + 1392.0, + 1056.0, + 1392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 420.0, + 1394.0, + 509.0, + 1394.0, + 509.0, + 1411.0, + 420.0, + 1411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 652.0, + 1389.0, + 711.0, + 1389.0, + 711.0, + 1415.0, + 652.0, + 1415.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 879.0, + 1394.0, + 927.0, + 1394.0, + 927.0, + 1411.0, + 879.0, + 1411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1103.0, + 1392.0, + 1139.0, + 1392.0, + 1139.0, + 1411.0, + 1103.0, + 1411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1316.0, + 1394.0, + 1367.0, + 1394.0, + 1367.0, + 1411.0, + 1316.0, + 1411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 1408.0, + 340.0, + 1408.0, + 340.0, + 1426.0, + 315.0, + 1426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 534.0, + 1423.0, + 560.0, + 1423.0, + 560.0, + 1442.0, + 534.0, + 1442.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 689.0, + 1420.0, + 719.0, + 1420.0, + 719.0, + 1432.0, + 689.0, + 1432.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1415.0, + 779.0, + 1415.0, + 779.0, + 1433.0, + 757.0, + 1433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 1421.0, + 995.0, + 1421.0, + 995.0, + 1437.0, + 979.0, + 1437.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1418.0, + 1217.0, + 1418.0, + 1217.0, + 1438.0, + 1195.0, + 1438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1432.0, + 340.0, + 1432.0, + 340.0, + 1452.0, + 313.0, + 1452.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1449.0, + 780.0, + 1449.0, + 780.0, + 1467.0, + 757.0, + 1467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1197.0, + 1450.0, + 1214.0, + 1450.0, + 1214.0, + 1464.0, + 1197.0, + 1464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1457.0, + 340.0, + 1457.0, + 340.0, + 1477.0, + 313.0, + 1477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1464.0, + 563.0, + 1464.0, + 563.0, + 1483.0, + 538.0, + 1483.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 1462.0, + 996.0, + 1462.0, + 996.0, + 1477.0, + 979.0, + 1477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 1477.0, + 346.0, + 1477.0, + 346.0, + 1515.0, + 298.0, + 1515.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 522.0, + 1481.0, + 539.0, + 1481.0, + 539.0, + 1510.0, + 522.0, + 1510.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 741.0, + 1476.0, + 786.0, + 1476.0, + 786.0, + 1515.0, + 741.0, + 1515.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 958.0, + 1478.0, + 982.0, + 1478.0, + 982.0, + 1509.0, + 958.0, + 1509.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 1510.0, + 340.0, + 1510.0, + 340.0, + 1528.0, + 315.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1506.0, + 561.0, + 1506.0, + 561.0, + 1525.0, + 538.0, + 1525.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 1503.0, + 995.0, + 1503.0, + 995.0, + 1520.0, + 979.0, + 1520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1506.0, + 1215.0, + 1506.0, + 1215.0, + 1527.0, + 1195.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1518.0, + 780.0, + 1518.0, + 780.0, + 1537.0, + 757.0, + 1537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 1527.0, + 461.0, + 1527.0, + 461.0, + 1544.0, + 373.0, + 1544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 594.0, + 1527.0, + 680.0, + 1527.0, + 680.0, + 1544.0, + 594.0, + 1544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 813.0, + 1527.0, + 898.0, + 1527.0, + 898.0, + 1544.0, + 813.0, + 1544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1030.0, + 1527.0, + 1117.0, + 1527.0, + 1117.0, + 1544.0, + 1030.0, + 1544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1253.0, + 1527.0, + 1336.0, + 1527.0, + 1336.0, + 1544.0, + 1253.0, + 1544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1534.0, + 342.0, + 1534.0, + 342.0, + 1579.0, + 313.0, + 1579.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1557.0, + 368.0, + 1557.0, + 368.0, + 1569.0, + 354.0, + 1569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 1539.0, + 437.0, + 1539.0, + 437.0, + 1574.0, + 373.0, + 1574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 541.0, + 1550.0, + 558.0, + 1550.0, + 558.0, + 1566.0, + 541.0, + 1566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 1539.0, + 655.0, + 1539.0, + 655.0, + 1574.0, + 592.0, + 1574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1552.0, + 779.0, + 1552.0, + 779.0, + 1571.0, + 757.0, + 1571.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 809.0, + 1539.0, + 874.0, + 1539.0, + 874.0, + 1574.0, + 809.0, + 1574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1542.0, + 998.0, + 1542.0, + 998.0, + 1562.0, + 976.0, + 1562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 1540.0, + 1090.0, + 1540.0, + 1090.0, + 1574.0, + 1028.0, + 1574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1539.0, + 1217.0, + 1539.0, + 1217.0, + 1557.0, + 1195.0, + 1557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 1539.0, + 1312.0, + 1539.0, + 1312.0, + 1574.0, + 1248.0, + 1574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1579.0, + 366.0, + 1579.0, + 366.0, + 1593.0, + 352.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1579.0, + 385.0, + 1579.0, + 385.0, + 1595.0, + 369.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 388.0, + 1579.0, + 403.0, + 1579.0, + 403.0, + 1593.0, + 388.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 408.0, + 1579.0, + 424.0, + 1579.0, + 424.0, + 1593.0, + 408.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 427.0, + 1579.0, + 439.0, + 1579.0, + 439.0, + 1593.0, + 427.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 447.0, + 1581.0, + 456.0, + 1581.0, + 456.0, + 1590.0, + 447.0, + 1590.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 463.0, + 1579.0, + 480.0, + 1579.0, + 480.0, + 1595.0, + 463.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 481.0, + 1578.0, + 517.0, + 1578.0, + 517.0, + 1596.0, + 481.0, + 1596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 1579.0, + 587.0, + 1579.0, + 587.0, + 1593.0, + 570.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 590.0, + 1579.0, + 604.0, + 1579.0, + 604.0, + 1593.0, + 590.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 607.0, + 1579.0, + 621.0, + 1579.0, + 621.0, + 1595.0, + 607.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 626.0, + 1579.0, + 641.0, + 1579.0, + 641.0, + 1593.0, + 626.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 1579.0, + 658.0, + 1579.0, + 658.0, + 1593.0, + 645.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 667.0, + 1581.0, + 675.0, + 1581.0, + 675.0, + 1590.0, + 667.0, + 1590.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 682.0, + 1579.0, + 695.0, + 1579.0, + 695.0, + 1593.0, + 682.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 697.0, + 1578.0, + 735.0, + 1578.0, + 735.0, + 1595.0, + 697.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 791.0, + 1579.0, + 804.0, + 1579.0, + 804.0, + 1595.0, + 791.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 808.0, + 1579.0, + 825.0, + 1579.0, + 825.0, + 1593.0, + 808.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 826.0, + 1579.0, + 840.0, + 1579.0, + 840.0, + 1595.0, + 826.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 845.0, + 1579.0, + 859.0, + 1579.0, + 859.0, + 1593.0, + 845.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 864.0, + 1579.0, + 879.0, + 1579.0, + 879.0, + 1593.0, + 864.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 1579.0, + 896.0, + 1579.0, + 896.0, + 1591.0, + 882.0, + 1591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 1579.0, + 915.0, + 1579.0, + 915.0, + 1593.0, + 899.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 916.0, + 1578.0, + 954.0, + 1578.0, + 954.0, + 1595.0, + 916.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1008.0, + 1579.0, + 1023.0, + 1579.0, + 1023.0, + 1595.0, + 1008.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1025.0, + 1579.0, + 1042.0, + 1579.0, + 1042.0, + 1593.0, + 1025.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1045.0, + 1579.0, + 1059.0, + 1579.0, + 1059.0, + 1593.0, + 1045.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1083.0, + 1579.0, + 1096.0, + 1579.0, + 1096.0, + 1593.0, + 1083.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 1578.0, + 1117.0, + 1578.0, + 1117.0, + 1593.0, + 1100.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1120.0, + 1578.0, + 1173.0, + 1578.0, + 1173.0, + 1596.0, + 1120.0, + 1596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1197.0, + 1569.0, + 1219.0, + 1569.0, + 1219.0, + 1588.0, + 1197.0, + 1588.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 1579.0, + 1243.0, + 1579.0, + 1243.0, + 1593.0, + 1229.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1244.0, + 1579.0, + 1260.0, + 1579.0, + 1260.0, + 1595.0, + 1244.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1265.0, + 1579.0, + 1278.0, + 1579.0, + 1278.0, + 1593.0, + 1265.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1302.0, + 1579.0, + 1314.0, + 1579.0, + 1314.0, + 1593.0, + 1302.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1319.0, + 1576.0, + 1392.0, + 1576.0, + 1392.0, + 1598.0, + 1319.0, + 1598.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 400.0, + 1588.0, + 447.0, + 1588.0, + 447.0, + 1610.0, + 400.0, + 1610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 617.0, + 1588.0, + 667.0, + 1588.0, + 667.0, + 1610.0, + 617.0, + 1610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 1588.0, + 886.0, + 1588.0, + 886.0, + 1610.0, + 838.0, + 1610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 1591.0, + 1103.0, + 1591.0, + 1103.0, + 1608.0, + 1057.0, + 1608.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1275.0, + 1588.0, + 1324.0, + 1588.0, + 1324.0, + 1610.0, + 1275.0, + 1610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 422.0, + 1613.0, + 510.0, + 1613.0, + 510.0, + 1630.0, + 422.0, + 1630.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 658.0, + 1608.0, + 713.0, + 1608.0, + 713.0, + 1634.0, + 658.0, + 1634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 750.0, + 1618.0, + 779.0, + 1618.0, + 779.0, + 1639.0, + 750.0, + 1639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1107.0, + 1612.0, + 1141.0, + 1612.0, + 1141.0, + 1630.0, + 1107.0, + 1630.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1316.0, + 1610.0, + 1368.0, + 1610.0, + 1368.0, + 1632.0, + 1316.0, + 1632.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1627.0, + 340.0, + 1627.0, + 340.0, + 1647.0, + 313.0, + 1647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 534.0, + 1627.0, + 558.0, + 1627.0, + 558.0, + 1646.0, + 534.0, + 1646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 706.0, + 1635.0, + 721.0, + 1635.0, + 721.0, + 1649.0, + 706.0, + 1649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 746.0, + 1639.0, + 780.0, + 1639.0, + 780.0, + 1663.0, + 746.0, + 1663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1629.0, + 996.0, + 1629.0, + 996.0, + 1649.0, + 976.0, + 1649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1642.0, + 1217.0, + 1642.0, + 1217.0, + 1663.0, + 1195.0, + 1663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1652.0, + 340.0, + 1652.0, + 340.0, + 1673.0, + 313.0, + 1673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1651.0, + 561.0, + 1651.0, + 561.0, + 1671.0, + 538.0, + 1671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1652.0, + 998.0, + 1652.0, + 998.0, + 1673.0, + 976.0, + 1673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1678.0, + 342.0, + 1678.0, + 342.0, + 1698.0, + 313.0, + 1698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1676.0, + 561.0, + 1676.0, + 561.0, + 1697.0, + 538.0, + 1697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 746.0, + 1661.0, + 780.0, + 1661.0, + 780.0, + 1703.0, + 746.0, + 1703.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 1680.0, + 996.0, + 1680.0, + 996.0, + 1695.0, + 979.0, + 1695.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1669.0, + 1215.0, + 1669.0, + 1215.0, + 1690.0, + 1195.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 1695.0, + 346.0, + 1695.0, + 346.0, + 1734.0, + 298.0, + 1734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 522.0, + 1700.0, + 560.0, + 1700.0, + 560.0, + 1729.0, + 522.0, + 1729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 733.0, + 1695.0, + 784.0, + 1695.0, + 784.0, + 1732.0, + 733.0, + 1732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 961.0, + 1695.0, + 1003.0, + 1695.0, + 1003.0, + 1736.0, + 961.0, + 1736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1181.0, + 1695.0, + 1219.0, + 1695.0, + 1219.0, + 1732.0, + 1181.0, + 1732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 1729.0, + 340.0, + 1729.0, + 340.0, + 1749.0, + 318.0, + 1749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1725.0, + 560.0, + 1725.0, + 560.0, + 1746.0, + 538.0, + 1746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 746.0, + 1725.0, + 780.0, + 1725.0, + 780.0, + 1749.0, + 746.0, + 1749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 1729.0, + 996.0, + 1729.0, + 996.0, + 1744.0, + 979.0, + 1744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 1729.0, + 1214.0, + 1729.0, + 1214.0, + 1742.0, + 1198.0, + 1742.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 1756.0, + 340.0, + 1756.0, + 340.0, + 1775.0, + 318.0, + 1775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 1742.0, + 461.0, + 1742.0, + 461.0, + 1776.0, + 373.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1751.0, + 560.0, + 1751.0, + 560.0, + 1771.0, + 538.0, + 1771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 594.0, + 1742.0, + 680.0, + 1742.0, + 680.0, + 1776.0, + 594.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 748.0, + 1751.0, + 780.0, + 1751.0, + 780.0, + 1770.0, + 748.0, + 1770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 809.0, + 1745.0, + 899.0, + 1745.0, + 899.0, + 1776.0, + 809.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1751.0, + 998.0, + 1751.0, + 998.0, + 1771.0, + 976.0, + 1771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1026.0, + 1745.0, + 1116.0, + 1745.0, + 1116.0, + 1776.0, + 1026.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1754.0, + 1217.0, + 1754.0, + 1217.0, + 1775.0, + 1195.0, + 1775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1249.0, + 1746.0, + 1334.0, + 1746.0, + 1334.0, + 1776.0, + 1249.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1783.0, + 339.0, + 1783.0, + 339.0, + 1797.0, + 322.0, + 1797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1775.0, + 369.0, + 1775.0, + 369.0, + 1788.0, + 352.0, + 1788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 1771.0, + 430.0, + 1771.0, + 430.0, + 1793.0, + 373.0, + 1793.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 541.0, + 1778.0, + 556.0, + 1778.0, + 556.0, + 1792.0, + 541.0, + 1792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 1771.0, + 650.0, + 1771.0, + 650.0, + 1793.0, + 592.0, + 1793.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 750.0, + 1773.0, + 780.0, + 1773.0, + 780.0, + 1792.0, + 750.0, + 1792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 1775.0, + 808.0, + 1775.0, + 808.0, + 1788.0, + 789.0, + 1788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 809.0, + 1771.0, + 869.0, + 1771.0, + 869.0, + 1793.0, + 809.0, + 1793.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 1776.0, + 996.0, + 1776.0, + 996.0, + 1792.0, + 979.0, + 1792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1011.0, + 1776.0, + 1027.0, + 1776.0, + 1027.0, + 1788.0, + 1011.0, + 1788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 1771.0, + 1088.0, + 1771.0, + 1088.0, + 1793.0, + 1028.0, + 1793.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1783.0, + 1217.0, + 1783.0, + 1217.0, + 1802.0, + 1195.0, + 1802.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 1776.0, + 1246.0, + 1776.0, + 1246.0, + 1788.0, + 1229.0, + 1788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 1771.0, + 1307.0, + 1771.0, + 1307.0, + 1793.0, + 1248.0, + 1793.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1799.0, + 366.0, + 1799.0, + 366.0, + 1812.0, + 352.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1799.0, + 385.0, + 1799.0, + 385.0, + 1812.0, + 368.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 386.0, + 1799.0, + 403.0, + 1799.0, + 403.0, + 1812.0, + 386.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 407.0, + 1799.0, + 441.0, + 1799.0, + 441.0, + 1812.0, + 407.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 444.0, + 1797.0, + 517.0, + 1797.0, + 517.0, + 1814.0, + 444.0, + 1814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 1799.0, + 587.0, + 1799.0, + 587.0, + 1812.0, + 570.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 590.0, + 1799.0, + 604.0, + 1799.0, + 604.0, + 1812.0, + 590.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 605.0, + 1799.0, + 622.0, + 1799.0, + 622.0, + 1812.0, + 605.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 624.0, + 1799.0, + 660.0, + 1799.0, + 660.0, + 1812.0, + 624.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 663.0, + 1793.0, + 736.0, + 1793.0, + 736.0, + 1816.0, + 663.0, + 1816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 1799.0, + 804.0, + 1799.0, + 804.0, + 1812.0, + 789.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 806.0, + 1797.0, + 860.0, + 1797.0, + 860.0, + 1814.0, + 806.0, + 1814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 862.0, + 1799.0, + 879.0, + 1799.0, + 879.0, + 1812.0, + 862.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 896.0, + 1797.0, + 954.0, + 1797.0, + 954.0, + 1814.0, + 896.0, + 1814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1008.0, + 1799.0, + 1023.0, + 1799.0, + 1023.0, + 1812.0, + 1008.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1025.0, + 1799.0, + 1061.0, + 1799.0, + 1061.0, + 1812.0, + 1025.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1062.0, + 1799.0, + 1078.0, + 1799.0, + 1078.0, + 1812.0, + 1062.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1081.0, + 1799.0, + 1098.0, + 1799.0, + 1098.0, + 1812.0, + 1081.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 1793.0, + 1175.0, + 1793.0, + 1175.0, + 1816.0, + 1100.0, + 1816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1212.0, + 1800.0, + 1220.0, + 1800.0, + 1220.0, + 1809.0, + 1212.0, + 1809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1224.0, + 1793.0, + 1392.0, + 1793.0, + 1392.0, + 1816.0, + 1224.0, + 1816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 1807.0, + 447.0, + 1807.0, + 447.0, + 1829.0, + 398.0, + 1829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 837.0, + 1807.0, + 886.0, + 1807.0, + 886.0, + 1829.0, + 837.0, + 1829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1275.0, + 1807.0, + 1324.0, + 1807.0, + 1324.0, + 1829.0, + 1275.0, + 1829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.25, + 299.0, + 424.25, + 299.0, + 424.25, + 323.0, + 331.25, + 323.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 569.0, + 297.5, + 660.0, + 297.5, + 660.0, + 325.0, + 569.0, + 325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 790.0, + 297.5, + 883.0, + 297.5, + 883.0, + 325.0, + 790.0, + 325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1014.25, + 299.0, + 1107.25, + 299.0, + 1107.25, + 323.0, + 1014.25, + 323.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1227.0, + 297.5, + 1318.0, + 297.5, + 1318.0, + 325.0, + 1227.0, + 325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 754.0, + 393.0, + 766.0, + 393.0, + 766.0, + 418.0, + 754.0, + 418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 518.0, + 421.0, + 518.0, + 421.0, + 540.5, + 333.0, + 540.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.75, + 518.0, + 658.75, + 518.0, + 658.75, + 540.5, + 570.75, + 540.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 658.75, + 517.5, + 711.75, + 517.5, + 711.75, + 538.0, + 658.75, + 538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.25, + 518.0, + 881.25, + 518.0, + 881.25, + 540.5, + 793.25, + 540.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1017.75, + 518.0, + 1103.75, + 518.0, + 1103.75, + 540.5, + 1017.75, + 540.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1228.0, + 518.0, + 1315.0, + 518.0, + 1315.0, + 540.5, + 1228.0, + 540.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1277.75, + 704.5, + 1317.75, + 704.5, + 1317.75, + 720.0, + 1277.75, + 720.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 618.0, + 714.5, + 666.0, + 714.5, + 666.0, + 733.5, + 618.0, + 733.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1055.0, + 714.5, + 1104.0, + 714.5, + 1104.0, + 733.5, + 1055.0, + 733.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 300.0, + 821.0, + 342.0, + 821.0, + 342.0, + 859.5, + 300.0, + 859.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 949.0, + 823.5, + 1007.0, + 823.5, + 1007.0, + 846.5, + 949.0, + 846.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 896.5, + 1087.0, + 896.5, + 1087.0, + 918.5, + 1028.0, + 918.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1247.0, + 896.5, + 1306.0, + 896.5, + 1306.0, + 918.5, + 1247.0, + 918.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 335.25, + 954.0, + 420.25, + 954.0, + 420.25, + 978.0, + 335.25, + 978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 571.0, + 954.0, + 657.0, + 954.0, + 657.0, + 978.0, + 571.0, + 978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.0, + 955.0, + 880.0, + 955.0, + 880.0, + 977.0, + 793.0, + 977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 956.0, + 927.0, + 956.0, + 927.0, + 974.5, + 882.0, + 974.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1018.25, + 954.0, + 1103.25, + 954.0, + 1103.25, + 978.0, + 1018.25, + 978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 952.5, + 1314.0, + 952.5, + 1314.0, + 978.0, + 1229.0, + 978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 512.0, + 1035.5, + 570.0, + 1035.5, + 570.0, + 1074.0, + 512.0, + 1074.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1192.0, + 1051.5, + 1204.0, + 1051.5, + 1204.0, + 1072.5, + 1192.0, + 1072.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 729.0, + 1040.0, + 791.0, + 1040.0, + 791.0, + 1065.5, + 729.0, + 1065.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1059.0, + 1143.0, + 1080.0, + 1143.0, + 1080.0, + 1155.0, + 1059.0, + 1155.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 1172.5, + 419.0, + 1172.5, + 419.0, + 1196.0, + 333.0, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.75, + 1174.0, + 658.75, + 1174.0, + 658.75, + 1196.0, + 570.75, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 658.0, + 1171.5, + 713.0, + 1171.5, + 713.0, + 1193.5, + 658.0, + 1193.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.25, + 1174.0, + 881.25, + 1174.0, + 881.25, + 1196.0, + 793.25, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1017.75, + 1174.0, + 1105.75, + 1174.0, + 1105.75, + 1196.0, + 1017.75, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1228.0, + 1174.0, + 1315.0, + 1174.0, + 1315.0, + 1196.0, + 1228.0, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 300.0, + 1257.5, + 342.0, + 1257.5, + 342.0, + 1296.5, + 300.0, + 1296.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1240.0, + 1361.5, + 1265.0, + 1361.5, + 1265.0, + 1373.5, + 1240.0, + 1373.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1272.0, + 1368.5, + 1325.0, + 1368.5, + 1325.0, + 1390.5, + 1272.0, + 1390.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 336.25, + 1392.0, + 419.25, + 1392.0, + 419.25, + 1414.0, + 336.25, + 1414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 573.0, + 1390.0, + 657.0, + 1390.0, + 657.0, + 1414.0, + 573.0, + 1414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 795.0, + 1392.0, + 878.0, + 1392.0, + 878.0, + 1414.0, + 795.0, + 1414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1019.25, + 1392.0, + 1102.25, + 1392.0, + 1102.25, + 1415.5, + 1019.25, + 1415.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.75, + 1392.0, + 1312.75, + 1392.0, + 1312.75, + 1414.0, + 1229.75, + 1414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1169.0, + 1477.0, + 1226.0, + 1477.0, + 1226.0, + 1503.0, + 1169.0, + 1503.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 335.0, + 1609.0, + 421.0, + 1609.0, + 421.0, + 1633.0, + 335.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 569.0, + 1609.0, + 660.0, + 1609.0, + 660.0, + 1636.0, + 569.0, + 1636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.0, + 1611.0, + 880.0, + 1611.0, + 880.0, + 1633.0, + 793.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 882.0, + 1611.5, + 925.0, + 1611.5, + 925.0, + 1632.0, + 882.0, + 1632.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1018.0, + 1610.0, + 1102.0, + 1610.0, + 1102.0, + 1634.0, + 1018.0, + 1634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1227.0, + 1608.5, + 1316.0, + 1608.5, + 1316.0, + 1635.5, + 1227.0, + 1635.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 617.0, + 1807.0, + 666.0, + 1807.0, + 666.0, + 1827.5, + 617.0, + 1827.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1053.0, + 1805.5, + 1104.0, + 1805.5, + 1104.0, + 1827.5, + 1053.0, + 1827.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1835.0, + 1404.0, + 1835.0, + 1404.0, + 1868.0, + 294.0, + 1868.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1863.0, + 1107.0, + 1863.0, + 1107.0, + 1906.0, + 292.0, + 1906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2095.0, + 832.0, + 2095.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 17, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 754, + 1405, + 754, + 1405, + 1089, + 297, + 1089 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 297, + 1495, + 1406, + 1495, + 1406, + 1801, + 297, + 1801 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 1200, + 1405, + 1200, + 1405, + 1384, + 297, + 1384 + ], + "score": 0.981 + }, + { + "category_id": 5, + "poly": [ + 519, + 266, + 1179, + 266, + 1179, + 482, + 519, + 482 + ], + "score": 0.978, + "html": "
TransferCOCO CaptionVQAv2
linearbestlinearbest
FlanT5base → OPT350M110.8136.540.444.2
FlanT5xL →OPT2.7B132.1139.350.450.3
OPT350M→FlanT5base1.1122.134.049.9
OPT2.7B→FlanT5xL106.5133.251.353.5
" + }, + { + "category_id": 1, + "poly": [ + 299, + 601, + 1400, + 601, + 1400, + 665, + 299, + 665 + ], + "score": 0.953 + }, + { + "category_id": 6, + "poly": [ + 296, + 194, + 1397, + 194, + 1397, + 255, + 296, + 255 + ], + "score": 0.936 + }, + { + "category_id": 0, + "poly": [ + 296, + 1131, + 983, + 1131, + 983, + 1170, + 296, + 1170 + ], + "score": 0.935 + }, + { + "category_id": 0, + "poly": [ + 298, + 531, + 750, + 531, + 750, + 571, + 298, + 571 + ], + "score": 0.915 + }, + { + "category_id": 0, + "poly": [ + 302, + 1426, + 825, + 1426, + 825, + 1464, + 302, + 1464 + ], + "score": 0.913 + }, + { + "category_id": 0, + "poly": [ + 299, + 697, + 1115, + 697, + 1115, + 732, + 299, + 732 + ], + "score": 0.905 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.855 + }, + { + "category_id": 13, + "poly": [ + 294, + 999, + 428, + 999, + 428, + 1029, + 294, + 1029 + ], + "score": 0.88, + "latex": " \\mathrm { O P T } _ { 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 1016, + 908, + 1111, + 908, + 1111, + 939, + 1016, + 939 + ], + "score": 0.88, + "latex": "\\mathrm { O P T } _ { 2 . 7 \\mathrm { B } }" + }, + { + "category_id": 13, + "poly": [ + 724, + 877, + 832, + 877, + 832, + 908, + 724, + 908 + ], + "score": 0.72, + "latex": "( \\mathrm { O P T } _ { 3 5 0 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 1245, + 938, + 1404, + 938, + 1404, + 968, + 1245, + 968 + ], + "score": 0.7, + "latex": "\\mathrm { F l a n T 5 _ { b a s e } }" + }, + { + "category_id": 13, + "poly": [ + 297, + 968, + 402, + 968, + 402, + 998, + 297, + 998 + ], + "score": 0.7, + "latex": "\\mathrm { O P T } _ { 3 5 0 \\mathrm { M } }" + }, + { + "category_id": 13, + "poly": [ + 888, + 877, + 1009, + 877, + 1009, + 908, + 888, + 908 + ], + "score": 0.46, + "latex": "{ \\mathrm { F l a n T } } 5 _ { \\mathrm { b a s e } }" + }, + { + "category_id": 13, + "poly": [ + 1136, + 601, + 1211, + 601, + 1211, + 634, + 1136, + 634 + ], + "score": 0.45, + "latex": "( c f . \\ \\ S 5 )" + }, + { + "category_id": 13, + "poly": [ + 1359, + 970, + 1402, + 970, + 1402, + 999, + 1359, + 999 + ], + "score": 0.43, + "latex": "5 _ { \\mathrm { X L } }" + }, + { + "category_id": 13, + "poly": [ + 1236, + 908, + 1277, + 908, + 1277, + 938, + 1236, + 938 + ], + "score": 0.37, + "latex": "5 _ { \\mathrm { X L } }" + }, + { + "category_id": 13, + "poly": [ + 966, + 878, + 1009, + 878, + 1009, + 908, + 966, + 908 + ], + "score": 0.34, + "latex": "\\dot { \\mathfrak { b } } _ { \\mathrm { a s e } }" + }, + { + "category_id": 13, + "poly": [ + 1166, + 907, + 1277, + 907, + 1277, + 939, + 1166, + 939 + ], + "score": 0.33, + "latex": "\\mathrm { F l a n T 5 } _ { \\mathrm { X L } }" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 192.0, + 1403.0, + 192.0, + 1403.0, + 231.0, + 292.0, + 231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 222.0, + 519.0, + 222.0, + 519.0, + 257.0, + 295.0, + 257.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1127.0, + 987.0, + 1127.0, + 987.0, + 1178.0, + 291.0, + 1178.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 528.0, + 752.0, + 528.0, + 752.0, + 580.0, + 291.0, + 580.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1426.0, + 830.0, + 1426.0, + 830.0, + 1470.0, + 294.0, + 1470.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 693.0, + 1116.0, + 693.0, + 1116.0, + 739.0, + 292.0, + 739.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 755.0, + 1405.0, + 755.0, + 1405.0, + 789.0, + 294.0, + 789.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 783.0, + 1405.0, + 783.0, + 1405.0, + 820.0, + 292.0, + 820.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 816.0, + 1406.0, + 816.0, + 1406.0, + 851.0, + 294.0, + 851.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 842.0, + 1406.0, + 842.0, + 1406.0, + 884.0, + 292.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 876.0, + 723.0, + 876.0, + 723.0, + 910.0, + 295.0, + 910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 833.0, + 876.0, + 887.0, + 876.0, + 887.0, + 910.0, + 833.0, + 910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1010.0, + 876.0, + 1405.0, + 876.0, + 1405.0, + 910.0, + 1010.0, + 910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 903.0, + 1015.0, + 903.0, + 1015.0, + 944.0, + 292.0, + 944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1112.0, + 903.0, + 1165.0, + 903.0, + 1165.0, + 944.0, + 1112.0, + 944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1278.0, + 903.0, + 1408.0, + 903.0, + 1408.0, + 944.0, + 1278.0, + 944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 934.0, + 1244.0, + 934.0, + 1244.0, + 975.0, + 292.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 403.0, + 962.0, + 1358.0, + 962.0, + 1358.0, + 1008.0, + 403.0, + 1008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 429.0, + 993.0, + 1407.0, + 993.0, + 1407.0, + 1036.0, + 429.0, + 1036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1027.0, + 1405.0, + 1027.0, + 1405.0, + 1064.0, + 294.0, + 1064.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1060.0, + 765.0, + 1060.0, + 765.0, + 1092.0, + 293.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1494.0, + 1407.0, + 1494.0, + 1407.0, + 1531.0, + 295.0, + 1531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1526.0, + 1404.0, + 1526.0, + 1404.0, + 1562.0, + 295.0, + 1562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1554.0, + 1404.0, + 1554.0, + 1404.0, + 1590.0, + 295.0, + 1590.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1587.0, + 1404.0, + 1587.0, + 1404.0, + 1623.0, + 295.0, + 1623.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1618.0, + 1406.0, + 1618.0, + 1406.0, + 1651.0, + 292.0, + 1651.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1647.0, + 1406.0, + 1647.0, + 1406.0, + 1681.0, + 292.0, + 1681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1677.0, + 1408.0, + 1677.0, + 1408.0, + 1713.0, + 294.0, + 1713.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1706.0, + 1409.0, + 1706.0, + 1409.0, + 1743.0, + 292.0, + 1743.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1738.0, + 1407.0, + 1738.0, + 1407.0, + 1774.0, + 295.0, + 1774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1770.0, + 513.0, + 1770.0, + 513.0, + 1802.0, + 295.0, + 1802.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1199.0, + 1406.0, + 1199.0, + 1406.0, + 1236.0, + 294.0, + 1236.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1227.0, + 1405.0, + 1227.0, + 1405.0, + 1268.0, + 294.0, + 1268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1264.0, + 1402.0, + 1264.0, + 1402.0, + 1296.0, + 296.0, + 1296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1291.0, + 1405.0, + 1291.0, + 1405.0, + 1328.0, + 292.0, + 1328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1322.0, + 1406.0, + 1322.0, + 1406.0, + 1358.0, + 295.0, + 1358.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1353.0, + 773.0, + 1353.0, + 773.0, + 1387.0, + 295.0, + 1387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 598.0, + 1135.0, + 598.0, + 1135.0, + 638.0, + 293.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1212.0, + 598.0, + 1405.0, + 598.0, + 1405.0, + 638.0, + 1212.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 632.0, + 717.0, + 632.0, + 717.0, + 668.0, + 294.0, + 668.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 18, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 3, + "poly": [ + 296, + 629, + 1393, + 629, + 1393, + 1533, + 296, + 1533 + ], + "score": 0.973 + }, + { + "category_id": 4, + "poly": [ + 293, + 1546, + 1403, + 1546, + 1403, + 1611, + 293, + 1611 + ], + "score": 0.929 + }, + { + "category_id": 2, + "poly": [ + 834, + 2061, + 866, + 2061, + 866, + 2086, + 834, + 2086 + ], + "score": 0.778 + }, + { + "category_id": 2, + "poly": [ + 834, + 2061, + 866, + 2061, + 866, + 2086, + 834, + 2086 + ], + "score": 0.199 + }, + { + "category_id": 15, + "poly": [ + 376.0, + 624.0, + 473.0, + 624.0, + 473.0, + 652.0, + 376.0, + 652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 836.0, + 627.0, + 889.0, + 627.0, + 889.0, + 652.0, + 836.0, + 652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1060.0, + 628.0, + 1099.0, + 628.0, + 1099.0, + 651.0, + 1060.0, + 651.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1271.0, + 628.0, + 1327.0, + 628.0, + 1327.0, + 649.0, + 1271.0, + 649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 640.0, + 500.0, + 640.0, + 500.0, + 668.0, + 351.0, + 668.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 567.0, + 638.0, + 719.0, + 638.0, + 719.0, + 670.0, + 567.0, + 670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 785.0, + 638.0, + 937.0, + 638.0, + 937.0, + 670.0, + 785.0, + 670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1003.0, + 638.0, + 1155.0, + 638.0, + 1155.0, + 670.0, + 1003.0, + 670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1221.0, + 638.0, + 1373.0, + 638.0, + 1373.0, + 670.0, + 1221.0, + 670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 664.0, + 343.0, + 664.0, + 343.0, + 687.0, + 313.0, + 687.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 533.0, + 664.0, + 560.0, + 664.0, + 560.0, + 683.0, + 533.0, + 683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 688.0, + 343.0, + 688.0, + 343.0, + 711.0, + 313.0, + 711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 677.0, + 780.0, + 677.0, + 780.0, + 697.0, + 756.0, + 697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 683.0, + 997.0, + 683.0, + 997.0, + 703.0, + 974.0, + 703.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1192.0, + 672.0, + 1215.0, + 672.0, + 1215.0, + 693.0, + 1192.0, + 693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 711.0, + 343.0, + 711.0, + 343.0, + 735.0, + 313.0, + 735.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 705.0, + 560.0, + 705.0, + 560.0, + 723.0, + 538.0, + 723.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 585.0, + 715.0, + 596.0, + 715.0, + 596.0, + 727.0, + 585.0, + 727.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 599.0, + 706.0, + 605.0, + 706.0, + 605.0, + 712.0, + 599.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 706.0, + 778.0, + 706.0, + 778.0, + 725.0, + 756.0, + 725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1192.0, + 701.0, + 1215.0, + 701.0, + 1215.0, + 722.0, + 1192.0, + 722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 721.0, + 348.0, + 721.0, + 348.0, + 766.0, + 296.0, + 766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 519.0, + 726.0, + 560.0, + 726.0, + 560.0, + 763.0, + 519.0, + 763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 740.0, + 722.0, + 783.0, + 722.0, + 783.0, + 765.0, + 740.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 963.0, + 720.0, + 997.0, + 720.0, + 997.0, + 760.0, + 963.0, + 760.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1175.0, + 721.0, + 1222.0, + 721.0, + 1222.0, + 766.0, + 1175.0, + 766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 758.0, + 352.0, + 758.0, + 352.0, + 781.0, + 313.0, + 781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 762.0, + 780.0, + 762.0, + 780.0, + 782.0, + 756.0, + 782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 756.0, + 997.0, + 756.0, + 997.0, + 776.0, + 974.0, + 776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1194.0, + 760.0, + 1215.0, + 760.0, + 1215.0, + 780.0, + 1194.0, + 780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 781.0, + 345.0, + 781.0, + 345.0, + 805.0, + 313.0, + 805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 357.0, + 797.0, + 379.0, + 797.0, + 379.0, + 811.0, + 357.0, + 811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 382.0, + 777.0, + 492.0, + 777.0, + 492.0, + 816.0, + 382.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 785.0, + 560.0, + 785.0, + 560.0, + 805.0, + 538.0, + 805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 578.0, + 800.0, + 593.0, + 800.0, + 593.0, + 809.0, + 578.0, + 809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 777.0, + 709.0, + 777.0, + 709.0, + 816.0, + 600.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 791.0, + 780.0, + 791.0, + 780.0, + 811.0, + 756.0, + 811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 791.0, + 797.0, + 813.0, + 797.0, + 813.0, + 811.0, + 791.0, + 811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 818.0, + 777.0, + 927.0, + 777.0, + 927.0, + 816.0, + 818.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 791.0, + 998.0, + 791.0, + 998.0, + 812.0, + 974.0, + 812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1017.0, + 800.0, + 1030.0, + 800.0, + 1030.0, + 809.0, + 1017.0, + 809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1037.0, + 777.0, + 1145.0, + 777.0, + 1145.0, + 816.0, + 1037.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1194.0, + 789.0, + 1216.0, + 789.0, + 1216.0, + 809.0, + 1194.0, + 809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1230.0, + 797.0, + 1252.0, + 797.0, + 1252.0, + 811.0, + 1230.0, + 811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1255.0, + 777.0, + 1364.0, + 777.0, + 1364.0, + 816.0, + 1255.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 805.0, + 343.0, + 805.0, + 343.0, + 827.0, + 313.0, + 827.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 335.0, + 820.0, + 522.0, + 820.0, + 522.0, + 844.0, + 335.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 554.0, + 821.0, + 740.0, + 821.0, + 740.0, + 843.0, + 554.0, + 843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 768.0, + 821.0, + 957.0, + 821.0, + 957.0, + 843.0, + 768.0, + 843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 992.0, + 820.0, + 1175.0, + 820.0, + 1175.0, + 844.0, + 992.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1210.0, + 820.0, + 1394.0, + 820.0, + 1394.0, + 844.0, + 1210.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 397.0, + 833.0, + 454.0, + 833.0, + 454.0, + 858.0, + 397.0, + 858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 616.0, + 835.0, + 671.0, + 835.0, + 671.0, + 856.0, + 616.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 835.0, + 835.0, + 889.0, + 835.0, + 889.0, + 856.0, + 835.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1053.0, + 835.0, + 1108.0, + 835.0, + 1108.0, + 856.0, + 1053.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1271.0, + 835.0, + 1326.0, + 835.0, + 1326.0, + 856.0, + 1271.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 377.0, + 855.0, + 474.0, + 855.0, + 474.0, + 879.0, + 377.0, + 879.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 614.0, + 852.0, + 674.0, + 852.0, + 674.0, + 881.0, + 614.0, + 881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 835.0, + 854.0, + 889.0, + 854.0, + 889.0, + 879.0, + 835.0, + 879.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1060.0, + 855.0, + 1100.0, + 855.0, + 1100.0, + 878.0, + 1060.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1270.0, + 854.0, + 1328.0, + 854.0, + 1328.0, + 879.0, + 1270.0, + 879.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 866.0, + 501.0, + 866.0, + 501.0, + 897.0, + 350.0, + 897.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 569.0, + 868.0, + 717.0, + 868.0, + 717.0, + 895.0, + 569.0, + 895.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 786.0, + 865.0, + 937.0, + 865.0, + 937.0, + 897.0, + 786.0, + 897.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1005.0, + 868.0, + 1155.0, + 868.0, + 1155.0, + 895.0, + 1005.0, + 895.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1223.0, + 865.0, + 1372.0, + 865.0, + 1372.0, + 897.0, + 1223.0, + 897.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 889.0, + 342.0, + 889.0, + 342.0, + 909.0, + 316.0, + 909.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 891.0, + 560.0, + 891.0, + 560.0, + 912.0, + 538.0, + 912.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 886.0, + 780.0, + 886.0, + 780.0, + 907.0, + 756.0, + 907.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 886.0, + 999.0, + 886.0, + 999.0, + 910.0, + 964.0, + 910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1182.0, + 883.0, + 1217.0, + 883.0, + 1217.0, + 905.0, + 1182.0, + 905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 912.0, + 343.0, + 912.0, + 343.0, + 934.0, + 313.0, + 934.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 963.0, + 912.0, + 999.0, + 912.0, + 999.0, + 935.0, + 963.0, + 935.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1182.0, + 907.0, + 1217.0, + 907.0, + 1217.0, + 930.0, + 1182.0, + 930.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 937.0, + 342.0, + 937.0, + 342.0, + 958.0, + 320.0, + 958.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 928.0, + 560.0, + 928.0, + 560.0, + 948.0, + 538.0, + 948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 925.0, + 780.0, + 925.0, + 780.0, + 945.0, + 756.0, + 945.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 963.0, + 935.0, + 999.0, + 935.0, + 999.0, + 959.0, + 963.0, + 959.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1181.0, + 932.0, + 1217.0, + 932.0, + 1217.0, + 954.0, + 1181.0, + 954.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1241.0, + 929.0, + 1258.0, + 929.0, + 1258.0, + 973.0, + 1241.0, + 973.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 301.0, + 953.0, + 342.0, + 953.0, + 342.0, + 988.0, + 301.0, + 988.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 517.0, + 948.0, + 564.0, + 948.0, + 564.0, + 989.0, + 517.0, + 989.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 740.0, + 949.0, + 785.0, + 949.0, + 785.0, + 991.0, + 740.0, + 991.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 948.0, + 948.0, + 1004.0, + 948.0, + 1004.0, + 992.0, + 948.0, + 992.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1024.0, + 962.0, + 1033.0, + 962.0, + 1033.0, + 972.0, + 1024.0, + 972.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1165.0, + 949.0, + 1221.0, + 949.0, + 1221.0, + 993.0, + 1165.0, + 993.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 987.0, + 342.0, + 987.0, + 342.0, + 1006.0, + 320.0, + 1006.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 984.0, + 999.0, + 984.0, + 999.0, + 1008.0, + 964.0, + 1008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1184.0, + 981.0, + 1217.0, + 981.0, + 1217.0, + 1003.0, + 1184.0, + 1003.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1011.0, + 342.0, + 1011.0, + 342.0, + 1053.0, + 320.0, + 1053.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 383.0, + 1004.0, + 490.0, + 1004.0, + 490.0, + 1042.0, + 383.0, + 1042.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 999.0, + 560.0, + 999.0, + 560.0, + 1053.0, + 538.0, + 1053.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 1004.0, + 709.0, + 1004.0, + 709.0, + 1046.0, + 600.0, + 1046.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 755.0, + 1001.0, + 788.0, + 1001.0, + 788.0, + 1023.0, + 755.0, + 1023.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 818.0, + 1004.0, + 927.0, + 1004.0, + 927.0, + 1046.0, + 818.0, + 1046.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 963.0, + 1008.0, + 1000.0, + 1008.0, + 1000.0, + 1056.0, + 963.0, + 1056.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1036.0, + 1004.0, + 1145.0, + 1004.0, + 1145.0, + 1046.0, + 1036.0, + 1046.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1182.0, + 1006.0, + 1219.0, + 1006.0, + 1219.0, + 1052.0, + 1182.0, + 1052.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1255.0, + 1004.0, + 1364.0, + 1004.0, + 1364.0, + 1043.0, + 1255.0, + 1043.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 337.0, + 1047.0, + 522.0, + 1047.0, + 522.0, + 1071.0, + 337.0, + 1071.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 553.0, + 1047.0, + 741.0, + 1047.0, + 741.0, + 1071.0, + 553.0, + 1071.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 783.0, + 1047.0, + 959.0, + 1047.0, + 959.0, + 1071.0, + 783.0, + 1071.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 988.0, + 1046.0, + 1176.0, + 1046.0, + 1176.0, + 1070.0, + 988.0, + 1070.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1204.0, + 1047.0, + 1394.0, + 1047.0, + 1394.0, + 1071.0, + 1204.0, + 1071.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 399.0, + 1062.0, + 452.0, + 1062.0, + 452.0, + 1084.0, + 399.0, + 1084.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 616.0, + 1062.0, + 670.0, + 1062.0, + 670.0, + 1084.0, + 616.0, + 1084.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 835.0, + 1062.0, + 888.0, + 1062.0, + 888.0, + 1084.0, + 835.0, + 1084.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1053.0, + 1061.0, + 1108.0, + 1061.0, + 1108.0, + 1086.0, + 1053.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1272.0, + 1062.0, + 1325.0, + 1062.0, + 1325.0, + 1084.0, + 1272.0, + 1084.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 376.0, + 1080.0, + 473.0, + 1080.0, + 473.0, + 1105.0, + 376.0, + 1105.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 615.0, + 1084.0, + 672.0, + 1084.0, + 672.0, + 1105.0, + 615.0, + 1105.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 834.0, + 1079.0, + 889.0, + 1079.0, + 889.0, + 1108.0, + 834.0, + 1108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1059.0, + 1082.0, + 1100.0, + 1082.0, + 1100.0, + 1105.0, + 1059.0, + 1105.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1270.0, + 1082.0, + 1326.0, + 1082.0, + 1326.0, + 1104.0, + 1270.0, + 1104.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 1111.0, + 342.0, + 1111.0, + 342.0, + 1130.0, + 315.0, + 1130.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 357.0, + 1095.0, + 492.0, + 1095.0, + 492.0, + 1123.0, + 357.0, + 1123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 530.0, + 1105.0, + 561.0, + 1105.0, + 561.0, + 1129.0, + 530.0, + 1129.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 576.0, + 1095.0, + 710.0, + 1095.0, + 710.0, + 1123.0, + 576.0, + 1123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 1111.0, + 778.0, + 1111.0, + 778.0, + 1130.0, + 756.0, + 1130.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 794.0, + 1095.0, + 928.0, + 1095.0, + 928.0, + 1123.0, + 794.0, + 1123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1012.0, + 1095.0, + 1149.0, + 1095.0, + 1149.0, + 1122.0, + 1012.0, + 1122.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1230.0, + 1095.0, + 1365.0, + 1095.0, + 1365.0, + 1123.0, + 1230.0, + 1123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1128.0, + 348.0, + 1128.0, + 348.0, + 1222.0, + 297.0, + 1222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 519.0, + 1140.0, + 561.0, + 1140.0, + 561.0, + 1215.0, + 519.0, + 1215.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 739.0, + 1135.0, + 786.0, + 1135.0, + 786.0, + 1220.0, + 739.0, + 1220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 955.0, + 1122.0, + 1003.0, + 1122.0, + 1003.0, + 1220.0, + 955.0, + 1220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1174.0, + 1127.0, + 1224.0, + 1127.0, + 1224.0, + 1224.0, + 1174.0, + 1224.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1215.0, + 345.0, + 1215.0, + 345.0, + 1239.0, + 313.0, + 1239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1214.0, + 560.0, + 1214.0, + 560.0, + 1234.0, + 538.0, + 1234.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 1208.0, + 780.0, + 1208.0, + 780.0, + 1228.0, + 756.0, + 1228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 973.0, + 1209.0, + 998.0, + 1209.0, + 998.0, + 1229.0, + 973.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1191.0, + 1222.0, + 1216.0, + 1222.0, + 1216.0, + 1242.0, + 1191.0, + 1242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 312.0, + 1232.0, + 522.0, + 1232.0, + 522.0, + 1298.0, + 312.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 1232.0, + 740.0, + 1232.0, + 740.0, + 1298.0, + 537.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 755.0, + 1230.0, + 958.0, + 1230.0, + 958.0, + 1298.0, + 755.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 1232.0, + 1176.0, + 1232.0, + 1176.0, + 1298.0, + 974.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1190.0, + 1232.0, + 1394.0, + 1232.0, + 1394.0, + 1298.0, + 1190.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 1289.0, + 452.0, + 1289.0, + 452.0, + 1311.0, + 398.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 616.0, + 1289.0, + 670.0, + 1289.0, + 670.0, + 1311.0, + 616.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 835.0, + 1289.0, + 888.0, + 1289.0, + 888.0, + 1311.0, + 835.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1054.0, + 1289.0, + 1106.0, + 1289.0, + 1106.0, + 1311.0, + 1054.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1271.0, + 1289.0, + 1326.0, + 1289.0, + 1326.0, + 1311.0, + 1271.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1309.0, + 492.0, + 1309.0, + 492.0, + 1352.0, + 356.0, + 1352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 572.0, + 1306.0, + 711.0, + 1306.0, + 711.0, + 1354.0, + 572.0, + 1354.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 745.0, + 1332.0, + 781.0, + 1332.0, + 781.0, + 1355.0, + 745.0, + 1355.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 791.0, + 1308.0, + 928.0, + 1308.0, + 928.0, + 1352.0, + 791.0, + 1352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 1333.0, + 998.0, + 1333.0, + 998.0, + 1353.0, + 974.0, + 1353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1010.0, + 1308.0, + 1145.0, + 1308.0, + 1145.0, + 1350.0, + 1010.0, + 1350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 1308.0, + 1364.0, + 1308.0, + 1364.0, + 1351.0, + 1229.0, + 1351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1347.0, + 343.0, + 1347.0, + 343.0, + 1393.0, + 313.0, + 1393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 1343.0, + 561.0, + 1343.0, + 561.0, + 1388.0, + 532.0, + 1388.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 745.0, + 1356.0, + 781.0, + 1356.0, + 781.0, + 1378.0, + 745.0, + 1378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 973.0, + 1356.0, + 999.0, + 1356.0, + 999.0, + 1381.0, + 973.0, + 1381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1192.0, + 1363.0, + 1216.0, + 1363.0, + 1216.0, + 1383.0, + 1192.0, + 1383.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1393.0, + 348.0, + 1393.0, + 348.0, + 1447.0, + 296.0, + 1447.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 521.0, + 1391.0, + 561.0, + 1391.0, + 561.0, + 1444.0, + 521.0, + 1444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 726.0, + 1380.0, + 790.0, + 1380.0, + 790.0, + 1460.0, + 726.0, + 1460.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 955.0, + 1383.0, + 1003.0, + 1383.0, + 1003.0, + 1447.0, + 955.0, + 1447.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1192.0, + 1395.0, + 1216.0, + 1395.0, + 1216.0, + 1415.0, + 1192.0, + 1415.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1442.0, + 343.0, + 1442.0, + 343.0, + 1463.0, + 320.0, + 1463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1437.0, + 560.0, + 1437.0, + 560.0, + 1459.0, + 538.0, + 1459.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 1432.0, + 998.0, + 1432.0, + 998.0, + 1454.0, + 974.0, + 1454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1465.0, + 343.0, + 1465.0, + 343.0, + 1509.0, + 320.0, + 1509.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 361.0, + 1483.0, + 377.0, + 1483.0, + 377.0, + 1493.0, + 361.0, + 1493.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 380.0, + 1457.0, + 493.0, + 1457.0, + 493.0, + 1502.0, + 380.0, + 1502.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 535.0, + 1460.0, + 563.0, + 1460.0, + 563.0, + 1505.0, + 535.0, + 1505.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 598.0, + 1456.0, + 709.0, + 1456.0, + 709.0, + 1501.0, + 598.0, + 1501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 745.0, + 1452.0, + 781.0, + 1452.0, + 781.0, + 1500.0, + 745.0, + 1500.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 791.0, + 1480.0, + 815.0, + 1480.0, + 815.0, + 1494.0, + 791.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 817.0, + 1460.0, + 927.0, + 1460.0, + 927.0, + 1501.0, + 817.0, + 1501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 973.0, + 1457.0, + 998.0, + 1457.0, + 998.0, + 1503.0, + 973.0, + 1503.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1036.0, + 1460.0, + 1145.0, + 1460.0, + 1145.0, + 1501.0, + 1036.0, + 1501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1192.0, + 1460.0, + 1216.0, + 1460.0, + 1216.0, + 1511.0, + 1192.0, + 1511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1232.0, + 1481.0, + 1252.0, + 1481.0, + 1252.0, + 1494.0, + 1232.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1254.0, + 1460.0, + 1364.0, + 1460.0, + 1364.0, + 1501.0, + 1254.0, + 1501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 335.0, + 1501.0, + 523.0, + 1501.0, + 523.0, + 1525.0, + 335.0, + 1525.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 553.0, + 1503.0, + 740.0, + 1503.0, + 740.0, + 1525.0, + 553.0, + 1525.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 768.0, + 1504.0, + 958.0, + 1504.0, + 958.0, + 1524.0, + 768.0, + 1524.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 989.0, + 1505.0, + 1042.0, + 1505.0, + 1042.0, + 1523.0, + 989.0, + 1523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1043.0, + 1504.0, + 1175.0, + 1504.0, + 1175.0, + 1524.0, + 1043.0, + 1524.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1207.0, + 1501.0, + 1394.0, + 1501.0, + 1394.0, + 1525.0, + 1207.0, + 1525.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 1516.0, + 453.0, + 1516.0, + 453.0, + 1538.0, + 398.0, + 1538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 616.0, + 1516.0, + 671.0, + 1516.0, + 671.0, + 1538.0, + 616.0, + 1538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 835.0, + 1516.0, + 889.0, + 1516.0, + 889.0, + 1538.0, + 835.0, + 1538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1053.0, + 1516.0, + 1108.0, + 1516.0, + 1108.0, + 1538.0, + 1053.0, + 1538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1272.0, + 1516.0, + 1326.0, + 1516.0, + 1326.0, + 1538.0, + 1272.0, + 1538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 615.0, + 628.5, + 670.0, + 628.5, + 670.0, + 651.5, + 615.0, + 651.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 814.75, + 930.5, + 845.75, + 930.5, + 845.75, + 943.0, + 814.75, + 943.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1243.0, + 918.0, + 1287.0, + 918.0, + 1287.0, + 935.5, + 1243.0, + 935.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 959.0, + 619.0, + 959.0, + 619.0, + 968.0, + 600.0, + 968.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 761.0, + 1041.0, + 785.0, + 1041.0, + 785.0, + 1066.0, + 761.0, + 1066.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 1269.0, + 1219.0, + 1269.0, + 1219.0, + 1292.0, + 1198.0, + 1292.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1190.75, + 1416.5, + 1197.75, + 1416.5, + 1197.75, + 1443.5, + 1190.75, + 1443.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1545.0, + 1404.0, + 1545.0, + 1404.0, + 1581.0, + 295.0, + 1581.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1573.0, + 1104.0, + 1573.0, + 1104.0, + 1616.0, + 293.0, + 1616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 828.0, + 2056.0, + 872.0, + 2056.0, + 872.0, + 2096.0, + 828.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 828.0, + 2056.0, + 872.0, + 2056.0, + 872.0, + 2096.0, + 828.0, + 2096.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 19, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 3, + "poly": [ + 302, + 255, + 1406, + 255, + 1406, + 1403, + 302, + 1403 + ], + "score": 0.937 + }, + { + "category_id": 3, + "poly": [ + 300, + 1511, + 1396, + 1511, + 1396, + 1920, + 300, + 1920 + ], + "score": 0.91 + }, + { + "category_id": 4, + "poly": [ + 610, + 1946, + 1091, + 1946, + 1091, + 1979, + 610, + 1979 + ], + "score": 0.903 + }, + { + "category_id": 4, + "poly": [ + 485, + 1414, + 1212, + 1414, + 1212, + 1447, + 485, + 1447 + ], + "score": 0.892 + }, + { + "category_id": 2, + "poly": [ + 834, + 2061, + 863, + 2061, + 863, + 2085, + 834, + 2085 + ], + "score": 0.842 + }, + { + "category_id": 2, + "poly": [ + 514, + 238, + 629, + 238, + 629, + 263, + 514, + 263 + ], + "score": 0.42 + }, + { + "category_id": 2, + "poly": [ + 1075, + 238, + 1187, + 238, + 1187, + 264, + 1075, + 264 + ], + "score": 0.376 + }, + { + "category_id": 1, + "poly": [ + 1075, + 238, + 1187, + 238, + 1187, + 264, + 1075, + 264 + ], + "score": 0.146 + }, + { + "category_id": 15, + "poly": [ + 635.0, + 325.0, + 729.0, + 325.0, + 729.0, + 386.0, + 635.0, + 386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 495.0, + 472.0, + 777.0, + 472.0, + 777.0, + 504.0, + 495.0, + 504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 785.0, + 465.0, + 820.0, + 465.0, + 820.0, + 505.0, + 785.0, + 505.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1053.0, + 474.0, + 1337.0, + 474.0, + 1337.0, + 507.0, + 1053.0, + 507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1342.0, + 466.0, + 1383.0, + 466.0, + 1383.0, + 512.0, + 1342.0, + 512.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 319.0, + 529.0, + 349.0, + 529.0, + 349.0, + 548.0, + 319.0, + 548.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 516.0, + 719.0, + 516.0, + 719.0, + 564.0, + 369.0, + 564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 927.0, + 522.0, + 1173.0, + 522.0, + 1173.0, + 555.0, + 927.0, + 555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 558.0, + 575.0, + 744.0, + 575.0, + 744.0, + 604.0, + 558.0, + 604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 783.0, + 564.0, + 825.0, + 564.0, + 825.0, + 611.0, + 783.0, + 611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1138.0, + 574.0, + 1328.0, + 574.0, + 1328.0, + 607.0, + 1138.0, + 607.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1342.0, + 565.0, + 1385.0, + 565.0, + 1385.0, + 611.0, + 1342.0, + 611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 621.0, + 752.0, + 621.0, + 752.0, + 651.0, + 373.0, + 651.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 931.0, + 618.0, + 1325.0, + 618.0, + 1325.0, + 647.0, + 931.0, + 647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 371.0, + 643.0, + 755.0, + 643.0, + 755.0, + 672.0, + 371.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 928.0, + 638.0, + 1320.0, + 638.0, + 1320.0, + 669.0, + 928.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 371.0, + 664.0, + 740.0, + 664.0, + 740.0, + 693.0, + 371.0, + 693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 927.0, + 658.0, + 1039.0, + 658.0, + 1039.0, + 693.0, + 927.0, + 693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 685.0, + 762.0, + 685.0, + 762.0, + 713.0, + 373.0, + 713.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 705.0, + 631.0, + 705.0, + 631.0, + 734.0, + 373.0, + 734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 561.0, + 976.0, + 752.0, + 976.0, + 752.0, + 1005.0, + 561.0, + 1005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 785.0, + 967.0, + 828.0, + 967.0, + 828.0, + 1014.0, + 785.0, + 1014.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1115.0, + 973.0, + 1310.0, + 973.0, + 1310.0, + 1000.0, + 1115.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1343.0, + 960.0, + 1387.0, + 960.0, + 1387.0, + 1011.0, + 1343.0, + 1011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 1020.0, + 361.0, + 1020.0, + 361.0, + 1063.0, + 314.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 386.0, + 1040.0, + 744.0, + 1040.0, + 744.0, + 1075.0, + 386.0, + 1075.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 940.0, + 1037.0, + 1300.0, + 1037.0, + 1300.0, + 1071.0, + 940.0, + 1071.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 1064.0, + 715.0, + 1064.0, + 715.0, + 1093.0, + 387.0, + 1093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1060.0, + 1303.0, + 1060.0, + 1303.0, + 1089.0, + 941.0, + 1089.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 386.0, + 1085.0, + 736.0, + 1085.0, + 736.0, + 1112.0, + 386.0, + 1112.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 938.0, + 1080.0, + 1295.0, + 1080.0, + 1295.0, + 1110.0, + 938.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 1106.0, + 715.0, + 1106.0, + 715.0, + 1131.0, + 387.0, + 1131.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1101.0, + 1230.0, + 1101.0, + 1230.0, + 1129.0, + 941.0, + 1129.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 940.0, + 1119.0, + 1291.0, + 1119.0, + 1291.0, + 1151.0, + 940.0, + 1151.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 1146.0, + 763.0, + 1146.0, + 763.0, + 1175.0, + 387.0, + 1175.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1144.0, + 992.0, + 1144.0, + 992.0, + 1170.0, + 941.0, + 1170.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 384.0, + 1166.0, + 766.0, + 1166.0, + 766.0, + 1196.0, + 384.0, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 937.0, + 1159.0, + 1312.0, + 1159.0, + 1312.0, + 1192.0, + 937.0, + 1192.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 383.0, + 1185.0, + 772.0, + 1185.0, + 772.0, + 1219.0, + 383.0, + 1219.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 942.0, + 1185.0, + 1017.0, + 1185.0, + 1017.0, + 1211.0, + 942.0, + 1211.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 1211.0, + 746.0, + 1211.0, + 746.0, + 1236.0, + 387.0, + 1236.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1203.0, + 1313.0, + 1203.0, + 1313.0, + 1232.0, + 941.0, + 1232.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 1231.0, + 716.0, + 1231.0, + 716.0, + 1255.0, + 387.0, + 1255.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 940.0, + 1224.0, + 1120.0, + 1224.0, + 1120.0, + 1253.0, + 940.0, + 1253.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 386.0, + 1249.0, + 759.0, + 1249.0, + 759.0, + 1277.0, + 386.0, + 1277.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 938.0, + 1244.0, + 1259.0, + 1244.0, + 1259.0, + 1275.0, + 938.0, + 1275.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 386.0, + 1268.0, + 744.0, + 1268.0, + 744.0, + 1298.0, + 386.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 940.0, + 1266.0, + 1309.0, + 1266.0, + 1309.0, + 1294.0, + 940.0, + 1294.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 386.0, + 1289.0, + 740.0, + 1289.0, + 740.0, + 1320.0, + 386.0, + 1320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 939.0, + 1284.0, + 1133.0, + 1284.0, + 1133.0, + 1318.0, + 939.0, + 1318.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 1310.0, + 519.0, + 1310.0, + 519.0, + 1341.0, + 387.0, + 1341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 391.0, + 1341.0, + 426.0, + 1341.0, + 426.0, + 1358.0, + 391.0, + 1358.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 629.25, + 374.0, + 737.25, + 374.0, + 737.25, + 407.5, + 629.25, + 407.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1205.25, + 373.5, + 1312.25, + 373.5, + 1312.25, + 407.5, + 1205.25, + 407.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 443.0, + 1665.0, + 665.0, + 1665.0, + 665.0, + 1700.0, + 443.0, + 1700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 777.0, + 1657.0, + 822.0, + 1657.0, + 822.0, + 1706.0, + 777.0, + 1706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1009.0, + 1668.0, + 1242.0, + 1668.0, + 1242.0, + 1698.0, + 1009.0, + 1698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1351.0, + 1665.0, + 1374.0, + 1665.0, + 1374.0, + 1695.0, + 1351.0, + 1695.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 1744.0, + 755.0, + 1744.0, + 755.0, + 1771.0, + 372.0, + 1771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 877.0, + 1744.0, + 915.0, + 1744.0, + 915.0, + 1766.0, + 877.0, + 1766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 930.0, + 1739.0, + 1336.0, + 1739.0, + 1336.0, + 1769.0, + 930.0, + 1769.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 1766.0, + 741.0, + 1766.0, + 741.0, + 1789.0, + 372.0, + 1789.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 931.0, + 1762.0, + 1316.0, + 1762.0, + 1316.0, + 1789.0, + 931.0, + 1789.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 371.0, + 1784.0, + 762.0, + 1784.0, + 762.0, + 1811.0, + 371.0, + 1811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 929.0, + 1779.0, + 1326.0, + 1779.0, + 1326.0, + 1811.0, + 929.0, + 1811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 370.0, + 1806.0, + 770.0, + 1806.0, + 770.0, + 1833.0, + 370.0, + 1833.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 931.0, + 1804.0, + 1328.0, + 1804.0, + 1328.0, + 1831.0, + 931.0, + 1831.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1828.0, + 737.0, + 1828.0, + 737.0, + 1852.0, + 368.0, + 1852.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 928.0, + 1826.0, + 1121.0, + 1826.0, + 1121.0, + 1850.0, + 928.0, + 1850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 370.0, + 1845.0, + 758.0, + 1845.0, + 758.0, + 1873.0, + 370.0, + 1873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 1867.0, + 576.0, + 1867.0, + 576.0, + 1894.0, + 367.0, + 1894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 604.0, + 1944.0, + 1093.0, + 1944.0, + 1093.0, + 1981.0, + 604.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 484.0, + 1411.0, + 1215.0, + 1411.0, + 1215.0, + 1451.0, + 484.0, + 1451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 830.0, + 2057.0, + 868.0, + 2057.0, + 868.0, + 2097.0, + 830.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 512.0, + 237.0, + 632.0, + 237.0, + 632.0, + 268.0, + 512.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1074.0, + 234.0, + 1190.0, + 234.0, + 1190.0, + 269.0, + 1074.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1074.0, + 234.0, + 1190.0, + 234.0, + 1190.0, + 269.0, + 1074.0, + 269.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 20, + "width": 1700, + "height": 2200 + } + } +] \ No newline at end of file diff --git a/parse/dev/9h3KsOVXhLZ/images/1bb8bb00097fc003ed50452c518b133bb61db7f0945a63e73169aacad6d425e8.jpg b/parse/dev/9h3KsOVXhLZ/images/1bb8bb00097fc003ed50452c518b133bb61db7f0945a63e73169aacad6d425e8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8cae391ce55aff17361327e5b50bc612ddd82f04 --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/1bb8bb00097fc003ed50452c518b133bb61db7f0945a63e73169aacad6d425e8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4762abb72e674e695d8f5d83c584e0a4db906de22ae0ae52435aa88695a052d5 +size 45867 diff --git a/parse/dev/9h3KsOVXhLZ/images/3bba2b7d42766c9ff3a4f32481a61ce95bd523e3be23e40ee2e48340e3add62a.jpg b/parse/dev/9h3KsOVXhLZ/images/3bba2b7d42766c9ff3a4f32481a61ce95bd523e3be23e40ee2e48340e3add62a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8826ede4ea05424ad543145e33293c0bd772ac84 --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/3bba2b7d42766c9ff3a4f32481a61ce95bd523e3be23e40ee2e48340e3add62a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e1e36381626b96bff78e767c5f3e6fc69293253f500c649a281941356229d60a +size 14503 diff --git a/parse/dev/9h3KsOVXhLZ/images/430b51ce3631780d398a75dc7794e5aba609c20ad0b960bb5a5dca0210f1fdbb.jpg b/parse/dev/9h3KsOVXhLZ/images/430b51ce3631780d398a75dc7794e5aba609c20ad0b960bb5a5dca0210f1fdbb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ebf0d8883e3d09bbd24fa07eaca2e33fd54c84d8 --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/430b51ce3631780d398a75dc7794e5aba609c20ad0b960bb5a5dca0210f1fdbb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6b36e75fc68f8aaa1aa288e63762ea1f613d4ae91c21ba86ca94a551fd7efb7 +size 5674 diff --git a/parse/dev/9h3KsOVXhLZ/images/73828994b44f600185e85d70e76927cce9520e5ccd019e4f95bdf479fb194855.jpg b/parse/dev/9h3KsOVXhLZ/images/73828994b44f600185e85d70e76927cce9520e5ccd019e4f95bdf479fb194855.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fda58f2b49fd814a63acb3e963a866b5c8954fd5 --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/73828994b44f600185e85d70e76927cce9520e5ccd019e4f95bdf479fb194855.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be5ea837483fa7beafb2d1c8453c60cee601e76ecbdeac49d1d0465952947512 +size 32311 diff --git a/parse/dev/9h3KsOVXhLZ/images/74ca43b58bc04880ee39b4edeb852b7901f8ee82b6bd92c8f29c4e183958a2b7.jpg b/parse/dev/9h3KsOVXhLZ/images/74ca43b58bc04880ee39b4edeb852b7901f8ee82b6bd92c8f29c4e183958a2b7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9bcea49da83e1b02b1e162a8c2ef1b4165043795 --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/74ca43b58bc04880ee39b4edeb852b7901f8ee82b6bd92c8f29c4e183958a2b7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:700693d14b67633c2fdf382d586c8b7cb896913866ead2034fd6fcb4ced967b1 +size 3667 diff --git a/parse/dev/9h3KsOVXhLZ/images/875790817d6f8f66d0085f280abbbf833d6cc0a510a1acc60b86080a75f38e4e.jpg b/parse/dev/9h3KsOVXhLZ/images/875790817d6f8f66d0085f280abbbf833d6cc0a510a1acc60b86080a75f38e4e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3f21c9527621ce7c40d862cb7b7a85527391c970 --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/875790817d6f8f66d0085f280abbbf833d6cc0a510a1acc60b86080a75f38e4e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:268810369e1d2219cb91ffdcca66040e44255da3af30d1b70e0cb50a69f88cfd +size 12246 diff --git a/parse/dev/9h3KsOVXhLZ/images/a44e35c837372cddc6d07e6849d74323e28bf2af4ffd3ab4776283ad76cf9d69.jpg b/parse/dev/9h3KsOVXhLZ/images/a44e35c837372cddc6d07e6849d74323e28bf2af4ffd3ab4776283ad76cf9d69.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7c26a00b92ea74d1d0cafcd7fd232d84749d1a86 --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/a44e35c837372cddc6d07e6849d74323e28bf2af4ffd3ab4776283ad76cf9d69.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5a03a851a5e89befe9f7c9f6ed0ebd30c713eb169ce869a8e85cfc0d721388b +size 34422 diff --git a/parse/dev/9h3KsOVXhLZ/images/a87f4b17ae97d04fa2ccec404354fa712c08104ebdca9eedde6266d69c1cc000.jpg b/parse/dev/9h3KsOVXhLZ/images/a87f4b17ae97d04fa2ccec404354fa712c08104ebdca9eedde6266d69c1cc000.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ed96ca9f8298b2fcc5cb2664f98f8315cd96b42a --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/a87f4b17ae97d04fa2ccec404354fa712c08104ebdca9eedde6266d69c1cc000.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e3abdcc53cdacc53758d29724b782ea3a97ddcf7ddae90c53bde13db6a5d614 +size 6623 diff --git a/parse/dev/9h3KsOVXhLZ/images/b7a6628efd2eae2119392aade7e89ec4e64d2435bd65751772febfd9330744d7.jpg b/parse/dev/9h3KsOVXhLZ/images/b7a6628efd2eae2119392aade7e89ec4e64d2435bd65751772febfd9330744d7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d0e3390c2e2f42ac131e3abfaca43f17114c8f12 --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/b7a6628efd2eae2119392aade7e89ec4e64d2435bd65751772febfd9330744d7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:006a19290f90342a0a034dea70dfb8f85cd54588aa1961ed977366da7565af89 +size 15976 diff --git a/parse/dev/9h3KsOVXhLZ/images/bbec3b18183201f9ed3782f1477a1e4082f6dcb762564bffeab908496dfdcf90.jpg b/parse/dev/9h3KsOVXhLZ/images/bbec3b18183201f9ed3782f1477a1e4082f6dcb762564bffeab908496dfdcf90.jpg new file mode 100644 index 0000000000000000000000000000000000000000..02d3cd5ccca55e7e6dae96c418010d7c449a54ca --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/bbec3b18183201f9ed3782f1477a1e4082f6dcb762564bffeab908496dfdcf90.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9666f7e8ac1c67fbdfe5f4edb5f7b58ebae4b71a5beb462487145f3f24428c29 +size 97733 diff --git a/parse/dev/9h3KsOVXhLZ/images/bff5b4163414025c27ee09c6e9bfb93d55e373aa201c681052eec140b863d86c.jpg b/parse/dev/9h3KsOVXhLZ/images/bff5b4163414025c27ee09c6e9bfb93d55e373aa201c681052eec140b863d86c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7372fb2c168eee1520530d992c3e24d9f80ed4aa --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/bff5b4163414025c27ee09c6e9bfb93d55e373aa201c681052eec140b863d86c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4440fbd64fc7f9d848bc004282c1d3df059bf67a907dae00535b51f172c01cf0 +size 60386 diff --git a/parse/dev/9h3KsOVXhLZ/images/ceb1ea2031ef2c2e53af6c2d1dc3d1053361e969b449e58a2d842f8df5d040e9.jpg b/parse/dev/9h3KsOVXhLZ/images/ceb1ea2031ef2c2e53af6c2d1dc3d1053361e969b449e58a2d842f8df5d040e9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..38b2e091b187a6ca71a3f80e782bf29552cd5b44 --- /dev/null +++ b/parse/dev/9h3KsOVXhLZ/images/ceb1ea2031ef2c2e53af6c2d1dc3d1053361e969b449e58a2d842f8df5d040e9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c7ac39ec6646d3eb416defc4994cdca404650bab49c52b0f5f70a69730349e1 +size 68710 diff --git a/parse/dev/E01k9048soZ/images/2f8c99510591bfdc0bffc66a6d5d85eed91d57bc162512dc17a717a4f67899d4.jpg b/parse/dev/E01k9048soZ/images/2f8c99510591bfdc0bffc66a6d5d85eed91d57bc162512dc17a717a4f67899d4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6c9ab00b38a3f226dbd5aa6d232e45f3cc8e74b4 --- /dev/null +++ b/parse/dev/E01k9048soZ/images/2f8c99510591bfdc0bffc66a6d5d85eed91d57bc162512dc17a717a4f67899d4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8350be935ebdc7be4ff5b45230739552563d22aa256b3e3e2f197d8d035c2083 +size 21788 diff --git a/parse/dev/E01k9048soZ/images/32c0745c0c071d9f734e64c926d92b88237b585ab3ee8a3554a588cf7e9b215a.jpg b/parse/dev/E01k9048soZ/images/32c0745c0c071d9f734e64c926d92b88237b585ab3ee8a3554a588cf7e9b215a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e5dc04818d3cc2ccb9ea4d66baddc4e516e4850c --- /dev/null +++ b/parse/dev/E01k9048soZ/images/32c0745c0c071d9f734e64c926d92b88237b585ab3ee8a3554a588cf7e9b215a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:343a931bff3b4ce1f0315d258a0c217412367ff4a7c854f5c8b875c897b46926 +size 33805 diff --git a/parse/dev/E01k9048soZ/images/340219cfe46bf17e80123caf70353582275bc280b521242e7f513b62951e2ff3.jpg b/parse/dev/E01k9048soZ/images/340219cfe46bf17e80123caf70353582275bc280b521242e7f513b62951e2ff3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..595b6ff3ddb408f87f559f3e0e0df765a2e49752 --- /dev/null +++ b/parse/dev/E01k9048soZ/images/340219cfe46bf17e80123caf70353582275bc280b521242e7f513b62951e2ff3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2f43eaa1c9a8827df3d03b5168f8f06eb70ca58c490aa76ae6a1ef3a6a600a3a +size 23805 diff --git a/parse/dev/E01k9048soZ/images/49815df8fbbf3ec01b60ee349bebb1489c7894830b85fde0e4ed0c9f7d3af5ed.jpg b/parse/dev/E01k9048soZ/images/49815df8fbbf3ec01b60ee349bebb1489c7894830b85fde0e4ed0c9f7d3af5ed.jpg new file mode 100644 index 0000000000000000000000000000000000000000..27f85b5d7eb6c5a5ae376fabd3f8bc8ab9ee86fd --- /dev/null +++ b/parse/dev/E01k9048soZ/images/49815df8fbbf3ec01b60ee349bebb1489c7894830b85fde0e4ed0c9f7d3af5ed.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82c485f512a026fd1d722e2d393a863b536f9d12eadb41d24c3b7c7944dd4af4 +size 23425 diff --git a/parse/dev/E01k9048soZ/images/699b0f1210062183978fe46e8634cccdbedef691e47a3e9eb136e8d34185926b.jpg b/parse/dev/E01k9048soZ/images/699b0f1210062183978fe46e8634cccdbedef691e47a3e9eb136e8d34185926b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7f6ee6a1500c76272c442b12016f2d1ff63eca4c --- /dev/null +++ b/parse/dev/E01k9048soZ/images/699b0f1210062183978fe46e8634cccdbedef691e47a3e9eb136e8d34185926b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24e97beec8927a2bbc77ece4e62d6c6307d31726570aeeb61fb13a3f9439637c +size 122171 diff --git a/parse/dev/E01k9048soZ/images/722bb9e2215091818c8566ad42a5ddb3d8099e83665a41e5a73a563c77131c7c.jpg b/parse/dev/E01k9048soZ/images/722bb9e2215091818c8566ad42a5ddb3d8099e83665a41e5a73a563c77131c7c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..64e715ade996951fc277da2954ee13b74a947711 --- /dev/null +++ b/parse/dev/E01k9048soZ/images/722bb9e2215091818c8566ad42a5ddb3d8099e83665a41e5a73a563c77131c7c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3ae513455b72879e03e0d9dbbab6b7d215a4f87b5d5a41dd9152475beab80baf +size 67444 diff --git a/parse/dev/E01k9048soZ/images/88e9bbd9246837d3da1f486ea70459d3c34b69dba4dfd7830f2fc84e02bbb626.jpg b/parse/dev/E01k9048soZ/images/88e9bbd9246837d3da1f486ea70459d3c34b69dba4dfd7830f2fc84e02bbb626.jpg new file mode 100644 index 0000000000000000000000000000000000000000..75a08db5fe38c8acb6a043dececbd20a7e54053b --- /dev/null +++ b/parse/dev/E01k9048soZ/images/88e9bbd9246837d3da1f486ea70459d3c34b69dba4dfd7830f2fc84e02bbb626.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:61bb2161c8a74577c572acc7535b80941aa071ab3bb4b4c863a3c9af9098bba4 +size 20216 diff --git a/parse/dev/E01k9048soZ/images/b3e20e8881f8e5699ab32cce15964dcf3f7941a1c0cb3be5dea2032a85938f7e.jpg b/parse/dev/E01k9048soZ/images/b3e20e8881f8e5699ab32cce15964dcf3f7941a1c0cb3be5dea2032a85938f7e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0521e13073dae96c558e813f9c5f9ffcd5bef2f8 --- /dev/null +++ b/parse/dev/E01k9048soZ/images/b3e20e8881f8e5699ab32cce15964dcf3f7941a1c0cb3be5dea2032a85938f7e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a27b3540bcd849980ce25efe558fcf14532d155b5b93f97343adc96c9b26efb +size 29363 diff --git a/parse/dev/E01k9048soZ/images/c7054abff91937690c1564a2fa81d48b2c2b155cd48ffc42e6d517d0d4d2b56f.jpg b/parse/dev/E01k9048soZ/images/c7054abff91937690c1564a2fa81d48b2c2b155cd48ffc42e6d517d0d4d2b56f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c0879b8f07f1827297da5d0ecf75600dbdb48008 --- /dev/null +++ b/parse/dev/E01k9048soZ/images/c7054abff91937690c1564a2fa81d48b2c2b155cd48ffc42e6d517d0d4d2b56f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23b649dce1c3157999e34548f7f515e323fadcb40dd5631e769e0533e5c55f8c +size 109894 diff --git a/parse/dev/E01k9048soZ/images/f36332e5ad41732e4e14f2aec2f8157ff441af2a7bda0e05e650bbf15b129693.jpg b/parse/dev/E01k9048soZ/images/f36332e5ad41732e4e14f2aec2f8157ff441af2a7bda0e05e650bbf15b129693.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e406637a8426d510b02fd455888cdc1e636aeeb2 --- /dev/null +++ b/parse/dev/E01k9048soZ/images/f36332e5ad41732e4e14f2aec2f8157ff441af2a7bda0e05e650bbf15b129693.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:515c9caa2872e1f338d8293d6b4496db2c3a32b0135ad4fee6e4980229113ba9 +size 212154 diff --git a/parse/dev/EcGGFkNTxdJ/images/0ac287e63f9a8b3eb326e180bc923f589249b906612a99dd6387abf3c279dad2.jpg b/parse/dev/EcGGFkNTxdJ/images/0ac287e63f9a8b3eb326e180bc923f589249b906612a99dd6387abf3c279dad2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..97e65cec2435cba9039ade7e30481a8683ed5006 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/0ac287e63f9a8b3eb326e180bc923f589249b906612a99dd6387abf3c279dad2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14dd36c98728448c3e34bc7481960a0a6d6454d4585b73c9fc55a6b92f0d419e +size 5514 diff --git a/parse/dev/EcGGFkNTxdJ/images/109dd7203aebce18e3a21fb48981227cfaf7e655c2fc9f6515e5afa46f18b83f.jpg b/parse/dev/EcGGFkNTxdJ/images/109dd7203aebce18e3a21fb48981227cfaf7e655c2fc9f6515e5afa46f18b83f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7d2f2d9945e7a312044131a98c1904b02b89b191 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/109dd7203aebce18e3a21fb48981227cfaf7e655c2fc9f6515e5afa46f18b83f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6bc56fdd698959d50f3dfee56e41423946e3b785333985928a289241eef43d1a +size 4399 diff --git a/parse/dev/EcGGFkNTxdJ/images/10e11e170c2d55e1d97023be2d4345eae0d9af69e83ea03475ef969be26c5009.jpg b/parse/dev/EcGGFkNTxdJ/images/10e11e170c2d55e1d97023be2d4345eae0d9af69e83ea03475ef969be26c5009.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9dabf21bcab90c38ac58ba2b7651f4376c3f6e9f --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/10e11e170c2d55e1d97023be2d4345eae0d9af69e83ea03475ef969be26c5009.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c0293d5e1e2b941d91577216d554128c430c7050f4c6efea726e895b89f54e34 +size 6704 diff --git a/parse/dev/EcGGFkNTxdJ/images/1452c170b412454eb6dd873f2f328454cb120118ab9b06195da39d6bd27a20f7.jpg b/parse/dev/EcGGFkNTxdJ/images/1452c170b412454eb6dd873f2f328454cb120118ab9b06195da39d6bd27a20f7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e0ba04b27455affaed8c955e19d7a3a7bae8c443 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/1452c170b412454eb6dd873f2f328454cb120118ab9b06195da39d6bd27a20f7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3c0949d15351e460e36fcd676056a8ff838b599074ae8ed4fc07e27a74af9896 +size 10427 diff --git a/parse/dev/EcGGFkNTxdJ/images/185850e23755a962234386c92de53a5f84881290a79af4e891a1795ed8d575c5.jpg b/parse/dev/EcGGFkNTxdJ/images/185850e23755a962234386c92de53a5f84881290a79af4e891a1795ed8d575c5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d598b6bb93abebba2dbcb4e68c8eaa395d9a1aac --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/185850e23755a962234386c92de53a5f84881290a79af4e891a1795ed8d575c5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:90e53dc55bfa3dadb202c92a2502c5c7855c597177c2bc4ca64e06085227719e +size 27812 diff --git a/parse/dev/EcGGFkNTxdJ/images/1b14dcb8d81300539901a7050d8d12253b8e8b7c62c4284f40c904a258719326.jpg b/parse/dev/EcGGFkNTxdJ/images/1b14dcb8d81300539901a7050d8d12253b8e8b7c62c4284f40c904a258719326.jpg new file mode 100644 index 0000000000000000000000000000000000000000..86f74c18bb43dfe0fd28172eb15a1291721aa54b --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/1b14dcb8d81300539901a7050d8d12253b8e8b7c62c4284f40c904a258719326.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b768abc6c7fe98857d2c0287742d9f47656ee1da4b37ac7345a256f7f591d388 +size 8702 diff --git a/parse/dev/EcGGFkNTxdJ/images/1d6c69f171c1b9238353477863507a76d315b894d9955bae53556095dd7c5548.jpg b/parse/dev/EcGGFkNTxdJ/images/1d6c69f171c1b9238353477863507a76d315b894d9955bae53556095dd7c5548.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2c6bbeda0339a5b40f8b8ca6cf409656394d5384 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/1d6c69f171c1b9238353477863507a76d315b894d9955bae53556095dd7c5548.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19fbe31236f8fcb164ab5acb6b80b88c6b5f69a9dd56d7672aaded68253ea7ed +size 6082 diff --git a/parse/dev/EcGGFkNTxdJ/images/1de86a621641e50e847765246cabf57dd2e97806e6852bf6cc35b7aaa6720c4e.jpg b/parse/dev/EcGGFkNTxdJ/images/1de86a621641e50e847765246cabf57dd2e97806e6852bf6cc35b7aaa6720c4e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a8e81a6653051a6eefbaec88e743756f262455d3 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/1de86a621641e50e847765246cabf57dd2e97806e6852bf6cc35b7aaa6720c4e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b441fe4e9d374b8037544540b9dd9d5a7920015b6cd6493fc5fbb09b39a9bc2 +size 15084 diff --git a/parse/dev/EcGGFkNTxdJ/images/2431b6ede9e6459eafad8e55eb103298cc15df1197574a392964954cdb8a6b42.jpg b/parse/dev/EcGGFkNTxdJ/images/2431b6ede9e6459eafad8e55eb103298cc15df1197574a392964954cdb8a6b42.jpg new file mode 100644 index 0000000000000000000000000000000000000000..111597e27b16b0d44c83625fe77aedf1b2d473fd --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/2431b6ede9e6459eafad8e55eb103298cc15df1197574a392964954cdb8a6b42.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a68755ffcb5d6564df35ca861152210a111dc9ebb5482c4f4ecbc5425562c7ba +size 21037 diff --git a/parse/dev/EcGGFkNTxdJ/images/2455ee657015e108afc7bce52f94abcc0d0417d00b89b57010bb619c988c3609.jpg b/parse/dev/EcGGFkNTxdJ/images/2455ee657015e108afc7bce52f94abcc0d0417d00b89b57010bb619c988c3609.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7ad712561a799b8b712df08620cfc75de8232ea5 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/2455ee657015e108afc7bce52f94abcc0d0417d00b89b57010bb619c988c3609.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27f6cfee0e1da11b623572c3d30229def60700c1a567b3ad34ac4bc5c73fa3e9 +size 6008 diff --git a/parse/dev/EcGGFkNTxdJ/images/24b6a552a211a1112c0a0d5cf9460ea9821a01134a2435e173b2307119c62125.jpg b/parse/dev/EcGGFkNTxdJ/images/24b6a552a211a1112c0a0d5cf9460ea9821a01134a2435e173b2307119c62125.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7be7d6f7687afa184990945428fd4b0c74edbe9c --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/24b6a552a211a1112c0a0d5cf9460ea9821a01134a2435e173b2307119c62125.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:177088bf1c34bbb642443130ee8cb26f82f1fb623231e6d208cf25d64217cfc2 +size 48530 diff --git a/parse/dev/EcGGFkNTxdJ/images/26ad445b4959ec57c79735b1f22403b5c55f3e91f3de3313f672fe0d050dd826.jpg b/parse/dev/EcGGFkNTxdJ/images/26ad445b4959ec57c79735b1f22403b5c55f3e91f3de3313f672fe0d050dd826.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1407c5d2fe403477cecaa17499e56b0465cfc02d --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/26ad445b4959ec57c79735b1f22403b5c55f3e91f3de3313f672fe0d050dd826.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9546248610efab6c6a14f00e70ade92276c48e0627bc5720e52665e7247e9a32 +size 4213 diff --git a/parse/dev/EcGGFkNTxdJ/images/27a9e715219ce194ab8f7fdde31746fb93cb6706252a2f88658b820df6d17ea4.jpg b/parse/dev/EcGGFkNTxdJ/images/27a9e715219ce194ab8f7fdde31746fb93cb6706252a2f88658b820df6d17ea4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ec0ac4403006194638bf787984a6cf2f95049131 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/27a9e715219ce194ab8f7fdde31746fb93cb6706252a2f88658b820df6d17ea4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7bcf2e599dd6833ca5cbcb4657544a4703f409322f51812ef0b4f3d709e2130b +size 8532 diff --git a/parse/dev/EcGGFkNTxdJ/images/2a4d4ba88795875ca0629e584f21bc68074af2398bc2876aac0df5815dfee42c.jpg b/parse/dev/EcGGFkNTxdJ/images/2a4d4ba88795875ca0629e584f21bc68074af2398bc2876aac0df5815dfee42c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cc972b207773b3d0bc79f824a08ffb2cb0a7b1a7 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/2a4d4ba88795875ca0629e584f21bc68074af2398bc2876aac0df5815dfee42c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20272c9961c5ebf71d2f3563f6bd97e8849c127e759ed44b6d9b10cac211465f +size 32957 diff --git a/parse/dev/EcGGFkNTxdJ/images/322337707658ee88873f127c397e4de76e5ffff585527789b75703a3f8787b13.jpg b/parse/dev/EcGGFkNTxdJ/images/322337707658ee88873f127c397e4de76e5ffff585527789b75703a3f8787b13.jpg new file mode 100644 index 0000000000000000000000000000000000000000..038a4ed5292053f0f1fc75fd879f756e1bf6e731 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/322337707658ee88873f127c397e4de76e5ffff585527789b75703a3f8787b13.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3dd67cc3d075ef76ba4111db5365de463216816c8e56e000eae732ca4226acb3 +size 11627 diff --git a/parse/dev/EcGGFkNTxdJ/images/3283f952f5d73bf5641da6012c6c84ab616e4e7412f4872fee6224c57e79d60a.jpg b/parse/dev/EcGGFkNTxdJ/images/3283f952f5d73bf5641da6012c6c84ab616e4e7412f4872fee6224c57e79d60a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bf239566f31f21795906e8ccd3bbe0ec2bb70bc8 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/3283f952f5d73bf5641da6012c6c84ab616e4e7412f4872fee6224c57e79d60a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:873cd6cf8bdc9b95baab55952e70f968141443f9f18c90d47fd98595a95d7422 +size 4576 diff --git a/parse/dev/EcGGFkNTxdJ/images/3686de664f8fd93f87626f7cd60c8c3319e8fa2826169e891d59ab0ee78b3e81.jpg b/parse/dev/EcGGFkNTxdJ/images/3686de664f8fd93f87626f7cd60c8c3319e8fa2826169e891d59ab0ee78b3e81.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1018f64fa58ba4e3243a3e1049e53d3275d6d950 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/3686de664f8fd93f87626f7cd60c8c3319e8fa2826169e891d59ab0ee78b3e81.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:554541daa9bfb507bde327845f49a150c7f4ba8d6e8b2a438777a0f88ac47ff3 +size 5366 diff --git a/parse/dev/EcGGFkNTxdJ/images/37bc48cebe49f98ba87d8b6e216814420ae79f4b65c1f13a82c4d130efb655d8.jpg b/parse/dev/EcGGFkNTxdJ/images/37bc48cebe49f98ba87d8b6e216814420ae79f4b65c1f13a82c4d130efb655d8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..067af83c3b3dbc580a165c3a8a626c5a1fa42ad5 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/37bc48cebe49f98ba87d8b6e216814420ae79f4b65c1f13a82c4d130efb655d8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f85f668f56f52a4f0bbf3680c99c5f5749f46019b173c94a6ec7af66c048fe94 +size 4529 diff --git a/parse/dev/EcGGFkNTxdJ/images/39f0e3de2d21529cea9f151a42be732c32fa3fdeed5588bcf597c3312ef4a9ee.jpg b/parse/dev/EcGGFkNTxdJ/images/39f0e3de2d21529cea9f151a42be732c32fa3fdeed5588bcf597c3312ef4a9ee.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d1a12837b1bf9b5d0b2940f7df794592d5ffd8a1 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/39f0e3de2d21529cea9f151a42be732c32fa3fdeed5588bcf597c3312ef4a9ee.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b937a73f11497a59ea657086c4a61d94d6ea369238897b6bd248360ce239e9c +size 5480 diff --git a/parse/dev/EcGGFkNTxdJ/images/3e95346dd3e4da1c089b4540f29da64335a84a4cdda64d6b92e42105f85d0685.jpg b/parse/dev/EcGGFkNTxdJ/images/3e95346dd3e4da1c089b4540f29da64335a84a4cdda64d6b92e42105f85d0685.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c1c55d142c88d4f0d8d5c8ed2907f37444a4054e --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/3e95346dd3e4da1c089b4540f29da64335a84a4cdda64d6b92e42105f85d0685.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04cad4861a36d6dcb2e990938e6117bd904d3435855f7f394d2a952c539c9449 +size 15837 diff --git a/parse/dev/EcGGFkNTxdJ/images/3ee1d24a8e996a98e5e18cfa0b156579ebc3a213b8fb0c2424ac89735a1aed14.jpg b/parse/dev/EcGGFkNTxdJ/images/3ee1d24a8e996a98e5e18cfa0b156579ebc3a213b8fb0c2424ac89735a1aed14.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a7b6f00cbdbb844e931556c575c411d72e28bc57 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/3ee1d24a8e996a98e5e18cfa0b156579ebc3a213b8fb0c2424ac89735a1aed14.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ecbcfe4e3f0f4c6282d9cd94f6be6762d07639cd521ed04e0060d2bfd1740920 +size 10011 diff --git a/parse/dev/EcGGFkNTxdJ/images/3ffb3967cd817938d081e4998f860801c4ff58e8ab095b1d8a5f1ea84acd8ef3.jpg b/parse/dev/EcGGFkNTxdJ/images/3ffb3967cd817938d081e4998f860801c4ff58e8ab095b1d8a5f1ea84acd8ef3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2f299b8b324f461077d7f0fdc62ec2a7508d9217 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/3ffb3967cd817938d081e4998f860801c4ff58e8ab095b1d8a5f1ea84acd8ef3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9d4a2d0910bb72d05267fc553cc58dc5c8a57cac93eba19b2efc2af0426cb84e +size 2196 diff --git a/parse/dev/EcGGFkNTxdJ/images/47d7438c4489866cc0308225cc633486eaf82272ff3cc6be78601fce18449e22.jpg b/parse/dev/EcGGFkNTxdJ/images/47d7438c4489866cc0308225cc633486eaf82272ff3cc6be78601fce18449e22.jpg new file mode 100644 index 0000000000000000000000000000000000000000..79f68a99b9703c6344aedba428e1492d541222bd --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/47d7438c4489866cc0308225cc633486eaf82272ff3cc6be78601fce18449e22.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b991f933abc85c46e2f90141ebe7695b89d13d1ae51f8e1b7a1590445b765d19 +size 7176 diff --git a/parse/dev/EcGGFkNTxdJ/images/4ac528bf0345c88d4ca22c5cb9fbbc5124dc053c5d74b2dc1d88cd4f9b79a7fb.jpg b/parse/dev/EcGGFkNTxdJ/images/4ac528bf0345c88d4ca22c5cb9fbbc5124dc053c5d74b2dc1d88cd4f9b79a7fb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cd01938a708c0945be67f9188bc3bdf6c8ec8766 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/4ac528bf0345c88d4ca22c5cb9fbbc5124dc053c5d74b2dc1d88cd4f9b79a7fb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a7b7f54fee787b6ec7ffb2bf36f379c99fc9215f731ecdae7ec28fed5e1aa2e8 +size 15525 diff --git a/parse/dev/EcGGFkNTxdJ/images/4f52edf18ea00b419e7a2062cde9f96227668cabdb846f8fd1383955778b489b.jpg b/parse/dev/EcGGFkNTxdJ/images/4f52edf18ea00b419e7a2062cde9f96227668cabdb846f8fd1383955778b489b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..697e9a5b62a2eda50995a6dda41fb68af3093d59 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/4f52edf18ea00b419e7a2062cde9f96227668cabdb846f8fd1383955778b489b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b5179294c858279e068e3347828b5506a445c2b0f01d267c827f251c5545f9c +size 18521 diff --git a/parse/dev/EcGGFkNTxdJ/images/51453831f14124be4ddde250a3eb42b0676185b0c1d69c19e8b6fd0a122d7dbe.jpg b/parse/dev/EcGGFkNTxdJ/images/51453831f14124be4ddde250a3eb42b0676185b0c1d69c19e8b6fd0a122d7dbe.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2f9d4d0b2ba255a7ed3160f67158a10777f68957 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/51453831f14124be4ddde250a3eb42b0676185b0c1d69c19e8b6fd0a122d7dbe.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d27583a91e281a37eaec203ba264abbe3c4000ec7478e3d62b697f29b62294f +size 53452 diff --git a/parse/dev/EcGGFkNTxdJ/images/526b4ff2b766cbfda8efceb4a4eb64ea4e1151da1ff8cd722a5260f091c707b7.jpg b/parse/dev/EcGGFkNTxdJ/images/526b4ff2b766cbfda8efceb4a4eb64ea4e1151da1ff8cd722a5260f091c707b7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..261a5c7e416a26e695178f7332ea594754413f7c --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/526b4ff2b766cbfda8efceb4a4eb64ea4e1151da1ff8cd722a5260f091c707b7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:754ad717707f94593430f558a1bd20ca75d0a7b57d48fcdb766f86086d3c2cb2 +size 7224 diff --git a/parse/dev/EcGGFkNTxdJ/images/5687c6a71d30adb2005d5b682e5db69727e74984e1ed0e614aec8633d95b93a7.jpg b/parse/dev/EcGGFkNTxdJ/images/5687c6a71d30adb2005d5b682e5db69727e74984e1ed0e614aec8633d95b93a7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..41148689dd51e46e4db4894e9a3c37de86b180bb --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/5687c6a71d30adb2005d5b682e5db69727e74984e1ed0e614aec8633d95b93a7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:712f56ff95e1e751d8a7a11c7541c578d5825565ae0b7c2ce04c7b4649f178d5 +size 5812 diff --git a/parse/dev/EcGGFkNTxdJ/images/6185e213b730a9cf68528c2707c834c529e0774c9c48eb6527b35d2934870343.jpg b/parse/dev/EcGGFkNTxdJ/images/6185e213b730a9cf68528c2707c834c529e0774c9c48eb6527b35d2934870343.jpg new file mode 100644 index 0000000000000000000000000000000000000000..26680ef68df0bbfc78f9d5c3c28d296281747b56 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/6185e213b730a9cf68528c2707c834c529e0774c9c48eb6527b35d2934870343.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3e71197e5104350a57bacc551cda7d429592d3e2dab5439449486c32c907449 +size 8009 diff --git a/parse/dev/EcGGFkNTxdJ/images/6219bc681b9cffee219105a061a005721d46ac7e0778f218ff03915b332ff7ea.jpg b/parse/dev/EcGGFkNTxdJ/images/6219bc681b9cffee219105a061a005721d46ac7e0778f218ff03915b332ff7ea.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b55ecaddc7c5b74324e004e6e1e7bb07a4826b96 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/6219bc681b9cffee219105a061a005721d46ac7e0778f218ff03915b332ff7ea.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b199714106d422a35b5cdc64b670c415067487b447c108fac1c81aef258c844 +size 228852 diff --git a/parse/dev/EcGGFkNTxdJ/images/63c7bfd86eaf56acc3fa93602ef21fcae2652f5e5dc2f09cd11017cc41c8b7b5.jpg b/parse/dev/EcGGFkNTxdJ/images/63c7bfd86eaf56acc3fa93602ef21fcae2652f5e5dc2f09cd11017cc41c8b7b5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..db88d26d9a33e9a85fa418b1b5cb9a2a58f5b2fd --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/63c7bfd86eaf56acc3fa93602ef21fcae2652f5e5dc2f09cd11017cc41c8b7b5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f701bbeaf18d1e3bbcbb2496c77f1f2b0a3cdd4b1407c68b2361e93a0879e0bf +size 12738 diff --git a/parse/dev/EcGGFkNTxdJ/images/6498ec7b1be8110093621a61297b862492708531c94943bbc9c8dbb68a4cefde.jpg b/parse/dev/EcGGFkNTxdJ/images/6498ec7b1be8110093621a61297b862492708531c94943bbc9c8dbb68a4cefde.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c21c9f106a2842a0bc00d0a7092a96c17b3176c3 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/6498ec7b1be8110093621a61297b862492708531c94943bbc9c8dbb68a4cefde.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07199bddf12b99d454313a033f64337b5c82ecd2eaeee1a31f718e3317de806e +size 47476 diff --git a/parse/dev/EcGGFkNTxdJ/images/658c257cdae97f15b52501179090c97b83272eeb7b66a0ea3472647c08292526.jpg b/parse/dev/EcGGFkNTxdJ/images/658c257cdae97f15b52501179090c97b83272eeb7b66a0ea3472647c08292526.jpg new file mode 100644 index 0000000000000000000000000000000000000000..03463655efa63de58a1709ec25ee49c828b609a2 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/658c257cdae97f15b52501179090c97b83272eeb7b66a0ea3472647c08292526.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ff6885a4e7a8e847b9812d94a53944a32449e6ada1717dedbe7911409a965c06 +size 8195 diff --git a/parse/dev/EcGGFkNTxdJ/images/69933ca8f36ab9354476fd712dabaa92dd52c49fba592557c8fd5ce94e649486.jpg b/parse/dev/EcGGFkNTxdJ/images/69933ca8f36ab9354476fd712dabaa92dd52c49fba592557c8fd5ce94e649486.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d74f534fab4fcac6679db9c58b78b9b42f94a263 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/69933ca8f36ab9354476fd712dabaa92dd52c49fba592557c8fd5ce94e649486.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14c8197444a96d03b9148103bb34c8733351dc77dde996160cc4303099da5c2d +size 3678 diff --git a/parse/dev/EcGGFkNTxdJ/images/6ca71e7f1845ce8452c0e36bea5a84006cc21510d01616ab9642b5d663655ad5.jpg b/parse/dev/EcGGFkNTxdJ/images/6ca71e7f1845ce8452c0e36bea5a84006cc21510d01616ab9642b5d663655ad5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..312fa8a1f49abef9278af7f76d7a01242a28e5e7 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/6ca71e7f1845ce8452c0e36bea5a84006cc21510d01616ab9642b5d663655ad5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59f4a75dbd3aaefd5930515f49d2bc0d50479b9f84b5fc39bfe9f2cdbed2a2c4 +size 25014 diff --git a/parse/dev/EcGGFkNTxdJ/images/6cff0e2d6fa1596e0a2468866073a79f9f090fb2f0d2cdf4ecefd9c00b640d8b.jpg b/parse/dev/EcGGFkNTxdJ/images/6cff0e2d6fa1596e0a2468866073a79f9f090fb2f0d2cdf4ecefd9c00b640d8b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d93b8066e2a69835113c51cd721a526708ae9ba0 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/6cff0e2d6fa1596e0a2468866073a79f9f090fb2f0d2cdf4ecefd9c00b640d8b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:01cb922fe968520dacfaeb1d5d9807e35602d7f61122037476f950e7bc5b44e1 +size 9830 diff --git a/parse/dev/EcGGFkNTxdJ/images/6dd2bd39688e0dc65a2611d735d4b9ca209b045eae0c729c31ed4166be752793.jpg b/parse/dev/EcGGFkNTxdJ/images/6dd2bd39688e0dc65a2611d735d4b9ca209b045eae0c729c31ed4166be752793.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bb5a2fa1570e2339f9e38bab92139792bfb24bc4 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/6dd2bd39688e0dc65a2611d735d4b9ca209b045eae0c729c31ed4166be752793.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f971b549f18f5523bfdb543489722f881b669eb681f56658a1b17e0d0f67b1a +size 9880 diff --git a/parse/dev/EcGGFkNTxdJ/images/70aedec04c15f3091232e327d31b8584a07dea4f4850bf7fd75244cf223b9d1a.jpg b/parse/dev/EcGGFkNTxdJ/images/70aedec04c15f3091232e327d31b8584a07dea4f4850bf7fd75244cf223b9d1a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..11b2c3a631be992caaba04ac742165eaa0cb4886 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/70aedec04c15f3091232e327d31b8584a07dea4f4850bf7fd75244cf223b9d1a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb72f467b0e1521d7924ebb03b7bb79f9c780b9b9a411c85fa0e424bf3bf5caf +size 8822 diff --git a/parse/dev/EcGGFkNTxdJ/images/761a2b89d7535e08ff5132bb85a0f7829351ac4a5e5d1358438dddf002180592.jpg b/parse/dev/EcGGFkNTxdJ/images/761a2b89d7535e08ff5132bb85a0f7829351ac4a5e5d1358438dddf002180592.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a873d19018f01243a747979cce19f066f9173950 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/761a2b89d7535e08ff5132bb85a0f7829351ac4a5e5d1358438dddf002180592.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d1dbbccc29098e0ac911f0e1977c4cf6031fce023dabcb29c18024d7c519e6c2 +size 8727 diff --git a/parse/dev/EcGGFkNTxdJ/images/7817e1c041f55e0debc99c516d1499d1fa2504e478c42a7678c61317a9d1da05.jpg b/parse/dev/EcGGFkNTxdJ/images/7817e1c041f55e0debc99c516d1499d1fa2504e478c42a7678c61317a9d1da05.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1cfdc463f6a5bffc24841b9d1fbeb2cb8db58458 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/7817e1c041f55e0debc99c516d1499d1fa2504e478c42a7678c61317a9d1da05.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb7067d51cf864e09ce673f1df1a34349f0c06572e62887a8926b672051e03b7 +size 7972 diff --git a/parse/dev/EcGGFkNTxdJ/images/7924a3e8696e8590eb6fa9322acd006a3d663af161ffb513f8f0f68728377756.jpg b/parse/dev/EcGGFkNTxdJ/images/7924a3e8696e8590eb6fa9322acd006a3d663af161ffb513f8f0f68728377756.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2f7c699a5d5ec78dc97d151f80269704d353c9df --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/7924a3e8696e8590eb6fa9322acd006a3d663af161ffb513f8f0f68728377756.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fae3e4197424c750586bbe56b4c5a995ac5a3eabcd0f109c457bfec2e2b70c8b +size 7273 diff --git a/parse/dev/EcGGFkNTxdJ/images/7eb4e2920938e0371d02393c6aa96d5c0304456214a97778f07a93fa38befc1a.jpg b/parse/dev/EcGGFkNTxdJ/images/7eb4e2920938e0371d02393c6aa96d5c0304456214a97778f07a93fa38befc1a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..27a7b4c41e0282deca43885fd035acd203612a34 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/7eb4e2920938e0371d02393c6aa96d5c0304456214a97778f07a93fa38befc1a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a8240d38154f4ab8831724cec345a46a94313628aa3007a50cda5c074f240bc +size 5904 diff --git a/parse/dev/EcGGFkNTxdJ/images/7f5a853881e6ce656fd70dc47be5f3014d1277b84209d14929a62ce4729e6239.jpg b/parse/dev/EcGGFkNTxdJ/images/7f5a853881e6ce656fd70dc47be5f3014d1277b84209d14929a62ce4729e6239.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0dfb5979114af60d04da02d848a721957d7ca35b --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/7f5a853881e6ce656fd70dc47be5f3014d1277b84209d14929a62ce4729e6239.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77c4e557ef915e6e6e289dde2d8eaa289618ead6e87e36b56cc2f71b60656989 +size 45871 diff --git a/parse/dev/EcGGFkNTxdJ/images/8225ba1b78c77f9fd706eeb4eedb144f370d5ee75fffed17db8365ef71611c79.jpg b/parse/dev/EcGGFkNTxdJ/images/8225ba1b78c77f9fd706eeb4eedb144f370d5ee75fffed17db8365ef71611c79.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7b9a5958cc08252d2aada5fb74771bbc5f27b9e6 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/8225ba1b78c77f9fd706eeb4eedb144f370d5ee75fffed17db8365ef71611c79.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:253b900bd44bbe845e63eaddceb7e5447f63604c84cc4927b204e63baa846eca +size 38265 diff --git a/parse/dev/EcGGFkNTxdJ/images/85b559cbe923dc345a6c5a7f7cfee900e24c7deb985c4a2c8d83017ab6f2dd1d.jpg b/parse/dev/EcGGFkNTxdJ/images/85b559cbe923dc345a6c5a7f7cfee900e24c7deb985c4a2c8d83017ab6f2dd1d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..06fd24b3f1b2af06e1ec541121a33007e2bbfbfa --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/85b559cbe923dc345a6c5a7f7cfee900e24c7deb985c4a2c8d83017ab6f2dd1d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6ce951554fc0d37d166ee4df036c7fcd806c8bb50b163f4489a8f4c2cb6bef5 +size 14881 diff --git a/parse/dev/EcGGFkNTxdJ/images/862ea309a38178f270f6c2c127392719b54d968d4ca51aaecca7e8b425bfeb79.jpg b/parse/dev/EcGGFkNTxdJ/images/862ea309a38178f270f6c2c127392719b54d968d4ca51aaecca7e8b425bfeb79.jpg new file mode 100644 index 0000000000000000000000000000000000000000..727215995261625d7d6daf1f5adbd36fabdd81a1 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/862ea309a38178f270f6c2c127392719b54d968d4ca51aaecca7e8b425bfeb79.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6bfa1a4416fbed149b9667929fe06f63bd7298bd46793c0b0f951e5c93e4fd08 +size 60646 diff --git a/parse/dev/EcGGFkNTxdJ/images/90bebfb003a44772d2bee75defb9487f585bac230b344151e05bc96578206e40.jpg b/parse/dev/EcGGFkNTxdJ/images/90bebfb003a44772d2bee75defb9487f585bac230b344151e05bc96578206e40.jpg new file mode 100644 index 0000000000000000000000000000000000000000..07dcd13a43eccfd2e4f65ea111d794f8987604d1 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/90bebfb003a44772d2bee75defb9487f585bac230b344151e05bc96578206e40.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:736aa37fbf71c10316dbce4144251ee9c9d5b37a26035063335df0f5fab30885 +size 14156 diff --git a/parse/dev/EcGGFkNTxdJ/images/943c8ff1577a20bb7b9a663eba92977dffb1c116bb0be119f67e1e958082894d.jpg b/parse/dev/EcGGFkNTxdJ/images/943c8ff1577a20bb7b9a663eba92977dffb1c116bb0be119f67e1e958082894d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b495fe8de59d0eeaa166772c0c968290fbfba939 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/943c8ff1577a20bb7b9a663eba92977dffb1c116bb0be119f67e1e958082894d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:771d2160aacb21b486b6fbc514b391509d502ab54e15f2463c6ff5b5fe8501a8 +size 7750 diff --git a/parse/dev/EcGGFkNTxdJ/images/9bb3c1aa1d085ec246bc88f9cae03d5385193a9dacddda0abf0d5f4923609462.jpg b/parse/dev/EcGGFkNTxdJ/images/9bb3c1aa1d085ec246bc88f9cae03d5385193a9dacddda0abf0d5f4923609462.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f68e79b992198ddb890f932135508ea070aa2d1d --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/9bb3c1aa1d085ec246bc88f9cae03d5385193a9dacddda0abf0d5f4923609462.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b0fb2291ae2a4ba13e7edc9f740935f75c981b678434b4f850e0667f8dbf9f37 +size 4038 diff --git a/parse/dev/EcGGFkNTxdJ/images/9e023f47db794be5f1120b87db0d2d9df49e0a883efffa48f1dc3f5aa6b89af7.jpg b/parse/dev/EcGGFkNTxdJ/images/9e023f47db794be5f1120b87db0d2d9df49e0a883efffa48f1dc3f5aa6b89af7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5ddaafa279231059d67ea4e70968e3517c073728 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/9e023f47db794be5f1120b87db0d2d9df49e0a883efffa48f1dc3f5aa6b89af7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58c23f8b09ca8e959f71fa5c8fa5ff79a19094420a1e55398288967ef16aa17f +size 4161 diff --git a/parse/dev/EcGGFkNTxdJ/images/a393e10ff0604b50ed87ea7ca84d1c8914f223208b463dd08462e6b8e87eb832.jpg b/parse/dev/EcGGFkNTxdJ/images/a393e10ff0604b50ed87ea7ca84d1c8914f223208b463dd08462e6b8e87eb832.jpg new file mode 100644 index 0000000000000000000000000000000000000000..234ab5551c418192a7c50d30d6c347fb6ce1db41 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/a393e10ff0604b50ed87ea7ca84d1c8914f223208b463dd08462e6b8e87eb832.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf7e6b21bfef758b57e06b85e15313d5e4ac32a3f66aa6a44de30444f378dd68 +size 5458 diff --git a/parse/dev/EcGGFkNTxdJ/images/a604a3b6cb03792ce5e1a3469c333809c00eaee5bad9270752627079fb38df12.jpg b/parse/dev/EcGGFkNTxdJ/images/a604a3b6cb03792ce5e1a3469c333809c00eaee5bad9270752627079fb38df12.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6857d5dab231814fd095f7982b9ae8f5fd4ce815 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/a604a3b6cb03792ce5e1a3469c333809c00eaee5bad9270752627079fb38df12.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:90ead7a1402a7002150fd4d96b2c4d1519db0226cfc24d81d1da54cfc2f9a159 +size 5555 diff --git a/parse/dev/EcGGFkNTxdJ/images/b119467175a3cc619cc1dd991fd157d0e14965ae318f07fa889649b399de4135.jpg b/parse/dev/EcGGFkNTxdJ/images/b119467175a3cc619cc1dd991fd157d0e14965ae318f07fa889649b399de4135.jpg new file mode 100644 index 0000000000000000000000000000000000000000..52f14c59b47d87ca9758ca68060c2e2ace4ab568 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/b119467175a3cc619cc1dd991fd157d0e14965ae318f07fa889649b399de4135.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:513e22ea6a9ab59c3990f6d7380becba73d2a451af44b915b94e7a1ae91a7ea4 +size 15189 diff --git a/parse/dev/EcGGFkNTxdJ/images/b1940acd38e97ce6af68bd75807b4b13fedd1441c44db783366d139fb2b29cf2.jpg b/parse/dev/EcGGFkNTxdJ/images/b1940acd38e97ce6af68bd75807b4b13fedd1441c44db783366d139fb2b29cf2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d09141af5dc0c104936650bf526ed3d76d5713ab --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/b1940acd38e97ce6af68bd75807b4b13fedd1441c44db783366d139fb2b29cf2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:32829a809b1c2b9163751548f94c18f2f8c555e07e525b559158ca7111c03146 +size 6683 diff --git a/parse/dev/EcGGFkNTxdJ/images/b204de2e2a41e07427ddde4503789224ac02bb0ea311e2b97722adcb0170502a.jpg b/parse/dev/EcGGFkNTxdJ/images/b204de2e2a41e07427ddde4503789224ac02bb0ea311e2b97722adcb0170502a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f2e79e4dbaf23d6fcafa87ddf5296fbf3880195e --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/b204de2e2a41e07427ddde4503789224ac02bb0ea311e2b97722adcb0170502a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b4ba23b241a731a4843985be09871348364db83519c671e700b52e8e455da34 +size 5142 diff --git a/parse/dev/EcGGFkNTxdJ/images/b87d3b8120243ad32415189b963d268f4dd68f64b87fcb1e2ae2a5b641ad8e9e.jpg b/parse/dev/EcGGFkNTxdJ/images/b87d3b8120243ad32415189b963d268f4dd68f64b87fcb1e2ae2a5b641ad8e9e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e3a0d472040bee5f9cddbbd4f127a823ef95ba29 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/b87d3b8120243ad32415189b963d268f4dd68f64b87fcb1e2ae2a5b641ad8e9e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b07c14f0d3e1dcb5602026643df2b27663d8f7a997bde6052e771665b54e88a +size 6228 diff --git a/parse/dev/EcGGFkNTxdJ/images/ba434252ac0fcbeeb6c24242ffd3e4e7fed764ddba0a9cc8b80105a21f146a50.jpg b/parse/dev/EcGGFkNTxdJ/images/ba434252ac0fcbeeb6c24242ffd3e4e7fed764ddba0a9cc8b80105a21f146a50.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a3596ba2a5ad3823b99ab3d91c732b6d3fb05705 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/ba434252ac0fcbeeb6c24242ffd3e4e7fed764ddba0a9cc8b80105a21f146a50.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74c11355604ba965e9dc1d6f4d965551ff9ef4e7c45cbc5ebc6c97c937641606 +size 7528 diff --git a/parse/dev/EcGGFkNTxdJ/images/bda01f7603b68f31092333760715d9b01b395df42a1de203a0ebb0a8c93f408f.jpg b/parse/dev/EcGGFkNTxdJ/images/bda01f7603b68f31092333760715d9b01b395df42a1de203a0ebb0a8c93f408f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5cba1f82f7060658ac50d85dbcfa6ba9e9499ed2 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/bda01f7603b68f31092333760715d9b01b395df42a1de203a0ebb0a8c93f408f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c59e2889822e70a9b736ebe29eabf056860e5ba06113755ee91cfd3b3a7f430 +size 6093 diff --git a/parse/dev/EcGGFkNTxdJ/images/c0ab4ef814d89a42119d0ee4cc8d9cd94e5d617db3f4865e1be98fa2b6a90372.jpg b/parse/dev/EcGGFkNTxdJ/images/c0ab4ef814d89a42119d0ee4cc8d9cd94e5d617db3f4865e1be98fa2b6a90372.jpg new file mode 100644 index 0000000000000000000000000000000000000000..19a0ef39926f67809fa401c3e0b82fa5cc5a4f60 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/c0ab4ef814d89a42119d0ee4cc8d9cd94e5d617db3f4865e1be98fa2b6a90372.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fefe867b6344276937b613acf0583644b3139021acd04de1ab6f0407917215d7 +size 3575 diff --git a/parse/dev/EcGGFkNTxdJ/images/c5195d67d5129e95b53a59cc899ffa2f48deab85184a8f6fb932f17e3dbdc43a.jpg b/parse/dev/EcGGFkNTxdJ/images/c5195d67d5129e95b53a59cc899ffa2f48deab85184a8f6fb932f17e3dbdc43a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..970295b2d3542fac84333f73a09a46ccccac083d --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/c5195d67d5129e95b53a59cc899ffa2f48deab85184a8f6fb932f17e3dbdc43a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e081fca44bb3665d87e5da6469f86b2e9f2807af2c099eb1a0442337a381e993 +size 55070 diff --git a/parse/dev/EcGGFkNTxdJ/images/c60c7dc1b3f83530b76b65418dc91c8b693959ff8f259c2dca6ded0459c074af.jpg b/parse/dev/EcGGFkNTxdJ/images/c60c7dc1b3f83530b76b65418dc91c8b693959ff8f259c2dca6ded0459c074af.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3c843631183ff5874c19febfc6f00ad38a19cc2d --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/c60c7dc1b3f83530b76b65418dc91c8b693959ff8f259c2dca6ded0459c074af.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:90a6868f0e22b0e5bdb39790015f848fccdbf02e1c54e29953f89324e43ff5f7 +size 5401 diff --git a/parse/dev/EcGGFkNTxdJ/images/c7c2bc77b4e5f071b61d4f1f9b29b6ee10538c6632e035217b59f879345b0c9d.jpg b/parse/dev/EcGGFkNTxdJ/images/c7c2bc77b4e5f071b61d4f1f9b29b6ee10538c6632e035217b59f879345b0c9d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..54afd69450c35d7f76a47b6886bf1ed259744a12 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/c7c2bc77b4e5f071b61d4f1f9b29b6ee10538c6632e035217b59f879345b0c9d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:28bc39357e292fad358c56d93ffbed46da35881e39451a7fe928db46c074fb25 +size 52769 diff --git a/parse/dev/EcGGFkNTxdJ/images/c94cd1944c14c664de4b6a7d71de0ccc288b9a7947e852a7d18177fd79cf0660.jpg b/parse/dev/EcGGFkNTxdJ/images/c94cd1944c14c664de4b6a7d71de0ccc288b9a7947e852a7d18177fd79cf0660.jpg new file mode 100644 index 0000000000000000000000000000000000000000..00a297d725c628b63e9a917e33f5a679f08dd4ad --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/c94cd1944c14c664de4b6a7d71de0ccc288b9a7947e852a7d18177fd79cf0660.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:550f9411624d7a58156eda00dc34c5fdb1561e408c0921274e7b835060f2ce97 +size 51926 diff --git a/parse/dev/EcGGFkNTxdJ/images/cad8fe38ad7a4381b12836f6b8208e2717d154c1f41237e1c48beaaf19acf568.jpg b/parse/dev/EcGGFkNTxdJ/images/cad8fe38ad7a4381b12836f6b8208e2717d154c1f41237e1c48beaaf19acf568.jpg new file mode 100644 index 0000000000000000000000000000000000000000..07a0167cb6cd9b33f6c2385bef120958aa6fd6f6 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/cad8fe38ad7a4381b12836f6b8208e2717d154c1f41237e1c48beaaf19acf568.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:abb16473f1acea0e04c634ad3d42a0916eeea95d2309a1bb6025beb6ea4dc441 +size 5732 diff --git a/parse/dev/EcGGFkNTxdJ/images/d4cf7fefe34c7508b0fd1b074cb73316c788d7bd7de4e103bc877686082d42e0.jpg b/parse/dev/EcGGFkNTxdJ/images/d4cf7fefe34c7508b0fd1b074cb73316c788d7bd7de4e103bc877686082d42e0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e7706aa6d95411c992113184d696cb6d405f7cda --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/d4cf7fefe34c7508b0fd1b074cb73316c788d7bd7de4e103bc877686082d42e0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9b244ba3f30b85ba81f21c926d9025d2a62d23786a58486a39dfe48a9facd45 +size 9257 diff --git a/parse/dev/EcGGFkNTxdJ/images/d64ed546d25353c4726316f4962cf8c1a3e94e86015705e40b1d2166f19925ab.jpg b/parse/dev/EcGGFkNTxdJ/images/d64ed546d25353c4726316f4962cf8c1a3e94e86015705e40b1d2166f19925ab.jpg new file mode 100644 index 0000000000000000000000000000000000000000..34c0d01aa89d46849f4b8dbaf8eaa78baf3bbb0b --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/d64ed546d25353c4726316f4962cf8c1a3e94e86015705e40b1d2166f19925ab.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fbacdc5be6daa85574c691a8fba49d7514b361c9947f8a8bd88626949875b10f +size 5382 diff --git a/parse/dev/EcGGFkNTxdJ/images/d8b6b686912f3cc8efc0d914cbde563abd107157eab38590f4cf5776aa06901a.jpg b/parse/dev/EcGGFkNTxdJ/images/d8b6b686912f3cc8efc0d914cbde563abd107157eab38590f4cf5776aa06901a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..04b6807dec1b4df832f1a19e707e0ef89c439750 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/d8b6b686912f3cc8efc0d914cbde563abd107157eab38590f4cf5776aa06901a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8585fa69844c58ee284d66ee6359079ad88c54fe6976831d8cbdaf16ad2c208f +size 7513 diff --git a/parse/dev/EcGGFkNTxdJ/images/d9684efb239882a0a42a56091db931331007d454f57a081619a64ca2beb4004a.jpg b/parse/dev/EcGGFkNTxdJ/images/d9684efb239882a0a42a56091db931331007d454f57a081619a64ca2beb4004a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..84d872bf4d364a8397c2451486f0ee67ed451638 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/d9684efb239882a0a42a56091db931331007d454f57a081619a64ca2beb4004a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1a3a6728a5806d2c7eea530280c753ea00a4bb7ba1febc4e01f77cd8f1aba34 +size 40612 diff --git a/parse/dev/EcGGFkNTxdJ/images/dc3377e0553d4b47a061f1c184045fefbe8cbc1263e73ec8f0b61c656f97e882.jpg b/parse/dev/EcGGFkNTxdJ/images/dc3377e0553d4b47a061f1c184045fefbe8cbc1263e73ec8f0b61c656f97e882.jpg new file mode 100644 index 0000000000000000000000000000000000000000..91e9445fa8998791eff2098aa31a75e0ebc64f42 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/dc3377e0553d4b47a061f1c184045fefbe8cbc1263e73ec8f0b61c656f97e882.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5cb261606dadafd84bf4a58356d201eb6933ba5430babd17e74f02a48cc60d8e +size 5503 diff --git a/parse/dev/EcGGFkNTxdJ/images/dcc43dea08c05ab9ba9a037a4621f44e601434b10b20c8a5ac4a642927f3d90f.jpg b/parse/dev/EcGGFkNTxdJ/images/dcc43dea08c05ab9ba9a037a4621f44e601434b10b20c8a5ac4a642927f3d90f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8426f7012dfeb57979166aacd2fd5190c5c5e6a0 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/dcc43dea08c05ab9ba9a037a4621f44e601434b10b20c8a5ac4a642927f3d90f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb1e4eb2da1b16d5f5261dbe6ec3e3b814e7f4591581d6d44cb8fc607816b3b3 +size 26278 diff --git a/parse/dev/EcGGFkNTxdJ/images/dd386f129487761e2d0440f4cab1b3e1d2f52798e944ef884be3fcde089392f3.jpg b/parse/dev/EcGGFkNTxdJ/images/dd386f129487761e2d0440f4cab1b3e1d2f52798e944ef884be3fcde089392f3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4e307863f110df69354a9270c0782e771875e2a5 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/dd386f129487761e2d0440f4cab1b3e1d2f52798e944ef884be3fcde089392f3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8a990900177948982bd9851ba1ef5ab5d1a95fd6acb0f0dc550ef4132a7babf3 +size 19743 diff --git a/parse/dev/EcGGFkNTxdJ/images/e151627b967ca14e12a0b65ec507f2e4bd4acb4f598cd5028782f142f2336a49.jpg b/parse/dev/EcGGFkNTxdJ/images/e151627b967ca14e12a0b65ec507f2e4bd4acb4f598cd5028782f142f2336a49.jpg new file mode 100644 index 0000000000000000000000000000000000000000..78e3f6f44da13604726debad3f8fb2a249006619 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/e151627b967ca14e12a0b65ec507f2e4bd4acb4f598cd5028782f142f2336a49.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2499bbd25459a5a901fac98bb1acc560ee1429de185fadce173ab55b448e8419 +size 2242 diff --git a/parse/dev/EcGGFkNTxdJ/images/e57d75721e0c55a637d15501ce89899af166bcbf0b1d9a6a8ea166b10f911084.jpg b/parse/dev/EcGGFkNTxdJ/images/e57d75721e0c55a637d15501ce89899af166bcbf0b1d9a6a8ea166b10f911084.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e908732ddf5c5631ec987a6dfda4a6e1e4e1e2ec --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/e57d75721e0c55a637d15501ce89899af166bcbf0b1d9a6a8ea166b10f911084.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c41858ac9a696dd5eb7eebf847fe905f2ffa34218d5b2a82f3f68f78dc573cc +size 41949 diff --git a/parse/dev/EcGGFkNTxdJ/images/e7477959dd71fb8642c1e424e1c4ecd5ad55c8e2ae3c5bb0d9653e6fe15e40d5.jpg b/parse/dev/EcGGFkNTxdJ/images/e7477959dd71fb8642c1e424e1c4ecd5ad55c8e2ae3c5bb0d9653e6fe15e40d5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9e9e93447c2f384c74e2d83a7ef15cd8f75562d5 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/e7477959dd71fb8642c1e424e1c4ecd5ad55c8e2ae3c5bb0d9653e6fe15e40d5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e10b11175fff857e61bb29f7c501ea795081e746b2fc1b11a3a353298b61a31f +size 3109 diff --git a/parse/dev/EcGGFkNTxdJ/images/e7547e528980c44a05baac4c30539e35794f46da62baf389f64c1373502eb179.jpg b/parse/dev/EcGGFkNTxdJ/images/e7547e528980c44a05baac4c30539e35794f46da62baf389f64c1373502eb179.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6756cd40099f79d8a231b2f07bc757bd6bc7265c --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/e7547e528980c44a05baac4c30539e35794f46da62baf389f64c1373502eb179.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f933133e8d15f465a0091c3adb0d41b50663cea969def44ea9ba7f6837b5aad3 +size 8333 diff --git a/parse/dev/EcGGFkNTxdJ/images/ed4ec727393e853c6cfe043e6e9c6141297c2df1a835fb5de962353423329048.jpg b/parse/dev/EcGGFkNTxdJ/images/ed4ec727393e853c6cfe043e6e9c6141297c2df1a835fb5de962353423329048.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5a441f3787eec9499198e917b46ea830959bca71 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/ed4ec727393e853c6cfe043e6e9c6141297c2df1a835fb5de962353423329048.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:922a24f3c5e2eec8ec4822445faa4f7686d67408692c169493c78262ec0dc582 +size 32077 diff --git a/parse/dev/EcGGFkNTxdJ/images/ede0faa24a6f1b0eeb739ce8ad05979346019a3108810a1a1a7959757ddc2e84.jpg b/parse/dev/EcGGFkNTxdJ/images/ede0faa24a6f1b0eeb739ce8ad05979346019a3108810a1a1a7959757ddc2e84.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a6e25753c8f878d92873295ffcb53b681ae28a01 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/ede0faa24a6f1b0eeb739ce8ad05979346019a3108810a1a1a7959757ddc2e84.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77c7de0ef85c654bd4d7cf3c0b859b93d2d39bf571ac30dcc3cc03bf5cf5fee9 +size 9732 diff --git a/parse/dev/EcGGFkNTxdJ/images/f0f280e5c027898beceef5a365d22c27e86e2da4a68e2fbf8fc25c6c78d764f6.jpg b/parse/dev/EcGGFkNTxdJ/images/f0f280e5c027898beceef5a365d22c27e86e2da4a68e2fbf8fc25c6c78d764f6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..801fb716f97cc04a38f4b3657b2a7f46a38a87ef --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/f0f280e5c027898beceef5a365d22c27e86e2da4a68e2fbf8fc25c6c78d764f6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3130119fd97b9858fa451f4a255e6277219395658d907f728e2f109da33013da +size 5308 diff --git a/parse/dev/EcGGFkNTxdJ/images/f2fdae7d0a94ea030ecaae03de6645ab2f9279dc99e1f7a135988c2b8a4c5dfb.jpg b/parse/dev/EcGGFkNTxdJ/images/f2fdae7d0a94ea030ecaae03de6645ab2f9279dc99e1f7a135988c2b8a4c5dfb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..addb9403e6a1113c2b822749e22f777b4d966cd4 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/f2fdae7d0a94ea030ecaae03de6645ab2f9279dc99e1f7a135988c2b8a4c5dfb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:95966319683b276aee5e643d729c36f154ce7936acee33e85f7bac326d504b6f +size 10264 diff --git a/parse/dev/EcGGFkNTxdJ/images/f33444a0c07e5ac5e4b58ca2c09cd41f3bff4ea0ef2406d44ec69a4b7ccbe1f4.jpg b/parse/dev/EcGGFkNTxdJ/images/f33444a0c07e5ac5e4b58ca2c09cd41f3bff4ea0ef2406d44ec69a4b7ccbe1f4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3a1fab685ec8ec328518d8eb14faf5f40f93d69b --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/f33444a0c07e5ac5e4b58ca2c09cd41f3bff4ea0ef2406d44ec69a4b7ccbe1f4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e7a7988a6aefad676b50425aaaaf1c3235a51c32ce0459727026e2959cb1835 +size 7254 diff --git a/parse/dev/EcGGFkNTxdJ/images/f69f01f4b0af5e4ecf0ce2e296bedc597706b1fa2ad3d8e9f0be5a42bd168d04.jpg b/parse/dev/EcGGFkNTxdJ/images/f69f01f4b0af5e4ecf0ce2e296bedc597706b1fa2ad3d8e9f0be5a42bd168d04.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cea3687bd7d458ba7839bf17ed66c9d5da6ddc57 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/f69f01f4b0af5e4ecf0ce2e296bedc597706b1fa2ad3d8e9f0be5a42bd168d04.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:93f8baab8696d43bea59b792f7b93e00601a684fc28362d81db4426f58253a05 +size 7139 diff --git a/parse/dev/EcGGFkNTxdJ/images/f841c7d963a6c6f1591164bc0ac0a74529ac8bd9d3c5abffbdb350de92a12530.jpg b/parse/dev/EcGGFkNTxdJ/images/f841c7d963a6c6f1591164bc0ac0a74529ac8bd9d3c5abffbdb350de92a12530.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4f01af64c4aace95c2e881f5946b0d33b04c08bf --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/f841c7d963a6c6f1591164bc0ac0a74529ac8bd9d3c5abffbdb350de92a12530.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f3f1871277836759e72edb0090898f1edde574cab6d8612e7d4fc819d044eb9 +size 29682 diff --git a/parse/dev/EcGGFkNTxdJ/images/f8bd567247e7cab29cec0124a14a9b4c30ab12642abd4d02a5623dd259ca9953.jpg b/parse/dev/EcGGFkNTxdJ/images/f8bd567247e7cab29cec0124a14a9b4c30ab12642abd4d02a5623dd259ca9953.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fadc8757df354f1c9a5f68293742d12ede343591 --- /dev/null +++ b/parse/dev/EcGGFkNTxdJ/images/f8bd567247e7cab29cec0124a14a9b4c30ab12642abd4d02a5623dd259ca9953.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1179e479b6c2ad91f5457038e42b59019863afd63386764b585ff2abed82e6c +size 20894 diff --git a/parse/dev/FELWgMjxZJj/FELWgMjxZJj_content_list.json b/parse/dev/FELWgMjxZJj/FELWgMjxZJj_content_list.json new file mode 100644 index 0000000000000000000000000000000000000000..e6a17a5b06a21e5ef0810c8983af0721b9fd8c65 --- /dev/null +++ b/parse/dev/FELWgMjxZJj/FELWgMjxZJj_content_list.json @@ -0,0 +1,1526 @@ +[ + { + "type": "text", + "text": "OPEN-VOCABULARY SEMANTIC SEGMENTATION WITH MASK-ADAPTED CLIP ", + "text_level": 1, + "bbox": [ + 176, + 98, + 761, + 146 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Anonymous authors Paper under double-blind review ", + "bbox": [ + 183, + 171, + 398, + 198 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "ABSTRACT ", + "text_level": 1, + "bbox": [ + 454, + 234, + 544, + 251 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Open-vocabulary semantic segmentation aims to segment an image into semantic regions according to text descriptions, which may not have been seen during training. Recent two-stage methods first generate class-agnostic mask proposals and then leverage pre-trained vision-language models, e.g., CLIP, to classify masked regions. We identify the performance bottleneck of this paradigm to be the pre-trained CLIP model, since it does not perform well on masked images. To address this, we propose to finetune CLIP on a collection of masked image regions and their corresponding text descriptions. We collect training data by mining an existing image-caption dataset (e.g., COCO Captions), using CLIP to match masked image regions to nouns in the image captions. Compared with the more precise and manually annotated segmentation labels with fixed classes (e.g., COCO-Stuff), we find our noisy but diverse dataset can better retain CLIP’s generalization ability. Along with finetuning the entire model, we utilize the “blank” areas in masked images using a method we dub mask prompt tuning. Experiments demonstrate mask prompt tuning brings significant improvement without modifying any weights of CLIP, and it can further improve a fully finetuned model. In particular, when trained on COCO and evaluated on ADE20K-150, our best model achieves $2 9 . 6 \\%$ mIoU, which is $+ 8 . 5 \\%$ higher than the previous state-of-the-art. For the first time, open-vocabulary generalist models match the performance of supervised specialist models in 2017 without dataset specific adaptations. ", + "bbox": [ + 233, + 270, + 764, + 547 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "1 INTRODUCTION ", + "text_level": 1, + "bbox": [ + 176, + 583, + 336, + 598 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Semantic segmentation aims to group pixels into meaningful regions with corresponding semantic categories. Although remarkable progress has been made (Long et al., 2015; Chen et al., 2017; 2018; Zhao et al., 2017; Cheng et al., 2021), modern semantic segmentation models are mainly trained with pre-defined categories, failing to generalize to unseen classes. On the contrary, humans understand scenes in an open-vocabulary manner, typically with thousands of categories (Biederman, 1987). To approach human-level perception, this paper studies open-vocabulary semantic segmentation where the model segments an image by arbitrary categories described by texts. ", + "bbox": [ + 174, + 617, + 825, + 715 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Vision-language models, e.g., CLIP (Radford et al., 2021), learn rich multi-modal features from billion-scale image-text pairs. Witnessing its superior open-vocabulary classification ability, prior works propose to use pre-trained vision-language models for open-vocabulary segmentation (Li et al., 2022; Xu et al., 2021; Ding et al., 2022; Ghiasi et al., 2021). Among them, two-stage approaches have shown great potential: they first generate class-agnostic mask proposals and then leverage pre-trained CLIP to perform open-vocabulary classification (see Figure 1(b)). Their success relies on two assumptions: (1) the model can generate class-agnostic mask proposals (2) pre-trained CLIP can transfer its classification performance to masked image proposals. ", + "bbox": [ + 174, + 722, + 825, + 833 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "To examine these two assumptions, we conduct the following analysis. First, we assume an “oracle” mask generator and an ordinary CLIP classifier. We use ground-truth masks as region proposals, and feed masked images to a pre-trained CLIP for classification. This model only reaches an mIoU of $2 0 . 1 \\%$ on the ADE20K-150 dataset. Next, we assume an “oracle” classifier but an ordinary mask proposal generator – a MaskFormer (Cheng et al. (2021)) pre-trained on the COCO dataset. We first extract masked region proposals, then compare each region with ground-truth object masks, find the object with the highest overlap, and assign the object label to this extracted region. This model, despite imperfect region proposals, reaches a significantly higher mIoU of $6 6 . 5 \\%$ . ", + "bbox": [ + 174, + 840, + 823, + 924 + ], + "page_idx": 0 + }, + { + "type": "image", + "img_path": "images/4dbc502b48b3c12c7d6f43c0302de369b5be7ca5cde8c785213170dd1b35a746.jpg", + "image_caption": [ + "Figure 1: (a) CLIP is pre-trained with natural images with little data augmentation. (b) Two-stage open-vocabulary semantic segmentation approaches first generate class-agnostic mask proposals and then leverage pre-trained CLIP to do open-vocabulary classification. The input of the CLIP model is cropped masked images, which have huge domain gap from the natural images. (c) Our analysis reveals that pre-trained CLIP does not work well on masked images. " + ], + "image_footnote": [], + "bbox": [ + 179, + 99, + 818, + 248 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "", + "bbox": [ + 171, + 353, + 821, + 382 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "This analysis clearly shows that pre-trained CLIP can not perform satisfactory classification over masked images, and it is the performance bottleneck of two-stage open-vocabulary segmentation models. We hypothesize that this is caused by the significant domain gap between masked images and CLIP’s training images. CLIP is pre-trained on natural images with minimal data augmentation (Radford et al., 2021). On the other hand, mask proposals are cropped and re-sized from original images, and are further corrupted by noisy segmentation masks, see examples in Figure 1 (b). ", + "bbox": [ + 174, + 388, + 825, + 472 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "To address this, we propose to adapt CLIP by finetuning it on masked images and corresponding text labels. One direct solution is to use segmentation labels, e.g., from the COCO-stuff dataset. However, this leads to bad generalization to unseen classes (Section 4.3.1). Such manually annotated masks are accurate but classes are limited to a closed set (e.g., 171 classes for COCO-stuff). We hypothesize that the lack of text diversity causes the finetuned CLIP to lose the generalization ability to open vocabulary concepts. Instead, we collect training data by mining an existing image-caption dataset (e.g., COCO Captions). Given an image-caption pair, we first extract nouns in the caption, and generate class-agnostic masked region proposals using a pre-trained segmentation model. Then, with a pre-trained CLIP model, we assign the best-matching proposal to each extracted noun. By learning from this weakly-supervised alignments between masked images and novel categories, the adapted CLIP better retains its generalization ability for open vocabulary classification. ", + "bbox": [ + 174, + 479, + 825, + 632 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "The next question is how to effectively finetune CLIP? The most notable difference between a masked image and a natural image is that background pixels in a masked image are masked out, leading to many blank areas, which will be converted to “zero tokens” when feeding to CLIP transformers. Such tokens not only contain no useful information, but also bring domain distribution shift to the model (since such tokens don’t exist in natural images) and cause performance degradation. To mitigate this, we propose mask prompt tuning, a la visual prompt tuning (Jia et al., 2022). ´ When tokenizing a masked image, we replace the “zero tokens” with learnable prompt tokens. During finetuning, we either train prompts only and freeze CLIP’s weights, or train both of them. We find that mask prompt tuning alone significantly improves CLIP’s performance on masked images. This is a crucial property for multi-task scenarios where we cannot change CLIP’s weight since it is shared with other tasks. When combined with full model finetuning, mask prompt tuning can further improve the performance by a non-trivial margin (see Section 4.3.2). ", + "bbox": [ + 174, + 638, + 825, + 805 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "In our experiments, we measure the open-vocabulary segmentation performances on holdout segmentation datasets in a “zero-shot” manner – we do not adapt the model for each evaluation dataset. We train our model using COCO-stuff (Caesar et al., 2018) dataset with captions from Chen et al. (2015), and test on challenging ADE20K (A-150, A-847 for 150/846 categories) (Zhou et al., 2019), Pascal Context (PC-59, PC-459 for 59/459 categories) (Mottaghi et al., 2014) and PASCAL VOC (PAS-20) (Everingham et al., 2010). Our best model achieves $2 9 . 6 \\%$ mIoU on A-150, which is $+ 8 . 5 \\%$ than the state-of-the-art OpenSeg (Ghiasi et al., 2021) under the same setting. On more challenging A-847 and PC-459, our model sets up a new state-of-the-art of $9 . 0 \\%$ , $1 2 . 4 \\%$ mIoU, surpassing the previous best solution by $+ 2 . 7 \\%$ and $3 . 4 \\%$ . Moreover, for the first time, we show openvocabulary generalist models can match the performance of supervised specialist models (Long et al., 2015; Chen et al., 2017; Zoph et al., 2020) without dataset specific adaptations. ", + "bbox": [ + 174, + 813, + 825, + 924 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 103, + 821, + 146 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "In summary our contributions include: (1) Our analysis reveals the pre-trained CLIP does not perform well on mask proposals, making it the performance bottleneck of two-stage approaches. (2) We collect diverse mask-category pairs from captions to adapt CLIP for masked images and retain its generalization ability. (3) We propose mask prompt tuning specifically for masked image adaptation. This method does not change CLIP’s weight, enabling multi-task weight sharing. (4) For the first time, we show open-vocabulary generalist models can match the performance of supervised specialist models without dataset specific adaptations. ", + "bbox": [ + 174, + 152, + 823, + 251 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "2 RELATED WORK ", + "text_level": 1, + "bbox": [ + 176, + 273, + 344, + 290 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Pre-trained vision-language models (Radford et al., 2021; Jia et al., 2021; Li et al., 2021) connect the visual concepts with textual description. Pre-trained CLIP (Radford et al., 2021) has strong openvocabulary classification ability, i.e., classifying an image with arbitrary categories described by language. Pre-trained CLIP has empowered many computer vision tasks with the language ability, such as image manipulation (Patashnik et al., 2021), image generation (Crowson et al., 2022), object detection (Gu et al., 2021; Zhong et al., 2022) and image segmentation (Li et al., 2022; Xu et al., 2021; Ding et al., 2022; Ghiasi et al., 2021). Our work is similar with RegionCLIP (Zhong et al., 2022), which adapts CLIP for object detection by finetuning on region proposals. Our method differs from RegionCLIP since (1) we adapt CLIP to process masked images wile RegionCLIP process complete region crops; (2) We leverage blank areas in masked images and propose mask prompt tuning, which adapts CLIP without changing its weights. This enables sharing CLIP’s weight with other tasks in multi-task scenarios. This is not supported by RegionCLIP. ", + "bbox": [ + 174, + 309, + 825, + 476 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Open-vocabulary segmentation aims to understand an image with arbitrary categories described by texts. Pioneering work ZS3Net (Bucher et al., 2019) uses generative models to synthesize pixellevel features by word embeddings of unseen class. SPNet (Xian et al., 2019) utilizes the word embeddings, e.g., word2vec (Mikolov et al., 2013), to align the semantic meaning with visual features. More recently, researchers propose to leverage the pre-trained CLIP (Radford et al., 2021) for open-vocabulary semantic segmentation. LSeg (Li et al., 2022) aligns pixel embeddings to the text embedding of the corresponding semantic class, which is generated by CLIP’s text encoder. Unlike pixel-level LSeg, OpenSeg (Ghiasi et al., 2021) proposes to align the segment-level visual features with text embedding via region-word grounding. Our approach falls into the category of two-stage approaches, such as ZSSeg (Xu et al., 2021) and ZegFormer (Ding et al., 2022): they first generate class-agnostic mask proposals and then utilize pre-trained CLIP to perform open-vocabulary classification. Unlike ZSSeg and ZegFormer which directly use the original CLIP for masked image classification, we propose to adapt CLIP to achieve better results. ", + "bbox": [ + 174, + 483, + 825, + 662 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Prompt tuning is a strategy to adapt large-scale pre-trained models to new tasks. The idea originated from natural language processing (Liu et al., 2021a; Li & Liang, 2021; Lester et al., 2021), and recent work extends prompt tuning to computer vision. CoOp (Zhou et al., 2022) pre-appends the category words with learnable vectors to adapt CLIP for many recognition tasks. The textual prompt tuning is also widely used in open-vocabulary object detection (Du et al., 2022) and semantic segmentation $\\mathrm { X u }$ et al. (2021). Our mask prompt tuning is more relevant to prompt tuning in the visual domain (Bahng et al., 2022; Jia et al., 2022) where learnable vectors are applied to the image domain. Unlike visual prompt tuning (Jia et al., 2022) that inserts additional tokens before the actual image tokens, we replace masked tokens with learnable prompts. Furthermore, mask prompt tuning brings additional improvement over a fully finetuned model (Section 4.3.2). Such additional improvements have not been reported by prior work. ", + "bbox": [ + 173, + 670, + 825, + 823 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "3 METHOD ", + "text_level": 1, + "bbox": [ + 176, + 847, + 281, + 863 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "In this section, we first revisit the two-stage open-vocabulary segmentation methods (Xu et al., 2021; \nDing et al., 2022). Then we discuss how to obtain a dataset of mask-category pairs to finetune CLIP. \nLast, we discuss the proposed mask prompt tuning technique to adapt CLIP for masked images. ", + "bbox": [ + 176, + 882, + 823, + 924 + ], + "page_idx": 2 + }, + { + "type": "image", + "img_path": "images/902e1db0e01f8c692bcbdff40193c4d268f6bcfba7b6dc25c7dc80e9059a5f34.jpg", + "image_caption": [ + "Figure 2: Two-stage approaches consist of one segmentation model, e.g., MaskFormer, and one CLIP model. Firstly, the modified MaskFormer is trained with CLIP’s text embeddings so as to perform open-vocabulary segmentation. (Section 3.1). We then use the pre-trained segmentation model to generate class-agnostic proposals and align proposals with extracted nouns from corresponding captions (Section 3.2). After collecting diverse mask-category pairs, we finetune CLIP with the proposed mask prompt tuning (Section 3.3). " + ], + "image_footnote": [], + "bbox": [ + 186, + 74, + 799, + 270 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "3.1 TWO-STAGE MODELS FOR OPEN-VOCABULARY SEMANTIC SEGMENTATION ", + "text_level": 1, + "bbox": [ + 176, + 377, + 727, + 390 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "Our two-stage open-vocabulary semantic segmentation model is shown in Figure 2. It consists of a segmentation model that generates mask proposals, and an open vocabulary classification model. ", + "bbox": [ + 173, + 404, + 823, + 433 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "Following Xu et al. (2021); Ding et al. (2022), we choose MaskFormer (Cheng et al., 2021) as the segmentation model. Unlike per-pixel segmentation models (Long et al., 2015; Chen et al., 2017), MaskFormer predicts a set of $N$ mask proposals and corresponding class predictions. Each proposal is represented by an $H \\times W$ binary mask, indicating the location of the target object. The class prediction is a $C$ -dimensional distribution, where $C$ is the number of classes in the training set. Following (Xu et al., 2021), we modify MaskFormer such that for each mask, it generates a $C$ -dimensional proposal embedding, where $C$ is the embedding dimension of a CLIP model (512 for ViT-B/16 and 768 for ViT-L/14). This change allows MaskFormer to perform open-vocabulary segmentation. Specifically, suppose we would like to classify the mask to $K$ categories, we can first use a CLIP model’s text encoder to generate $\\mathrm { K }$ text embeddings for each class as $\\{ t _ { k } | t _ { k } \\in$ $ \\mathbf { R } ^ { C } \\} _ { k = 1 , \\cdots , K }$ . Next, we compare each mask embedding $v _ { i }$ with the text embedding, and predict the class- $k$ probability as $\\begin{array} { r } { p _ { i , k } = \\exp ( \\sigma ( v _ { i } , t _ { k } ) / \\tau ) / \\sum _ { k } ( \\exp ( \\sigma ( v _ { i } , t _ { k } ) / \\tau ) ) } \\end{array}$ . Here $\\sigma ( \\cdot , \\cdot )$ denotes the cosine similarity between two embedding vectors, and $\\tau$ is the temperature coefficient (Radford et al., 2021). We train the modified MaskFormer on the COCO-Stuff dataset (Caesar et al., 2018) with 171 classes. We use CLIP’s text encoder to process class names to generate text embeddings. We also append a learnable embedding $\\varnothing$ to represent the category of “no object”. For other training settings, we follow the original MaskFormer (Cheng et al., 2021). ", + "bbox": [ + 174, + 439, + 825, + 675 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "Note that the mask proposal generator trained this way is not strictly “class-agnostic”, as the definition of an object is determined by the class definitions in the training set. For example, if the training set only contains ”person” as a class, it is not likely the model will automatically segment a person into “face”, “hand”, “body”, or finer body parts. How to train a general and class agnostic model to generate mask proposals is an important topic but is beyond the scope of this paper. ", + "bbox": [ + 174, + 681, + 825, + 752 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "In addition to MaskFormer’s prediction, following (Ding et al., 2022; Xu et al., 2021), we add a parallel prediction branch using CLIP. MaskFormer generates mask proposals $\\{ M _ { i } | M _ { i } \\ \\in$ $\\{ 0 , 1 \\} ^ { \\dot { H } \\times W } \\} _ { i = 1 , \\cdots , N }$ where 1 and 0 denotes foreground and background. For each mask, we select a tight bounding box that includes all foreground pixels, crop the image, mask out backgrounds, and re-size to CLIP’s resolution. We feed mask proposal- $i$ to CLIP and compute class- $k$ probability as $\\hat { p } _ { i , k }$ . We ensemble both predictions to compute final prediction as $p _ { i , k } ^ { ( 1 - \\bar { \\lambda } ) } * \\hat { p } _ { i , k } ^ { \\lambda }$ where $\\lambda \\in [ 0 , 1 ]$ . We fuse mask-wise predictions to semantic segmentation using MaskFormer’s fusion module. ", + "bbox": [ + 174, + 758, + 825, + 861 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "As discussed in Section 1 and Figure 1 (c), our analysis show that CLIP does not work well on such masked images. Specifically, CLIP is trained on natural images with little data augmentation (Radford et al., 2021). However, masked images as shown in Figure 1 (b) contain a lot of “blank regions”. Such a significant domain gap makes it difficult for CLIP to transfer its classification performance. We also tried cropping the proposals without masking out background pixels. However, we observe worse performance (see Appendix A.1). We conjecture that keeping background pixels makes it more confusing for CLIP to correctly classify the foreground. ", + "bbox": [ + 174, + 867, + 823, + 924 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "", + "bbox": [ + 176, + 103, + 823, + 146 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "3.2 COLLECTING DIVERSE MASK-CATEGORY PAIRS FROM CAPTIONS ", + "text_level": 1, + "bbox": [ + 176, + 165, + 660, + 178 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "To adapt CLIP to better process masked images, we propose to finetune CLIP on a dataset consists of masked image and text pairs. One direct solution is to leverage manually annotated segmentation labels, e.g., from COCO-Stuff. Such labels are accurate, but have a closed set of categories. We try this solution and collect 965K mask-category pairs spanning 171 classes (e.g., banana, orange) from COCO-Stuff. Then we finetune the CLIP’s image encoder, while freezing the text encoder, following Zhong et al. (2022). However, we observe that this naive approach limits the generalization ability of CLIP, as the performance drops if there are more unseen classes (see Section 4.3.1). We hypothesize that due to the limited text vocabulary, the finetuned CLIP over-fits to the 171 classes, losing the ability to generalize to unseen categories. ", + "bbox": [ + 173, + 189, + 825, + 315 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Compared with segmentation labels, image captions contain much richer information about images and involve a much larger vocabulary. For example, in Figure 3, the image caption is \"There are apple and orange and teapot.\". Though \"apple\" and \"orange\" are valid classes in COCO-Stuff, other concepts are not valid classes and are ignored. ", + "bbox": [ + 174, + 321, + 496, + 433 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Based on this observation, we designed a selflabeling strategy (Ghiasi et al., 2021; Zhong et al., 2022) to extract mask-category pairs. As in Figure 3, given an image, we first use a pre-trained MaskFormer to extract masked proposals. Meanwhile, from the corresponding image caption, we extract all nouns using an off-the-shelf language parser (Bird et al., 2009), and treat them as potential classes. Then, we use CLIP to pair the most matching mask proposal to each class. From COCO-Captions Chen et al. (2015), we collect ", + "bbox": [ + 174, + 440, + 496, + 592 + ], + "page_idx": 4 + }, + { + "type": "image", + "img_path": "images/1d43a9fbee5e73fcda5c740587ceec91de8b3210c727ff52a7d84f9e5335e779.jpg", + "image_caption": [ + "Figure 3: For the given image-cation pair, only \"apple\" and \"orange\" are categories in COCO. By extracting nouns from captions, we can also get a novel \"teapot\" category. " + ], + "image_footnote": [], + "bbox": [ + 509, + 327, + 823, + 520 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "1.3M mask-category pairs with 27K unique nouns using 5 captions per image, or 440K pairs with 12K nouns using 1 caption per image. Experiments show this noisy but diverse mask-category dataset leads to significantly better performance than manual segmentation labels (see Section 4.3.1). ", + "bbox": [ + 176, + 593, + 825, + 635 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "3.3 MASK PROMPT TUNING ", + "text_level": 1, + "bbox": [ + 176, + 652, + 379, + 666 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "After collecting the dataset, a natural question is how to finetune CLIP effectively? The most notable difference between a masked image and a natural image is that background pixels in a masked images are set to zeros, leading to many “blank areas”. When feeding masked images to CLIP, images will be divided into non-overlapping patches and subsequently tokenized. Those blank areas will then become zero tokens. Such tokens not only contain no useful information, but also bring domain distribution shift to the model (since such tokens don’t exist in natural images) and cause performance degradation. To mitigate this, we propose a technique called mask prompt tuning, a\\` la visual prompt tuning (Jia et al., 2022). Specifically, when feeding into CLIP, a masked image will be tokenized to a tensor $T \\in \\mathbf { R } ^ { N _ { p } \\times E }$ , where $N _ { p }$ is the number of patches, and $E$ is the token dimension. The masked image also comes with a condensed binary mask $M _ { p } \\in \\{ 0 , 1 \\} ^ { N _ { p } }$ , where each element indicating whether a given patch is kept of masked out. We allocate a learnable tensor representing prompt tokens as $P \\in \\mathbf { R } ^ { \\dot { N } _ { p } \\times E }$ . Finally, the final input to the transformer is computed as $T \\otimes M _ { p } + P \\otimes ( 1 - M _ { p } )$ , where $\\otimes$ denotes element-wise multiplication. Following the “deep prompts” setting in Jia et al. (2022), we can add such prompt tokens to deeper layers of the transformer. This is also illustrated in Figure 4. ", + "bbox": [ + 174, + 679, + 825, + 888 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Compared with fully finetuning the entire model (Zhong et al., 2022), mask prompt tuning has several advantages. First, it is specifically designed for segmentation tasks, where parts of input images are masked. Next, compared with full model finetuning, the amount of trainable parameters in mask prompt tuning is orders of magnitude smaller, leading to much better training efficiency. Moreover, as a foundational model, CLIP may be simultaneously used for many tasks, and we may not be allowed to tune CLIP’s weights. Mask prompt tuning does not require changing weights of CLIP, thus is suitable for such multi-task scenarios. Lastly, our experiments show that mask prompt tuning alone leads to significant improvement. And if applied together with full model finetuning, it can further improve the open-vocabulary segmentation performance (Section 4.3.2). ", + "bbox": [ + 174, + 895, + 821, + 924 + ], + "page_idx": 4 + }, + { + "type": "image", + "img_path": "images/718708aaa18f8fa2ca1f4b308b7854826f33e07fcbef12d53dacd8e3e3aa8c98.jpg", + "image_caption": [ + "Figure 4: The proposed mask prompt tuning can adapt CLIP to masked images without changing its weights. We replace the zero tokens from masked patches to learnable mask prompt. " + ], + "image_footnote": [], + "bbox": [ + 173, + 99, + 790, + 253 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 301, + 825, + 400 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "4 EXPERIMENTS ", + "text_level": 1, + "bbox": [ + 176, + 414, + 326, + 429 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "4.1 EXPERIMENTAL SETUP ", + "text_level": 1, + "bbox": [ + 176, + 445, + 374, + 459 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Training Dataset We train our model on the COCO dataset (Lin et al., 2014). We first train the modified MaskFormer using the segmentation labels from COCO-Stuff (Caesar et al., 2018). Next, we finetune CLIP on the mask-category dataset that we obtained from COCO Captions (Chen et al., 2015). There are 118k training images labeled with 171 valid categories in the dataset, ranging from things (e.g., orange, car) to stuffs (e.g., sky, road). If not specified otherwise, we use all the 171 categories data during training. ", + "bbox": [ + 174, + 470, + 825, + 555 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Evaluation Dataset Our open-vocabulary model is able to perform zero-shot segmentation on arbitrary datasets without dataset-specific adaption. Thus, we test our model on challenging ADE20K (Zhou et al., 2019), Pascal VOC (Everingham et al., 2010) and Pascal Context (Mottaghi et al., 2014) datasets. ADE20K is a densely pixel-wise annotated dataset for scene understanding, which spans diverse annotations of indoor and outdoor scenes. There are 2K images in its validation set. We choose two versions of categories, one with 150 frequently used categories (A-150) and one with more diverse 847 categories (A-847). Pascal VOC is a classical dataset for segmentation. We evaluate on the 1.5K validation images with 20 categories (PAS-20). Pascal Context is a set of additional annotations for PASCAL VOC 2010. It goes beyond the original PASCAL semantic segmentation task by providing annotations for the whole scene. There are 5K images in its validation set. We also choose two versions of categories, one with 59 frequently used categories (PC-59) and one with the whole 459 categories (PC-459). ", + "bbox": [ + 174, + 563, + 825, + 728 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Implementation Details As indicated before, our model consists of two part: one segmentation model based on MaskFormer (Cheng et al., 2021) and one mask-adapted CLIP model (Radford et al., 2021). For the segmentation model, we have two backbone choices, ResNet-101c (Chen et al., 2017) and Swin-Base (Liu et al., 2021b). For the CLIP model, we have two choices: ViTB/16 and ViT-L/14 (Dosovitskiy et al., 2020). We detail our largest model setting here, while the training recipe of the R101c model can be found in Appendix A.2. For Swin-Base segmentation model, the backbone weights are initialized from an ImageNet-21K pre-trained model. We use AdamW (Loshchilov & Hutter, 2017) optimizer with the poly learning rate schedule (Chen et al., 2017). The initial learning rate and weight decay are set to ${ \\bar { 6 } } \\cdot 1 0 ^ { - 5 }$ and $1 0 ^ { - 2 }$ , respectively. We use a crop size of $6 4 0 \\times 6 4 0$ , a batch size of 32 and train the model for 120K iterations. For data augmentations and other hyper-parameters, we mainly following the setting of Cheng et al. (2021). ", + "bbox": [ + 173, + 736, + 825, + 888 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "For adapting CLIP ViT-L/14 model, we utilize the OpenCLIP (Ilharco et al., 2021) implementation. After collecting 440K mask-category pairs from captions (see Section 3.2), we propose three ways to adapt CLIP: mask prompt tuning (MPT) only, full model fine-tuning (FT) only and joint MPT $+ \\ \\mathrm { F T } .$ . For MPT only, we initialize the CLIP model with official OpenAI weights (Radford et al., 2021) and the learnable tokens are randomly initialized. We also use the deep prompts as proposed in Jia et al. (2022). The prompt depth is set to 3 if not specified otherwise. The training optimizer is AdamW with initial learning rate $2 \\cdot 1 0 ^ { - 2 } $ and weight decay 0. The cosine annealing scheduler is adopted to adjust the learning rate. The model is trained with input size of $2 2 4 \\times 2 2 4$ , a batch size of 256 for 5 epochs. For FT only, we keep similar training procedure but with a much lower learning rate $5 \\cdot 1 0 ^ { - 6 }$ and larger weight decay 0.2. For $\\mathbf { M P T } + \\mathbf { F T }$ , we first initialize the CLIP with fully finetuned model and then apply the mask prompt tuning over it, which we fined more stable and effective (see Appendix A.3). All other hyper-parameters are the same with MPT only. The text encoder of CLIP is frozen in all our experiments. ", + "bbox": [ + 174, + 895, + 823, + 924 + ], + "page_idx": 5 + }, + { + "type": "table", + "img_path": "images/af4b497aa3db86c868f4353a17d55d305b1766a341b921326b56bffd337e97a6.jpg", + "table_caption": [ + "Table 1: The mIoU results of our model and previous open-vocabulary generalist models and supervised specialist models. Results for SPNet and ZS3Net on PAS-20 are reported from Li et al. (2022). Results for ZegFormer on PAS-20 are recalculated by us. Under the R101c model scale, our model significantly outperforms other open-vocabulary models. Our large Swin-Base model can match the performance of some supervised specialist models. " + ], + "table_footnote": [], + "table_body": "
methodbackbonetraining datasetA-847PC-459A-150PC-59PAS-20
Open-vocabulary generalist models.
SPNet (Xian et al., 2019)R-101PASCAL-15=24.318.3
ZS3Net (Bucher et al., 2019)R-101PASCAL-1519.438.3
LSeg (Li et al., 2022)R-101PASCAL-15=-47.4
LSeg+ (Ghiasi et al., 2021)R-101COCO Panoptic2.55.213.036.059.0
SimBaseline (Xu et al.,2021)R-101cCOCO-Stuff-156--15.3-74.5
ZegFormer (Ding et al.,2022)R-50COCO-Stuff-156==16.4-80.7
OpenSeg (Ghiasi et al., 2021)R-101COCO Panoptic4.06.515.336.960.0
OVSeg (Ours)R-101cCOCO-Stuff-1567.010.424.051.789.2
OVSeg (Ours)R-101cCOCO-Stuff-1717.111.024.853.392.6
LSeg+ (Ghiasi et al., 2021)Eff-B7COCO Panoptic3.87.818.046.5
OpenSeg (Ghiasi et al., 2021)Eff-B7COCO Panoptic6.39.021.142.1
OVSeg (Ours)Swin-BCOCO-Stuff-1719.012.429.655.794.5
Supervised specialist models.
FCN (Long et al., 2015)FCN-8sSame as test29.437.81
Deeplab (Chen et al., 2017)R-101Same as test=45.777.7
SelfTrain (Zoph et al., 2020)Eff-L2Same as test=--90.0
MaskFormer (Cheng et al., 2021)R-101cSame as test17.446.0-=
", + "bbox": [ + 171, + 162, + 838, + 434 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "", + "bbox": [ + 173, + 448, + 825, + 602 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "4.2 MAIN RESULTS ON OPEN VOCABULARY SEMANTIC SEGMENTATION ", + "text_level": 1, + "bbox": [ + 176, + 612, + 678, + 626 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "OVSeg achieves best performance among open-vocabulary models. We conduct the comparison with other open-vocabulary generalist models using the common ResNet-101 (R-101) model scale in Table 1. If not specified otherwise, our best performance is achieved using joint mask prompt tuning and fine-tuning (see Section 4.3.2). First of all, compared with per-pixel approaches (SPNet (Xian et al., 2019), ZS3Net (Bucher et al., 2019), LSeg (Li et al., 2022) and ${ \\mathrm { L S e g } } +$ (Ghiasi et al., 2021)), proposal-based approaches (OpenSeg (Ghiasi et al., 2021), SimBaseline (Xu et al., 2021) and ZegFormer (Ding et al., 2022)) show better performance. Our OVSeg also falls into the proposal-based category. Compared with other proposal-based approaches, our model shows significant improvements across all five benchmarks. In particular, our R101c model achieves $7 . 1 \\%$ and $1 1 . 0 \\%$ mIoU on challenging A-847 and PC-459, which even performs better than the EfficientNetB7 based OpenSeg model. All proposal-based approaches are using the same COCO images, i.e., the 2017 splits with 118K images, but with different annotations. Our experiments show different annotations result in relatively small performance differences: we only observe a ${ \\sim } 1 \\%$ performance drop on A-150 when changing COCO-Stuff-171 to COCO-Stuff-156. ", + "bbox": [ + 173, + 638, + 825, + 833 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "Largest OVSeg model sets up new SOTA results on zero-shot benchmarks. When we scale up the model, our method can further achieve better results. With Swin-Base (Swin-B) backbone and CLIP ViT-L/14, our model can achieve $2 9 . 6 \\%$ and $5 5 . 5 \\%$ mIoU on A-150 and Pascal PC-59, which is $+ 8 . 5 \\%$ and $+ 1 3 . 6 \\%$ higher than the SOTA zero-shot results. On the challenging A-847 and PC459, our model sets up a new zero-shot state-of-the-art $9 . 0 \\%$ and $12 . 4 \\%$ mIoU. We further detail the class-wise IoU of A-150 categories in Appendix A.4. ", + "bbox": [ + 173, + 840, + 823, + 924 + ], + "page_idx": 6 + }, + { + "type": "table", + "img_path": "images/dc5e4bb7c5f0e26427ff1c8feb2013a0179d46ac03d1c03f02c6d4db1ccd691c.jpg", + "table_caption": [ + "Table 2: Ablation on mask-category pairs. The baseline is MaskFormer Swin-Base with original CLIP ViT-L/14. The masks come from ground-truth (GT) or generated proposals. The category nouns come from ground-truth (GT) classes or captions. We also calculate the statistics (number of pairs and unique nouns) of collected pairs. " + ], + "table_footnote": [], + "table_body": "
CaseSourceStatisticsA-847A-150PC-59
MaskCategoryPairsUnique nouns
Baseline11-17.321.851.4
(1)GTGT965K1715.3 (-2.0)23.0 (+1.2)57.3 (+5.9)
(2)GT1 caption440K12K7.9 (+0.6)24.2 (+2.4)53.2 (+1.8)
(3)proposals1 caption440K12K8.8 (+1.5)28.8 (+7.0)55.7 (+4.3)
(4)proposals5 captions1.3M27K8.8 (+1.5)28.6 (+6.8)55.5 (+4.1)
", + "bbox": [ + 179, + 148, + 815, + 263 + ], + "page_idx": 7 + }, + { + "type": "table", + "img_path": "images/09bf78328ac989401de7e896ac571717e065febf116b8a454b2a975f2e7c02a8.jpg", + "table_caption": [ + "Table 3: Ablation on mask prompt tuning. The baseline is MaskFormer Swin-Base with CLIP ViTL/14. We report the zero-shot mIoU on representative ADE-847, ADE-150 and PC-59 datasets. All the improvements are measured upon the baseline model. " + ], + "table_footnote": [], + "table_body": "
caseFinetuning methodA-847A-150PC-59
mask promptfull model
Baseline7.321.851.4
(a)√8.4 (+1.1)26.5 (+4.7)55.4 (+4.0)
(b)√8.8 (+1.5)28.8 (+7.0)55.7 (+4.3)
(c)√√9.0 (+1.7)29.6 (+7.8)55.7 (+4.3)
", + "bbox": [ + 246, + 325, + 746, + 426 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "Open-vocabulary generalist models can match supervised specialist models in 2017. We show our generalist model can achieve competitive performance without the need of any dataset specific training. On the challenging A-150, our model achieves similar performance with fully supervised FCN-8s (Long et al., 2015). On the PAS-20, our model achieves $9 4 . 5 \\%$ mIoU, which is even $+ 4 . 5 \\%$ than the SOTA specialist model (Zoph et al., 2020). We note OVSeg is not directly comparable with supervised models because OVSeg is not trained on evaluation datasets. OVSeg also has different backbones and segmentation model architectures. Thus, comparison with supervised models is for reference purposes only. Our generalist model still underperforms the advanced specialist models, such as supervised MaskFormer (Cheng et al., 2021). ", + "bbox": [ + 173, + 440, + 825, + 565 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "4.3 ABLATION STUDY ", + "text_level": 1, + "bbox": [ + 176, + 583, + 339, + 597 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "4.3.1 COLLECTING MASK-CATEGORY PAIRS ", + "text_level": 1, + "bbox": [ + 176, + 609, + 490, + 623 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "We discuss the impact of finetuning data in Table 2. The baseline model is MaskFormer Swin-Base with the original CLIP ViT-L/14. Our initial trial (case (1)) is collecting ground-truth (GT) masks with supervised GT classes. We can collecting 965K mask-category pairs with 171 unique nouns (the number of classes defined in COCO-stuff). Then we finetune the CLIP model with the collected pairs. We observe a $- 2 . 0 \\%$ performance drop on the A-847 dataset. This is because the adapted CLIP is over-fitting to the 171 GT classes. Although the model achieves good results on PC-59 (whose categories are highly overlapped with COCO-Stuff), it perform badly for more diverse concepts in A-847. As detailed in Section 3.2, we propose to utilize captions (Chen et al., 2015) to collect diverse mask-category pairs. After parsing the nouns in the caption, we match the nouns with GT masks (case (2)) or proposals (case (3)) generated by the baseline model. By replacing the GT masks with proposals, the A-150 mIoU is significantly improved (from $2 4 . 2 \\%$ to $2 8 . 8 \\%$ ) We conjecture that many regions are not labeled as GT masks (see examples in Figure 3), and are therefore ignored. In contrast, the generated proposals (usually 100) can cover most of regions-of-interest in the image, leading to better performance. If all the 5 captions per image are used (case (4)), we observe a mild $. 0 . 2 \\%$ degradation on A-150 and PC-59 We hypothesis that 12K nouns are adequate for the CLIP to retain its open-vocabulary ability. Thus, we use 1 caption per image as our default setting. ", + "bbox": [ + 174, + 633, + 825, + 854 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "4.3.2 MASK PROMPT TUNING ", + "text_level": 1, + "bbox": [ + 176, + 871, + 392, + 885 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "We ablate the effect of mask prompt tuning in Table 3. The baseline model is MaskFormer SwinBase with CLIP ViT-L/14. If we only use mask prompt tuning (case (a)), our model outperforms the baseline by a large $+ 4 . 7 \\%$ and $+ 4 . 0 \\%$ mIoU improvement on ADE-150 and PC-59, respectively. Case (b) shows the result of full model fine-tuning. Although it achieves the best accuracy, the trainable parameters are orders of magnitude higher. In contrast, the proposed mask prompt tuning only modifies the input without changing CLIP’s weight. Furthermore, mask prompt tuning can further improve over a fully finetuned model, as shown in case (c). Case (c) achieves $2 9 . 6 \\%$ mIoU ADE-150, which further improves the fully finetuned model by a considerable margin of $+ 0 . 8 \\%$ . ", + "bbox": [ + 176, + 895, + 821, + 924 + ], + "page_idx": 7 + }, + { + "type": "image", + "img_path": "images/eab20d161b109f263f4a76fdc4fc604d480770e4b5eb0df443268e1d2470466f.jpg", + "image_caption": [ + "Figure 5: Open-vocabulary segmentation with user-defined queries. Our model accurately segments unseen categories, such as the Saturn V rocket, Oculus headset, and Golden gate bridge. " + ], + "image_footnote": [], + "bbox": [ + 174, + 90, + 818, + 222 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 276, + 825, + 361 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "4.4 DISCUSSIONS ", + "text_level": 1, + "bbox": [ + 174, + 377, + 310, + 392 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "4.4.1 SEGMENTATION WITH USER-DEFINED QUERIES. ", + "text_level": 1, + "bbox": [ + 176, + 405, + 552, + 417 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "Our method allows users to define arbitrary queries and search the query in the image, see Figure 5. Without training our models to learn specific concepts, our model can locate and segment Saturn V as the lego rocket, Oculus as the VR headset, and golden gate as the bridge in corresponding images. This demonstrates the strong potentials of open vocabulary semantic segmentation. ", + "bbox": [ + 174, + 428, + 825, + 484 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "4.4.2 AMBIGUITY OF OPEN VOCABULARY SEGMENTATION EVALUATION ", + "text_level": 1, + "bbox": [ + 174, + 501, + 684, + 513 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "We show some “failure” predictions from the A-150 dataset in Figure 6. For the left figure, the ground-truth category is “building” while our model predicts “skyscrapers”. The “skyscrapers” is a reasonable description, but the standard A-150 evaluation protocol will treat it as a wrong prediction. A similar case happens in the right figure, the ground-truth “rail” is recognized as “road”. This is caused by the fact that language defined categories are ambiguous and can overlap with each other. Designing a better evaluation metric for openvocabulary segmentation models is an important topic for our future research. Note that due to IP constraints, we use our own images, in", + "bbox": [ + 174, + 523, + 483, + 732 + ], + "page_idx": 8 + }, + { + "type": "image", + "img_path": "images/2d8f5fc671546954aa5ef4320c106d9b6ae07356d1bb906c1d52f0a1f7d1d051.jpg", + "image_caption": [ + "Figure 6: Ambiguity of the class definition. " + ], + "image_footnote": [], + "bbox": [ + 516, + 522, + 821, + 689 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "stead of ADE20K images in Figure 6. But this phenomenon widely exists on ADE20K images. ", + "bbox": [ + 171, + 732, + 799, + 746 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "5 CONCLUSION ", + "text_level": 1, + "bbox": [ + 176, + 766, + 318, + 781 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "This paper studies open-vocabulary semantic segmentation where the model segments an image by arbitrary categories described by texts. We identify the performance bottleneck of current two-stage methods to be the pre-trained CLIP, since it doesn’t perform well on masked images. We propose to adapt CLIP for masked images. To retain CLIP’s open-vocabulary classification ability, we adapt CLIP with diverse mask-category pairs mined from image-caption dataset. We further propose mask prompt tuning, a method can adapt CLIP without changing its original weights. The proposed model is general and can do zero-shot segmentation on arbitrary datasets without dataset-specific adaption. For the first time, we showopen-vocabulary generalist models can match the performance of supervised specialist models. ", + "bbox": [ + 174, + 797, + 825, + 922 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "ETHICS STATEMENT ", + "text_level": 1, + "bbox": [ + 176, + 102, + 343, + 118 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "We only use the public computer vision datasets (COCO, ADE20K, Pascal) and leverage the opensourced vision-language models (CLIP) for our experiments. To the best of our knowledge, we do not foresee our approach as being inherently subject to concerns of discrimination / bias / fairness, inappropriate potential applications, impact, privacy and security issues, research integrity or research practice issues. However, the public datasets and pre-trained models may be subject to bias that may be inherited by models trained with our approach. ", + "bbox": [ + 174, + 133, + 825, + 217 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "REPRODUCIBILITY STATEMENT ", + "text_level": 1, + "bbox": [ + 176, + 238, + 436, + 253 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Our code is reproducible and can be implemented based on the method description in Section 3 as well as training details in Section 4.1. ", + "bbox": [ + 174, + 268, + 823, + 297 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "REFERENCES ", + "text_level": 1, + "bbox": [ + 174, + 319, + 285, + 333 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "H Bahng, A Jahanian, S Sankaranarayanan, and P Isola. Exploring visual prompts for adapting large-scale models. arXiv preprint arXiv:2203.17274, pp. 2022, 2022. ", + "bbox": [ + 173, + 342, + 823, + 371 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Irving Biederman. Recognition-by-components: a theory of human image understanding. Psychological review, 94(2):115, 1987. ", + "bbox": [ + 174, + 378, + 821, + 407 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Steven Bird, Ewan Klein, and Edward Loper. Natural language processing with Python: analyzing text with the natural language toolkit. ” O’Reilly Media, Inc.”, 2009. ", + "bbox": [ + 173, + 416, + 823, + 445 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Maxime Bucher, Tuan-Hung Vu, Matthieu Cord, and Patrick Perez. Zero-shot semantic segmenta- ´ tion. Advances in Neural Information Processing Systems, 32, 2019. ", + "bbox": [ + 171, + 454, + 823, + 483 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Holger Caesar, Jasper Uijlings, and Vittorio Ferrari. Coco-stuff: Thing and stuff classes in context. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1209– 1218, 2018. ", + "bbox": [ + 174, + 492, + 823, + 535 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos, Kevin Murphy, and Alan L Yuille. Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs. IEEE transactions on pattern analysis and machine intelligence, 40(4): 834–848, 2017. ", + "bbox": [ + 173, + 544, + 825, + 601 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam. Encoderdecoder with atrous separable convolution for semantic image segmentation. In Proceedings of the European conference on computer vision (ECCV), pp. 801–818, 2018. ", + "bbox": [ + 174, + 609, + 823, + 652 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Xinlei Chen, Hao Fang, Tsung-Yi Lin, Ramakrishna Vedantam, Saurabh Gupta, Piotr Dollar, and ´ C Lawrence Zitnick. Microsoft coco captions: Data collection and evaluation server. arXiv preprint arXiv:1504.00325, 2015. ", + "bbox": [ + 176, + 661, + 821, + 704 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Bowen Cheng, Alex Schwing, and Alexander Kirillov. Per-pixel classification is not all you need for semantic segmentation. Advances in Neural Information Processing Systems, 34:17864–17875, 2021. ", + "bbox": [ + 173, + 712, + 823, + 755 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Katherine Crowson, Stella Biderman, Daniel Kornis, Dashiell Stander, Eric Hallahan, Louis Castricato, and Edward Raff. Vqgan-clip: Open domain image generation and editing with natural language guidance. arXiv preprint arXiv:2204.08583, 2022. ", + "bbox": [ + 174, + 763, + 823, + 808 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Jian Ding, Nan Xue, Gui-Song Xia, and Dengxin Dai. Decoupling zero-shot semantic segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11583–11592, 2022. ", + "bbox": [ + 173, + 815, + 821, + 858 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. ", + "bbox": [ + 176, + 867, + 825, + 922 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Yu Du, Fangyun Wei, Zihe Zhang, Miaojing Shi, Yue Gao, and Guoqi Li. Learning to prompt for open-vocabulary object detection with vision-language model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 14084–14093, 2022. ", + "bbox": [ + 178, + 103, + 823, + 146 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Mark Everingham, Luc Van Gool, Christopher KI Williams, John Winn, and Andrew Zisserman. The pascal visual object classes (voc) challenge. International journal of computer vision, 88(2): 303–338, 2010. ", + "bbox": [ + 173, + 156, + 825, + 199 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Golnaz Ghiasi, Xiuye Gu, Yin Cui, and Tsung-Yi Lin. Open-vocabulary image segmentation. arXiv preprint arXiv:2112.12143, 2021. ", + "bbox": [ + 174, + 210, + 823, + 239 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Xiuye Gu, Tsung-Yi Lin, Weicheng Kuo, and Yin Cui. Open-vocabulary object detection via vision and language knowledge distillation. arXiv preprint arXiv:2104.13921, 2021. ", + "bbox": [ + 173, + 251, + 823, + 280 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Gabriel Ilharco, Mitchell Wortsman, Ross Wightman, Cade Gordon, Nicholas Carlini, Rohan Taori, Achal Dave, Vaishaal Shankar, Hongseok Namkoong, John Miller, Hannaneh Hajishirzi, Ali Farhadi, and Ludwig Schmidt. Openclip, July 2021. URL https://doi.org/10.5281/ zenodo.5143773. If you use this software, please cite it as below. ", + "bbox": [ + 173, + 290, + 826, + 347 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc Le, Yun-Hsuan Sung, Zhen Li, and Tom Duerig. Scaling up visual and vision-language representation learning with noisy text supervision. In International Conference on Machine Learning, pp. 4904–4916. PMLR, 2021. ", + "bbox": [ + 173, + 358, + 825, + 415 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, and Ser-Nam Lim. Visual prompt tuning. arXiv preprint arXiv:2203.12119, 2022. ", + "bbox": [ + 173, + 425, + 823, + 455 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Brian Lester, Rami Al-Rfou, and Noah Constant. The power of scale for parameter-efficient prompt tuning. arXiv preprint arXiv:2104.08691, 2021. ", + "bbox": [ + 171, + 465, + 823, + 496 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Boyi Li, Kilian Q Weinberger, Serge Belongie, Vladlen Koltun, and Rene Ranftl. Language-driven ´ semantic segmentation. arXiv preprint arXiv:2201.03546, 2022. ", + "bbox": [ + 169, + 506, + 823, + 535 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Xiang Lisa Li and Percy Liang. Prefix-tuning: Optimizing continuous prompts for generation. arXiv preprint arXiv:2101.00190, 2021. ", + "bbox": [ + 173, + 545, + 823, + 575 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Yangguang Li, Feng Liang, Lichen Zhao, Yufeng Cui, Wanli Ouyang, Jing Shao, Fengwei Yu, and Junjie Yan. Supervision exists everywhere: A data efficient contrastive language-image pretraining paradigm. arXiv preprint arXiv:2110.05208, 2021. ", + "bbox": [ + 174, + 585, + 825, + 630 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In ´ European conference on computer vision, pp. 740–755. Springer, 2014. ", + "bbox": [ + 176, + 640, + 825, + 683 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Pengfei Liu, Weizhe Yuan, Jinlan Fu, Zhengbao Jiang, Hiroaki Hayashi, and Graham Neubig. Pretrain, prompt, and predict: A systematic survey of prompting methods in natural language processing. arXiv preprint arXiv:2107.13586, 2021a. ", + "bbox": [ + 173, + 693, + 823, + 737 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 10012–10022, 2021b. ", + "bbox": [ + 173, + 747, + 825, + 790 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Jonathan Long, Evan Shelhamer, and Trevor Darrell. Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 3431–3440, 2015. ", + "bbox": [ + 171, + 801, + 823, + 844 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017. ", + "bbox": [ + 169, + 854, + 825, + 883 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781, 2013. ", + "bbox": [ + 173, + 895, + 821, + 924 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Roozbeh Mottaghi, Xianjie Chen, Xiaobai Liu, Nam-Gyu Cho, Seong-Whan Lee, Sanja Fidler, Raquel Urtasun, and Alan Yuille. The role of context for object detection and semantic segmentation in the wild. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 891–898, 2014. ", + "bbox": [ + 174, + 103, + 825, + 160 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Or Patashnik, Zongze Wu, Eli Shechtman, Daniel Cohen-Or, and Dani Lischinski. Styleclip: Textdriven manipulation of stylegan imagery. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 2085–2094, 2021. ", + "bbox": [ + 176, + 167, + 820, + 212 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International Conference on Machine Learning, pp. 8748–8763. PMLR, 2021. ", + "bbox": [ + 173, + 219, + 825, + 276 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Yongqin Xian, Subhabrata Choudhury, Yang He, Bernt Schiele, and Zeynep Akata. Semantic projection network for zero-and few-label semantic segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 8256–8265, 2019. ", + "bbox": [ + 173, + 285, + 823, + 329 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Mengde Xu, Zheng Zhang, Fangyun Wei, Yutong Lin, Yue Cao, Han Hu, and Xiang Bai. A simple baseline for zero-shot semantic segmentation with pre-trained vision-language model. arXiv preprint arXiv:2112.14757, 2021. ", + "bbox": [ + 173, + 337, + 823, + 380 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Hengshuang Zhao, Jianping Shi, Xiaojuan Qi, Xiaogang Wang, and Jiaya Jia. Pyramid scene parsing network. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2881–2890, 2017. ", + "bbox": [ + 173, + 388, + 823, + 431 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Yiwu Zhong, Jianwei Yang, Pengchuan Zhang, Chunyuan Li, Noel Codella, Liunian Harold Li, Luowei Zhou, Xiyang Dai, Lu Yuan, Yin Li, et al. Regionclip: Region-based language-image pretraining. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 16793–16803, 2022. ", + "bbox": [ + 174, + 440, + 825, + 496 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Bolei Zhou, Hang Zhao, Xavier Puig, Tete Xiao, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Semantic understanding of scenes through the ade20k dataset. International Journal of Computer Vision, 127(3):302–321, 2019. ", + "bbox": [ + 174, + 506, + 823, + 549 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Kaiyang Zhou, Jingkang Yang, Chen Change Loy, and Ziwei Liu. Learning to prompt for visionlanguage models. International Journal of Computer Vision, 130(9):2337–2348, 2022. ", + "bbox": [ + 173, + 556, + 821, + 587 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Barret Zoph, Golnaz Ghiasi, Tsung-Yi Lin, Yin Cui, Hanxiao Liu, Ekin Dogus Cubuk, and Quoc Le. Rethinking pre-training and self-training. Advances in neural information processing systems, 33: 3833–3845, 2020. ", + "bbox": [ + 174, + 594, + 825, + 637 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "A APPENDIX ", + "text_level": 1, + "bbox": [ + 176, + 102, + 297, + 117 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "A.1 CROP WITH OR WITHOUT MASK ", + "text_level": 1, + "bbox": [ + 176, + 135, + 437, + 148 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "In the paper, we use the default crop with mask (see the left of Figure 7). We also try the direct crop without mask (see the right of Figure 7). Following the bottleneck analysis in Figure 1, we feed the unmasked crops a pre-trained CLIP for classification. This experiment gives a $1 3 . 8 \\%$ mIoU, which is - $- 6 . 3 \\%$ worse than using the masked crops. We hypothesize that the crop with mask introduces many background pixels, making the prediction more difficult. For the example in the right of Figure 7, the “orange” will also be an appropriate category for the unmasked crop. ", + "bbox": [ + 174, + 161, + 516, + 327 + ], + "page_idx": 12 + }, + { + "type": "image", + "img_path": "images/3777c50a6d4d2c5e9ed3c9bd2ff3eefe0094ec2326b1ed0cb8d9cc5013293687.jpg", + "image_caption": [ + "Figure 7: Crop without mask will introduce background pixels, making the prediction more difficult. " + ], + "image_footnote": [], + "bbox": [ + 531, + 160, + 820, + 286 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "done a ablation study about different strategies to obtain final crop. We have the similar conclusions. ", + "bbox": [ + 169, + 349, + 821, + 363 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "A.2 TRAINING HYPERPARAMS OF R101C MODEL ", + "text_level": 1, + "bbox": [ + 174, + 382, + 526, + 395 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "Our small model is MaskFormer R101c with CLIP ViT-B/16. For MaskFormer training, the backbone weights are initialized from an ImageNet-1K pre-trained model. We use AdamW optimizer with the poly learning rate schedule. The initial learning rate and weight decay are set to $\\bar { 2 } \\cdot 1 0 ^ { - 4 }$ and $1 0 ^ { - 4 }$ , respectively. We also use a learning rate multiplier 0.1 on the backbone. We use a crop size of $5 1 2 \\times 5 1 2$ , a batch size of 32 and train the model for 120K iterations. For data augmentations and other hyper-parameters, we majorly following the setting of Cheng et al. (2021). For adapting CLIP ViT-B/16 model, we basically follow the hyperparameters of finetuning ViT-L/16 except we use a larger batch size 1024. ", + "bbox": [ + 174, + 407, + 825, + 518 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "A.3 MORE ABLATION STUDIES ON MASK PROMPT TUNING ", + "text_level": 1, + "bbox": [ + 174, + 539, + 589, + 553 + ], + "page_idx": 12 + }, + { + "type": "table", + "img_path": "images/c0d6e417e14d841b3eeb0757d8cbb06906b3eea3ee3ae914e515e024aadeb29d.jpg", + "table_caption": [ + "Table 4: Ablation on combining mask prompt tuning (MPT) and fine-tuning (FT). FT - ${ \\bf \\cdot } { \\bf > } { \\bf M } { \\bf P } { \\bf T }$ indicates first FT and then MPT, and vice versa. FT $+ { \\mathbf { M P T } }$ sim. means optimizing prompts and CLIP simultaneously. " + ], + "table_footnote": [], + "table_body": "
combinationA-847A-150
FT ->MPT (default)9.029.6
MPT ->FT8.5 (-0.5)28.1 (-1.5)
FT + MPT sim.8.8 (-0.2)29.0 (-0.6)
", + "bbox": [ + 173, + 656, + 500, + 731 + ], + "page_idx": 12 + }, + { + "type": "table", + "img_path": "images/2fcac108374fa59b823f747b0df33caf7b992ec94c55f024cb7902f031619e6e.jpg", + "table_caption": [ + "Table 5: Ablation on prompt depth. We test with and without fully fine-tuned (FT) model. " + ], + "table_footnote": [], + "table_body": "
prompt depthA-150
1w/o FT 25.7w/FT 29.3
3 (default)26.5
26.829.6 29.4
6 1226.829.3
", + "bbox": [ + 550, + 619, + 789, + 727 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "We explore two other ways to combine mask prompt tuning (MPT) and fine-tuning (FT) as in Table 4. Our default setting (F $\\mathrm { \\mathbf { \\hat { T } } \\mathbf { \\Phi } } \\mathrm { \\mathbf { - } } \\mathrm { \\mathbf { \\mathrm { > } } } \\mathbf { M } \\mathbf { P } \\mathrm { T } \\mathrm { \\mathbf { \\Phi } }$ ) is first doing FT and then applying MPT to the already fine-tuned model. We don’t change the weights of fine-tuned CLIP. The other option is first doing MPT and then doing FT with fixed mask prompts (MPT - ${ \\mathrm { - } } { \\mathrm { P T } }$ ). This combination produces poor results (- $. 1 . 5 \\%$ drop on A-150). We conjecture the mask prompts learned with original CLIP provide a bad prior when we fune-tune the entire CLIP model. We also explore learning mask prompts and fine-tune CLIP weight simultaneously $\\mathrm { ( F T + M P T }$ sim.). This doesn’t bring favorable results. ", + "bbox": [ + 173, + 750, + 825, + 848 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "We further ablate the effects of prompt depth in Table 5. The depth can be selected from $\\{ 1 , 3 , 6 , 1 2 \\}$ . We use two different scenarios: without fine-tuning (w/o FT) for mask prompt tuning only, with finetuning (w/ FT) for applying mask prompt tuning over a already fine-tuned model. For w/o FT case, one layer prompt can bring significant improvement, e.g., from baseline’s $2 1 . 8 \\%$ to $2 5 . 7 \\%$ . Deeper prompts result in better performance, because more parameters are introduced with more prompts. ", + "bbox": [ + 174, + 853, + 823, + 924 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "Interestingly, deeper prompts (going from 3 to 12) don’t bring further improvement for w/ FT case. \nWe choose prompt depth as 3 for default setting. ", + "bbox": [ + 174, + 103, + 823, + 132 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "A.4 CLASS-WISE IOU OVER SEEN AND UNSEEN CATEGORIES. ", + "text_level": 1, + "bbox": [ + 176, + 148, + 612, + 162 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "We detail the class IoU on all 150 categories in ADE20K-150 (model trained on COCO) in Figure 8, and we annotated seen vs. unseen classes and their IoUs. Seen categories mean there are similar categories in COCO-stuff training set. Unseen categories denotes the novel categories in ADE20K. The average IoU of seen and unseen categories are $3 7 . 6 \\%$ and $2 1 . 9 \\%$ , respectively, showing that our model performs better on seen categories. This is also observed in other open vocabulary segmentation work, such as (Ding et al., 2022). ", + "bbox": [ + 174, + 174, + 825, + 258 + ], + "page_idx": 13 + }, + { + "type": "image", + "img_path": "images/036880ad502f2f39d96cfa158711ec8183689044e31f4d8c132f21e1ebe3db00.jpg", + "image_caption": [ + "Figure 8: Class IoU on all 150 categories in ADE20K (model trained on COCO). It is expected the model performs better on seen categories in training set. " + ], + "image_footnote": [], + "bbox": [ + 333, + 20, + 656, + 885 + ], + "page_idx": 14 + } +] \ No newline at end of file diff --git a/parse/dev/FELWgMjxZJj/FELWgMjxZJj_model.json b/parse/dev/FELWgMjxZJj/FELWgMjxZJj_model.json new file mode 100644 index 0000000000000000000000000000000000000000..e8b42cc044b05eabe0c974026ef9a0bdb5a1c484 --- /dev/null +++ b/parse/dev/FELWgMjxZJj/FELWgMjxZJj_model.json @@ -0,0 +1,19862 @@ +[ + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 299, + 1589, + 1403, + 1589, + 1403, + 1835, + 299, + 1835 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 1360, + 1403, + 1360, + 1403, + 1575, + 298, + 1575 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 299, + 1850, + 1402, + 1850, + 1402, + 2034, + 299, + 2034 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 398, + 595, + 1302, + 595, + 1302, + 1207, + 398, + 1207 + ], + "score": 0.973 + }, + { + "category_id": 0, + "poly": [ + 301, + 219, + 1295, + 219, + 1295, + 323, + 301, + 323 + ], + "score": 0.964 + }, + { + "category_id": 1, + "poly": [ + 313, + 378, + 680, + 378, + 680, + 438, + 313, + 438 + ], + "score": 0.931 + }, + { + "category_id": 0, + "poly": [ + 302, + 1284, + 573, + 1284, + 573, + 1319, + 302, + 1319 + ], + "score": 0.898 + }, + { + "category_id": 0, + "poly": [ + 773, + 519, + 926, + 519, + 926, + 553, + 773, + 553 + ], + "score": 0.875 + }, + { + "category_id": 2, + "poly": [ + 298, + 75, + 856, + 75, + 856, + 104, + 298, + 104 + ], + "score": 0.858 + }, + { + "category_id": 2, + "poly": [ + 842, + 2089, + 857, + 2089, + 857, + 2112, + 842, + 2112 + ], + "score": 0.611 + }, + { + "category_id": 13, + "poly": [ + 762, + 1115, + 840, + 1115, + 840, + 1144, + 762, + 1144 + ], + "score": 0.9, + "latex": "+ 8 . 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 326, + 1942, + 402, + 1942, + 402, + 1972, + 326, + 1972 + ], + "score": 0.87, + "latex": "2 0 . 1 \\%" + }, + { + "category_id": 13, + "poly": [ + 499, + 1115, + 575, + 1115, + 575, + 1144, + 499, + 1144 + ], + "score": 0.85, + "latex": "2 9 . 6 \\%" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 216.0, + 1304.0, + 216.0, + 1304.0, + 272.0, + 296.0, + 272.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 272.0, + 890.0, + 272.0, + 890.0, + 327.0, + 296.0, + 327.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1281.0, + 579.0, + 1281.0, + 579.0, + 1328.0, + 294.0, + 1328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 769.0, + 518.0, + 932.0, + 518.0, + 932.0, + 556.0, + 769.0, + 556.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 72.0, + 858.0, + 72.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 841.0, + 2088.0, + 860.0, + 2088.0, + 860.0, + 2118.0, + 841.0, + 2118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1589.0, + 1407.0, + 1589.0, + 1407.0, + 1627.0, + 295.0, + 1627.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1619.0, + 1405.0, + 1619.0, + 1405.0, + 1656.0, + 294.0, + 1656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1652.0, + 1404.0, + 1652.0, + 1404.0, + 1686.0, + 294.0, + 1686.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1676.0, + 1405.0, + 1676.0, + 1405.0, + 1722.0, + 290.0, + 1722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1712.0, + 1407.0, + 1712.0, + 1407.0, + 1748.0, + 293.0, + 1748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1742.0, + 1407.0, + 1742.0, + 1407.0, + 1779.0, + 293.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1774.0, + 1405.0, + 1774.0, + 1405.0, + 1808.0, + 294.0, + 1808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1802.0, + 1146.0, + 1802.0, + 1146.0, + 1838.0, + 294.0, + 1838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1359.0, + 1405.0, + 1359.0, + 1405.0, + 1397.0, + 294.0, + 1397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1391.0, + 1405.0, + 1391.0, + 1405.0, + 1426.0, + 293.0, + 1426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1420.0, + 1407.0, + 1420.0, + 1407.0, + 1458.0, + 293.0, + 1458.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1453.0, + 1406.0, + 1453.0, + 1406.0, + 1487.0, + 292.0, + 1487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1484.0, + 1405.0, + 1484.0, + 1405.0, + 1515.0, + 293.0, + 1515.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1514.0, + 1407.0, + 1514.0, + 1407.0, + 1550.0, + 293.0, + 1550.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1545.0, + 1099.0, + 1545.0, + 1099.0, + 1579.0, + 296.0, + 1579.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1848.0, + 1404.0, + 1848.0, + 1404.0, + 1884.0, + 294.0, + 1884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1881.0, + 1406.0, + 1881.0, + 1406.0, + 1917.0, + 295.0, + 1917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1912.0, + 1403.0, + 1912.0, + 1403.0, + 1943.0, + 294.0, + 1943.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1942.0, + 325.0, + 1942.0, + 325.0, + 1978.0, + 294.0, + 1978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 403.0, + 1942.0, + 1407.0, + 1942.0, + 1407.0, + 1978.0, + 403.0, + 1978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1972.0, + 1406.0, + 1972.0, + 1406.0, + 2007.0, + 293.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 2003.0, + 1406.0, + 2003.0, + 1406.0, + 2037.0, + 293.0, + 2037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 397.0, + 597.0, + 1306.0, + 597.0, + 1306.0, + 633.0, + 397.0, + 633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 627.0, + 1306.0, + 627.0, + 1306.0, + 665.0, + 393.0, + 665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 659.0, + 1303.0, + 659.0, + 1303.0, + 692.0, + 395.0, + 692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 687.0, + 1305.0, + 687.0, + 1305.0, + 726.0, + 393.0, + 726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 720.0, + 1305.0, + 720.0, + 1305.0, + 753.0, + 394.0, + 753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 751.0, + 1305.0, + 751.0, + 1305.0, + 783.0, + 394.0, + 783.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 778.0, + 1306.0, + 778.0, + 1306.0, + 816.0, + 393.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 810.0, + 1305.0, + 810.0, + 1305.0, + 846.0, + 394.0, + 846.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 841.0, + 1306.0, + 841.0, + 1306.0, + 877.0, + 394.0, + 877.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 870.0, + 1306.0, + 870.0, + 1306.0, + 906.0, + 393.0, + 906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 391.0, + 898.0, + 1307.0, + 898.0, + 1307.0, + 940.0, + 391.0, + 940.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 931.0, + 1305.0, + 931.0, + 1305.0, + 966.0, + 394.0, + 966.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 961.0, + 1306.0, + 961.0, + 1306.0, + 996.0, + 393.0, + 996.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 993.0, + 1306.0, + 993.0, + 1306.0, + 1029.0, + 393.0, + 1029.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1023.0, + 1306.0, + 1023.0, + 1306.0, + 1059.0, + 394.0, + 1059.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1055.0, + 1306.0, + 1055.0, + 1306.0, + 1086.0, + 393.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1084.0, + 1305.0, + 1084.0, + 1305.0, + 1116.0, + 393.0, + 1116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1113.0, + 498.0, + 1113.0, + 498.0, + 1149.0, + 394.0, + 1149.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 576.0, + 1113.0, + 761.0, + 1113.0, + 761.0, + 1149.0, + 576.0, + 1149.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 841.0, + 1113.0, + 1304.0, + 1113.0, + 1304.0, + 1149.0, + 841.0, + 1149.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1144.0, + 1305.0, + 1144.0, + 1305.0, + 1180.0, + 394.0, + 1180.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1173.0, + 1219.0, + 1173.0, + 1219.0, + 1212.0, + 392.0, + 1212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 315.0, + 379.0, + 560.0, + 379.0, + 560.0, + 411.0, + 315.0, + 411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 312.0, + 407.0, + 681.0, + 407.0, + 681.0, + 442.0, + 312.0, + 442.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 0, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1406, + 1404, + 1406, + 1404, + 1774, + 298, + 1774 + ], + "score": 0.985 + }, + { + "category_id": 1, + "poly": [ + 298, + 1056, + 1404, + 1056, + 1404, + 1392, + 298, + 1392 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 298, + 856, + 1403, + 856, + 1403, + 1041, + 298, + 1041 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 1789, + 1403, + 1789, + 1403, + 2034, + 298, + 2034 + ], + "score": 0.978 + }, + { + "category_id": 3, + "poly": [ + 307, + 220, + 1392, + 220, + 1392, + 549, + 307, + 549 + ], + "score": 0.971 + }, + { + "category_id": 4, + "poly": [ + 297, + 580, + 1404, + 580, + 1404, + 735, + 297, + 735 + ], + "score": 0.965 + }, + { + "category_id": 1, + "poly": [ + 294, + 779, + 1399, + 779, + 1399, + 842, + 294, + 842 + ], + "score": 0.952 + }, + { + "category_id": 2, + "poly": [ + 298, + 76, + 857, + 76, + 857, + 104, + 298, + 104 + ], + "score": 0.877 + }, + { + "category_id": 2, + "poly": [ + 841, + 2088, + 858, + 2088, + 858, + 2112, + 841, + 2112 + ], + "score": 0.721 + }, + { + "category_id": 2, + "poly": [ + 841, + 2088, + 859, + 2088, + 859, + 2112, + 841, + 2112 + ], + "score": 0.096 + }, + { + "category_id": 13, + "poly": [ + 297, + 1973, + 374, + 1973, + 374, + 2002, + 297, + 2002 + ], + "score": 0.88, + "latex": "+ 8 . 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 1125, + 811, + 1199, + 811, + 1199, + 840, + 1125, + 840 + ], + "score": 0.87, + "latex": "6 6 . 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 1018, + 1943, + 1095, + 1943, + 1095, + 1972, + 1018, + 1972 + ], + "score": 0.86, + "latex": "2 9 . 6 \\%" + }, + { + "category_id": 13, + "poly": [ + 1200, + 2003, + 1274, + 2003, + 1274, + 2032, + 1200, + 2032 + ], + "score": 0.86, + "latex": "1 2 . 4 \\%" + }, + { + "category_id": 13, + "poly": [ + 1128, + 2003, + 1188, + 2003, + 1188, + 2032, + 1128, + 2032 + ], + "score": 0.83, + "latex": "9 . 0 \\%" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 230.0, + 457.0, + 230.0, + 457.0, + 258.0, + 320.0, + 258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 577.0, + 232.0, + 692.0, + 232.0, + 692.0, + 260.0, + 577.0, + 260.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1182.0, + 221.0, + 1395.0, + 221.0, + 1395.0, + 253.0, + 1182.0, + 253.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 249.0, + 457.0, + 249.0, + 457.0, + 277.0, + 321.0, + 277.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 575.0, + 252.0, + 690.0, + 252.0, + 690.0, + 280.0, + 575.0, + 280.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1183.0, + 249.0, + 1377.0, + 249.0, + 1377.0, + 283.0, + 1183.0, + 283.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 267.0, + 444.0, + 267.0, + 444.0, + 302.0, + 320.0, + 302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 780.0, + 269.0, + 929.0, + 269.0, + 929.0, + 303.0, + 780.0, + 303.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1145.0, + 288.0, + 1177.0, + 288.0, + 1177.0, + 506.0, + 1145.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1298.0, + 299.0, + 1353.0, + 299.0, + 1353.0, + 333.0, + 1298.0, + 333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 358.0, + 337.0, + 422.0, + 337.0, + 422.0, + 371.0, + 358.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 595.0, + 356.0, + 614.0, + 356.0, + 614.0, + 371.0, + 595.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1064.0, + 331.0, + 1141.0, + 331.0, + 1141.0, + 376.0, + 1064.0, + 376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1180.0, + 335.0, + 1219.0, + 335.0, + 1219.0, + 367.0, + 1180.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 682.0, + 372.0, + 776.0, + 372.0, + 776.0, + 433.0, + 682.0, + 433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 927.0, + 372.0, + 1058.0, + 372.0, + 1058.0, + 436.0, + 927.0, + 436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1084.0, + 388.0, + 1101.0, + 388.0, + 1101.0, + 422.0, + 1084.0, + 422.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1179.0, + 378.0, + 1219.0, + 378.0, + 1219.0, + 414.0, + 1179.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1230.0, + 411.0, + 1284.0, + 411.0, + 1284.0, + 445.0, + 1230.0, + 445.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 681.0, + 431.0, + 779.0, + 431.0, + 779.0, + 458.0, + 681.0, + 458.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1180.0, + 432.0, + 1217.0, + 432.0, + 1217.0, + 461.0, + 1180.0, + 461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 301.0, + 493.0, + 506.0, + 493.0, + 506.0, + 525.0, + 301.0, + 525.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1228.0, + 486.0, + 1352.0, + 486.0, + 1352.0, + 514.0, + 1228.0, + 514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 516.0, + 482.0, + 516.0, + 482.0, + 554.0, + 297.0, + 554.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 695.0, + 514.0, + 1033.0, + 514.0, + 1033.0, + 548.0, + 695.0, + 548.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1170.0, + 515.0, + 1384.0, + 515.0, + 1384.0, + 550.0, + 1170.0, + 550.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1065.0, + 444.0, + 1102.0, + 444.0, + 1102.0, + 476.0, + 1065.0, + 476.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 579.0, + 1404.0, + 579.0, + 1404.0, + 619.0, + 294.0, + 619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 612.0, + 1404.0, + 612.0, + 1404.0, + 647.0, + 293.0, + 647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 642.0, + 1405.0, + 642.0, + 1405.0, + 676.0, + 293.0, + 676.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 673.0, + 1405.0, + 673.0, + 1405.0, + 709.0, + 292.0, + 709.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 700.0, + 1061.0, + 700.0, + 1061.0, + 741.0, + 292.0, + 741.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 857.0, + 73.0, + 857.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2085.0, + 862.0, + 2085.0, + 862.0, + 2120.0, + 839.0, + 2120.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2085.0, + 862.0, + 2085.0, + 862.0, + 2120.0, + 838.0, + 2120.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1405.0, + 1409.0, + 1405.0, + 1409.0, + 1444.0, + 293.0, + 1444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1439.0, + 1405.0, + 1439.0, + 1405.0, + 1473.0, + 294.0, + 1473.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1466.0, + 1405.0, + 1466.0, + 1405.0, + 1506.0, + 292.0, + 1506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1500.0, + 1404.0, + 1500.0, + 1404.0, + 1534.0, + 294.0, + 1534.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1528.0, + 1404.0, + 1528.0, + 1404.0, + 1564.0, + 293.0, + 1564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1560.0, + 1402.0, + 1560.0, + 1402.0, + 1594.0, + 294.0, + 1594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1591.0, + 1402.0, + 1591.0, + 1402.0, + 1625.0, + 294.0, + 1625.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1620.0, + 1406.0, + 1620.0, + 1406.0, + 1653.0, + 293.0, + 1653.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1650.0, + 1404.0, + 1650.0, + 1404.0, + 1688.0, + 293.0, + 1688.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1678.0, + 1406.0, + 1678.0, + 1406.0, + 1718.0, + 292.0, + 1718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1711.0, + 1405.0, + 1711.0, + 1405.0, + 1748.0, + 293.0, + 1748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1742.0, + 1063.0, + 1742.0, + 1063.0, + 1776.0, + 294.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1053.0, + 1405.0, + 1053.0, + 1405.0, + 1094.0, + 293.0, + 1094.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1085.0, + 1405.0, + 1085.0, + 1405.0, + 1124.0, + 292.0, + 1124.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1119.0, + 1402.0, + 1119.0, + 1402.0, + 1150.0, + 296.0, + 1150.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1146.0, + 1406.0, + 1146.0, + 1406.0, + 1184.0, + 293.0, + 1184.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1176.0, + 1404.0, + 1176.0, + 1404.0, + 1215.0, + 292.0, + 1215.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1208.0, + 1404.0, + 1208.0, + 1404.0, + 1243.0, + 294.0, + 1243.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1240.0, + 1404.0, + 1240.0, + 1404.0, + 1275.0, + 294.0, + 1275.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1270.0, + 1405.0, + 1270.0, + 1405.0, + 1305.0, + 293.0, + 1305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1298.0, + 1405.0, + 1298.0, + 1405.0, + 1337.0, + 294.0, + 1337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1332.0, + 1405.0, + 1332.0, + 1405.0, + 1366.0, + 294.0, + 1366.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1360.0, + 1269.0, + 1360.0, + 1269.0, + 1399.0, + 294.0, + 1399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 857.0, + 1407.0, + 857.0, + 1407.0, + 893.0, + 296.0, + 893.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 886.0, + 1404.0, + 886.0, + 1404.0, + 922.0, + 294.0, + 922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 917.0, + 1405.0, + 917.0, + 1405.0, + 955.0, + 293.0, + 955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 950.0, + 1404.0, + 950.0, + 1404.0, + 982.0, + 296.0, + 982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 976.0, + 1404.0, + 976.0, + 1404.0, + 1012.0, + 294.0, + 1012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1010.0, + 1338.0, + 1010.0, + 1338.0, + 1045.0, + 294.0, + 1045.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1787.0, + 1404.0, + 1787.0, + 1404.0, + 1828.0, + 292.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1820.0, + 1405.0, + 1820.0, + 1405.0, + 1854.0, + 294.0, + 1854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1849.0, + 1407.0, + 1849.0, + 1407.0, + 1886.0, + 294.0, + 1886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1882.0, + 1405.0, + 1882.0, + 1405.0, + 1915.0, + 296.0, + 1915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1911.0, + 1405.0, + 1911.0, + 1405.0, + 1945.0, + 293.0, + 1945.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1943.0, + 1017.0, + 1943.0, + 1017.0, + 1976.0, + 294.0, + 1976.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1096.0, + 1943.0, + 1407.0, + 1943.0, + 1407.0, + 1976.0, + 1096.0, + 1976.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1970.0, + 296.0, + 1970.0, + 296.0, + 2009.0, + 293.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 375.0, + 1970.0, + 1407.0, + 1970.0, + 1407.0, + 2009.0, + 375.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2003.0, + 1127.0, + 2003.0, + 1127.0, + 2036.0, + 294.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1189.0, + 2003.0, + 1199.0, + 2003.0, + 1199.0, + 2036.0, + 1189.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1275.0, + 2003.0, + 1405.0, + 2003.0, + 1405.0, + 2036.0, + 1275.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 776.0, + 1404.0, + 776.0, + 1404.0, + 816.0, + 294.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 809.0, + 1124.0, + 809.0, + 1124.0, + 845.0, + 294.0, + 845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1200.0, + 809.0, + 1210.0, + 809.0, + 1210.0, + 845.0, + 1200.0, + 845.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 1, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 681, + 1403, + 681, + 1403, + 1049, + 298, + 1049 + ], + "score": 0.984 + }, + { + "category_id": 1, + "poly": [ + 298, + 1064, + 1404, + 1064, + 1404, + 1461, + 298, + 1461 + ], + "score": 0.984 + }, + { + "category_id": 1, + "poly": [ + 297, + 1476, + 1404, + 1476, + 1404, + 1812, + 297, + 1812 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 298, + 337, + 1402, + 337, + 1402, + 553, + 298, + 553 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 300, + 1942, + 1402, + 1942, + 1402, + 2034, + 300, + 2034 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 299, + 229, + 1399, + 229, + 1399, + 323, + 299, + 323 + ], + "score": 0.955 + }, + { + "category_id": 0, + "poly": [ + 301, + 605, + 587, + 605, + 587, + 641, + 301, + 641 + ], + "score": 0.903 + }, + { + "category_id": 0, + "poly": [ + 300, + 1866, + 480, + 1866, + 480, + 1902, + 300, + 1902 + ], + "score": 0.889 + }, + { + "category_id": 2, + "poly": [ + 298, + 76, + 856, + 76, + 856, + 104, + 298, + 104 + ], + "score": 0.855 + }, + { + "category_id": 2, + "poly": [ + 842, + 2088, + 858, + 2088, + 858, + 2111, + 842, + 2111 + ], + "score": 0.57 + }, + { + "category_id": 2, + "poly": [ + 842, + 2088, + 859, + 2088, + 859, + 2111, + 842, + 2111 + ], + "score": 0.409 + }, + { + "category_id": 2, + "poly": [ + 300, + 76, + 855, + 76, + 855, + 104, + 300, + 104 + ], + "score": 0.153 + }, + { + "category_id": 13, + "poly": [ + 712, + 231, + 790, + 231, + 790, + 260, + 712, + 260 + ], + "score": 0.89, + "latex": "+ 2 . 7 \\%" + }, + { + "category_id": 13, + "poly": [ + 840, + 231, + 901, + 231, + 901, + 260, + 840, + 260 + ], + "score": 0.86, + "latex": "3 . 4 \\%" + }, + { + "category_id": 13, + "poly": [ + 453, + 1630, + 492, + 1630, + 492, + 1658, + 453, + 1658 + ], + "score": 0.34, + "latex": "\\mathrm { X u }" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 602.0, + 594.0, + 602.0, + 594.0, + 649.0, + 291.0, + 649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1860.0, + 486.0, + 1860.0, + 486.0, + 1910.0, + 291.0, + 1910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2086.0, + 860.0, + 2086.0, + 860.0, + 2117.0, + 839.0, + 2117.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2086.0, + 860.0, + 2086.0, + 860.0, + 2117.0, + 839.0, + 2117.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 297.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 682.0, + 1405.0, + 682.0, + 1405.0, + 716.0, + 294.0, + 716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 709.0, + 1405.0, + 709.0, + 1405.0, + 749.0, + 292.0, + 749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 743.0, + 1404.0, + 743.0, + 1404.0, + 781.0, + 294.0, + 781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 771.0, + 1407.0, + 771.0, + 1407.0, + 811.0, + 292.0, + 811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 805.0, + 1405.0, + 805.0, + 1405.0, + 839.0, + 294.0, + 839.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 832.0, + 1405.0, + 832.0, + 1405.0, + 870.0, + 293.0, + 870.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 861.0, + 1405.0, + 861.0, + 1405.0, + 902.0, + 292.0, + 902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 894.0, + 1407.0, + 894.0, + 1407.0, + 932.0, + 293.0, + 932.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 925.0, + 1405.0, + 925.0, + 1405.0, + 962.0, + 293.0, + 962.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 955.0, + 1404.0, + 955.0, + 1404.0, + 993.0, + 293.0, + 993.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 986.0, + 1405.0, + 986.0, + 1405.0, + 1023.0, + 293.0, + 1023.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1014.0, + 1113.0, + 1014.0, + 1113.0, + 1054.0, + 292.0, + 1054.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1063.0, + 1405.0, + 1063.0, + 1405.0, + 1099.0, + 293.0, + 1099.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1094.0, + 1404.0, + 1094.0, + 1404.0, + 1129.0, + 293.0, + 1129.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1127.0, + 1404.0, + 1127.0, + 1404.0, + 1159.0, + 294.0, + 1159.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1155.0, + 1404.0, + 1155.0, + 1404.0, + 1191.0, + 294.0, + 1191.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1185.0, + 1406.0, + 1185.0, + 1406.0, + 1221.0, + 292.0, + 1221.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1216.0, + 1406.0, + 1216.0, + 1406.0, + 1252.0, + 293.0, + 1252.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1247.0, + 1406.0, + 1247.0, + 1406.0, + 1283.0, + 293.0, + 1283.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1275.0, + 1408.0, + 1275.0, + 1408.0, + 1316.0, + 293.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1307.0, + 1405.0, + 1307.0, + 1405.0, + 1347.0, + 293.0, + 1347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1338.0, + 1405.0, + 1338.0, + 1405.0, + 1376.0, + 292.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1365.0, + 1405.0, + 1365.0, + 1405.0, + 1407.0, + 292.0, + 1407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1394.0, + 1405.0, + 1394.0, + 1405.0, + 1439.0, + 292.0, + 1439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1429.0, + 1026.0, + 1429.0, + 1026.0, + 1466.0, + 294.0, + 1466.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1477.0, + 1405.0, + 1477.0, + 1405.0, + 1511.0, + 295.0, + 1511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1505.0, + 1406.0, + 1505.0, + 1406.0, + 1542.0, + 292.0, + 1542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1537.0, + 1404.0, + 1537.0, + 1404.0, + 1572.0, + 292.0, + 1572.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1567.0, + 1405.0, + 1567.0, + 1405.0, + 1602.0, + 294.0, + 1602.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1598.0, + 1405.0, + 1598.0, + 1405.0, + 1634.0, + 291.0, + 1634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1628.0, + 452.0, + 1628.0, + 452.0, + 1665.0, + 291.0, + 1665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 493.0, + 1628.0, + 1406.0, + 1628.0, + 1406.0, + 1665.0, + 493.0, + 1665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1656.0, + 1406.0, + 1656.0, + 1406.0, + 1697.0, + 292.0, + 1697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1690.0, + 1406.0, + 1690.0, + 1406.0, + 1725.0, + 295.0, + 1725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1721.0, + 1404.0, + 1721.0, + 1404.0, + 1755.0, + 294.0, + 1755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1751.0, + 1406.0, + 1751.0, + 1406.0, + 1785.0, + 294.0, + 1785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1782.0, + 886.0, + 1782.0, + 886.0, + 1816.0, + 293.0, + 1816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 339.0, + 1404.0, + 339.0, + 1404.0, + 373.0, + 294.0, + 373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 367.0, + 1404.0, + 367.0, + 1404.0, + 404.0, + 292.0, + 404.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 396.0, + 1407.0, + 396.0, + 1407.0, + 436.0, + 293.0, + 436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 429.0, + 1406.0, + 429.0, + 1406.0, + 467.0, + 293.0, + 467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 460.0, + 1404.0, + 460.0, + 1404.0, + 494.0, + 293.0, + 494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 489.0, + 1407.0, + 489.0, + 1407.0, + 528.0, + 293.0, + 528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 523.0, + 903.0, + 523.0, + 903.0, + 557.0, + 294.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1941.0, + 1406.0, + 1941.0, + 1406.0, + 1978.0, + 293.0, + 1978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1971.0, + 1404.0, + 1971.0, + 1404.0, + 2008.0, + 293.0, + 2008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 2000.0, + 1362.0, + 2000.0, + 1362.0, + 2040.0, + 291.0, + 2040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 228.0, + 711.0, + 228.0, + 711.0, + 267.0, + 291.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 791.0, + 228.0, + 839.0, + 228.0, + 839.0, + 267.0, + 791.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 902.0, + 228.0, + 1404.0, + 228.0, + 1404.0, + 267.0, + 902.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 259.0, + 1405.0, + 259.0, + 1405.0, + 299.0, + 294.0, + 299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 292.0, + 1248.0, + 292.0, + 1248.0, + 326.0, + 294.0, + 326.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 2, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 967, + 1404, + 967, + 1404, + 1488, + 298, + 1488 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 298, + 1671, + 1403, + 1671, + 1403, + 1896, + 298, + 1896 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 1502, + 1403, + 1502, + 1403, + 1656, + 298, + 1656 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 299, + 1911, + 1401, + 1911, + 1401, + 2034, + 299, + 2034 + ], + "score": 0.971 + }, + { + "category_id": 3, + "poly": [ + 319, + 165, + 1359, + 165, + 1359, + 595, + 319, + 595 + ], + "score": 0.971 + }, + { + "category_id": 4, + "poly": [ + 296, + 605, + 1405, + 605, + 1405, + 790, + 296, + 790 + ], + "score": 0.964 + }, + { + "category_id": 1, + "poly": [ + 296, + 889, + 1402, + 889, + 1402, + 953, + 296, + 953 + ], + "score": 0.947 + }, + { + "category_id": 2, + "poly": [ + 297, + 76, + 857, + 76, + 857, + 104, + 297, + 104 + ], + "score": 0.889 + }, + { + "category_id": 0, + "poly": [ + 301, + 831, + 1238, + 831, + 1238, + 860, + 301, + 860 + ], + "score": 0.867 + }, + { + "category_id": 2, + "poly": [ + 841, + 2089, + 858, + 2089, + 858, + 2111, + 841, + 2111 + ], + "score": 0.757 + }, + { + "category_id": 13, + "poly": [ + 1050, + 1823, + 1198, + 1823, + 1198, + 1869, + 1050, + 1869 + ], + "score": 0.93, + "latex": "p _ { i , k } ^ { ( 1 - \\bar { \\lambda } ) } * \\hat { p } _ { i , k } ^ { \\lambda }" + }, + { + "category_id": 13, + "poly": [ + 298, + 1729, + 536, + 1729, + 536, + 1766, + 298, + 1766 + ], + "score": 0.92, + "latex": "\\{ 0 , 1 \\} ^ { \\dot { H } \\times W } \\} _ { i = 1 , \\cdots , N }" + }, + { + "category_id": 13, + "poly": [ + 1236, + 1303, + 1306, + 1303, + 1306, + 1337, + 1236, + 1337 + ], + "score": 0.92, + "latex": "\\sigma ( \\cdot , \\cdot )" + }, + { + "category_id": 13, + "poly": [ + 1272, + 1702, + 1402, + 1702, + 1402, + 1735, + 1272, + 1735 + ], + "score": 0.91, + "latex": "\\{ M _ { i } | M _ { i } \\ \\in" + }, + { + "category_id": 13, + "poly": [ + 298, + 1270, + 454, + 1270, + 454, + 1306, + 298, + 1306 + ], + "score": 0.91, + "latex": " \\mathbf { R } ^ { C } \\} _ { k = 1 , \\cdots , K }" + }, + { + "category_id": 13, + "poly": [ + 1279, + 1829, + 1393, + 1829, + 1393, + 1864, + 1279, + 1864 + ], + "score": 0.91, + "latex": "\\lambda \\in [ 0 , 1 ]" + }, + { + "category_id": 13, + "poly": [ + 628, + 1060, + 721, + 1060, + 721, + 1089, + 628, + 1089 + ], + "score": 0.91, + "latex": "H \\times W" + }, + { + "category_id": 13, + "poly": [ + 1297, + 1242, + 1403, + 1242, + 1403, + 1276, + 1297, + 1276 + ], + "score": 0.91, + "latex": "\\{ t _ { k } | t _ { k } \\in" + }, + { + "category_id": 13, + "poly": [ + 297, + 1831, + 342, + 1831, + 342, + 1865, + 297, + 1865 + ], + "score": 0.89, + "latex": "\\hat { p } _ { i , k }" + }, + { + "category_id": 13, + "poly": [ + 593, + 1303, + 1150, + 1303, + 1150, + 1338, + 593, + 1338 + ], + "score": 0.84, + "latex": "\\begin{array} { r } { p _ { i , k } = \\exp ( \\sigma ( v _ { i } , t _ { k } ) / \\tau ) / \\sum _ { k } ( \\exp ( \\sigma ( v _ { i } , t _ { k } ) / \\tau ) ) } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 713, + 1031, + 741, + 1031, + 741, + 1057, + 713, + 1057 + ], + "score": 0.84, + "latex": "N" + }, + { + "category_id": 13, + "poly": [ + 733, + 1425, + 752, + 1425, + 752, + 1455, + 733, + 1455 + ], + "score": 0.83, + "latex": "\\varnothing" + }, + { + "category_id": 13, + "poly": [ + 530, + 1091, + 555, + 1091, + 555, + 1119, + 530, + 1119 + ], + "score": 0.82, + "latex": "C" + }, + { + "category_id": 13, + "poly": [ + 949, + 1277, + 975, + 1277, + 975, + 1303, + 949, + 1303 + ], + "score": 0.82, + "latex": "v _ { i }" + }, + { + "category_id": 13, + "poly": [ + 926, + 1091, + 951, + 1091, + 951, + 1119, + 926, + 1119 + ], + "score": 0.81, + "latex": "C" + }, + { + "category_id": 13, + "poly": [ + 791, + 1153, + 817, + 1153, + 817, + 1179, + 791, + 1179 + ], + "score": 0.81, + "latex": "C" + }, + { + "category_id": 13, + "poly": [ + 298, + 1152, + 322, + 1152, + 322, + 1179, + 298, + 1179 + ], + "score": 0.79, + "latex": "C" + }, + { + "category_id": 13, + "poly": [ + 1152, + 1212, + 1183, + 1212, + 1183, + 1240, + 1152, + 1240 + ], + "score": 0.79, + "latex": "K" + }, + { + "category_id": 13, + "poly": [ + 1222, + 1795, + 1241, + 1795, + 1241, + 1821, + 1222, + 1821 + ], + "score": 0.78, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 405, + 1306, + 424, + 1306, + 424, + 1332, + 405, + 1332 + ], + "score": 0.78, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 946, + 1340, + 966, + 1340, + 966, + 1361, + 946, + 1361 + ], + "score": 0.77, + "latex": "\\tau" + }, + { + "category_id": 13, + "poly": [ + 894, + 1796, + 907, + 1796, + 907, + 1821, + 894, + 1821 + ], + "score": 0.7, + "latex": "i" + }, + { + "category_id": 13, + "poly": [ + 798, + 553, + 831, + 553, + 831, + 586, + 798, + 586 + ], + "score": 0.67, + "latex": "\\textcircled{2}" + }, + { + "category_id": 13, + "poly": [ + 392, + 553, + 425, + 553, + 425, + 585, + 392, + 585 + ], + "score": 0.43, + "latex": "\\textcircled{1}" + }, + { + "category_id": 13, + "poly": [ + 871, + 1243, + 896, + 1243, + 896, + 1271, + 871, + 1271 + ], + "score": 0.29, + "latex": "\\mathrm { K }" + }, + { + "category_id": 15, + "poly": [ + 1122.0, + 161.0, + 1217.0, + 161.0, + 1217.0, + 197.0, + 1122.0, + 197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 609.0, + 192.0, + 668.0, + 192.0, + 668.0, + 231.0, + 609.0, + 231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1069.0, + 198.0, + 1100.0, + 198.0, + 1100.0, + 227.0, + 1069.0, + 227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1173.0, + 203.0, + 1189.0, + 203.0, + 1189.0, + 218.0, + 1173.0, + 218.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1254.0, + 193.0, + 1342.0, + 193.0, + 1342.0, + 233.0, + 1254.0, + 233.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1173.0, + 233.0, + 1189.0, + 233.0, + 1189.0, + 246.0, + 1173.0, + 246.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 487.0, + 239.0, + 703.0, + 239.0, + 703.0, + 280.0, + 487.0, + 280.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1073.0, + 257.0, + 1092.0, + 257.0, + 1092.0, + 281.0, + 1073.0, + 281.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1139.0, + 261.0, + 1157.0, + 261.0, + 1157.0, + 281.0, + 1139.0, + 281.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1200.0, + 262.0, + 1213.0, + 262.0, + 1213.0, + 278.0, + 1200.0, + 278.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1233.0, + 261.0, + 1248.0, + 261.0, + 1248.0, + 279.0, + 1233.0, + 279.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1067.0, + 284.0, + 1105.0, + 284.0, + 1105.0, + 317.0, + 1067.0, + 317.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1250.0, + 285.0, + 1331.0, + 285.0, + 1331.0, + 320.0, + 1250.0, + 320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 784.0, + 320.0, + 910.0, + 320.0, + 910.0, + 373.0, + 784.0, + 373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 960.0, + 341.0, + 970.0, + 341.0, + 970.0, + 352.0, + 960.0, + 352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 984.0, + 323.0, + 1039.0, + 323.0, + 1039.0, + 356.0, + 984.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1092.0, + 327.0, + 1352.0, + 327.0, + 1352.0, + 369.0, + 1092.0, + 369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 536.0, + 357.0, + 648.0, + 357.0, + 648.0, + 388.0, + 536.0, + 388.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 985.0, + 352.0, + 1035.0, + 352.0, + 1035.0, + 381.0, + 985.0, + 381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 509.0, + 380.0, + 677.0, + 380.0, + 677.0, + 414.0, + 509.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1067.0, + 384.0, + 1100.0, + 384.0, + 1100.0, + 419.0, + 1067.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1174.0, + 395.0, + 1189.0, + 395.0, + 1189.0, + 408.0, + 1174.0, + 408.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1068.0, + 412.0, + 1101.0, + 412.0, + 1101.0, + 450.0, + 1068.0, + 450.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 603.0, + 449.0, + 694.0, + 449.0, + 694.0, + 483.0, + 603.0, + 483.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 709.0, + 481.0, + 784.0, + 481.0, + 784.0, + 519.0, + 709.0, + 519.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 968.0, + 429.0, + 1042.0, + 429.0, + 1042.0, + 491.0, + 968.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1067.0, + 452.0, + 1104.0, + 452.0, + 1104.0, + 507.0, + 1067.0, + 507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1115.0, + 452.0, + 1128.0, + 452.0, + 1128.0, + 469.0, + 1115.0, + 469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1143.0, + 453.0, + 1156.0, + 453.0, + 1156.0, + 469.0, + 1143.0, + 469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1174.0, + 486.0, + 1188.0, + 486.0, + 1188.0, + 495.0, + 1174.0, + 495.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1202.0, + 455.0, + 1212.0, + 455.0, + 1212.0, + 468.0, + 1202.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1232.0, + 451.0, + 1333.0, + 451.0, + 1333.0, + 510.0, + 1232.0, + 510.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 507.0, + 490.0, + 679.0, + 490.0, + 679.0, + 531.0, + 507.0, + 531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 509.0, + 771.0, + 509.0, + 771.0, + 539.0, + 721.0, + 539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1121.0, + 501.0, + 1217.0, + 501.0, + 1217.0, + 541.0, + 1121.0, + 541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 426.0, + 549.0, + 716.0, + 549.0, + 716.0, + 591.0, + 426.0, + 591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 791.0, + 549.0, + 797.0, + 549.0, + 797.0, + 595.0, + 791.0, + 595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 549.0, + 1349.0, + 549.0, + 1349.0, + 595.0, + 832.0, + 595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1071.0, + 226.0, + 1098.0, + 226.0, + 1098.0, + 256.0, + 1071.0, + 256.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1256.0, + 242.0, + 1354.0, + 242.0, + 1354.0, + 277.0, + 1256.0, + 277.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1107.75, + 262.5, + 1130.75, + 262.5, + 1130.75, + 277.5, + 1107.75, + 277.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1251.0, + 381.0, + 1343.0, + 381.0, + 1343.0, + 420.0, + 1251.0, + 420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1256.0, + 428.0, + 1354.0, + 428.0, + 1354.0, + 467.0, + 1256.0, + 467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 371.0, + 495.0, + 462.0, + 495.0, + 462.0, + 530.0, + 371.0, + 530.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 605.0, + 1405.0, + 605.0, + 1405.0, + 640.0, + 295.0, + 640.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 634.0, + 1405.0, + 634.0, + 1405.0, + 672.0, + 293.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 666.0, + 1405.0, + 666.0, + 1405.0, + 701.0, + 294.0, + 701.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 695.0, + 1406.0, + 695.0, + 1406.0, + 736.0, + 290.0, + 736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 724.0, + 1403.0, + 724.0, + 1403.0, + 764.0, + 293.0, + 764.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 759.0, + 755.0, + 759.0, + 755.0, + 792.0, + 293.0, + 792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 297.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 828.0, + 1244.0, + 828.0, + 1244.0, + 863.0, + 295.0, + 863.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2086.0, + 862.0, + 2086.0, + 862.0, + 2118.0, + 838.0, + 2118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 967.0, + 1405.0, + 967.0, + 1405.0, + 1004.0, + 293.0, + 1004.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 999.0, + 1405.0, + 999.0, + 1405.0, + 1035.0, + 294.0, + 1035.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1026.0, + 712.0, + 1026.0, + 712.0, + 1067.0, + 293.0, + 1067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 742.0, + 1026.0, + 1405.0, + 1026.0, + 1405.0, + 1067.0, + 742.0, + 1067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1058.0, + 627.0, + 1058.0, + 627.0, + 1096.0, + 292.0, + 1096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 722.0, + 1058.0, + 1405.0, + 1058.0, + 1405.0, + 1096.0, + 722.0, + 1096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1085.0, + 529.0, + 1085.0, + 529.0, + 1131.0, + 292.0, + 1131.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 556.0, + 1085.0, + 925.0, + 1085.0, + 925.0, + 1131.0, + 556.0, + 1131.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 952.0, + 1085.0, + 1408.0, + 1085.0, + 1408.0, + 1131.0, + 952.0, + 1131.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1119.0, + 1408.0, + 1119.0, + 1408.0, + 1156.0, + 293.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1149.0, + 790.0, + 1149.0, + 790.0, + 1186.0, + 323.0, + 1186.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 818.0, + 1149.0, + 1406.0, + 1149.0, + 1406.0, + 1186.0, + 818.0, + 1186.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1178.0, + 1404.0, + 1178.0, + 1404.0, + 1218.0, + 292.0, + 1218.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1212.0, + 1151.0, + 1212.0, + 1151.0, + 1248.0, + 293.0, + 1248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1184.0, + 1212.0, + 1406.0, + 1212.0, + 1406.0, + 1248.0, + 1184.0, + 1248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1240.0, + 870.0, + 1240.0, + 870.0, + 1279.0, + 292.0, + 1279.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 897.0, + 1240.0, + 1296.0, + 1240.0, + 1296.0, + 1279.0, + 897.0, + 1279.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1266.0, + 297.0, + 1266.0, + 297.0, + 1314.0, + 291.0, + 1314.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 455.0, + 1266.0, + 948.0, + 1266.0, + 948.0, + 1314.0, + 455.0, + 1314.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1266.0, + 1409.0, + 1266.0, + 1409.0, + 1314.0, + 976.0, + 1314.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1300.0, + 404.0, + 1300.0, + 404.0, + 1344.0, + 292.0, + 1344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 425.0, + 1300.0, + 592.0, + 1300.0, + 592.0, + 1344.0, + 425.0, + 1344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1151.0, + 1300.0, + 1235.0, + 1300.0, + 1235.0, + 1344.0, + 1151.0, + 1344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1307.0, + 1300.0, + 1409.0, + 1300.0, + 1409.0, + 1344.0, + 1307.0, + 1344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1331.0, + 945.0, + 1331.0, + 945.0, + 1370.0, + 291.0, + 1370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 967.0, + 1331.0, + 1406.0, + 1331.0, + 1406.0, + 1370.0, + 967.0, + 1370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1363.0, + 1405.0, + 1363.0, + 1405.0, + 1400.0, + 293.0, + 1400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1389.0, + 1405.0, + 1389.0, + 1405.0, + 1433.0, + 291.0, + 1433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1420.0, + 732.0, + 1420.0, + 732.0, + 1464.0, + 293.0, + 1464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 753.0, + 1420.0, + 1405.0, + 1420.0, + 1405.0, + 1464.0, + 753.0, + 1464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1455.0, + 1031.0, + 1455.0, + 1031.0, + 1490.0, + 293.0, + 1490.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1669.0, + 1404.0, + 1669.0, + 1404.0, + 1705.0, + 293.0, + 1705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1701.0, + 1271.0, + 1701.0, + 1271.0, + 1737.0, + 294.0, + 1737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1720.0, + 297.0, + 1720.0, + 297.0, + 1772.0, + 291.0, + 1772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 1720.0, + 1409.0, + 1720.0, + 1409.0, + 1772.0, + 537.0, + 1772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1762.0, + 1405.0, + 1762.0, + 1405.0, + 1797.0, + 294.0, + 1797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1791.0, + 893.0, + 1791.0, + 893.0, + 1829.0, + 292.0, + 1829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 908.0, + 1791.0, + 1221.0, + 1791.0, + 1221.0, + 1829.0, + 908.0, + 1829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1242.0, + 1791.0, + 1405.0, + 1791.0, + 1405.0, + 1829.0, + 1242.0, + 1829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 287.0, + 1815.0, + 296.0, + 1815.0, + 296.0, + 1881.0, + 287.0, + 1881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 343.0, + 1815.0, + 1049.0, + 1815.0, + 1049.0, + 1881.0, + 343.0, + 1881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1199.0, + 1815.0, + 1278.0, + 1815.0, + 1278.0, + 1881.0, + 1199.0, + 1881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1394.0, + 1815.0, + 1404.0, + 1815.0, + 1404.0, + 1881.0, + 1394.0, + 1881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1862.0, + 1348.0, + 1862.0, + 1348.0, + 1902.0, + 294.0, + 1902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1500.0, + 1402.0, + 1500.0, + 1402.0, + 1537.0, + 294.0, + 1537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1527.0, + 1405.0, + 1527.0, + 1405.0, + 1571.0, + 292.0, + 1571.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1562.0, + 1405.0, + 1562.0, + 1405.0, + 1600.0, + 292.0, + 1600.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1590.0, + 1405.0, + 1590.0, + 1405.0, + 1629.0, + 293.0, + 1629.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1623.0, + 1228.0, + 1623.0, + 1228.0, + 1664.0, + 293.0, + 1664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1909.0, + 1406.0, + 1909.0, + 1406.0, + 1946.0, + 294.0, + 1946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1943.0, + 1405.0, + 1943.0, + 1405.0, + 1976.0, + 295.0, + 1976.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1974.0, + 1405.0, + 1974.0, + 1405.0, + 2006.0, + 295.0, + 2006.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 2002.0, + 1405.0, + 2002.0, + 1405.0, + 2038.0, + 293.0, + 2038.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 889.0, + 1405.0, + 889.0, + 1405.0, + 924.0, + 295.0, + 924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 921.0, + 1375.0, + 921.0, + 1375.0, + 956.0, + 294.0, + 956.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 3, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1497, + 1404, + 1497, + 1404, + 1957, + 298, + 1957 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 297, + 419, + 1404, + 419, + 1404, + 695, + 297, + 695 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 299, + 971, + 845, + 971, + 845, + 1304, + 299, + 1304 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 299, + 711, + 846, + 711, + 846, + 955, + 299, + 955 + ], + "score": 0.979 + }, + { + "category_id": 3, + "poly": [ + 869, + 722, + 1401, + 722, + 1401, + 1147, + 869, + 1147 + ], + "score": 0.966 + }, + { + "category_id": 4, + "poly": [ + 867, + 1166, + 1402, + 1166, + 1402, + 1290, + 867, + 1290 + ], + "score": 0.963 + }, + { + "category_id": 1, + "poly": [ + 301, + 1307, + 1403, + 1307, + 1403, + 1398, + 301, + 1398 + ], + "score": 0.958 + }, + { + "category_id": 1, + "poly": [ + 300, + 229, + 1401, + 229, + 1401, + 323, + 300, + 323 + ], + "score": 0.956 + }, + { + "category_id": 1, + "poly": [ + 298, + 1972, + 1399, + 1972, + 1399, + 2034, + 298, + 2034 + ], + "score": 0.938 + }, + { + "category_id": 0, + "poly": [ + 301, + 1437, + 645, + 1437, + 645, + 1468, + 301, + 1468 + ], + "score": 0.885 + }, + { + "category_id": 2, + "poly": [ + 298, + 76, + 856, + 76, + 856, + 104, + 298, + 104 + ], + "score": 0.875 + }, + { + "category_id": 0, + "poly": [ + 300, + 364, + 1123, + 364, + 1123, + 392, + 300, + 392 + ], + "score": 0.867 + }, + { + "category_id": 2, + "poly": [ + 841, + 2088, + 858, + 2088, + 858, + 2112, + 841, + 2112 + ], + "score": 0.714 + }, + { + "category_id": 13, + "poly": [ + 642, + 1737, + 802, + 1737, + 802, + 1769, + 642, + 1769 + ], + "score": 0.92, + "latex": "T \\in \\mathbf { R } ^ { N _ { p } \\times E }" + }, + { + "category_id": 13, + "poly": [ + 1212, + 1770, + 1393, + 1770, + 1393, + 1807, + 1212, + 1807 + ], + "score": 0.92, + "latex": "M _ { p } \\in \\{ 0 , 1 \\} ^ { N _ { p } }" + }, + { + "category_id": 13, + "poly": [ + 725, + 1831, + 882, + 1831, + 882, + 1863, + 725, + 1863 + ], + "score": 0.92, + "latex": "P \\in \\mathbf { R } ^ { \\dot { N } _ { p } \\times E }" + }, + { + "category_id": 13, + "poly": [ + 445, + 1865, + 746, + 1865, + 746, + 1898, + 445, + 1898 + ], + "score": 0.91, + "latex": "T \\otimes M _ { p } + P \\otimes ( 1 - M _ { p } )" + }, + { + "category_id": 13, + "poly": [ + 897, + 1740, + 936, + 1740, + 936, + 1774, + 897, + 1774 + ], + "score": 0.9, + "latex": "N _ { p }" + }, + { + "category_id": 13, + "poly": [ + 1365, + 1019, + 1385, + 1019, + 1385, + 1039, + 1365, + 1039 + ], + "score": 0.83, + "latex": "t _ { 3 }" + }, + { + "category_id": 13, + "poly": [ + 833, + 1868, + 860, + 1868, + 860, + 1894, + 833, + 1894 + ], + "score": 0.82, + "latex": "\\otimes" + }, + { + "category_id": 13, + "poly": [ + 1301, + 1741, + 1327, + 1741, + 1327, + 1768, + 1301, + 1768 + ], + "score": 0.81, + "latex": "E" + }, + { + "category_id": 13, + "poly": [ + 1323, + 1019, + 1342, + 1019, + 1342, + 1039, + 1323, + 1039 + ], + "score": 0.71, + "latex": "t _ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 1207, + 955, + 1234, + 955, + 1234, + 978, + 1207, + 978 + ], + "score": 0.65, + "latex": "{ \\hat { v } } _ { N }" + }, + { + "category_id": 13, + "poly": [ + 1274, + 1019, + 1293, + 1019, + 1293, + 1039, + 1274, + 1039 + ], + "score": 0.47, + "latex": "t _ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 1209, + 808, + 1233, + 808, + 1233, + 832, + 1209, + 832 + ], + "score": 0.4, + "latex": "\\hat { v } _ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 1209, + 755, + 1233, + 755, + 1233, + 781, + 1209, + 781 + ], + "score": 0.27, + "latex": "\\widehat { v } _ { 1 }" + }, + { + "category_id": 15, + "poly": [ + 1056.0, + 725.0, + 1133.0, + 725.0, + 1133.0, + 856.0, + 1056.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1203.0, + 748.0, + 1208.0, + 748.0, + 1208.0, + 788.0, + 1203.0, + 788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1234.0, + 748.0, + 1239.0, + 748.0, + 1239.0, + 788.0, + 1234.0, + 788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1262.0, + 757.0, + 1305.0, + 757.0, + 1305.0, + 788.0, + 1262.0, + 788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1308.0, + 758.0, + 1350.0, + 758.0, + 1350.0, + 788.0, + 1308.0, + 788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1353.0, + 758.0, + 1396.0, + 758.0, + 1396.0, + 788.0, + 1353.0, + 788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1201.0, + 797.0, + 1208.0, + 797.0, + 1208.0, + 842.0, + 1201.0, + 842.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1234.0, + 797.0, + 1241.0, + 797.0, + 1241.0, + 842.0, + 1234.0, + 842.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1262.0, + 807.0, + 1305.0, + 807.0, + 1305.0, + 837.0, + 1262.0, + 837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1309.0, + 807.0, + 1350.0, + 807.0, + 1350.0, + 837.0, + 1309.0, + 837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1354.0, + 807.0, + 1396.0, + 807.0, + 1396.0, + 837.0, + 1354.0, + 837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 997.0, + 835.0, + 1052.0, + 835.0, + 1052.0, + 866.0, + 997.0, + 866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1085.0, + 845.0, + 1099.0, + 845.0, + 1099.0, + 869.0, + 1085.0, + 869.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1129.0, + 839.0, + 1183.0, + 839.0, + 1183.0, + 871.0, + 1129.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 991.0, + 857.0, + 1057.0, + 857.0, + 1057.0, + 889.0, + 991.0, + 889.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1210.0, + 858.0, + 1228.0, + 858.0, + 1228.0, + 884.0, + 1210.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1269.0, + 858.0, + 1287.0, + 858.0, + 1287.0, + 885.0, + 1269.0, + 885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1317.0, + 860.0, + 1332.0, + 860.0, + 1332.0, + 884.0, + 1317.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1361.0, + 859.0, + 1380.0, + 859.0, + 1380.0, + 885.0, + 1361.0, + 885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1101.0, + 901.0, + 1118.0, + 901.0, + 1118.0, + 913.0, + 1101.0, + 913.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1090.0, + 910.0, + 1101.0, + 910.0, + 1101.0, + 917.0, + 1090.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1262.0, + 909.0, + 1305.0, + 909.0, + 1305.0, + 939.0, + 1262.0, + 939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1309.0, + 908.0, + 1351.0, + 908.0, + 1351.0, + 938.0, + 1309.0, + 938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1353.0, + 909.0, + 1396.0, + 909.0, + 1396.0, + 939.0, + 1353.0, + 939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1061.0, + 923.0, + 1126.0, + 923.0, + 1126.0, + 995.0, + 1061.0, + 995.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1200.0, + 944.0, + 1206.0, + 944.0, + 1206.0, + 987.0, + 1200.0, + 987.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1235.0, + 944.0, + 1242.0, + 944.0, + 1242.0, + 987.0, + 1235.0, + 987.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1262.0, + 958.0, + 1304.0, + 958.0, + 1304.0, + 989.0, + 1262.0, + 989.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1309.0, + 958.0, + 1350.0, + 958.0, + 1350.0, + 989.0, + 1309.0, + 989.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1353.0, + 957.0, + 1397.0, + 957.0, + 1397.0, + 991.0, + 1353.0, + 991.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 874.0, + 1037.0, + 968.0, + 1037.0, + 968.0, + 1068.0, + 874.0, + 1068.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 991.0, + 1049.0, + 1048.0, + 1049.0, + 1048.0, + 1085.0, + 991.0, + 1085.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1055.0, + 1031.0, + 1115.0, + 1031.0, + 1115.0, + 1055.0, + 1055.0, + 1055.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1267.0, + 1009.0, + 1273.0, + 1009.0, + 1273.0, + 1045.0, + 1267.0, + 1045.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1294.0, + 1009.0, + 1300.0, + 1009.0, + 1300.0, + 1045.0, + 1294.0, + 1045.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1317.0, + 1009.0, + 1322.0, + 1009.0, + 1322.0, + 1066.0, + 1317.0, + 1066.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1343.0, + 1009.0, + 1349.0, + 1009.0, + 1349.0, + 1066.0, + 1343.0, + 1066.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1358.0, + 1010.0, + 1364.0, + 1010.0, + 1364.0, + 1046.0, + 1358.0, + 1046.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1386.0, + 1010.0, + 1392.0, + 1010.0, + 1392.0, + 1046.0, + 1386.0, + 1046.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 871.0, + 1058.0, + 978.0, + 1058.0, + 978.0, + 1125.0, + 871.0, + 1125.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 990.0, + 1072.0, + 1049.0, + 1072.0, + 1049.0, + 1105.0, + 990.0, + 1105.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1056.0, + 1067.0, + 1124.0, + 1067.0, + 1124.0, + 1133.0, + 1056.0, + 1133.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1130.0, + 1058.0, + 1182.0, + 1058.0, + 1182.0, + 1109.0, + 1130.0, + 1109.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1127.0, + 862.5, + 1185.0, + 862.5, + 1185.0, + 894.0, + 1127.0, + 894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1196.0, + 905.5, + 1247.0, + 905.5, + 1247.0, + 947.5, + 1196.0, + 947.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1270.0, + 1084.0, + 1297.0, + 1084.0, + 1297.0, + 1102.0, + 1270.0, + 1102.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1317.75, + 1086.5, + 1346.75, + 1086.5, + 1346.75, + 1106.0, + 1317.75, + 1106.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1361.0, + 1086.0, + 1393.0, + 1086.0, + 1393.0, + 1104.5, + 1361.0, + 1104.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 866.0, + 1163.0, + 1403.0, + 1163.0, + 1403.0, + 1203.0, + 866.0, + 1203.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 869.0, + 1197.0, + 1403.0, + 1197.0, + 1403.0, + 1231.0, + 869.0, + 1231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 868.0, + 1226.0, + 1405.0, + 1226.0, + 1405.0, + 1262.0, + 868.0, + 1262.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 866.0, + 1255.0, + 1341.0, + 1255.0, + 1341.0, + 1297.0, + 866.0, + 1297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1437.0, + 646.0, + 1437.0, + 646.0, + 1470.0, + 296.0, + 1470.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 361.0, + 1127.0, + 361.0, + 1127.0, + 398.0, + 293.0, + 398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2085.0, + 861.0, + 2085.0, + 861.0, + 2120.0, + 839.0, + 2120.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1494.0, + 1405.0, + 1494.0, + 1405.0, + 1530.0, + 296.0, + 1530.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1526.0, + 1405.0, + 1526.0, + 1405.0, + 1561.0, + 294.0, + 1561.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1556.0, + 1405.0, + 1556.0, + 1405.0, + 1591.0, + 294.0, + 1591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1586.0, + 1408.0, + 1586.0, + 1408.0, + 1625.0, + 293.0, + 1625.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1612.0, + 1408.0, + 1612.0, + 1408.0, + 1656.0, + 293.0, + 1656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1647.0, + 1406.0, + 1647.0, + 1406.0, + 1682.0, + 294.0, + 1682.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1677.0, + 1407.0, + 1677.0, + 1407.0, + 1714.0, + 292.0, + 1714.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1704.0, + 1406.0, + 1704.0, + 1406.0, + 1746.0, + 292.0, + 1746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1728.0, + 641.0, + 1728.0, + 641.0, + 1780.0, + 290.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 803.0, + 1728.0, + 896.0, + 1728.0, + 896.0, + 1780.0, + 803.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 937.0, + 1728.0, + 1300.0, + 1728.0, + 1300.0, + 1780.0, + 937.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1328.0, + 1728.0, + 1409.0, + 1728.0, + 1409.0, + 1780.0, + 1328.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1770.0, + 1211.0, + 1770.0, + 1211.0, + 1807.0, + 292.0, + 1807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1394.0, + 1770.0, + 1404.0, + 1770.0, + 1404.0, + 1807.0, + 1394.0, + 1807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1803.0, + 1406.0, + 1803.0, + 1406.0, + 1838.0, + 296.0, + 1838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1824.0, + 724.0, + 1824.0, + 724.0, + 1872.0, + 291.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 883.0, + 1824.0, + 1408.0, + 1824.0, + 1408.0, + 1872.0, + 883.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1859.0, + 444.0, + 1859.0, + 444.0, + 1905.0, + 292.0, + 1905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 747.0, + 1859.0, + 832.0, + 1859.0, + 832.0, + 1905.0, + 747.0, + 1905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 861.0, + 1859.0, + 1406.0, + 1859.0, + 1406.0, + 1905.0, + 861.0, + 1905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1894.0, + 1406.0, + 1894.0, + 1406.0, + 1929.0, + 294.0, + 1929.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1927.0, + 867.0, + 1927.0, + 867.0, + 1958.0, + 296.0, + 1958.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 417.0, + 1405.0, + 417.0, + 1405.0, + 456.0, + 294.0, + 456.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 449.0, + 1404.0, + 449.0, + 1404.0, + 487.0, + 294.0, + 487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 476.0, + 1405.0, + 476.0, + 1405.0, + 522.0, + 291.0, + 522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 507.0, + 1405.0, + 507.0, + 1405.0, + 549.0, + 292.0, + 549.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 540.0, + 1405.0, + 540.0, + 1405.0, + 579.0, + 292.0, + 579.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 571.0, + 1405.0, + 571.0, + 1405.0, + 608.0, + 295.0, + 608.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 600.0, + 1405.0, + 600.0, + 1405.0, + 638.0, + 294.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 631.0, + 1404.0, + 631.0, + 1404.0, + 671.0, + 292.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 661.0, + 763.0, + 661.0, + 763.0, + 702.0, + 294.0, + 702.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 970.0, + 848.0, + 970.0, + 848.0, + 1004.0, + 295.0, + 1004.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1001.0, + 848.0, + 1001.0, + 848.0, + 1035.0, + 295.0, + 1035.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1029.0, + 848.0, + 1029.0, + 848.0, + 1066.0, + 295.0, + 1066.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1063.0, + 849.0, + 1063.0, + 849.0, + 1096.0, + 294.0, + 1096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1090.0, + 849.0, + 1090.0, + 849.0, + 1126.0, + 294.0, + 1126.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1122.0, + 849.0, + 1122.0, + 849.0, + 1158.0, + 295.0, + 1158.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1150.0, + 850.0, + 1150.0, + 850.0, + 1189.0, + 292.0, + 1189.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1183.0, + 850.0, + 1183.0, + 850.0, + 1219.0, + 291.0, + 1219.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1214.0, + 850.0, + 1214.0, + 850.0, + 1247.0, + 293.0, + 1247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1245.0, + 850.0, + 1245.0, + 850.0, + 1278.0, + 294.0, + 1278.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1274.0, + 850.0, + 1274.0, + 850.0, + 1309.0, + 295.0, + 1309.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 705.0, + 849.0, + 705.0, + 849.0, + 747.0, + 295.0, + 747.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 742.0, + 848.0, + 742.0, + 848.0, + 772.0, + 296.0, + 772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 770.0, + 849.0, + 770.0, + 849.0, + 806.0, + 293.0, + 806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 801.0, + 849.0, + 801.0, + 849.0, + 836.0, + 293.0, + 836.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 831.0, + 851.0, + 831.0, + 851.0, + 866.0, + 292.0, + 866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 863.0, + 848.0, + 863.0, + 848.0, + 897.0, + 293.0, + 897.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 891.0, + 850.0, + 891.0, + 850.0, + 929.0, + 293.0, + 929.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 922.0, + 711.0, + 922.0, + 711.0, + 958.0, + 294.0, + 958.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1302.0, + 1407.0, + 1302.0, + 1407.0, + 1343.0, + 295.0, + 1343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1335.0, + 1404.0, + 1335.0, + 1404.0, + 1374.0, + 295.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1367.0, + 1403.0, + 1367.0, + 1403.0, + 1400.0, + 297.0, + 1400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 226.0, + 1405.0, + 226.0, + 1405.0, + 268.0, + 294.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 263.0, + 1403.0, + 263.0, + 1403.0, + 297.0, + 296.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 291.0, + 1086.0, + 291.0, + 1086.0, + 328.0, + 295.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1971.0, + 1404.0, + 1971.0, + 1404.0, + 2007.0, + 297.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2003.0, + 1404.0, + 2003.0, + 1404.0, + 2038.0, + 294.0, + 2038.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 4, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1239, + 1404, + 1239, + 1404, + 1604, + 298, + 1604 + ], + "score": 0.984 + }, + { + "category_id": 1, + "poly": [ + 297, + 1620, + 1404, + 1620, + 1404, + 1957, + 297, + 1957 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 666, + 1403, + 666, + 1403, + 881, + 298, + 881 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 298, + 1038, + 1404, + 1038, + 1404, + 1224, + 298, + 1224 + ], + "score": 0.98 + }, + { + "category_id": 3, + "poly": [ + 296, + 220, + 1345, + 220, + 1345, + 559, + 296, + 559 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 298, + 1971, + 1401, + 1971, + 1401, + 2035, + 298, + 2035 + ], + "score": 0.937 + }, + { + "category_id": 0, + "poly": [ + 300, + 983, + 637, + 983, + 637, + 1013, + 300, + 1013 + ], + "score": 0.887 + }, + { + "category_id": 0, + "poly": [ + 300, + 913, + 557, + 913, + 557, + 947, + 300, + 947 + ], + "score": 0.881 + }, + { + "category_id": 4, + "poly": [ + 296, + 588, + 1401, + 588, + 1401, + 652, + 296, + 652 + ], + "score": 0.831 + }, + { + "category_id": 2, + "poly": [ + 298, + 76, + 857, + 76, + 857, + 104, + 298, + 104 + ], + "score": 0.816 + }, + { + "category_id": 2, + "poly": [ + 840, + 2089, + 858, + 2089, + 858, + 2112, + 840, + 2112 + ], + "score": 0.777 + }, + { + "category_id": 2, + "poly": [ + 300, + 76, + 854, + 76, + 854, + 104, + 300, + 104 + ], + "score": 0.336 + }, + { + "category_id": 1, + "poly": [ + 296, + 588, + 1401, + 588, + 1401, + 652, + 296, + 652 + ], + "score": 0.137 + }, + { + "category_id": 13, + "poly": [ + 981, + 1863, + 1079, + 1863, + 1079, + 1893, + 981, + 1893 + ], + "score": 0.91, + "latex": "{ \\bar { 6 } } \\cdot 1 0 ^ { - 5 }" + }, + { + "category_id": 13, + "poly": [ + 503, + 1895, + 626, + 1895, + 626, + 1924, + 503, + 1924 + ], + "score": 0.91, + "latex": "6 4 0 \\times 6 4 0" + }, + { + "category_id": 13, + "poly": [ + 1134, + 1863, + 1194, + 1863, + 1194, + 1893, + 1134, + 1893 + ], + "score": 0.89, + "latex": "1 0 ^ { - 2 }" + }, + { + "category_id": 13, + "poly": [ + 1193, + 229, + 1224, + 229, + 1224, + 259, + 1193, + 259 + ], + "score": 0.56, + "latex": "\\textcircled{ R }" + }, + { + "category_id": 15, + "poly": [ + 336.0, + 227.0, + 511.0, + 227.0, + 511.0, + 264.0, + 336.0, + 264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 578.0, + 230.0, + 998.0, + 230.0, + 998.0, + 257.0, + 578.0, + 257.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1037.0, + 228.0, + 1192.0, + 228.0, + 1192.0, + 258.0, + 1037.0, + 258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1225.0, + 228.0, + 1350.0, + 228.0, + 1350.0, + 258.0, + 1225.0, + 258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1012.0, + 281.0, + 1047.0, + 281.0, + 1047.0, + 303.0, + 1012.0, + 303.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1075.0, + 282.0, + 1114.0, + 282.0, + 1114.0, + 305.0, + 1075.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1142.0, + 280.0, + 1188.0, + 280.0, + 1188.0, + 305.0, + 1142.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1249.0, + 267.0, + 1320.0, + 267.0, + 1320.0, + 336.0, + 1249.0, + 336.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 342.0, + 313.0, + 453.0, + 313.0, + 453.0, + 342.0, + 342.0, + 342.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 307.0, + 652.0, + 307.0, + 652.0, + 338.0, + 600.0, + 338.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1010.0, + 303.0, + 1114.0, + 303.0, + 1114.0, + 337.0, + 1010.0, + 337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1145.0, + 311.0, + 1183.0, + 311.0, + 1183.0, + 332.0, + 1145.0, + 332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 343.0, + 333.0, + 452.0, + 333.0, + 452.0, + 362.0, + 343.0, + 362.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 337.0, + 649.0, + 337.0, + 649.0, + 367.0, + 600.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 327.0, + 1115.0, + 327.0, + 1115.0, + 366.0, + 1006.0, + 366.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1145.0, + 337.0, + 1181.0, + 337.0, + 1181.0, + 358.0, + 1145.0, + 358.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1181.0, + 358.0, + 1208.0, + 358.0, + 1208.0, + 468.0, + 1181.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1221.0, + 358.0, + 1249.0, + 358.0, + 1249.0, + 468.0, + 1221.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1114.0, + 361.0, + 1141.0, + 361.0, + 1141.0, + 472.0, + 1114.0, + 472.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1297.0, + 368.0, + 1306.0, + 368.0, + 1306.0, + 381.0, + 1297.0, + 381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 573.0, + 375.0, + 675.0, + 375.0, + 675.0, + 433.0, + 573.0, + 433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 944.0, + 403.0, + 1006.0, + 403.0, + 1006.0, + 438.0, + 944.0, + 438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1014.0, + 413.0, + 1075.0, + 413.0, + 1075.0, + 439.0, + 1014.0, + 439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1077.0, + 414.0, + 1113.0, + 414.0, + 1113.0, + 436.0, + 1077.0, + 436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1147.0, + 417.0, + 1178.0, + 417.0, + 1178.0, + 434.0, + 1147.0, + 434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1261.0, + 379.0, + 1343.0, + 379.0, + 1343.0, + 416.0, + 1261.0, + 416.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 617.0, + 426.0, + 633.0, + 426.0, + 633.0, + 443.0, + 617.0, + 443.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 946.0, + 429.0, + 1004.0, + 429.0, + 1004.0, + 464.0, + 946.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1010.0, + 434.0, + 1113.0, + 434.0, + 1113.0, + 469.0, + 1010.0, + 469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1145.0, + 439.0, + 1181.0, + 439.0, + 1181.0, + 461.0, + 1145.0, + 461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 593.0, + 450.0, + 661.0, + 450.0, + 661.0, + 490.0, + 593.0, + 490.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1020.0, + 466.0, + 1034.0, + 466.0, + 1034.0, + 483.0, + 1020.0, + 483.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 483.0, + 649.0, + 483.0, + 649.0, + 513.0, + 600.0, + 513.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 740.0, + 507.0, + 828.0, + 507.0, + 828.0, + 544.0, + 740.0, + 544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1013.0, + 513.0, + 1113.0, + 513.0, + 1113.0, + 537.0, + 1013.0, + 537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1145.0, + 515.0, + 1181.0, + 515.0, + 1181.0, + 537.0, + 1145.0, + 537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 423.0, + 526.0, + 552.0, + 526.0, + 552.0, + 563.0, + 423.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 983.0, + 639.0, + 983.0, + 639.0, + 1015.0, + 296.0, + 1015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 910.0, + 561.0, + 910.0, + 561.0, + 953.0, + 293.0, + 953.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 585.0, + 1405.0, + 585.0, + 1405.0, + 626.0, + 292.0, + 626.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 620.0, + 1239.0, + 620.0, + 1239.0, + 656.0, + 296.0, + 656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 857.0, + 73.0, + 857.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2087.0, + 861.0, + 2087.0, + 861.0, + 2118.0, + 840.0, + 2118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 297.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1237.0, + 1405.0, + 1237.0, + 1405.0, + 1274.0, + 293.0, + 1274.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1266.0, + 1405.0, + 1266.0, + 1405.0, + 1306.0, + 292.0, + 1306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1298.0, + 1404.0, + 1298.0, + 1404.0, + 1334.0, + 293.0, + 1334.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1327.0, + 1405.0, + 1327.0, + 1405.0, + 1368.0, + 291.0, + 1368.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1361.0, + 1404.0, + 1361.0, + 1404.0, + 1395.0, + 294.0, + 1395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1392.0, + 1405.0, + 1392.0, + 1405.0, + 1426.0, + 293.0, + 1426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1424.0, + 1402.0, + 1424.0, + 1402.0, + 1454.0, + 294.0, + 1454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1453.0, + 1405.0, + 1453.0, + 1405.0, + 1487.0, + 296.0, + 1487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1478.0, + 1404.0, + 1478.0, + 1404.0, + 1521.0, + 292.0, + 1521.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1514.0, + 1404.0, + 1514.0, + 1404.0, + 1547.0, + 294.0, + 1547.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1543.0, + 1405.0, + 1543.0, + 1405.0, + 1580.0, + 293.0, + 1580.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1574.0, + 801.0, + 1574.0, + 801.0, + 1608.0, + 294.0, + 1608.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1618.0, + 1404.0, + 1618.0, + 1404.0, + 1654.0, + 294.0, + 1654.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1651.0, + 1405.0, + 1651.0, + 1405.0, + 1684.0, + 292.0, + 1684.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1680.0, + 1405.0, + 1680.0, + 1405.0, + 1715.0, + 294.0, + 1715.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1713.0, + 1402.0, + 1713.0, + 1402.0, + 1744.0, + 294.0, + 1744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1740.0, + 1405.0, + 1740.0, + 1405.0, + 1776.0, + 292.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1774.0, + 1406.0, + 1774.0, + 1406.0, + 1808.0, + 295.0, + 1808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1803.0, + 1406.0, + 1803.0, + 1406.0, + 1838.0, + 294.0, + 1838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1832.0, + 1404.0, + 1832.0, + 1404.0, + 1869.0, + 294.0, + 1869.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1862.0, + 980.0, + 1862.0, + 980.0, + 1900.0, + 294.0, + 1900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1080.0, + 1862.0, + 1133.0, + 1862.0, + 1133.0, + 1900.0, + 1080.0, + 1900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1862.0, + 1406.0, + 1862.0, + 1406.0, + 1900.0, + 1195.0, + 1900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1893.0, + 502.0, + 1893.0, + 502.0, + 1928.0, + 295.0, + 1928.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 627.0, + 1893.0, + 1405.0, + 1893.0, + 1405.0, + 1928.0, + 627.0, + 1928.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1926.0, + 1404.0, + 1926.0, + 1404.0, + 1961.0, + 296.0, + 1961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 668.0, + 1405.0, + 668.0, + 1405.0, + 702.0, + 296.0, + 702.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 695.0, + 1404.0, + 695.0, + 1404.0, + 735.0, + 292.0, + 735.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 725.0, + 1403.0, + 725.0, + 1403.0, + 765.0, + 293.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 757.0, + 1407.0, + 757.0, + 1407.0, + 794.0, + 292.0, + 794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 788.0, + 1405.0, + 788.0, + 1405.0, + 825.0, + 293.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 819.0, + 1408.0, + 819.0, + 1408.0, + 857.0, + 293.0, + 857.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 850.0, + 1235.0, + 850.0, + 1235.0, + 884.0, + 294.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1040.0, + 1404.0, + 1040.0, + 1404.0, + 1072.0, + 297.0, + 1072.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1068.0, + 1405.0, + 1068.0, + 1405.0, + 1105.0, + 293.0, + 1105.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1097.0, + 1406.0, + 1097.0, + 1406.0, + 1137.0, + 292.0, + 1137.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1128.0, + 1404.0, + 1128.0, + 1404.0, + 1167.0, + 293.0, + 1167.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1159.0, + 1405.0, + 1159.0, + 1405.0, + 1199.0, + 293.0, + 1199.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1190.0, + 648.0, + 1190.0, + 648.0, + 1231.0, + 295.0, + 1231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1971.0, + 1403.0, + 1971.0, + 1403.0, + 2007.0, + 296.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2000.0, + 1405.0, + 2000.0, + 1405.0, + 2039.0, + 294.0, + 2039.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 585.0, + 1405.0, + 585.0, + 1405.0, + 626.0, + 292.0, + 626.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 620.0, + 1239.0, + 620.0, + 1239.0, + 656.0, + 296.0, + 656.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 5, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 292, + 359, + 1425, + 359, + 1425, + 957, + 292, + 957 + ], + "score": 0.985, + "html": "
methodbackbonetraining datasetA-847PC-459A-150PC-59PAS-20
Open-vocabulary generalist models.
SPNet (Xian et al., 2019)R-101PASCAL-15=24.318.3
ZS3Net (Bucher et al., 2019)R-101PASCAL-1519.438.3
LSeg (Li et al., 2022)R-101PASCAL-15=-47.4
LSeg+ (Ghiasi et al., 2021)R-101COCO Panoptic2.55.213.036.059.0
SimBaseline (Xu et al.,2021)R-101cCOCO-Stuff-156--15.3-74.5
ZegFormer (Ding et al.,2022)R-50COCO-Stuff-156==16.4-80.7
OpenSeg (Ghiasi et al., 2021)R-101COCO Panoptic4.06.515.336.960.0
OVSeg (Ours)R-101cCOCO-Stuff-1567.010.424.051.789.2
OVSeg (Ours)R-101cCOCO-Stuff-1717.111.024.853.392.6
LSeg+ (Ghiasi et al., 2021)Eff-B7COCO Panoptic3.87.818.046.5
OpenSeg (Ghiasi et al., 2021)Eff-B7COCO Panoptic6.39.021.142.1
OVSeg (Ours)Swin-BCOCO-Stuff-1719.012.429.655.794.5
Supervised specialist models.
FCN (Long et al., 2015)FCN-8sSame as test29.437.81
Deeplab (Chen et al., 2017)R-101Same as test=45.777.7
SelfTrain (Zoph et al., 2020)Eff-L2Same as test=--90.0
MaskFormer (Cheng et al., 2021)R-101cSame as test17.446.0-=
" + }, + { + "category_id": 1, + "poly": [ + 297, + 1407, + 1404, + 1407, + 1404, + 1834, + 297, + 1834 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 297, + 988, + 1405, + 988, + 1405, + 1325, + 297, + 1325 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 297, + 1850, + 1402, + 1850, + 1402, + 2034, + 297, + 2034 + ], + "score": 0.978 + }, + { + "category_id": 6, + "poly": [ + 296, + 193, + 1405, + 193, + 1405, + 348, + 296, + 348 + ], + "score": 0.949 + }, + { + "category_id": 2, + "poly": [ + 298, + 76, + 856, + 76, + 856, + 104, + 298, + 104 + ], + "score": 0.895 + }, + { + "category_id": 0, + "poly": [ + 301, + 1349, + 1153, + 1349, + 1153, + 1378, + 301, + 1378 + ], + "score": 0.89 + }, + { + "category_id": 2, + "poly": [ + 842, + 2088, + 858, + 2088, + 858, + 2111, + 842, + 2111 + ], + "score": 0.646 + }, + { + "category_id": 2, + "poly": [ + 842, + 2088, + 859, + 2088, + 859, + 2111, + 842, + 2111 + ], + "score": 0.277 + }, + { + "category_id": 13, + "poly": [ + 705, + 1110, + 802, + 1110, + 802, + 1140, + 705, + 1140 + ], + "score": 0.91, + "latex": "2 \\cdot 1 0 ^ { - 2 } " + }, + { + "category_id": 13, + "poly": [ + 1181, + 1143, + 1303, + 1143, + 1303, + 1172, + 1181, + 1172 + ], + "score": 0.9, + "latex": "2 2 4 \\times 2 2 4" + }, + { + "category_id": 13, + "poly": [ + 444, + 1201, + 539, + 1201, + 539, + 1232, + 444, + 1232 + ], + "score": 0.9, + "latex": "5 \\cdot 1 0 ^ { - 6 }" + }, + { + "category_id": 13, + "poly": [ + 452, + 1943, + 545, + 1943, + 545, + 1972, + 452, + 1972 + ], + "score": 0.9, + "latex": "+ 1 3 . 6 \\%" + }, + { + "category_id": 13, + "poly": [ + 323, + 1943, + 401, + 1943, + 401, + 1972, + 323, + 1972 + ], + "score": 0.89, + "latex": "+ 8 . 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 1232, + 1531, + 1311, + 1531, + 1311, + 1561, + 1232, + 1561 + ], + "score": 0.88, + "latex": "{ \\mathrm { L S e g } } +" + }, + { + "category_id": 13, + "poly": [ + 734, + 1912, + 810, + 1912, + 810, + 1941, + 734, + 1941 + ], + "score": 0.88, + "latex": "2 9 . 6 \\%" + }, + { + "category_id": 13, + "poly": [ + 1193, + 1774, + 1255, + 1774, + 1255, + 1802, + 1193, + 1802 + ], + "score": 0.87, + "latex": "{ \\sim } 1 \\%" + }, + { + "category_id": 13, + "poly": [ + 860, + 1912, + 935, + 1912, + 935, + 1941, + 860, + 1941 + ], + "score": 0.87, + "latex": "5 5 . 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 904, + 1973, + 966, + 1973, + 966, + 2002, + 904, + 2002 + ], + "score": 0.87, + "latex": "9 . 0 \\%" + }, + { + "category_id": 13, + "poly": [ + 1017, + 1973, + 1090, + 1973, + 1090, + 2002, + 1017, + 2002 + ], + "score": 0.86, + "latex": "12 . 4 \\%" + }, + { + "category_id": 13, + "poly": [ + 1291, + 1652, + 1353, + 1652, + 1353, + 1681, + 1291, + 1681 + ], + "score": 0.86, + "latex": "7 . 1 \\%" + }, + { + "category_id": 13, + "poly": [ + 298, + 1682, + 372, + 1682, + 372, + 1711, + 298, + 1711 + ], + "score": 0.85, + "latex": "1 1 . 0 \\%" + }, + { + "category_id": 13, + "poly": [ + 914, + 1203, + 1035, + 1203, + 1035, + 1232, + 914, + 1232 + ], + "score": 0.7, + "latex": "\\mathbf { M P T } + \\mathbf { F T }" + }, + { + "category_id": 13, + "poly": [ + 307, + 533, + 380, + 533, + 380, + 560, + 307, + 560 + ], + "score": 0.54, + "latex": "{ \\mathrm { L S e g } } +" + }, + { + "category_id": 13, + "poly": [ + 296, + 1022, + 320, + 1022, + 320, + 1049, + 296, + 1049 + ], + "score": 0.37, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 294, + 1021, + 358, + 1021, + 358, + 1050, + 294, + 1050 + ], + "score": 0.32, + "latex": "+ \\ \\mathrm { F T } ." + }, + { + "category_id": 15, + "poly": [ + 294.0, + 191.0, + 1405.0, + 191.0, + 1405.0, + 229.0, + 294.0, + 229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 224.0, + 1402.0, + 224.0, + 1402.0, + 257.0, + 296.0, + 257.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 256.0, + 1403.0, + 256.0, + 1403.0, + 289.0, + 295.0, + 289.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 284.0, + 1406.0, + 284.0, + 1406.0, + 320.0, + 293.0, + 320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 316.0, + 867.0, + 316.0, + 867.0, + 353.0, + 293.0, + 353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1344.0, + 1158.0, + 1344.0, + 1158.0, + 1382.0, + 295.0, + 1382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2086.0, + 860.0, + 2086.0, + 860.0, + 2119.0, + 840.0, + 2119.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2087.0, + 861.0, + 2087.0, + 861.0, + 2119.0, + 840.0, + 2119.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1405.0, + 1405.0, + 1405.0, + 1405.0, + 1443.0, + 294.0, + 1443.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1438.0, + 1405.0, + 1438.0, + 1405.0, + 1472.0, + 294.0, + 1472.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1465.0, + 1406.0, + 1465.0, + 1406.0, + 1507.0, + 291.0, + 1507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1500.0, + 1404.0, + 1500.0, + 1404.0, + 1536.0, + 294.0, + 1536.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1529.0, + 1231.0, + 1529.0, + 1231.0, + 1565.0, + 294.0, + 1565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1312.0, + 1529.0, + 1405.0, + 1529.0, + 1405.0, + 1565.0, + 1312.0, + 1565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1561.0, + 1405.0, + 1561.0, + 1405.0, + 1596.0, + 294.0, + 1596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1590.0, + 1406.0, + 1590.0, + 1406.0, + 1626.0, + 295.0, + 1626.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1619.0, + 1405.0, + 1619.0, + 1405.0, + 1656.0, + 291.0, + 1656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1651.0, + 1290.0, + 1651.0, + 1290.0, + 1685.0, + 294.0, + 1685.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1354.0, + 1651.0, + 1406.0, + 1651.0, + 1406.0, + 1685.0, + 1354.0, + 1685.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1677.0, + 297.0, + 1677.0, + 297.0, + 1718.0, + 292.0, + 1718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 1677.0, + 1405.0, + 1677.0, + 1405.0, + 1718.0, + 373.0, + 1718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1709.0, + 1406.0, + 1709.0, + 1406.0, + 1750.0, + 291.0, + 1750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1742.0, + 1404.0, + 1742.0, + 1404.0, + 1778.0, + 295.0, + 1778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1773.0, + 1192.0, + 1773.0, + 1192.0, + 1809.0, + 294.0, + 1809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1256.0, + 1773.0, + 1405.0, + 1773.0, + 1405.0, + 1809.0, + 1256.0, + 1809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1804.0, + 1075.0, + 1804.0, + 1075.0, + 1837.0, + 293.0, + 1837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 988.0, + 1406.0, + 988.0, + 1406.0, + 1026.0, + 294.0, + 1026.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 359.0, + 1020.0, + 1403.0, + 1020.0, + 1403.0, + 1055.0, + 359.0, + 1055.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1048.0, + 1405.0, + 1048.0, + 1405.0, + 1086.0, + 294.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1079.0, + 1406.0, + 1079.0, + 1406.0, + 1117.0, + 292.0, + 1117.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1109.0, + 704.0, + 1109.0, + 704.0, + 1144.0, + 294.0, + 1144.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 803.0, + 1109.0, + 1405.0, + 1109.0, + 1405.0, + 1144.0, + 803.0, + 1144.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1144.0, + 1180.0, + 1144.0, + 1180.0, + 1175.0, + 295.0, + 1175.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1304.0, + 1144.0, + 1405.0, + 1144.0, + 1405.0, + 1175.0, + 1304.0, + 1175.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1173.0, + 1407.0, + 1173.0, + 1407.0, + 1207.0, + 294.0, + 1207.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1200.0, + 443.0, + 1200.0, + 443.0, + 1238.0, + 292.0, + 1238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 540.0, + 1200.0, + 913.0, + 1200.0, + 913.0, + 1238.0, + 540.0, + 1238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1036.0, + 1200.0, + 1406.0, + 1200.0, + 1406.0, + 1238.0, + 1036.0, + 1238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1233.0, + 1405.0, + 1233.0, + 1405.0, + 1268.0, + 295.0, + 1268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1262.0, + 1406.0, + 1262.0, + 1406.0, + 1299.0, + 294.0, + 1299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1294.0, + 846.0, + 1294.0, + 846.0, + 1328.0, + 295.0, + 1328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1847.0, + 1407.0, + 1847.0, + 1407.0, + 1887.0, + 291.0, + 1887.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1882.0, + 1404.0, + 1882.0, + 1404.0, + 1914.0, + 295.0, + 1914.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1912.0, + 733.0, + 1912.0, + 733.0, + 1943.0, + 296.0, + 1943.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 811.0, + 1912.0, + 859.0, + 1912.0, + 859.0, + 1943.0, + 811.0, + 1943.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 936.0, + 1912.0, + 1404.0, + 1912.0, + 1404.0, + 1943.0, + 936.0, + 1943.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1941.0, + 322.0, + 1941.0, + 322.0, + 1976.0, + 293.0, + 1976.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 402.0, + 1941.0, + 451.0, + 1941.0, + 451.0, + 1976.0, + 402.0, + 1976.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 546.0, + 1941.0, + 1406.0, + 1941.0, + 1406.0, + 1976.0, + 546.0, + 1976.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1970.0, + 903.0, + 1970.0, + 903.0, + 2005.0, + 293.0, + 2005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 967.0, + 1970.0, + 1016.0, + 1970.0, + 1016.0, + 2005.0, + 967.0, + 2005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1091.0, + 1970.0, + 1404.0, + 1970.0, + 1404.0, + 2005.0, + 1091.0, + 2005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 2002.0, + 893.0, + 2002.0, + 893.0, + 2037.0, + 295.0, + 2037.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 6, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1395, + 1404, + 1395, + 1404, + 1883, + 298, + 1883 + ], + "score": 0.985 + }, + { + "category_id": 5, + "poly": [ + 308, + 330, + 1387, + 330, + 1387, + 581, + 308, + 581 + ], + "score": 0.981, + "html": "
CaseSourceStatisticsA-847A-150PC-59
MaskCategoryPairsUnique nouns
Baseline11-17.321.851.4
(1)GTGT965K1715.3 (-2.0)23.0 (+1.2)57.3 (+5.9)
(2)GT1 caption440K12K7.9 (+0.6)24.2 (+2.4)53.2 (+1.8)
(3)proposals1 caption440K12K8.8 (+1.5)28.8 (+7.0)55.7 (+4.3)
(4)proposals5 captions1.3M27K8.8 (+1.5)28.6 (+6.8)55.5 (+4.1)
" + }, + { + "category_id": 5, + "poly": [ + 421, + 717, + 1272, + 717, + 1272, + 939, + 421, + 939 + ], + "score": 0.98, + "html": "
caseFinetuning methodA-847A-150PC-59
mask promptfull model
Baseline7.321.851.4
(a)√8.4 (+1.1)26.5 (+4.7)55.4 (+4.0)
(b)√8.8 (+1.5)28.8 (+7.0)55.7 (+4.3)
(c)√√9.0 (+1.7)29.6 (+7.8)55.7 (+4.3)
" + }, + { + "category_id": 6, + "poly": [ + 296, + 193, + 1403, + 193, + 1403, + 317, + 296, + 317 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 297, + 970, + 1404, + 970, + 1404, + 1247, + 297, + 1247 + ], + "score": 0.944 + }, + { + "category_id": 1, + "poly": [ + 300, + 1972, + 1399, + 1972, + 1399, + 2034, + 300, + 2034 + ], + "score": 0.943 + }, + { + "category_id": 2, + "poly": [ + 298, + 75, + 856, + 75, + 856, + 105, + 298, + 105 + ], + "score": 0.914 + }, + { + "category_id": 6, + "poly": [ + 298, + 610, + 1401, + 610, + 1401, + 704, + 298, + 704 + ], + "score": 0.906 + }, + { + "category_id": 0, + "poly": [ + 300, + 1918, + 668, + 1918, + 668, + 1949, + 300, + 1949 + ], + "score": 0.895 + }, + { + "category_id": 0, + "poly": [ + 300, + 1286, + 578, + 1286, + 578, + 1315, + 300, + 1315 + ], + "score": 0.866 + }, + { + "category_id": 0, + "poly": [ + 300, + 1343, + 834, + 1343, + 834, + 1373, + 300, + 1373 + ], + "score": 0.841 + }, + { + "category_id": 2, + "poly": [ + 841, + 2089, + 858, + 2089, + 858, + 2112, + 841, + 2112 + ], + "score": 0.788 + }, + { + "category_id": 13, + "poly": [ + 1324, + 1063, + 1401, + 1063, + 1401, + 1092, + 1324, + 1092 + ], + "score": 0.89, + "latex": "+ 4 . 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 1011, + 1063, + 1086, + 1063, + 1086, + 1092, + 1011, + 1092 + ], + "score": 0.87, + "latex": "9 4 . 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 981, + 1701, + 1057, + 1701, + 1057, + 1730, + 981, + 1730 + ], + "score": 0.86, + "latex": "2 4 . 2 \\%" + }, + { + "category_id": 13, + "poly": [ + 521, + 1518, + 588, + 1518, + 588, + 1547, + 521, + 1547 + ], + "score": 0.84, + "latex": "- 2 . 0 \\%" + }, + { + "category_id": 13, + "poly": [ + 306, + 1822, + 368, + 1822, + 368, + 1852, + 306, + 1852 + ], + "score": 0.83, + "latex": ". 0 . 2 \\%" + }, + { + "category_id": 13, + "poly": [ + 1094, + 1701, + 1172, + 1701, + 1172, + 1731, + 1094, + 1731 + ], + "score": 0.83, + "latex": "2 8 . 8 \\%" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 193.0, + 1405.0, + 193.0, + 1405.0, + 229.0, + 295.0, + 229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 221.0, + 1405.0, + 221.0, + 1405.0, + 261.0, + 294.0, + 261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 257.0, + 1406.0, + 257.0, + 1406.0, + 288.0, + 292.0, + 288.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 288.0, + 773.0, + 288.0, + 773.0, + 320.0, + 295.0, + 320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 608.0, + 1405.0, + 608.0, + 1405.0, + 646.0, + 293.0, + 646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 640.0, + 1405.0, + 640.0, + 1405.0, + 674.0, + 294.0, + 674.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 670.0, + 940.0, + 670.0, + 940.0, + 708.0, + 293.0, + 708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1918.0, + 670.0, + 1918.0, + 670.0, + 1951.0, + 297.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1282.0, + 582.0, + 1282.0, + 582.0, + 1321.0, + 294.0, + 1321.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1341.0, + 840.0, + 1341.0, + 840.0, + 1377.0, + 295.0, + 1377.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2086.0, + 860.0, + 2086.0, + 860.0, + 2116.0, + 839.0, + 2116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1395.0, + 1404.0, + 1395.0, + 1404.0, + 1430.0, + 296.0, + 1430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1425.0, + 1404.0, + 1425.0, + 1404.0, + 1461.0, + 293.0, + 1461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1455.0, + 1406.0, + 1455.0, + 1406.0, + 1492.0, + 293.0, + 1492.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1487.0, + 1405.0, + 1487.0, + 1405.0, + 1521.0, + 294.0, + 1521.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1517.0, + 520.0, + 1517.0, + 520.0, + 1552.0, + 292.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 589.0, + 1517.0, + 1405.0, + 1517.0, + 1405.0, + 1552.0, + 589.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1547.0, + 1405.0, + 1547.0, + 1405.0, + 1581.0, + 293.0, + 1581.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1580.0, + 1404.0, + 1580.0, + 1404.0, + 1613.0, + 293.0, + 1613.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1610.0, + 1405.0, + 1610.0, + 1405.0, + 1644.0, + 294.0, + 1644.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1640.0, + 1404.0, + 1640.0, + 1404.0, + 1675.0, + 294.0, + 1675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1671.0, + 1404.0, + 1671.0, + 1404.0, + 1703.0, + 293.0, + 1703.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1700.0, + 980.0, + 1700.0, + 980.0, + 1736.0, + 292.0, + 1736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1058.0, + 1700.0, + 1093.0, + 1700.0, + 1093.0, + 1736.0, + 1058.0, + 1736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1173.0, + 1700.0, + 1407.0, + 1700.0, + 1407.0, + 1736.0, + 1173.0, + 1736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1730.0, + 1406.0, + 1730.0, + 1406.0, + 1766.0, + 291.0, + 1766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1763.0, + 1405.0, + 1763.0, + 1405.0, + 1797.0, + 293.0, + 1797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1791.0, + 1404.0, + 1791.0, + 1404.0, + 1825.0, + 294.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1822.0, + 305.0, + 1822.0, + 305.0, + 1859.0, + 294.0, + 1859.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1822.0, + 1408.0, + 1822.0, + 1408.0, + 1859.0, + 369.0, + 1859.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1851.0, + 1300.0, + 1851.0, + 1300.0, + 1891.0, + 291.0, + 1891.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 971.0, + 1405.0, + 971.0, + 1405.0, + 1008.0, + 296.0, + 1008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1002.0, + 1406.0, + 1002.0, + 1406.0, + 1038.0, + 294.0, + 1038.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1034.0, + 1405.0, + 1034.0, + 1405.0, + 1067.0, + 296.0, + 1067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1061.0, + 1010.0, + 1061.0, + 1010.0, + 1098.0, + 294.0, + 1098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1087.0, + 1061.0, + 1323.0, + 1061.0, + 1323.0, + 1098.0, + 1087.0, + 1098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 1061.0, + 1405.0, + 1061.0, + 1405.0, + 1098.0, + 1402.0, + 1098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1093.0, + 1405.0, + 1093.0, + 1405.0, + 1129.0, + 294.0, + 1129.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1122.0, + 1407.0, + 1122.0, + 1407.0, + 1161.0, + 292.0, + 1161.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1156.0, + 1404.0, + 1156.0, + 1404.0, + 1188.0, + 296.0, + 1188.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1184.0, + 1406.0, + 1184.0, + 1406.0, + 1221.0, + 294.0, + 1221.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1216.0, + 894.0, + 1216.0, + 894.0, + 1250.0, + 293.0, + 1250.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1971.0, + 1403.0, + 1971.0, + 1403.0, + 2007.0, + 296.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2000.0, + 1403.0, + 2000.0, + 1403.0, + 2038.0, + 294.0, + 2038.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 7, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1757, + 1404, + 1757, + 1404, + 2032, + 298, + 2032 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 299, + 1155, + 823, + 1155, + 823, + 1613, + 299, + 1613 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 944, + 1403, + 944, + 1403, + 1067, + 298, + 1067 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 298, + 611, + 1404, + 611, + 1404, + 796, + 298, + 796 + ], + "score": 0.974 + }, + { + "category_id": 3, + "poly": [ + 298, + 200, + 1392, + 200, + 1392, + 490, + 298, + 490 + ], + "score": 0.971 + }, + { + "category_id": 3, + "poly": [ + 880, + 1152, + 1398, + 1152, + 1398, + 1518, + 880, + 1518 + ], + "score": 0.964 + }, + { + "category_id": 4, + "poly": [ + 297, + 520, + 1399, + 520, + 1399, + 583, + 297, + 583 + ], + "score": 0.935 + }, + { + "category_id": 4, + "poly": [ + 887, + 1538, + 1367, + 1538, + 1367, + 1570, + 887, + 1570 + ], + "score": 0.91 + }, + { + "category_id": 0, + "poly": [ + 299, + 1104, + 1165, + 1104, + 1165, + 1131, + 299, + 1131 + ], + "score": 0.905 + }, + { + "category_id": 1, + "poly": [ + 293, + 1612, + 1361, + 1612, + 1361, + 1644, + 293, + 1644 + ], + "score": 0.896 + }, + { + "category_id": 0, + "poly": [ + 299, + 833, + 529, + 833, + 529, + 865, + 299, + 865 + ], + "score": 0.896 + }, + { + "category_id": 0, + "poly": [ + 300, + 1687, + 543, + 1687, + 543, + 1722, + 300, + 1722 + ], + "score": 0.884 + }, + { + "category_id": 0, + "poly": [ + 300, + 892, + 939, + 892, + 939, + 922, + 300, + 922 + ], + "score": 0.867 + }, + { + "category_id": 2, + "poly": [ + 298, + 76, + 856, + 76, + 856, + 104, + 298, + 104 + ], + "score": 0.856 + }, + { + "category_id": 2, + "poly": [ + 840, + 2088, + 858, + 2088, + 858, + 2111, + 840, + 2111 + ], + "score": 0.789 + }, + { + "category_id": 2, + "poly": [ + 299, + 76, + 854, + 76, + 854, + 104, + 299, + 104 + ], + "score": 0.134 + }, + { + "category_id": 13, + "poly": [ + 549, + 613, + 627, + 613, + 627, + 642, + 549, + 642 + ], + "score": 0.89, + "latex": "+ 4 . 7 \\%" + }, + { + "category_id": 13, + "poly": [ + 675, + 613, + 753, + 613, + 753, + 642, + 675, + 642 + ], + "score": 0.89, + "latex": "+ 4 . 0 \\%" + }, + { + "category_id": 13, + "poly": [ + 1286, + 766, + 1362, + 766, + 1362, + 794, + 1286, + 794 + ], + "score": 0.88, + "latex": "+ 0 . 8 \\%" + }, + { + "category_id": 13, + "poly": [ + 1253, + 735, + 1329, + 735, + 1329, + 763, + 1253, + 763 + ], + "score": 0.84, + "latex": "2 9 . 6 \\%" + }, + { + "category_id": 15, + "poly": [ + 823.0, + 243.0, + 874.0, + 243.0, + 874.0, + 263.0, + 823.0, + 263.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 283.0, + 367.0, + 283.0, + 367.0, + 302.0, + 313.0, + 302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 665.0, + 277.0, + 681.0, + 277.0, + 681.0, + 328.0, + 665.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 880.0, + 335.0, + 927.0, + 335.0, + 927.0, + 356.0, + 880.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1231.0, + 330.0, + 1304.0, + 330.0, + 1304.0, + 350.0, + 1231.0, + 350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 528.0, + 371.0, + 588.0, + 371.0, + 588.0, + 400.0, + 528.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 339.0, + 458.0, + 603.0, + 458.0, + 603.0, + 491.0, + 339.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 743.0, + 458.0, + 985.0, + 458.0, + 985.0, + 491.0, + 743.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1096.0, + 458.0, + 1364.0, + 458.0, + 1364.0, + 494.0, + 1096.0, + 494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1105.25, + 378.5, + 1149.25, + 378.5, + 1149.25, + 400.5, + 1105.25, + 400.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 940.0, + 1195.0, + 967.0, + 1195.0, + 967.0, + 1214.0, + 940.0, + 1214.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1244.0, + 1225.0, + 1272.0, + 1225.0, + 1272.0, + 1244.0, + 1244.0, + 1244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 998.0, + 1331.0, + 1066.0, + 1331.0, + 1066.0, + 1351.0, + 998.0, + 1351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1292.0, + 1328.0, + 1351.0, + 1328.0, + 1351.0, + 1346.0, + 1292.0, + 1346.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1004.0, + 1385.0, + 1060.0, + 1385.0, + 1060.0, + 1408.0, + 1004.0, + 1408.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1178.0, + 1383.0, + 1198.0, + 1383.0, + 1198.0, + 1392.0, + 1178.0, + 1392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1145.0, + 1401.0, + 1180.0, + 1401.0, + 1180.0, + 1418.0, + 1145.0, + 1418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1168.0, + 1432.0, + 1240.0, + 1432.0, + 1240.0, + 1451.0, + 1168.0, + 1451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1274.0, + 1442.0, + 1291.0, + 1442.0, + 1291.0, + 1453.0, + 1274.0, + 1453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1264.0, + 1457.0, + 1290.0, + 1457.0, + 1290.0, + 1475.0, + 1264.0, + 1475.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 952.0, + 1472.0, + 977.0, + 1472.0, + 977.0, + 1486.0, + 952.0, + 1486.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 881.0, + 1492.0, + 1139.0, + 1492.0, + 1139.0, + 1519.0, + 881.0, + 1519.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1174.0, + 1491.0, + 1368.0, + 1491.0, + 1368.0, + 1519.0, + 1174.0, + 1519.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 520.0, + 1403.0, + 520.0, + 1403.0, + 555.0, + 296.0, + 555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 552.0, + 1381.0, + 552.0, + 1381.0, + 587.0, + 296.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 881.0, + 1535.0, + 1368.0, + 1535.0, + 1368.0, + 1575.0, + 881.0, + 1575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1102.0, + 1167.0, + 1102.0, + 1167.0, + 1135.0, + 295.0, + 1135.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 831.0, + 533.0, + 831.0, + 533.0, + 869.0, + 294.0, + 869.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1683.0, + 549.0, + 1683.0, + 549.0, + 1731.0, + 292.0, + 1731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 890.0, + 945.0, + 890.0, + 945.0, + 927.0, + 294.0, + 927.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2087.0, + 860.0, + 2087.0, + 860.0, + 2117.0, + 839.0, + 2117.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 297.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1755.0, + 1404.0, + 1755.0, + 1404.0, + 1794.0, + 295.0, + 1794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1784.0, + 1405.0, + 1784.0, + 1405.0, + 1827.0, + 292.0, + 1827.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1818.0, + 1406.0, + 1818.0, + 1406.0, + 1854.0, + 294.0, + 1854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1846.0, + 1405.0, + 1846.0, + 1405.0, + 1885.0, + 293.0, + 1885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1876.0, + 1408.0, + 1876.0, + 1408.0, + 1918.0, + 292.0, + 1918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1908.0, + 1406.0, + 1908.0, + 1406.0, + 1948.0, + 293.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1940.0, + 1404.0, + 1940.0, + 1404.0, + 1977.0, + 293.0, + 1977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1969.0, + 1405.0, + 1969.0, + 1405.0, + 2008.0, + 293.0, + 2008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2001.0, + 655.0, + 2001.0, + 655.0, + 2038.0, + 294.0, + 2038.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1151.0, + 826.0, + 1151.0, + 826.0, + 1188.0, + 296.0, + 1188.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1182.0, + 825.0, + 1182.0, + 825.0, + 1221.0, + 296.0, + 1221.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1215.0, + 823.0, + 1215.0, + 823.0, + 1251.0, + 295.0, + 1251.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1245.0, + 827.0, + 1245.0, + 827.0, + 1280.0, + 296.0, + 1280.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1273.0, + 828.0, + 1273.0, + 828.0, + 1311.0, + 295.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1306.0, + 827.0, + 1306.0, + 827.0, + 1341.0, + 295.0, + 1341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1337.0, + 827.0, + 1337.0, + 827.0, + 1371.0, + 295.0, + 1371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1368.0, + 826.0, + 1368.0, + 826.0, + 1399.0, + 294.0, + 1399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1393.0, + 828.0, + 1393.0, + 828.0, + 1434.0, + 293.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1428.0, + 827.0, + 1428.0, + 827.0, + 1463.0, + 296.0, + 1463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1459.0, + 827.0, + 1459.0, + 827.0, + 1494.0, + 296.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1490.0, + 827.0, + 1490.0, + 827.0, + 1525.0, + 295.0, + 1525.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1521.0, + 825.0, + 1521.0, + 825.0, + 1553.0, + 296.0, + 1553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1551.0, + 825.0, + 1551.0, + 825.0, + 1583.0, + 296.0, + 1583.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1581.0, + 825.0, + 1581.0, + 825.0, + 1615.0, + 295.0, + 1615.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 942.0, + 1405.0, + 942.0, + 1405.0, + 980.0, + 293.0, + 980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 976.0, + 1404.0, + 976.0, + 1404.0, + 1009.0, + 296.0, + 1009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1003.0, + 1404.0, + 1003.0, + 1404.0, + 1043.0, + 292.0, + 1043.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1037.0, + 1316.0, + 1037.0, + 1316.0, + 1069.0, + 294.0, + 1069.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 612.0, + 548.0, + 612.0, + 548.0, + 648.0, + 294.0, + 648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 628.0, + 612.0, + 674.0, + 612.0, + 674.0, + 648.0, + 628.0, + 648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 754.0, + 612.0, + 1402.0, + 612.0, + 1402.0, + 648.0, + 754.0, + 648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 642.0, + 1404.0, + 642.0, + 1404.0, + 679.0, + 293.0, + 679.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 669.0, + 1405.0, + 669.0, + 1405.0, + 712.0, + 293.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 701.0, + 1405.0, + 701.0, + 1405.0, + 742.0, + 293.0, + 742.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 734.0, + 1252.0, + 734.0, + 1252.0, + 770.0, + 293.0, + 770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1330.0, + 734.0, + 1405.0, + 734.0, + 1405.0, + 770.0, + 1330.0, + 770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 766.0, + 1285.0, + 766.0, + 1285.0, + 798.0, + 296.0, + 798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1363.0, + 766.0, + 1372.0, + 766.0, + 1372.0, + 798.0, + 1363.0, + 798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1605.0, + 1366.0, + 1605.0, + 1366.0, + 1651.0, + 293.0, + 1651.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 8, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 295, + 1403, + 295, + 1403, + 480, + 298, + 480 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 298, + 594, + 1401, + 594, + 1401, + 656, + 298, + 656 + ], + "score": 0.937 + }, + { + "category_id": 2, + "poly": [ + 298, + 75, + 856, + 75, + 856, + 105, + 298, + 105 + ], + "score": 0.907 + }, + { + "category_id": 0, + "poly": [ + 299, + 703, + 488, + 703, + 488, + 736, + 299, + 736 + ], + "score": 0.867 + }, + { + "category_id": 0, + "poly": [ + 301, + 525, + 744, + 525, + 744, + 560, + 301, + 560 + ], + "score": 0.859 + }, + { + "category_id": 2, + "poly": [ + 837, + 2088, + 865, + 2088, + 865, + 2112, + 837, + 2112 + ], + "score": 0.823 + }, + { + "category_id": 0, + "poly": [ + 300, + 226, + 586, + 226, + 586, + 262, + 300, + 262 + ], + "score": 0.818 + }, + { + "category_id": 1, + "poly": [ + 298, + 1343, + 1400, + 1343, + 1400, + 1437, + 298, + 1437 + ], + "score": 0.543 + }, + { + "category_id": 1, + "poly": [ + 295, + 1198, + 1403, + 1198, + 1403, + 1323, + 295, + 1323 + ], + "score": 0.53 + }, + { + "category_id": 1, + "poly": [ + 299, + 1084, + 1401, + 1084, + 1401, + 1179, + 299, + 1179 + ], + "score": 0.494 + }, + { + "category_id": 1, + "poly": [ + 300, + 1456, + 1399, + 1456, + 1399, + 1551, + 300, + 1551 + ], + "score": 0.478 + }, + { + "category_id": 1, + "poly": [ + 297, + 754, + 1402, + 754, + 1402, + 818, + 297, + 818 + ], + "score": 0.454 + }, + { + "category_id": 1, + "poly": [ + 293, + 1001, + 1401, + 1001, + 1401, + 1066, + 293, + 1066 + ], + "score": 0.444 + }, + { + "category_id": 1, + "poly": [ + 297, + 1569, + 1402, + 1569, + 1402, + 1662, + 297, + 1662 + ], + "score": 0.422 + }, + { + "category_id": 1, + "poly": [ + 299, + 836, + 1399, + 836, + 1399, + 899, + 299, + 899 + ], + "score": 0.409 + }, + { + "category_id": 1, + "poly": [ + 296, + 919, + 1401, + 919, + 1401, + 983, + 296, + 983 + ], + "score": 0.404 + }, + { + "category_id": 1, + "poly": [ + 297, + 1796, + 1399, + 1796, + 1399, + 1890, + 297, + 1890 + ], + "score": 0.379 + }, + { + "category_id": 1, + "poly": [ + 299, + 1682, + 1400, + 1682, + 1400, + 1779, + 299, + 1779 + ], + "score": 0.343 + }, + { + "category_id": 1, + "poly": [ + 300, + 1910, + 1403, + 1910, + 1403, + 2033, + 300, + 2033 + ], + "score": 0.26 + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 703.0, + 490.0, + 703.0, + 490.0, + 740.0, + 297.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 524.0, + 750.0, + 524.0, + 750.0, + 564.0, + 296.0, + 564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 831.0, + 2084.0, + 871.0, + 2084.0, + 871.0, + 2125.0, + 831.0, + 2125.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 225.0, + 592.0, + 225.0, + 592.0, + 265.0, + 296.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 295.0, + 1405.0, + 295.0, + 1405.0, + 332.0, + 294.0, + 332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 326.0, + 1405.0, + 326.0, + 1405.0, + 361.0, + 293.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 353.0, + 1407.0, + 353.0, + 1407.0, + 395.0, + 291.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 386.0, + 1407.0, + 386.0, + 1407.0, + 426.0, + 293.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 418.0, + 1405.0, + 418.0, + 1405.0, + 454.0, + 294.0, + 454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 450.0, + 957.0, + 450.0, + 957.0, + 482.0, + 296.0, + 482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 593.0, + 1405.0, + 593.0, + 1405.0, + 629.0, + 296.0, + 629.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 624.0, + 720.0, + 624.0, + 720.0, + 658.0, + 294.0, + 658.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1343.0, + 1404.0, + 1343.0, + 1404.0, + 1380.0, + 294.0, + 1380.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1372.0, + 1406.0, + 1372.0, + 1406.0, + 1412.0, + 322.0, + 1412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1403.0, + 1153.0, + 1403.0, + 1153.0, + 1441.0, + 323.0, + 1441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1201.0, + 1404.0, + 1201.0, + 1404.0, + 1234.0, + 295.0, + 1234.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 1228.0, + 1406.0, + 1228.0, + 1406.0, + 1266.0, + 321.0, + 1266.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 319.0, + 1259.0, + 1405.0, + 1259.0, + 1405.0, + 1298.0, + 319.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1292.0, + 505.0, + 1292.0, + 505.0, + 1323.0, + 323.0, + 1323.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1084.0, + 1403.0, + 1084.0, + 1403.0, + 1122.0, + 294.0, + 1122.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1116.0, + 1405.0, + 1116.0, + 1405.0, + 1152.0, + 320.0, + 1152.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1145.0, + 464.0, + 1145.0, + 464.0, + 1180.0, + 323.0, + 1180.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1457.0, + 1404.0, + 1457.0, + 1404.0, + 1491.0, + 295.0, + 1491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1488.0, + 1404.0, + 1488.0, + 1404.0, + 1522.0, + 324.0, + 1522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1519.0, + 707.0, + 1519.0, + 707.0, + 1551.0, + 320.0, + 1551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 747.0, + 1406.0, + 747.0, + 1406.0, + 794.0, + 292.0, + 794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 782.0, + 1107.0, + 782.0, + 1107.0, + 818.0, + 324.0, + 818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1000.0, + 1403.0, + 1000.0, + 1403.0, + 1039.0, + 293.0, + 1039.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1032.0, + 1091.0, + 1032.0, + 1091.0, + 1067.0, + 322.0, + 1067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1571.0, + 1404.0, + 1571.0, + 1404.0, + 1605.0, + 295.0, + 1605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1601.0, + 1406.0, + 1601.0, + 1406.0, + 1635.0, + 320.0, + 1635.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 1630.0, + 396.0, + 1630.0, + 396.0, + 1664.0, + 321.0, + 1664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 834.0, + 1403.0, + 834.0, + 1403.0, + 874.0, + 294.0, + 874.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 868.0, + 687.0, + 868.0, + 687.0, + 900.0, + 321.0, + 900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 916.0, + 1406.0, + 916.0, + 1406.0, + 959.0, + 294.0, + 959.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 950.0, + 1093.0, + 950.0, + 1093.0, + 984.0, + 321.0, + 984.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1796.0, + 1403.0, + 1796.0, + 1403.0, + 1835.0, + 293.0, + 1835.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 1824.0, + 1404.0, + 1824.0, + 1404.0, + 1867.0, + 321.0, + 1867.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1858.0, + 560.0, + 1858.0, + 560.0, + 1892.0, + 323.0, + 1892.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1685.0, + 1401.0, + 1685.0, + 1401.0, + 1716.0, + 297.0, + 1716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1716.0, + 1404.0, + 1716.0, + 1404.0, + 1750.0, + 324.0, + 1750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1744.0, + 1001.0, + 1744.0, + 1001.0, + 1782.0, + 323.0, + 1782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1912.0, + 1404.0, + 1912.0, + 1404.0, + 1945.0, + 296.0, + 1945.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1941.0, + 1404.0, + 1941.0, + 1404.0, + 1977.0, + 323.0, + 1977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1971.0, + 1405.0, + 1971.0, + 1405.0, + 2010.0, + 320.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 2002.0, + 610.0, + 2002.0, + 610.0, + 2035.0, + 323.0, + 2035.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 9, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2088, + 863, + 2088, + 863, + 2113, + 836, + 2113 + ], + "score": 0.831 + }, + { + "category_id": 1, + "poly": [ + 298, + 465, + 1400, + 465, + 1400, + 529, + 298, + 529 + ], + "score": 0.783 + }, + { + "category_id": 1, + "poly": [ + 297, + 553, + 1400, + 553, + 1400, + 618, + 297, + 618 + ], + "score": 0.769 + }, + { + "category_id": 1, + "poly": [ + 297, + 641, + 1406, + 641, + 1406, + 766, + 297, + 766 + ], + "score": 0.756 + }, + { + "category_id": 1, + "poly": [ + 297, + 1527, + 1401, + 1527, + 1401, + 1623, + 297, + 1623 + ], + "score": 0.752 + }, + { + "category_id": 1, + "poly": [ + 300, + 1409, + 1404, + 1409, + 1404, + 1505, + 300, + 1505 + ], + "score": 0.726 + }, + { + "category_id": 1, + "poly": [ + 303, + 228, + 1401, + 228, + 1401, + 324, + 303, + 324 + ], + "score": 0.726 + }, + { + "category_id": 1, + "poly": [ + 297, + 1202, + 1400, + 1202, + 1400, + 1267, + 297, + 1267 + ], + "score": 0.724 + }, + { + "category_id": 1, + "poly": [ + 297, + 346, + 1405, + 346, + 1405, + 441, + 297, + 441 + ], + "score": 0.717 + }, + { + "category_id": 1, + "poly": [ + 291, + 1114, + 1402, + 1114, + 1402, + 1180, + 291, + 1180 + ], + "score": 0.711 + }, + { + "category_id": 1, + "poly": [ + 294, + 1026, + 1402, + 1026, + 1402, + 1092, + 294, + 1092 + ], + "score": 0.702 + }, + { + "category_id": 1, + "poly": [ + 293, + 1765, + 1400, + 1765, + 1400, + 1859, + 293, + 1859 + ], + "score": 0.671 + }, + { + "category_id": 1, + "poly": [ + 295, + 938, + 1400, + 938, + 1400, + 1005, + 295, + 1005 + ], + "score": 0.667 + }, + { + "category_id": 1, + "poly": [ + 298, + 1290, + 1404, + 1290, + 1404, + 1387, + 298, + 1387 + ], + "score": 0.659 + }, + { + "category_id": 1, + "poly": [ + 296, + 1645, + 1405, + 1645, + 1405, + 1741, + 296, + 1741 + ], + "score": 0.658 + }, + { + "category_id": 1, + "poly": [ + 297, + 1971, + 1399, + 1971, + 1399, + 2034, + 297, + 2034 + ], + "score": 0.651 + }, + { + "category_id": 1, + "poly": [ + 297, + 790, + 1404, + 790, + 1404, + 914, + 297, + 914 + ], + "score": 0.622 + }, + { + "category_id": 2, + "poly": [ + 298, + 75, + 857, + 75, + 857, + 105, + 298, + 105 + ], + "score": 0.621 + }, + { + "category_id": 1, + "poly": [ + 289, + 1883, + 1405, + 1883, + 1405, + 1946, + 289, + 1946 + ], + "score": 0.617 + }, + { + "category_id": 2, + "poly": [ + 298, + 75, + 857, + 75, + 857, + 105, + 298, + 105 + ], + "score": 0.291 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 868.0, + 2085.0, + 868.0, + 2125.0, + 832.0, + 2125.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 857.0, + 73.0, + 857.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 857.0, + 73.0, + 857.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 462.0, + 1405.0, + 462.0, + 1405.0, + 505.0, + 293.0, + 505.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 319.0, + 498.0, + 707.0, + 498.0, + 707.0, + 529.0, + 319.0, + 529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 555.0, + 1405.0, + 555.0, + 1405.0, + 591.0, + 296.0, + 591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 586.0, + 1193.0, + 586.0, + 1193.0, + 619.0, + 322.0, + 619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 643.0, + 1404.0, + 643.0, + 1404.0, + 675.0, + 296.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 674.0, + 1404.0, + 674.0, + 1404.0, + 707.0, + 324.0, + 707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 706.0, + 1403.0, + 706.0, + 1403.0, + 739.0, + 323.0, + 739.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 736.0, + 1102.0, + 736.0, + 1102.0, + 768.0, + 323.0, + 768.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1526.0, + 1402.0, + 1526.0, + 1402.0, + 1565.0, + 293.0, + 1565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 1558.0, + 1405.0, + 1558.0, + 1405.0, + 1597.0, + 321.0, + 1597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1592.0, + 888.0, + 1592.0, + 888.0, + 1624.0, + 322.0, + 1624.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1408.0, + 1406.0, + 1408.0, + 1406.0, + 1446.0, + 295.0, + 1446.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1440.0, + 1404.0, + 1440.0, + 1404.0, + 1477.0, + 323.0, + 1477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1470.0, + 1009.0, + 1470.0, + 1009.0, + 1507.0, + 323.0, + 1507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 228.0, + 1405.0, + 228.0, + 1405.0, + 266.0, + 297.0, + 266.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 261.0, + 1406.0, + 261.0, + 1406.0, + 296.0, + 322.0, + 296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 288.0, + 1246.0, + 288.0, + 1246.0, + 330.0, + 322.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1204.0, + 1404.0, + 1204.0, + 1404.0, + 1240.0, + 295.0, + 1240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 1236.0, + 708.0, + 1236.0, + 708.0, + 1269.0, + 321.0, + 1269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 345.0, + 1402.0, + 345.0, + 1402.0, + 384.0, + 294.0, + 384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 376.0, + 1403.0, + 376.0, + 1403.0, + 417.0, + 321.0, + 417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 409.0, + 504.0, + 409.0, + 504.0, + 441.0, + 324.0, + 441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1113.0, + 1404.0, + 1113.0, + 1404.0, + 1152.0, + 294.0, + 1152.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1147.0, + 1043.0, + 1147.0, + 1043.0, + 1181.0, + 322.0, + 1181.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1024.0, + 1404.0, + 1024.0, + 1404.0, + 1064.0, + 294.0, + 1064.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1059.0, + 863.0, + 1059.0, + 863.0, + 1093.0, + 322.0, + 1093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1766.0, + 1403.0, + 1766.0, + 1403.0, + 1800.0, + 295.0, + 1800.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1796.0, + 1405.0, + 1796.0, + 1405.0, + 1833.0, + 322.0, + 1833.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 1829.0, + 574.0, + 1829.0, + 574.0, + 1858.0, + 321.0, + 1858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 939.0, + 1405.0, + 939.0, + 1405.0, + 975.0, + 294.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 970.0, + 1195.0, + 970.0, + 1195.0, + 1005.0, + 322.0, + 1005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1290.0, + 1404.0, + 1290.0, + 1404.0, + 1329.0, + 297.0, + 1329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1321.0, + 1403.0, + 1321.0, + 1403.0, + 1359.0, + 323.0, + 1359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1355.0, + 990.0, + 1355.0, + 990.0, + 1386.0, + 323.0, + 1386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1645.0, + 1403.0, + 1645.0, + 1403.0, + 1683.0, + 294.0, + 1683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1678.0, + 1405.0, + 1678.0, + 1405.0, + 1713.0, + 323.0, + 1713.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1709.0, + 1262.0, + 1709.0, + 1262.0, + 1743.0, + 323.0, + 1743.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1969.0, + 1401.0, + 1969.0, + 1401.0, + 2008.0, + 295.0, + 2008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 2003.0, + 1026.0, + 2003.0, + 1026.0, + 2036.0, + 322.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 789.0, + 1405.0, + 789.0, + 1405.0, + 828.0, + 294.0, + 828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 820.0, + 1405.0, + 820.0, + 1405.0, + 859.0, + 320.0, + 859.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 852.0, + 1404.0, + 852.0, + 1404.0, + 888.0, + 322.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 880.0, + 485.0, + 880.0, + 485.0, + 916.0, + 320.0, + 916.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1881.0, + 1404.0, + 1881.0, + 1404.0, + 1921.0, + 293.0, + 1921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1912.0, + 614.0, + 1912.0, + 614.0, + 1946.0, + 320.0, + 1946.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 10, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 299, + 1115, + 1402, + 1115, + 1402, + 1209, + 299, + 1209 + ], + "score": 0.853 + }, + { + "category_id": 2, + "poly": [ + 836, + 2088, + 865, + 2088, + 865, + 2113, + 836, + 2113 + ], + "score": 0.836 + }, + { + "category_id": 1, + "poly": [ + 296, + 1227, + 1399, + 1227, + 1399, + 1293, + 296, + 1293 + ], + "score": 0.817 + }, + { + "category_id": 1, + "poly": [ + 298, + 1311, + 1404, + 1311, + 1404, + 1405, + 298, + 1405 + ], + "score": 0.787 + }, + { + "category_id": 1, + "poly": [ + 298, + 971, + 1403, + 971, + 1403, + 1094, + 298, + 1094 + ], + "score": 0.731 + }, + { + "category_id": 1, + "poly": [ + 295, + 856, + 1401, + 856, + 1401, + 951, + 295, + 951 + ], + "score": 0.703 + }, + { + "category_id": 1, + "poly": [ + 297, + 744, + 1401, + 744, + 1401, + 838, + 297, + 838 + ], + "score": 0.664 + }, + { + "category_id": 1, + "poly": [ + 297, + 630, + 1401, + 630, + 1401, + 725, + 297, + 725 + ], + "score": 0.66 + }, + { + "category_id": 1, + "poly": [ + 298, + 229, + 1404, + 229, + 1404, + 353, + 298, + 353 + ], + "score": 0.626 + }, + { + "category_id": 1, + "poly": [ + 300, + 372, + 1397, + 372, + 1397, + 467, + 300, + 467 + ], + "score": 0.622 + }, + { + "category_id": 1, + "poly": [ + 296, + 485, + 1403, + 485, + 1403, + 611, + 296, + 611 + ], + "score": 0.578 + }, + { + "category_id": 2, + "poly": [ + 298, + 75, + 856, + 75, + 856, + 105, + 298, + 105 + ], + "score": 0.557 + }, + { + "category_id": 2, + "poly": [ + 298, + 75, + 856, + 75, + 856, + 105, + 298, + 105 + ], + "score": 0.352 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 870.0, + 2085.0, + 870.0, + 2124.0, + 832.0, + 2124.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1116.0, + 1402.0, + 1116.0, + 1402.0, + 1150.0, + 294.0, + 1150.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1147.0, + 1404.0, + 1147.0, + 1404.0, + 1181.0, + 324.0, + 1181.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 327.0, + 1177.0, + 669.0, + 1177.0, + 669.0, + 1211.0, + 327.0, + 1211.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1228.0, + 1404.0, + 1228.0, + 1404.0, + 1264.0, + 295.0, + 1264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 1259.0, + 1291.0, + 1259.0, + 1291.0, + 1294.0, + 321.0, + 1294.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1309.0, + 1403.0, + 1309.0, + 1403.0, + 1348.0, + 292.0, + 1348.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 1342.0, + 1403.0, + 1342.0, + 1403.0, + 1379.0, + 321.0, + 1379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 325.0, + 1374.0, + 531.0, + 1374.0, + 531.0, + 1404.0, + 325.0, + 1404.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 971.0, + 1405.0, + 971.0, + 1405.0, + 1007.0, + 296.0, + 1007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 1000.0, + 1405.0, + 1000.0, + 1405.0, + 1038.0, + 318.0, + 1038.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1034.0, + 1404.0, + 1034.0, + 1404.0, + 1070.0, + 323.0, + 1070.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1064.0, + 678.0, + 1064.0, + 678.0, + 1096.0, + 322.0, + 1096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 855.0, + 1405.0, + 855.0, + 1405.0, + 896.0, + 293.0, + 896.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 885.0, + 1405.0, + 885.0, + 1405.0, + 929.0, + 321.0, + 929.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 919.0, + 534.0, + 919.0, + 534.0, + 950.0, + 324.0, + 950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 744.0, + 1402.0, + 744.0, + 1402.0, + 781.0, + 295.0, + 781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 776.0, + 1403.0, + 776.0, + 1403.0, + 810.0, + 324.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 808.0, + 707.0, + 808.0, + 707.0, + 838.0, + 320.0, + 838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 629.0, + 1403.0, + 629.0, + 1403.0, + 668.0, + 296.0, + 668.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 662.0, + 1406.0, + 662.0, + 1406.0, + 697.0, + 320.0, + 697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 691.0, + 1220.0, + 691.0, + 1220.0, + 729.0, + 322.0, + 729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 227.0, + 1405.0, + 227.0, + 1405.0, + 265.0, + 292.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 256.0, + 1406.0, + 256.0, + 1406.0, + 300.0, + 318.0, + 300.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 293.0, + 1404.0, + 293.0, + 1404.0, + 326.0, + 322.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 323.0, + 686.0, + 323.0, + 686.0, + 355.0, + 323.0, + 355.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 371.0, + 1402.0, + 371.0, + 1402.0, + 410.0, + 294.0, + 410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 325.0, + 405.0, + 1401.0, + 405.0, + 1401.0, + 439.0, + 325.0, + 439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 437.0, + 867.0, + 437.0, + 867.0, + 468.0, + 324.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 486.0, + 1404.0, + 486.0, + 1404.0, + 522.0, + 295.0, + 522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 519.0, + 1404.0, + 519.0, + 1404.0, + 552.0, + 323.0, + 552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 547.0, + 1406.0, + 547.0, + 1406.0, + 586.0, + 320.0, + 586.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 580.0, + 666.0, + 580.0, + 666.0, + 613.0, + 320.0, + 613.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 11, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 899, + 1403, + 899, + 1403, + 1144, + 298, + 1144 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 299, + 356, + 878, + 356, + 878, + 722, + 299, + 722 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 297, + 1650, + 1403, + 1650, + 1403, + 1867, + 297, + 1867 + ], + "score": 0.977 + }, + { + "category_id": 5, + "poly": [ + 296, + 1447, + 852, + 1447, + 852, + 1611, + 296, + 1611 + ], + "score": 0.977, + "html": "
combinationA-847A-150
FT ->MPT (default)9.029.6
MPT ->FT8.5 (-0.5)28.1 (-1.5)
FT + MPT sim.8.8 (-0.2)29.0 (-0.6)
" + }, + { + "category_id": 1, + "poly": [ + 299, + 1880, + 1402, + 1880, + 1402, + 2034, + 299, + 2034 + ], + "score": 0.976 + }, + { + "category_id": 3, + "poly": [ + 904, + 353, + 1397, + 353, + 1397, + 632, + 904, + 632 + ], + "score": 0.972 + }, + { + "category_id": 5, + "poly": [ + 937, + 1364, + 1343, + 1364, + 1343, + 1601, + 937, + 1601 + ], + "score": 0.972, + "html": "
prompt depthA-150
1w/o FT 25.7w/FT 29.3
3 (default)26.5
26.829.6 29.4
6 1226.829.3
" + }, + { + "category_id": 4, + "poly": [ + 902, + 650, + 1400, + 650, + 1400, + 742, + 902, + 742 + ], + "score": 0.954 + }, + { + "category_id": 6, + "poly": [ + 295, + 1280, + 852, + 1280, + 852, + 1434, + 295, + 1434 + ], + "score": 0.953 + }, + { + "category_id": 6, + "poly": [ + 880, + 1289, + 1402, + 1289, + 1402, + 1353, + 880, + 1353 + ], + "score": 0.944 + }, + { + "category_id": 0, + "poly": [ + 302, + 298, + 745, + 298, + 745, + 328, + 302, + 328 + ], + "score": 0.896 + }, + { + "category_id": 2, + "poly": [ + 298, + 75, + 856, + 75, + 856, + 104, + 298, + 104 + ], + "score": 0.889 + }, + { + "category_id": 0, + "poly": [ + 299, + 1188, + 1004, + 1188, + 1004, + 1217, + 299, + 1217 + ], + "score": 0.872 + }, + { + "category_id": 0, + "poly": [ + 301, + 227, + 507, + 227, + 507, + 261, + 301, + 261 + ], + "score": 0.871 + }, + { + "category_id": 2, + "poly": [ + 837, + 2088, + 864, + 2088, + 864, + 2112, + 837, + 2112 + ], + "score": 0.841 + }, + { + "category_id": 0, + "poly": [ + 299, + 842, + 897, + 842, + 897, + 872, + 299, + 872 + ], + "score": 0.807 + }, + { + "category_id": 1, + "poly": [ + 290, + 770, + 1398, + 770, + 1398, + 801, + 290, + 801 + ], + "score": 0.48 + }, + { + "category_id": 13, + "poly": [ + 1258, + 1880, + 1393, + 1880, + 1393, + 1915, + 1258, + 1915 + ], + "score": 0.9, + "latex": "\\{ 1 , 3 , 6 , 1 2 \\}" + }, + { + "category_id": 13, + "poly": [ + 376, + 1023, + 490, + 1023, + 490, + 1052, + 376, + 1052 + ], + "score": 0.9, + "latex": "5 1 2 \\times 5 1 2" + }, + { + "category_id": 13, + "poly": [ + 345, + 991, + 405, + 991, + 405, + 1020, + 345, + 1020 + ], + "score": 0.9, + "latex": "1 0 ^ { - 4 }" + }, + { + "category_id": 13, + "poly": [ + 1304, + 960, + 1400, + 960, + 1400, + 990, + 1304, + 990 + ], + "score": 0.89, + "latex": "\\bar { 2 } \\cdot 1 0 ^ { - 4 }" + }, + { + "category_id": 13, + "poly": [ + 1229, + 1973, + 1304, + 1973, + 1304, + 2002, + 1229, + 2002 + ], + "score": 0.87, + "latex": "2 5 . 7 \\%" + }, + { + "category_id": 13, + "poly": [ + 1121, + 1973, + 1197, + 1973, + 1197, + 2002, + 1121, + 2002 + ], + "score": 0.87, + "latex": "2 1 . 8 \\%" + }, + { + "category_id": 13, + "poly": [ + 326, + 540, + 393, + 540, + 393, + 569, + 326, + 569 + ], + "score": 0.86, + "latex": "- 6 . 3 \\%" + }, + { + "category_id": 13, + "poly": [ + 646, + 509, + 720, + 509, + 720, + 539, + 646, + 539 + ], + "score": 0.84, + "latex": "1 3 . 8 \\%" + }, + { + "category_id": 13, + "poly": [ + 394, + 1774, + 456, + 1774, + 456, + 1803, + 394, + 1803 + ], + "score": 0.83, + "latex": ". 1 . 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 952, + 1743, + 1013, + 1743, + 1013, + 1773, + 952, + 1773 + ], + "score": 0.78, + "latex": "{ \\mathrm { - } } { \\mathrm { P T } }" + }, + { + "category_id": 13, + "poly": [ + 731, + 1835, + 857, + 1835, + 857, + 1865, + 731, + 1865 + ], + "score": 0.6, + "latex": "\\mathrm { ( F T + M P T }" + }, + { + "category_id": 13, + "poly": [ + 626, + 1683, + 748, + 1683, + 748, + 1711, + 626, + 1711 + ], + "score": 0.59, + "latex": "\\mathrm { \\mathbf { \\hat { T } } \\mathbf { \\Phi } } \\mathrm { \\mathbf { - } } \\mathrm { \\mathbf { \\mathrm { > } } } \\mathbf { M } \\mathbf { P } \\mathrm { T } \\mathrm { \\mathbf { \\Phi } }" + }, + { + "category_id": 13, + "poly": [ + 708, + 1312, + 792, + 1312, + 792, + 1341, + 708, + 1341 + ], + "score": 0.58, + "latex": "{ \\bf \\cdot } { \\bf > } { \\bf M } { \\bf P } { \\bf T }" + }, + { + "category_id": 13, + "poly": [ + 296, + 1375, + 319, + 1375, + 319, + 1401, + 296, + 1401 + ], + "score": 0.42, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 295, + 1373, + 381, + 1373, + 381, + 1402, + 295, + 1402 + ], + "score": 0.37, + "latex": "+ { \\mathbf { M P T } }" + }, + { + "category_id": 15, + "poly": [ + 1128.0, + 338.0, + 1421.0, + 338.0, + 1421.0, + 607.0, + 1128.0, + 607.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 348.0, + 1135.0, + 348.0, + 1135.0, + 593.0, + 899.0, + 593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 915.0, + 589.0, + 1124.0, + 589.0, + 1124.0, + 634.0, + 915.0, + 634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1178.0, + 588.0, + 1401.0, + 588.0, + 1401.0, + 633.0, + 1178.0, + 633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 900.0, + 649.0, + 1403.0, + 649.0, + 1403.0, + 683.0, + 900.0, + 683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 900.0, + 680.0, + 1402.0, + 680.0, + 1402.0, + 716.0, + 900.0, + 716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 902.0, + 713.0, + 1116.0, + 713.0, + 1116.0, + 742.0, + 902.0, + 742.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1279.0, + 853.0, + 1279.0, + 853.0, + 1315.0, + 295.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1311.0, + 707.0, + 1311.0, + 707.0, + 1343.0, + 295.0, + 1343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.0, + 1311.0, + 854.0, + 1311.0, + 854.0, + 1343.0, + 793.0, + 1343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1341.0, + 855.0, + 1341.0, + 855.0, + 1375.0, + 295.0, + 1375.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 382.0, + 1371.0, + 855.0, + 1371.0, + 855.0, + 1407.0, + 382.0, + 1407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1400.0, + 476.0, + 1400.0, + 476.0, + 1437.0, + 295.0, + 1437.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 878.0, + 1288.0, + 1404.0, + 1288.0, + 1404.0, + 1323.0, + 878.0, + 1323.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 879.0, + 1319.0, + 1394.0, + 1319.0, + 1394.0, + 1355.0, + 879.0, + 1355.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 298.0, + 750.0, + 298.0, + 750.0, + 332.0, + 297.0, + 332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 72.0, + 858.0, + 72.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1185.0, + 1008.0, + 1185.0, + 1008.0, + 1220.0, + 296.0, + 1220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 222.0, + 513.0, + 222.0, + 513.0, + 268.0, + 294.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 830.0, + 2086.0, + 868.0, + 2086.0, + 868.0, + 2121.0, + 830.0, + 2121.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 841.0, + 900.0, + 841.0, + 900.0, + 877.0, + 295.0, + 877.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 901.0, + 1404.0, + 901.0, + 1404.0, + 934.0, + 294.0, + 934.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 933.0, + 1404.0, + 933.0, + 1404.0, + 963.0, + 296.0, + 963.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 959.0, + 1303.0, + 959.0, + 1303.0, + 996.0, + 292.0, + 996.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1401.0, + 959.0, + 1408.0, + 959.0, + 1408.0, + 996.0, + 1401.0, + 996.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 987.0, + 344.0, + 987.0, + 344.0, + 1030.0, + 291.0, + 1030.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 406.0, + 987.0, + 1408.0, + 987.0, + 1408.0, + 1030.0, + 406.0, + 1030.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1020.0, + 375.0, + 1020.0, + 375.0, + 1057.0, + 293.0, + 1057.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 491.0, + 1020.0, + 1407.0, + 1020.0, + 1407.0, + 1057.0, + 491.0, + 1057.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1047.0, + 1405.0, + 1047.0, + 1405.0, + 1092.0, + 291.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1080.0, + 1407.0, + 1080.0, + 1407.0, + 1121.0, + 292.0, + 1121.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1114.0, + 618.0, + 1114.0, + 618.0, + 1147.0, + 294.0, + 1147.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 356.0, + 882.0, + 356.0, + 882.0, + 389.0, + 294.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 382.0, + 883.0, + 382.0, + 883.0, + 424.0, + 294.0, + 424.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 417.0, + 882.0, + 417.0, + 882.0, + 450.0, + 295.0, + 450.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 449.0, + 881.0, + 449.0, + 881.0, + 480.0, + 294.0, + 480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 479.0, + 881.0, + 479.0, + 881.0, + 510.0, + 294.0, + 510.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 508.0, + 645.0, + 508.0, + 645.0, + 542.0, + 293.0, + 542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 508.0, + 882.0, + 508.0, + 882.0, + 542.0, + 721.0, + 542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 538.0, + 325.0, + 538.0, + 325.0, + 574.0, + 292.0, + 574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 538.0, + 882.0, + 538.0, + 882.0, + 574.0, + 394.0, + 574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 568.0, + 883.0, + 568.0, + 883.0, + 605.0, + 292.0, + 605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 599.0, + 881.0, + 599.0, + 881.0, + 632.0, + 293.0, + 632.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 629.0, + 882.0, + 629.0, + 882.0, + 663.0, + 294.0, + 663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 657.0, + 883.0, + 657.0, + 883.0, + 698.0, + 292.0, + 698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 688.0, + 481.0, + 688.0, + 481.0, + 728.0, + 295.0, + 728.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1650.0, + 1405.0, + 1650.0, + 1405.0, + 1687.0, + 295.0, + 1687.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1682.0, + 625.0, + 1682.0, + 625.0, + 1716.0, + 295.0, + 1716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 749.0, + 1682.0, + 1403.0, + 1682.0, + 1403.0, + 1716.0, + 749.0, + 1716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1709.0, + 1407.0, + 1709.0, + 1407.0, + 1750.0, + 292.0, + 1750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1740.0, + 951.0, + 1740.0, + 951.0, + 1780.0, + 294.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1014.0, + 1740.0, + 1405.0, + 1740.0, + 1405.0, + 1780.0, + 1014.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1773.0, + 393.0, + 1773.0, + 393.0, + 1808.0, + 293.0, + 1808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 457.0, + 1773.0, + 1405.0, + 1773.0, + 1405.0, + 1808.0, + 457.0, + 1808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1804.0, + 1405.0, + 1804.0, + 1405.0, + 1839.0, + 292.0, + 1839.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1837.0, + 730.0, + 1837.0, + 730.0, + 1867.0, + 296.0, + 1867.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 858.0, + 1837.0, + 1335.0, + 1837.0, + 1335.0, + 1867.0, + 858.0, + 1867.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1882.0, + 1257.0, + 1882.0, + 1257.0, + 1915.0, + 297.0, + 1915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1394.0, + 1882.0, + 1403.0, + 1882.0, + 1403.0, + 1915.0, + 1394.0, + 1915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1908.0, + 1404.0, + 1908.0, + 1404.0, + 1948.0, + 294.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1942.0, + 1407.0, + 1942.0, + 1407.0, + 1979.0, + 295.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1972.0, + 1120.0, + 1972.0, + 1120.0, + 2009.0, + 294.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 1972.0, + 1228.0, + 1972.0, + 1228.0, + 2009.0, + 1198.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1305.0, + 1972.0, + 1406.0, + 1972.0, + 1406.0, + 2009.0, + 1305.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 2002.0, + 1406.0, + 2002.0, + 1406.0, + 2040.0, + 292.0, + 2040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 766.0, + 1403.0, + 766.0, + 1403.0, + 805.0, + 294.0, + 805.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 12, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 299, + 386, + 1404, + 386, + 1404, + 571, + 299, + 571 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 298, + 229, + 1400, + 229, + 1400, + 293, + 298, + 293 + ], + "score": 0.947 + }, + { + "category_id": 2, + "poly": [ + 298, + 75, + 856, + 75, + 856, + 104, + 298, + 104 + ], + "score": 0.888 + }, + { + "category_id": 2, + "poly": [ + 836, + 2089, + 864, + 2089, + 864, + 2112, + 836, + 2112 + ], + "score": 0.838 + }, + { + "category_id": 0, + "poly": [ + 300, + 330, + 1042, + 330, + 1042, + 361, + 300, + 361 + ], + "score": 0.553 + }, + { + "category_id": 1, + "poly": [ + 300, + 330, + 1042, + 330, + 1042, + 361, + 300, + 361 + ], + "score": 0.379 + }, + { + "category_id": 13, + "poly": [ + 860, + 479, + 935, + 479, + 935, + 508, + 860, + 508 + ], + "score": 0.88, + "latex": "3 7 . 6 \\%" + }, + { + "category_id": 13, + "poly": [ + 984, + 479, + 1058, + 479, + 1058, + 508, + 984, + 508 + ], + "score": 0.87, + "latex": "2 1 . 9 \\%" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 72.0, + 858.0, + 72.0, + 858.0, + 108.0, + 298.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 870.0, + 2085.0, + 870.0, + 2124.0, + 832.0, + 2124.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 329.0, + 1047.0, + 329.0, + 1047.0, + 365.0, + 297.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 384.0, + 1404.0, + 384.0, + 1404.0, + 421.0, + 295.0, + 421.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 419.0, + 1405.0, + 419.0, + 1405.0, + 451.0, + 297.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 448.0, + 1404.0, + 448.0, + 1404.0, + 484.0, + 294.0, + 484.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 476.0, + 859.0, + 476.0, + 859.0, + 514.0, + 294.0, + 514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 936.0, + 476.0, + 983.0, + 476.0, + 983.0, + 514.0, + 936.0, + 514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1059.0, + 476.0, + 1404.0, + 476.0, + 1404.0, + 514.0, + 1059.0, + 514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 508.0, + 1404.0, + 508.0, + 1404.0, + 546.0, + 294.0, + 546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 538.0, + 728.0, + 538.0, + 728.0, + 574.0, + 295.0, + 574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 226.0, + 1404.0, + 226.0, + 1404.0, + 267.0, + 293.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 258.0, + 842.0, + 258.0, + 842.0, + 298.0, + 296.0, + 298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 329.0, + 1047.0, + 329.0, + 1047.0, + 365.0, + 297.0, + 365.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 13, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 4, + "poly": [ + 299, + 1960, + 1403, + 1960, + 1403, + 2025, + 299, + 2025 + ], + "score": 0.905 + }, + { + "category_id": 2, + "poly": [ + 836, + 2088, + 864, + 2088, + 864, + 2113, + 836, + 2113 + ], + "score": 0.858 + }, + { + "category_id": 2, + "poly": [ + 299, + 76, + 855, + 76, + 855, + 104, + 299, + 104 + ], + "score": 0.839 + }, + { + "category_id": 3, + "poly": [ + 569, + 47, + 1119, + 47, + 1119, + 1949, + 569, + 1949 + ], + "score": 0.8 + }, + { + "category_id": 3, + "poly": [ + 568, + 88, + 1120, + 88, + 1120, + 1958, + 568, + 1958 + ], + "score": 0.528 + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1960.0, + 1403.0, + 1960.0, + 1403.0, + 1996.0, + 295.0, + 1996.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1990.0, + 924.0, + 1990.0, + 924.0, + 2026.0, + 295.0, + 2026.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 870.0, + 2085.0, + 870.0, + 2124.0, + 832.0, + 2124.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 73.0, + 858.0, + 73.0, + 858.0, + 108.0, + 297.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 874.0, + 114.0, + 984.0, + 114.0, + 984.0, + 147.0, + 874.0, + 147.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 724.0, + 153.0, + 746.0, + 153.0, + 746.0, + 191.0, + 724.0, + 191.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 803.0, + 141.0, + 836.0, + 141.0, + 836.0, + 181.0, + 803.0, + 181.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 889.0, + 141.0, + 921.0, + 141.0, + 921.0, + 181.0, + 889.0, + 181.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 141.0, + 1004.0, + 141.0, + 1004.0, + 181.0, + 974.0, + 181.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1059.0, + 141.0, + 1090.0, + 141.0, + 1090.0, + 181.0, + 1059.0, + 181.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 677.0, + 197.0, + 698.0, + 197.0, + 698.0, + 216.0, + 677.0, + 216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 619.0, + 201.0, + 698.0, + 201.0, + 698.0, + 243.0, + 619.0, + 243.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 671.0, + 212.0, + 718.0, + 212.0, + 718.0, + 256.0, + 671.0, + 256.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 574.0, + 224.0, + 684.0, + 224.0, + 684.0, + 258.0, + 574.0, + 258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 708.0, + 256.0, + 722.0, + 256.0, + 722.0, + 281.0, + 708.0, + 281.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 688.0, + 268.0, + 706.0, + 268.0, + 706.0, + 283.0, + 688.0, + 283.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 661.0, + 287.0, + 696.0, + 287.0, + 696.0, + 312.0, + 661.0, + 312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 694.0, + 291.0, + 710.0, + 291.0, + 710.0, + 306.0, + 694.0, + 306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 577.0, + 305.0, + 724.0, + 305.0, + 724.0, + 386.0, + 577.0, + 386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 312.0, + 745.0, + 312.0, + 745.0, + 414.0, + 645.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 688.0, + 383.0, + 708.0, + 383.0, + 708.0, + 402.0, + 688.0, + 402.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 728.0, + 395.0, + 742.0, + 395.0, + 742.0, + 435.0, + 728.0, + 435.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 641.0, + 397.0, + 702.0, + 397.0, + 702.0, + 443.0, + 641.0, + 443.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 641.0, + 445.0, + 696.0, + 445.0, + 696.0, + 511.0, + 641.0, + 511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 728.0, + 454.0, + 742.0, + 454.0, + 742.0, + 523.0, + 728.0, + 523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 493.0, + 688.0, + 493.0, + 688.0, + 523.0, + 645.0, + 523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 644.0, + 516.0, + 746.0, + 516.0, + 746.0, + 750.0, + 644.0, + 750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 649.0, + 573.0, + 704.0, + 573.0, + 704.0, + 612.0, + 649.0, + 612.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 641.0, + 598.0, + 702.0, + 598.0, + 702.0, + 627.0, + 641.0, + 627.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 730.0, + 618.0, + 744.0, + 618.0, + 744.0, + 666.0, + 730.0, + 666.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 625.0, + 646.0, + 716.0, + 646.0, + 716.0, + 719.0, + 625.0, + 719.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 675.0, + 694.0, + 710.0, + 694.0, + 710.0, + 727.0, + 675.0, + 727.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 690.0, + 719.0, + 716.0, + 719.0, + 716.0, + 777.0, + 690.0, + 777.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 728.0, + 719.0, + 742.0, + 719.0, + 742.0, + 760.0, + 728.0, + 760.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 633.0, + 727.0, + 692.0, + 727.0, + 692.0, + 754.0, + 633.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 609.0, + 741.0, + 699.0, + 741.0, + 699.0, + 776.0, + 609.0, + 776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 612.0, + 783.0, + 681.0, + 783.0, + 681.0, + 810.0, + 612.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 728.0, + 827.0, + 742.0, + 827.0, + 742.0, + 852.0, + 728.0, + 852.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 625.0, + 873.0, + 746.0, + 873.0, + 746.0, + 967.0, + 625.0, + 967.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 710.0, + 940.0, + 724.0, + 940.0, + 724.0, + 967.0, + 710.0, + 967.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 653.0, + 948.0, + 690.0, + 948.0, + 690.0, + 971.0, + 653.0, + 971.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 631.0, + 967.0, + 708.0, + 967.0, + 708.0, + 1011.0, + 631.0, + 1011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 639.0, + 1004.0, + 696.0, + 1004.0, + 696.0, + 1031.0, + 639.0, + 1031.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 692.0, + 1021.0, + 714.0, + 1021.0, + 714.0, + 1038.0, + 692.0, + 1038.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 643.0, + 1029.0, + 686.0, + 1029.0, + 686.0, + 1052.0, + 643.0, + 1052.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 698.0, + 1098.0, + 728.0, + 1098.0, + 728.0, + 1134.0, + 698.0, + 1134.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 610.0, + 1106.0, + 690.0, + 1106.0, + 690.0, + 1138.0, + 610.0, + 1138.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 602.0, + 1131.0, + 694.0, + 1131.0, + 694.0, + 1217.0, + 602.0, + 1217.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 677.0, + 1171.0, + 694.0, + 1171.0, + 694.0, + 1190.0, + 677.0, + 1190.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 692.0, + 1200.0, + 730.0, + 1200.0, + 730.0, + 1248.0, + 692.0, + 1248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 657.0, + 1209.0, + 679.0, + 1209.0, + 679.0, + 1238.0, + 657.0, + 1238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 864.0, + 1225.0, + 917.0, + 1225.0, + 917.0, + 1590.0, + 864.0, + 1590.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 621.0, + 1234.0, + 681.0, + 1234.0, + 681.0, + 1263.0, + 621.0, + 1263.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 663.0, + 1267.0, + 681.0, + 1267.0, + 681.0, + 1282.0, + 663.0, + 1282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 673.0, + 1300.0, + 694.0, + 1300.0, + 694.0, + 1317.0, + 673.0, + 1317.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 560.0, + 1330.0, + 702.0, + 1330.0, + 702.0, + 1403.0, + 560.0, + 1403.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 673.0, + 1405.0, + 694.0, + 1405.0, + 694.0, + 1419.0, + 673.0, + 1419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 605.0, + 1423.0, + 721.0, + 1423.0, + 721.0, + 1522.0, + 605.0, + 1522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 651.0, + 1423.0, + 690.0, + 1423.0, + 690.0, + 1446.0, + 651.0, + 1446.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 584.0, + 1471.0, + 759.0, + 1471.0, + 759.0, + 1630.0, + 584.0, + 1630.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 673.0, + 1496.0, + 720.0, + 1496.0, + 720.0, + 1511.0, + 673.0, + 1511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1064.0, + 1500.0, + 1116.0, + 1500.0, + 1116.0, + 1885.0, + 1064.0, + 1885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1027.0, + 1507.0, + 1078.0, + 1507.0, + 1078.0, + 1926.0, + 1027.0, + 1926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 681.0, + 1557.0, + 714.0, + 1557.0, + 714.0, + 1578.0, + 681.0, + 1578.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 730.0, + 1588.0, + 753.0, + 1588.0, + 753.0, + 1619.0, + 730.0, + 1619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 1602.0, + 765.0, + 1602.0, + 765.0, + 1939.0, + 721.0, + 1939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 633.0, + 1753.0, + 692.0, + 1753.0, + 692.0, + 1801.0, + 633.0, + 1801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 686.0, + 1776.0, + 700.0, + 1776.0, + 700.0, + 1793.0, + 686.0, + 1793.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 608.0, + 1799.0, + 708.0, + 1799.0, + 708.0, + 1849.0, + 608.0, + 1849.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 669.0, + 1832.0, + 688.0, + 1832.0, + 688.0, + 1851.0, + 669.0, + 1851.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 692.0, + 1843.0, + 718.0, + 1843.0, + 718.0, + 1872.0, + 692.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 659.0, + 1853.0, + 692.0, + 1853.0, + 692.0, + 1876.0, + 659.0, + 1876.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 694.0, + 1868.0, + 712.0, + 1868.0, + 712.0, + 1884.0, + 694.0, + 1884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 667.0, + 1891.0, + 692.0, + 1891.0, + 692.0, + 1922.0, + 667.0, + 1922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 673.0, + 1916.0, + 692.0, + 1916.0, + 692.0, + 1930.0, + 673.0, + 1930.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 557.0, + 70.0, + 863.0, + 70.0, + 863.0, + 107.5, + 557.0, + 107.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 634.0, + 239.5, + 701.0, + 239.5, + 701.0, + 279.0, + 634.0, + 279.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 666.0, + 425.0, + 711.0, + 425.0, + 711.0, + 447.0, + 666.0, + 447.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 634.25, + 825.0, + 695.25, + 825.0, + 695.25, + 863.0, + 634.25, + 863.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 679.0, + 868.0, + 682.0, + 868.0, + 682.0, + 892.5, + 679.0, + 892.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 690.0, + 997.5, + 714.0, + 997.5, + 714.0, + 1012.5, + 690.0, + 1012.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 588.0, + 1560.0, + 728.0, + 1560.0, + 728.0, + 1625.0, + 588.0, + 1625.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 633.75, + 1734.0, + 674.75, + 1734.0, + 674.75, + 1763.5, + 633.75, + 1763.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 594.75, + 1875.5, + 675.75, + 1875.5, + 675.75, + 1923.0, + 594.75, + 1923.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 563.0, + 79.0, + 860.0, + 79.0, + 860.0, + 106.0, + 563.0, + 106.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 873.0, + 110.0, + 983.0, + 110.0, + 983.0, + 145.0, + 873.0, + 145.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 726.0, + 157.0, + 744.0, + 157.0, + 744.0, + 175.0, + 726.0, + 175.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 805.0, + 141.0, + 836.0, + 141.0, + 836.0, + 180.0, + 805.0, + 180.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 893.0, + 145.0, + 917.0, + 145.0, + 917.0, + 178.0, + 893.0, + 178.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 143.0, + 1001.0, + 143.0, + 1001.0, + 178.0, + 976.0, + 178.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1058.0, + 141.0, + 1088.0, + 141.0, + 1088.0, + 180.0, + 1058.0, + 180.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 685.0, + 180.0, + 730.0, + 180.0, + 730.0, + 245.0, + 685.0, + 245.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 616.0, + 198.0, + 703.0, + 198.0, + 703.0, + 247.0, + 616.0, + 247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 677.0, + 198.0, + 695.0, + 198.0, + 695.0, + 212.0, + 677.0, + 212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 577.0, + 227.0, + 689.0, + 227.0, + 689.0, + 253.0, + 577.0, + 253.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 691.0, + 233.0, + 707.0, + 233.0, + 707.0, + 247.0, + 691.0, + 247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 635.0, + 242.0, + 690.0, + 242.0, + 690.0, + 275.0, + 635.0, + 275.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 705.0, + 243.0, + 720.0, + 243.0, + 720.0, + 260.0, + 705.0, + 260.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 707.0, + 256.0, + 722.0, + 256.0, + 722.0, + 280.0, + 707.0, + 280.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 685.0, + 266.0, + 701.0, + 266.0, + 701.0, + 280.0, + 685.0, + 280.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 663.0, + 286.0, + 695.0, + 286.0, + 695.0, + 311.0, + 663.0, + 311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 585.0, + 306.0, + 717.0, + 306.0, + 717.0, + 384.0, + 585.0, + 384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 689.0, + 321.0, + 722.0, + 321.0, + 722.0, + 356.0, + 689.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 684.0, + 336.0, + 732.0, + 336.0, + 732.0, + 387.0, + 684.0, + 387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 640.0, + 393.0, + 734.0, + 393.0, + 734.0, + 444.0, + 640.0, + 444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 690.0, + 434.0, + 725.0, + 434.0, + 725.0, + 506.0, + 690.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 648.0, + 444.0, + 697.0, + 444.0, + 697.0, + 496.0, + 648.0, + 496.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 683.0, + 463.0, + 707.0, + 463.0, + 707.0, + 496.0, + 683.0, + 496.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 655.0, + 481.0, + 707.0, + 481.0, + 707.0, + 508.0, + 655.0, + 508.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 646.0, + 494.0, + 689.0, + 494.0, + 689.0, + 522.0, + 646.0, + 522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 701.0, + 502.0, + 720.0, + 502.0, + 720.0, + 516.0, + 701.0, + 516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 728.0, + 510.0, + 742.0, + 510.0, + 742.0, + 551.0, + 728.0, + 551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 673.0, + 541.0, + 705.0, + 541.0, + 705.0, + 563.0, + 673.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 632.0, + 567.0, + 707.0, + 567.0, + 707.0, + 633.0, + 632.0, + 633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 681.0, + 623.0, + 703.0, + 623.0, + 703.0, + 656.0, + 681.0, + 656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 628.0, + 650.0, + 714.0, + 650.0, + 714.0, + 711.0, + 628.0, + 711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 646.0, + 678.0, + 693.0, + 678.0, + 693.0, + 707.0, + 646.0, + 707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 627.0, + 724.0, + 695.0, + 724.0, + 695.0, + 753.0, + 627.0, + 753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 609.0, + 784.0, + 680.0, + 784.0, + 680.0, + 807.0, + 609.0, + 807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 693.0, + 791.0, + 707.0, + 791.0, + 707.0, + 816.0, + 693.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 707.0, + 795.0, + 728.0, + 795.0, + 728.0, + 840.0, + 707.0, + 840.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 730.0, + 801.0, + 744.0, + 801.0, + 744.0, + 818.0, + 730.0, + 818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 634.0, + 830.0, + 695.0, + 830.0, + 695.0, + 859.0, + 634.0, + 859.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 663.0, + 867.0, + 693.0, + 867.0, + 693.0, + 890.0, + 663.0, + 890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 638.0, + 880.0, + 707.0, + 880.0, + 707.0, + 941.0, + 638.0, + 941.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 689.0, + 892.0, + 728.0, + 892.0, + 728.0, + 912.0, + 689.0, + 912.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 710.0, + 906.0, + 724.0, + 906.0, + 724.0, + 937.0, + 710.0, + 937.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 643.0, + 930.0, + 734.0, + 930.0, + 734.0, + 975.0, + 643.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 640.0, + 1005.0, + 695.0, + 1005.0, + 695.0, + 1031.0, + 640.0, + 1031.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 642.0, + 1029.0, + 699.0, + 1029.0, + 699.0, + 1056.0, + 642.0, + 1056.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 701.0, + 1078.0, + 726.0, + 1078.0, + 726.0, + 1093.0, + 701.0, + 1093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 701.0, + 1093.0, + 722.0, + 1093.0, + 722.0, + 1107.0, + 701.0, + 1107.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 603.0, + 1094.0, + 708.0, + 1094.0, + 708.0, + 1137.0, + 603.0, + 1137.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 695.0, + 1097.0, + 730.0, + 1097.0, + 730.0, + 1171.0, + 695.0, + 1171.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 632.0, + 1144.0, + 695.0, + 1144.0, + 695.0, + 1171.0, + 632.0, + 1171.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 612.0, + 1165.0, + 685.0, + 1165.0, + 685.0, + 1208.0, + 612.0, + 1208.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 646.0, + 1177.0, + 703.0, + 1177.0, + 703.0, + 1247.0, + 646.0, + 1247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 864.0, + 1226.0, + 917.0, + 1226.0, + 917.0, + 1589.0, + 864.0, + 1589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 620.0, + 1234.0, + 687.0, + 1234.0, + 687.0, + 1263.0, + 620.0, + 1263.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 673.0, + 1298.0, + 693.0, + 1298.0, + 693.0, + 1316.0, + 673.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 636.0, + 1341.0, + 689.0, + 1341.0, + 689.0, + 1368.0, + 636.0, + 1368.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 565.0, + 1349.0, + 699.0, + 1349.0, + 699.0, + 1396.0, + 565.0, + 1396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 673.0, + 1399.0, + 691.0, + 1399.0, + 691.0, + 1419.0, + 673.0, + 1419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 650.0, + 1421.0, + 701.0, + 1421.0, + 701.0, + 1450.0, + 650.0, + 1450.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 703.0, + 1454.0, + 714.0, + 1454.0, + 714.0, + 1464.0, + 703.0, + 1464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 675.0, + 1495.0, + 712.0, + 1495.0, + 712.0, + 1516.0, + 675.0, + 1516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 602.0, + 1500.0, + 739.0, + 1500.0, + 739.0, + 1578.0, + 602.0, + 1578.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1064.0, + 1503.0, + 1115.0, + 1503.0, + 1115.0, + 1885.0, + 1064.0, + 1885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1023.0, + 1504.0, + 1074.0, + 1504.0, + 1074.0, + 1925.0, + 1023.0, + 1925.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 675.0, + 1505.0, + 722.0, + 1505.0, + 722.0, + 1528.0, + 675.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 646.0, + 1563.0, + 679.0, + 1563.0, + 679.0, + 1577.0, + 646.0, + 1577.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 687.0, + 1583.0, + 720.0, + 1583.0, + 720.0, + 1604.0, + 687.0, + 1604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 722.0, + 1616.0, + 763.0, + 1616.0, + 763.0, + 1926.0, + 722.0, + 1926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 689.0, + 1764.0, + 705.0, + 1764.0, + 705.0, + 1780.0, + 689.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 685.0, + 1776.0, + 703.0, + 1776.0, + 703.0, + 1790.0, + 685.0, + 1790.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 665.0, + 1819.0, + 695.0, + 1819.0, + 695.0, + 1856.0, + 665.0, + 1856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 691.0, + 1836.0, + 732.0, + 1836.0, + 732.0, + 1938.0, + 691.0, + 1938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 659.0, + 1854.0, + 691.0, + 1854.0, + 691.0, + 1877.0, + 659.0, + 1877.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 683.0, + 1864.0, + 699.0, + 1864.0, + 699.0, + 1881.0, + 683.0, + 1881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 586.0, + 1869.0, + 707.0, + 1869.0, + 707.0, + 1948.0, + 586.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 702.0, + 291.0, + 720.0, + 291.0, + 720.0, + 320.0, + 702.0, + 320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 662.0, + 356.5, + 724.0, + 356.5, + 724.0, + 410.0, + 662.0, + 410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 611.0, + 743.0, + 700.0, + 743.0, + 700.0, + 778.0, + 611.0, + 778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 634.25, + 963.5, + 693.25, + 963.5, + 693.25, + 1006.5, + 634.25, + 1006.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 636.0, + 1429.0, + 704.0, + 1429.0, + 704.0, + 1507.0, + 636.0, + 1507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 694.0, + 1637.5, + 717.0, + 1637.5, + 717.0, + 1674.5, + 694.0, + 1674.5 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 14, + "width": 1700, + "height": 2200 + } + } +] \ No newline at end of file diff --git a/parse/dev/I9xE1Jsjfx/images/03ce01940e766fb80345a613f8c4980eb82e17fddd7a760e7db341f30921c6b9.jpg b/parse/dev/I9xE1Jsjfx/images/03ce01940e766fb80345a613f8c4980eb82e17fddd7a760e7db341f30921c6b9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2311093a0addbd3630f3c0c5a64f28cf737de426 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/03ce01940e766fb80345a613f8c4980eb82e17fddd7a760e7db341f30921c6b9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50e3f728386c9ba29d487a930f78367c6cdab590dd15cc65299b1eca9e403a05 +size 6940 diff --git a/parse/dev/I9xE1Jsjfx/images/1d4a06045635b34483fce62aa60352abccb47239e1c716e56f84185edd966087.jpg b/parse/dev/I9xE1Jsjfx/images/1d4a06045635b34483fce62aa60352abccb47239e1c716e56f84185edd966087.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cef8d1bc68cd8c891e0d6ad14a243b492b0716dc --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/1d4a06045635b34483fce62aa60352abccb47239e1c716e56f84185edd966087.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4900bc3a49233a8525ded9ae6d05d8ab08defa54a58520583e48459f29859d4 +size 80539 diff --git a/parse/dev/I9xE1Jsjfx/images/2bdd9785a9ba2c2f9c758143020175742cb0bbe3357211d3d4e660f98717baae.jpg b/parse/dev/I9xE1Jsjfx/images/2bdd9785a9ba2c2f9c758143020175742cb0bbe3357211d3d4e660f98717baae.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a6eccae10b4a708225aba6c43d19cf4705c73a04 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/2bdd9785a9ba2c2f9c758143020175742cb0bbe3357211d3d4e660f98717baae.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b374ffc40b83849cd47fe7debc3d63c4a4dd48caf2fc6231bc2ec3d7be7476df +size 3999 diff --git a/parse/dev/I9xE1Jsjfx/images/531d417bc63f69d40dd51e235a71dc8c85ad872964ab1ef6cbf46f7ddef440fd.jpg b/parse/dev/I9xE1Jsjfx/images/531d417bc63f69d40dd51e235a71dc8c85ad872964ab1ef6cbf46f7ddef440fd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..794fc433aca63b3b327b52024e605a42e0ad9a5f --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/531d417bc63f69d40dd51e235a71dc8c85ad872964ab1ef6cbf46f7ddef440fd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e55d0367e84172c0305b0d994176e1406a56d6d3db3f15b070d09314af43faac +size 55875 diff --git a/parse/dev/I9xE1Jsjfx/images/546d111d30cd8f4cbe57cdd904ad8d27fc502a826e6a4608f19c4123333ff5e0.jpg b/parse/dev/I9xE1Jsjfx/images/546d111d30cd8f4cbe57cdd904ad8d27fc502a826e6a4608f19c4123333ff5e0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3dd36030d85de962dba4474bf2b146cbb73e00b7 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/546d111d30cd8f4cbe57cdd904ad8d27fc502a826e6a4608f19c4123333ff5e0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b4511b5af233f5a4bdc3ade3ab773b33beda04d0fb93e6d9bbc67e44a28f5ad5 +size 27594 diff --git a/parse/dev/I9xE1Jsjfx/images/5b4a4ff471d85d69588df21015071d814484296cf34cb2c44d73b8ecd7983c2f.jpg b/parse/dev/I9xE1Jsjfx/images/5b4a4ff471d85d69588df21015071d814484296cf34cb2c44d73b8ecd7983c2f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..14b387d12e1bcac493263f429401a6d033ade80c --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/5b4a4ff471d85d69588df21015071d814484296cf34cb2c44d73b8ecd7983c2f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df08576c7c2eedb20d6e8f2b78a1f1dad5f98421a219c0c1dd54e388a2078796 +size 57509 diff --git a/parse/dev/I9xE1Jsjfx/images/72908e44327621ef80512524f924c0ee10e568b5d472904aec66cbc266d29c03.jpg b/parse/dev/I9xE1Jsjfx/images/72908e44327621ef80512524f924c0ee10e568b5d472904aec66cbc266d29c03.jpg new file mode 100644 index 0000000000000000000000000000000000000000..03017d074e6487e843c3b18730e3b82e6a55a640 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/72908e44327621ef80512524f924c0ee10e568b5d472904aec66cbc266d29c03.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24f1dcda62fdd35975791ebd486b8c8eab072baf6c50af928eb368533d7fd768 +size 57754 diff --git a/parse/dev/I9xE1Jsjfx/images/8247f8ddc1246f78ea3faa8900b2a4fc67d2985704d41e4ecd8c8c447f2652fa.jpg b/parse/dev/I9xE1Jsjfx/images/8247f8ddc1246f78ea3faa8900b2a4fc67d2985704d41e4ecd8c8c447f2652fa.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f6c0ac1189183779ed54f4aea1d4775fade29e63 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/8247f8ddc1246f78ea3faa8900b2a4fc67d2985704d41e4ecd8c8c447f2652fa.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:508ac03c714ce2e8fd840990bc744eec391e65571f8789ea03629329a158f97c +size 43808 diff --git a/parse/dev/I9xE1Jsjfx/images/929cb3ac1e376a2a76b0c3cb6da0c702804e7684966c143ac1aa19207fa4d6bb.jpg b/parse/dev/I9xE1Jsjfx/images/929cb3ac1e376a2a76b0c3cb6da0c702804e7684966c143ac1aa19207fa4d6bb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4f16ed2372bb0ef3b5fe549b112f168dff3d1641 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/929cb3ac1e376a2a76b0c3cb6da0c702804e7684966c143ac1aa19207fa4d6bb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d55cf71464fc4deba6324c0b77d06c6d5299c9c5e1c2d664b2398fbf0e9b4b3 +size 186222 diff --git a/parse/dev/I9xE1Jsjfx/images/9c539fff16a55003586255fe3b2c8fc933b73816543ca91973223a369a4c99eb.jpg b/parse/dev/I9xE1Jsjfx/images/9c539fff16a55003586255fe3b2c8fc933b73816543ca91973223a369a4c99eb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0e057aa2aa7a0763af21f7c43d7c8c6fc471d9b3 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/9c539fff16a55003586255fe3b2c8fc933b73816543ca91973223a369a4c99eb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c89c8a7e310b6bb54007f4bcf66f5215956b0998675450f11f40f79b21f361f +size 77102 diff --git a/parse/dev/I9xE1Jsjfx/images/a5515dc0d0d64addb78bc780b5034ff1b1ee727b9244a45c4667c319f16d4261.jpg b/parse/dev/I9xE1Jsjfx/images/a5515dc0d0d64addb78bc780b5034ff1b1ee727b9244a45c4667c319f16d4261.jpg new file mode 100644 index 0000000000000000000000000000000000000000..317d2b213b41acbc40f87658fd237470c68ea296 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/a5515dc0d0d64addb78bc780b5034ff1b1ee727b9244a45c4667c319f16d4261.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c1aaa9525aad02f6e8a994765dc6ae12a3185de63b2b007c9fcf1cce80eea65b +size 91272 diff --git a/parse/dev/I9xE1Jsjfx/images/b94e8274753b8d4b2a2f220f91837b8b41a3cc630192847a78b31b5ec54cb4c3.jpg b/parse/dev/I9xE1Jsjfx/images/b94e8274753b8d4b2a2f220f91837b8b41a3cc630192847a78b31b5ec54cb4c3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c348f42ef25d19a3f70945d94ab3e83cab6b6953 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/b94e8274753b8d4b2a2f220f91837b8b41a3cc630192847a78b31b5ec54cb4c3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:efb51c7513c3f54cc2f55b42ec8b2ea00951b45c26d277bc5896e7c4b08c5e97 +size 66526 diff --git a/parse/dev/I9xE1Jsjfx/images/c62243be2b9dff4b59fe0b315d6be2046ad7a185bb575215e6f2cac5af69dc25.jpg b/parse/dev/I9xE1Jsjfx/images/c62243be2b9dff4b59fe0b315d6be2046ad7a185bb575215e6f2cac5af69dc25.jpg new file mode 100644 index 0000000000000000000000000000000000000000..69aad49d6b2a51a817ce2a9b6f8c104ed8ccc047 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/c62243be2b9dff4b59fe0b315d6be2046ad7a185bb575215e6f2cac5af69dc25.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1f5a958aa70f14233843b27a361cc07384201d065691a322820fe89f7c52368 +size 93136 diff --git a/parse/dev/I9xE1Jsjfx/images/ce8de2016b4d824a33ee0a88c4afe6e7073ebcc33cfaa19af3605acc8837fa00.jpg b/parse/dev/I9xE1Jsjfx/images/ce8de2016b4d824a33ee0a88c4afe6e7073ebcc33cfaa19af3605acc8837fa00.jpg new file mode 100644 index 0000000000000000000000000000000000000000..423218f97e26f4158608f54efdbf6d7e3c2b9a96 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/ce8de2016b4d824a33ee0a88c4afe6e7073ebcc33cfaa19af3605acc8837fa00.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8173d4587814eeebc855d1a534471820abbc515ef5dc352a17caf1f3f3c8cfb3 +size 56618 diff --git a/parse/dev/I9xE1Jsjfx/images/cf5ed91b98ddb1a85ffe2c8fc54590ab6f4cb2127e5baaa083aca7c927bf2718.jpg b/parse/dev/I9xE1Jsjfx/images/cf5ed91b98ddb1a85ffe2c8fc54590ab6f4cb2127e5baaa083aca7c927bf2718.jpg new file mode 100644 index 0000000000000000000000000000000000000000..12497c923a0d276b628d9f1963fb9b3de7684aa6 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/cf5ed91b98ddb1a85ffe2c8fc54590ab6f4cb2127e5baaa083aca7c927bf2718.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89efd95b9713fec385354eb6fd2328720023d17d73bbee03130fa0da4f8e9ea9 +size 56286 diff --git a/parse/dev/I9xE1Jsjfx/images/fb89bae86831b2c88004e6e95d68daf94ba4300ad13d2fdd49a2ae3801d6dc5b.jpg b/parse/dev/I9xE1Jsjfx/images/fb89bae86831b2c88004e6e95d68daf94ba4300ad13d2fdd49a2ae3801d6dc5b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1e121a198ee5a8ec8d724ee2af67479a056337d0 --- /dev/null +++ b/parse/dev/I9xE1Jsjfx/images/fb89bae86831b2c88004e6e95d68daf94ba4300ad13d2fdd49a2ae3801d6dc5b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19619a125527c9b35428a32012031b51289624d4a0ef59e376c8e8fef514a4af +size 75383 diff --git a/parse/dev/Ih2bG6h1r4S/images/028d13c2b6b7713311944022490eca8111753bed108284b61d20c89e04ca0f30.jpg b/parse/dev/Ih2bG6h1r4S/images/028d13c2b6b7713311944022490eca8111753bed108284b61d20c89e04ca0f30.jpg new file mode 100644 index 0000000000000000000000000000000000000000..957e8ef9fc4d2a3cf4ce6f395b065a188768fe0b --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/028d13c2b6b7713311944022490eca8111753bed108284b61d20c89e04ca0f30.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:55ed2e8f3cc2aae5b247936c765968693dbacf1db8ea45eeb130b6e9990c4ff0 +size 49991 diff --git a/parse/dev/Ih2bG6h1r4S/images/172d3f5b955b90be7aafbce26e43f6e6536ef2fde2ba78db92373afd0148de66.jpg b/parse/dev/Ih2bG6h1r4S/images/172d3f5b955b90be7aafbce26e43f6e6536ef2fde2ba78db92373afd0148de66.jpg new file mode 100644 index 0000000000000000000000000000000000000000..55e8fa77884a536dd4045f112d3d1e534b1da643 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/172d3f5b955b90be7aafbce26e43f6e6536ef2fde2ba78db92373afd0148de66.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50a22493b650c0e89d23ac7d15d1b28c55ca3589ed2907e9b8fc513326dccbf8 +size 8119 diff --git a/parse/dev/Ih2bG6h1r4S/images/17e5c01c81a8228b695cb473c4143b12814436d94c854560a38cb0d03395832e.jpg b/parse/dev/Ih2bG6h1r4S/images/17e5c01c81a8228b695cb473c4143b12814436d94c854560a38cb0d03395832e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..01884d729f876b97b1ba3eb557c0b8b202dae023 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/17e5c01c81a8228b695cb473c4143b12814436d94c854560a38cb0d03395832e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:37fbb651317ec1810d46d570be514ceeccd6694f422cf75b3d67712a58b9c34a +size 6999 diff --git a/parse/dev/Ih2bG6h1r4S/images/1dc0147e883359e959f6be2022d913184433c48f1218f6f1d9bf3d66d296daf5.jpg b/parse/dev/Ih2bG6h1r4S/images/1dc0147e883359e959f6be2022d913184433c48f1218f6f1d9bf3d66d296daf5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2a419d1a931eb3014aea342e1b8298d0a4fdd112 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/1dc0147e883359e959f6be2022d913184433c48f1218f6f1d9bf3d66d296daf5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1bbb3aa78d8ece042930424e1b3a66c011b75f5e6760d25f1bb784aa803074a6 +size 9405 diff --git a/parse/dev/Ih2bG6h1r4S/images/21400ea6e71ae3699c131c816277d737f1eaa31e57fdaa0b327c33a825e1ab57.jpg b/parse/dev/Ih2bG6h1r4S/images/21400ea6e71ae3699c131c816277d737f1eaa31e57fdaa0b327c33a825e1ab57.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8257864586adc7aa7de4a61872105cb8a39c92ab --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/21400ea6e71ae3699c131c816277d737f1eaa31e57fdaa0b327c33a825e1ab57.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24613668c2c5614de705b83835911def1ef3da03709bb240d8a6b6a47ac2f0f8 +size 8427 diff --git a/parse/dev/Ih2bG6h1r4S/images/37ecf4172c618d313c9c69a080f62cc0f347d6201f000ae3a9355d77e4d4c362.jpg b/parse/dev/Ih2bG6h1r4S/images/37ecf4172c618d313c9c69a080f62cc0f347d6201f000ae3a9355d77e4d4c362.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0b98110681a235567cefa2af2d28790be3f26300 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/37ecf4172c618d313c9c69a080f62cc0f347d6201f000ae3a9355d77e4d4c362.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b3fafeb750891c5cd3d0470bba36b31d4a2129c41c62dd15d3a0a7af8b1268b +size 24930 diff --git a/parse/dev/Ih2bG6h1r4S/images/3f37be69d2a2d63f8cc9e2e0b505e672fb0a48d4b91739a97364c03ed4a72070.jpg b/parse/dev/Ih2bG6h1r4S/images/3f37be69d2a2d63f8cc9e2e0b505e672fb0a48d4b91739a97364c03ed4a72070.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a1f6bc705a6a00e9752934061e20231187bf8885 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/3f37be69d2a2d63f8cc9e2e0b505e672fb0a48d4b91739a97364c03ed4a72070.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:154860f9aae7f2cef00fcf16d4938799460f3b96469aa667c26f81fca48f23f3 +size 7670 diff --git a/parse/dev/Ih2bG6h1r4S/images/48a779adaf44eba98e2fe1a3bd4690138d9bcace728802639941cafdcd7a32b1.jpg b/parse/dev/Ih2bG6h1r4S/images/48a779adaf44eba98e2fe1a3bd4690138d9bcace728802639941cafdcd7a32b1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8ebb96ead151ab2398c676188862cf267198e602 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/48a779adaf44eba98e2fe1a3bd4690138d9bcace728802639941cafdcd7a32b1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6bb8d6ec828fda4a87e9666cec34796323cd7b64bdd46f374222047bd1276ec7 +size 15605 diff --git a/parse/dev/Ih2bG6h1r4S/images/62dab8029220c676c3074b985a2ab8483df5ee915ca30fac5e960ad5618cf15b.jpg b/parse/dev/Ih2bG6h1r4S/images/62dab8029220c676c3074b985a2ab8483df5ee915ca30fac5e960ad5618cf15b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..63c48b2cac04f4f4f858b7ac3eb473b3b221e6ec --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/62dab8029220c676c3074b985a2ab8483df5ee915ca30fac5e960ad5618cf15b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5f4420659c99115c5acba3bd8df2a2018e14c533dd7e0b704d92a05ab613759f +size 4053 diff --git a/parse/dev/Ih2bG6h1r4S/images/69fd83e684314d7b31ae10ecb0bc79671538bdac7605d3ff1ef68c887e93f4ad.jpg b/parse/dev/Ih2bG6h1r4S/images/69fd83e684314d7b31ae10ecb0bc79671538bdac7605d3ff1ef68c887e93f4ad.jpg new file mode 100644 index 0000000000000000000000000000000000000000..24013b4a63d4da128e1546d17dd5dc5964a61f53 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/69fd83e684314d7b31ae10ecb0bc79671538bdac7605d3ff1ef68c887e93f4ad.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3487f30bf7176756fe568453ff9401274f868169be51255bec823dc2078e55e0 +size 11620 diff --git a/parse/dev/Ih2bG6h1r4S/images/a69d9601804d8eee790f1274360694637d2a944d8cca3e75d34005b28d78ef9d.jpg b/parse/dev/Ih2bG6h1r4S/images/a69d9601804d8eee790f1274360694637d2a944d8cca3e75d34005b28d78ef9d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..129c46e3304cf090a84db18c3bdbf94ee2ae6b18 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/a69d9601804d8eee790f1274360694637d2a944d8cca3e75d34005b28d78ef9d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7341974bc763a7632724d1e641dfce56d2b03ce4bde871464d1532736d89d028 +size 33790 diff --git a/parse/dev/Ih2bG6h1r4S/images/abc047c8daa2d7d4508d0e7b12e7c5544158182f326a1f2c57252235ea6718be.jpg b/parse/dev/Ih2bG6h1r4S/images/abc047c8daa2d7d4508d0e7b12e7c5544158182f326a1f2c57252235ea6718be.jpg new file mode 100644 index 0000000000000000000000000000000000000000..03679cd81541d4c4899e9004ef683badd34266a6 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/abc047c8daa2d7d4508d0e7b12e7c5544158182f326a1f2c57252235ea6718be.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75c6bb7d5f74ba30168db7325c9b14ba952646984a6b8e18ffdc4a8cfa883bcc +size 56875 diff --git a/parse/dev/Ih2bG6h1r4S/images/af43d47f42b3aa8d8e5003e0391e47b8901ff329dc0d6a4498d6ca58c85159d1.jpg b/parse/dev/Ih2bG6h1r4S/images/af43d47f42b3aa8d8e5003e0391e47b8901ff329dc0d6a4498d6ca58c85159d1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..12b842727d175fbba3a985a153ce6566faa2009d --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/af43d47f42b3aa8d8e5003e0391e47b8901ff329dc0d6a4498d6ca58c85159d1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:55889f9d5ef4c808037e7cd1bbc48682f82be71bd73486e736e54d5faed24655 +size 9581 diff --git a/parse/dev/Ih2bG6h1r4S/images/c5f2017dcb9a32255e16051d74b2a19c166c0b5d374771d11f8d2cb9bd66bf5b.jpg b/parse/dev/Ih2bG6h1r4S/images/c5f2017dcb9a32255e16051d74b2a19c166c0b5d374771d11f8d2cb9bd66bf5b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b9d5dd01b0bea817610b21ee5a53e5f976b2ec7f --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/c5f2017dcb9a32255e16051d74b2a19c166c0b5d374771d11f8d2cb9bd66bf5b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:64140b0ffad8f2e41beeba5ba4d7e2dd731d4344d032990672abce5d93688a54 +size 25389 diff --git a/parse/dev/Ih2bG6h1r4S/images/c77a7e07edeec0d0abb8e373e1dd0afc01f91476965d858ad5c7886ad07d469c.jpg b/parse/dev/Ih2bG6h1r4S/images/c77a7e07edeec0d0abb8e373e1dd0afc01f91476965d858ad5c7886ad07d469c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8f607e967a3e8d70c067bc46dddbf91e3330fc65 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/c77a7e07edeec0d0abb8e373e1dd0afc01f91476965d858ad5c7886ad07d469c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d71abd96bd9a4fcd6f9353a2972b49046232e105b0ac71152d9e7e06fd63178 +size 5324 diff --git a/parse/dev/Ih2bG6h1r4S/images/c962bae81aa58e95b5d3e1c0a7d5ccbeaa795b9215016102e5552b539f5e1b50.jpg b/parse/dev/Ih2bG6h1r4S/images/c962bae81aa58e95b5d3e1c0a7d5ccbeaa795b9215016102e5552b539f5e1b50.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f4501129c2ee19d980e64ab01a602a4467a325d3 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/c962bae81aa58e95b5d3e1c0a7d5ccbeaa795b9215016102e5552b539f5e1b50.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b90ec558d65a19e9c9ddc22d1fb79c239edbd51238726fb78c25d8cafd2ccc5 +size 11514 diff --git a/parse/dev/Ih2bG6h1r4S/images/caa86e7cdb535c32b2b4442e1e8d257886bd9cea5bc4b88e38f938a6dc1900cc.jpg b/parse/dev/Ih2bG6h1r4S/images/caa86e7cdb535c32b2b4442e1e8d257886bd9cea5bc4b88e38f938a6dc1900cc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3d36ee4d78a9d5421c9495a9023001ab7c2d44ce --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/caa86e7cdb535c32b2b4442e1e8d257886bd9cea5bc4b88e38f938a6dc1900cc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa01d0514a7e4dcd0241ce7c036781cdf19a22757fda4a68bba460886855b3b6 +size 7667 diff --git a/parse/dev/Ih2bG6h1r4S/images/cf84bbc72c04c479ea06bbf48f63219072ab102b5c79bd6c876f7ed2bd38ae92.jpg b/parse/dev/Ih2bG6h1r4S/images/cf84bbc72c04c479ea06bbf48f63219072ab102b5c79bd6c876f7ed2bd38ae92.jpg new file mode 100644 index 0000000000000000000000000000000000000000..722e327a12a2fc05885c3f353f78195c666c7369 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/cf84bbc72c04c479ea06bbf48f63219072ab102b5c79bd6c876f7ed2bd38ae92.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13ccacbfb567e39dc65686081f37a25867d99228d3db279ab9e9c466f1603dfe +size 9393 diff --git a/parse/dev/Ih2bG6h1r4S/images/da9313910da0a408c010fc9a69dbb2ed85a4c7adf2661ead6e9efba059b37682.jpg b/parse/dev/Ih2bG6h1r4S/images/da9313910da0a408c010fc9a69dbb2ed85a4c7adf2661ead6e9efba059b37682.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ffed652ec8180950d6392e99d7c4946a6677a7c3 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/da9313910da0a408c010fc9a69dbb2ed85a4c7adf2661ead6e9efba059b37682.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b9b8e15ff4ec520054e4e9fcbd21badbb205c5bec2fe6a6b1d9b23d098cd52f2 +size 26317 diff --git a/parse/dev/Ih2bG6h1r4S/images/f4951e3b55afa1c7336dcdfa76ab152c7dd5f251cc69568c09843213b221d889.jpg b/parse/dev/Ih2bG6h1r4S/images/f4951e3b55afa1c7336dcdfa76ab152c7dd5f251cc69568c09843213b221d889.jpg new file mode 100644 index 0000000000000000000000000000000000000000..17e49ef6a4aca5f85ea890433f1536914f0b425d --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/f4951e3b55afa1c7336dcdfa76ab152c7dd5f251cc69568c09843213b221d889.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b4521cf263d1646298e91f98e8a9f36c77440f2ab30bbf5d7e1c7432c820ec20 +size 21755 diff --git a/parse/dev/Ih2bG6h1r4S/images/f5b41c215b545eed9c1d0d7f767355817a439db8fe08f105e81662c748816dd4.jpg b/parse/dev/Ih2bG6h1r4S/images/f5b41c215b545eed9c1d0d7f767355817a439db8fe08f105e81662c748816dd4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5bb88c634a98b7a54ebd94411adc4010628613c4 --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/f5b41c215b545eed9c1d0d7f767355817a439db8fe08f105e81662c748816dd4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ab2b07066cff07614e862f52da2c4dc8fa72a726d7fcca1a9d569ef9aecc173 +size 68882 diff --git a/parse/dev/Ih2bG6h1r4S/images/f5fd907a7fc8180b698937683a16d888a3d1dfdea2c78db1bd7b746b2e82ec33.jpg b/parse/dev/Ih2bG6h1r4S/images/f5fd907a7fc8180b698937683a16d888a3d1dfdea2c78db1bd7b746b2e82ec33.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7736c4cf03f0911c72d165c407b881ea551b59eb --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/f5fd907a7fc8180b698937683a16d888a3d1dfdea2c78db1bd7b746b2e82ec33.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:005eaf4ce0413eb648f00dedea54a83aaca9290900217cca686c55d3621d0621 +size 6164 diff --git a/parse/dev/Ih2bG6h1r4S/images/f6440d186ac7e09e72087844941b503a7a7dd8e137813e52eb5f03041577d570.jpg b/parse/dev/Ih2bG6h1r4S/images/f6440d186ac7e09e72087844941b503a7a7dd8e137813e52eb5f03041577d570.jpg new file mode 100644 index 0000000000000000000000000000000000000000..04c76e4a70c967a66c6be8e5efff3da58ca025ce --- /dev/null +++ b/parse/dev/Ih2bG6h1r4S/images/f6440d186ac7e09e72087844941b503a7a7dd8e137813e52eb5f03041577d570.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0a2cedcfd42ed8df1d42ae5c167ffc6725a579938c06680fe440ca76006c82c +size 4672 diff --git a/parse/dev/JprM0p-q0Co/images/03eb85937fe74958741a2b98bd056a5dcae026adc6639c77d586da5ef627cbd3.jpg b/parse/dev/JprM0p-q0Co/images/03eb85937fe74958741a2b98bd056a5dcae026adc6639c77d586da5ef627cbd3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e4b72b0b3f58561e8432100482bd1c8fdf1d1810 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/03eb85937fe74958741a2b98bd056a5dcae026adc6639c77d586da5ef627cbd3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:929a9e674096e2e03166cac467f6126addccf22eb4b9068f2d1bbf169791623d +size 26774 diff --git a/parse/dev/JprM0p-q0Co/images/07196b744ee36b1686a80b8d3929381102a579b1735f28f8dc5d2106a23d36b4.jpg b/parse/dev/JprM0p-q0Co/images/07196b744ee36b1686a80b8d3929381102a579b1735f28f8dc5d2106a23d36b4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d84be722cd3d4dfb46ba3f4feeafbf967982f18c --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/07196b744ee36b1686a80b8d3929381102a579b1735f28f8dc5d2106a23d36b4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e100a48e2456b1e2761d8746c4f7b559feb30479674a0058bba20db76457bdba +size 63062 diff --git a/parse/dev/JprM0p-q0Co/images/082996502209cbdc4bc84b7c4ede50a2a6ae0d7e28101538fd0fce8b0b0c4a25.jpg b/parse/dev/JprM0p-q0Co/images/082996502209cbdc4bc84b7c4ede50a2a6ae0d7e28101538fd0fce8b0b0c4a25.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9cf6f7e33d179c4aac36c37754647074c4639608 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/082996502209cbdc4bc84b7c4ede50a2a6ae0d7e28101538fd0fce8b0b0c4a25.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:973d55726b8089c08ee0b7f2f59a6b5236450eb09af1338110880ca716c782e1 +size 150631 diff --git a/parse/dev/JprM0p-q0Co/images/0ecd0f4d64686ca6a4a5f192539609c17016e9f41a96c4bb7f97dd5c73d91aa5.jpg b/parse/dev/JprM0p-q0Co/images/0ecd0f4d64686ca6a4a5f192539609c17016e9f41a96c4bb7f97dd5c73d91aa5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..76aa114743f0a8ba6716df669cd1997001843438 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/0ecd0f4d64686ca6a4a5f192539609c17016e9f41a96c4bb7f97dd5c73d91aa5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2cf275b45d1052058c0a3fb1ca780f412f39cb14f33eec8e79922551144e22dc +size 45347 diff --git a/parse/dev/JprM0p-q0Co/images/10acc8bebe9cf5c069809fd10ad839d19a1f1b0c952aa04729c74242dec4991b.jpg b/parse/dev/JprM0p-q0Co/images/10acc8bebe9cf5c069809fd10ad839d19a1f1b0c952aa04729c74242dec4991b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dc8ffe735cfee7e5ff20f61953a09c3549753e04 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/10acc8bebe9cf5c069809fd10ad839d19a1f1b0c952aa04729c74242dec4991b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:973e571b7861353ec59caf9ac3a768a23b7bbffe49bf962665f2c69ee80bfd06 +size 5559 diff --git a/parse/dev/JprM0p-q0Co/images/18850958e9b722cfb7a14882640fa796ea7c666e8e3da1bff9b490c3361882a5.jpg b/parse/dev/JprM0p-q0Co/images/18850958e9b722cfb7a14882640fa796ea7c666e8e3da1bff9b490c3361882a5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e613ecb29acc3e0e4a0dbebc431ef61764587bcf --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/18850958e9b722cfb7a14882640fa796ea7c666e8e3da1bff9b490c3361882a5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c2998cfa89305a243d7ee9164c24ac11b7506b591eff2e3c040c5f2f6e4d590a +size 8250 diff --git a/parse/dev/JprM0p-q0Co/images/18c1b88b35731d857d9e4b41ed765c097c87a53d89697eb49a1667b747f3381d.jpg b/parse/dev/JprM0p-q0Co/images/18c1b88b35731d857d9e4b41ed765c097c87a53d89697eb49a1667b747f3381d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..918f367a4751e922d46facf0521aacfa310f1c07 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/18c1b88b35731d857d9e4b41ed765c097c87a53d89697eb49a1667b747f3381d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b6350de16a81b5d4092a4da98f77607cf9866fe638d48973fc57f3ecfcdd5c1 +size 6694 diff --git a/parse/dev/JprM0p-q0Co/images/1bb6291c0269933c0cb38c1c38d82e48f037b74cdc915e08e48010586f0612a9.jpg b/parse/dev/JprM0p-q0Co/images/1bb6291c0269933c0cb38c1c38d82e48f037b74cdc915e08e48010586f0612a9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..435c2ea1be1333adff57abd5f385ecf60bea3c0b --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/1bb6291c0269933c0cb38c1c38d82e48f037b74cdc915e08e48010586f0612a9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:52fe0b737494ccbd99d881035123cd3966fc96ab2ed1b1f9ccca6bdccf05e385 +size 35715 diff --git a/parse/dev/JprM0p-q0Co/images/22793ad6ddc557c48399c000c07d889f5d2caafc0b046a27863cc5981140e3fc.jpg b/parse/dev/JprM0p-q0Co/images/22793ad6ddc557c48399c000c07d889f5d2caafc0b046a27863cc5981140e3fc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..886a1853d4da82953a56fb302bac40bdac97be1a --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/22793ad6ddc557c48399c000c07d889f5d2caafc0b046a27863cc5981140e3fc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4dc59b70ace350d31db5ad891336c6a73e5cbe1a3c2943ab47146512c2a13c82 +size 22196 diff --git a/parse/dev/JprM0p-q0Co/images/2286bf6438ddba04055296741845f08ba6ecb11db64ca52fbcac43ce3bc0c761.jpg b/parse/dev/JprM0p-q0Co/images/2286bf6438ddba04055296741845f08ba6ecb11db64ca52fbcac43ce3bc0c761.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8d6caee18845f0f034ce58c60e406d6667037873 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/2286bf6438ddba04055296741845f08ba6ecb11db64ca52fbcac43ce3bc0c761.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d80e6224158f5812a3fd8029b9e7f264de53f5d6cec191d6badd399936a9d641 +size 250186 diff --git a/parse/dev/JprM0p-q0Co/images/25d855fe3741af4dbfa3f689e0063841cc05f3826353d95d8276afee92e1390f.jpg b/parse/dev/JprM0p-q0Co/images/25d855fe3741af4dbfa3f689e0063841cc05f3826353d95d8276afee92e1390f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9ba812fb77094158e18844897e64ecdeec931229 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/25d855fe3741af4dbfa3f689e0063841cc05f3826353d95d8276afee92e1390f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:016dd7ad7a16b9b5ad11a88ea33e1d2ea16be136edd47a46abd3e295fe322b6a +size 29673 diff --git a/parse/dev/JprM0p-q0Co/images/27c47e1e44020a7bcce80a51bf1b125f4f64e148bda9117b753e51dcf4d5354c.jpg b/parse/dev/JprM0p-q0Co/images/27c47e1e44020a7bcce80a51bf1b125f4f64e148bda9117b753e51dcf4d5354c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4370aff2256b2eba4a3ec2a24b87c63656abe5bf --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/27c47e1e44020a7bcce80a51bf1b125f4f64e148bda9117b753e51dcf4d5354c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b0329b6db9efb437fe5dc05d6132da643f49f99d24f6ef79c1d4a3fe4f367cf +size 4829 diff --git a/parse/dev/JprM0p-q0Co/images/29b2677b73f28886735c1788ab5c4c2975a26358b689704b0eaae553ae9d76f6.jpg b/parse/dev/JprM0p-q0Co/images/29b2677b73f28886735c1788ab5c4c2975a26358b689704b0eaae553ae9d76f6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ce090b6f77cf6f6896ef1fb5545857786ebd2739 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/29b2677b73f28886735c1788ab5c4c2975a26358b689704b0eaae553ae9d76f6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2fad4a316d73435d98f8581f4a16dd06bda937f05d13d43418790d4a47051806 +size 11109 diff --git a/parse/dev/JprM0p-q0Co/images/423ceff83912d3a4ed91fc4f500cf665922d67b1950f2741da4398e027982b41.jpg b/parse/dev/JprM0p-q0Co/images/423ceff83912d3a4ed91fc4f500cf665922d67b1950f2741da4398e027982b41.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f54c8f6609921832a1a884a5357576bf85d90592 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/423ceff83912d3a4ed91fc4f500cf665922d67b1950f2741da4398e027982b41.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d6a1f1aa3920b47c87370dceceb1d021d5d0611dd55900d329077b7f438d453 +size 11988 diff --git a/parse/dev/JprM0p-q0Co/images/59609abd441c3da356ea55a886e0b1eba63cbadd994f8d41d3c454cd5d470c62.jpg b/parse/dev/JprM0p-q0Co/images/59609abd441c3da356ea55a886e0b1eba63cbadd994f8d41d3c454cd5d470c62.jpg new file mode 100644 index 0000000000000000000000000000000000000000..523bb6c0aa46b64427a641d11e653302af103709 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/59609abd441c3da356ea55a886e0b1eba63cbadd994f8d41d3c454cd5d470c62.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b95f675d872aa08c340c8225e1147e1f3693835678f16bf58d6f5b7e3376b911 +size 7176 diff --git a/parse/dev/JprM0p-q0Co/images/5d3bb1ed92703ec8f2daee3131b59a437597b55d196606d67110e5feb8f1ae2f.jpg b/parse/dev/JprM0p-q0Co/images/5d3bb1ed92703ec8f2daee3131b59a437597b55d196606d67110e5feb8f1ae2f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6ff541c3c9cd4e052ac54e9f41773f87ce14026f --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/5d3bb1ed92703ec8f2daee3131b59a437597b55d196606d67110e5feb8f1ae2f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c5e65777860b2baec265e5c409de5ff60db846f66ad0487612a396d622ae416 +size 9306 diff --git a/parse/dev/JprM0p-q0Co/images/5fc515c2d57acb32a17b863956f9dbb507835e258655059683edfac950bec81a.jpg b/parse/dev/JprM0p-q0Co/images/5fc515c2d57acb32a17b863956f9dbb507835e258655059683edfac950bec81a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..164307d3472b5fa6dc6fd38ea25c065290f5bf4e --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/5fc515c2d57acb32a17b863956f9dbb507835e258655059683edfac950bec81a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:369c9c929cf3dbdb1048c2973fe2a9d0b2a90a3cd50f842bcf64f6043e774e4c +size 38587 diff --git a/parse/dev/JprM0p-q0Co/images/6169f1959b255f0af726deed22981091b05bfb930e2b4e8de2a63c5c4867095a.jpg b/parse/dev/JprM0p-q0Co/images/6169f1959b255f0af726deed22981091b05bfb930e2b4e8de2a63c5c4867095a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4330a0caa5c42392306e269f180f0e7a6cfcb098 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/6169f1959b255f0af726deed22981091b05bfb930e2b4e8de2a63c5c4867095a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2abcc53af25e1bcf9fa13502d9e7fe8de2fe0b2b2a56d5015869c800d5ceaa72 +size 245799 diff --git a/parse/dev/JprM0p-q0Co/images/64fe4e6bd071a944679fa7909742ea86b790524d906bd440c2591b41657877d1.jpg b/parse/dev/JprM0p-q0Co/images/64fe4e6bd071a944679fa7909742ea86b790524d906bd440c2591b41657877d1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f232b22eaebd3d3df15154278305a14baea5b629 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/64fe4e6bd071a944679fa7909742ea86b790524d906bd440c2591b41657877d1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:137f706fe5968702b921480b094ecafeaf5809a67621342534f274abd9c14a1b +size 34864 diff --git a/parse/dev/JprM0p-q0Co/images/6a20c924d47a2d94dfa7bcf92bb736f162630e21a206cf929f5dea20dcf355d4.jpg b/parse/dev/JprM0p-q0Co/images/6a20c924d47a2d94dfa7bcf92bb736f162630e21a206cf929f5dea20dcf355d4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1ae291dfe815d0a513a76c38fe059936c9205c6e --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/6a20c924d47a2d94dfa7bcf92bb736f162630e21a206cf929f5dea20dcf355d4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1e9ac90174eff156eed8f01273cb5999cf828f62bb4bb0e1dd146b838d5cf9e +size 9016 diff --git a/parse/dev/JprM0p-q0Co/images/6b09c473c3d632932e4dc875b2a761d8347abed8c5df96180dace288cf111265.jpg b/parse/dev/JprM0p-q0Co/images/6b09c473c3d632932e4dc875b2a761d8347abed8c5df96180dace288cf111265.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a4ce645ee8bfe23acbad22bf4dcd9e30415612a6 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/6b09c473c3d632932e4dc875b2a761d8347abed8c5df96180dace288cf111265.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91372f14e9130450c02d7aed1dfa9b440a40db21ae911f047b5bd729eb88c104 +size 272028 diff --git a/parse/dev/JprM0p-q0Co/images/6da7b35299a1ffae82b84707c228d91064e1de614083fbcca7755a2d8dbda68e.jpg b/parse/dev/JprM0p-q0Co/images/6da7b35299a1ffae82b84707c228d91064e1de614083fbcca7755a2d8dbda68e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c6f393dea8d9a1e141f42599c131f6ba09705290 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/6da7b35299a1ffae82b84707c228d91064e1de614083fbcca7755a2d8dbda68e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dfef6e29d1e8e7ca032d13f7e0bb1bd837a842cd8c279d240c9ae74cf8b4839d +size 39385 diff --git a/parse/dev/JprM0p-q0Co/images/7681f20791a11379285a2b054253347a96f5a0d0361f9d1a92d641bec7ae1bd1.jpg b/parse/dev/JprM0p-q0Co/images/7681f20791a11379285a2b054253347a96f5a0d0361f9d1a92d641bec7ae1bd1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..baafce0f6edc688385b64d6474db41b81c88b541 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/7681f20791a11379285a2b054253347a96f5a0d0361f9d1a92d641bec7ae1bd1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:94da5c352d42ad14f66bdf8b22f81eb05216c10cf72eef0b693a654d0d2cc690 +size 9273 diff --git a/parse/dev/JprM0p-q0Co/images/79232f4d96e5d614c68595063e6e236a7f91631adb499901e0dcbae244cd416b.jpg b/parse/dev/JprM0p-q0Co/images/79232f4d96e5d614c68595063e6e236a7f91631adb499901e0dcbae244cd416b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..aacba098a6c4e6f18d10158ce23878300e5c35c6 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/79232f4d96e5d614c68595063e6e236a7f91631adb499901e0dcbae244cd416b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:64e84dcfbf1eee2da77429a52fa7f18527ade648832b2fa0060febe4e9f1131b +size 41399 diff --git a/parse/dev/JprM0p-q0Co/images/7a9d5080a43d1243414c65373f0a7010bb11108d537987841b8ec6716e179796.jpg b/parse/dev/JprM0p-q0Co/images/7a9d5080a43d1243414c65373f0a7010bb11108d537987841b8ec6716e179796.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6d493d4aa9f5f637f898af89f664001dedf0a0c9 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/7a9d5080a43d1243414c65373f0a7010bb11108d537987841b8ec6716e179796.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1bedac28bb500c9b6091c1f274d4fce9158f719995698a230469406f7ea9e34c +size 10528 diff --git a/parse/dev/JprM0p-q0Co/images/7db5de04593da86e2a7700fbb7325e30d9c8897ba3f9154c4416821b9586d5ed.jpg b/parse/dev/JprM0p-q0Co/images/7db5de04593da86e2a7700fbb7325e30d9c8897ba3f9154c4416821b9586d5ed.jpg new file mode 100644 index 0000000000000000000000000000000000000000..68b9b36bee71dae37b8717f54f9e089c4aad15cd --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/7db5de04593da86e2a7700fbb7325e30d9c8897ba3f9154c4416821b9586d5ed.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee04e8ec717995616beb2d475fddb306c847ada8423797f8add00d9f064d384d +size 336060 diff --git a/parse/dev/JprM0p-q0Co/images/845f116ad2b98bb73fd2e21146fcc7c3405a626dcd21e38e48e74061c57c4609.jpg b/parse/dev/JprM0p-q0Co/images/845f116ad2b98bb73fd2e21146fcc7c3405a626dcd21e38e48e74061c57c4609.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6949a0861b27ff98ba515eb84b25134ee0bea69e --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/845f116ad2b98bb73fd2e21146fcc7c3405a626dcd21e38e48e74061c57c4609.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2efdadb90fc2d14936937c39b68c7da6412d118a50ccc46f6ad67eb857cd0dd9 +size 44427 diff --git a/parse/dev/JprM0p-q0Co/images/857372cbbb19ed4744f6e6376cb69f65a352e21e75f5cb12b6aa519281e0f549.jpg b/parse/dev/JprM0p-q0Co/images/857372cbbb19ed4744f6e6376cb69f65a352e21e75f5cb12b6aa519281e0f549.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2b55b794d893590a09e10dc19f76e56cd7b20885 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/857372cbbb19ed4744f6e6376cb69f65a352e21e75f5cb12b6aa519281e0f549.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe89b03dae9f4b263fef9b526f713baeb2e1f9a3acf904a3eeb5bf43101fc5aa +size 199596 diff --git a/parse/dev/JprM0p-q0Co/images/892c8380ccc4102e6d67c720b451efe6e37a8e0edbc50c38d231d40549fdf6be.jpg b/parse/dev/JprM0p-q0Co/images/892c8380ccc4102e6d67c720b451efe6e37a8e0edbc50c38d231d40549fdf6be.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5de0c0eb396d245280d59585309b367b2532cb55 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/892c8380ccc4102e6d67c720b451efe6e37a8e0edbc50c38d231d40549fdf6be.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b812d1951acc3a6b9b49c5b4c898c11593115364751224b3fd0b6b97c6793c2 +size 47880 diff --git a/parse/dev/JprM0p-q0Co/images/8cd1617eb2ccc30683948eaa03c054b5effcb93330724077d6ac1326ba04b356.jpg b/parse/dev/JprM0p-q0Co/images/8cd1617eb2ccc30683948eaa03c054b5effcb93330724077d6ac1326ba04b356.jpg new file mode 100644 index 0000000000000000000000000000000000000000..30a0628791ae782f97fad702421f69632b595bae --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/8cd1617eb2ccc30683948eaa03c054b5effcb93330724077d6ac1326ba04b356.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be3ec799a4f8d090c6ac7bf3b9fdda69a58c1f04217455660180b3cec5489d5a +size 37817 diff --git a/parse/dev/JprM0p-q0Co/images/93faa9ece25fd6e1b91d43462bfac94ba950757f4581965b28f4042d9087cfa2.jpg b/parse/dev/JprM0p-q0Co/images/93faa9ece25fd6e1b91d43462bfac94ba950757f4581965b28f4042d9087cfa2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..41a0642b7bef32ca6cdda4de2cd4b0d6c396f8e4 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/93faa9ece25fd6e1b91d43462bfac94ba950757f4581965b28f4042d9087cfa2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f8bfb602f1ef62f0aa08823ff33e26c4b5745cf5689c3ff20e2d7c92d4e7993 +size 29395 diff --git a/parse/dev/JprM0p-q0Co/images/9a853c9907b0b4ffca4ac8add7de01e2609d5475a867e464dfb438d4e0bee0b5.jpg b/parse/dev/JprM0p-q0Co/images/9a853c9907b0b4ffca4ac8add7de01e2609d5475a867e464dfb438d4e0bee0b5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4ff3b83a1efa956b7cdea64541eee89ccf2d1653 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/9a853c9907b0b4ffca4ac8add7de01e2609d5475a867e464dfb438d4e0bee0b5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b818941181fb16a0f0a446105c0a680735666248550a319ce2c2e21ffc80f97f +size 28823 diff --git a/parse/dev/JprM0p-q0Co/images/a116958748a4fe3cb0f5104e44075967bc6f7be0d418cc15a6f72f17988dc914.jpg b/parse/dev/JprM0p-q0Co/images/a116958748a4fe3cb0f5104e44075967bc6f7be0d418cc15a6f72f17988dc914.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d607f579789d3d4c42343d6ccb5accbbd31ce175 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/a116958748a4fe3cb0f5104e44075967bc6f7be0d418cc15a6f72f17988dc914.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0183682afc55d00c5ae9596935997aa26af28ef345e163f6514338ab820b9254 +size 21926 diff --git a/parse/dev/JprM0p-q0Co/images/af5a1cf4344803932c24fc6b7ca4545014964ce23eb0caf6f3775a4ea50603b8.jpg b/parse/dev/JprM0p-q0Co/images/af5a1cf4344803932c24fc6b7ca4545014964ce23eb0caf6f3775a4ea50603b8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b5c064bd87e34c66edb4fc6f7f375426e4bbc736 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/af5a1cf4344803932c24fc6b7ca4545014964ce23eb0caf6f3775a4ea50603b8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5178fa988e257b8ed0fc200f2e44345eb74216073ac3dfff226bf694b8e4ad19 +size 10231 diff --git a/parse/dev/JprM0p-q0Co/images/b84a997eab30f2f0979dc36bbb3ef8ad0714aad662cc20587115f10d50469bec.jpg b/parse/dev/JprM0p-q0Co/images/b84a997eab30f2f0979dc36bbb3ef8ad0714aad662cc20587115f10d50469bec.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d6c1ea1f865804e26ac2628e9be04be1c7ca8e40 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/b84a997eab30f2f0979dc36bbb3ef8ad0714aad662cc20587115f10d50469bec.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8afcb8582e5021e9702a041e2d399772615d8a60bd853f26c3482076713cdb14 +size 7532 diff --git a/parse/dev/JprM0p-q0Co/images/c0abf2664971a12b945104d8691e0bb9509c3f46da0a605f5af0b8e440d974ef.jpg b/parse/dev/JprM0p-q0Co/images/c0abf2664971a12b945104d8691e0bb9509c3f46da0a605f5af0b8e440d974ef.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a4d6c88f0e6e25a4bc458b18572f599922761cab --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/c0abf2664971a12b945104d8691e0bb9509c3f46da0a605f5af0b8e440d974ef.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3c66b00fb89f7718319502579dab4c66fdea0133bc9ea64177b2781b1ce926e +size 210801 diff --git a/parse/dev/JprM0p-q0Co/images/c0e312d7030cda214a9f37a93e79138a391c205f22f1cc6648d949b71a4e4c85.jpg b/parse/dev/JprM0p-q0Co/images/c0e312d7030cda214a9f37a93e79138a391c205f22f1cc6648d949b71a4e4c85.jpg new file mode 100644 index 0000000000000000000000000000000000000000..16d6a1d11ad28db767a9af150e13f93be36f3f54 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/c0e312d7030cda214a9f37a93e79138a391c205f22f1cc6648d949b71a4e4c85.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59a4a1f554d2fa948900ddc1bfdc302d08149b82c9ee01c5b48bf436e9a999bd +size 28932 diff --git a/parse/dev/JprM0p-q0Co/images/c20d251bd704c260ecf8cbd4384425a05cea94ff0ce3bfbbc6265c0474c02120.jpg b/parse/dev/JprM0p-q0Co/images/c20d251bd704c260ecf8cbd4384425a05cea94ff0ce3bfbbc6265c0474c02120.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c0bae7d6867a7cb9f37dbc7c82f4e17bcbf4285a --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/c20d251bd704c260ecf8cbd4384425a05cea94ff0ce3bfbbc6265c0474c02120.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:40bed9d98d851dd5a756418ac654419a76c2f09a15788a813185b2534e479063 +size 9795 diff --git a/parse/dev/JprM0p-q0Co/images/cfa8b199b3af3ade5cb9c0a94fc3c092f9063bfdac677cacfccfddbc988718e2.jpg b/parse/dev/JprM0p-q0Co/images/cfa8b199b3af3ade5cb9c0a94fc3c092f9063bfdac677cacfccfddbc988718e2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f3dc29dc8fd8378a22dffe3dde7a6438b55e49da --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/cfa8b199b3af3ade5cb9c0a94fc3c092f9063bfdac677cacfccfddbc988718e2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e4565234de098ab676125ba1917eab6a56f3473912a10bc87797d4a5a1132fa2 +size 23379 diff --git a/parse/dev/JprM0p-q0Co/images/d1259007a02273d2832e3080369c17cb598b5e903d988a81282b4aa455229642.jpg b/parse/dev/JprM0p-q0Co/images/d1259007a02273d2832e3080369c17cb598b5e903d988a81282b4aa455229642.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f52cde9c21e2281abb92c977bd0b453ff8704f2b --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/d1259007a02273d2832e3080369c17cb598b5e903d988a81282b4aa455229642.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d9f70bc18d5f836a6027dfd14fc5dbb302b0f5d58a2520646470d4832738e6a +size 11901 diff --git a/parse/dev/JprM0p-q0Co/images/d87d38b6bd780d55bc10c9a6d92d01216eeb26b5e4c7010f6195e2b1888b20d5.jpg b/parse/dev/JprM0p-q0Co/images/d87d38b6bd780d55bc10c9a6d92d01216eeb26b5e4c7010f6195e2b1888b20d5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ef0819bf6ec12e9c5d9b2aada63309e90d116051 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/d87d38b6bd780d55bc10c9a6d92d01216eeb26b5e4c7010f6195e2b1888b20d5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8a45838779a5a5a23fd09bf5a3e4c40ada0209592464763314d4ca451b70a4e7 +size 11893 diff --git a/parse/dev/JprM0p-q0Co/images/da78922349e342b885df023c6b9fbd63dc2aa86e9041b4eb70c4ba742ef33c3a.jpg b/parse/dev/JprM0p-q0Co/images/da78922349e342b885df023c6b9fbd63dc2aa86e9041b4eb70c4ba742ef33c3a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..33b271b1912e18b5cca876177450d26331747270 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/da78922349e342b885df023c6b9fbd63dc2aa86e9041b4eb70c4ba742ef33c3a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:492f3b3796521acbf3a14fb5429e7bb811fab3bc808819a42f36c842e8163424 +size 182272 diff --git a/parse/dev/JprM0p-q0Co/images/def1fdb167c8133456452f25a3f5a8a30a46fd37d9657f0f2afe8b94aacf6fc6.jpg b/parse/dev/JprM0p-q0Co/images/def1fdb167c8133456452f25a3f5a8a30a46fd37d9657f0f2afe8b94aacf6fc6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4ff898718a00c3b89fd7e9532a37e87f1c6a10f2 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/def1fdb167c8133456452f25a3f5a8a30a46fd37d9657f0f2afe8b94aacf6fc6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44a45a7e714d25c20acd9fbe87c4491853115c766e938c3b2aa47614cdcd49be +size 39369 diff --git a/parse/dev/JprM0p-q0Co/images/df7021cb38fb1c017cc830738c8aa9a2adaf821b22417fd3b0bc27c0e4c51bf5.jpg b/parse/dev/JprM0p-q0Co/images/df7021cb38fb1c017cc830738c8aa9a2adaf821b22417fd3b0bc27c0e4c51bf5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9e084be6a022ab40f196b74f478960ae8c760cb2 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/df7021cb38fb1c017cc830738c8aa9a2adaf821b22417fd3b0bc27c0e4c51bf5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cec4cfcaa00bcefe2ccf97c825423ff18ebcf4e5c5f95da6499214ac4ba4b5b1 +size 46043 diff --git a/parse/dev/JprM0p-q0Co/images/e0563cc8f7c6261f6874db0dde4fc179497e65f711c7e609e22a9229ed7e31b0.jpg b/parse/dev/JprM0p-q0Co/images/e0563cc8f7c6261f6874db0dde4fc179497e65f711c7e609e22a9229ed7e31b0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fbd08bc1832f4a2a83a1277ba8f3760ef062fca1 --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/e0563cc8f7c6261f6874db0dde4fc179497e65f711c7e609e22a9229ed7e31b0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5bad630ac168a1d23dc46d80d67bee14a459319c8dc75d500b37efba01e07e0c +size 4070 diff --git a/parse/dev/JprM0p-q0Co/images/eaaa63a6c2932d48732a9f9ce4418e72b53237cb76e0ad136214e3974382a315.jpg b/parse/dev/JprM0p-q0Co/images/eaaa63a6c2932d48732a9f9ce4418e72b53237cb76e0ad136214e3974382a315.jpg new file mode 100644 index 0000000000000000000000000000000000000000..69ed3744f2430a24bd81109139431fecb8f7bdbe --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/eaaa63a6c2932d48732a9f9ce4418e72b53237cb76e0ad136214e3974382a315.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79ebde6501ec3dab89a2d1e7bcd70a58e2a10a1ea0d981c8fd96b50ae0d5e12d +size 4822 diff --git a/parse/dev/JprM0p-q0Co/images/f3e4bef2107cb261b5f213ba36a5c90ab70ed2176ef807e4e54dd4033fcc9ed8.jpg b/parse/dev/JprM0p-q0Co/images/f3e4bef2107cb261b5f213ba36a5c90ab70ed2176ef807e4e54dd4033fcc9ed8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..96f2d34fff448ef46d98abd19c1346ebaf408bac --- /dev/null +++ b/parse/dev/JprM0p-q0Co/images/f3e4bef2107cb261b5f213ba36a5c90ab70ed2176ef807e4e54dd4033fcc9ed8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:33479440ff9b827695c6761ef42872c017bb0a12ad8c3a1ca3cd66a6e975239b +size 3057 diff --git a/parse/dev/K2OixmPDou3/K2OixmPDou3_content_list.json b/parse/dev/K2OixmPDou3/K2OixmPDou3_content_list.json new file mode 100644 index 0000000000000000000000000000000000000000..1cc5c675264abc1a43d955dd04af55c99d159189 --- /dev/null +++ b/parse/dev/K2OixmPDou3/K2OixmPDou3_content_list.json @@ -0,0 +1,2770 @@ +[ + { + "type": "text", + "text": "UNLEASHING MASK: EXPLORE THE INTRINSIC OUTOF-DISTRIBUTION DETECTION CAPABILITY ", + "text_level": 1, + "bbox": [ + 176, + 98, + 826, + 146 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Anonymous authors Paper under double-blind review ", + "bbox": [ + 183, + 170, + 398, + 198 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "ABSTRACT ", + "text_level": 1, + "bbox": [ + 454, + 234, + 544, + 251 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Out-of-distribution (OOD) detection is an important aspect of safely deploying machine learning models in real-world applications. Previous approaches either design better scoring functions or utilize the knowledge of outliers to equip the well-trained models with the ability of OOD detection. However, few of them explore to excavate the intrinsic OOD detection capability of a given model. In this work, we discover the existence of an intermediate stage of a model trained on in-distribution data having higher OOD detection performance than that of its final stage across different settings and further identify the critical attribution to be learning with atypical samples. Based on such empirical insights, we propose a new method, Unleashing Mask (UM), that restores the OOD discriminative capabilities of the model. To be specific, we utilize the mask to figure out the memorized atypical samples and fine-tune the model to forget them. Extensive experiments have been conducted to characterize and verify the effectiveness of our method. ", + "bbox": [ + 233, + 270, + 764, + 450 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "1 INTRODUCTION ", + "text_level": 1, + "bbox": [ + 176, + 486, + 334, + 502 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Out-of-distribution (OOD) detection has drawn increasing attention when deploying machine learning models into the open-world scenarios (Nguyen et al., 2015; Lee et al., 2018a). Since the test samples can naturally arise from a label-different distribution, identifying OOD inputs is important, especially for those safety-critical applications like autonomous driving and medical intelligence. Previous studies focus on designing a series of scoring functions (Hendrycks & Gimpel, 2017b; Liang et al., 2018; Lee et al., 2018a; Liu et al., 2020; Sun et al., 2021; 2022) for OOD uncertainty estimation or finetuning with auxiliary outlier data to better distinguish the OOD inputs (Hendrycks et al., $2 0 1 9 \\mathrm { c }$ ; Tack et al., 2020; Mohseni et al., 2020; Sehwag et al., 2021; Wei et al., 2022; Ming et al., 2022). ", + "bbox": [ + 174, + 520, + 825, + 632 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Despite the promising results achieved by previous methods (Hendrycks & Gimpel, 2017a; Hendrycks et al., $2 0 1 9 \\mathrm { c }$ ; Liu et al., 2020; Ming et al., 2022), little attention is paid to considering whether the well-trained given model is the most appropriate for OOD detection. In general, models deployed for various applications have different targets (e.g., multi-class classification) (Goodfellow et al., 2016) instead of OOD detection (Nguyen et al., 2015; Lee et al., 2018a). However, most representative score functions, e.g., MSP (Hendrycks et al., 2019c), ODIN(Liang et al., 2018), and Energy (Liu et al., 2020), uniformly leverage the given models for OOD detection. Considering the target-oriented discrepancy, it arises a critical question: does the well-trained given model have the optimal OOD detection capability? If not, how can we find a more appropriate model for OOD detection? ", + "bbox": [ + 174, + 638, + 825, + 763 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "In this work, we start by revealing an important observation (as illustrated in Figure 1), i.e., there exists a historical training stage where the model has a higher OOD detection performance than the final well-trained one. This is generally true across different OOD/ID datasets (Netzer et al., 2011; Van Horn et al., 2018; Cimpoi et al., 2014), learning rate schedules (Loshchilov & Hutter, 2017), and model structures (Huang et al., 2017; Zagoruyko & Komodakis, 2016). The empirical results of Figure 1 reflect the inconsistency between gaining better OOD detection capability (Nguyen et al., 2015) and pursuing better performance on ID data. We delve into the differences between the intermediate model and the final model by visualizing the misclassified examples. As shown in Figure 2, one possible attribution for covering the detection capability should be memorizing the atypical samples (at the semantic level) that are hard to learn for the model. Seeking zero error on those samples makes the model more confident on OOD data (see Figures 1(b) and 1(c)). ", + "bbox": [ + 174, + 771, + 825, + 924 + ], + "page_idx": 0 + }, + { + "type": "image", + "img_path": "images/0999e8dc8619d56385f0f50f3492c4db278ff3ca6514f7f0f6a51a3233af7890.jpg", + "image_caption": [ + "Figure 1: (a) the curves of FPR95 (false positive rate of OOD examples when the true positive rate of in-distribution examples is at $9 5 \\%$ ) based on the Energy score (Liu et al., 2020) across three different OOD datasets during the training on the CIFAR-10 dataset. (b) comparison between ID and OOD distribution at Epoch 60. (c) comparison between ID and OOD distribution at Epoch 100. All the experiments testing for OOD detection performance have been conducted multiple times. By backtracking the training phase, we can observe the existence of the model stage with better OOD detection capability using the Energy score to distinguish the OOD inputs. When zooming in the ID and OOD distributions at Epoch 60 and Epoch 100 respectively, it can be seen that, along with the training at the later stage, the overlap between them grows. Figure 2 contains further exploration. " + ], + "image_footnote": [], + "bbox": [ + 181, + 106, + 810, + 257 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "The above analysis inspires us to propose a new strategy, namely, Unleashing Mask (UM), to excavate the once-covered detection capability of a well-trained given model by alleviating the memorization of those atypical samples (as illustrated in Figure 3) of ID data. In general, we aim to backtrack its previous stage with a better OOD detection capability. To achieve this target, there are two essential issues: (1) the model that is well-trained on ID data has already memorized some atypical samples; (2) how to forget those memorized atypical samples considering the given model? Accordingly, our proposed UM contains two parts utilizing different insights to address the above problems. First, as atypical samples are more sensitive to the change of model parameters, we initialize a mask with the specific cutting rate to mine these samples with constructed discrepancy. Then, with the loss reference estimated by the mask, we conduct the constrained gradient ascent (i.e., Eq. 3) for model forgetting. It will encourage the model to finally stabilize around the optimal stage. To avoid the severe sacrifice of the original task performance on ID data, we further propose UM Adopts Pruning (UMAP) which performs the tuning on the introduced mask with the newly designed objective. ", + "bbox": [ + 173, + 425, + 825, + 606 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "For our proposed methods, we conduct extensive experiments to characterize and understand the working mechanism (in Section 4 and Appendix F). The comprehensive results accordingly demonstrate their effectiveness. We have verified the effectiveness of UM with a series of OOD detection benchmarks considering the two different ID datasets, i.e., CIFAR-10 and CIFAR-100. Under the various evaluation metrics, our UM, as well as UMAP, can indeed excavate the better OOD detection capability of given models and the averaged FPR95 can be reduced by a significant margin. Finally, a range of ablation studies and further discussions related to our proposed strategy are provided. We summarize our main contributions as follows, ", + "bbox": [ + 174, + 613, + 825, + 724 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "• Conceptually, we explore the OOD detection performance via a new perspective, i.e., backtracking the model training phase without regularizing by any auxiliary outliers, different from most previous works that start with the well-trained model on ID data. Empirically, we reveal the potential detection capability of the well-trained model. We observe the general existence of an intermediate stage where the model has more appropriate discriminative features that can be utilized for OOD detection. Technically, we introduce a new strategy, i.e., Unleashing Mask, to excavate the once-covered OOD detection capability of a given model. By introducing the mask, we estimate the loss constraint for forgetting the atypical samples and empower the detection performance. Experimentally, we conduct extensive explorations to verify the general effectiveness on improving the OOD detection performance of our methods. Using various ID and OOD benchmarks, we provide comprehensive results across different setups and further discussion. ", + "bbox": [ + 217, + 737, + 825, + 924 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "2 BACKGROUND ", + "text_level": 1, + "bbox": [ + 176, + 102, + 326, + 118 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "In this section, we briefly introduce the preliminaries and related work about OOD detection. ", + "bbox": [ + 169, + 133, + 781, + 148 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "2.1 PRELIMINARIES ", + "text_level": 1, + "bbox": [ + 174, + 167, + 326, + 183 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "We consider multi-class classification as the training task, where $\\mathcal { X } \\subset \\mathbb { R } ^ { d }$ denotes the input space and $\\mathcal { Y } = \\{ 1 , \\ldots , C \\}$ denotes the label space. Given the model deployed in the real world, the reliable classifier is expected to figure out the OOD input, which can be formulated as a binary classification problem. Given $\\mathcal { P }$ , the distribution over $\\mathcal { X } \\times \\mathcal { V }$ , we consider $D _ { \\mathrm { i n } }$ as the marginal distribution of $\\mathcal { P }$ for $\\mathcal { X }$ , namely, the distribution of ID data. At test time, the environment can present a distribution $D _ { \\mathrm { o u t } }$ over $\\mathcal { X }$ of OOD data. In general, the OOD distribution $D _ { \\mathrm { o u t } }$ is defined as an irrelevant distribution of which the label set has no intersection with $\\mathcal { V }$ Yang et al. (2021) and therefore should not be predicted by the model. The decision can be made with the threshold $\\lambda$ : ", + "bbox": [ + 173, + 194, + 825, + 306 + ], + "page_idx": 2 + }, + { + "type": "equation", + "img_path": "images/e87448befed0059cc41d5d7312eecc2686852e4d9c83a86ca7d1674ebc8041fd.jpg", + "text": "$$\nD _ { \\lambda } ( x ; f ) = { \\left\\{ \\begin{array} { l l } { \\mathrm { I D } } & { S ( x ) \\geq \\lambda } \\\\ { \\mathrm { O O D } } & { S ( x ) < \\lambda ^ { \\prime } } \\end{array} \\right. }\n$$", + "text_format": "latex", + "bbox": [ + 387, + 313, + 607, + 349 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Building upon the model $f \\in \\mathcal { H } : \\mathcal { X } \\to \\mathbb { R } ^ { c }$ trained on ID data with the logit outputs, the goal of decision is to utilize the scoring function $S : \\mathcal { X } \\mathbb { R }$ to distinguish the inputs of $D _ { \\mathrm { i n } }$ from that of $D _ { \\mathrm { o u t } }$ by $S ( x )$ . Typically, if the score value is larger than the threshold $\\lambda$ , the associated input $x$ is classified as ID and vice versa. We consider several representative scoring functions designed for OOD detection, e.g., MSP (Hendrycks & Gimpel, 2017b), ODIN (Liang et al., 2018), and Energy (Liu et al., 2020). More detailed definitions and implementation can refer to Appendix A. ", + "bbox": [ + 173, + 356, + 825, + 441 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "To mitigate the issue of over-confident predictions for (Hendrycks & Gimpel, 2017b; Liu et al., 2020) some OOD data, recent works (Hendrycks et al., $2 0 1 9 \\mathrm { c }$ ; Tack et al., 2020) utilize the auxiliary unlabeled dataset to regularize the model behavior. Among them, one representative baseline is outlier exposure (OE) (Hendrycks et al., 2019c). OE can further improve the detection performance by making the model $f ( \\cdot )$ finetuned from a surrogate OOD distribution $D _ { \\mathrm { o u t } } ^ { \\mathrm { s } }$ , and its corresponding learning objective is defined as follows, ", + "bbox": [ + 174, + 446, + 825, + 531 + ], + "page_idx": 2 + }, + { + "type": "equation", + "img_path": "images/ebe06b3345ccb0c28753e1d45a18df14d8791140aca800b8ba6c4a50bc0fa038.jpg", + "text": "$$\n\\mathcal { L } _ { f } = \\mathbb { E } _ { D _ { \\mathrm { i n } } } \\left[ \\ell _ { \\mathrm { C E } } ( f ( x ) , y ) \\right] + \\lambda \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } \\left[ \\ell _ { \\mathrm { O E } } ( f ( x ) ) \\right] ,\n$$", + "text_format": "latex", + "bbox": [ + 339, + 540, + 656, + 558 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "where $\\lambda$ is the balancing parameter, $\\ell _ { \\mathrm { C E } } ( \\cdot )$ is the Cross-Entropy (CE) loss, and $\\ell _ { \\mathrm { O E } } ( \\cdot )$ is the Kullback-Leibler divergence to the uniform distribution, which can be written as $\\ell _ { { \\bf { O E } } } ( h ( { \\bf { x } } ) ) =$ $- \\sum k$ softmaxk $f ( x ) { \\bar { / C } }$ , where $\\mathtt { s o f t m a x } _ { k } ( \\cdot )$ denotes the $k$ -th element of a softmax output. The OE loss $\\ell _ { \\mathrm { O E } } ( \\cdot )$ is designed for model regularization, making the model learn from surrogate OOD inputs to return low-confident predictions. The general formulation of $\\operatorname { E q } 2$ is also adopted in other related works for designing better tuning objectives that use different auxiliary outlier data. ", + "bbox": [ + 173, + 564, + 825, + 650 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Although previous works show promising results via designing scoring functions or regularizing models based on the model $f$ trained on ID data, few of them investigated the original detection capability of the well-trained given model. In this work, we introduce the layer-wise mask $m$ (Han et al., 2016; Ramanujan et al., 2020) to mine the atypical samples that memorized by the model. Accordingly, the decision can be written as $D ( x ; m \\odot f )$ , and the output of masked model is $m \\odot f ( x )$ . ", + "bbox": [ + 174, + 655, + 825, + 727 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "2.2 RELATED WORK ", + "text_level": 1, + "bbox": [ + 174, + 744, + 330, + 758 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Out-of-distribution Detection without auxiliary data. (Hendrycks & Gimpel, 2017a) formally shed light on out-of-distribution detection, proposing to use softmax prediction probability as a baseline which is demonstrated to be unsuitable for OOD detection. Subsequent works keep focusing on designing post-hoc metrics to distinguish ID samples from OOD samples, among which ODIN (Liang et al., 2018) introduces small perturbations into input images to facilitate the separation of softmax score, Mahalanobis distance-based confidence score (Lee et al., 2018b) exploits the feature space by obtaining conditional Gaussian distributions, energy-based score (Liu et al., 2020) aligns better with the probability density. Besides designing score functions, many other works pay attention to various aspects to enhance the OOD detection such that LogitNorm (Wei et al., 2022) produces confidence scores by training with a constant vector norm on the logits, and DICE (Sun & Li, 2022) reduces the variance of the output distribution by leveraging the sparsification of the model. ", + "bbox": [ + 173, + 770, + 825, + 924 + ], + "page_idx": 2 + }, + { + "type": "image", + "img_path": "images/3eb50f356c32cc81abd5826ebf220eef6e801f6950413747092d4d1cf0fbe6f9.jpg", + "image_caption": [ + "Figure 2: We train on CIFAR-10 for the original multi-class classification and check the details as follows: (a) training/testing loss on ID data; (b) training/testing accuracy on ID data; (c) curves of FPR95 based on ODIN score; (d) curves of FPR95 based on Energy score; (e) visualization of correct classified samples at Epoch 60; (f) visualization of misclassified samples at Epoch 60. We investigate the once-covered OOD detection capability by checking the model behavior during its training phase. We take a closer look at the corresponding training and testing performance with the OOD detection capability indicated by two different scores. Through comparison, we find that achieving a reasonably small loss value (at round Epoch 60) on ID data is enough for OOD detection. However, continually optimizing on those atypical samples can impair the detection performance. " + ], + "image_footnote": [], + "bbox": [ + 181, + 104, + 813, + 353 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "Out-of-distribution Detection with auxiliary data. Another promising direction towards OOD detection involves the auxiliary outliers for model regularization. On one hand, some works generate virtual outliers such that Lee et al. (2018a) uses generative adversarial networks to generate boundary samples, VOS (Du et al., 2022a) regularizes the decision boundary by adaptively sampling virtual outliers from the low-likelihood region. On the other hand, other works tend to exploit information from natural outliers, such that outlier exposure (OE) is introduced by Hendrycks et al. (2019b), given that diverse data are available in enormous quantities. (Yu & Aizawa, 2019) train an additional \"head\" and maximizes the discrepancy of decision boundaries of the two heads to detect OOD samples. Energy-bounded learning (Liu et al., 2020) fine-tunes the neural network to widen the energy gap by adding an energy loss term to the objective. Some other works also highlight the sampling strategy, such that ATOM (Chen et al., 2021) greedily utilizes informative auxiliary data to tighten the decision boundary for OOD detection, and POEM (Ming et al., 2022) adopts Thompson sampling to contour the decision boundary precisely. The performance of training with outliers is usually superior to that without outliers, shown in many recent works (Mohseni et al., 2020; Liu et al., 2020; Fort et al., 2021; Sun et al., 2021; Sehwag et al., 2021; Yang et al., 2021; Chen et al., 2021; Salehi et al., 2021). ", + "bbox": [ + 174, + 502, + 825, + 710 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "3 PROPOSED METHOD: UNLEASHING MASK ", + "text_level": 1, + "bbox": [ + 174, + 733, + 560, + 750 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "In this section, we introduce our new method, i.e., Unleashing Mask (UM), to reveal the potential OOD detection capability of the well-trained model. First, we present and discuss the important observation that inspires our methods (Section 3.1). Second, we provide the insights behind the two critical parts of our UM (Section 3.2). Lastly, we describe the framework and the learning objective of UM that incorporates the previous component, as well as its variant, i.e., UMAP (Section 3.3). ", + "bbox": [ + 174, + 765, + 825, + 835 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "3.1 ONCE-COVERED OOD DETECTION CAPABILITY ", + "text_level": 1, + "bbox": [ + 174, + 854, + 547, + 869 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "First, we present the phenomenon of the inconsistent trend between a better OOD detection capability and smaller training error during training. Empirically, as shown in Figure 1, we plot the OOD detection performance during the model training after multiple runs of the experiments. Across three different OOD testing datasets, we can observe the existence of a better detection capability using the index of FPR95 metric based on the Energy (Liu et al., 2020) or ODIN (Liang et al., 2018) score. The generality has also been verified using different learning schedules and model structures in our experimental sections. We further show the comparison of the ID/OOD distributions in Figures 1(b) and 1(c). To be specific, the statics of the two distributions indicate that the gap between the ID and OOD data gets narrow as their overlap grows along with the training. After Epoch 60, although the model becomes more confident on ID data which satisfies a part of the calibration target (Hendrycks et al., 2019a), its predictions on the OOD data also become more confident which is unexpected. Without seeing any auxiliary outliers, it motivates us to explore how the model achieves that. ", + "bbox": [ + 176, + 882, + 825, + 924 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 103, + 825, + 228 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "We take a closer look at the model behaviors in Figure 2, where we check its corresponding training/testing loss and accuracy. We find that the training loss has reached a reasonably small value at Epoch 60 where its detection performance also achieves a satisfactory level. However, if we further minimize the training loss, the trend of the FPR95 curve shows almost opposite directions with both training and testing loss or accuracy (see Figures 1(a), 2(b), and 2(c)). Accordingly, we extract those samples that were learned by the model at this period. As shown in Figures 2(e) and 2(f), the misclassified samples learned after Epoch 60 present much atypical semantic features. As deep neural networks tend to first learn the data with typical features (Arpit et al., 2017), we attribute the inconsistent trend to memorizing those atypical data at the later stage. In Appendix C, we provide a detailed discussion between it with the concept of conventional overfitting (Goodfellow et al., 2016). ", + "bbox": [ + 174, + 236, + 825, + 375 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "3.2 UNLEASHING THE INTRINSIC DETECTION POWER ", + "text_level": 1, + "bbox": [ + 176, + 392, + 550, + 405 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "In general, the models that are developed for the original classification tasks are always seeking better performance (i.e., higher testing accuracy and lower training loss) in practice. However, the inconsistent trend revealed before indicates that the intrinsic OOD detection capability maybe once-covered during the training. It gives us a chance to unleash the potential detection power only considering the ID data in training. To this end, we have two important issues that need to address: (1) the model that is well-trained on ID data may have already memorized some atypical samples which can not be figured out; (2) how to forget those atypical samples considering the given model? ", + "bbox": [ + 174, + 417, + 825, + 515 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Mining the atypical samples with constructed discrepancy. According to Figures 2(a) and 2(b), both training accuracy and loss provide limited information that can differentiate the typical and atypical data. Inspired by the learning dynamics (Goodfellow et al., 2016; Arpit et al., 2017) of deep neural networks and the pathway conjecture (Barham et al., 2022) for inference, we try to manually construct the parameter discrepancy to mine the atypical samples from a well-trained model. To be specific, we introduce a novel layer-wise mask to achieve the goal. The masks are applied to all layers, which is consistent with the mask generation in the conventional pruning pipeline (Han et al., 2016). In Figure 3(c), we provide empirical evidence to show that we can figure out the atypical samples via enlarging the mask rate. Utilizing the masked output for loss computation, the atypical samples can be better differentiated. We also provide more discussion about the intuition in Appendix B. ", + "bbox": [ + 174, + 530, + 825, + 669 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Forgetting the atypical samples with gradient ascent. As the training loss achieves zero at the final stage of the given model, we need extra optimization signals to forget those memorized atypical samples. Considering the previous consistent trend before the potential optimal stage (e.g., before Epoch 60 in Figure 1(a)), the optimization signal also needs to control the model update not to be too greedy to drop the discriminative features for OOD detection. Starting with the given model, we can employ the gradient ascent (Sorg et al., 2010; Ishida et al., 2020) to forget the targeted samples, while the tuning phase should also prevent further updates if the model can achieve the expected stage. ", + "bbox": [ + 174, + 684, + 825, + 781 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "3.3 METHOD REALIZATION ", + "text_level": 1, + "bbox": [ + 176, + 799, + 379, + 813 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Based on previous insights, we present our overall framework as well as the learning objective of the proposed Unleashing Mask for OOD detection. Lastly, we discuss its compatibility with either the fundamental scoring functions or the outlier exposure approaches utilizing auxiliary outlier data. ", + "bbox": [ + 176, + 824, + 825, + 866 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Framework. As illustrated in Figure 3(a), our framework consists of two critical components for uncovering the intrinsic OOD detection capability: (1) the initialized mask with a specific pruning rate for constructing the output discrepancy with the original model; (2) the fine-tuning procedure for ", + "bbox": [ + 176, + 882, + 823, + 924 + ], + "page_idx": 4 + }, + { + "type": "image", + "img_path": "images/449c5426c25f0d6c449c60c781928f496a0fbe06ead281d568841a5ca4b5a2d5.jpg", + "image_caption": [ + "Figure 3: (a) a brief illustration of the proposed Unleashing Mask (UM); (b) the mask rate w.r.t. loss value using the masked outputs; (c) examples of misclassified samples after masking the original well-trained model. As for our framework, given a well-trained model, we initialize an extra mask for mining the atypical samples that are sensitive to the changes in model parameters. Then we fine-tune the original model or adopt pruning with the estimated forgetting threshold, i.e., the loss value estimated by the UM. The final model can serve as the base of various score functions to utilize the discriminative features and also as the new initialization of fine-tuning with the auxiliary outliers. " + ], + "image_footnote": [], + "bbox": [ + 192, + 108, + 816, + 252 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "alleviating the memorization of atypical samples. The overall workflow starts with obtaining the loss value of misclassifying those atypical samples and then conducts tuning with the model to forget. ", + "bbox": [ + 173, + 371, + 823, + 401 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Objective for forgetting. Based on our framework, we introduce the forgetting objective as, ", + "bbox": [ + 174, + 415, + 790, + 429 + ], + "page_idx": 5 + }, + { + "type": "equation", + "img_path": "images/7bcd5f75397cf8a7cb7797e49cb807f703141cc6b6347834dccb36a08a06d097.jpg", + "text": "$$\n\\operatorname* { m i n } \\mathcal { L } _ { \\mathrm { U M } } = \\operatorname* { m i n } _ { m _ { \\delta } \\in [ 0 , 1 ] ^ { n } } | \\ell _ { \\mathrm { C E } } ( f ) - \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) | + \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) ,\n$$", + "text_format": "latex", + "bbox": [ + 281, + 433, + 714, + 460 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "where $m _ { \\delta }$ is our proposed layer-wise mask with the pruning rate $\\delta$ , $\\ell _ { \\mathrm { C E } }$ is the CE loss, $\\widehat { \\ell } _ { \\mathrm { C E } }$ is the averaged CE loss over the ID training data, $| \\cdot |$ indicates the computation for absolute value and $m _ { \\delta } \\odot f ^ { * }$ denotes the masked output of the fixed pretrained model that is used to estimate the loss constraint for the learning objective of forgetting, which would be a constant value during the whole finetuning process. Concretely, the well-trained model will start to optimize itself again if it memorized the atypical samples and achieved almost zero loss value. We provide a positive gradient signal when the current loss value is lower than the estimated one and vice versa. The model is expected to finally stabilize around the stage that can forget those atypical samples. ", + "bbox": [ + 173, + 467, + 825, + 580 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Unleashing Mask Adopts Pruning (UMAP). Considering the potential negative effect on the original task performance when conducting tuning for forgetting, we further propose a variant of UM Adopts Pruning, i.e., UMAP, to conduct tuning based on the masked output (e.g., replace $\\ell _ { \\mathrm { C E } } ( f )$ to $\\ell _ { \\mathrm { C E } } ( \\hat { m } _ { p } \\odot f ^ { * } )$ in Eq 3) using the different mask $\\hat { m } _ { p }$ with its pruning rate $p$ as follows, ", + "bbox": [ + 173, + 594, + 825, + 651 + ], + "page_idx": 5 + }, + { + "type": "equation", + "img_path": "images/2377b492a02f03724b79b5e55c19c0be886d00f7a5d0399b6677c53238470b9d.jpg", + "text": "$$\n\\operatorname* { m i n } \\mathcal { L } _ { \\mathrm { U M A P } } = \\operatorname* { m i n } _ { \\hat { m } _ { p } \\in [ 0 , 1 ] ^ { n } , m _ { \\delta } \\in [ 0 , 1 ] ^ { n } } | \\ell _ { \\mathrm { C E } } ( \\hat { m } _ { p } \\odot f ^ { * } ) - \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) | + \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) ,\n$$", + "text_format": "latex", + "bbox": [ + 218, + 655, + 777, + 683 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Different from the objective of UM (i.e., Eq 3) that minimizes the loss value over the model parameter, the objective of UMAP minimizes the loss over the mask to achieve the target of forgetting those atypical samples. UMAP provides an extra mask to restore the detection capacity but doesn’t affect the model parameter for the inference on original tasks, indicating that UMAP is a more practical choice in real-world applications (as empirically verified in our experiments like Table 1). We summarize the algorithms of UM (in Algorithm 1) and UMAP (in Algorithm 2) in Appendix D. ", + "bbox": [ + 174, + 685, + 825, + 770 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Compatible to other methods. As we explore the original OOD detection capability of the welltrained model, it is orthogonal and compatible with those promising methods that equip the given model with better detection ability. To be specific, through our proposed methods, we reveal the oncecovered OOD detection capability via tuning the original model towards its intermediate training stage. The discriminative feature learned at that stage can be utilized by different scoring functions (Huang et al., 2021; Sun & Li, 2022; Wei et al., 2022), like ODIN (Liang et al., 2018) adopted in Figure 2(c). For those methods (Hendrycks et al., 2019a; Liu et al., 2020; Ming et al., 2022) utilizing the auxiliary outliers to regularize the model, our fine-tuned model obtained by UM and UMAP can also serve as their starting point or adjustment. As our strategy does not require any auxiliary outlier data to be involved in training, adjusting the model using ID data during its developing phase is practical. ", + "bbox": [ + 173, + 784, + 826, + 924 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "4 EXPERIMENTS ", + "text_level": 1, + "bbox": [ + 174, + 102, + 326, + 117 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "In this section, we present the performance comparison of the proposed method in the OOD detection scenario. Specifically, we verify the effectiveness of our UM and UMAP with two mainstreams of OOD detection approaches: (i) fundamental scoring function methods; (ii) outlier exposure methods involving auxiliary samples. To better understand and characterize our proposed method, we further conduct extensive explorations on the ablation study and provide the corresponding discussion on each sub-aspect considered in our work. More details and additional results can also refer to Appendix F. ", + "bbox": [ + 174, + 135, + 825, + 218 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "4.1 EXPERIMENTAL SETUPS ", + "text_level": 1, + "bbox": [ + 174, + 238, + 383, + 252 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "Datasets. Following the common benchmarks used in previous work (Liu et al., 2020; Ming et al., 2022), we adopt CIFAR-10, CIFAR-100 (Krizhevsky, 2009) as our ID datasets. We use a series of different image datasets as the OOD datasets, namely Textures (Cimpoi et al., 2014), Places365 (Zhou et al., 2017), SUN (Xiao et al., 2010), LSUN (Yu et al., 2015), and iNaturalist (Van Horn et al., 2018). We also use the other ID dataset as OOD dataset when training on a specific ID dataset, given that none of them shares any same classes (Yang et al., 2021). e.g. we treat CIFAR-100 as the OOD dataset when training on CIFAR-10 in our experiments for comparison. ", + "bbox": [ + 174, + 266, + 825, + 363 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "Training details. We conduct all major experiments on DenseNet-101 (Huang et al., 2017) with training epochs fixed to 100. We also include experiment results on other types of models in the Appendix F. The models are trained using stochastic gradient descent (Kiefer & Wolfowitz, 1952) with Nesterov momentum (Duchi et al., 2011). We adopt Cosine Annealing (Loshchilov & Hutter, 2017) to schedule the learning rate which begins at 0.1. We set the momentum and weight decay to be 0.9 and $1 0 ^ { - 4 }$ respectively throughout all experiments. The size of the mini-batch is 64 for both ID samples (when training and testing) and OOD samples (when testing). More details and further discussion about choosing the mask ratio in experiments can be referred to at the end of Appendix F. ", + "bbox": [ + 174, + 381, + 825, + 492 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "Evaluation metrics. We employ the following three common metrics to evaluate the performance of OOD detection: (i) Area Under the Receiver Operating Characteristic curve (AUROC) (Davis & Goadrich, 2006) can be interpreted as the probability for a positive sample to have a higher discriminating score than a negative sample (Fawcett, 2006); (ii) Area Under the Precision-Recall curve (AUPR) (Manning & Schütze, 1999) is an ideal metric to adjust the extreme difference between positive and negative base rates; (iii) False Positive Rate (FPR) at $9 5 \\%$ True Positive Rate (TPR) (Liang et al., 2018) indicates the probability for a negative sample to be misclassified as positive when the true positive rate is at $9 5 \\%$ . We also include in-distribution testing accuracy (ID-ACC) to reflect the preservation level of the performance for the original classification task on ID data. ", + "bbox": [ + 174, + 511, + 825, + 636 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "OOD detection baselines. We compare the proposed method with several competitive baselines in the two directions. Specifically, we adopt Maximum Softmax Probability (MSP) (Hendrycks & Gimpel, 2017a), ODIN (Liang et al., 2018), Mahalanobis score (Lee et al., 2018b), and Energy score (Liu et al., 2020) as scoring function baselines; We adopt OE (Hendrycks et al., 2019b), Energybounded learning (Liu et al., 2020), and POEM (Ming et al., 2022) as baselines with outliers. For all scoring function methods, we assume the accessibility of well-trained models. For all methods involving outliers, we constrain all major experiments to a fine-tuning scenario, which is more practical in real cases. Different from training a dual-task model at the very beginning, equipping deployed models with OOD Detection ability is a much more common circumstance, considering the millions of existing and running deep learning systems. We leave more details in Appendix A. ", + "bbox": [ + 174, + 654, + 825, + 792 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "4.2 PERFORMANCE COMPARISON ", + "text_level": 1, + "bbox": [ + 176, + 813, + 419, + 827 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "In this part, we present the performance comparison with some representative baseline methods to demonstrate the effectiveness of our UM and UMAP. Our proposed UM is designed for excavating the potential OOD detection capability of the given model. Here we consider several scoring functions to compare the detection performance, and also some outlier exposure methods to further regularize the given model and boost the OOD detection ability. In each category, we choose one with the best detection performance to adopt UM/UMAP and check the detection results with the ID-ACC. ", + "bbox": [ + 174, + 840, + 825, + 922 + ], + "page_idx": 6 + }, + { + "type": "table", + "img_path": "images/dbf3cde50a959ea56cfbc57c0fe25efe3f04e4f79e781821b4936a380e3c5ee4.jpg", + "table_caption": [ + "Table 1: Main Results $( \\% )$ . Comparison with competitive OOD detection baselines. " + ], + "table_footnote": [], + "table_body": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑w./w.o Daux
CIFAR-10MSP(Hendrycks & Gimpel,2017a) ODIN(Liang et al.,2018)89.90 ±0.30 91.46 ± 0.5691.48 ± 0.43 91.67 ± 0.5860.08 ±0.76 42.31 ± 1.3894.01 ±0.08 94.01 ± 0.08
Mahalanobis(Lee et al.,2018b)75.10 ± 1.0472.32 ± 1.9261.35 ± 1.2594.01±0.08
Energy(Liu et al.,2020)92.07 ± 0.2292.72 ± 0.3942.69 ± 1.3194.01± 0.08
Energy+UM(ours)93.73 ±0.3694.27 ± 0.6033.29 ± 1.7092.80±0.47
Energy+UMAP(ours)93.97 ± 0.1194.38 ± 0.0630.71 ± 1.9494.01± 0.08
OE(Hendrycks et al., 2019b)97.07 ± 0.0197.31± 0.0513.80 ± 0.2892.59 ± 0.32
Energy (w.Daux)(Liu et al.,2020)94.58 ± 0.6494.69 ± 0.6518.79 ± 2.3180.91 ± 3.13
POEM(Ming et al.,2022) OE+UM(ours)94.37 ± 0.07 97.60 ± 0.0394.51± 0.06 97.87 ± 0.0218.50 ± 0.33 11.22 ± 0.1677.24 ± 2.22 93.66 ±0.12>>>>
OE+UMAP(ours)97.48 ± 0.0197.74 ± 0.0012.21± 0.0994.01 ± 0.08√
MSP(Hendrycks & Gimpel,2017a) ODIN(Liang et al., 2018)74.06 ± 0.6975.37 ± 0.7383.14± 0.8774.86 ± 0.21
Mahalanobis(Lee et al.,2018b)76.18 ± 0.14 63.90 ± 1.9176.49 ± 0.20 64.31 ± 0.9178.93 ± 0.31 78.79± 0.5074.86 ± 0.21
Energy(Liu et al.,2020)76.29 ±0.2477.06± 0.5578.46± 0.0674.86± 0.21 74.86 ± 0.21
Energy+UM(ours)76.22 ± 0.4276.39 ± 1.0374.05 ± 0.5564.55± 0.24
Energy+UMAP(ours)75.57 ± 0.5975.66 ±0.0772.21 ± 1.4674.86 ± 0.21
OE(Hendrycks et al., 2019b)
90.55 ± 0.87 88.92 ± 0.5790.34 ± 0.94 89.13 ± 0.5634.73 ± 3.85 37.90 ± 2.5973.59 ± 0.30
Energy (w.Daux)(Liu et al.,2020)57.85 ± 2.65广
POEM(Ming et al.,2022)88.95 ± 0.5488.94 ± 0.3138.10 ± 1.3056.18 ± 1.92√
OE+UM(ours)91.04±0.1191.13 ± 0.2434.71 ± 0.8175.15 ± 0.18√
OE+UMAP (ours)91.10 ± 0.1690.99 ±0.2333.62 ±0.2674.76 ± 0.11√
", + "bbox": [ + 176, + 125, + 825, + 383 + ], + "page_idx": 7 + }, + { + "type": "table", + "img_path": "images/8fb71e78539bf2c0d61786777ad61aa9dee9b855b204b5fc5b22ef8d595be04e.jpg", + "table_caption": [ + "Table 2: Fine-grained Results $( \\% )$ . Comparison on different OOD benchmark datasets. " + ], + "table_footnote": [], + "table_body": "
ID datasetMethodOOD dataset
CIFAR-100TexturesPlaces365
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-10MSP66.43 ± 1.2587.73± 0.0265.20±1.3388.06 ±0.6161.34 ± 0.6089.63±0.15
ODIN55.31± 0.8587.75 ± 0.3753.11 ± 4.8487.13 ± 2.0443.77 ±0.2091.70 ± 0.30
Mahalanobis81.61 ± 0.9664.52 ±0..7320.04 ± 1.4394.38 ± 0.7886.21±1.3664.00 ± 1.21
Energy54.65 ± 1.2489.01 ± 1.1857.09 ±3.5287.51 ± 1.4338.62 ± 1.6493.03±0.20
Energy+UM(ours)54.62 ± 1.1688.30±0.3041.61 ± 3.6791.31 ± 0.0130.85 ±0.5894.27 ±0.16
MethodSUN FPR95↓LSUNiNaturalist
AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
MSP60.27 ± 0.6690.00±0.2436.43 ± 1.9495.17 ± 0.3267.53 ± 1.6488.01±0.82
ODIN41.14 ± 1.2992.34±0.625.16 ±0.7698.96 ±0.0954.41 ± 0.9190.17 ± 0.19
Mahalanobis84.56 ± 1.5166.41 ± 4.5769.18 ± 3.5266.41 ± 4.5780.76± 2.4871.77 ± 1.12
Energy36.73 ± 1.7293.63 ± 0.34 94.83 ± 0.116.25 ± 0.4398.77±0.07 99.22 ± 0.1159.11 ± 1.1889.71 ± 0.06 92.75 ± 0.80
Energy+UM(ours)27.88 ±0.732.91 ± 0.5346.27 ± 2.74
", + "bbox": [ + 176, + 422, + 825, + 587 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "In Table 1, we summarize the results of different OOD test sets using different methods. Note that, here the evaluation results are obtained by averaging several OOD test datasets across multiple independent trials. For the scoring-based methods, our UM can further improve the overall detection performance by alleviating the memorization of atypical ID data, when the ID-ACC keeps comparable with the baseline. For the complex CIFAR-100 dataset, our UMAP can be adopted as a practical way to empower the detection performance and simultaneously avoid affecting the original performance on ID data. As for those methods of the second category (i.e., involving auxiliary outlier $D _ { a u x }$ sampling from ImageNet), since we consider a practical workflow, i.e., fine-tuning, on the given model, OE achieves the best performance on the task. Due to the special optimization characteristic, Energy $( \\mathbf { w . } D _ { a u x } )$ and POEM focus more on the energy loss on differentiating OOD data while performing not well on the preservation of ID-ACC. Without sacrificing much performance on ID data, OE with our UM can still achieve better detection performance. In Table 3, the fine-grained detection performance on each OOD testing set demonstrates the general effectiveness of UM. We have comprehensively verified the significant improvement (up to $1 8 \\%$ reduced on averaged FPR95) in OOD detection of our methods across different setups in Appendix F, the complete results can refer to Tables 8 to 20. More fine-grained results of the experiments on CIFAR-100 is provided in Table 16 and 19. In addition, we also provide similar results using another model structure in Table 18. ", + "bbox": [ + 173, + 602, + 825, + 837 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "4.3 ABLATION AND FURTHER DISCUSSION ", + "text_level": 1, + "bbox": [ + 176, + 856, + 483, + 869 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "In this part, we conduct various explorations to provide a thorough understanding of our presented Unleashing Mask from different perspectives. To be specific, we first present the general existence of the once-covered intrinsic OOD detection capability across different settings. Second, we present the exploration about the mask, which helps us to characterize its effect on figuring out the atypical samples. Third, we provide further exploration on excavating the detection power via pruning. ", + "bbox": [ + 176, + 882, + 823, + 924 + ], + "page_idx": 7 + }, + { + "type": "image", + "img_path": "images/ddbd0f1c22b91a329ea8bf02dce6b6e526822b5945af1d952567d8f633227704.jpg", + "image_caption": [ + "Figure 4: Ablation studies. (a)-(b) exploring the existence of potential OOD detection capability with different learning rate schedules; (c) comparison of loss value using original output with masked output (the $\\mathbf { X }$ -axis represents the index of samples within a mini-batch); (d) effects of using different masking ratios in UM; (e) comparison of using original pruning with our proposed UMAP. " + ], + "image_footnote": [], + "bbox": [ + 184, + 108, + 815, + 210 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 295, + 821, + 324 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "General existence of once-covered OOD detection capability. In Figure 4(a) and Figure 4(b), we explore 4 learning rate schedules to demonstrate the general existence of once-covered OOD detection capability. To be specific, the OOD performance (indicated by FPR95) is evaluated along with the training in every 5 epoch, in which the model takes CIFAR-10 as ID data and SVHN as OOD data. As shown by the curves, a middle stage exists with better OOD performance than that of the final stage across different schedules. We empirically verify the existence of this phenomenon without schedule specificity. More explorations on the other ID dataset and model structure reveal similar results in Figures 5, 6, and 7. The detailed information and discussion can refer to Appendix F. ", + "bbox": [ + 174, + 338, + 825, + 450 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "Effects of the mask on mining atypical samples. Following our previous illustration of Figure 3(c), we scrutinize the change of training loss on a random batch of the training set in Figure 4(c). The results further explain why the loss value estimated by the UM can be used to force the model to forget atypical samples. It can be seen from Figure 4(c) that the loss is proportionally increased by randomly knocking out $2 . 5 \\%$ weights. In this case, the estimated loss is more influenced by those who have a higher initial loss and are what we termed as atypical samples. By controlling the training loss to the estimated value, the model is encouraged to backtrack to a middle training stage where samples with high loss value have little influence on the forgetting process of the gradient ascent. ", + "bbox": [ + 174, + 465, + 825, + 578 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "Exploration on revealing detection capability with model pruning. Although the large constrain on training loss can help reveal the model’s OOD performance, the ID-ACC is undermined under such circumstances. Generally speaking, the proposed UM forces the model to forget the atypical samples and may result in lower test performance. To mitigate this issue, we further adopt pruning as a countermeasure to learn a mask instead of tuning the model parameters directly. In Figure 4(e), we experiment with various prune rates $p$ and demonstrate that we can achieve the same or better OOD performance also by pruning. Specifically, our UMAP can achieve a lower FPR95 than pure pruning with the original objective. The prune rate can be selected from a wide range (e.g. $p \\in [ 0 . 3 , 0 . 9 ] )$ ) to guarantee a fast convergence and effectiveness. Since pruning doesn’t change the well-trained model parameters, it can preserve the performance of the original task. We also provide additional empirical results and corresponding discussion about the effectiveness of our UMAP in Appendix F. ", + "bbox": [ + 174, + 593, + 825, + 746 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "5 CONCLUSION ", + "text_level": 1, + "bbox": [ + 176, + 766, + 318, + 782 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "In this work, we explore the intrinsic OOD detection capability of a well-trained model. Without involving any auxiliary outliers in training, we reveal the inconsistent trend between minimizing original training loss and gaining OOD detection capability. We further attribute it to the memorization behavior of atypical samples. To excavate the once-covered capability, we propose a new method, namely, Unleashing Mask (UM). Through UM, we construct model-level discrepancy that figures out the memorized atypical samples and utilizes the constrained gradient ascent to encourage forgetting. It better utilizes the given model for OOD detection via backtracking or sub-structure pruning. We hope our work could provide new insights for revisiting the model development in OOD detection. ", + "bbox": [ + 174, + 796, + 825, + 909 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "REPRODUCIBILITY STATEMENT ", + "text_level": 1, + "bbox": [ + 176, + 102, + 437, + 118 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "For the experimental setups, we have provided the details in Section 4.1 and Appendixes A and E. We will also provide the anonymous repository about our source codes in the discussion phase for reviewing purposes to ensure the reproducibility of our experimental results. ", + "bbox": [ + 174, + 133, + 825, + 176 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "REFERENCES ", + "text_level": 1, + "bbox": [ + 174, + 196, + 287, + 212 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Devansh Arpit, Stanislaw Jastrzebski, Nicolas Ballas, David Krueger, Emmanuel Bengio, Maxinder S. Kanwal, Tegan Maharaj, Asja Fischer, Aaron C. Courville, Yoshua Bengio, and Simon LacosteJulien. A closer look at memorization in deep networks. In ICML, 2017. ", + "bbox": [ + 176, + 220, + 826, + 262 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Paul Barham, Aakanksha Chowdhery, Jeff Dean, Sanjay Ghemawat, Steven Hand, Daniel Hurt, Michael Isard, Hyeontaek Lim, Ruoming Pang, Sudip Roy, et al. Pathways: Asynchronous distributed dataflow for ml. In MLSys, 2022. ", + "bbox": [ + 174, + 272, + 826, + 315 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Mikhail Belkin, Daniel Hsu, Siyuan Ma, and Soumik Mandal. Reconciling modern machine-learning practice and the classical bias–variance trade-off. In Proceedings of the National Academy of Sciences, 2019. ", + "bbox": [ + 173, + 325, + 825, + 367 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Jiefeng Chen, Yixuan Li, Xi Wu, Yingyu Liang, and Somesh Jha. Atom: Robustifying out-ofdistribution detection using outlier mining. In ECML PKDD, 2021. ", + "bbox": [ + 169, + 377, + 825, + 406 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Mircea Cimpoi, Subhransu Maji, Iasonas Kokkinos, Sammy Mohamed, and Andrea Vedaldi. Describing textures in the wild. In CVPR, 2014. ", + "bbox": [ + 173, + 415, + 825, + 445 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Jesse Davis and Mark Goadrich. The relationship between precision-recall and roc curves. In ICML, 2006. ", + "bbox": [ + 171, + 454, + 825, + 483 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In CVPR, 2009. ", + "bbox": [ + 171, + 492, + 823, + 522 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Xuefeng Du, Zhaoning Wang, Mu Cai, and Sharon Li. Towards unknown-aware learning with virtual outlier synthesis. In ICLR, 2022a. ", + "bbox": [ + 173, + 531, + 823, + 560 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Xuefeng Du, Zhaoning Wang, Mu Cai, and Yixuan Li. VOS: learning what you don’t know by virtual outlier synthesis. In ICLR, 2022b. ", + "bbox": [ + 173, + 569, + 823, + 599 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "John Duchi, Elad Hazan, and Yoram Singer. Adaptive subgradient methods for online learning and stochastic optimization. In JMLR, 2011. ", + "bbox": [ + 173, + 608, + 823, + 637 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Tom Fawcett. An introduction to roc analysis. In Pattern Recognition Letters, 2006. ", + "bbox": [ + 174, + 647, + 725, + 662 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Stanislav Fort, Jie Ren, and Balaji Lakshminarayanan. Exploring the limits of out-of-distribution detection. In NeurIPS, 2021. ", + "bbox": [ + 176, + 671, + 825, + 700 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Jonathan Frankle and Michael Carbin. The lottery ticket hypothesis: Training pruned neural networks. In ICLR, 2019. ", + "bbox": [ + 173, + 710, + 823, + 739 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Ian Goodfellow, Yoshua Bengio, Aaron Courville, and Yoshua Bengio. Deep learning. MIT Press, 2016. ", + "bbox": [ + 173, + 748, + 823, + 777 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Ian J. Goodfellow, Dumitru Erhan, Pierre Luc Carrier, Aaron Courville, Mehdi Mirza, Ben Hamner, Will Cukierski, Yichuan Tang, David Thaler, Dong-Hyun Lee, Yingbo Zhou, Chetan Ramaiah, Fangxiang Feng, Ruifan Li, Xiaojie Wang, Dimitris Athanasakis, John Shawe-Taylor, Maxim Milakov, John Park, Radu Ionescu, Marius Popescu, Cristian Grozea, James Bergstra, Jingjing Xie, Lukasz Romaszko, Bing Xu, Zhang Chuang, and Yoshua Bengio. Challenges in representation learning: A report on three machine learning contests. In Minho Lee, Akira Hirose, Zeng-Guang Hou, and Rhee Man Kil (eds.), NeurIPS. Springer Berlin Heidelberg, 2013. ", + "bbox": [ + 174, + 787, + 826, + 886 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Song Han, Huizi Mao, and William J Dally. Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding. In ICLR, 2016. ", + "bbox": [ + 178, + 895, + 823, + 924 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Trevor Hastie, Robert Tibshirani, and Jerome Friedman. The elements of statistical learning: Data mining, inference, and prediction, second edition. Springer New York, NY, 2009. ", + "bbox": [ + 171, + 103, + 823, + 132 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution examples in neural networks. In ICLR, 2017a. ", + "bbox": [ + 173, + 141, + 823, + 171 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution examples in neural networks. In ICLR, 2017b. ", + "bbox": [ + 173, + 179, + 823, + 209 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Dan Hendrycks, Kimin Lee, and Mantas Mazeika. Using pre-training can improve model robustness and uncertainty. In ICML, 2019a. ", + "bbox": [ + 173, + 218, + 823, + 247 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier exposure. In ICLR, 2019b. ", + "bbox": [ + 174, + 256, + 823, + 286 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Dan Hendrycks, Mantas Mazeika, and Thomas G. Dietterich. Deep anomaly detection with outlier exposure. In ICLR, 2019c. ", + "bbox": [ + 174, + 295, + 823, + 324 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Gao Huang, Zhuang Liu, Laurens Van Der Maaten, and Kilian Q Weinberger. Densely connected convolutional networks. In CVPR, 2017. ", + "bbox": [ + 174, + 333, + 823, + 363 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Rui Huang, Andrew Geng, and Yixuan Li. On the importance of gradients for detecting distributional shifts in the wild. In NeurIPS, 2021. ", + "bbox": [ + 173, + 371, + 823, + 401 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Takashi Ishida, Ikko Yamane, Tomoya Sakai, Gang Niu, and Masashi Sugiyama. Do we need zero training loss after achieving zero training error? In ICML, 2020. ", + "bbox": [ + 173, + 410, + 823, + 440 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "J. Kiefer and J. Wolfowitz. Stochastic estimation of the maximum of a regression function. In Ann. Math. Statist., 1952. ", + "bbox": [ + 173, + 448, + 825, + 478 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Alex Krizhevsky. Learning multiple layers of features from tiny images. In arXiv, 2009. ", + "bbox": [ + 169, + 486, + 754, + 502 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Kimin Lee, Honglak Lee, Kibok Lee, and Jinwoo Shin. Training confidence-calibrated classifiers for detecting out-of-distribution samples. In ICLR, 2018a. ", + "bbox": [ + 173, + 511, + 823, + 541 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. A simple unified framework for detecting out-of-distribution samples and adversarial attacks. In NeurIPS, 2018b. ", + "bbox": [ + 174, + 549, + 823, + 579 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Shiyu Liang, Yixuan Li, and R. Srikant. Enhancing the reliability of out-of-distribution image detection in neural networks. In ICLR, 2018. ", + "bbox": [ + 174, + 588, + 823, + 617 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Ziqian Lin, Sreya Dutta Roy, and Yixuan Li. Mood: Multi-level out-of-distribution detection. In CVPR, 2021. ", + "bbox": [ + 174, + 626, + 825, + 656 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Weitang Liu, Xiaoyun Wang, John D Owens, and Yixuan Li. Energy-based out-of-distribution detection. In NeurIPS, 2020. ", + "bbox": [ + 173, + 664, + 825, + 694 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Ilya Loshchilov and Frank Hutter. SGDR: stochastic gradient descent with warm restarts. In ICLR, 2017. ", + "bbox": [ + 171, + 703, + 826, + 733 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Christopher D. Manning and Hinrich Schütze. Foundations of Statistical Natural Language Processing. The MIT Press, Cambridge, Massachusetts, 1999. ", + "bbox": [ + 173, + 741, + 826, + 771 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Yifei Ming, Ying Fan, and Yixuan Li. Poem: Out-of-distribution detection with posterior sampling. In ICML, 2022. ", + "bbox": [ + 173, + 779, + 825, + 809 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Sina Mohseni, Mandar Pitale, J. B. S. Yadawa, and Zhangyang Wang. Self-supervised learning for generalizable out-of-distribution detection. In AAAI, 2020. ", + "bbox": [ + 171, + 818, + 825, + 848 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Yuval Netzer, Tao Wang, Adam Coates, Alessandro Bissacco, Bo Wu, and Andrew Y Ng. Reading digits in natural images with unsupervised feature learning. In NeurIPS Workshop, 2011. ", + "bbox": [ + 174, + 856, + 821, + 886 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Anh Nguyen, Jason Yosinski, and Jeff Clune. Deep neural networks are easily fooled: High confidence predictions for unrecognizable images. In CVPR, 2015. ", + "bbox": [ + 176, + 895, + 820, + 924 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Vivek Ramanujan, Mitchell Wortsman, Aniruddha Kembhavi, Ali Farhadi, and Mohammad Rastegari. What’s hidden in a randomly weighted neural network? In CVPR, 2020. ", + "bbox": [ + 171, + 103, + 825, + 132 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Mohammadreza Salehi, Hossein Mirzaei, Dan Hendrycks, Yixuan Li, Mohammad Hossein Rohban, and Mohammad Sabokrou. A unified survey on anomaly, novelty, open-set, and out-of-distribution detection: Solutions and future challenges. In arXiv, 2021. ", + "bbox": [ + 178, + 140, + 823, + 184 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Vikash Sehwag, Mung Chiang, and Prateek Mittal. SSD: A unified framework for self-supervised outlier detection. In ICLR, 2021. ", + "bbox": [ + 171, + 193, + 823, + 222 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Jonathan Sorg, Richard L Lewis, and Satinder Singh. Reward design via online gradient ascent. In NeurIPS, 2010. ", + "bbox": [ + 173, + 229, + 825, + 258 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Nitish Srivastava, Geoffrey E. Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. Dropout: a simple way to prevent neural networks from overfitting. In J. Mach. Learn. Res., 2014. ", + "bbox": [ + 173, + 267, + 825, + 297 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Yiyou Sun and Yixuan Li. Dice: Leveraging sparsification for out-of-distribution detection. In ECCV, 2022. ", + "bbox": [ + 174, + 306, + 825, + 335 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Yiyou Sun, Chuan Guo, and Yixuan Li. React: Out-of-distribution detection with rectified activations. In NeurIPS, 2021. ", + "bbox": [ + 174, + 343, + 825, + 372 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Yiyou Sun, Yifei Ming, Xiaojin Zhu, and Yixuan Li. Out-of-distribution detection with deep nearest neighbors. In ICML, 2022. ", + "bbox": [ + 176, + 381, + 823, + 410 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Jihoon Tack, Sangwoo Mo, Jongheon Jeong, and Jinwoo Shin. CSI: novelty detection via contrastive learning on distributionally shifted instances. In NeurIPS, 2020. ", + "bbox": [ + 173, + 419, + 823, + 448 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Amirhossein Tavanaei. Embedded encoder-decoder in convolutional networks towards explainable AI. In arXiv, 2020. ", + "bbox": [ + 176, + 457, + 823, + 486 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "William R Thompson. On the likelihood that one unknown probability exceeds another in view of the evidence of two sample. In Biometrika, 1933. ", + "bbox": [ + 174, + 494, + 821, + 523 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Grant Van Horn, Oisin Mac Aodha, Yang Song, Yin Cui, Chen Sun, Alex Shepard, Hartwig Adam, Pietro Perona, and Serge Belongie. The inaturalist species classification and detection dataset. In CVPR, 2018. ", + "bbox": [ + 174, + 531, + 823, + 575 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Hongxin Wei, Renchunzi Xie, Hao Cheng, Lei Feng, Bo An, and Yixuan Li. Mitigating neural network overconfidence with logit normalization. In ICML, 2022. ", + "bbox": [ + 173, + 583, + 821, + 613 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Jianxiong Xiao, James Hays, Krista A. Ehinger, Aude Oliva, and Antonio Torralba. Sun database: Large-scale scene recognition from abbey to zoo. In IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2010. ", + "bbox": [ + 173, + 621, + 823, + 665 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Jingkang Yang, Kaiyang Zhou, Yixuan Li, and Ziwei Liu. Generalized out-of-distribution detection: A survey. In arXiv, 2021. ", + "bbox": [ + 173, + 672, + 825, + 702 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Fisher Yu, Ari Seff, Yinda Zhang, Shuran Song, Thomas Funkhouser, and Jianxiong Xiao. Lsun: Construction of a large-scale image dataset using deep learning with humans in the loop. In arXiv, 2015. ", + "bbox": [ + 174, + 710, + 825, + 753 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Qing Yu and Kiyoharu Aizawa. Unsupervised out-of-distribution detection by maximum classifier discrepancy. In ICCV, 2019. ", + "bbox": [ + 173, + 762, + 823, + 791 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Sergey Zagoruyko and Nikos Komodakis. Wide residual networks. In BMVC, 2016. ", + "bbox": [ + 173, + 799, + 727, + 815 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Bolei Zhou, Agata Lapedriza, Antonio Torralba, and Aude Oliva. Places: An image database for deep scene understanding. In Journal of Vision, 2017. ", + "bbox": [ + 174, + 824, + 825, + 853 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "APPENDIX ", + "text_level": 1, + "bbox": [ + 176, + 103, + 263, + 117 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "REPRODUCIBILITY STATEMENT ", + "text_level": 1, + "bbox": [ + 176, + 135, + 437, + 151 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "We will provide the anonymous repository about our source codes in the discussion phase for reviewing purposes to ensure the reproducibility of our experimental results. Below we summarize some critical aspects to facilitate the reproducible results: ", + "bbox": [ + 174, + 166, + 825, + 208 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "• Datasets. The datasets we used are all publicly accessible, which is introduced in Section 4.1. For methods involving auxiliary outliers, we strictly follow previous works (Sun et al., 2021; Du et al., 2022b) to avoid overlap between the auxiliary dataset (ImageNet-1k) (Deng et al., 2009) and any other OOD datasets. \n• Assumption. We set our experiments to a post-hoc scenario where a well-trained model is available, and some parts of training samples are also available for subsequent fine-tuning. \n• Open source. The code repository will be available in an anonymous repository for reviewing purposes. We provide a backbone of our experiments as well as several auxiliary components, such as score estimation. \n• Environment. All experiments are conducted multiple runs on NVIDIA Tesla V100-SXM2- 32GB GPUs with Python 3.6 and PyTorch 1.8. ", + "bbox": [ + 217, + 219, + 825, + 387 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "A DETAILED INFORMATION ABOUT THE USED BASELINES AND METRICS ", + "text_level": 1, + "bbox": [ + 176, + 407, + 787, + 422 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "In this section, we provide the details about the baselines for the scoring functions and finetuning with auxiliary outliers, and the corresponding hyper-parameters that are considered in our work. ", + "bbox": [ + 173, + 436, + 825, + 467 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "Maximum Softmax Probability (MSP). (Hendrycks & Gimpel, 2017a) proposes to use maximum softmax probability to discriminate ID and OOD samples. The score is defined as follows, ", + "bbox": [ + 173, + 479, + 823, + 508 + ], + "page_idx": 12 + }, + { + "type": "equation", + "img_path": "images/eb647d90be010a2a4bc2264b9e3f27009ae5a64b4cbb043708a540be775f9884.jpg", + "text": "$$\nS _ { \\mathrm { M S P } } ( x ; f ) = \\operatorname* { m a x } _ { c } P ( y = c | x ; f ) = \\operatorname* { m a x } { \\ s o f \\ t m a x } ( f ( x ) )\n$$", + "text_format": "latex", + "bbox": [ + 300, + 512, + 697, + 535 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "where $f$ represents the given well-trained model and $c$ is one of the classes $\\mathcal { V } = \\{ 1 , \\ldots , C \\}$ . The larger softmax score indicates the larger probability for a sample to be ID data, reflecting the model’s confidence on the sample. ", + "bbox": [ + 174, + 540, + 823, + 582 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "ODIN. (Liang et al., 2018) designed the ODIN score, leveraging the temperature scaling and tiny perturbations to widen the gap between the distributions of ID and OOD samples. The ODIN score is defined as follows, ", + "bbox": [ + 173, + 597, + 825, + 638 + ], + "page_idx": 12 + }, + { + "type": "equation", + "img_path": "images/58226108e1cc1addb3f70429e71b26864132fcdcf209f7b1ce63d32ddfcb0abd.jpg", + "text": "$$\nS _ { \\mathrm { O D I N } } ( x ; f ) = \\operatorname* { m a x } _ { c } P ( y = c | \\tilde { x } ; f ) = \\operatorname* { m a x } \\ s \\ o f \\tan \\alpha ( \\frac { f ( \\tilde { x } ) } { T } )\n$$", + "text_format": "latex", + "bbox": [ + 295, + 642, + 702, + 672 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "where $\\tilde { x }$ represents the perturbed samples (controled by $\\dot { \\epsilon }$ ), $T$ represents the temperature. For fair comparison, we adopt the suggested hyperparameters (Liang et al., 2018): $\\epsilon \\doteq 1 . 4 \\times 1 0 ^ { - 3 }$ $T = 1 . 0 { \\dot { \\times } } 1 0 ^ { 4 }$ . ", + "bbox": [ + 174, + 676, + 823, + 719 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "Mahalanobis. (Lee et al., 2018b) introduces a Mahalanobis distance-based confidence score, exploiting the feature space of the neural networks by inspecting the class conditional Gaussian distributions. The Mahalanobis distance score is defined as follows, ", + "bbox": [ + 173, + 733, + 825, + 775 + ], + "page_idx": 12 + }, + { + "type": "equation", + "img_path": "images/891b3f112444c2d329b35cc3e1e45abd0703a13b9d9b5a9a91c0a457b3486865.jpg", + "text": "$$\nS _ { \\mathrm { M a h a l a n o b i s } } ( x ; f ) = \\operatorname* { m a x } _ { c } - ( f ( x ) - \\hat { \\mu } _ { c } ) ^ { T } \\hat { \\Sigma } ^ { - 1 } ( f ( x ) - \\hat { \\mu } _ { c } )\n$$", + "text_format": "latex", + "bbox": [ + 308, + 777, + 689, + 803 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "where $\\hat { \\mu } _ { c }$ represents the estimated mean of multivariate Gaussian distribution of class $c$ , $\\hat { \\Sigma }$ represents the estimated tied covariance of the $C$ class-conditional Gaussian distributions. ", + "bbox": [ + 173, + 809, + 825, + 838 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "Energy. (Liu et al., 2020) proposes to use the Energy of the predicted logits to distinguish the ID and OOD samples. The Energy score is defined as follows, ", + "bbox": [ + 171, + 852, + 823, + 881 + ], + "page_idx": 12 + }, + { + "type": "equation", + "img_path": "images/943af2b0f9767b994b1104b7822273e96ffa5715dea3e0fd413a2e99977a8a73.jpg", + "text": "$$\nS _ { \\mathrm { E n e r g y } } ( x ; f ) = - T \\log \\sum _ { c = 1 } ^ { C } e ^ { f ( x ) _ { c } / T }\n$$", + "text_format": "latex", + "bbox": [ + 379, + 886, + 619, + 928 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "where $T$ represents the temperature parameter. As theoretically illustrated in Liu et al. (2020), a lower Energy score indicates a higher probability for a sample to be ID. Following (Liu et al., 2020), we fix the $T$ to 1.0 throughout all experiments. ", + "bbox": [ + 173, + 103, + 823, + 146 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "Outlier Exposure (OE). (Hendrycks et al., 2019b) initiates a promising approach towards OOD detections by involving outliers to force apart the distributions of ID and OOD samples. In the experiments, we use the cross-entropy from $f ( x _ { \\mathrm { o u t } } )$ to the uniform distribution as the $\\mathcal { L } _ { \\mathrm { O E } }$ (Lee et al., 2018a), ", + "bbox": [ + 174, + 161, + 825, + 217 + ], + "page_idx": 13 + }, + { + "type": "equation", + "img_path": "images/e370174786b5ffabd541014636396d5283038b86fad99a11cfbc378251857514.jpg", + "text": "$$\n\\mathcal { L } _ { f } = \\mathbb { E } _ { D _ { \\mathrm { i n } } } \\left[ \\ell _ { \\mathrm { C E } } ( f ( x ) , y ) \\right] + \\lambda \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } \\left[ \\log \\sum _ { c = 1 } ^ { C } e ^ { f ( x ) _ { c } } - \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } ( f ( x ) ) \\right]\n$$", + "text_format": "latex", + "bbox": [ + 277, + 217, + 720, + 261 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "Energy (w. $D _ { \\mathbf { a u x } } )$ ). In addition to using the Energy as a post-hoc score to distinguish $\\mathrm { I D }$ and OOD samples, (Liu et al., 2020) proposes an Energy-bounded objective to further separate the two distributions. The OE objective is as follows, ", + "bbox": [ + 174, + 275, + 825, + 316 + ], + "page_idx": 13 + }, + { + "type": "equation", + "img_path": "images/54a9f407cbd94084751fcdf26413676f1b8a4a35835da9c373810c40b9ce2039.jpg", + "text": "$$\n\\mathcal { L } _ { \\mathrm { O E } } = \\mathbb { E } _ { D _ { \\mathrm { i n } } ^ { s } } ( \\operatorname* { m a x } ( 0 , S _ { \\mathrm { E n e r g y } } ( x , f ) - m _ { \\mathrm { i n } } ) ) ^ { 2 } + \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } ( \\operatorname* { m a x } ( 0 , m _ { \\mathrm { o u t } } - S _ { \\mathrm { E n e r g y } } ( x , f ) ) ) ^ { 2 }\n$$", + "text_format": "latex", + "bbox": [ + 209, + 323, + 759, + 343 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "We keep the thresholds same to (Liu et al., 2020): $m _ { \\mathrm { i n } } = - 2 5 . 0$ , $m _ { \\mathrm { o u t } } = - 7 . 0$ . ", + "bbox": [ + 176, + 348, + 691, + 364 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "POEM. (Ming et al., 2022) explores the Thompson sampling strategy (Thompson, 1933) to make the most use of outliers to learn a tight decision boundary. Though given the POEM’s nature to be orthogonal to other OE methods, we use the Energy(w. $D _ { \\mathrm { a u x . } }$ ) as the backbone, which is the same as Eq.( 10) in Liu et al. (2020). The details of Thompson sampling can refer to Ming et al. (2022). ", + "bbox": [ + 173, + 380, + 823, + 438 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "Detailed formulations of FPR and TPR. Suppose we have a binary classification task (to predict an image to be an ID or OOD sample in this paper). There are two possible outputs: a positive result (the model predicts an image to be an ID sample); a negative result (the model predicts an image to be an OOD sample). Since we have two possible labels and two possible outputs, we can form a confusion matrix with all possible outputs as follows. ", + "bbox": [ + 174, + 452, + 825, + 523 + ], + "page_idx": 13 + }, + { + "type": "table", + "img_path": "", + "table_caption": [ + "Table 3: Confusion Matrix. " + ], + "table_footnote": [], + "page_idx": 13 + }, + { + "type": "equation", + "img_path": "images/ca2f2a2fa96503d7c6f7f24da26223a214941e13ae665bd918c6f60ee89e3b68.jpg", + "text": "$$\n\\frac { | \\mathrm { ~ { T r u t h : I D } ~ } | \\mathrm { { T r u t h : O O D } } } { \\mathrm { { P r e d i c t : ~ O O D ~ | \\mathrm { ~ { ~ T r u e ~ P o s i t i v e ~ ( T P ) ~ } | { \\mathrm { ~ F a l s e ~ P o s i t i v e ~ ( F P ) ~ } } } } } }\n$$", + "text_format": "latex", + "bbox": [ + 302, + 556, + 689, + 616 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "Therefore, the false positive rate (FPR) is calculated as : ", + "bbox": [ + 174, + 628, + 539, + 643 + ], + "page_idx": 13 + }, + { + "type": "equation", + "img_path": "images/6de9bd865babdf43ef851a8d2f51cca43a0e5ddbbc75bd8c1eb583ef265e7786.jpg", + "text": "$$\nF P R = { \\frac { F P } { F P + T N } }\n$$", + "text_format": "latex", + "bbox": [ + 428, + 650, + 568, + 681 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "The true positive rate (TPR) is calculated as: ", + "bbox": [ + 176, + 695, + 465, + 710 + ], + "page_idx": 13 + }, + { + "type": "equation", + "img_path": "images/1175e2c392fa210f94055024b476db4a2fb3df5a12672ac7bfe9beb3d5388cde.jpg", + "text": "$$\nT P R = { \\frac { T P } { T P + F N } }\n$$", + "text_format": "latex", + "bbox": [ + 429, + 717, + 568, + 750 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "B ADDITIONAL EXPLANATION TOWARDS MINING THE ATYPICAL SAMPLES", + "text_level": 1, + "bbox": [ + 171, + 767, + 803, + 782 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "First, for identifying those atypical samples using a layer-wise mask with the well-pre-trained model, the core intuition behind is constructing the parameter-level discrepancy to mine the atypical samples. It is inspired by and based on the evidence drawn from previous literature about learning behaviors (Arpit et al., 2017; Goodfellow et al., 2016) of deep neural networks (DNNs) and sparse representation (Frankle & Carbin, 2019; Goodfellow et al., 2013; Barham et al., 2022). To be specific, the atypical samples tend to be learned by the DNNs later than those typical samples (Arpit et al., 2017), and are relatively more sensitive to the changes of the model parameter as the model does not generalize well on that. By the layer-wise mask, the constructed discrepancy can make the model misclassify the atypical samples and estimate loss constraint for the forgetting objective, as visualized in Figure 3(c). ", + "bbox": [ + 173, + 797, + 825, + 924 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "Second, introducing the layer-wise mask has several advantages for achieving the staged target of mining atypical samples in our proposed method, while we would also admit that the layer-wise mask is not an irreplaceable option or maybe not optimal. On the one hand, considering that the model has been trained to approach the zero error on training data, utilizing the layer-wise mask is an integrated strategy to 1) figure out the atypical samples and 2) obtain the loss value computed by the masked output that misclassifies them. The loss constraint is later used in the forgetting objective to fine-tune the model. On the other hand, the layer-wise mask is also compatible with the proposed UMAP to generate a flexible mask for restoring the detection capability of the original model. ", + "bbox": [ + 174, + 103, + 825, + 215 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "Third, we also adopt the unit/weight mask and visualize the misclassified samples in Figure 12. We think they can also be used to mine the atypical samples and can be extended or improved to be a more flexible choice. Further investigating the specific effect of different methods that construct the parameter-level discrepancy would be an interesting sub-topic in future work. For the value of CE loss, although the atypical samples tend to have high CE loss value, they are already memorized and correctly classified as indicated by the zero training error. Only using the high CE error can not provide the loss estimation when the model does not correctly classify those samples. ", + "bbox": [ + 174, + 222, + 825, + 319 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "C CONCEPTUAL AND EMPIRICAL COMPARISON WITH OVERFITTING ", + "text_level": 1, + "bbox": [ + 173, + 347, + 748, + 361 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "First of all, we would refer to the concept of the conventional overfitting (Goodfellow et al., 2016; Belkin et al., 2019), i.e., the model \"overfit\" the training data but fail to generalize and perform well on the test data that is unseen during training. The common empirical reflection of overfitting is that the training error is decreasing while the test error is increasing at the same time, which enlarges the generalization gap of the model. It has been empirically confirmed not the case in our observation as observed in Figure 2(a) and 2(b). To be specific, for the original classification task, there is no conventional overfitting observed as the test performance is still improved at the later training stage, which is a general pursuit of the model development phase on the original tasks. ", + "bbox": [ + 174, + 381, + 825, + 492 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "Then, when we consider the OOD detection performance of the well-pretrained model, our unique observation is about the inconsistency between gaining better OOD detection capability and pursuing better performance on the original classification task for the in-distribution (ID) data. It is worth noting that here the training task is not the binary classification of OOD detection, but the classification task on ID data. It is out of the rigorous concept of the conventional overfitting and has received limited focus and discussion in the previous literature about OOD detection to the best of our knowledge. Considering the practical scenario that exists target-level discrepancy, our revealed observation may encourage us to revisit the detection capability of the well-trained model. ", + "bbox": [ + 174, + 498, + 825, + 611 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "Third, through empirical observation, those strategies designed for preventing the conventional overfitting may need to change the target to the OOD detection based on the important observation. In our experiments, for all the baseline models including that used in Figure 1, we have adopted those strategies (Srivastava et al., 2014; Hastie et al., 2009) (e.g., drop-out, weight decay) to reduce overfitting. It is found to be not enough to restore the OOD detection performance. For another shared issue, on the CIFAR-100 dataset, our UM restore the OOD detection capability of the well-trained model with a significant sacrifice on \"ID-ACC\". Using those strategies for reducing overfitting in the model development phase maybe not be acceptable to the users that it achieves such a lower performance on the original task. In contrast, our proposed UMAP can be a more practical and flexible way to restore detection performance. ", + "bbox": [ + 174, + 617, + 825, + 756 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "We conduct the extra comparisons between our UM and UMAP with those methods for reducing overfitting. The results are summarized in the following Tables 4, 5, 6 and 7. According to our extra experiments, most conventional methods proposed to prevent conventional overfitting show limited benefits on gaining better OOD detection performance. Based on our important observation, the effective criterion, i.e., early stopping, also need to change its validation target to be the OOD data. However, most of them suffer from higher sacrifice on the performance of the original task and maybe not compatible and practical in the current general setting, i.e., starting from a well-trained model. ", + "bbox": [ + 174, + 763, + 825, + 861 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "Given the concept discrepancy aforementioned, one conclusive message is that \"memorization of the atypical samples\" are not \"memorization in overfitting\". Those atypical samples are empirically beneficial in improving the performance on the original classification task as shown in Figure 2. However, this part of knowledge is not very necessary and even harmful to the OOD detection task as the detection performance of the model is drop significantly. Based on the training and test curves in our observation, the memorization in overfitting is expected to happen later than the final stage in which the test performance would drop. Since we have already used some strategies to prevent overfitting, it does not exist. Intuitively, the \"atypical samples\" identified in our work are relative to the OOD detection task. The memorization of \"atypical samples\" indicates that the model may not be able to draw the general information of the ID distribution through further learning on those atypical samples through the original classification task Since we mainly provide the empirical observation and understanding of the proposed algorithm in this work, further analysis from other views or theoretically would be an interesting and a major part of future work. ", + "bbox": [ + 176, + 868, + 825, + 924 + ], + "page_idx": 14 + }, + { + "type": "table", + "img_path": "images/f430d358f55a03ab246fefbad236a7c78ca3462d7a9d197666430a7d1e0c2f71.jpg", + "table_caption": [ + "Table 4: Comparison among overfitting methods and ODIN with DenseNet-101 $( \\% )$ ). $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline91.6791.8940.7493.67
Early Stopping w. ACC92.1392.4638.8693.69
Early Stopping w. OOD92.9593.2635.2393.18
Weight Decay 0.186.6486.6760.0788.53
Weight Decay 0.0190.7691.2544.2092.07
Weight Decay 0.00188.9388.2548.9594.26
Drop Rate 0.391.1492.2146.5890.05
Drop Rate 0.484.9586.6262.5282.55
Drop Rate 0.583.7585.1762.1775.31
UM (ours)92.4593.0637.1392.76
UMAP (ours)91.9292.8837.6993.69
", + "bbox": [ + 228, + 137, + 764, + 308 + ], + "page_idx": 15 + }, + { + "type": "table", + "img_path": "images/1d838572164e14370718419838c8f7295c19a674645d6d63804cae4c92da9386.jpg", + "table_caption": [ + "Table 5: Comparison among overfitting methods and Energy with DenseNet-101 $( \\% )$ . $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline92.7293.4838.3093.67
Early Stopping w. ACC92.7593.5437.8493.69
Early Stopping w. OOD93.0493.7836.5693.18
Weight Decay 0.186.7888.0465.0888.53
Weight Decay 0.0190.8691.7747.6492.07
Weight Decay 0.00190.6890.9047.3894.26
Drop Rate 0.390.5291.7951.2390.05
Drop Rate 0.484.2986.4368.1782.55
Drop Rate 0.583.2985.1468.1775.31
UM (ours)93.5894.1433.6692.76
UMAP (ours)93.1793.8736.1193.69
", + "bbox": [ + 228, + 366, + 764, + 535 + ], + "page_idx": 15 + }, + { + "type": "table", + "img_path": "images/fb5b51d30e40d5894a6888f9d323b424b639886b2073340516d18dddee62178d.jpg", + "table_caption": [ + "Table 6: Comparison among overfitting methods and ODIN with WRN-40-4 $( \\% )$ . $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline86.2485.9060.1393.86
Early Stopping w. ACC83.8083.3065.1393.99
Early Stopping w. OOD89.9290.9449.8791.47
Weight Decay 0.184.3884.7565.7589.88
Weight Decay 0.0188.0888.4555.1693.16
Weight Decay 0.00186.3486.3857.4294.91
Drop Rate 0.387.5387.2556.1294.22
Drop Rate 0.488.2488.4154.6294.20
Drop Rate 0.589.1389.9953.0793.91
UM (ours)89.6191.1350.9792.68
UMAP (ours)90.4391.7346.9693.86
", + "bbox": [ + 228, + 594, + 764, + 763 + ], + "page_idx": 15 + }, + { + "type": "text", + "text": "", + "bbox": [ + 173, + 797, + 825, + 924 + ], + "page_idx": 15 + }, + { + "type": "table", + "img_path": "images/6ddca4b36c8e630d21ddb3e64e00889c3562f9ae052ec1b20ad4eb06c9b47932.jpg", + "table_caption": [ + "Table 7: Comparison among overfitting methods and Energy with WRN-40-4 $( \\% )$ . $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline87.6988.1658.4793.86
Early Stopping w. ACC88.0788.6567.6193.99
Early Stopping w. OOD90.9291.9446.6391.47
Weight Decay 0.186.9788.5163.5489.88
Weight Decay 0.0189.7789.8250.2393.16
Weight Decay 0.00189.2589.8450.9593.91
Drop Rate 0.389.7490.0752.1693.22
Drop Rate 0.489.9490.5351.1394.20
Drop Rate 0.590.0991.0452.7693.91
UM (ours)91.7492.6740.4092.68
UMAP (ours)88.8489.3150.2393.86
", + "bbox": [ + 228, + 137, + 764, + 308 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "D DETAILED REALIZATION OF THE PROPOSED ALGORITHMS ", + "text_level": 1, + "bbox": [ + 173, + 333, + 687, + 348 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "In this section, we provide the detailed realization of our proposed Unleashing Mask (UM) (i.e., in Algorithm 1) and Unleashing Mask Adopt Pruning (UMAP) (i.e., in Algorithm 2). ", + "bbox": [ + 173, + 364, + 823, + 393 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "To estimate the loss constrain $\\zeta$ (i.e. $\\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } )$ in Eq 3 with the fixed given model $f ^ { * }$ ) for forgetting, we need to randomly knock out parts of weights according to the given mask ratio $\\delta$ . To be specific, we sample a score from a Gaussian distribution for every weight. Then we initialize a unit matrix for every layer of the model concerning the size of the layer. We formulate the mask $m _ { \\delta }$ according to the sampled scores. Find the threshold for each layer that is smaller than the score of the given mask ratio in that layer (termed as quantile). Then set all the ones, whose corresponding scores are more significant than the layers’ thresholds, to zeros. In our algorithms, the fine-tuning epochs $k$ is the epochs we fine-tune after we get the well-trained model. ", + "bbox": [ + 174, + 401, + 825, + 513 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "We dot-multiply every layer’s weights with the formulated binary matrix as if we delete some parts of the weights. We input a batch of training samples to the masked model and treat the mean value of the outputs’ cross-entropy loss as the loss constraint. After all of these have been done, we begin to fine-tune the model’s weights with the loss constraint applied to the cross-entropy loss. ", + "bbox": [ + 174, + 521, + 825, + 577 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "For UMAP, the only difference from UM is that, instead of fine-tuning the weights, we generate a popup score for every weight, and force the gradients to pass through the scores. In every iteration, we need to formulate a binary mask according to the given prune rate $p$ . This is just what we do when estimating the loss constraint. For more details, it can refer to (Ramanujan et al., 2020). In Table 8, we summarize the overall comparison results of UM and UMAP to show their effectiveness. ", + "bbox": [ + 174, + 583, + 825, + 654 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "E ADDITIONAL SETUPS OF THE EXPERIMENTS ", + "text_level": 1, + "bbox": [ + 174, + 676, + 570, + 691 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "In this section, we describe more details about the experimental setups for our exploration. ", + "bbox": [ + 171, + 708, + 764, + 723 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "Model setups. For DenseNet-101, we fix the growth rate and reduce rate to 12 and 0.5 respectively with the bottleneck block included in the backbone. We also explore the proposed UM on WideResNet (Zagoruyko & Komodakis, 2016) with 40 depth and 4 widen factor, which is termed as WRN-40-4. The batch size for both ID and OOD testing samples is 64, and the batch size of auxiliary samples is 2000. The $\\lambda$ in Eq.(9) is 0.5 to keep the OE loss comparable to the CE loss. As for the strategy of sampling outliers, we randomly retrieve 50000 samples from ImageNet-1k (Deng et al., 2009) for OE and Energy (w. $D _ { \\mathrm { a u x . } }$ ) and 50000 samples using Thompson sampling (Ming et al., 2022) for POEM. ", + "bbox": [ + 174, + 739, + 825, + 838 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "Learning rate schedules. We use 4 different learning rate schedules to demonstrate the existence of the once-covered OOD detection capability. For cosine annealing, we follow the common setups in Loshchilov & Hutter (2017); for linear schedule, the learning rate remains the same in the first one-third epochs, decreases linearly to the tenth of the initial rate in the middle one-third epochs, and decrease linearly to $1 \\%$ of the initial rate in the last one-third epochs; for the multiple decay ", + "bbox": [ + 174, + 854, + 825, + 924 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "Algorithm 1 Unleashing Mask (UM) ", + "text_level": 1, + "bbox": [ + 174, + 103, + 419, + 118 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "Input: well-trained model : $\\theta$ , mask ratio: $\\delta \\in [ 0 , 1 ]$ , fine-tuning epochs of UM: $k$ , training samples: $x \\sim D _ { \\mathrm { i n } } ^ { \\mathrm { s } }$ ; ", + "bbox": [ + 171, + 121, + 823, + 150 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "Output: fine-tuned model $\\hat { \\theta }$ ", + "bbox": [ + 174, + 151, + 364, + 166 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "1: Initialize a popup score for every weight \n2: for $w \\in { \\boldsymbol { \\theta } }$ do \n3: $s _ { w } \\sim N ( \\mu , \\sigma ^ { 2 } )$ \n4: end for \n5: Generate mask by the popup scores \n6: for l ∈ θlayers do \n7: $m _ { \\delta } ^ { l } = \\dot { s } ^ { l } >$ quantile $( s ^ { l } , \\delta )$ \n8: end for \n9: Estimate loss constrain \n10: $\\zeta = \\mathbb { E } _ { x \\sim D _ { \\mathrm { i n } } ^ { s } } ( \\hat { \\mathcal { L } } _ { \\mathrm { C E } } ( x , m _ { \\delta } \\odot \\theta ) )$ \n11: Unleashing Mask: fine-tuning \n12: for $t \\in ( 1 , \\ldots , k )$ do \n13: θ(t+1) = θ(t) − η ∂(|LCE(x,θ)−ζ|+ζ) \n14: end for ", + "bbox": [ + 179, + 169, + 477, + 387 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "Algorithm 2 Unleashing Mask Adopt Pruning (UMAP) ", + "text_level": 1, + "bbox": [ + 174, + 405, + 545, + 420 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "Input: well-trained model : $\\theta$ , mask ratio: $\\delta \\in [ 0 , 1 ]$ , fine-tuning epochs of UM: $k$ , training samples: $x \\sim D _ { \\mathrm { i n } } ^ { \\mathrm { s } }$ , prune rate: $p$ ; ", + "bbox": [ + 173, + 422, + 823, + 452 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "Output: learnt binary mask $\\hat { m } _ { p }$ ; ", + "bbox": [ + 174, + 452, + 390, + 465 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "1: Initialize a popup score for every weight \n2: for $w \\in { \\boldsymbol { \\theta } }$ do \n3: $s _ { w } \\sim N ( \\mu , \\sigma ^ { 2 } )$ \n4: end for \n5: Generate mask by the popup scores \n6: for $l \\in \\theta _ { \\mathrm { l a y e r s } }$ do \n7: $m _ { \\delta } ^ { l } = s ^ { l } > \\mathrm { q u a n t i l e } ( s ^ { l } , \\delta )$ \n8: end for \n9: Estimate loss constrain \n10: $\\zeta = \\mathbb { E } _ { x \\sim D _ { \\mathrm { i n } } ^ { s } } ( \\hat { \\mathcal { L } } _ { \\mathrm { C E } } ( x , m _ { \\delta } \\odot \\theta ) )$ \n11: Unleashing Mask Adopt Pruning: fine-tuning \n12: $s ^ { ( 1 ) } \\sim N ( \\mu , \\sigma ^ { 2 } )$ \n13: for $t \\in ( 1 , \\ldots , k )$ do \n14: for $l \\in \\theta _ { \\mathrm { l a y e r s } }$ do \n15: $\\hat { m } _ { p } ^ { l } = \\dot { s ^ { l ( t ) } } > \\mathrm { q u a n t i l e } ( s ^ { l ( t ) } , p )$ \n16: end for \n17: $\\begin{array} { r } { s ^ { ( t + 1 ) } = s ^ { ( t ) } - \\eta \\frac { \\partial ( | \\mathcal { L } _ { \\mathrm { C E } } ( x , \\hat { m } _ { p } \\odot \\theta ) - \\zeta | + \\zeta ) } { \\partial \\theta } } \\end{array}$ \n18: end for ", + "bbox": [ + 178, + 468, + 511, + 743 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "schedule, the learning rate decreases $1 0 \\%$ of the initial rate (0.01) every $1 0 \\%$ epochs (10 epochs); for the multiple step schedule, the learning rate decreases to $1 0 \\%$ of the current rate every 30 epochs. All those learning rate schedules for our experiments are intuitively illustrated in Figure 4(a). ", + "bbox": [ + 174, + 772, + 825, + 814 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "F ADDITIONAL EXPERIMENT RESULTS AND ABLATION STUDY ", + "text_level": 1, + "bbox": [ + 174, + 837, + 697, + 852 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "In this section, we provide more experiment results. We first show the fine-grained results on CIFAR10 and CIFAR-100, then conduct the experiments under a different model structure (i.e., WRN-40-4), and finally apply an additional ablation study on the proposed UM and UMAP. The mean and variance of all metrics (ID-ACC, AUROC, AUPR, FPR95) are reported based on multiple independent trials. ", + "bbox": [ + 174, + 867, + 826, + 924 + ], + "page_idx": 17 + }, + { + "type": "table", + "img_path": "images/062ca41fb7ee42b648df09bd68a14fa3e91b58cf9f5c49dbf4812e0737189050.jpg", + "table_caption": [ + "Table 8: Completed Results $( \\% )$ . Comparison with competitive OOD detection baselines. $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinMethodAUROC↑AUPR个FPR95↓ID-ACC↑w./w.o Daux
CIFAR-10MSP(Hendrycks & Gimpel, 2017a)89.90 ±0.30 91.46 ± 0.5691.48 ± 0.43 91.67 ± 0.5860.08 ±0.76 42.31 ± 1.3894.01 ± 0.08 94.01 ± 0.08
ODIN(Liang et al.,2018)
Mahalanobis(Lee et al.,2018b)75.10 ± 1.0472.32 ± 1.9261.35 ± 1.2594.01 ± 0.08
Energy(Liu et al.,2020)92.07±0.2292.72 ± 0.3942.69 ± 1.3194.01 ± 0.08
Energy+UM(ours)93.73±0.3694.27 ± 0.6033.29 ± 1.7092.80±0.47
Energy+UMAP(ours)93.97 ± 0.1194.38 ± 0.0630.71 ± 1.9494.01 ± 0.08
OE(Hendrycks et al.,2019b)97.07 ± 0.0197.31± 0.0513.80 ± 0.2892.59 ± 0.32
Energy (w.Daux)(Liu et al.,2020)94.58 ± 0.6494.69 ± 0.6518.79 ± 2.3180.91 ± 3.13>>>>
POEM(Ming et al.,2022) OE+UM(ours)94.37 ± 0.0794.51± 0.0618.50 ± 0.3377.24 ± 2.22
Energy+UM(ours)97.60 ± 0.03 93.02±0.4297.87 ± 0.02 92.36 ± 0.3811.22 ± 0.16 24.41 ± 1.6593.66 ± 0.12 71.97 ± 0.92
POEM+UM(ours)
OE+UMAP(ours)93.04± 0.0292.99 ± 0.0223.52 ±0.1667.41 ± 0.27
Energy+UMAP(ours)95.63 ± 1.1597.48 ± 0.0197.74±0.00 95.92 ± 1.1712.21 ± 0.09 17.51 ± 2.5993.44± 0.21>>>>
POEM+UMAP (ours)94.18 ± 2.9894.15 ± 3.4620.55± 8.7088.12 ± 4.22 76.62 ± 17.95√
MSP(Hendrycks & Gimpel,2017a)74.06 ±0.6975.37 ± 0.7383.14 ± 0.8774.86 ± 0.21
ODIN(Liang et al.,2018)76.18 ± 0.1476.49 ± 0.2078.93 ± 0.3174.86 ± 0.21
Mahalanobis(Lee et al.,2018b)63.90 ± 1.9164.31 ± 0.9178.79 ± 0.5074.86 ± 0.21
Energy(Liu et al.,2020)76.29 ±0.2477.06± 0.5578.46 ± 0.0674.86 ± 0.21
Energy+UM(ours)76.22 ± 0.4276.39 ±1.0374.05 ± 0.5564.55 ± 0.24
Energy+UMAP(ours)75.57 ± 0.5975.66 ± 0.0772.21 ± 1.4674.86 ± 0.21
OE(Hendrycks et al.,2019b)90.55 ± 0.8790.34± 0.9434.73 ±3.8573.59 ± 0.30
Energy(w.Daux)(Liu et al.,2020)88.92±0.5789.13± 0.5637.90 ± 2.5957.85± 2.65>>>>
POEM(Ming et al.,2022)88.95 ±0.5488.94± 0.3138.10 ±1.3056.18 ± 1.92
91.04± 0.1191.13 ± 0.2434.71 ± 0.8175.15 ± 0.18
OE+UM(ours)90.14± 0.4532.65 ± 3.13
Energy+UM(ours)90.39 ± 0.40 91.18 ± 0.3591.45 ± 0.2730.78 ± 1.7671.95 ± 0.23 70.17 ± 0.01
POEM+UM(ours) OE+UMAP (ours)91.10 ± 0.1690.99±0.2333.62±0.2674.76 ± 0.11
Energy+UMAP(ours)90.52 ±0.2690.46 ±0.5032.17 ± 0.3072.76 ± 0.18√
POEM+UMAP (ours)91.10 ±0.2991.41 ± 0.2831.02 ± 1.7071.05 ± 0.04√
", + "bbox": [ + 174, + 137, + 825, + 478 + ], + "page_idx": 18 + }, + { + "type": "text", + "text": "Empirical verification on typical/atypical data. In the following Tables 9, 10, 11, 12, and 13, we further conduct the experiments to identify the negative effect of learning on those atypical samples by comparing with a counterpart that learning only with the typical samples. The results confirm that the degeneration on detection performance is more likely to come from learning atypical samples. ", + "bbox": [ + 174, + 515, + 825, + 570 + ], + "page_idx": 18 + }, + { + "type": "text", + "text": "In Table 9, we provide the main results for verification on typical/atypical samples. Intuitively, we intend to separate the training dataset into a typical set and an atypical set and train respectively on these two sets to see whether it is learning atypical samples that causes OOD performance to decrease during the latter part of the training phase. We force training samples through the model (DenseNet-101) of the 60th epoch and get the CE loss for separation. We provide the ACC of the generated sets on the model of the 60th epoch (ACC in the tables). The extremely low ACCs of the atypical sets show that the model of the 60th epoch can hardly predict the samples, which meets our definition of the atypical sample. We then fine-tune the model of the 60th epoch with the generated dataset and report the OOD performance. The results show learning from only those atypical data fail to gain better detection performance than its counterpart, i.e., learning from only those typical data. Learning on those atypical samples fails to draw the suitable features for the OOD detection task, though it still can improve the original task performance. The experiments provide a conceptual verification of our conjecture which links our observation and the proposed method. ", + "bbox": [ + 174, + 578, + 825, + 757 + ], + "page_idx": 18 + }, + { + "type": "text", + "text": "Results of fine-tuning for less epochs. UM adopts finetuning on the proposed objective for forgetting has shown the advantages of being cost-effective compared with train-from-scratch. For the tuning epochs, we show in Figures 9 and 10 that fine-tuning using UM can converge within about 20 epochs, indicating that we can apply our UM/UMAP for far less than 100 epochs (compared with train-from-scratch) to restore the better detection performance of the original well-trained model. It is intuitively reasonable that finetuning with the newly designed objective would benefit from the well-trained model, allowing a faster convergence since the two phases consider the same task with the same training data. As for the major experiments conducted in our work, finetuning adopts 100 epochs for better exploring and understanding its learning dynamics for research purposes, this configuration is indicated in the training details of Section 4.1. ", + "bbox": [ + 174, + 785, + 825, + 924 + ], + "page_idx": 18 + }, + { + "type": "table", + "img_path": "images/5aed1946c9645f06bcccbbf0abdca6c6cad363a926c926ad7e6d449c9e74e6fc.jpg", + "table_caption": [ + "Table 9: Fine-tuning on typical/atypical samples with different model structures $( \\% )$ . $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinDataset SizeStructureAtypical/TypicalAUROC↑AUPR↑FPR95↓
CIFAR-10200DenseNet-101Atypical Typical81.45 82.8682.40 84.3862.10 60.01
WRN-40-4Atypical Typical85.13 86.2686.57 86.8966.41
DenseNet-101Atypical Typical71.96 74.7973.16 75.8359.93 85.57
CIFAR-1001000WRN-40-4Atypical Typical66.64 71.9567.41 72.0280.97 86.92 80.00
", + "bbox": [ + 176, + 137, + 825, + 295 + ], + "page_idx": 19 + }, + { + "type": "table", + "img_path": "images/41341d0a19719a641a9201de0fb28a0d33c42b7e5f9abb12cd8ce42a17e3219c.jpg", + "table_caption": [ + "Table 10: Fine-tuning on typical/atypical CIFAR-10 samples with DenseNet-101 $( \\% )$ . $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinDataset SizeAtypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-10200Atypical Typical3.50 100.0081.45 82.8682.40 83.4862.10 60.01
350Atypical11.14 100.0085.9086.0155.10
500Typical Atypical16.8085.90 84.9486.16 85.3352.81 59.27
Typical100.0085.5386.1058.74
", + "bbox": [ + 178, + 345, + 825, + 481 + ], + "page_idx": 19 + }, + { + "type": "table", + "img_path": "images/04631571d0fae549eb9e4a58dda2f7a5e481b5a024a6e1c3734b8ec291637d1c.jpg", + "table_caption": [ + "Table 11: Fine-tuning on typical/atypical CIFAR-10 samples with WRN-40-4 $( \\% )$ . $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinDataset SizeAtypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-10200Atypical Typical3.50 100.0085.13 86.2686.57 86.8966.41 59.93
350Atypical11.1482.9284.2468.57
Typical Atypical100.00 16.8085.8287.8465.54
500Typical100.0082.88 87.3883.22 87.9366.75 52.27
", + "bbox": [ + 178, + 534, + 825, + 667 + ], + "page_idx": 19 + }, + { + "type": "table", + "img_path": "images/3b7899f69d285eacdfc1b81f218d626770641869566f8065f253fcbd065face8.jpg", + "table_caption": [ + "Table 12: Fine-tuning on typical/atypical CIFAR-100 samples with DenseNet-101 $( \\% )$ . ↑ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinDataset Size Atypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-100500Atypical Typical1.00 100.0072.69 74.0773.28 75.2080.71 80.19
800Atypical2.8869.7471.1585.46
Typical Atypical100.00 3.5072.49 71.9673.17 73.1681.97
1000Typical100.0074.7975.8385.57 80.97
", + "bbox": [ + 178, + 719, + 825, + 853 + ], + "page_idx": 19 + }, + { + "type": "text", + "text": "We also provide an extra comparison to show the relative efficiency of our proposed UM/UMAP in the following Table 14 and Table 15. The results show that UM and UMAP can efficiently restore detection performance compared with the baseline. It is intuitively reasonable that fine-tuning would benefit from the well-trained model, allowing a faster convergence as the two phases consider the same task and training data. Considering the significance of the OOD awareness for those safetycritical areas, it is worthwhile to further excavate the OOD detection capability of the deployed well-trained model using our UM and UMAP. ", + "bbox": [ + 174, + 882, + 825, + 924 + ], + "page_idx": 19 + }, + { + "type": "table", + "img_path": "images/0e79c48e21bf60c61eb8bca60e53b9117c40a5168d1316ab43e92c5ae5305b26.jpg", + "table_caption": [ + "Table 13: Fine-tuning on typical/atypical CIFAR-100 samples with WRN-40-4 $( \\% )$ . $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinDataset Size Atypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-100500Atypical Typical1.00 100.0066.03 68.6066.17 69.9389.56 86.53
800Atypical2.8867.5968.6685.61
1000Typical Atypical100.00 3.5070.25 66.6468.95 67.4179.66 86.92
Typical100.0071.9572.0280.00
", + "bbox": [ + 176, + 137, + 825, + 271 + ], + "page_idx": 20 + }, + { + "type": "text", + "text": "", + "bbox": [ + 173, + 299, + 825, + 354 + ], + "page_idx": 20 + }, + { + "type": "table", + "img_path": "images/809c900b629a6946687476f96bee55ce8ee51dbace08c9351de633a026c6f197.jpg", + "table_caption": [ + "Table 14: Fine-tuning for 20 epochs with DenseNet-101 $( \\% )$ . $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinEpochMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10100MSP89.9091.4860.0894.01
ODIN91.4691.6742.3194.01
Energy92.0792.7242.6994.01
Energy +UM93.7394.2733.2992.80
Energy + UMAP93.9794.3830.7194.01
MSP +UM90.3191.9953.6191.70
20ODIN +UM94.0894.6731.0191.70
Energy +UM93.6094.3233.0391.70
MSP+UMAP88.7090.3957.6994.01
ODIN+UMAP92.8893.3335.19
Energy + UMAP92.8893.3935.6094.01
94.01
", + "bbox": [ + 220, + 405, + 774, + 580 + ], + "page_idx": 20 + }, + { + "type": "table", + "img_path": "images/7aa54b6d164b551fcc233f0d1a80763b50c75dd589734ab7a9c8593a303ca70f.jpg", + "table_caption": [ + "Table 15: Fine-tuning for 20 epochs with WRN-40-4 $( \\% )$ . $\\uparrow$ indicates higher values are better, and ↓ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinEpochMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10100MSP87.1287.8468.2993.86
ODIN83.2982.7465.6893.86
Energy87.6988.1658.4793.86
Energy + UM91.7492.6740.4092.68
Energy + UMAP88.8489.3150.2393.86
MSP +UM89.8691.3251.6291.96
20ODIN +UM91.9792.5841.7891.96
Energy +UM92.9593.6436.2191.96
MSP+UMAP88.7790.6161.6093.86
ODIN + UMAP90.8591.8945.7093.86
Energy + UMAP91.6692.4942.9493.86
", + "bbox": [ + 220, + 640, + 774, + 815 + ], + "page_idx": 20 + }, + { + "type": "text", + "text": "Fine-grained results on OOD data. In order to further figure out the effectiveness of the proposed UM and UMAP on different OOD datasets, we further report the fine-grained results of our experiments on CIFAR-10 and CIFAR-100 with 6 OOD datasets (CIFAR-10/CIFAR-100, textures, Places365, SUN, LSUN, iNaturalist). The results on the 6 OOD datasets show the general effectiveness of the proposed UM as well as UMAP. In Table 16, $\\mathbf { O E + U M }$ can outperform all the OOD baselines, and further improve the OOD performance even though the original detection performance is already well. By equipping with our proposed UM and UMAP, the baselines can outperform their counterparts on most of the OOD datasets. For instance, the FPR95 can decrease from 1.91 to 1.42. In Table 17, we also take a closer check about results on CIFAR-100 with 6 OOD datasets. Our proposed method can almost improve all competitive baselines (either the scoring functions or the finetuning with auxiliary outliers) on the 6 OOD datasets. In both w. $D _ { a u x }$ and w.o. $D _ { a u x }$ scenarios, Unleashing Mask can significantly excavate the intrinsic OOD detection capability of the model. In addition to unleashing the excellent OOD performance, UMAP can also maintain the high ID-ACC by learning a binary mask instead of tuning the well-trained original parameters directly. ", + "bbox": [ + 173, + 839, + 826, + 924 + ], + "page_idx": 20 + }, + { + "type": "table", + "img_path": "images/51c49194022a3f450f05525beb0e80c2781701dee1b944e1944f81d7233e6b6e.jpg", + "table_caption": [ + "Table 16: Fine-grained Results $( \\% )$ of DenseNet-101 on CIFAR-10. Comparison on different OOD benchmark datasets respectively. $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
ID datasetMethodOOD dataset TexturesPlaces365
CIFAR-100 FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
Energy+UMAP54.95 ± 2.6187.72 ±1.0533.59 ±1.3292.67 ±0.2332.80 ± 4.1493.57 ± 1.12
OE59.29 ± 1.3088.51 ±0.222.89 ± 0.3099.16 ±0.0511.14 ± 1.1197.50 ±0.19
Energy (w. Dauz)79.88 ± 2.4774.99 ± 2.404.27 ± 0.5798.80 ±0.1914.22 ± 3.9997.07 ±0.72
POEM82.30 ± 1.5772.74 ± 1.421.91 ± 0.4199.40 ±0.1011.24 ± 2.7096.67 ±0.48
OE+UM(ours)55.74 ± 1.4789.53 ± 0.181.42 ± 0.1599.49± 0.047.77 ±0.6998.15 ± 0.08
Energy (w.Daux)+ UM(ours)84.52 ± 0.0170.09 ± 0.478.30 ±0.8897.76 ±0.0520.27 ±1.3096.06 ±0.31
POEM+UM(ours) OE+UMAP(ours)84.87 ± 1.5668.97 ±0.394.73 ± 0.5298.88 ±0.1319.83 ± 0.3496.35 ±0.09
Energy(w.Daur)+UMAP(ours)59.05 ± 1.4189.14 ± 0.141.86 ± 0.0799.35 ±0.008.21±0.1298.07 ±0.03
POEM+UMAP(ours)75.18 ± 4.96 79.33 ± 4.1480.93 ± 4.49 76.89 ± 4.862.24 ± 1.34 2.10 ± 1.3799.25 ±0.29 99.34 ±0.309.30 ± 2.12 9.94 ± 6.9297.90 ±0.40 98.01 ± 1.11
OEMethodSUNLSUNiNaturalist
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
Energy+UMAP29.05 ±2.7894.41 ± 0.732.31± 0.8899.42 ± 0.0447.22 ± 14.0392.63 ± 2.14
8.38±0.7198.00 ±0.145.90 ± 1.4398.60± 0.215.09 ± 0.6498.76 ± 0.11
Energy (w. Dauz)10.30 ±3.8297.77 ± 0.6412.80 ± 4.6796.08 ± 1.386.93 ± 1.8698.40 ±0.32
POEM8.39±2.4298.16 ±0.439.69 ± 1.8997.25±0.583.78 ±0.9098.99 ±0.17
OE+UM(ours)5.51 ± 0.4498.55 ± 0.073.51 ± 0.4398.93 ± 0.09 92.11 ± 0.942.87 ± 0.4999.14 ± 0.09
Energy(w.Daux)+ UM(ours) POEM+UM(ours)16.13 ± 1.86 16.16 ± 0.5796.84±0.30 97.01±0.0823.27 ± 2.40 25.69 ± 0.1593.38 ± 0.2711.20 ± 2.35 9.30 ± 1.6097.54±0.42 98.05 ±0.24
OE+UMAP(ours)6.16 ±0.0298.49 ± 0.014.53 ± 0.1698.86±0.063.40 ± 0.7498.96 ±0.09
Energy(w.Dauz)+UMAP(ours)6.67 ± 1.5098.40±0.3223.50 ± 5.6194.78 ± 2.043.77 ± 2.1498.93 ±0.46
94.74 ±3.67
POEM+UMAP(ours)7.00±5.8098.46±0.9621.17 ± 12.843.63±2.7898.99 ±0.54
", + "bbox": [ + 176, + 155, + 825, + 383 + ], + "page_idx": 21 + }, + { + "type": "table", + "img_path": "images/bbcdb0726730fb08f597974785e84bf8ec7f9af7e92c5725afbf29197e39ed92.jpg", + "table_caption": [ + "Table 17: Fine-grained Results $( \\% )$ of DenseNet-101 on CIFAR-100. Comparison on different OOD benchmark datasets respectively. $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
ID datasetMethodOOD datasetPlaces365
CIFAR-10 FPR95↓AUROC↑Textures FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-100MSP ODIN Mahalanobis Energy83.53± 0.33 85.29 ± 0.1775.11 ± 0.27 73.31 ± 0.24 49.60 ± 1.5186.90 ±0.18 86.45 ± 1.27 33.06± 3.7671.45 ± 0.40 71.91 ± 0.2785.83 ±0.48 84.35 ± 0.6470.54 ± 0.42 73.58 ± 0.51
98.25 ±0.0590.19 ± 1.2195.20 ±0.4953.69 ± 1.55
Energy+UM(ours)82.16 ± 0.5975.31 ± 0.2190.20 ±0.3068.98 ±0.34 65.39 ± 1.4482.39 ±0.97 77.30 ± 2.0873.78 ±0.66 76.06 ± 1.05
OE89.62 ±0.0766.12 ± 0.9386.99 ± 1.22
Energy (w.Daux) POEM90.97 ±0.46 96.14 ±0.0669.02 ± 0.39 57.52 ± 0.8814.36 ± 0.25 9.02 ±0.0695.92 ±0.12 97.16 ±0.2640.19 ± 6.97 35.18 ± 4.7390.70 ± 2.01 93.29 ± 1.14
OE+UM (ours)96.19 ±0.1655.82 ± 1.057.63 ± 1.4097.69 ±0.0632.67 ± 3.7393.94±0.68
89.61±0.0871.24 ± 0.0816.78 ± 0.2595.60 ±0.0839.77 ± 0.3491.07 ± 0.13
Energy(w.Daux)+UM(ours)95.38 ± 0.4563.41 ± 0.146.41 ± 0.8397.77 ± 0.3330.96 ± 3.6192.85 ±0.69
POEM+UM(ours)95.78 ± 0.1460.23 ± 0.705.17 ±0.1898.53± 0.0323.90 ±0.8495.45 ± 0.11
OE+UMAP(ours) Energy(w.Daux)+UMAP(ours)90.72±0.3569.76 ± 0.2515.32 ± 0.2395.72 ± 0.0136.42 ± 1.9192.08 ±0.49
POEM+UMAP(ours)95.39±0.10 95.69 ± 0.1763.26 ±0.18 61.62 ± 0.246.52 ±0.44 5.23 ± 0.5897.83 ±0.18 98.52 ± 0.0131.18 ± 0.43 26.06 ± 1.1693.13±0.41 94.91 ± 0.25
MethodSUNLSUN
FPR95↓ AUROC↑FPR95↓AUROC↑iNaturalist FPR95↓AUROC↑
MSP88.75 ±0.2366.75 ± 0.2567.83 ±1.3782.94± 0.3285.00 ± 0.7376.62 ± 0.25
ODIN88.49 ± 0.9969.64± 0.61 54.37 ± 1.3534.80 ± 2.55 89.31 ± 4.8393.92 ±0.75 43.19 ± 16.3681.67 ± 2.77 93.63 ± 1.1978.36 ± 1.57
Mahalanobis95.53±0.3749.60 ± 1.51
Energy Energy+UM(ours)97.17 ± 0.92 69.04 ± 0.8335.09 ± 3.1793.49 ±0.8785.70 ± 2.1475.82 ± 1.72
81.96 ± 2.2671.47 ± 1.8822.54± 5.9394.98 ± 1.7574.28 ± 3.7280.72 ± 3.75
OE44.47 ± 9.10 90.70 ± 2.015.75 ± 1.1898.57 ± 0.1325.51 ± 4.1294.46 ± 0.88
Energy (w. Daux)32.69 ± 5.6993.63 ± 1.4855.75 ± 4.3187.96 ± 1.0317.34 ± 4.5496.50 ±0.81
POEM30.45 ± 5.1194.26 ± 0.9046.68 ± 3.5990.30 ± 2.1716.50 ± 2.0996.63±0.23
OE+UM(ours)44.23±0.20 90.28±0.035.80 ±0.3398.63 ±0.0326.72 ± 1.9594.51 ± 0.43
93.18 ± 0.6991.98 ± 0.8310.83 ± 2.0697.09 ±0.69
Energy(w.Daux)+UM(ours) POEM+UM(ours)28.98 ±3.1537.56 ± 4.8194.43 ± 1.18
21.34 ± 1.0795.76 ±0.2233.74 ±6.2298.64 ± 0.128.85±0.2397.93 ± 0.03 95.08 ±0.22
OE+UMAP(ours) Energy(w.Daux)+UMAP(ours)39.58 ± 2.0291.37 ± 0.52 93.38 ± 0.195.77 ± 0.7192.08 ±0.3923.33 ± 1.24 13.96 ± 2.4896.87 ± 0.15
29.65 ± 1.0635.94 ± 0.7593.57 ±0.88
POEM+UMAP(ours)23.73 ± 0.7195.25 ± 0.0633.09 ± 5.949.76 ± 1.0997.77 ± 0.24
", + "bbox": [ + 178, + 473, + 825, + 765 + ], + "page_idx": 21 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 811, + 825, + 924 + ], + "page_idx": 21 + }, + { + "type": "table", + "img_path": "images/4bcf6ecf1f82bd834d5a6547f5e9543455c4aec84f74493ab4a1e5067ca38c27.jpg", + "table_caption": [ + "Table 18: Results $( \\% )$ of WRN-40-4. Comparison with competitive OOD detection baselines. We respectively train WRN-40-4 on CIFAR-10 and CIFAR-100. For those methods involving outliers, we retrieve 5000 samples from ImageNet-1k. $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10MSP(Hendrycks & Gimpel,2017a)87.12 ± 0.2587.84± 0.3068.29 ±0.9693.86 ± 0.19
ODIN(Liang et al.,2018)83.29 ±0.7282.74±0.7965.68 ± 0.7793.86 ± 0.19
Mahalanobis(Lee et al., 2018b)77.57 ± 0.2876.11 ±0.1061.18 ±0.1093.86 ± 0.19
Energy(Liu et al.,2020)87.69 ± 0.5488.16 ±0.6958.47 ± 1.9493.86 ± 0.19
Energy+UM(ours)91.74 ± 0.4392.67 ± 0.5240.40±1.3292.68±0.23
Energy+UMAP(ours)88.84± 1.0289.31 ± 1.4450.23 ±2.2593.86 ± 0.19
CIFAR-100MSP(Hendrycks & Gimpel,2017a)72.34± 0.6372.69 ± 0.4485.40 ±0.5975.01 ±0.07
ODIN(Liang et al., 2018)68.78±0.6766.92 ± 0.7285.28 ± 0.6475.01 ± 0.07
Mahalanobis(Lee et al., 2018b)68.20±0.9968.30 ± 1.1576.46 ± 2.0275.01 ± 0.07
Energy(Liu et al.,2020)74.00 ± 0.4173.02 ± 0.4781.37 ±0.0875.01 ± 0.07
Energy+UM(ours)76.07 ± 0.0476.94 ± 0.0674.29 ± 1.6659.08 ± 2.75
Energy+UMAP(ours)77.35 ± 0.7877.43 ± 0.9168.20 ± 0.0675.01 ±0.07
", + "bbox": [ + 176, + 165, + 825, + 335 + ], + "page_idx": 22 + }, + { + "type": "table", + "img_path": "images/5ce55c54c2423aee8aef57faccdc3d8b2906c7115943d3777e81a97a06580fd9.jpg", + "table_caption": [ + "Table 19: Fine-grained Results $( \\% )$ of WRN-40-4 on CIFAR-10. Comparison on different OOD benchmark datasets. $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
ID datasetMethodOOD dataset
CIFAR-100TexturesPlaces365
MSPFPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-10ODIN70.96 ±0.7086.08 ±0.0868.81 ± 1.2986.53 ±0.8368.31 ± 0.2586.71 ±0.13
64.97 ±0.0883.36 ±0.1166.86 ±2.2481.34 ± 0.8166.49 ±1.1683.47 ±0.93
Mahalanobis79.84±0.5570.33 ± 0.2422.56 ± 0.0894.07 ± 0.0485.09 ± 0.5967.90 ± 0.37
Energy61.09±0.5886.66 ±0.0464.29 ± 1.7285.56 ±0.5355.32 ± 0.1388.29 ±0.26
Energy+UM(ours)57.21 ± 1.4187.56± 0.1546.49 ± 1.0389.74±0.4540.68±4.4692.51± 0.97 90.43± 0.47
Energy+UMAP(ours)65.45 ± 1.1084.65 ±0.9559.14 ± 1.6485.27 ± 1.7448.16 ± 1.89
MethodSUNLSUNiNaturalist
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
MSP68.62 ±0.5086.95 ±0.2352.97± 3.0792.41 ±0.1876.05 ± 0.0183.44±0.36
ODIN65.47 ±0.7883.79 ± 1.1031.89 ± 3.4494.34 ±0.8979.28 ±0.1879.80 ±0.35
Mahalanobis82.92 ±0.2870.52 ± 0.4764.31 ± 0.5767.75 ±0.5581.50 ± 2.9174.97 ± 2.91
Energy54.88 ±0.1888.67±0.3024.99 ± 1.3895.98± 0.3775.89 ± 0.8582.40±0.22
Energy+UM(ours)38.92 ± 3.4692.98 ± 0.958.38 ±0.7798.18 ± 0.1666.02 ± 6.7085.22 ±3.42
Energy+UMAP(ours)45.94 ± 2.6491.27±0.5614.10 ± 0.0497.46 ± 0.1474.69±0.1581.13 ± 1.12
", + "bbox": [ + 176, + 393, + 825, + 574 + ], + "page_idx": 22 + }, + { + "type": "text", + "text": "Experiment on different model structure. Following 4.2, we additionally conduct critical experiments on the WRN-40-4 (Lin et al., 2021) backbone to demonstrate the effectiveness of the proposed UM and UMAP. In Figure 7, we can find during the model training phase on ID data, there also exists the once-covered OOD detection capability can be explored in later development. In Table 18, we show the comparison of multiple OOD detection baselines, evaluating the OOD performance on the 7 OOD datasets mentioned in Section 4.1. The results again demonstrate that our proposed method indeed excavates the intrinsic detection capability and improves the performance. ", + "bbox": [ + 173, + 604, + 825, + 703 + ], + "page_idx": 22 + }, + { + "type": "text", + "text": "As for the fine-grained results of WRN-40-4, we report results on 6 OOD datasets respectively. When trained on CIFAR-10, UM can outstrip all the scoring function baselines on 5 OOD datasets except Textures on which Mahalanobis performs better while UMAP still has excellent OOD performance ranking only second to UM. When trained on CIFAR-100, UM and UMAP can also outperform the baselines on most OOD datasets. The fine-grained results of WRN-40-4 further demonstrate the effectiveness of the proposed UM/UMAP on other architectures. ", + "bbox": [ + 174, + 709, + 825, + 792 + ], + "page_idx": 22 + }, + { + "type": "text", + "text": "Additional results about the general existence of once-covered OOD detection capability. In Section 4.3, we display the once-covered OOD detection capability on CIFAR-10 using SVHN as the OOD dataset. Here, we additionally verify the previously observed trend during training when training DenseNet-101 on CIFAR-100 using iNaturalist as an OOD dataset. In Figure 5, we trace the three evaluation metrics during training on CIFAR-100 using 4 different learning rate schedules. Consistent with the original experiment, we still use iNaturalist as the OOD dataset. It can be seen for all the three metrics that exists a middle stage where the model has the better OOD detection capability (For FPR95, it is smaller (better) in the middle stage; for AUROC and AUPR, they are higher (better) in the middle stage). Besides that, we also look into the change of OOD performance on other architecture (e.g., WRN-40-4) in Figure 6 and Figure 7. In Figure 6, we display the curves of three metrics of WRN-40-4 when trained on CIFAR-10 with SVHN and Textures as OOD datasets. The trend that the OOD performance first goes better and then converges to worse OOD performance can be reflected. In Figure 7, we continually provide curves of the three metrics of WRN-40-4 during training on CIFAR-100 with iNaturalist, Places365, and SUN as OOD datasets. A clear better middle stage can still be excavated in this scenario. ", + "bbox": [ + 174, + 811, + 825, + 924 + ], + "page_idx": 22 + }, + { + "type": "table", + "img_path": "images/db8ab563d2f7b2f21e22411a59359b785cf8ebd0be133a80a03550c6303ee097.jpg", + "table_caption": [ + "Table 20: Fine-grained Results $( \\% )$ of WRN-40-4 on CIFAR-100. Comparison on different OOD benchmark datasets. $\\uparrow$ indicates higher values are better, and $\\downarrow$ indicates lower values are better. " + ], + "table_footnote": [], + "table_body": "
ID datasetMethodOOD dataset
CIFAR-10TexturesPlaces365
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-100MSP83.83 ±0.2975.50±0.2186.15 ± 0.2372.36 ± 0.4086.72 ±0.2969.60± 0.27
ODIN83.70 ±0.3074.32 ±0.0381.57 ± 1.7471.67±0.2888.07 ± 0.1164.83 ± 1.36
Mahalanobis96.89 ± 0.1168.78 ±0.6731.02 ± 2.0491.85 ± 0.9193.34 ± 0.4461.28 ± 1.62
Energy81.32 ± 0.4777.49 ± 0.2686.38 ±0.4973.50±0.4584.45 ± 0.3869.82 ± 0.63
Energy+UM(ours)89.23 ± 1.5163.85 ± 1.7378.90 ±0.0772.58 ± 1.3380.46 ± 1.9970.49 ± 1.01
Energy+UMAP(ours)94.11 ± 0.7260.77 ± 0.9666.94 ± 4.4975.82 ± 5.3382.59±0.2771.92 ± 3.58
MethodSUNLSUNiNaturalist
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
MSP88.88±0.8365.22 ± 0.8578.56±0.6679.10 ± 0.5386.72 ± 0.2973,75 ±0.56
ODIN91.00 ±0.1059.06 ± 1.8170.14 ± 2.4284.03±0.8687.86±1.1364.52 ± 1.51
Mahalanobis94.22 ± 0.0160.09 ± 1.5689.73± 2.8740.81± 3.0787.25±3.2874.98 ± 2.85
Energy88.35±0.5264.04 ±0.7659.84 ±0.0687.91 ± 0.5388.91±0.7867.81 ± 0.91
Energy+UM(ours)84.04 ±0.0967.19 ± 0.1433.87 ± 1.2192.29 ± 0.0276.91 ± 6.0779.28 ± 4.17
Energy+UMAP(ours)80.53 ± 1.3172.68± 4.5027.79 ± 2.1993.39 ± 0.5755.53± 4.6185.65± 0.83
", + "bbox": [ + 176, + 141, + 825, + 319 + ], + "page_idx": 23 + }, + { + "type": "image", + "img_path": "images/1674497873272ebcef525d5b25ffb65e99d34866ce9d3e4d699f013f99848ca6.jpg", + "image_caption": [ + "Figure 5: Ablation studies on three metrics with 4 different learning rate schedules. The model is DenseNet-101 trained on CIFAR-100 with iNaturalist as the OOD dataset. (a) change of FPR95 throughout the pruning phase when training on CIFAR-100; (b) change of AUROC throughout the pruning phase when training on CIFAR-100; (c) change of AUPR throughout the pruning phase when training on CIFAR-100. It demonstrates a better middle stage exists according to the three metrics. " + ], + "image_footnote": [], + "bbox": [ + 192, + 344, + 805, + 506 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "", + "bbox": [ + 173, + 626, + 825, + 724 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "UMAP: adopting pruning on UM. We conduct various experiments to see whether pruning has an impact on Unleashing Mask itself. To be specific, we expect the pruning to learn a mask on the given model while not impairing the excellent OOD performance that UM brings. In Figure 8, it presents that pruning from a wide range (e.g. $p \\in [ 0 . 3 , 0 . 9 ] )$ ) can well maintain the effectiveness of UM while possessing a terrific convergence trend. For simplicity, we use prune to indicate the original pruning approach and UMAP indicate UM with pruning on the mask with our newly designed forgetting objective in Figure 8. In Figure 8(a), the solid lines represent the proposed UMAP and the dashed lines represent only pruning the well-trained model at prune rates 0.2, 0.5, and 0.8. While the model’s OOD performance can’t be improved (not better than the baseline) through only pruning, using our proposed forgetting objective for the loss constrain can significantly bring out better OOD performance at a wide range of mask rates (e.g. $p \\in [ 0 . 5 , 0 . 8 ] \\rangle$ ). In Figure 8(b), we intuitively reflect the effect of the estimated loss constraint by the initialized mask which redirects the gradients when the loss reaches the value, while the loss will just approach 0 when pruning only. In Figure 8(c), we can see that ID-ACC for both UMAP and Prune can converge to approximately the same high level $( 9 2 \\sim 9 4 $ ), though we can simply remove the learned mask to recover the original ID-ACC. ", + "bbox": [ + 173, + 743, + 825, + 924 + ], + "page_idx": 23 + }, + { + "type": "image", + "img_path": "images/7f407edbb5ffbadd5e84855897933477d5ac0f0382566b843728bbc0b1f1e5c4.jpg", + "image_caption": [ + "Figure 6: Ablation studies on three metrics of WRN-40-4 with CIFAR-10 as ID dataset, SVHN, and Textures as OOD datasets. (a) change of FPR95 throughout the pruning phase when training on CIFAR-10; (b) change of AUROC throughout the pruning phase when training on CIFAR-10; (c) change of AUPR throughout the pruning phase when training on CIFAR-10. It demonstrates a better middle stage exists according to the three metrics. " + ], + "image_footnote": [], + "bbox": [ + 187, + 106, + 810, + 270 + ], + "page_idx": 24 + }, + { + "type": "image", + "img_path": "images/24ec436949fcf37fd12376ca16998e59459938abcbe0fd87b4d16de8c5362499.jpg", + "image_caption": [ + "Figure 7: Ablation studies on three metrics of WRN-40-4 with CIFAR-100 as ID dataset, iNaturalist, Places365, and SUN as OOD datasets. (a) change of FPR95 throughout the pruning phase when training on CIFAR-100; (b) change of AUROC throughout the pruning phase when training on CIFAR-100; (c) change of AUPR throughout the pruning phase when training on CIFAR-100. It demonstrates a better middle stage exists according to the three metrics. " + ], + "image_footnote": [], + "bbox": [ + 192, + 382, + 803, + 545 + ], + "page_idx": 24 + }, + { + "type": "text", + "text": "", + "bbox": [ + 173, + 664, + 823, + 693 + ], + "page_idx": 24 + }, + { + "type": "text", + "text": "The effectiveness of UM. In Figure 9, we present the FPR95, AUROC, and AUPR curves during training to show the comparison of the original training and our proposed UM on ID data. We observe that training using UM can consistently outperform than the vanilla model training, either for the final stage or the middle stage with the best OOD detection performance indicated by the FPR95 curve. In Figure 10, we also adopt different mask rates for the initialized loss constraint estimation for forgetting the atypical samples. The results show that a wide range of mask ratios (i.e., from $96 \\%$ to $9 9 \\%$ ) to estimate the loss constraint used in Eq.3 can gain better OOD detection performance than the baseline. It shows the mask ratio would be robust to hyper-parameter selection under a certain value. The principle intuition behind this is our revealed important observation as indicated in Figures 1(a), 2(b), and 2(c). With the guidance of the general mechanism, empirically choosing the hyper-parameter using the validation set is supportable and valuable for excavating better OOD detection capability of the model as conducted by previous literature (Hendrycks et al., 2019b; Liu et al., 2020; Sun et al., 2021). ", + "bbox": [ + 174, + 708, + 825, + 888 + ], + "page_idx": 24 + }, + { + "type": "text", + "text": "In our experiments, we empirically determine the value of our proposed UM and UMAP by examining the training loss on the masked output. For CIFAR-10 as ID datasets, the value of mask rate is $9 7 . 5 \\%$ and the estimated loss constraint for forgetting is 0.10 for our tuning until the convergence; For CIFAR-100, the value of mask rate is $9 7 \\%$ and the estimated loss constraint for forgetting is 1.20 for our tuning until the convergence. ", + "bbox": [ + 173, + 896, + 823, + 924 + ], + "page_idx": 24 + }, + { + "type": "image", + "img_path": "images/be945de072e51170925a45b431f1d3d1c0191ad541bc3cedc09a89a54d7509db.jpg", + "image_caption": [ + "Figure 8: Ablation studies on Prune Rate of UMAP. (a) change of OOD performance throughout the pruning phase; (b) training loss converges to estimated loss constraint properly; (c) though ID-ACC is not taken into consideration for UMAP, it still raise high after training for 100 epochs. " + ], + "image_footnote": [], + "bbox": [ + 194, + 107, + 803, + 270 + ], + "page_idx": 25 + }, + { + "type": "image", + "img_path": "images/70757eca6d628ecb94634d260d5782c79827c5486109a38181ba093662388d29.jpg", + "image_caption": [ + "Figure 9: Ablation studies to reflect the effectiveness of UM. The mask ratio of UM is $9 9 . 5 \\%$ . (a) change of FPR95 throughout the training phase on CIFAR-10; (b) change of AUROC throughout the training phase on CIFAR-10; (c) change of AUPR throughout the training phase on CIFAR-10. " + ], + "image_footnote": [], + "bbox": [ + 196, + 354, + 800, + 517 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 608, + 825, + 650 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "To choose the parameters of the estimated loss constraint, we use the TinyImageNet (Tavanaei, 2020) dataset as the validation set, which is not seen during training and is not considered in our evaluation of OOD detection performance. Since the core intuition behind our method is to restore the OOD detection performance starting from the well-trained model stage, forgetting a relatively small portion (empirically found around $9 7 \\%$ mask ratio) of atypical samples can be beneficial. To find the optimal parameter for tuning, more advanced searching techniques like AutoML or validation design based on the important observation in our work may be further employed in the future. ", + "bbox": [ + 173, + 657, + 825, + 755 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "Fine-grained comparison of the model weights. We display the weights of the original model, pruned model, and the UMAP model respectively in Figure 11. The histograms show that the adopted pruning algorithm tends to choose weights far from 0 for the first convolution layer, shown in Figure 11(a). However, for almost all layers (from the 2nd to the 98th), the pruning chooses weights with no respect to the value of weights, shown in Figure 11(b). For the \"head\" of the model (the fully connected layer), the pruning algorithm itself still keeps its behavior on the first layer, while UMAP forces the prune algorithm to choose weights near 0, shown in Figure 11(c), indicating that forgetting learned atypical samples doesn’t necessarily correspond to larger weights or smaller weights. ", + "bbox": [ + 173, + 770, + 825, + 882 + ], + "page_idx": 25 + }, + { + "type": "image", + "img_path": "images/88729bef6a9f4ab2516da4709ab0cf6ff01bd447fe513e8f4350607f94bf5553.jpg", + "image_caption": [ + "Figure 10: Ablation studies of WRN-40-4 on various Mask Ratios. The mask rate is from $9 5 . 8 \\%$ to $9 9 . 0 \\%$ . (a) change of FPR95 throughout the training on CIFAR-10; (b) change of AUROC throughout the training on CIFAR-10; (c) change of AUPR throughout the training on CIFAR-10. " + ], + "image_footnote": [], + "bbox": [ + 186, + 85, + 808, + 616 + ], + "page_idx": 26 + }, + { + "type": "image", + "img_path": "images/6123471821ec415b46dc126c81c28e0604d37b4633d536b6328c02e27cb00670.jpg", + "image_caption": [ + "Figure 11: Histograms of different layers for the original model, pruned model, and UMAP model. The model is DenseNet-101 with a prune rate of $5 0 \\%$ . (a) the histogram of the first convolution layer; (b) the histogram of the 50th convolution layer; (c) the histogram of the last (fully connected) layer. " + ], + "image_footnote": [], + "bbox": [ + 187, + 691, + 808, + 856 + ], + "page_idx": 26 + }, + { + "type": "image", + "img_path": "images/a2e38ac79e2f9708309fa1f31e758448e2d469ae77bb8ce0661afd67144d1919.jpg", + "image_caption": [ + "Figure 12: Examples of misclassified samples after masking the original well-trained model. The scores are estimated according to uniform distribution. (a) layer-wise masking scores; (b) layer-wise masking weights directly; (c) model-wise masking scores; (d) model-wise masking weights directly. " + ], + "image_footnote": [], + "bbox": [ + 236, + 69, + 758, + 883 + ], + "page_idx": 27 + } +] \ No newline at end of file diff --git a/parse/dev/K2OixmPDou3/K2OixmPDou3_middle.json b/parse/dev/K2OixmPDou3/K2OixmPDou3_middle.json new file mode 100644 index 0000000000000000000000000000000000000000..6e87ceee3a6f729c5bec1709552ac53227a26119 --- /dev/null +++ b/parse/dev/K2OixmPDou3/K2OixmPDou3_middle.json @@ -0,0 +1,62474 @@ +{ + "pdf_info": [ + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 78, + 506, + 116 + ], + "lines": [ + { + "bbox": [ + 106, + 78, + 508, + 97 + ], + "spans": [ + { + "bbox": [ + 106, + 78, + 508, + 97 + ], + "score": 1.0, + "content": "UNLEASHING MASK: EXPLORE THE INTRINSIC OUT-", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 99, + 431, + 118 + ], + "spans": [ + { + "bbox": [ + 106, + 99, + 431, + 118 + ], + "score": 1.0, + "content": "OF-DISTRIBUTION DETECTION CAPABILITY", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 112, + 135, + 244, + 157 + ], + "lines": [ + { + "bbox": [ + 113, + 135, + 201, + 147 + ], + "spans": [ + { + "bbox": [ + 113, + 135, + 201, + 147 + ], + "score": 1.0, + "content": "Anonymous authors", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 111, + 146, + 245, + 158 + ], + "spans": [ + { + "bbox": [ + 111, + 146, + 245, + 158 + ], + "score": 1.0, + "content": "Paper under double-blind review", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5 + }, + { + "type": "title", + "bbox": [ + 278, + 186, + 333, + 199 + ], + "lines": [ + { + "bbox": [ + 276, + 185, + 336, + 200 + ], + "spans": [ + { + "bbox": [ + 276, + 185, + 336, + 200 + ], + "score": 1.0, + "content": "ABSTRACT", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 143, + 214, + 468, + 357 + ], + "lines": [ + { + "bbox": [ + 141, + 212, + 470, + 229 + ], + "spans": [ + { + "bbox": [ + 141, + 212, + 470, + 229 + ], + "score": 1.0, + "content": "Out-of-distribution (OOD) detection is an important aspect of safely deploying", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 141, + 225, + 469, + 237 + ], + "spans": [ + { + "bbox": [ + 141, + 225, + 469, + 237 + ], + "score": 1.0, + "content": "machine learning models in real-world applications. Previous approaches either", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 142, + 237, + 469, + 248 + ], + "spans": [ + { + "bbox": [ + 142, + 237, + 469, + 248 + ], + "score": 1.0, + "content": "design better scoring functions or utilize the knowledge of outliers to equip the", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 141, + 247, + 469, + 259 + ], + "spans": [ + { + "bbox": [ + 141, + 247, + 469, + 259 + ], + "score": 1.0, + "content": "well-trained models with the ability of OOD detection. However, few of them", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 142, + 259, + 469, + 270 + ], + "spans": [ + { + "bbox": [ + 142, + 259, + 469, + 270 + ], + "score": 1.0, + "content": "explore to excavate the intrinsic OOD detection capability of a given model. In", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 142, + 270, + 469, + 281 + ], + "spans": [ + { + "bbox": [ + 142, + 270, + 469, + 281 + ], + "score": 1.0, + "content": "this work, we discover the existence of an intermediate stage of a model trained", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 280, + 469, + 292 + ], + "spans": [ + { + "bbox": [ + 141, + 280, + 469, + 292 + ], + "score": 1.0, + "content": "on in-distribution data having higher OOD detection performance than that of its", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 290, + 470, + 304 + ], + "spans": [ + { + "bbox": [ + 141, + 290, + 470, + 304 + ], + "score": 1.0, + "content": "final stage across different settings and further identify the critical attribution to be", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 142, + 303, + 470, + 314 + ], + "spans": [ + { + "bbox": [ + 142, + 303, + 470, + 314 + ], + "score": 1.0, + "content": "learning with atypical samples. Based on such empirical insights, we propose a new", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 142, + 313, + 470, + 325 + ], + "spans": [ + { + "bbox": [ + 142, + 313, + 470, + 325 + ], + "score": 1.0, + "content": "method, Unleashing Mask (UM), that restores the OOD discriminative capabilities", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 323, + 469, + 336 + ], + "spans": [ + { + "bbox": [ + 141, + 323, + 469, + 336 + ], + "score": 1.0, + "content": "of the model. To be specific, we utilize the mask to figure out the memorized", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 335, + 469, + 347 + ], + "spans": [ + { + "bbox": [ + 141, + 335, + 469, + 347 + ], + "score": 1.0, + "content": "atypical samples and fine-tune the model to forget them. Extensive experiments", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 346, + 461, + 357 + ], + "spans": [ + { + "bbox": [ + 141, + 346, + 461, + 357 + ], + "score": 1.0, + "content": "have been conducted to characterize and verify the effectiveness of our method.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 11 + }, + { + "type": "title", + "bbox": [ + 108, + 385, + 205, + 398 + ], + "lines": [ + { + "bbox": [ + 105, + 384, + 208, + 401 + ], + "spans": [ + { + "bbox": [ + 105, + 384, + 208, + 401 + ], + "score": 1.0, + "content": "1 INTRODUCTION", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 107, + 412, + 505, + 501 + ], + "lines": [ + { + "bbox": [ + 105, + 411, + 506, + 426 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 506, + 426 + ], + "score": 1.0, + "content": "Out-of-distribution (OOD) detection has drawn increasing attention when deploying machine learning", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 424, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 505, + 436 + ], + "score": 1.0, + "content": "models into the open-world scenarios (Nguyen et al., 2015; Lee et al., 2018a). Since the test samples", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 434, + 505, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 505, + 447 + ], + "score": 1.0, + "content": "can naturally arise from a label-different distribution, identifying OOD inputs is important, especially", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 446, + 505, + 458 + ], + "spans": [ + { + "bbox": [ + 106, + 446, + 505, + 458 + ], + "score": 1.0, + "content": "for those safety-critical applications like autonomous driving and medical intelligence. Previous", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 456, + 507, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 507, + 470 + ], + "score": 1.0, + "content": "studies focus on designing a series of scoring functions (Hendrycks & Gimpel, 2017b; Liang et al.,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 467, + 506, + 480 + ], + "spans": [ + { + "bbox": [ + 105, + 467, + 506, + 480 + ], + "score": 1.0, + "content": "2018; Lee et al., 2018a; Liu et al., 2020; Sun et al., 2021; 2022) for OOD uncertainty estimation or", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 478, + 507, + 491 + ], + "spans": [ + { + "bbox": [ + 105, + 479, + 476, + 491 + ], + "score": 1.0, + "content": "finetuning with auxiliary outlier data to better distinguish the OOD inputs (Hendrycks et al.,", + "type": "text" + }, + { + "bbox": [ + 476, + 478, + 502, + 489 + ], + "score": 0.35, + "content": "2 0 1 9 \\mathrm { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 502, + 479, + 507, + 491 + ], + "score": 1.0, + "content": ";", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 489, + 492, + 502 + ], + "spans": [ + { + "bbox": [ + 105, + 489, + 492, + 502 + ], + "score": 1.0, + "content": "Tack et al., 2020; Mohseni et al., 2020; Sehwag et al., 2021; Wei et al., 2022; Ming et al., 2022).", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 22.5 + }, + { + "type": "text", + "bbox": [ + 107, + 506, + 505, + 605 + ], + "lines": [ + { + "bbox": [ + 105, + 506, + 505, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 505, + 519 + ], + "score": 1.0, + "content": "Despite the promising results achieved by previous methods (Hendrycks & Gimpel, 2017a; Hendrycks", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 516, + 506, + 531 + ], + "spans": [ + { + "bbox": [ + 105, + 516, + 131, + 531 + ], + "score": 1.0, + "content": "et al.,", + "type": "text" + }, + { + "bbox": [ + 131, + 518, + 157, + 528 + ], + "score": 0.32, + "content": "2 0 1 9 \\mathrm { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 158, + 516, + 506, + 531 + ], + "score": 1.0, + "content": "; Liu et al., 2020; Ming et al., 2022), little attention is paid to considering whether the", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 529, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 529, + 505, + 541 + ], + "score": 1.0, + "content": "well-trained given model is the most appropriate for OOD detection. In general, models deployed for", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 539, + 506, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 539, + 506, + 552 + ], + "score": 1.0, + "content": "various applications have different targets (e.g., multi-class classification) (Goodfellow et al., 2016)", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 550, + 506, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 550, + 506, + 563 + ], + "score": 1.0, + "content": "instead of OOD detection (Nguyen et al., 2015; Lee et al., 2018a). However, most representative", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 561, + 505, + 574 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 505, + 574 + ], + "score": 1.0, + "content": "score functions, e.g., MSP (Hendrycks et al., 2019c), ODIN(Liang et al., 2018), and Energy (Liu", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 571, + 506, + 585 + ], + "spans": [ + { + "bbox": [ + 105, + 571, + 506, + 585 + ], + "score": 1.0, + "content": "et al., 2020), uniformly leverage the given models for OOD detection. Considering the target-oriented", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 582, + 506, + 595 + ], + "spans": [ + { + "bbox": [ + 105, + 582, + 506, + 595 + ], + "score": 1.0, + "content": "discrepancy, it arises a critical question: does the well-trained given model have the optimal OOD", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 593, + 477, + 608 + ], + "spans": [ + { + "bbox": [ + 105, + 593, + 477, + 608 + ], + "score": 1.0, + "content": "detection capability? If not, how can we find a more appropriate model for OOD detection?", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 107, + 611, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "score": 1.0, + "content": "In this work, we start by revealing an important observation (as illustrated in Figure 1), i.e., there", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 622, + 505, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 622, + 505, + 634 + ], + "score": 1.0, + "content": "exists a historical training stage where the model has a higher OOD detection performance than", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 632, + 507, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 507, + 645 + ], + "score": 1.0, + "content": "the final well-trained one. This is generally true across different OOD/ID datasets (Netzer et al.,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 642, + 507, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 507, + 657 + ], + "score": 1.0, + "content": "2011; Van Horn et al., 2018; Cimpoi et al., 2014), learning rate schedules (Loshchilov & Hutter,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 653, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 506, + 668 + ], + "score": 1.0, + "content": "2017), and model structures (Huang et al., 2017; Zagoruyko & Komodakis, 2016). The empirical", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "score": 1.0, + "content": "results of Figure 1 reflect the inconsistency between gaining better OOD detection capability (Nguyen", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "score": 1.0, + "content": "et al., 2015) and pursuing better performance on ID data. We delve into the differences between", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "score": 1.0, + "content": "the intermediate model and the final model by visualizing the misclassified examples. As shown in", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "Figure 2, one possible attribution for covering the detection capability should be memorizing the", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 710, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 506, + 722 + ], + "score": 1.0, + "content": "atypical samples (at the semantic level) that are hard to learn for the model. Seeking zero error on", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 721, + 464, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 464, + 733 + ], + "score": 1.0, + "content": "those samples makes the model more confident on OOD data (see Figures 1(b) and 1(c)).", + "type": "text" + } + ], + "index": 46 + } + ], + "index": 41 + } + ], + "page_idx": 0, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 308, + 761 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 308, + 761 + ], + "score": 1.0, + "content": "1", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 78, + 506, + 116 + ], + "lines": [ + { + "bbox": [ + 106, + 78, + 508, + 97 + ], + "spans": [ + { + "bbox": [ + 106, + 78, + 508, + 97 + ], + "score": 1.0, + "content": "UNLEASHING MASK: EXPLORE THE INTRINSIC OUT-", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 99, + 431, + 118 + ], + "spans": [ + { + "bbox": [ + 106, + 99, + 431, + 118 + ], + "score": 1.0, + "content": "OF-DISTRIBUTION DETECTION CAPABILITY", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 112, + 135, + 244, + 157 + ], + "lines": [ + { + "bbox": [ + 113, + 135, + 201, + 147 + ], + "spans": [ + { + "bbox": [ + 113, + 135, + 201, + 147 + ], + "score": 1.0, + "content": "Anonymous authors", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 111, + 146, + 245, + 158 + ], + "spans": [ + { + "bbox": [ + 111, + 146, + 245, + 158 + ], + "score": 1.0, + "content": "Paper under double-blind review", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5, + "bbox_fs": [ + 111, + 135, + 245, + 158 + ] + }, + { + "type": "title", + "bbox": [ + 278, + 186, + 333, + 199 + ], + "lines": [ + { + "bbox": [ + 276, + 185, + 336, + 200 + ], + "spans": [ + { + "bbox": [ + 276, + 185, + 336, + 200 + ], + "score": 1.0, + "content": "ABSTRACT", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 143, + 214, + 468, + 357 + ], + "lines": [ + { + "bbox": [ + 141, + 212, + 470, + 229 + ], + "spans": [ + { + "bbox": [ + 141, + 212, + 470, + 229 + ], + "score": 1.0, + "content": "Out-of-distribution (OOD) detection is an important aspect of safely deploying", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 141, + 225, + 469, + 237 + ], + "spans": [ + { + "bbox": [ + 141, + 225, + 469, + 237 + ], + "score": 1.0, + "content": "machine learning models in real-world applications. Previous approaches either", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 142, + 237, + 469, + 248 + ], + "spans": [ + { + "bbox": [ + 142, + 237, + 469, + 248 + ], + "score": 1.0, + "content": "design better scoring functions or utilize the knowledge of outliers to equip the", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 141, + 247, + 469, + 259 + ], + "spans": [ + { + "bbox": [ + 141, + 247, + 469, + 259 + ], + "score": 1.0, + "content": "well-trained models with the ability of OOD detection. However, few of them", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 142, + 259, + 469, + 270 + ], + "spans": [ + { + "bbox": [ + 142, + 259, + 469, + 270 + ], + "score": 1.0, + "content": "explore to excavate the intrinsic OOD detection capability of a given model. In", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 142, + 270, + 469, + 281 + ], + "spans": [ + { + "bbox": [ + 142, + 270, + 469, + 281 + ], + "score": 1.0, + "content": "this work, we discover the existence of an intermediate stage of a model trained", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 280, + 469, + 292 + ], + "spans": [ + { + "bbox": [ + 141, + 280, + 469, + 292 + ], + "score": 1.0, + "content": "on in-distribution data having higher OOD detection performance than that of its", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 290, + 470, + 304 + ], + "spans": [ + { + "bbox": [ + 141, + 290, + 470, + 304 + ], + "score": 1.0, + "content": "final stage across different settings and further identify the critical attribution to be", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 142, + 303, + 470, + 314 + ], + "spans": [ + { + "bbox": [ + 142, + 303, + 470, + 314 + ], + "score": 1.0, + "content": "learning with atypical samples. Based on such empirical insights, we propose a new", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 142, + 313, + 470, + 325 + ], + "spans": [ + { + "bbox": [ + 142, + 313, + 470, + 325 + ], + "score": 1.0, + "content": "method, Unleashing Mask (UM), that restores the OOD discriminative capabilities", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 323, + 469, + 336 + ], + "spans": [ + { + "bbox": [ + 141, + 323, + 469, + 336 + ], + "score": 1.0, + "content": "of the model. To be specific, we utilize the mask to figure out the memorized", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 335, + 469, + 347 + ], + "spans": [ + { + "bbox": [ + 141, + 335, + 469, + 347 + ], + "score": 1.0, + "content": "atypical samples and fine-tune the model to forget them. Extensive experiments", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 346, + 461, + 357 + ], + "spans": [ + { + "bbox": [ + 141, + 346, + 461, + 357 + ], + "score": 1.0, + "content": "have been conducted to characterize and verify the effectiveness of our method.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 11, + "bbox_fs": [ + 141, + 212, + 470, + 357 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 385, + 205, + 398 + ], + "lines": [ + { + "bbox": [ + 105, + 384, + 208, + 401 + ], + "spans": [ + { + "bbox": [ + 105, + 384, + 208, + 401 + ], + "score": 1.0, + "content": "1 INTRODUCTION", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 107, + 412, + 505, + 501 + ], + "lines": [ + { + "bbox": [ + 105, + 411, + 506, + 426 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 506, + 426 + ], + "score": 1.0, + "content": "Out-of-distribution (OOD) detection has drawn increasing attention when deploying machine learning", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 424, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 505, + 436 + ], + "score": 1.0, + "content": "models into the open-world scenarios (Nguyen et al., 2015; Lee et al., 2018a). Since the test samples", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 434, + 505, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 505, + 447 + ], + "score": 1.0, + "content": "can naturally arise from a label-different distribution, identifying OOD inputs is important, especially", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 446, + 505, + 458 + ], + "spans": [ + { + "bbox": [ + 106, + 446, + 505, + 458 + ], + "score": 1.0, + "content": "for those safety-critical applications like autonomous driving and medical intelligence. Previous", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 456, + 507, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 507, + 470 + ], + "score": 1.0, + "content": "studies focus on designing a series of scoring functions (Hendrycks & Gimpel, 2017b; Liang et al.,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 467, + 506, + 480 + ], + "spans": [ + { + "bbox": [ + 105, + 467, + 506, + 480 + ], + "score": 1.0, + "content": "2018; Lee et al., 2018a; Liu et al., 2020; Sun et al., 2021; 2022) for OOD uncertainty estimation or", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 478, + 507, + 491 + ], + "spans": [ + { + "bbox": [ + 105, + 479, + 476, + 491 + ], + "score": 1.0, + "content": "finetuning with auxiliary outlier data to better distinguish the OOD inputs (Hendrycks et al.,", + "type": "text" + }, + { + "bbox": [ + 476, + 478, + 502, + 489 + ], + "score": 0.35, + "content": "2 0 1 9 \\mathrm { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 502, + 479, + 507, + 491 + ], + "score": 1.0, + "content": ";", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 489, + 492, + 502 + ], + "spans": [ + { + "bbox": [ + 105, + 489, + 492, + 502 + ], + "score": 1.0, + "content": "Tack et al., 2020; Mohseni et al., 2020; Sehwag et al., 2021; Wei et al., 2022; Ming et al., 2022).", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 22.5, + "bbox_fs": [ + 105, + 411, + 507, + 502 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 506, + 505, + 605 + ], + "lines": [ + { + "bbox": [ + 105, + 506, + 505, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 505, + 519 + ], + "score": 1.0, + "content": "Despite the promising results achieved by previous methods (Hendrycks & Gimpel, 2017a; Hendrycks", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 516, + 506, + 531 + ], + "spans": [ + { + "bbox": [ + 105, + 516, + 131, + 531 + ], + "score": 1.0, + "content": "et al.,", + "type": "text" + }, + { + "bbox": [ + 131, + 518, + 157, + 528 + ], + "score": 0.32, + "content": "2 0 1 9 \\mathrm { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 158, + 516, + 506, + 531 + ], + "score": 1.0, + "content": "; Liu et al., 2020; Ming et al., 2022), little attention is paid to considering whether the", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 529, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 529, + 505, + 541 + ], + "score": 1.0, + "content": "well-trained given model is the most appropriate for OOD detection. In general, models deployed for", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 539, + 506, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 539, + 506, + 552 + ], + "score": 1.0, + "content": "various applications have different targets (e.g., multi-class classification) (Goodfellow et al., 2016)", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 550, + 506, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 550, + 506, + 563 + ], + "score": 1.0, + "content": "instead of OOD detection (Nguyen et al., 2015; Lee et al., 2018a). However, most representative", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 561, + 505, + 574 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 505, + 574 + ], + "score": 1.0, + "content": "score functions, e.g., MSP (Hendrycks et al., 2019c), ODIN(Liang et al., 2018), and Energy (Liu", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 571, + 506, + 585 + ], + "spans": [ + { + "bbox": [ + 105, + 571, + 506, + 585 + ], + "score": 1.0, + "content": "et al., 2020), uniformly leverage the given models for OOD detection. Considering the target-oriented", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 582, + 506, + 595 + ], + "spans": [ + { + "bbox": [ + 105, + 582, + 506, + 595 + ], + "score": 1.0, + "content": "discrepancy, it arises a critical question: does the well-trained given model have the optimal OOD", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 593, + 477, + 608 + ], + "spans": [ + { + "bbox": [ + 105, + 593, + 477, + 608 + ], + "score": 1.0, + "content": "detection capability? If not, how can we find a more appropriate model for OOD detection?", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 31, + "bbox_fs": [ + 105, + 506, + 506, + 608 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 611, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "score": 1.0, + "content": "In this work, we start by revealing an important observation (as illustrated in Figure 1), i.e., there", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 622, + 505, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 622, + 505, + 634 + ], + "score": 1.0, + "content": "exists a historical training stage where the model has a higher OOD detection performance than", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 632, + 507, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 507, + 645 + ], + "score": 1.0, + "content": "the final well-trained one. This is generally true across different OOD/ID datasets (Netzer et al.,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 642, + 507, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 507, + 657 + ], + "score": 1.0, + "content": "2011; Van Horn et al., 2018; Cimpoi et al., 2014), learning rate schedules (Loshchilov & Hutter,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 653, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 506, + 668 + ], + "score": 1.0, + "content": "2017), and model structures (Huang et al., 2017; Zagoruyko & Komodakis, 2016). The empirical", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "score": 1.0, + "content": "results of Figure 1 reflect the inconsistency between gaining better OOD detection capability (Nguyen", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "score": 1.0, + "content": "et al., 2015) and pursuing better performance on ID data. We delve into the differences between", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "score": 1.0, + "content": "the intermediate model and the final model by visualizing the misclassified examples. As shown in", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "Figure 2, one possible attribution for covering the detection capability should be memorizing the", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 710, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 506, + 722 + ], + "score": 1.0, + "content": "atypical samples (at the semantic level) that are hard to learn for the model. Seeking zero error on", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 721, + 464, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 464, + 733 + ], + "score": 1.0, + "content": "those samples makes the model more confident on OOD data (see Figures 1(b) and 1(c)).", + "type": "text" + } + ], + "index": 46 + } + ], + "index": 41, + "bbox_fs": [ + 105, + 611, + 507, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 111, + 84, + 496, + 204 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 111, + 84, + 496, + 204 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 111, + 84, + 496, + 204 + ], + "spans": [ + { + "bbox": [ + 111, + 84, + 496, + 204 + ], + "score": 0.968, + "type": "image", + "image_path": "0999e8dc8619d56385f0f50f3492c4db278ff3ca6514f7f0f6a51a3233af7890.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 111, + 84, + 496, + 124.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 111, + 124.0, + 496, + 164.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 111, + 164.0, + 496, + 204.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 215, + 506, + 315 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 214, + 506, + 227 + ], + "spans": [ + { + "bbox": [ + 105, + 214, + 506, + 227 + ], + "score": 1.0, + "content": "Figure 1: (a) the curves of FPR95 (false positive rate of OOD examples when the true positive rate", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 226, + 506, + 239 + ], + "spans": [ + { + "bbox": [ + 105, + 226, + 241, + 239 + ], + "score": 1.0, + "content": "of in-distribution examples is at", + "type": "text" + }, + { + "bbox": [ + 241, + 227, + 262, + 237 + ], + "score": 0.87, + "content": "9 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 262, + 226, + 506, + 239 + ], + "score": 1.0, + "content": ") based on the Energy score (Liu et al., 2020) across three", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 237, + 505, + 249 + ], + "spans": [ + { + "bbox": [ + 106, + 237, + 505, + 249 + ], + "score": 1.0, + "content": "different OOD datasets during the training on the CIFAR-10 dataset. (b) comparison between ID", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 248, + 506, + 260 + ], + "spans": [ + { + "bbox": [ + 106, + 248, + 506, + 260 + ], + "score": 1.0, + "content": "and OOD distribution at Epoch 60. (c) comparison between ID and OOD distribution at Epoch 100.", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 258, + 506, + 272 + ], + "spans": [ + { + "bbox": [ + 105, + 258, + 506, + 272 + ], + "score": 1.0, + "content": "All the experiments testing for OOD detection performance have been conducted multiple times. By", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 269, + 505, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 269, + 505, + 282 + ], + "score": 1.0, + "content": "backtracking the training phase, we can observe the existence of the model stage with better OOD", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 280, + 505, + 294 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 505, + 294 + ], + "score": 1.0, + "content": "detection capability using the Energy score to distinguish the OOD inputs. When zooming in the ID", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 292, + 505, + 304 + ], + "spans": [ + { + "bbox": [ + 106, + 292, + 505, + 304 + ], + "score": 1.0, + "content": "and OOD distributions at Epoch 60 and Epoch 100 respectively, it can be seen that, along with the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 303, + 495, + 315 + ], + "spans": [ + { + "bbox": [ + 106, + 303, + 495, + 315 + ], + "score": 1.0, + "content": "training at the later stage, the overlap between them grows. Figure 2 contains further exploration.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 7 + } + ], + "index": 4.0 + }, + { + "type": "text", + "bbox": [ + 106, + 337, + 505, + 480 + ], + "lines": [ + { + "bbox": [ + 105, + 336, + 506, + 351 + ], + "spans": [ + { + "bbox": [ + 105, + 336, + 506, + 351 + ], + "score": 1.0, + "content": "The above analysis inspires us to propose a new strategy, namely, Unleashing Mask (UM), to excavate", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 348, + 506, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 348, + 506, + 361 + ], + "score": 1.0, + "content": "the once-covered detection capability of a well-trained given model by alleviating the memorization", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 360, + 505, + 372 + ], + "spans": [ + { + "bbox": [ + 106, + 360, + 505, + 372 + ], + "score": 1.0, + "content": "of those atypical samples (as illustrated in Figure 3) of ID data. In general, we aim to backtrack its", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 371, + 505, + 383 + ], + "spans": [ + { + "bbox": [ + 105, + 371, + 505, + 383 + ], + "score": 1.0, + "content": "previous stage with a better OOD detection capability. To achieve this target, there are two essential", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 381, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 381, + 506, + 394 + ], + "score": 1.0, + "content": "issues: (1) the model that is well-trained on ID data has already memorized some atypical samples;", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 392, + 506, + 406 + ], + "spans": [ + { + "bbox": [ + 105, + 392, + 506, + 406 + ], + "score": 1.0, + "content": "(2) how to forget those memorized atypical samples considering the given model? Accordingly, our", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 104, + 403, + 506, + 416 + ], + "spans": [ + { + "bbox": [ + 104, + 403, + 506, + 416 + ], + "score": 1.0, + "content": "proposed UM contains two parts utilizing different insights to address the above problems. First, as", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 104, + 414, + 506, + 427 + ], + "spans": [ + { + "bbox": [ + 104, + 414, + 506, + 427 + ], + "score": 1.0, + "content": "atypical samples are more sensitive to the change of model parameters, we initialize a mask with", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 425, + 506, + 438 + ], + "spans": [ + { + "bbox": [ + 106, + 425, + 506, + 438 + ], + "score": 1.0, + "content": "the specific cutting rate to mine these samples with constructed discrepancy. Then, with the loss", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 437, + 505, + 448 + ], + "spans": [ + { + "bbox": [ + 106, + 437, + 505, + 448 + ], + "score": 1.0, + "content": "reference estimated by the mask, we conduct the constrained gradient ascent (i.e., Eq. 3) for model", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 447, + 505, + 460 + ], + "spans": [ + { + "bbox": [ + 106, + 447, + 505, + 460 + ], + "score": 1.0, + "content": "forgetting. It will encourage the model to finally stabilize around the optimal stage. To avoid the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 104, + 456, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 104, + 456, + 505, + 473 + ], + "score": 1.0, + "content": "severe sacrifice of the original task performance on ID data, we further propose UM Adopts Pruning", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 470, + 488, + 481 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 488, + 481 + ], + "score": 1.0, + "content": "(UMAP) which performs the tuning on the introduced mask with the newly designed objective.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 107, + 486, + 505, + 574 + ], + "lines": [ + { + "bbox": [ + 105, + 486, + 506, + 499 + ], + "spans": [ + { + "bbox": [ + 105, + 486, + 506, + 499 + ], + "score": 1.0, + "content": "For our proposed methods, we conduct extensive experiments to characterize and understand the", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 497, + 506, + 509 + ], + "spans": [ + { + "bbox": [ + 106, + 497, + 506, + 509 + ], + "score": 1.0, + "content": "working mechanism (in Section 4 and Appendix F). The comprehensive results accordingly demon-", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 508, + 505, + 520 + ], + "spans": [ + { + "bbox": [ + 106, + 508, + 505, + 520 + ], + "score": 1.0, + "content": "strate their effectiveness. We have verified the effectiveness of UM with a series of OOD detection", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 519, + 505, + 531 + ], + "spans": [ + { + "bbox": [ + 106, + 519, + 505, + 531 + ], + "score": 1.0, + "content": "benchmarks considering the two different ID datasets, i.e., CIFAR-10 and CIFAR-100. Under the", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 529, + 505, + 542 + ], + "spans": [ + { + "bbox": [ + 106, + 529, + 505, + 542 + ], + "score": 1.0, + "content": "various evaluation metrics, our UM, as well as UMAP, can indeed excavate the better OOD detection", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 540, + 506, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 540, + 506, + 554 + ], + "score": 1.0, + "content": "capability of given models and the averaged FPR95 can be reduced by a significant margin. Finally, a", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 551, + 506, + 564 + ], + "spans": [ + { + "bbox": [ + 105, + 551, + 506, + 564 + ], + "score": 1.0, + "content": "range of ablation studies and further discussions related to our proposed strategy are provided. We", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 563, + 291, + 575 + ], + "spans": [ + { + "bbox": [ + 106, + 563, + 291, + 575 + ], + "score": 1.0, + "content": "summarize our main contributions as follows,", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 28.5 + }, + { + "type": "text", + "bbox": [ + 133, + 584, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 133, + 584, + 506, + 597 + ], + "spans": [ + { + "bbox": [ + 133, + 584, + 506, + 597 + ], + "score": 1.0, + "content": "• Conceptually, we explore the OOD detection performance via a new perspective, i.e., back-", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 141, + 595, + 505, + 608 + ], + "spans": [ + { + "bbox": [ + 141, + 595, + 505, + 608 + ], + "score": 1.0, + "content": "tracking the model training phase without regularizing by any auxiliary outliers, different", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 141, + 607, + 446, + 618 + ], + "spans": [ + { + "bbox": [ + 141, + 607, + 446, + 618 + ], + "score": 1.0, + "content": "from most previous works that start with the well-trained model on ID data.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 138, + 622, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 138, + 622, + 505, + 635 + ], + "score": 1.0, + "content": "Empirically, we reveal the potential detection capability of the well-trained model. We", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 142, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 142, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "observe the general existence of an intermediate stage where the model has more appropriate", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 142, + 644, + 393, + 656 + ], + "spans": [ + { + "bbox": [ + 142, + 644, + 393, + 656 + ], + "score": 1.0, + "content": "discriminative features that can be utilized for OOD detection.", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 138, + 659, + 506, + 674 + ], + "spans": [ + { + "bbox": [ + 138, + 659, + 506, + 674 + ], + "score": 1.0, + "content": "Technically, we introduce a new strategy, i.e., Unleashing Mask, to excavate the once-covered", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 142, + 672, + 505, + 684 + ], + "spans": [ + { + "bbox": [ + 142, + 672, + 505, + 684 + ], + "score": 1.0, + "content": "OOD detection capability of a given model. By introducing the mask, we estimate the loss", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 142, + 683, + 488, + 695 + ], + "spans": [ + { + "bbox": [ + 142, + 683, + 488, + 695 + ], + "score": 1.0, + "content": "constraint for forgetting the atypical samples and empower the detection performance.", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 138, + 698, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 138, + 698, + 505, + 712 + ], + "score": 1.0, + "content": "Experimentally, we conduct extensive explorations to verify the general effectiveness on", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 141, + 710, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 141, + 710, + 505, + 722 + ], + "score": 1.0, + "content": "improving the OOD detection performance of our methods. Using various ID and OOD", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 141, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 141, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "benchmarks, we provide comprehensive results across different setups and further discussion.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 38.5 + } + ], + "page_idx": 1, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 763 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 763 + ], + "score": 1.0, + "content": "2", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 111, + 84, + 496, + 204 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 111, + 84, + 496, + 204 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 111, + 84, + 496, + 204 + ], + "spans": [ + { + "bbox": [ + 111, + 84, + 496, + 204 + ], + "score": 0.968, + "type": "image", + "image_path": "0999e8dc8619d56385f0f50f3492c4db278ff3ca6514f7f0f6a51a3233af7890.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 111, + 84, + 496, + 124.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 111, + 124.0, + 496, + 164.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 111, + 164.0, + 496, + 204.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 215, + 506, + 315 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 214, + 506, + 227 + ], + "spans": [ + { + "bbox": [ + 105, + 214, + 506, + 227 + ], + "score": 1.0, + "content": "Figure 1: (a) the curves of FPR95 (false positive rate of OOD examples when the true positive rate", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 226, + 506, + 239 + ], + "spans": [ + { + "bbox": [ + 105, + 226, + 241, + 239 + ], + "score": 1.0, + "content": "of in-distribution examples is at", + "type": "text" + }, + { + "bbox": [ + 241, + 227, + 262, + 237 + ], + "score": 0.87, + "content": "9 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 262, + 226, + 506, + 239 + ], + "score": 1.0, + "content": ") based on the Energy score (Liu et al., 2020) across three", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 237, + 505, + 249 + ], + "spans": [ + { + "bbox": [ + 106, + 237, + 505, + 249 + ], + "score": 1.0, + "content": "different OOD datasets during the training on the CIFAR-10 dataset. (b) comparison between ID", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 248, + 506, + 260 + ], + "spans": [ + { + "bbox": [ + 106, + 248, + 506, + 260 + ], + "score": 1.0, + "content": "and OOD distribution at Epoch 60. (c) comparison between ID and OOD distribution at Epoch 100.", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 258, + 506, + 272 + ], + "spans": [ + { + "bbox": [ + 105, + 258, + 506, + 272 + ], + "score": 1.0, + "content": "All the experiments testing for OOD detection performance have been conducted multiple times. By", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 269, + 505, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 269, + 505, + 282 + ], + "score": 1.0, + "content": "backtracking the training phase, we can observe the existence of the model stage with better OOD", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 280, + 505, + 294 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 505, + 294 + ], + "score": 1.0, + "content": "detection capability using the Energy score to distinguish the OOD inputs. When zooming in the ID", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 292, + 505, + 304 + ], + "spans": [ + { + "bbox": [ + 106, + 292, + 505, + 304 + ], + "score": 1.0, + "content": "and OOD distributions at Epoch 60 and Epoch 100 respectively, it can be seen that, along with the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 303, + 495, + 315 + ], + "spans": [ + { + "bbox": [ + 106, + 303, + 495, + 315 + ], + "score": 1.0, + "content": "training at the later stage, the overlap between them grows. Figure 2 contains further exploration.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 7 + } + ], + "index": 4.0 + }, + { + "type": "text", + "bbox": [ + 106, + 337, + 505, + 480 + ], + "lines": [ + { + "bbox": [ + 105, + 336, + 506, + 351 + ], + "spans": [ + { + "bbox": [ + 105, + 336, + 506, + 351 + ], + "score": 1.0, + "content": "The above analysis inspires us to propose a new strategy, namely, Unleashing Mask (UM), to excavate", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 348, + 506, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 348, + 506, + 361 + ], + "score": 1.0, + "content": "the once-covered detection capability of a well-trained given model by alleviating the memorization", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 360, + 505, + 372 + ], + "spans": [ + { + "bbox": [ + 106, + 360, + 505, + 372 + ], + "score": 1.0, + "content": "of those atypical samples (as illustrated in Figure 3) of ID data. In general, we aim to backtrack its", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 371, + 505, + 383 + ], + "spans": [ + { + "bbox": [ + 105, + 371, + 505, + 383 + ], + "score": 1.0, + "content": "previous stage with a better OOD detection capability. To achieve this target, there are two essential", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 381, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 381, + 506, + 394 + ], + "score": 1.0, + "content": "issues: (1) the model that is well-trained on ID data has already memorized some atypical samples;", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 392, + 506, + 406 + ], + "spans": [ + { + "bbox": [ + 105, + 392, + 506, + 406 + ], + "score": 1.0, + "content": "(2) how to forget those memorized atypical samples considering the given model? Accordingly, our", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 104, + 403, + 506, + 416 + ], + "spans": [ + { + "bbox": [ + 104, + 403, + 506, + 416 + ], + "score": 1.0, + "content": "proposed UM contains two parts utilizing different insights to address the above problems. First, as", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 104, + 414, + 506, + 427 + ], + "spans": [ + { + "bbox": [ + 104, + 414, + 506, + 427 + ], + "score": 1.0, + "content": "atypical samples are more sensitive to the change of model parameters, we initialize a mask with", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 425, + 506, + 438 + ], + "spans": [ + { + "bbox": [ + 106, + 425, + 506, + 438 + ], + "score": 1.0, + "content": "the specific cutting rate to mine these samples with constructed discrepancy. Then, with the loss", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 437, + 505, + 448 + ], + "spans": [ + { + "bbox": [ + 106, + 437, + 505, + 448 + ], + "score": 1.0, + "content": "reference estimated by the mask, we conduct the constrained gradient ascent (i.e., Eq. 3) for model", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 447, + 505, + 460 + ], + "spans": [ + { + "bbox": [ + 106, + 447, + 505, + 460 + ], + "score": 1.0, + "content": "forgetting. It will encourage the model to finally stabilize around the optimal stage. To avoid the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 104, + 456, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 104, + 456, + 505, + 473 + ], + "score": 1.0, + "content": "severe sacrifice of the original task performance on ID data, we further propose UM Adopts Pruning", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 470, + 488, + 481 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 488, + 481 + ], + "score": 1.0, + "content": "(UMAP) which performs the tuning on the introduced mask with the newly designed objective.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 18, + "bbox_fs": [ + 104, + 336, + 506, + 481 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 486, + 505, + 574 + ], + "lines": [ + { + "bbox": [ + 105, + 486, + 506, + 499 + ], + "spans": [ + { + "bbox": [ + 105, + 486, + 506, + 499 + ], + "score": 1.0, + "content": "For our proposed methods, we conduct extensive experiments to characterize and understand the", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 497, + 506, + 509 + ], + "spans": [ + { + "bbox": [ + 106, + 497, + 506, + 509 + ], + "score": 1.0, + "content": "working mechanism (in Section 4 and Appendix F). The comprehensive results accordingly demon-", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 508, + 505, + 520 + ], + "spans": [ + { + "bbox": [ + 106, + 508, + 505, + 520 + ], + "score": 1.0, + "content": "strate their effectiveness. We have verified the effectiveness of UM with a series of OOD detection", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 519, + 505, + 531 + ], + "spans": [ + { + "bbox": [ + 106, + 519, + 505, + 531 + ], + "score": 1.0, + "content": "benchmarks considering the two different ID datasets, i.e., CIFAR-10 and CIFAR-100. Under the", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 529, + 505, + 542 + ], + "spans": [ + { + "bbox": [ + 106, + 529, + 505, + 542 + ], + "score": 1.0, + "content": "various evaluation metrics, our UM, as well as UMAP, can indeed excavate the better OOD detection", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 540, + 506, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 540, + 506, + 554 + ], + "score": 1.0, + "content": "capability of given models and the averaged FPR95 can be reduced by a significant margin. Finally, a", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 551, + 506, + 564 + ], + "spans": [ + { + "bbox": [ + 105, + 551, + 506, + 564 + ], + "score": 1.0, + "content": "range of ablation studies and further discussions related to our proposed strategy are provided. We", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 563, + 291, + 575 + ], + "spans": [ + { + "bbox": [ + 106, + 563, + 291, + 575 + ], + "score": 1.0, + "content": "summarize our main contributions as follows,", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 28.5, + "bbox_fs": [ + 105, + 486, + 506, + 575 + ] + }, + { + "type": "text", + "bbox": [ + 133, + 584, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 133, + 584, + 506, + 597 + ], + "spans": [ + { + "bbox": [ + 133, + 584, + 506, + 597 + ], + "score": 1.0, + "content": "• Conceptually, we explore the OOD detection performance via a new perspective, i.e., back-", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 141, + 595, + 505, + 608 + ], + "spans": [ + { + "bbox": [ + 141, + 595, + 505, + 608 + ], + "score": 1.0, + "content": "tracking the model training phase without regularizing by any auxiliary outliers, different", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 141, + 607, + 446, + 618 + ], + "spans": [ + { + "bbox": [ + 141, + 607, + 446, + 618 + ], + "score": 1.0, + "content": "from most previous works that start with the well-trained model on ID data.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 138, + 622, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 138, + 622, + 505, + 635 + ], + "score": 1.0, + "content": "Empirically, we reveal the potential detection capability of the well-trained model. We", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 142, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 142, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "observe the general existence of an intermediate stage where the model has more appropriate", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 142, + 644, + 393, + 656 + ], + "spans": [ + { + "bbox": [ + 142, + 644, + 393, + 656 + ], + "score": 1.0, + "content": "discriminative features that can be utilized for OOD detection.", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 138, + 659, + 506, + 674 + ], + "spans": [ + { + "bbox": [ + 138, + 659, + 506, + 674 + ], + "score": 1.0, + "content": "Technically, we introduce a new strategy, i.e., Unleashing Mask, to excavate the once-covered", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 142, + 672, + 505, + 684 + ], + "spans": [ + { + "bbox": [ + 142, + 672, + 505, + 684 + ], + "score": 1.0, + "content": "OOD detection capability of a given model. By introducing the mask, we estimate the loss", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 142, + 683, + 488, + 695 + ], + "spans": [ + { + "bbox": [ + 142, + 683, + 488, + 695 + ], + "score": 1.0, + "content": "constraint for forgetting the atypical samples and empower the detection performance.", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 138, + 698, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 138, + 698, + 505, + 712 + ], + "score": 1.0, + "content": "Experimentally, we conduct extensive explorations to verify the general effectiveness on", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 141, + 710, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 141, + 710, + 505, + 722 + ], + "score": 1.0, + "content": "improving the OOD detection performance of our methods. Using various ID and OOD", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 141, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 141, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "benchmarks, we provide comprehensive results across different setups and further discussion.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 38.5, + "bbox_fs": [ + 133, + 584, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 81, + 200, + 94 + ], + "lines": [ + { + "bbox": [ + 104, + 79, + 202, + 97 + ], + "spans": [ + { + "bbox": [ + 104, + 79, + 202, + 97 + ], + "score": 1.0, + "content": "2 BACKGROUND", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 104, + 106, + 478, + 118 + ], + "lines": [ + { + "bbox": [ + 105, + 106, + 479, + 119 + ], + "spans": [ + { + "bbox": [ + 105, + 106, + 479, + 119 + ], + "score": 1.0, + "content": "In this section, we briefly introduce the preliminaries and related work about OOD detection.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1 + }, + { + "type": "title", + "bbox": [ + 107, + 133, + 200, + 145 + ], + "lines": [ + { + "bbox": [ + 106, + 133, + 200, + 145 + ], + "spans": [ + { + "bbox": [ + 106, + 133, + 200, + 145 + ], + "score": 1.0, + "content": "2.1 PRELIMINARIES", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 2 + }, + { + "type": "text", + "bbox": [ + 106, + 154, + 505, + 243 + ], + "lines": [ + { + "bbox": [ + 106, + 154, + 506, + 166 + ], + "spans": [ + { + "bbox": [ + 106, + 154, + 359, + 166 + ], + "score": 1.0, + "content": "We consider multi-class classification as the training task, where", + "type": "text" + }, + { + "bbox": [ + 360, + 154, + 394, + 165 + ], + "score": 0.92, + "content": "\\mathcal { X } \\subset \\mathbb { R } ^ { d }", + "type": "inline_equation" + }, + { + "bbox": [ + 395, + 154, + 506, + 166 + ], + "score": 1.0, + "content": "denotes the input space and", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 107, + 165, + 505, + 177 + ], + "spans": [ + { + "bbox": [ + 107, + 165, + 174, + 177 + ], + "score": 0.93, + "content": "\\mathcal { Y } = \\{ 1 , \\ldots , C \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 174, + 165, + 505, + 177 + ], + "score": 1.0, + "content": "denotes the label space. Given the model deployed in the real world, the reliable", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 176, + 505, + 188 + ], + "spans": [ + { + "bbox": [ + 105, + 176, + 505, + 188 + ], + "score": 1.0, + "content": "classifier is expected to figure out the OOD input, which can be formulated as a binary classification", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 187, + 506, + 199 + ], + "spans": [ + { + "bbox": [ + 105, + 187, + 170, + 199 + ], + "score": 1.0, + "content": "problem. Given", + "type": "text" + }, + { + "bbox": [ + 170, + 187, + 179, + 198 + ], + "score": 0.8, + "content": "\\mathcal { P }", + "type": "inline_equation" + }, + { + "bbox": [ + 180, + 187, + 264, + 199 + ], + "score": 1.0, + "content": ", the distribution over", + "type": "text" + }, + { + "bbox": [ + 264, + 188, + 292, + 198 + ], + "score": 0.89, + "content": "\\mathcal { X } \\times \\mathcal { V }", + "type": "inline_equation" + }, + { + "bbox": [ + 293, + 187, + 345, + 199 + ], + "score": 1.0, + "content": ", we consider", + "type": "text" + }, + { + "bbox": [ + 345, + 187, + 361, + 198 + ], + "score": 0.89, + "content": "D _ { \\mathrm { i n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 361, + 187, + 481, + 199 + ], + "score": 1.0, + "content": "as the marginal distribution of", + "type": "text" + }, + { + "bbox": [ + 482, + 188, + 490, + 198 + ], + "score": 0.83, + "content": "\\mathcal { P }", + "type": "inline_equation" + }, + { + "bbox": [ + 490, + 187, + 506, + 199 + ], + "score": 1.0, + "content": "for", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 107, + 196, + 504, + 212 + ], + "spans": [ + { + "bbox": [ + 107, + 199, + 116, + 208 + ], + "score": 0.81, + "content": "\\mathcal { X }", + "type": "inline_equation" + }, + { + "bbox": [ + 117, + 196, + 485, + 212 + ], + "score": 1.0, + "content": ", namely, the distribution of ID data. At test time, the environment can present a distribution", + "type": "text" + }, + { + "bbox": [ + 485, + 199, + 504, + 209 + ], + "score": 0.88, + "content": "D _ { \\mathrm { o u t } }", + "type": "inline_equation" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 209, + 506, + 221 + ], + "spans": [ + { + "bbox": [ + 105, + 209, + 126, + 221 + ], + "score": 1.0, + "content": "over", + "type": "text" + }, + { + "bbox": [ + 126, + 210, + 136, + 219 + ], + "score": 0.82, + "content": "\\mathcal { X }", + "type": "inline_equation" + }, + { + "bbox": [ + 136, + 209, + 324, + 221 + ], + "score": 1.0, + "content": "of OOD data. In general, the OOD distribution", + "type": "text" + }, + { + "bbox": [ + 324, + 209, + 343, + 220 + ], + "score": 0.9, + "content": "D _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 343, + 209, + 506, + 221 + ], + "score": 1.0, + "content": "is defined as an irrelevant distribution of", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 220, + 505, + 232 + ], + "spans": [ + { + "bbox": [ + 105, + 220, + 276, + 232 + ], + "score": 1.0, + "content": "which the label set has no intersection with", + "type": "text" + }, + { + "bbox": [ + 276, + 221, + 285, + 231 + ], + "score": 0.81, + "content": "\\mathcal { V }", + "type": "inline_equation" + }, + { + "bbox": [ + 285, + 220, + 505, + 232 + ], + "score": 1.0, + "content": "Yang et al. (2021) and therefore should not be predicted", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 232, + 356, + 243 + ], + "spans": [ + { + "bbox": [ + 106, + 232, + 344, + 243 + ], + "score": 1.0, + "content": "by the model. The decision can be made with the threshold", + "type": "text" + }, + { + "bbox": [ + 345, + 232, + 351, + 241 + ], + "score": 0.81, + "content": "\\lambda", + "type": "inline_equation" + }, + { + "bbox": [ + 352, + 232, + 356, + 243 + ], + "score": 1.0, + "content": ":", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 6.5 + }, + { + "type": "interline_equation", + "bbox": [ + 237, + 248, + 372, + 277 + ], + "lines": [ + { + "bbox": [ + 237, + 248, + 372, + 277 + ], + "spans": [ + { + "bbox": [ + 237, + 248, + 372, + 277 + ], + "score": 0.94, + "content": "D _ { \\lambda } ( x ; f ) = { \\left\\{ \\begin{array} { l l } { \\mathrm { I D } } & { S ( x ) \\geq \\lambda } \\\\ { \\mathrm { O O D } } & { S ( x ) < \\lambda ^ { \\prime } } \\end{array} \\right. }", + "type": "interline_equation", + "image_path": "e87448befed0059cc41d5d7312eecc2686852e4d9c83a86ca7d1674ebc8041fd.jpg" + } + ] + } + ], + "index": 11.5, + "virtual_lines": [ + { + "bbox": [ + 237, + 248, + 372, + 262.5 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 237, + 262.5, + 372, + 277.0 + ], + "spans": [], + "index": 12 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 282, + 505, + 350 + ], + "lines": [ + { + "bbox": [ + 106, + 282, + 506, + 295 + ], + "spans": [ + { + "bbox": [ + 106, + 282, + 211, + 295 + ], + "score": 1.0, + "content": "Building upon the model", + "type": "text" + }, + { + "bbox": [ + 212, + 283, + 287, + 295 + ], + "score": 0.9, + "content": "f \\in \\mathcal { H } : \\mathcal { X } \\to \\mathbb { R } ^ { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 287, + 282, + 506, + 295 + ], + "score": 1.0, + "content": "trained on ID data with the logit outputs, the goal of", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 294, + 505, + 306 + ], + "spans": [ + { + "bbox": [ + 106, + 294, + 275, + 306 + ], + "score": 1.0, + "content": "decision is to utilize the scoring function", + "type": "text" + }, + { + "bbox": [ + 276, + 294, + 324, + 304 + ], + "score": 0.89, + "content": "S : \\mathcal { X } \\mathbb { R }", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 294, + 437, + 306 + ], + "score": 1.0, + "content": "to distinguish the inputs of", + "type": "text" + }, + { + "bbox": [ + 437, + 295, + 452, + 305 + ], + "score": 0.9, + "content": "D _ { \\mathrm { i n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 453, + 294, + 505, + 306 + ], + "score": 1.0, + "content": "from that of", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 107, + 305, + 506, + 318 + ], + "spans": [ + { + "bbox": [ + 107, + 306, + 126, + 317 + ], + "score": 0.89, + "content": "D _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 305, + 140, + 318 + ], + "score": 1.0, + "content": "by", + "type": "text" + }, + { + "bbox": [ + 140, + 305, + 161, + 317 + ], + "score": 0.9, + "content": "S ( x )", + "type": "inline_equation" + }, + { + "bbox": [ + 161, + 305, + 393, + 318 + ], + "score": 1.0, + "content": ". Typically, if the score value is larger than the threshold", + "type": "text" + }, + { + "bbox": [ + 393, + 306, + 400, + 315 + ], + "score": 0.79, + "content": "\\lambda", + "type": "inline_equation" + }, + { + "bbox": [ + 400, + 305, + 487, + 318 + ], + "score": 1.0, + "content": ", the associated input", + "type": "text" + }, + { + "bbox": [ + 487, + 307, + 495, + 315 + ], + "score": 0.77, + "content": "x", + "type": "inline_equation" + }, + { + "bbox": [ + 495, + 305, + 506, + 318 + ], + "score": 1.0, + "content": "is", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 315, + 506, + 329 + ], + "spans": [ + { + "bbox": [ + 105, + 315, + 506, + 329 + ], + "score": 1.0, + "content": "classified as ID and vice versa. We consider several representative scoring functions designed for", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 326, + 505, + 340 + ], + "spans": [ + { + "bbox": [ + 105, + 326, + 505, + 340 + ], + "score": 1.0, + "content": "OOD detection, e.g., MSP (Hendrycks & Gimpel, 2017b), ODIN (Liang et al., 2018), and Energy (Liu", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 338, + 446, + 351 + ], + "spans": [ + { + "bbox": [ + 105, + 338, + 446, + 351 + ], + "score": 1.0, + "content": "et al., 2020). More detailed definitions and implementation can refer to Appendix A.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 15.5 + }, + { + "type": "text", + "bbox": [ + 107, + 354, + 505, + 421 + ], + "lines": [ + { + "bbox": [ + 105, + 354, + 506, + 367 + ], + "spans": [ + { + "bbox": [ + 105, + 354, + 506, + 367 + ], + "score": 1.0, + "content": "To mitigate the issue of over-confident predictions for (Hendrycks & Gimpel, 2017b; Liu et al.,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 365, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 105, + 365, + 325, + 379 + ], + "score": 1.0, + "content": "2020) some OOD data, recent works (Hendrycks et al.,", + "type": "text" + }, + { + "bbox": [ + 326, + 366, + 351, + 377 + ], + "score": 0.66, + "content": "2 0 1 9 \\mathrm { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 352, + 365, + 505, + 379 + ], + "score": 1.0, + "content": "; Tack et al., 2020) utilize the auxiliary", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 376, + 506, + 389 + ], + "spans": [ + { + "bbox": [ + 106, + 376, + 506, + 389 + ], + "score": 1.0, + "content": "unlabeled dataset to regularize the model behavior. Among them, one representative baseline is", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 388, + 505, + 401 + ], + "spans": [ + { + "bbox": [ + 105, + 388, + 505, + 401 + ], + "score": 1.0, + "content": "outlier exposure (OE) (Hendrycks et al., 2019c). OE can further improve the detection performance", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 104, + 397, + 506, + 414 + ], + "spans": [ + { + "bbox": [ + 104, + 397, + 194, + 414 + ], + "score": 1.0, + "content": "by making the model", + "type": "text" + }, + { + "bbox": [ + 194, + 399, + 211, + 411 + ], + "score": 0.9, + "content": "f ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 212, + 397, + 394, + 414 + ], + "score": 1.0, + "content": "finetuned from a surrogate OOD distribution", + "type": "text" + }, + { + "bbox": [ + 394, + 399, + 412, + 411 + ], + "score": 0.91, + "content": "D _ { \\mathrm { o u t } } ^ { \\mathrm { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 413, + 397, + 506, + 414 + ], + "score": 1.0, + "content": ", and its corresponding", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 409, + 267, + 423 + ], + "spans": [ + { + "bbox": [ + 105, + 409, + 267, + 423 + ], + "score": 1.0, + "content": "learning objective is defined as follows,", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21.5 + }, + { + "type": "interline_equation", + "bbox": [ + 208, + 428, + 402, + 442 + ], + "lines": [ + { + "bbox": [ + 208, + 428, + 402, + 442 + ], + "spans": [ + { + "bbox": [ + 208, + 428, + 402, + 442 + ], + "score": 0.9, + "content": "\\mathcal { L } _ { f } = \\mathbb { E } _ { D _ { \\mathrm { i n } } } \\left[ \\ell _ { \\mathrm { C E } } ( f ( x ) , y ) \\right] + \\lambda \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } \\left[ \\ell _ { \\mathrm { O E } } ( f ( x ) ) \\right] ,", + "type": "interline_equation", + "image_path": "ebe06b3345ccb0c28753e1d45a18df14d8791140aca800b8ba6c4a50bc0fa038.jpg" + } + ] + } + ], + "index": 25, + "virtual_lines": [ + { + "bbox": [ + 208, + 428, + 402, + 442 + ], + "spans": [], + "index": 25 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 447, + 505, + 515 + ], + "lines": [ + { + "bbox": [ + 105, + 446, + 505, + 460 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 135, + 460 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 135, + 448, + 142, + 458 + ], + "score": 0.8, + "content": "\\lambda", + "type": "inline_equation" + }, + { + "bbox": [ + 143, + 446, + 264, + 460 + ], + "score": 1.0, + "content": "is the balancing parameter,", + "type": "text" + }, + { + "bbox": [ + 264, + 448, + 290, + 460 + ], + "score": 0.92, + "content": "\\ell _ { \\mathrm { C E } } ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 446, + 450, + 460 + ], + "score": 1.0, + "content": "is the Cross-Entropy (CE) loss, and", + "type": "text" + }, + { + "bbox": [ + 450, + 448, + 476, + 459 + ], + "score": 0.9, + "content": "\\ell _ { \\mathrm { O E } } ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 476, + 446, + 505, + 460 + ], + "score": 1.0, + "content": "is the", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 459, + 505, + 471 + ], + "spans": [ + { + "bbox": [ + 106, + 459, + 448, + 471 + ], + "score": 1.0, + "content": "Kullback-Leibler divergence to the uniform distribution, which can be written as", + "type": "text" + }, + { + "bbox": [ + 448, + 459, + 505, + 471 + ], + "score": 0.92, + "content": "\\ell _ { { \\bf { O E } } } ( h ( { \\bf { x } } ) ) =", + "type": "inline_equation" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 469, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 132, + 483 + ], + "score": 0.87, + "content": "- \\sum k", + "type": "inline_equation" + }, + { + "bbox": [ + 133, + 469, + 182, + 484 + ], + "score": 1.0, + "content": "softmaxk", + "type": "text" + }, + { + "bbox": [ + 182, + 470, + 216, + 482 + ], + "score": 0.73, + "content": "f ( x ) { \\bar { / C } }", + "type": "inline_equation" + }, + { + "bbox": [ + 216, + 469, + 246, + 484 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 246, + 470, + 304, + 482 + ], + "score": 0.59, + "content": "\\mathtt { s o f t m a x } _ { k } ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 305, + 469, + 352, + 484 + ], + "score": 1.0, + "content": "denotes the", + "type": "text" + }, + { + "bbox": [ + 352, + 470, + 359, + 480 + ], + "score": 0.81, + "content": "k", + "type": "inline_equation" + }, + { + "bbox": [ + 360, + 469, + 506, + 484 + ], + "score": 1.0, + "content": "-th element of a softmax output. The", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 480, + 505, + 493 + ], + "spans": [ + { + "bbox": [ + 106, + 480, + 140, + 493 + ], + "score": 1.0, + "content": "OE loss", + "type": "text" + }, + { + "bbox": [ + 141, + 481, + 166, + 493 + ], + "score": 0.91, + "content": "\\ell _ { \\mathrm { O E } } ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 480, + 505, + 493 + ], + "score": 1.0, + "content": "is designed for model regularization, making the model learn from surrogate OOD", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 492, + 506, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 388, + 504 + ], + "score": 1.0, + "content": "inputs to return low-confident predictions. The general formulation of", + "type": "text" + }, + { + "bbox": [ + 389, + 492, + 409, + 503 + ], + "score": 0.29, + "content": "\\operatorname { E q } 2", + "type": "inline_equation" + }, + { + "bbox": [ + 409, + 492, + 506, + 504 + ], + "score": 1.0, + "content": "is also adopted in other", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 502, + 471, + 515 + ], + "spans": [ + { + "bbox": [ + 105, + 502, + 471, + 515 + ], + "score": 1.0, + "content": "related works for designing better tuning objectives that use different auxiliary outlier data.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 28.5 + }, + { + "type": "text", + "bbox": [ + 107, + 519, + 505, + 576 + ], + "lines": [ + { + "bbox": [ + 105, + 519, + 505, + 533 + ], + "spans": [ + { + "bbox": [ + 105, + 519, + 505, + 533 + ], + "score": 1.0, + "content": "Although previous works show promising results via designing scoring functions or regularizing", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 531, + 505, + 543 + ], + "spans": [ + { + "bbox": [ + 106, + 531, + 221, + 543 + ], + "score": 1.0, + "content": "models based on the model", + "type": "text" + }, + { + "bbox": [ + 222, + 531, + 229, + 542 + ], + "score": 0.85, + "content": "f", + "type": "inline_equation" + }, + { + "bbox": [ + 230, + 531, + 505, + 543 + ], + "score": 1.0, + "content": "trained on ID data, few of them investigated the original detection", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 541, + 506, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 541, + 471, + 554 + ], + "score": 1.0, + "content": "capability of the well-trained given model. In this work, we introduce the layer-wise mask", + "type": "text" + }, + { + "bbox": [ + 471, + 543, + 481, + 552 + ], + "score": 0.74, + "content": "m", + "type": "inline_equation" + }, + { + "bbox": [ + 482, + 541, + 506, + 554 + ], + "score": 1.0, + "content": "(Han", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 553, + 506, + 565 + ], + "spans": [ + { + "bbox": [ + 106, + 553, + 506, + 565 + ], + "score": 1.0, + "content": "et al., 2016; Ramanujan et al., 2020) to mine the atypical samples that memorized by the model.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 564, + 506, + 576 + ], + "spans": [ + { + "bbox": [ + 106, + 564, + 275, + 576 + ], + "score": 1.0, + "content": "Accordingly, the decision can be written as", + "type": "text" + }, + { + "bbox": [ + 275, + 564, + 326, + 576 + ], + "score": 0.92, + "content": "D ( x ; m \\odot f )", + "type": "inline_equation" + }, + { + "bbox": [ + 326, + 564, + 464, + 576 + ], + "score": 1.0, + "content": ", and the output of masked model is", + "type": "text" + }, + { + "bbox": [ + 465, + 564, + 503, + 576 + ], + "score": 0.92, + "content": "m \\odot f ( x )", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 564, + 506, + 576 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34 + }, + { + "type": "title", + "bbox": [ + 107, + 590, + 202, + 601 + ], + "lines": [ + { + "bbox": [ + 105, + 588, + 205, + 603 + ], + "spans": [ + { + "bbox": [ + 105, + 588, + 205, + 603 + ], + "score": 1.0, + "content": "2.2 RELATED WORK", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 106, + 610, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "score": 1.0, + "content": "Out-of-distribution Detection without auxiliary data. (Hendrycks & Gimpel, 2017a) formally", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 621, + 506, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 506, + 635 + ], + "score": 1.0, + "content": "shed light on out-of-distribution detection, proposing to use softmax prediction probability as a", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 632, + 506, + 647 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 506, + 647 + ], + "score": 1.0, + "content": "baseline which is demonstrated to be unsuitable for OOD detection. Subsequent works keep focusing", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "score": 1.0, + "content": "on designing post-hoc metrics to distinguish ID samples from OOD samples, among which ODIN", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 655, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 506, + 668 + ], + "score": 1.0, + "content": "(Liang et al., 2018) introduces small perturbations into input images to facilitate the separation of", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 665, + 506, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 506, + 679 + ], + "score": 1.0, + "content": "softmax score, Mahalanobis distance-based confidence score (Lee et al., 2018b) exploits the feature", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "space by obtaining conditional Gaussian distributions, energy-based score (Liu et al., 2020) aligns", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "score": 1.0, + "content": "better with the probability density. Besides designing score functions, many other works pay attention", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "score": 1.0, + "content": "to various aspects to enhance the OOD detection such that LogitNorm (Wei et al., 2022) produces", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 710, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 506, + 722 + ], + "score": 1.0, + "content": "confidence scores by training with a constant vector norm on the logits, and DICE (Sun & Li, 2022)", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 721, + 474, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 474, + 733 + ], + "score": 1.0, + "content": "reduces the variance of the output distribution by leveraging the sparsification of the model.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 43 + } + ], + "page_idx": 2, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "score": 1.0, + "content": "3", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 81, + 200, + 94 + ], + "lines": [ + { + "bbox": [ + 104, + 79, + 202, + 97 + ], + "spans": [ + { + "bbox": [ + 104, + 79, + 202, + 97 + ], + "score": 1.0, + "content": "2 BACKGROUND", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 104, + 106, + 478, + 118 + ], + "lines": [ + { + "bbox": [ + 105, + 106, + 479, + 119 + ], + "spans": [ + { + "bbox": [ + 105, + 106, + 479, + 119 + ], + "score": 1.0, + "content": "In this section, we briefly introduce the preliminaries and related work about OOD detection.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1, + "bbox_fs": [ + 105, + 106, + 479, + 119 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 133, + 200, + 145 + ], + "lines": [ + { + "bbox": [ + 106, + 133, + 200, + 145 + ], + "spans": [ + { + "bbox": [ + 106, + 133, + 200, + 145 + ], + "score": 1.0, + "content": "2.1 PRELIMINARIES", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 2 + }, + { + "type": "text", + "bbox": [ + 106, + 154, + 505, + 243 + ], + "lines": [ + { + "bbox": [ + 106, + 154, + 506, + 166 + ], + "spans": [ + { + "bbox": [ + 106, + 154, + 359, + 166 + ], + "score": 1.0, + "content": "We consider multi-class classification as the training task, where", + "type": "text" + }, + { + "bbox": [ + 360, + 154, + 394, + 165 + ], + "score": 0.92, + "content": "\\mathcal { X } \\subset \\mathbb { R } ^ { d }", + "type": "inline_equation" + }, + { + "bbox": [ + 395, + 154, + 506, + 166 + ], + "score": 1.0, + "content": "denotes the input space and", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 107, + 165, + 505, + 177 + ], + "spans": [ + { + "bbox": [ + 107, + 165, + 174, + 177 + ], + "score": 0.93, + "content": "\\mathcal { Y } = \\{ 1 , \\ldots , C \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 174, + 165, + 505, + 177 + ], + "score": 1.0, + "content": "denotes the label space. Given the model deployed in the real world, the reliable", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 176, + 505, + 188 + ], + "spans": [ + { + "bbox": [ + 105, + 176, + 505, + 188 + ], + "score": 1.0, + "content": "classifier is expected to figure out the OOD input, which can be formulated as a binary classification", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 187, + 506, + 199 + ], + "spans": [ + { + "bbox": [ + 105, + 187, + 170, + 199 + ], + "score": 1.0, + "content": "problem. Given", + "type": "text" + }, + { + "bbox": [ + 170, + 187, + 179, + 198 + ], + "score": 0.8, + "content": "\\mathcal { P }", + "type": "inline_equation" + }, + { + "bbox": [ + 180, + 187, + 264, + 199 + ], + "score": 1.0, + "content": ", the distribution over", + "type": "text" + }, + { + "bbox": [ + 264, + 188, + 292, + 198 + ], + "score": 0.89, + "content": "\\mathcal { X } \\times \\mathcal { V }", + "type": "inline_equation" + }, + { + "bbox": [ + 293, + 187, + 345, + 199 + ], + "score": 1.0, + "content": ", we consider", + "type": "text" + }, + { + "bbox": [ + 345, + 187, + 361, + 198 + ], + "score": 0.89, + "content": "D _ { \\mathrm { i n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 361, + 187, + 481, + 199 + ], + "score": 1.0, + "content": "as the marginal distribution of", + "type": "text" + }, + { + "bbox": [ + 482, + 188, + 490, + 198 + ], + "score": 0.83, + "content": "\\mathcal { P }", + "type": "inline_equation" + }, + { + "bbox": [ + 490, + 187, + 506, + 199 + ], + "score": 1.0, + "content": "for", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 107, + 196, + 504, + 212 + ], + "spans": [ + { + "bbox": [ + 107, + 199, + 116, + 208 + ], + "score": 0.81, + "content": "\\mathcal { X }", + "type": "inline_equation" + }, + { + "bbox": [ + 117, + 196, + 485, + 212 + ], + "score": 1.0, + "content": ", namely, the distribution of ID data. At test time, the environment can present a distribution", + "type": "text" + }, + { + "bbox": [ + 485, + 199, + 504, + 209 + ], + "score": 0.88, + "content": "D _ { \\mathrm { o u t } }", + "type": "inline_equation" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 209, + 506, + 221 + ], + "spans": [ + { + "bbox": [ + 105, + 209, + 126, + 221 + ], + "score": 1.0, + "content": "over", + "type": "text" + }, + { + "bbox": [ + 126, + 210, + 136, + 219 + ], + "score": 0.82, + "content": "\\mathcal { X }", + "type": "inline_equation" + }, + { + "bbox": [ + 136, + 209, + 324, + 221 + ], + "score": 1.0, + "content": "of OOD data. In general, the OOD distribution", + "type": "text" + }, + { + "bbox": [ + 324, + 209, + 343, + 220 + ], + "score": 0.9, + "content": "D _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 343, + 209, + 506, + 221 + ], + "score": 1.0, + "content": "is defined as an irrelevant distribution of", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 220, + 505, + 232 + ], + "spans": [ + { + "bbox": [ + 105, + 220, + 276, + 232 + ], + "score": 1.0, + "content": "which the label set has no intersection with", + "type": "text" + }, + { + "bbox": [ + 276, + 221, + 285, + 231 + ], + "score": 0.81, + "content": "\\mathcal { V }", + "type": "inline_equation" + }, + { + "bbox": [ + 285, + 220, + 505, + 232 + ], + "score": 1.0, + "content": "Yang et al. (2021) and therefore should not be predicted", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 232, + 356, + 243 + ], + "spans": [ + { + "bbox": [ + 106, + 232, + 344, + 243 + ], + "score": 1.0, + "content": "by the model. The decision can be made with the threshold", + "type": "text" + }, + { + "bbox": [ + 345, + 232, + 351, + 241 + ], + "score": 0.81, + "content": "\\lambda", + "type": "inline_equation" + }, + { + "bbox": [ + 352, + 232, + 356, + 243 + ], + "score": 1.0, + "content": ":", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 6.5, + "bbox_fs": [ + 105, + 154, + 506, + 243 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 237, + 248, + 372, + 277 + ], + "lines": [ + { + "bbox": [ + 237, + 248, + 372, + 277 + ], + "spans": [ + { + "bbox": [ + 237, + 248, + 372, + 277 + ], + "score": 0.94, + "content": "D _ { \\lambda } ( x ; f ) = { \\left\\{ \\begin{array} { l l } { \\mathrm { I D } } & { S ( x ) \\geq \\lambda } \\\\ { \\mathrm { O O D } } & { S ( x ) < \\lambda ^ { \\prime } } \\end{array} \\right. }", + "type": "interline_equation", + "image_path": "e87448befed0059cc41d5d7312eecc2686852e4d9c83a86ca7d1674ebc8041fd.jpg" + } + ] + } + ], + "index": 11.5, + "virtual_lines": [ + { + "bbox": [ + 237, + 248, + 372, + 262.5 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 237, + 262.5, + 372, + 277.0 + ], + "spans": [], + "index": 12 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 282, + 505, + 350 + ], + "lines": [ + { + "bbox": [ + 106, + 282, + 506, + 295 + ], + "spans": [ + { + "bbox": [ + 106, + 282, + 211, + 295 + ], + "score": 1.0, + "content": "Building upon the model", + "type": "text" + }, + { + "bbox": [ + 212, + 283, + 287, + 295 + ], + "score": 0.9, + "content": "f \\in \\mathcal { H } : \\mathcal { X } \\to \\mathbb { R } ^ { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 287, + 282, + 506, + 295 + ], + "score": 1.0, + "content": "trained on ID data with the logit outputs, the goal of", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 294, + 505, + 306 + ], + "spans": [ + { + "bbox": [ + 106, + 294, + 275, + 306 + ], + "score": 1.0, + "content": "decision is to utilize the scoring function", + "type": "text" + }, + { + "bbox": [ + 276, + 294, + 324, + 304 + ], + "score": 0.89, + "content": "S : \\mathcal { X } \\mathbb { R }", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 294, + 437, + 306 + ], + "score": 1.0, + "content": "to distinguish the inputs of", + "type": "text" + }, + { + "bbox": [ + 437, + 295, + 452, + 305 + ], + "score": 0.9, + "content": "D _ { \\mathrm { i n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 453, + 294, + 505, + 306 + ], + "score": 1.0, + "content": "from that of", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 107, + 305, + 506, + 318 + ], + "spans": [ + { + "bbox": [ + 107, + 306, + 126, + 317 + ], + "score": 0.89, + "content": "D _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 305, + 140, + 318 + ], + "score": 1.0, + "content": "by", + "type": "text" + }, + { + "bbox": [ + 140, + 305, + 161, + 317 + ], + "score": 0.9, + "content": "S ( x )", + "type": "inline_equation" + }, + { + "bbox": [ + 161, + 305, + 393, + 318 + ], + "score": 1.0, + "content": ". Typically, if the score value is larger than the threshold", + "type": "text" + }, + { + "bbox": [ + 393, + 306, + 400, + 315 + ], + "score": 0.79, + "content": "\\lambda", + "type": "inline_equation" + }, + { + "bbox": [ + 400, + 305, + 487, + 318 + ], + "score": 1.0, + "content": ", the associated input", + "type": "text" + }, + { + "bbox": [ + 487, + 307, + 495, + 315 + ], + "score": 0.77, + "content": "x", + "type": "inline_equation" + }, + { + "bbox": [ + 495, + 305, + 506, + 318 + ], + "score": 1.0, + "content": "is", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 315, + 506, + 329 + ], + "spans": [ + { + "bbox": [ + 105, + 315, + 506, + 329 + ], + "score": 1.0, + "content": "classified as ID and vice versa. We consider several representative scoring functions designed for", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 326, + 505, + 340 + ], + "spans": [ + { + "bbox": [ + 105, + 326, + 505, + 340 + ], + "score": 1.0, + "content": "OOD detection, e.g., MSP (Hendrycks & Gimpel, 2017b), ODIN (Liang et al., 2018), and Energy (Liu", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 338, + 446, + 351 + ], + "spans": [ + { + "bbox": [ + 105, + 338, + 446, + 351 + ], + "score": 1.0, + "content": "et al., 2020). More detailed definitions and implementation can refer to Appendix A.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 15.5, + "bbox_fs": [ + 105, + 282, + 506, + 351 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 354, + 505, + 421 + ], + "lines": [ + { + "bbox": [ + 105, + 354, + 506, + 367 + ], + "spans": [ + { + "bbox": [ + 105, + 354, + 506, + 367 + ], + "score": 1.0, + "content": "To mitigate the issue of over-confident predictions for (Hendrycks & Gimpel, 2017b; Liu et al.,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 365, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 105, + 365, + 325, + 379 + ], + "score": 1.0, + "content": "2020) some OOD data, recent works (Hendrycks et al.,", + "type": "text" + }, + { + "bbox": [ + 326, + 366, + 351, + 377 + ], + "score": 0.66, + "content": "2 0 1 9 \\mathrm { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 352, + 365, + 505, + 379 + ], + "score": 1.0, + "content": "; Tack et al., 2020) utilize the auxiliary", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 376, + 506, + 389 + ], + "spans": [ + { + "bbox": [ + 106, + 376, + 506, + 389 + ], + "score": 1.0, + "content": "unlabeled dataset to regularize the model behavior. Among them, one representative baseline is", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 388, + 505, + 401 + ], + "spans": [ + { + "bbox": [ + 105, + 388, + 505, + 401 + ], + "score": 1.0, + "content": "outlier exposure (OE) (Hendrycks et al., 2019c). OE can further improve the detection performance", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 104, + 397, + 506, + 414 + ], + "spans": [ + { + "bbox": [ + 104, + 397, + 194, + 414 + ], + "score": 1.0, + "content": "by making the model", + "type": "text" + }, + { + "bbox": [ + 194, + 399, + 211, + 411 + ], + "score": 0.9, + "content": "f ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 212, + 397, + 394, + 414 + ], + "score": 1.0, + "content": "finetuned from a surrogate OOD distribution", + "type": "text" + }, + { + "bbox": [ + 394, + 399, + 412, + 411 + ], + "score": 0.91, + "content": "D _ { \\mathrm { o u t } } ^ { \\mathrm { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 413, + 397, + 506, + 414 + ], + "score": 1.0, + "content": ", and its corresponding", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 409, + 267, + 423 + ], + "spans": [ + { + "bbox": [ + 105, + 409, + 267, + 423 + ], + "score": 1.0, + "content": "learning objective is defined as follows,", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21.5, + "bbox_fs": [ + 104, + 354, + 506, + 423 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 208, + 428, + 402, + 442 + ], + "lines": [ + { + "bbox": [ + 208, + 428, + 402, + 442 + ], + "spans": [ + { + "bbox": [ + 208, + 428, + 402, + 442 + ], + "score": 0.9, + "content": "\\mathcal { L } _ { f } = \\mathbb { E } _ { D _ { \\mathrm { i n } } } \\left[ \\ell _ { \\mathrm { C E } } ( f ( x ) , y ) \\right] + \\lambda \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } \\left[ \\ell _ { \\mathrm { O E } } ( f ( x ) ) \\right] ,", + "type": "interline_equation", + "image_path": "ebe06b3345ccb0c28753e1d45a18df14d8791140aca800b8ba6c4a50bc0fa038.jpg" + } + ] + } + ], + "index": 25, + "virtual_lines": [ + { + "bbox": [ + 208, + 428, + 402, + 442 + ], + "spans": [], + "index": 25 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 447, + 505, + 515 + ], + "lines": [ + { + "bbox": [ + 105, + 446, + 505, + 460 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 135, + 460 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 135, + 448, + 142, + 458 + ], + "score": 0.8, + "content": "\\lambda", + "type": "inline_equation" + }, + { + "bbox": [ + 143, + 446, + 264, + 460 + ], + "score": 1.0, + "content": "is the balancing parameter,", + "type": "text" + }, + { + "bbox": [ + 264, + 448, + 290, + 460 + ], + "score": 0.92, + "content": "\\ell _ { \\mathrm { C E } } ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 446, + 450, + 460 + ], + "score": 1.0, + "content": "is the Cross-Entropy (CE) loss, and", + "type": "text" + }, + { + "bbox": [ + 450, + 448, + 476, + 459 + ], + "score": 0.9, + "content": "\\ell _ { \\mathrm { O E } } ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 476, + 446, + 505, + 460 + ], + "score": 1.0, + "content": "is the", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 459, + 505, + 471 + ], + "spans": [ + { + "bbox": [ + 106, + 459, + 448, + 471 + ], + "score": 1.0, + "content": "Kullback-Leibler divergence to the uniform distribution, which can be written as", + "type": "text" + }, + { + "bbox": [ + 448, + 459, + 505, + 471 + ], + "score": 0.92, + "content": "\\ell _ { { \\bf { O E } } } ( h ( { \\bf { x } } ) ) =", + "type": "inline_equation" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 469, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 132, + 483 + ], + "score": 0.87, + "content": "- \\sum k", + "type": "inline_equation" + }, + { + "bbox": [ + 133, + 469, + 182, + 484 + ], + "score": 1.0, + "content": "softmaxk", + "type": "text" + }, + { + "bbox": [ + 182, + 470, + 216, + 482 + ], + "score": 0.73, + "content": "f ( x ) { \\bar { / C } }", + "type": "inline_equation" + }, + { + "bbox": [ + 216, + 469, + 246, + 484 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 246, + 470, + 304, + 482 + ], + "score": 0.59, + "content": "\\mathtt { s o f t m a x } _ { k } ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 305, + 469, + 352, + 484 + ], + "score": 1.0, + "content": "denotes the", + "type": "text" + }, + { + "bbox": [ + 352, + 470, + 359, + 480 + ], + "score": 0.81, + "content": "k", + "type": "inline_equation" + }, + { + "bbox": [ + 360, + 469, + 506, + 484 + ], + "score": 1.0, + "content": "-th element of a softmax output. The", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 480, + 505, + 493 + ], + "spans": [ + { + "bbox": [ + 106, + 480, + 140, + 493 + ], + "score": 1.0, + "content": "OE loss", + "type": "text" + }, + { + "bbox": [ + 141, + 481, + 166, + 493 + ], + "score": 0.91, + "content": "\\ell _ { \\mathrm { O E } } ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 480, + 505, + 493 + ], + "score": 1.0, + "content": "is designed for model regularization, making the model learn from surrogate OOD", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 492, + 506, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 388, + 504 + ], + "score": 1.0, + "content": "inputs to return low-confident predictions. The general formulation of", + "type": "text" + }, + { + "bbox": [ + 389, + 492, + 409, + 503 + ], + "score": 0.29, + "content": "\\operatorname { E q } 2", + "type": "inline_equation" + }, + { + "bbox": [ + 409, + 492, + 506, + 504 + ], + "score": 1.0, + "content": "is also adopted in other", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 502, + 471, + 515 + ], + "spans": [ + { + "bbox": [ + 105, + 502, + 471, + 515 + ], + "score": 1.0, + "content": "related works for designing better tuning objectives that use different auxiliary outlier data.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 28.5, + "bbox_fs": [ + 105, + 446, + 506, + 515 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 519, + 505, + 576 + ], + "lines": [ + { + "bbox": [ + 105, + 519, + 505, + 533 + ], + "spans": [ + { + "bbox": [ + 105, + 519, + 505, + 533 + ], + "score": 1.0, + "content": "Although previous works show promising results via designing scoring functions or regularizing", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 531, + 505, + 543 + ], + "spans": [ + { + "bbox": [ + 106, + 531, + 221, + 543 + ], + "score": 1.0, + "content": "models based on the model", + "type": "text" + }, + { + "bbox": [ + 222, + 531, + 229, + 542 + ], + "score": 0.85, + "content": "f", + "type": "inline_equation" + }, + { + "bbox": [ + 230, + 531, + 505, + 543 + ], + "score": 1.0, + "content": "trained on ID data, few of them investigated the original detection", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 541, + 506, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 541, + 471, + 554 + ], + "score": 1.0, + "content": "capability of the well-trained given model. In this work, we introduce the layer-wise mask", + "type": "text" + }, + { + "bbox": [ + 471, + 543, + 481, + 552 + ], + "score": 0.74, + "content": "m", + "type": "inline_equation" + }, + { + "bbox": [ + 482, + 541, + 506, + 554 + ], + "score": 1.0, + "content": "(Han", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 553, + 506, + 565 + ], + "spans": [ + { + "bbox": [ + 106, + 553, + 506, + 565 + ], + "score": 1.0, + "content": "et al., 2016; Ramanujan et al., 2020) to mine the atypical samples that memorized by the model.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 564, + 506, + 576 + ], + "spans": [ + { + "bbox": [ + 106, + 564, + 275, + 576 + ], + "score": 1.0, + "content": "Accordingly, the decision can be written as", + "type": "text" + }, + { + "bbox": [ + 275, + 564, + 326, + 576 + ], + "score": 0.92, + "content": "D ( x ; m \\odot f )", + "type": "inline_equation" + }, + { + "bbox": [ + 326, + 564, + 464, + 576 + ], + "score": 1.0, + "content": ", and the output of masked model is", + "type": "text" + }, + { + "bbox": [ + 465, + 564, + 503, + 576 + ], + "score": 0.92, + "content": "m \\odot f ( x )", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 564, + 506, + 576 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34, + "bbox_fs": [ + 105, + 519, + 506, + 576 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 590, + 202, + 601 + ], + "lines": [ + { + "bbox": [ + 105, + 588, + 205, + 603 + ], + "spans": [ + { + "bbox": [ + 105, + 588, + 205, + 603 + ], + "score": 1.0, + "content": "2.2 RELATED WORK", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 106, + 610, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "score": 1.0, + "content": "Out-of-distribution Detection without auxiliary data. (Hendrycks & Gimpel, 2017a) formally", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 621, + 506, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 506, + 635 + ], + "score": 1.0, + "content": "shed light on out-of-distribution detection, proposing to use softmax prediction probability as a", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 632, + 506, + 647 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 506, + 647 + ], + "score": 1.0, + "content": "baseline which is demonstrated to be unsuitable for OOD detection. Subsequent works keep focusing", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "score": 1.0, + "content": "on designing post-hoc metrics to distinguish ID samples from OOD samples, among which ODIN", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 655, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 506, + 668 + ], + "score": 1.0, + "content": "(Liang et al., 2018) introduces small perturbations into input images to facilitate the separation of", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 665, + 506, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 506, + 679 + ], + "score": 1.0, + "content": "softmax score, Mahalanobis distance-based confidence score (Lee et al., 2018b) exploits the feature", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "space by obtaining conditional Gaussian distributions, energy-based score (Liu et al., 2020) aligns", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "score": 1.0, + "content": "better with the probability density. Besides designing score functions, many other works pay attention", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "score": 1.0, + "content": "to various aspects to enhance the OOD detection such that LogitNorm (Wei et al., 2022) produces", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 710, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 506, + 722 + ], + "score": 1.0, + "content": "confidence scores by training with a constant vector norm on the logits, and DICE (Sun & Li, 2022)", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 721, + 474, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 474, + 733 + ], + "score": 1.0, + "content": "reduces the variance of the output distribution by leveraging the sparsification of the model.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 43, + "bbox_fs": [ + 105, + 611, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 111, + 83, + 498, + 280 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 111, + 83, + 498, + 280 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 111, + 83, + 498, + 280 + ], + "spans": [ + { + "bbox": [ + 111, + 83, + 498, + 280 + ], + "score": 0.975, + "type": "image", + "image_path": "3eb50f356c32cc81abd5826ebf220eef6e801f6950413747092d4d1cf0fbe6f9.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 111, + 83, + 498, + 148.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 111, + 148.66666666666669, + 498, + 214.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 111, + 214.33333333333337, + 498, + 280.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 286, + 505, + 386 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 286, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 286, + 505, + 298 + ], + "score": 1.0, + "content": "Figure 2: We train on CIFAR-10 for the original multi-class classification and check the details", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 297, + 506, + 310 + ], + "spans": [ + { + "bbox": [ + 105, + 297, + 506, + 310 + ], + "score": 1.0, + "content": "as follows: (a) training/testing loss on ID data; (b) training/testing accuracy on ID data; (c) curves", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 308, + 505, + 320 + ], + "spans": [ + { + "bbox": [ + 106, + 308, + 505, + 320 + ], + "score": 1.0, + "content": "of FPR95 based on ODIN score; (d) curves of FPR95 based on Energy score; (e) visualization of", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 320, + 504, + 331 + ], + "spans": [ + { + "bbox": [ + 106, + 320, + 504, + 331 + ], + "score": 1.0, + "content": "correct classified samples at Epoch 60; (f) visualization of misclassified samples at Epoch 60. We", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 330, + 505, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 330, + 505, + 343 + ], + "score": 1.0, + "content": "investigate the once-covered OOD detection capability by checking the model behavior during its", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 341, + 505, + 354 + ], + "spans": [ + { + "bbox": [ + 106, + 341, + 505, + 354 + ], + "score": 1.0, + "content": "training phase. We take a closer look at the corresponding training and testing performance with", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 353, + 505, + 364 + ], + "spans": [ + { + "bbox": [ + 106, + 353, + 505, + 364 + ], + "score": 1.0, + "content": "the OOD detection capability indicated by two different scores. Through comparison, we find that", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 363, + 506, + 376 + ], + "spans": [ + { + "bbox": [ + 106, + 363, + 506, + 376 + ], + "score": 1.0, + "content": "achieving a reasonably small loss value (at round Epoch 60) on ID data is enough for OOD detection.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 374, + 498, + 387 + ], + "spans": [ + { + "bbox": [ + 106, + 374, + 498, + 387 + ], + "score": 1.0, + "content": "However, continually optimizing on those atypical samples can impair the detection performance.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 7 + } + ], + "index": 4.0 + }, + { + "type": "text", + "bbox": [ + 107, + 398, + 505, + 563 + ], + "lines": [ + { + "bbox": [ + 106, + 398, + 505, + 410 + ], + "spans": [ + { + "bbox": [ + 106, + 398, + 505, + 410 + ], + "score": 1.0, + "content": "Out-of-distribution Detection with auxiliary data. Another promising direction towards OOD", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 409, + 505, + 422 + ], + "spans": [ + { + "bbox": [ + 105, + 409, + 505, + 422 + ], + "score": 1.0, + "content": "detection involves the auxiliary outliers for model regularization. On one hand, some works generate", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 419, + 505, + 433 + ], + "spans": [ + { + "bbox": [ + 105, + 419, + 505, + 433 + ], + "score": 1.0, + "content": "virtual outliers such that Lee et al. (2018a) uses generative adversarial networks to generate boundary", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 431, + 504, + 443 + ], + "spans": [ + { + "bbox": [ + 105, + 431, + 504, + 443 + ], + "score": 1.0, + "content": "samples, VOS (Du et al., 2022a) regularizes the decision boundary by adaptively sampling virtual", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 441, + 505, + 455 + ], + "spans": [ + { + "bbox": [ + 105, + 441, + 505, + 455 + ], + "score": 1.0, + "content": "outliers from the low-likelihood region. On the other hand, other works tend to exploit information", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 452, + 505, + 465 + ], + "spans": [ + { + "bbox": [ + 105, + 452, + 505, + 465 + ], + "score": 1.0, + "content": "from natural outliers, such that outlier exposure (OE) is introduced by Hendrycks et al. (2019b), given", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 463, + 505, + 476 + ], + "spans": [ + { + "bbox": [ + 105, + 463, + 505, + 476 + ], + "score": 1.0, + "content": "that diverse data are available in enormous quantities. (Yu & Aizawa, 2019) train an additional \"head\"", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 474, + 506, + 487 + ], + "spans": [ + { + "bbox": [ + 105, + 474, + 506, + 487 + ], + "score": 1.0, + "content": "and maximizes the discrepancy of decision boundaries of the two heads to detect OOD samples.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 485, + 505, + 499 + ], + "spans": [ + { + "bbox": [ + 105, + 485, + 505, + 499 + ], + "score": 1.0, + "content": "Energy-bounded learning (Liu et al., 2020) fine-tunes the neural network to widen the energy gap by", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 496, + 507, + 510 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 507, + 510 + ], + "score": 1.0, + "content": "adding an energy loss term to the objective. Some other works also highlight the sampling strategy,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 507, + 505, + 520 + ], + "spans": [ + { + "bbox": [ + 105, + 507, + 505, + 520 + ], + "score": 1.0, + "content": "such that ATOM (Chen et al., 2021) greedily utilizes informative auxiliary data to tighten the decision", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 518, + 506, + 532 + ], + "spans": [ + { + "bbox": [ + 105, + 518, + 506, + 532 + ], + "score": 1.0, + "content": "boundary for OOD detection, and POEM (Ming et al., 2022) adopts Thompson sampling to contour", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 529, + 506, + 542 + ], + "spans": [ + { + "bbox": [ + 105, + 529, + 506, + 542 + ], + "score": 1.0, + "content": "the decision boundary precisely. The performance of training with outliers is usually superior to that", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 540, + 507, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 540, + 507, + 554 + ], + "score": 1.0, + "content": "without outliers, shown in many recent works (Mohseni et al., 2020; Liu et al., 2020; Fort et al., 2021;", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 551, + 483, + 564 + ], + "spans": [ + { + "bbox": [ + 106, + 551, + 483, + 564 + ], + "score": 1.0, + "content": "Sun et al., 2021; Sehwag et al., 2021; Yang et al., 2021; Chen et al., 2021; Salehi et al., 2021).", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 19 + }, + { + "type": "title", + "bbox": [ + 107, + 581, + 343, + 594 + ], + "lines": [ + { + "bbox": [ + 104, + 579, + 344, + 595 + ], + "spans": [ + { + "bbox": [ + 104, + 579, + 344, + 595 + ], + "score": 1.0, + "content": "3 PROPOSED METHOD: UNLEASHING MASK", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 107, + 606, + 505, + 662 + ], + "lines": [ + { + "bbox": [ + 106, + 607, + 505, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 607, + 505, + 619 + ], + "score": 1.0, + "content": "In this section, we introduce our new method, i.e., Unleashing Mask (UM), to reveal the potential", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "score": 1.0, + "content": "OOD detection capability of the well-trained model. First, we present and discuss the important", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 628, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 105, + 628, + 506, + 642 + ], + "score": 1.0, + "content": "observation that inspires our methods (Section 3.1). Second, we provide the insights behind the two", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 639, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 105, + 639, + 505, + 653 + ], + "score": 1.0, + "content": "critical parts of our UM (Section 3.2). Lastly, we describe the framework and the learning objective", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 650, + 496, + 663 + ], + "spans": [ + { + "bbox": [ + 106, + 650, + 496, + 663 + ], + "score": 1.0, + "content": "of UM that incorporates the previous component, as well as its variant, i.e., UMAP (Section 3.3).", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 30 + }, + { + "type": "title", + "bbox": [ + 107, + 677, + 335, + 689 + ], + "lines": [ + { + "bbox": [ + 105, + 677, + 336, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 336, + 690 + ], + "score": 1.0, + "content": "3.1 ONCE-COVERED OOD DETECTION CAPABILITY", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 33 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 712 + ], + "score": 1.0, + "content": "First, we present the phenomenon of the inconsistent trend between a better OOD detection capability", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "and smaller training error during training. Empirically, as shown in Figure 1, we plot the OOD", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 721, + 504, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 504, + 733 + ], + "score": 1.0, + "content": "detection performance during the model training after multiple runs of the experiments. Across three", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35 + } + ], + "page_idx": 3, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 307, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 759 + ], + "lines": [] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 111, + 83, + 498, + 280 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 111, + 83, + 498, + 280 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 111, + 83, + 498, + 280 + ], + "spans": [ + { + "bbox": [ + 111, + 83, + 498, + 280 + ], + "score": 0.975, + "type": "image", + "image_path": "3eb50f356c32cc81abd5826ebf220eef6e801f6950413747092d4d1cf0fbe6f9.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 111, + 83, + 498, + 148.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 111, + 148.66666666666669, + 498, + 214.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 111, + 214.33333333333337, + 498, + 280.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 286, + 505, + 386 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 286, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 286, + 505, + 298 + ], + "score": 1.0, + "content": "Figure 2: We train on CIFAR-10 for the original multi-class classification and check the details", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 297, + 506, + 310 + ], + "spans": [ + { + "bbox": [ + 105, + 297, + 506, + 310 + ], + "score": 1.0, + "content": "as follows: (a) training/testing loss on ID data; (b) training/testing accuracy on ID data; (c) curves", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 308, + 505, + 320 + ], + "spans": [ + { + "bbox": [ + 106, + 308, + 505, + 320 + ], + "score": 1.0, + "content": "of FPR95 based on ODIN score; (d) curves of FPR95 based on Energy score; (e) visualization of", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 320, + 504, + 331 + ], + "spans": [ + { + "bbox": [ + 106, + 320, + 504, + 331 + ], + "score": 1.0, + "content": "correct classified samples at Epoch 60; (f) visualization of misclassified samples at Epoch 60. We", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 330, + 505, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 330, + 505, + 343 + ], + "score": 1.0, + "content": "investigate the once-covered OOD detection capability by checking the model behavior during its", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 341, + 505, + 354 + ], + "spans": [ + { + "bbox": [ + 106, + 341, + 505, + 354 + ], + "score": 1.0, + "content": "training phase. We take a closer look at the corresponding training and testing performance with", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 353, + 505, + 364 + ], + "spans": [ + { + "bbox": [ + 106, + 353, + 505, + 364 + ], + "score": 1.0, + "content": "the OOD detection capability indicated by two different scores. Through comparison, we find that", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 363, + 506, + 376 + ], + "spans": [ + { + "bbox": [ + 106, + 363, + 506, + 376 + ], + "score": 1.0, + "content": "achieving a reasonably small loss value (at round Epoch 60) on ID data is enough for OOD detection.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 374, + 498, + 387 + ], + "spans": [ + { + "bbox": [ + 106, + 374, + 498, + 387 + ], + "score": 1.0, + "content": "However, continually optimizing on those atypical samples can impair the detection performance.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 7 + } + ], + "index": 4.0 + }, + { + "type": "text", + "bbox": [ + 107, + 398, + 505, + 563 + ], + "lines": [ + { + "bbox": [ + 106, + 398, + 505, + 410 + ], + "spans": [ + { + "bbox": [ + 106, + 398, + 505, + 410 + ], + "score": 1.0, + "content": "Out-of-distribution Detection with auxiliary data. Another promising direction towards OOD", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 409, + 505, + 422 + ], + "spans": [ + { + "bbox": [ + 105, + 409, + 505, + 422 + ], + "score": 1.0, + "content": "detection involves the auxiliary outliers for model regularization. On one hand, some works generate", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 419, + 505, + 433 + ], + "spans": [ + { + "bbox": [ + 105, + 419, + 505, + 433 + ], + "score": 1.0, + "content": "virtual outliers such that Lee et al. (2018a) uses generative adversarial networks to generate boundary", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 431, + 504, + 443 + ], + "spans": [ + { + "bbox": [ + 105, + 431, + 504, + 443 + ], + "score": 1.0, + "content": "samples, VOS (Du et al., 2022a) regularizes the decision boundary by adaptively sampling virtual", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 441, + 505, + 455 + ], + "spans": [ + { + "bbox": [ + 105, + 441, + 505, + 455 + ], + "score": 1.0, + "content": "outliers from the low-likelihood region. On the other hand, other works tend to exploit information", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 452, + 505, + 465 + ], + "spans": [ + { + "bbox": [ + 105, + 452, + 505, + 465 + ], + "score": 1.0, + "content": "from natural outliers, such that outlier exposure (OE) is introduced by Hendrycks et al. (2019b), given", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 463, + 505, + 476 + ], + "spans": [ + { + "bbox": [ + 105, + 463, + 505, + 476 + ], + "score": 1.0, + "content": "that diverse data are available in enormous quantities. (Yu & Aizawa, 2019) train an additional \"head\"", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 474, + 506, + 487 + ], + "spans": [ + { + "bbox": [ + 105, + 474, + 506, + 487 + ], + "score": 1.0, + "content": "and maximizes the discrepancy of decision boundaries of the two heads to detect OOD samples.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 485, + 505, + 499 + ], + "spans": [ + { + "bbox": [ + 105, + 485, + 505, + 499 + ], + "score": 1.0, + "content": "Energy-bounded learning (Liu et al., 2020) fine-tunes the neural network to widen the energy gap by", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 496, + 507, + 510 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 507, + 510 + ], + "score": 1.0, + "content": "adding an energy loss term to the objective. Some other works also highlight the sampling strategy,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 507, + 505, + 520 + ], + "spans": [ + { + "bbox": [ + 105, + 507, + 505, + 520 + ], + "score": 1.0, + "content": "such that ATOM (Chen et al., 2021) greedily utilizes informative auxiliary data to tighten the decision", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 518, + 506, + 532 + ], + "spans": [ + { + "bbox": [ + 105, + 518, + 506, + 532 + ], + "score": 1.0, + "content": "boundary for OOD detection, and POEM (Ming et al., 2022) adopts Thompson sampling to contour", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 529, + 506, + 542 + ], + "spans": [ + { + "bbox": [ + 105, + 529, + 506, + 542 + ], + "score": 1.0, + "content": "the decision boundary precisely. The performance of training with outliers is usually superior to that", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 540, + 507, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 540, + 507, + 554 + ], + "score": 1.0, + "content": "without outliers, shown in many recent works (Mohseni et al., 2020; Liu et al., 2020; Fort et al., 2021;", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 551, + 483, + 564 + ], + "spans": [ + { + "bbox": [ + 106, + 551, + 483, + 564 + ], + "score": 1.0, + "content": "Sun et al., 2021; Sehwag et al., 2021; Yang et al., 2021; Chen et al., 2021; Salehi et al., 2021).", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 19, + "bbox_fs": [ + 105, + 398, + 507, + 564 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 581, + 343, + 594 + ], + "lines": [ + { + "bbox": [ + 104, + 579, + 344, + 595 + ], + "spans": [ + { + "bbox": [ + 104, + 579, + 344, + 595 + ], + "score": 1.0, + "content": "3 PROPOSED METHOD: UNLEASHING MASK", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 107, + 606, + 505, + 662 + ], + "lines": [ + { + "bbox": [ + 106, + 607, + 505, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 607, + 505, + 619 + ], + "score": 1.0, + "content": "In this section, we introduce our new method, i.e., Unleashing Mask (UM), to reveal the potential", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "score": 1.0, + "content": "OOD detection capability of the well-trained model. First, we present and discuss the important", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 628, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 105, + 628, + 506, + 642 + ], + "score": 1.0, + "content": "observation that inspires our methods (Section 3.1). Second, we provide the insights behind the two", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 639, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 105, + 639, + 505, + 653 + ], + "score": 1.0, + "content": "critical parts of our UM (Section 3.2). Lastly, we describe the framework and the learning objective", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 650, + 496, + 663 + ], + "spans": [ + { + "bbox": [ + 106, + 650, + 496, + 663 + ], + "score": 1.0, + "content": "of UM that incorporates the previous component, as well as its variant, i.e., UMAP (Section 3.3).", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 30, + "bbox_fs": [ + 105, + 607, + 506, + 663 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 677, + 335, + 689 + ], + "lines": [ + { + "bbox": [ + 105, + 677, + 336, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 336, + 690 + ], + "score": 1.0, + "content": "3.1 ONCE-COVERED OOD DETECTION CAPABILITY", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 33 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 712 + ], + "score": 1.0, + "content": "First, we present the phenomenon of the inconsistent trend between a better OOD detection capability", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "and smaller training error during training. Empirically, as shown in Figure 1, we plot the OOD", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 721, + 504, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 504, + 733 + ], + "score": 1.0, + "content": "detection performance during the model training after multiple runs of the experiments. Across three", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 83, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 83, + 505, + 95 + ], + "score": 1.0, + "content": "different OOD testing datasets, we can observe the existence of a better detection capability using the", + "type": "text", + "cross_page": true + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 92, + 507, + 107 + ], + "spans": [ + { + "bbox": [ + 105, + 92, + 507, + 107 + ], + "score": 1.0, + "content": "index of FPR95 metric based on the Energy (Liu et al., 2020) or ODIN (Liang et al., 2018) score.", + "type": "text", + "cross_page": true + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 105, + 505, + 117 + ], + "spans": [ + { + "bbox": [ + 106, + 105, + 505, + 117 + ], + "score": 1.0, + "content": "The generality has also been verified using different learning schedules and model structures in our", + "type": "text", + "cross_page": true + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 115, + 506, + 128 + ], + "spans": [ + { + "bbox": [ + 105, + 115, + 506, + 128 + ], + "score": 1.0, + "content": "experimental sections. We further show the comparison of the ID/OOD distributions in Figures 1(b)", + "type": "text", + "cross_page": true + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 126, + 506, + 139 + ], + "spans": [ + { + "bbox": [ + 105, + 126, + 506, + 139 + ], + "score": 1.0, + "content": "and 1(c). To be specific, the statics of the two distributions indicate that the gap between the ID and", + "type": "text", + "cross_page": true + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "spans": [ + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "score": 1.0, + "content": "OOD data gets narrow as their overlap grows along with the training. After Epoch 60, although the", + "type": "text", + "cross_page": true + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 148, + 505, + 162 + ], + "spans": [ + { + "bbox": [ + 105, + 148, + 505, + 162 + ], + "score": 1.0, + "content": "model becomes more confident on ID data which satisfies a part of the calibration target (Hendrycks", + "type": "text", + "cross_page": true + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "score": 1.0, + "content": "et al., 2019a), its predictions on the OOD data also become more confident which is unexpected.", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 170, + 479, + 184 + ], + "spans": [ + { + "bbox": [ + 105, + 170, + 479, + 184 + ], + "score": 1.0, + "content": "Without seeing any auxiliary outliers, it motivates us to explore how the model achieves that.", + "type": "text", + "cross_page": true + } + ], + "index": 8 + } + ], + "index": 35, + "bbox_fs": [ + 105, + 698, + 505, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 181 + ], + "lines": [ + { + "bbox": [ + 106, + 83, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 83, + 505, + 95 + ], + "score": 1.0, + "content": "different OOD testing datasets, we can observe the existence of a better detection capability using the", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 92, + 507, + 107 + ], + "spans": [ + { + "bbox": [ + 105, + 92, + 507, + 107 + ], + "score": 1.0, + "content": "index of FPR95 metric based on the Energy (Liu et al., 2020) or ODIN (Liang et al., 2018) score.", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 105, + 505, + 117 + ], + "spans": [ + { + "bbox": [ + 106, + 105, + 505, + 117 + ], + "score": 1.0, + "content": "The generality has also been verified using different learning schedules and model structures in our", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 115, + 506, + 128 + ], + "spans": [ + { + "bbox": [ + 105, + 115, + 506, + 128 + ], + "score": 1.0, + "content": "experimental sections. We further show the comparison of the ID/OOD distributions in Figures 1(b)", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 126, + 506, + 139 + ], + "spans": [ + { + "bbox": [ + 105, + 126, + 506, + 139 + ], + "score": 1.0, + "content": "and 1(c). To be specific, the statics of the two distributions indicate that the gap between the ID and", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "spans": [ + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "score": 1.0, + "content": "OOD data gets narrow as their overlap grows along with the training. After Epoch 60, although the", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 148, + 505, + 162 + ], + "spans": [ + { + "bbox": [ + 105, + 148, + 505, + 162 + ], + "score": 1.0, + "content": "model becomes more confident on ID data which satisfies a part of the calibration target (Hendrycks", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "score": 1.0, + "content": "et al., 2019a), its predictions on the OOD data also become more confident which is unexpected.", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 170, + 479, + 184 + ], + "spans": [ + { + "bbox": [ + 105, + 170, + 479, + 184 + ], + "score": 1.0, + "content": "Without seeing any auxiliary outliers, it motivates us to explore how the model achieves that.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 107, + 187, + 505, + 297 + ], + "lines": [ + { + "bbox": [ + 105, + 186, + 507, + 200 + ], + "spans": [ + { + "bbox": [ + 105, + 186, + 507, + 200 + ], + "score": 1.0, + "content": "We take a closer look at the model behaviors in Figure 2, where we check its corresponding train-", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 198, + 506, + 210 + ], + "spans": [ + { + "bbox": [ + 105, + 198, + 506, + 210 + ], + "score": 1.0, + "content": "ing/testing loss and accuracy. We find that the training loss has reached a reasonably small value at", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 209, + 506, + 221 + ], + "spans": [ + { + "bbox": [ + 105, + 209, + 506, + 221 + ], + "score": 1.0, + "content": "Epoch 60 where its detection performance also achieves a satisfactory level. However, if we further", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 219, + 505, + 232 + ], + "spans": [ + { + "bbox": [ + 106, + 219, + 505, + 232 + ], + "score": 1.0, + "content": "minimize the training loss, the trend of the FPR95 curve shows almost opposite directions with both", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 231, + 506, + 243 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 506, + 243 + ], + "score": 1.0, + "content": "training and testing loss or accuracy (see Figures 1(a), 2(b), and 2(c)). Accordingly, we extract", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 241, + 506, + 254 + ], + "spans": [ + { + "bbox": [ + 106, + 241, + 506, + 254 + ], + "score": 1.0, + "content": "those samples that were learned by the model at this period. As shown in Figures 2(e) and 2(f),", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 252, + 505, + 266 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 505, + 266 + ], + "score": 1.0, + "content": "the misclassified samples learned after Epoch 60 present much atypical semantic features. As deep", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 264, + 505, + 277 + ], + "spans": [ + { + "bbox": [ + 106, + 264, + 505, + 277 + ], + "score": 1.0, + "content": "neural networks tend to first learn the data with typical features (Arpit et al., 2017), we attribute the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 274, + 506, + 287 + ], + "spans": [ + { + "bbox": [ + 105, + 274, + 506, + 287 + ], + "score": 1.0, + "content": "inconsistent trend to memorizing those atypical data at the later stage. In Appendix C, we provide a", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 285, + 506, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 285, + 506, + 298 + ], + "score": 1.0, + "content": "detailed discussion between it with the concept of conventional overfitting (Goodfellow et al., 2016).", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 13.5 + }, + { + "type": "title", + "bbox": [ + 108, + 311, + 337, + 321 + ], + "lines": [ + { + "bbox": [ + 105, + 310, + 339, + 323 + ], + "spans": [ + { + "bbox": [ + 105, + 310, + 339, + 323 + ], + "score": 1.0, + "content": "3.2 UNLEASHING THE INTRINSIC DETECTION POWER", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 107, + 331, + 505, + 408 + ], + "lines": [ + { + "bbox": [ + 105, + 330, + 506, + 344 + ], + "spans": [ + { + "bbox": [ + 105, + 330, + 506, + 344 + ], + "score": 1.0, + "content": "In general, the models that are developed for the original classification tasks are always seeking", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 342, + 507, + 355 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 507, + 355 + ], + "score": 1.0, + "content": "better performance (i.e., higher testing accuracy and lower training loss) in practice. However,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 353, + 505, + 365 + ], + "spans": [ + { + "bbox": [ + 106, + 353, + 505, + 365 + ], + "score": 1.0, + "content": "the inconsistent trend revealed before indicates that the intrinsic OOD detection capability maybe", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 364, + 505, + 377 + ], + "spans": [ + { + "bbox": [ + 105, + 364, + 505, + 377 + ], + "score": 1.0, + "content": "once-covered during the training. It gives us a chance to unleash the potential detection power only", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 375, + 506, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 506, + 387 + ], + "score": 1.0, + "content": "considering the ID data in training. To this end, we have two important issues that need to address:", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 385, + 505, + 398 + ], + "spans": [ + { + "bbox": [ + 106, + 385, + 505, + 398 + ], + "score": 1.0, + "content": "(1) the model that is well-trained on ID data may have already memorized some atypical samples", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 396, + 506, + 410 + ], + "spans": [ + { + "bbox": [ + 105, + 396, + 506, + 410 + ], + "score": 1.0, + "content": "which can not be figured out; (2) how to forget those atypical samples considering the given model?", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 23 + }, + { + "type": "text", + "bbox": [ + 107, + 420, + 505, + 530 + ], + "lines": [ + { + "bbox": [ + 105, + 420, + 506, + 432 + ], + "spans": [ + { + "bbox": [ + 105, + 420, + 506, + 432 + ], + "score": 1.0, + "content": "Mining the atypical samples with constructed discrepancy. According to Figures 2(a) and 2(b),", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 431, + 506, + 444 + ], + "spans": [ + { + "bbox": [ + 105, + 431, + 506, + 444 + ], + "score": 1.0, + "content": "both training accuracy and loss provide limited information that can differentiate the typical and", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 442, + 505, + 456 + ], + "spans": [ + { + "bbox": [ + 105, + 442, + 505, + 456 + ], + "score": 1.0, + "content": "atypical data. Inspired by the learning dynamics (Goodfellow et al., 2016; Arpit et al., 2017) of deep", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 453, + 505, + 465 + ], + "spans": [ + { + "bbox": [ + 105, + 453, + 505, + 465 + ], + "score": 1.0, + "content": "neural networks and the pathway conjecture (Barham et al., 2022) for inference, we try to manually", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 464, + 505, + 476 + ], + "spans": [ + { + "bbox": [ + 105, + 464, + 505, + 476 + ], + "score": 1.0, + "content": "construct the parameter discrepancy to mine the atypical samples from a well-trained model. To be", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 475, + 506, + 487 + ], + "spans": [ + { + "bbox": [ + 105, + 475, + 506, + 487 + ], + "score": 1.0, + "content": "specific, we introduce a novel layer-wise mask to achieve the goal. The masks are applied to all layers,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 486, + 506, + 498 + ], + "spans": [ + { + "bbox": [ + 105, + 486, + 506, + 498 + ], + "score": 1.0, + "content": "which is consistent with the mask generation in the conventional pruning pipeline (Han et al., 2016).", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 496, + 506, + 510 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 506, + 510 + ], + "score": 1.0, + "content": "In Figure 3(c), we provide empirical evidence to show that we can figure out the atypical samples via", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 508, + 504, + 520 + ], + "spans": [ + { + "bbox": [ + 106, + 508, + 504, + 520 + ], + "score": 1.0, + "content": "enlarging the mask rate. Utilizing the masked output for loss computation, the atypical samples can", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 518, + 475, + 531 + ], + "spans": [ + { + "bbox": [ + 105, + 518, + 475, + 531 + ], + "score": 1.0, + "content": "be better differentiated. We also provide more discussion about the intuition in Appendix B.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 107, + 542, + 505, + 619 + ], + "lines": [ + { + "bbox": [ + 105, + 542, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 505, + 555 + ], + "score": 1.0, + "content": "Forgetting the atypical samples with gradient ascent. As the training loss achieves zero at the", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 553, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 553, + 505, + 567 + ], + "score": 1.0, + "content": "final stage of the given model, we need extra optimization signals to forget those memorized atypical", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "score": 1.0, + "content": "samples. Considering the previous consistent trend before the potential optimal stage (e.g., before", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 575, + 506, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 575, + 506, + 588 + ], + "score": 1.0, + "content": "Epoch 60 in Figure 1(a)), the optimization signal also needs to control the model update not to be too", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 585, + 506, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 506, + 600 + ], + "score": 1.0, + "content": "greedy to drop the discriminative features for OOD detection. Starting with the given model, we can", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 597, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 106, + 597, + 505, + 610 + ], + "score": 1.0, + "content": "employ the gradient ascent (Sorg et al., 2010; Ishida et al., 2020) to forget the targeted samples, while", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 608, + 494, + 622 + ], + "spans": [ + { + "bbox": [ + 105, + 608, + 494, + 622 + ], + "score": 1.0, + "content": "the tuning phase should also prevent further updates if the model can achieve the expected stage.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 40 + }, + { + "type": "title", + "bbox": [ + 108, + 633, + 232, + 644 + ], + "lines": [ + { + "bbox": [ + 106, + 632, + 233, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 632, + 233, + 645 + ], + "score": 1.0, + "content": "3.3 METHOD REALIZATION", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 44 + }, + { + "type": "text", + "bbox": [ + 108, + 653, + 505, + 686 + ], + "lines": [ + { + "bbox": [ + 105, + 653, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 505, + 666 + ], + "score": 1.0, + "content": "Based on previous insights, we present our overall framework as well as the learning objective of the", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 664, + 505, + 677 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 505, + 677 + ], + "score": 1.0, + "content": "proposed Unleashing Mask for OOD detection. Lastly, we discuss its compatibility with either the", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 675, + 494, + 688 + ], + "spans": [ + { + "bbox": [ + 105, + 675, + 494, + 688 + ], + "score": 1.0, + "content": "fundamental scoring functions or the outlier exposure approaches utilizing auxiliary outlier data.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 46 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "Framework. As illustrated in Figure 3(a), our framework consists of two critical components for", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "score": 1.0, + "content": "uncovering the intrinsic OOD detection capability: (1) the initialized mask with a specific pruning", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "score": 1.0, + "content": "rate for constructing the output discrepancy with the original model; (2) the fine-tuning procedure for", + "type": "text" + } + ], + "index": 50 + } + ], + "index": 49 + } + ], + "page_idx": 4, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 759 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 309, + 763 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 309, + 763 + ], + "score": 1.0, + "content": "5", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 181 + ], + "lines": [], + "index": 4, + "bbox_fs": [ + 105, + 83, + 507, + 184 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 187, + 505, + 297 + ], + "lines": [ + { + "bbox": [ + 105, + 186, + 507, + 200 + ], + "spans": [ + { + "bbox": [ + 105, + 186, + 507, + 200 + ], + "score": 1.0, + "content": "We take a closer look at the model behaviors in Figure 2, where we check its corresponding train-", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 198, + 506, + 210 + ], + "spans": [ + { + "bbox": [ + 105, + 198, + 506, + 210 + ], + "score": 1.0, + "content": "ing/testing loss and accuracy. We find that the training loss has reached a reasonably small value at", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 209, + 506, + 221 + ], + "spans": [ + { + "bbox": [ + 105, + 209, + 506, + 221 + ], + "score": 1.0, + "content": "Epoch 60 where its detection performance also achieves a satisfactory level. However, if we further", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 219, + 505, + 232 + ], + "spans": [ + { + "bbox": [ + 106, + 219, + 505, + 232 + ], + "score": 1.0, + "content": "minimize the training loss, the trend of the FPR95 curve shows almost opposite directions with both", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 231, + 506, + 243 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 506, + 243 + ], + "score": 1.0, + "content": "training and testing loss or accuracy (see Figures 1(a), 2(b), and 2(c)). Accordingly, we extract", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 241, + 506, + 254 + ], + "spans": [ + { + "bbox": [ + 106, + 241, + 506, + 254 + ], + "score": 1.0, + "content": "those samples that were learned by the model at this period. As shown in Figures 2(e) and 2(f),", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 252, + 505, + 266 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 505, + 266 + ], + "score": 1.0, + "content": "the misclassified samples learned after Epoch 60 present much atypical semantic features. As deep", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 264, + 505, + 277 + ], + "spans": [ + { + "bbox": [ + 106, + 264, + 505, + 277 + ], + "score": 1.0, + "content": "neural networks tend to first learn the data with typical features (Arpit et al., 2017), we attribute the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 274, + 506, + 287 + ], + "spans": [ + { + "bbox": [ + 105, + 274, + 506, + 287 + ], + "score": 1.0, + "content": "inconsistent trend to memorizing those atypical data at the later stage. In Appendix C, we provide a", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 285, + 506, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 285, + 506, + 298 + ], + "score": 1.0, + "content": "detailed discussion between it with the concept of conventional overfitting (Goodfellow et al., 2016).", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 13.5, + "bbox_fs": [ + 105, + 186, + 507, + 298 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 311, + 337, + 321 + ], + "lines": [ + { + "bbox": [ + 105, + 310, + 339, + 323 + ], + "spans": [ + { + "bbox": [ + 105, + 310, + 339, + 323 + ], + "score": 1.0, + "content": "3.2 UNLEASHING THE INTRINSIC DETECTION POWER", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 107, + 331, + 505, + 408 + ], + "lines": [ + { + "bbox": [ + 105, + 330, + 506, + 344 + ], + "spans": [ + { + "bbox": [ + 105, + 330, + 506, + 344 + ], + "score": 1.0, + "content": "In general, the models that are developed for the original classification tasks are always seeking", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 342, + 507, + 355 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 507, + 355 + ], + "score": 1.0, + "content": "better performance (i.e., higher testing accuracy and lower training loss) in practice. However,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 353, + 505, + 365 + ], + "spans": [ + { + "bbox": [ + 106, + 353, + 505, + 365 + ], + "score": 1.0, + "content": "the inconsistent trend revealed before indicates that the intrinsic OOD detection capability maybe", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 364, + 505, + 377 + ], + "spans": [ + { + "bbox": [ + 105, + 364, + 505, + 377 + ], + "score": 1.0, + "content": "once-covered during the training. It gives us a chance to unleash the potential detection power only", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 375, + 506, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 506, + 387 + ], + "score": 1.0, + "content": "considering the ID data in training. To this end, we have two important issues that need to address:", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 385, + 505, + 398 + ], + "spans": [ + { + "bbox": [ + 106, + 385, + 505, + 398 + ], + "score": 1.0, + "content": "(1) the model that is well-trained on ID data may have already memorized some atypical samples", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 396, + 506, + 410 + ], + "spans": [ + { + "bbox": [ + 105, + 396, + 506, + 410 + ], + "score": 1.0, + "content": "which can not be figured out; (2) how to forget those atypical samples considering the given model?", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 23, + "bbox_fs": [ + 105, + 330, + 507, + 410 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 420, + 505, + 530 + ], + "lines": [ + { + "bbox": [ + 105, + 420, + 506, + 432 + ], + "spans": [ + { + "bbox": [ + 105, + 420, + 506, + 432 + ], + "score": 1.0, + "content": "Mining the atypical samples with constructed discrepancy. According to Figures 2(a) and 2(b),", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 431, + 506, + 444 + ], + "spans": [ + { + "bbox": [ + 105, + 431, + 506, + 444 + ], + "score": 1.0, + "content": "both training accuracy and loss provide limited information that can differentiate the typical and", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 442, + 505, + 456 + ], + "spans": [ + { + "bbox": [ + 105, + 442, + 505, + 456 + ], + "score": 1.0, + "content": "atypical data. Inspired by the learning dynamics (Goodfellow et al., 2016; Arpit et al., 2017) of deep", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 453, + 505, + 465 + ], + "spans": [ + { + "bbox": [ + 105, + 453, + 505, + 465 + ], + "score": 1.0, + "content": "neural networks and the pathway conjecture (Barham et al., 2022) for inference, we try to manually", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 464, + 505, + 476 + ], + "spans": [ + { + "bbox": [ + 105, + 464, + 505, + 476 + ], + "score": 1.0, + "content": "construct the parameter discrepancy to mine the atypical samples from a well-trained model. To be", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 475, + 506, + 487 + ], + "spans": [ + { + "bbox": [ + 105, + 475, + 506, + 487 + ], + "score": 1.0, + "content": "specific, we introduce a novel layer-wise mask to achieve the goal. The masks are applied to all layers,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 486, + 506, + 498 + ], + "spans": [ + { + "bbox": [ + 105, + 486, + 506, + 498 + ], + "score": 1.0, + "content": "which is consistent with the mask generation in the conventional pruning pipeline (Han et al., 2016).", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 496, + 506, + 510 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 506, + 510 + ], + "score": 1.0, + "content": "In Figure 3(c), we provide empirical evidence to show that we can figure out the atypical samples via", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 508, + 504, + 520 + ], + "spans": [ + { + "bbox": [ + 106, + 508, + 504, + 520 + ], + "score": 1.0, + "content": "enlarging the mask rate. Utilizing the masked output for loss computation, the atypical samples can", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 518, + 475, + 531 + ], + "spans": [ + { + "bbox": [ + 105, + 518, + 475, + 531 + ], + "score": 1.0, + "content": "be better differentiated. We also provide more discussion about the intuition in Appendix B.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 31.5, + "bbox_fs": [ + 105, + 420, + 506, + 531 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 542, + 505, + 619 + ], + "lines": [ + { + "bbox": [ + 105, + 542, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 505, + 555 + ], + "score": 1.0, + "content": "Forgetting the atypical samples with gradient ascent. As the training loss achieves zero at the", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 553, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 553, + 505, + 567 + ], + "score": 1.0, + "content": "final stage of the given model, we need extra optimization signals to forget those memorized atypical", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "score": 1.0, + "content": "samples. Considering the previous consistent trend before the potential optimal stage (e.g., before", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 575, + 506, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 575, + 506, + 588 + ], + "score": 1.0, + "content": "Epoch 60 in Figure 1(a)), the optimization signal also needs to control the model update not to be too", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 585, + 506, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 506, + 600 + ], + "score": 1.0, + "content": "greedy to drop the discriminative features for OOD detection. Starting with the given model, we can", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 597, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 106, + 597, + 505, + 610 + ], + "score": 1.0, + "content": "employ the gradient ascent (Sorg et al., 2010; Ishida et al., 2020) to forget the targeted samples, while", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 608, + 494, + 622 + ], + "spans": [ + { + "bbox": [ + 105, + 608, + 494, + 622 + ], + "score": 1.0, + "content": "the tuning phase should also prevent further updates if the model can achieve the expected stage.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 40, + "bbox_fs": [ + 105, + 542, + 506, + 622 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 633, + 232, + 644 + ], + "lines": [ + { + "bbox": [ + 106, + 632, + 233, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 632, + 233, + 645 + ], + "score": 1.0, + "content": "3.3 METHOD REALIZATION", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 44 + }, + { + "type": "text", + "bbox": [ + 108, + 653, + 505, + 686 + ], + "lines": [ + { + "bbox": [ + 105, + 653, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 505, + 666 + ], + "score": 1.0, + "content": "Based on previous insights, we present our overall framework as well as the learning objective of the", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 664, + 505, + 677 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 505, + 677 + ], + "score": 1.0, + "content": "proposed Unleashing Mask for OOD detection. Lastly, we discuss its compatibility with either the", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 675, + 494, + 688 + ], + "spans": [ + { + "bbox": [ + 105, + 675, + 494, + 688 + ], + "score": 1.0, + "content": "fundamental scoring functions or the outlier exposure approaches utilizing auxiliary outlier data.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 46, + "bbox_fs": [ + 105, + 653, + 505, + 688 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "Framework. As illustrated in Figure 3(a), our framework consists of two critical components for", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "score": 1.0, + "content": "uncovering the intrinsic OOD detection capability: (1) the initialized mask with a specific pruning", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "score": 1.0, + "content": "rate for constructing the output discrepancy with the original model; (2) the fine-tuning procedure for", + "type": "text" + } + ], + "index": 50 + } + ], + "index": 49, + "bbox_fs": [ + 105, + 698, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 118, + 86, + 500, + 200 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 118, + 86, + 500, + 200 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 118, + 86, + 500, + 200 + ], + "spans": [ + { + "bbox": [ + 118, + 86, + 500, + 200 + ], + "score": 0.955, + "type": "image", + "image_path": "449c5426c25f0d6c449c60c781928f496a0fbe06ead281d568841a5ca4b5a2d5.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 118, + 86, + 500, + 124.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 118, + 124.0, + 500, + 162.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 118, + 162.0, + 500, + 200.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 209, + 505, + 287 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 209, + 505, + 222 + ], + "spans": [ + { + "bbox": [ + 106, + 209, + 505, + 222 + ], + "score": 1.0, + "content": "Figure 3: (a) a brief illustration of the proposed Unleashing Mask (UM); (b) the mask rate w.r.t. loss", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 220, + 505, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 220, + 505, + 233 + ], + "score": 1.0, + "content": "value using the masked outputs; (c) examples of misclassified samples after masking the original", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 231, + 505, + 244 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 505, + 244 + ], + "score": 1.0, + "content": "well-trained model. As for our framework, given a well-trained model, we initialize an extra mask", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 242, + 506, + 255 + ], + "spans": [ + { + "bbox": [ + 105, + 242, + 506, + 255 + ], + "score": 1.0, + "content": "for mining the atypical samples that are sensitive to the changes in model parameters. Then we", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 253, + 505, + 266 + ], + "spans": [ + { + "bbox": [ + 105, + 253, + 505, + 266 + ], + "score": 1.0, + "content": "fine-tune the original model or adopt pruning with the estimated forgetting threshold, i.e., the loss", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 265, + 505, + 276 + ], + "spans": [ + { + "bbox": [ + 106, + 265, + 505, + 276 + ], + "score": 1.0, + "content": "value estimated by the UM. The final model can serve as the base of various score functions to utilize", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 274, + 505, + 288 + ], + "spans": [ + { + "bbox": [ + 105, + 274, + 505, + 288 + ], + "score": 1.0, + "content": "the discriminative features and also as the new initialization of fine-tuning with the auxiliary outliers.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 6 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 106, + 294, + 504, + 318 + ], + "lines": [ + { + "bbox": [ + 106, + 294, + 505, + 307 + ], + "spans": [ + { + "bbox": [ + 106, + 294, + 505, + 307 + ], + "score": 1.0, + "content": "alleviating the memorization of atypical samples. The overall workflow starts with obtaining the loss", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 306, + 496, + 319 + ], + "spans": [ + { + "bbox": [ + 106, + 306, + 496, + 319 + ], + "score": 1.0, + "content": "value of misclassifying those atypical samples and then conducts tuning with the model to forget.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5 + }, + { + "type": "text", + "bbox": [ + 107, + 329, + 484, + 340 + ], + "lines": [ + { + "bbox": [ + 106, + 326, + 485, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 326, + 485, + 343 + ], + "score": 1.0, + "content": "Objective for forgetting. Based on our framework, we introduce the forgetting objective as,", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 12 + }, + { + "type": "interline_equation", + "bbox": [ + 172, + 343, + 437, + 365 + ], + "lines": [ + { + "bbox": [ + 172, + 343, + 437, + 365 + ], + "spans": [ + { + "bbox": [ + 172, + 343, + 437, + 365 + ], + "score": 0.92, + "content": "\\operatorname* { m i n } \\mathcal { L } _ { \\mathrm { U M } } = \\operatorname* { m i n } _ { m _ { \\delta } \\in [ 0 , 1 ] ^ { n } } | \\ell _ { \\mathrm { C E } } ( f ) - \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) | + \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) ,", + "type": "interline_equation", + "image_path": "7bcd5f75397cf8a7cb7797e49cb807f703141cc6b6347834dccb36a08a06d097.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 172, + 343, + 437, + 365 + ], + "spans": [], + "index": 13 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 370, + 505, + 460 + ], + "lines": [ + { + "bbox": [ + 105, + 369, + 506, + 383 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 133, + 383 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 134, + 373, + 148, + 382 + ], + "score": 0.85, + "content": "m _ { \\delta }", + "type": "inline_equation" + }, + { + "bbox": [ + 149, + 370, + 375, + 383 + ], + "score": 1.0, + "content": "is our proposed layer-wise mask with the pruning rate", + "type": "text" + }, + { + "bbox": [ + 375, + 372, + 381, + 381 + ], + "score": 0.58, + "content": "\\delta", + "type": "inline_equation" + }, + { + "bbox": [ + 381, + 370, + 385, + 383 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 385, + 371, + 400, + 383 + ], + "score": 0.69, + "content": "\\ell _ { \\mathrm { C E } }", + "type": "inline_equation" + }, + { + "bbox": [ + 401, + 370, + 463, + 383 + ], + "score": 1.0, + "content": "is the CE loss,", + "type": "text" + }, + { + "bbox": [ + 464, + 369, + 479, + 383 + ], + "score": 0.9, + "content": "\\widehat { \\ell } _ { \\mathrm { C E } }", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 370, + 506, + 383 + ], + "score": 1.0, + "content": "is the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 382, + 504, + 394 + ], + "spans": [ + { + "bbox": [ + 106, + 382, + 287, + 393 + ], + "score": 1.0, + "content": "averaged CE loss over the ID training data,", + "type": "text" + }, + { + "bbox": [ + 288, + 382, + 302, + 394 + ], + "score": 0.84, + "content": "| \\cdot |", + "type": "inline_equation" + }, + { + "bbox": [ + 302, + 382, + 504, + 393 + ], + "score": 1.0, + "content": "indicates the computation for absolute value and", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 107, + 393, + 506, + 406 + ], + "spans": [ + { + "bbox": [ + 107, + 394, + 144, + 405 + ], + "score": 0.91, + "content": "m _ { \\delta } \\odot f ^ { * }", + "type": "inline_equation" + }, + { + "bbox": [ + 144, + 393, + 506, + 406 + ], + "score": 1.0, + "content": "denotes the masked output of the fixed pretrained model that is used to estimate the loss", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 404, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 106, + 404, + 505, + 416 + ], + "score": 1.0, + "content": "constraint for the learning objective of forgetting, which would be a constant value during the", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 416, + 505, + 426 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 505, + 426 + ], + "score": 1.0, + "content": "whole finetuning process. Concretely, the well-trained model will start to optimize itself again if it", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 426, + 505, + 438 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 505, + 438 + ], + "score": 1.0, + "content": "memorized the atypical samples and achieved almost zero loss value. We provide a positive gradient", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 437, + 506, + 449 + ], + "spans": [ + { + "bbox": [ + 105, + 437, + 506, + 449 + ], + "score": 1.0, + "content": "signal when the current loss value is lower than the estimated one and vice versa. The model is", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 448, + 441, + 460 + ], + "spans": [ + { + "bbox": [ + 105, + 448, + 441, + 460 + ], + "score": 1.0, + "content": "expected to finally stabilize around the stage that can forget those atypical samples.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 17.5 + }, + { + "type": "text", + "bbox": [ + 106, + 471, + 505, + 516 + ], + "lines": [ + { + "bbox": [ + 105, + 470, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 470, + 506, + 484 + ], + "score": 1.0, + "content": "Unleashing Mask Adopts Pruning (UMAP). Considering the potential negative effect on the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 482, + 506, + 494 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 506, + 494 + ], + "score": 1.0, + "content": "original task performance when conducting tuning for forgetting, we further propose a variant of UM", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 492, + 506, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 465, + 506 + ], + "score": 1.0, + "content": "Adopts Pruning, i.e., UMAP, to conduct tuning based on the masked output (e.g., replace", + "type": "text" + }, + { + "bbox": [ + 465, + 495, + 493, + 505 + ], + "score": 0.93, + "content": "\\ell _ { \\mathrm { C E } } ( f )", + "type": "inline_equation" + }, + { + "bbox": [ + 493, + 492, + 506, + 506 + ], + "score": 1.0, + "content": "to", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 107, + 504, + 452, + 517 + ], + "spans": [ + { + "bbox": [ + 107, + 505, + 165, + 516 + ], + "score": 0.91, + "content": "\\ell _ { \\mathrm { C E } } ( \\hat { m } _ { p } \\odot f ^ { * } )", + "type": "inline_equation" + }, + { + "bbox": [ + 165, + 504, + 300, + 517 + ], + "score": 1.0, + "content": "in Eq 3) using the different mask", + "type": "text" + }, + { + "bbox": [ + 300, + 505, + 314, + 516 + ], + "score": 0.89, + "content": "\\hat { m } _ { p }", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 504, + 398, + 517 + ], + "score": 1.0, + "content": "with its pruning rate", + "type": "text" + }, + { + "bbox": [ + 399, + 507, + 405, + 516 + ], + "score": 0.77, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 405, + 504, + 452, + 517 + ], + "score": 1.0, + "content": "as follows,", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 23.5 + }, + { + "type": "interline_equation", + "bbox": [ + 134, + 519, + 476, + 541 + ], + "lines": [ + { + "bbox": [ + 134, + 519, + 476, + 541 + ], + "spans": [ + { + "bbox": [ + 134, + 519, + 476, + 541 + ], + "score": 0.93, + "content": "\\operatorname* { m i n } \\mathcal { L } _ { \\mathrm { U M A P } } = \\operatorname* { m i n } _ { \\hat { m } _ { p } \\in [ 0 , 1 ] ^ { n } , m _ { \\delta } \\in [ 0 , 1 ] ^ { n } } | \\ell _ { \\mathrm { C E } } ( \\hat { m } _ { p } \\odot f ^ { * } ) - \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) | + \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) ,", + "type": "interline_equation", + "image_path": "2377b492a02f03724b79b5e55c19c0be886d00f7a5d0399b6677c53238470b9d.jpg" + } + ] + } + ], + "index": 26, + "virtual_lines": [ + { + "bbox": [ + 134, + 519, + 476, + 541 + ], + "spans": [], + "index": 26 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 543, + 505, + 610 + ], + "lines": [ + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "score": 1.0, + "content": "Different from the objective of UM (i.e., Eq 3) that minimizes the loss value over the model parameter,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 554, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 106, + 554, + 505, + 567 + ], + "score": 1.0, + "content": "the objective of UMAP minimizes the loss over the mask to achieve the target of forgetting those", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 565, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 505, + 577 + ], + "score": 1.0, + "content": "atypical samples. UMAP provides an extra mask to restore the detection capacity but doesn’t affect", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 577, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 106, + 577, + 505, + 588 + ], + "score": 1.0, + "content": "the model parameter for the inference on original tasks, indicating that UMAP is a more practical", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 587, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 587, + 505, + 600 + ], + "score": 1.0, + "content": "choice in real-world applications (as empirically verified in our experiments like Table 1). We", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 598, + 490, + 611 + ], + "spans": [ + { + "bbox": [ + 106, + 598, + 490, + 611 + ], + "score": 1.0, + "content": "summarize the algorithms of UM (in Algorithm 1) and UMAP (in Algorithm 2) in Appendix D.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 29.5 + }, + { + "type": "text", + "bbox": [ + 106, + 621, + 506, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 621, + 506, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 621, + 506, + 634 + ], + "score": 1.0, + "content": "Compatible to other methods. As we explore the original OOD detection capability of the well-", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 632, + 505, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 505, + 646 + ], + "score": 1.0, + "content": "trained model, it is orthogonal and compatible with those promising methods that equip the given", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 643, + 507, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 507, + 657 + ], + "score": 1.0, + "content": "model with better detection ability. To be specific, through our proposed methods, we reveal the once-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 104, + 654, + 506, + 669 + ], + "spans": [ + { + "bbox": [ + 104, + 654, + 506, + 669 + ], + "score": 1.0, + "content": "covered OOD detection capability via tuning the original model towards its intermediate training stage.", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 664, + 505, + 680 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 505, + 680 + ], + "score": 1.0, + "content": "The discriminative feature learned at that stage can be utilized by different scoring functions (Huang", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "et al., 2021; Sun & Li, 2022; Wei et al., 2022), like ODIN (Liang et al., 2018) adopted in Figure 2(c).", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 686, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 686, + 506, + 702 + ], + "score": 1.0, + "content": "For those methods (Hendrycks et al., 2019a; Liu et al., 2020; Ming et al., 2022) utilizing the auxiliary", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 699, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 712 + ], + "score": 1.0, + "content": "outliers to regularize the model, our fine-tuned model obtained by UM and UMAP can also serve as", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 710, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 506, + 722 + ], + "score": 1.0, + "content": "their starting point or adjustment. As our strategy does not require any auxiliary outlier data to be", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 721, + 486, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 486, + 732 + ], + "score": 1.0, + "content": "involved in training, adjusting the model using ID data during its developing phase is practical.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 37.5 + } + ], + "page_idx": 5, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 106, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 310, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 310, + 762 + ], + "score": 1.0, + "content": "6", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 118, + 86, + 500, + 200 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 118, + 86, + 500, + 200 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 118, + 86, + 500, + 200 + ], + "spans": [ + { + "bbox": [ + 118, + 86, + 500, + 200 + ], + "score": 0.955, + "type": "image", + "image_path": "449c5426c25f0d6c449c60c781928f496a0fbe06ead281d568841a5ca4b5a2d5.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 118, + 86, + 500, + 124.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 118, + 124.0, + 500, + 162.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 118, + 162.0, + 500, + 200.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 209, + 505, + 287 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 209, + 505, + 222 + ], + "spans": [ + { + "bbox": [ + 106, + 209, + 505, + 222 + ], + "score": 1.0, + "content": "Figure 3: (a) a brief illustration of the proposed Unleashing Mask (UM); (b) the mask rate w.r.t. loss", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 220, + 505, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 220, + 505, + 233 + ], + "score": 1.0, + "content": "value using the masked outputs; (c) examples of misclassified samples after masking the original", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 231, + 505, + 244 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 505, + 244 + ], + "score": 1.0, + "content": "well-trained model. As for our framework, given a well-trained model, we initialize an extra mask", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 242, + 506, + 255 + ], + "spans": [ + { + "bbox": [ + 105, + 242, + 506, + 255 + ], + "score": 1.0, + "content": "for mining the atypical samples that are sensitive to the changes in model parameters. Then we", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 253, + 505, + 266 + ], + "spans": [ + { + "bbox": [ + 105, + 253, + 505, + 266 + ], + "score": 1.0, + "content": "fine-tune the original model or adopt pruning with the estimated forgetting threshold, i.e., the loss", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 265, + 505, + 276 + ], + "spans": [ + { + "bbox": [ + 106, + 265, + 505, + 276 + ], + "score": 1.0, + "content": "value estimated by the UM. The final model can serve as the base of various score functions to utilize", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 274, + 505, + 288 + ], + "spans": [ + { + "bbox": [ + 105, + 274, + 505, + 288 + ], + "score": 1.0, + "content": "the discriminative features and also as the new initialization of fine-tuning with the auxiliary outliers.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 6 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 106, + 294, + 504, + 318 + ], + "lines": [ + { + "bbox": [ + 106, + 294, + 505, + 307 + ], + "spans": [ + { + "bbox": [ + 106, + 294, + 505, + 307 + ], + "score": 1.0, + "content": "alleviating the memorization of atypical samples. The overall workflow starts with obtaining the loss", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 306, + 496, + 319 + ], + "spans": [ + { + "bbox": [ + 106, + 306, + 496, + 319 + ], + "score": 1.0, + "content": "value of misclassifying those atypical samples and then conducts tuning with the model to forget.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5, + "bbox_fs": [ + 106, + 294, + 505, + 319 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 329, + 484, + 340 + ], + "lines": [ + { + "bbox": [ + 106, + 326, + 485, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 326, + 485, + 343 + ], + "score": 1.0, + "content": "Objective for forgetting. Based on our framework, we introduce the forgetting objective as,", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 12, + "bbox_fs": [ + 106, + 326, + 485, + 343 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 172, + 343, + 437, + 365 + ], + "lines": [ + { + "bbox": [ + 172, + 343, + 437, + 365 + ], + "spans": [ + { + "bbox": [ + 172, + 343, + 437, + 365 + ], + "score": 0.92, + "content": "\\operatorname* { m i n } \\mathcal { L } _ { \\mathrm { U M } } = \\operatorname* { m i n } _ { m _ { \\delta } \\in [ 0 , 1 ] ^ { n } } | \\ell _ { \\mathrm { C E } } ( f ) - \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) | + \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) ,", + "type": "interline_equation", + "image_path": "7bcd5f75397cf8a7cb7797e49cb807f703141cc6b6347834dccb36a08a06d097.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 172, + 343, + 437, + 365 + ], + "spans": [], + "index": 13 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 370, + 505, + 460 + ], + "lines": [ + { + "bbox": [ + 105, + 369, + 506, + 383 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 133, + 383 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 134, + 373, + 148, + 382 + ], + "score": 0.85, + "content": "m _ { \\delta }", + "type": "inline_equation" + }, + { + "bbox": [ + 149, + 370, + 375, + 383 + ], + "score": 1.0, + "content": "is our proposed layer-wise mask with the pruning rate", + "type": "text" + }, + { + "bbox": [ + 375, + 372, + 381, + 381 + ], + "score": 0.58, + "content": "\\delta", + "type": "inline_equation" + }, + { + "bbox": [ + 381, + 370, + 385, + 383 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 385, + 371, + 400, + 383 + ], + "score": 0.69, + "content": "\\ell _ { \\mathrm { C E } }", + "type": "inline_equation" + }, + { + "bbox": [ + 401, + 370, + 463, + 383 + ], + "score": 1.0, + "content": "is the CE loss,", + "type": "text" + }, + { + "bbox": [ + 464, + 369, + 479, + 383 + ], + "score": 0.9, + "content": "\\widehat { \\ell } _ { \\mathrm { C E } }", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 370, + 506, + 383 + ], + "score": 1.0, + "content": "is the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 382, + 504, + 394 + ], + "spans": [ + { + "bbox": [ + 106, + 382, + 287, + 393 + ], + "score": 1.0, + "content": "averaged CE loss over the ID training data,", + "type": "text" + }, + { + "bbox": [ + 288, + 382, + 302, + 394 + ], + "score": 0.84, + "content": "| \\cdot |", + "type": "inline_equation" + }, + { + "bbox": [ + 302, + 382, + 504, + 393 + ], + "score": 1.0, + "content": "indicates the computation for absolute value and", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 107, + 393, + 506, + 406 + ], + "spans": [ + { + "bbox": [ + 107, + 394, + 144, + 405 + ], + "score": 0.91, + "content": "m _ { \\delta } \\odot f ^ { * }", + "type": "inline_equation" + }, + { + "bbox": [ + 144, + 393, + 506, + 406 + ], + "score": 1.0, + "content": "denotes the masked output of the fixed pretrained model that is used to estimate the loss", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 404, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 106, + 404, + 505, + 416 + ], + "score": 1.0, + "content": "constraint for the learning objective of forgetting, which would be a constant value during the", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 416, + 505, + 426 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 505, + 426 + ], + "score": 1.0, + "content": "whole finetuning process. Concretely, the well-trained model will start to optimize itself again if it", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 426, + 505, + 438 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 505, + 438 + ], + "score": 1.0, + "content": "memorized the atypical samples and achieved almost zero loss value. We provide a positive gradient", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 437, + 506, + 449 + ], + "spans": [ + { + "bbox": [ + 105, + 437, + 506, + 449 + ], + "score": 1.0, + "content": "signal when the current loss value is lower than the estimated one and vice versa. The model is", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 448, + 441, + 460 + ], + "spans": [ + { + "bbox": [ + 105, + 448, + 441, + 460 + ], + "score": 1.0, + "content": "expected to finally stabilize around the stage that can forget those atypical samples.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 17.5, + "bbox_fs": [ + 105, + 369, + 506, + 460 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 471, + 505, + 516 + ], + "lines": [ + { + "bbox": [ + 105, + 470, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 470, + 506, + 484 + ], + "score": 1.0, + "content": "Unleashing Mask Adopts Pruning (UMAP). Considering the potential negative effect on the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 482, + 506, + 494 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 506, + 494 + ], + "score": 1.0, + "content": "original task performance when conducting tuning for forgetting, we further propose a variant of UM", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 492, + 506, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 465, + 506 + ], + "score": 1.0, + "content": "Adopts Pruning, i.e., UMAP, to conduct tuning based on the masked output (e.g., replace", + "type": "text" + }, + { + "bbox": [ + 465, + 495, + 493, + 505 + ], + "score": 0.93, + "content": "\\ell _ { \\mathrm { C E } } ( f )", + "type": "inline_equation" + }, + { + "bbox": [ + 493, + 492, + 506, + 506 + ], + "score": 1.0, + "content": "to", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 107, + 504, + 452, + 517 + ], + "spans": [ + { + "bbox": [ + 107, + 505, + 165, + 516 + ], + "score": 0.91, + "content": "\\ell _ { \\mathrm { C E } } ( \\hat { m } _ { p } \\odot f ^ { * } )", + "type": "inline_equation" + }, + { + "bbox": [ + 165, + 504, + 300, + 517 + ], + "score": 1.0, + "content": "in Eq 3) using the different mask", + "type": "text" + }, + { + "bbox": [ + 300, + 505, + 314, + 516 + ], + "score": 0.89, + "content": "\\hat { m } _ { p }", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 504, + 398, + 517 + ], + "score": 1.0, + "content": "with its pruning rate", + "type": "text" + }, + { + "bbox": [ + 399, + 507, + 405, + 516 + ], + "score": 0.77, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 405, + 504, + 452, + 517 + ], + "score": 1.0, + "content": "as follows,", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 23.5, + "bbox_fs": [ + 105, + 470, + 506, + 517 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 134, + 519, + 476, + 541 + ], + "lines": [ + { + "bbox": [ + 134, + 519, + 476, + 541 + ], + "spans": [ + { + "bbox": [ + 134, + 519, + 476, + 541 + ], + "score": 0.93, + "content": "\\operatorname* { m i n } \\mathcal { L } _ { \\mathrm { U M A P } } = \\operatorname* { m i n } _ { \\hat { m } _ { p } \\in [ 0 , 1 ] ^ { n } , m _ { \\delta } \\in [ 0 , 1 ] ^ { n } } | \\ell _ { \\mathrm { C E } } ( \\hat { m } _ { p } \\odot f ^ { * } ) - \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) | + \\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } ) ,", + "type": "interline_equation", + "image_path": "2377b492a02f03724b79b5e55c19c0be886d00f7a5d0399b6677c53238470b9d.jpg" + } + ] + } + ], + "index": 26, + "virtual_lines": [ + { + "bbox": [ + 134, + 519, + 476, + 541 + ], + "spans": [], + "index": 26 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 543, + 505, + 610 + ], + "lines": [ + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "score": 1.0, + "content": "Different from the objective of UM (i.e., Eq 3) that minimizes the loss value over the model parameter,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 554, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 106, + 554, + 505, + 567 + ], + "score": 1.0, + "content": "the objective of UMAP minimizes the loss over the mask to achieve the target of forgetting those", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 565, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 505, + 577 + ], + "score": 1.0, + "content": "atypical samples. UMAP provides an extra mask to restore the detection capacity but doesn’t affect", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 577, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 106, + 577, + 505, + 588 + ], + "score": 1.0, + "content": "the model parameter for the inference on original tasks, indicating that UMAP is a more practical", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 587, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 587, + 505, + 600 + ], + "score": 1.0, + "content": "choice in real-world applications (as empirically verified in our experiments like Table 1). We", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 598, + 490, + 611 + ], + "spans": [ + { + "bbox": [ + 106, + 598, + 490, + 611 + ], + "score": 1.0, + "content": "summarize the algorithms of UM (in Algorithm 1) and UMAP (in Algorithm 2) in Appendix D.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 29.5, + "bbox_fs": [ + 105, + 542, + 506, + 611 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 621, + 506, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 621, + 506, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 621, + 506, + 634 + ], + "score": 1.0, + "content": "Compatible to other methods. As we explore the original OOD detection capability of the well-", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 632, + 505, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 505, + 646 + ], + "score": 1.0, + "content": "trained model, it is orthogonal and compatible with those promising methods that equip the given", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 643, + 507, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 507, + 657 + ], + "score": 1.0, + "content": "model with better detection ability. To be specific, through our proposed methods, we reveal the once-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 104, + 654, + 506, + 669 + ], + "spans": [ + { + "bbox": [ + 104, + 654, + 506, + 669 + ], + "score": 1.0, + "content": "covered OOD detection capability via tuning the original model towards its intermediate training stage.", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 664, + 505, + 680 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 505, + 680 + ], + "score": 1.0, + "content": "The discriminative feature learned at that stage can be utilized by different scoring functions (Huang", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "et al., 2021; Sun & Li, 2022; Wei et al., 2022), like ODIN (Liang et al., 2018) adopted in Figure 2(c).", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 686, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 686, + 506, + 702 + ], + "score": 1.0, + "content": "For those methods (Hendrycks et al., 2019a; Liu et al., 2020; Ming et al., 2022) utilizing the auxiliary", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 699, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 712 + ], + "score": 1.0, + "content": "outliers to regularize the model, our fine-tuned model obtained by UM and UMAP can also serve as", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 710, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 506, + 722 + ], + "score": 1.0, + "content": "their starting point or adjustment. As our strategy does not require any auxiliary outlier data to be", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 721, + 486, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 486, + 732 + ], + "score": 1.0, + "content": "involved in training, adjusting the model using ID data during its developing phase is practical.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 37.5, + "bbox_fs": [ + 104, + 621, + 507, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 81, + 200, + 93 + ], + "lines": [ + { + "bbox": [ + 105, + 79, + 201, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 79, + 201, + 96 + ], + "score": 1.0, + "content": "4 EXPERIMENTS", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 107, + 505, + 173 + ], + "lines": [ + { + "bbox": [ + 105, + 107, + 506, + 120 + ], + "spans": [ + { + "bbox": [ + 105, + 107, + 506, + 120 + ], + "score": 1.0, + "content": "In this section, we present the performance comparison of the proposed method in the OOD detection", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 118, + 506, + 130 + ], + "spans": [ + { + "bbox": [ + 105, + 118, + 506, + 130 + ], + "score": 1.0, + "content": "scenario. Specifically, we verify the effectiveness of our UM and UMAP with two mainstreams of", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 130, + 505, + 141 + ], + "spans": [ + { + "bbox": [ + 106, + 130, + 505, + 141 + ], + "score": 1.0, + "content": "OOD detection approaches: (i) fundamental scoring function methods; (ii) outlier exposure methods", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 140, + 506, + 153 + ], + "spans": [ + { + "bbox": [ + 106, + 140, + 506, + 153 + ], + "score": 1.0, + "content": "involving auxiliary samples. To better understand and characterize our proposed method, we further", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 152, + 505, + 163 + ], + "spans": [ + { + "bbox": [ + 106, + 152, + 505, + 163 + ], + "score": 1.0, + "content": "conduct extensive explorations on the ablation study and provide the corresponding discussion on each", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 162, + 506, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 162, + 506, + 174 + ], + "score": 1.0, + "content": "sub-aspect considered in our work. More details and additional results can also refer to Appendix F.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 3.5 + }, + { + "type": "title", + "bbox": [ + 107, + 189, + 235, + 200 + ], + "lines": [ + { + "bbox": [ + 105, + 187, + 236, + 202 + ], + "spans": [ + { + "bbox": [ + 105, + 187, + 236, + 202 + ], + "score": 1.0, + "content": "4.1 EXPERIMENTAL SETUPS", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 107, + 211, + 505, + 288 + ], + "lines": [ + { + "bbox": [ + 105, + 210, + 506, + 224 + ], + "spans": [ + { + "bbox": [ + 105, + 210, + 506, + 224 + ], + "score": 1.0, + "content": "Datasets. Following the common benchmarks used in previous work (Liu et al., 2020; Ming et al.,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 221, + 506, + 234 + ], + "spans": [ + { + "bbox": [ + 105, + 221, + 506, + 234 + ], + "score": 1.0, + "content": "2022), we adopt CIFAR-10, CIFAR-100 (Krizhevsky, 2009) as our ID datasets. We use a series of", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 232, + 505, + 245 + ], + "spans": [ + { + "bbox": [ + 106, + 232, + 505, + 245 + ], + "score": 1.0, + "content": "different image datasets as the OOD datasets, namely Textures (Cimpoi et al., 2014), Places365", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 244, + 505, + 256 + ], + "spans": [ + { + "bbox": [ + 106, + 244, + 505, + 256 + ], + "score": 1.0, + "content": "(Zhou et al., 2017), SUN (Xiao et al., 2010), LSUN (Yu et al., 2015), and iNaturalist (Van Horn", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 253, + 506, + 267 + ], + "spans": [ + { + "bbox": [ + 105, + 253, + 506, + 267 + ], + "score": 1.0, + "content": "et al., 2018). We also use the other ID dataset as OOD dataset when training on a specific ID dataset,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 266, + 506, + 279 + ], + "spans": [ + { + "bbox": [ + 105, + 266, + 506, + 279 + ], + "score": 1.0, + "content": "given that none of them shares any same classes (Yang et al., 2021). e.g. we treat CIFAR-100 as", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 276, + 439, + 290 + ], + "spans": [ + { + "bbox": [ + 105, + 276, + 439, + 290 + ], + "score": 1.0, + "content": "the OOD dataset when training on CIFAR-10 in our experiments for comparison.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 107, + 302, + 505, + 390 + ], + "lines": [ + { + "bbox": [ + 105, + 301, + 505, + 315 + ], + "spans": [ + { + "bbox": [ + 105, + 301, + 505, + 315 + ], + "score": 1.0, + "content": "Training details. We conduct all major experiments on DenseNet-101 (Huang et al., 2017) with", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 313, + 506, + 325 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 506, + 325 + ], + "score": 1.0, + "content": "training epochs fixed to 100. We also include experiment results on other types of models in the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 325, + 505, + 336 + ], + "spans": [ + { + "bbox": [ + 106, + 325, + 505, + 336 + ], + "score": 1.0, + "content": "Appendix F. The models are trained using stochastic gradient descent (Kiefer & Wolfowitz, 1952)", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 335, + 506, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 335, + 506, + 347 + ], + "score": 1.0, + "content": "with Nesterov momentum (Duchi et al., 2011). We adopt Cosine Annealing (Loshchilov & Hutter,", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 345, + 506, + 359 + ], + "spans": [ + { + "bbox": [ + 105, + 345, + 506, + 359 + ], + "score": 1.0, + "content": "2017) to schedule the learning rate which begins at 0.1. We set the momentum and weight decay to", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 356, + 506, + 370 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 151, + 370 + ], + "score": 1.0, + "content": "be 0.9 and", + "type": "text" + }, + { + "bbox": [ + 151, + 357, + 173, + 368 + ], + "score": 0.9, + "content": "1 0 ^ { - 4 }", + "type": "inline_equation" + }, + { + "bbox": [ + 173, + 356, + 506, + 370 + ], + "score": 1.0, + "content": "respectively throughout all experiments. The size of the mini-batch is 64 for both", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 367, + 506, + 380 + ], + "spans": [ + { + "bbox": [ + 105, + 367, + 506, + 380 + ], + "score": 1.0, + "content": "ID samples (when training and testing) and OOD samples (when testing). More details and further", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 378, + 507, + 392 + ], + "spans": [ + { + "bbox": [ + 106, + 378, + 507, + 392 + ], + "score": 1.0, + "content": "discussion about choosing the mask ratio in experiments can be referred to at the end of Appendix F.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 18.5 + }, + { + "type": "text", + "bbox": [ + 107, + 405, + 505, + 504 + ], + "lines": [ + { + "bbox": [ + 105, + 404, + 506, + 417 + ], + "spans": [ + { + "bbox": [ + 105, + 404, + 506, + 417 + ], + "score": 1.0, + "content": "Evaluation metrics. We employ the following three common metrics to evaluate the performance", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "score": 1.0, + "content": "of OOD detection: (i) Area Under the Receiver Operating Characteristic curve (AUROC) (Davis", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 427, + 506, + 440 + ], + "spans": [ + { + "bbox": [ + 105, + 427, + 506, + 440 + ], + "score": 1.0, + "content": "& Goadrich, 2006) can be interpreted as the probability for a positive sample to have a higher", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 438, + 505, + 450 + ], + "spans": [ + { + "bbox": [ + 106, + 438, + 505, + 450 + ], + "score": 1.0, + "content": "discriminating score than a negative sample (Fawcett, 2006); (ii) Area Under the Precision-Recall", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 449, + 506, + 461 + ], + "spans": [ + { + "bbox": [ + 105, + 449, + 506, + 461 + ], + "score": 1.0, + "content": "curve (AUPR) (Manning & Schütze, 1999) is an ideal metric to adjust the extreme difference between", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 459, + 506, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 378, + 473 + ], + "score": 1.0, + "content": "positive and negative base rates; (iii) False Positive Rate (FPR) at", + "type": "text" + }, + { + "bbox": [ + 378, + 460, + 398, + 470 + ], + "score": 0.88, + "content": "9 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 398, + 459, + 506, + 473 + ], + "score": 1.0, + "content": "True Positive Rate (TPR)", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "(Liang et al., 2018) indicates the probability for a negative sample to be misclassified as positive", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 482, + 506, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 234, + 495 + ], + "score": 1.0, + "content": "when the true positive rate is at", + "type": "text" + }, + { + "bbox": [ + 234, + 482, + 253, + 492 + ], + "score": 0.87, + "content": "9 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 253, + 482, + 506, + 495 + ], + "score": 1.0, + "content": ". We also include in-distribution testing accuracy (ID-ACC) to", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 492, + 482, + 505 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 482, + 505 + ], + "score": 1.0, + "content": "reflect the preservation level of the performance for the original classification task on ID data.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 107, + 518, + 505, + 628 + ], + "lines": [ + { + "bbox": [ + 106, + 518, + 505, + 531 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 505, + 531 + ], + "score": 1.0, + "content": "OOD detection baselines. We compare the proposed method with several competitive baselines", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 529, + 506, + 542 + ], + "spans": [ + { + "bbox": [ + 105, + 529, + 506, + 542 + ], + "score": 1.0, + "content": "in the two directions. Specifically, we adopt Maximum Softmax Probability (MSP) (Hendrycks &", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 540, + 505, + 553 + ], + "spans": [ + { + "bbox": [ + 105, + 540, + 505, + 553 + ], + "score": 1.0, + "content": "Gimpel, 2017a), ODIN (Liang et al., 2018), Mahalanobis score (Lee et al., 2018b), and Energy score", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 550, + 507, + 565 + ], + "spans": [ + { + "bbox": [ + 105, + 550, + 507, + 565 + ], + "score": 1.0, + "content": "(Liu et al., 2020) as scoring function baselines; We adopt OE (Hendrycks et al., 2019b), Energy-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 561, + 506, + 575 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 506, + 575 + ], + "score": 1.0, + "content": "bounded learning (Liu et al., 2020), and POEM (Ming et al., 2022) as baselines with outliers. For", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 573, + 505, + 586 + ], + "spans": [ + { + "bbox": [ + 106, + 573, + 505, + 586 + ], + "score": 1.0, + "content": "all scoring function methods, we assume the accessibility of well-trained models. For all methods", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 583, + 505, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 583, + 505, + 597 + ], + "score": 1.0, + "content": "involving outliers, we constrain all major experiments to a fine-tuning scenario, which is more", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 594, + 505, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 594, + 505, + 609 + ], + "score": 1.0, + "content": "practical in real cases. Different from training a dual-task model at the very beginning, equipping", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 606, + 505, + 619 + ], + "spans": [ + { + "bbox": [ + 105, + 606, + 505, + 619 + ], + "score": 1.0, + "content": "deployed models with OOD Detection ability is a much more common circumstance, considering the", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 616, + 485, + 630 + ], + "spans": [ + { + "bbox": [ + 105, + 616, + 485, + 630 + ], + "score": 1.0, + "content": "millions of existing and running deep learning systems. We leave more details in Appendix A.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 36.5 + }, + { + "type": "title", + "bbox": [ + 108, + 644, + 257, + 655 + ], + "lines": [ + { + "bbox": [ + 105, + 644, + 259, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 259, + 657 + ], + "score": 1.0, + "content": "4.2 PERFORMANCE COMPARISON", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 42 + }, + { + "type": "text", + "bbox": [ + 107, + 666, + 505, + 731 + ], + "lines": [ + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "score": 1.0, + "content": "In this part, we present the performance comparison with some representative baseline methods to", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "demonstrate the effectiveness of our UM and UMAP. Our proposed UM is designed for excavating the", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 104, + 687, + 506, + 701 + ], + "spans": [ + { + "bbox": [ + 104, + 687, + 506, + 701 + ], + "score": 1.0, + "content": "potential OOD detection capability of the given model. Here we consider several scoring functions to", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "score": 1.0, + "content": "compare the detection performance, and also some outlier exposure methods to further regularize", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "the given model and boost the OOD detection ability. In each category, we choose one with the best", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 720, + 481, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 481, + 733 + ], + "score": 1.0, + "content": "detection performance to adopt UM/UMAP and check the detection results with the ID-ACC.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 45.5 + } + ], + "page_idx": 6, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 25, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 25, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 303, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 13, + "width": 7 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 81, + 200, + 93 + ], + "lines": [ + { + "bbox": [ + 105, + 79, + 201, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 79, + 201, + 96 + ], + "score": 1.0, + "content": "4 EXPERIMENTS", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 107, + 505, + 173 + ], + "lines": [ + { + "bbox": [ + 105, + 107, + 506, + 120 + ], + "spans": [ + { + "bbox": [ + 105, + 107, + 506, + 120 + ], + "score": 1.0, + "content": "In this section, we present the performance comparison of the proposed method in the OOD detection", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 118, + 506, + 130 + ], + "spans": [ + { + "bbox": [ + 105, + 118, + 506, + 130 + ], + "score": 1.0, + "content": "scenario. Specifically, we verify the effectiveness of our UM and UMAP with two mainstreams of", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 130, + 505, + 141 + ], + "spans": [ + { + "bbox": [ + 106, + 130, + 505, + 141 + ], + "score": 1.0, + "content": "OOD detection approaches: (i) fundamental scoring function methods; (ii) outlier exposure methods", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 140, + 506, + 153 + ], + "spans": [ + { + "bbox": [ + 106, + 140, + 506, + 153 + ], + "score": 1.0, + "content": "involving auxiliary samples. To better understand and characterize our proposed method, we further", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 152, + 505, + 163 + ], + "spans": [ + { + "bbox": [ + 106, + 152, + 505, + 163 + ], + "score": 1.0, + "content": "conduct extensive explorations on the ablation study and provide the corresponding discussion on each", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 162, + 506, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 162, + 506, + 174 + ], + "score": 1.0, + "content": "sub-aspect considered in our work. More details and additional results can also refer to Appendix F.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 3.5, + "bbox_fs": [ + 105, + 107, + 506, + 174 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 189, + 235, + 200 + ], + "lines": [ + { + "bbox": [ + 105, + 187, + 236, + 202 + ], + "spans": [ + { + "bbox": [ + 105, + 187, + 236, + 202 + ], + "score": 1.0, + "content": "4.1 EXPERIMENTAL SETUPS", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 107, + 211, + 505, + 288 + ], + "lines": [ + { + "bbox": [ + 105, + 210, + 506, + 224 + ], + "spans": [ + { + "bbox": [ + 105, + 210, + 506, + 224 + ], + "score": 1.0, + "content": "Datasets. Following the common benchmarks used in previous work (Liu et al., 2020; Ming et al.,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 221, + 506, + 234 + ], + "spans": [ + { + "bbox": [ + 105, + 221, + 506, + 234 + ], + "score": 1.0, + "content": "2022), we adopt CIFAR-10, CIFAR-100 (Krizhevsky, 2009) as our ID datasets. We use a series of", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 232, + 505, + 245 + ], + "spans": [ + { + "bbox": [ + 106, + 232, + 505, + 245 + ], + "score": 1.0, + "content": "different image datasets as the OOD datasets, namely Textures (Cimpoi et al., 2014), Places365", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 244, + 505, + 256 + ], + "spans": [ + { + "bbox": [ + 106, + 244, + 505, + 256 + ], + "score": 1.0, + "content": "(Zhou et al., 2017), SUN (Xiao et al., 2010), LSUN (Yu et al., 2015), and iNaturalist (Van Horn", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 253, + 506, + 267 + ], + "spans": [ + { + "bbox": [ + 105, + 253, + 506, + 267 + ], + "score": 1.0, + "content": "et al., 2018). We also use the other ID dataset as OOD dataset when training on a specific ID dataset,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 266, + 506, + 279 + ], + "spans": [ + { + "bbox": [ + 105, + 266, + 506, + 279 + ], + "score": 1.0, + "content": "given that none of them shares any same classes (Yang et al., 2021). e.g. we treat CIFAR-100 as", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 276, + 439, + 290 + ], + "spans": [ + { + "bbox": [ + 105, + 276, + 439, + 290 + ], + "score": 1.0, + "content": "the OOD dataset when training on CIFAR-10 in our experiments for comparison.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 210, + 506, + 290 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 302, + 505, + 390 + ], + "lines": [ + { + "bbox": [ + 105, + 301, + 505, + 315 + ], + "spans": [ + { + "bbox": [ + 105, + 301, + 505, + 315 + ], + "score": 1.0, + "content": "Training details. We conduct all major experiments on DenseNet-101 (Huang et al., 2017) with", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 313, + 506, + 325 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 506, + 325 + ], + "score": 1.0, + "content": "training epochs fixed to 100. We also include experiment results on other types of models in the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 325, + 505, + 336 + ], + "spans": [ + { + "bbox": [ + 106, + 325, + 505, + 336 + ], + "score": 1.0, + "content": "Appendix F. The models are trained using stochastic gradient descent (Kiefer & Wolfowitz, 1952)", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 335, + 506, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 335, + 506, + 347 + ], + "score": 1.0, + "content": "with Nesterov momentum (Duchi et al., 2011). We adopt Cosine Annealing (Loshchilov & Hutter,", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 345, + 506, + 359 + ], + "spans": [ + { + "bbox": [ + 105, + 345, + 506, + 359 + ], + "score": 1.0, + "content": "2017) to schedule the learning rate which begins at 0.1. We set the momentum and weight decay to", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 356, + 506, + 370 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 151, + 370 + ], + "score": 1.0, + "content": "be 0.9 and", + "type": "text" + }, + { + "bbox": [ + 151, + 357, + 173, + 368 + ], + "score": 0.9, + "content": "1 0 ^ { - 4 }", + "type": "inline_equation" + }, + { + "bbox": [ + 173, + 356, + 506, + 370 + ], + "score": 1.0, + "content": "respectively throughout all experiments. The size of the mini-batch is 64 for both", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 367, + 506, + 380 + ], + "spans": [ + { + "bbox": [ + 105, + 367, + 506, + 380 + ], + "score": 1.0, + "content": "ID samples (when training and testing) and OOD samples (when testing). More details and further", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 378, + 507, + 392 + ], + "spans": [ + { + "bbox": [ + 106, + 378, + 507, + 392 + ], + "score": 1.0, + "content": "discussion about choosing the mask ratio in experiments can be referred to at the end of Appendix F.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 18.5, + "bbox_fs": [ + 105, + 301, + 507, + 392 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 405, + 505, + 504 + ], + "lines": [ + { + "bbox": [ + 105, + 404, + 506, + 417 + ], + "spans": [ + { + "bbox": [ + 105, + 404, + 506, + 417 + ], + "score": 1.0, + "content": "Evaluation metrics. We employ the following three common metrics to evaluate the performance", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "score": 1.0, + "content": "of OOD detection: (i) Area Under the Receiver Operating Characteristic curve (AUROC) (Davis", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 427, + 506, + 440 + ], + "spans": [ + { + "bbox": [ + 105, + 427, + 506, + 440 + ], + "score": 1.0, + "content": "& Goadrich, 2006) can be interpreted as the probability for a positive sample to have a higher", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 438, + 505, + 450 + ], + "spans": [ + { + "bbox": [ + 106, + 438, + 505, + 450 + ], + "score": 1.0, + "content": "discriminating score than a negative sample (Fawcett, 2006); (ii) Area Under the Precision-Recall", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 449, + 506, + 461 + ], + "spans": [ + { + "bbox": [ + 105, + 449, + 506, + 461 + ], + "score": 1.0, + "content": "curve (AUPR) (Manning & Schütze, 1999) is an ideal metric to adjust the extreme difference between", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 459, + 506, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 378, + 473 + ], + "score": 1.0, + "content": "positive and negative base rates; (iii) False Positive Rate (FPR) at", + "type": "text" + }, + { + "bbox": [ + 378, + 460, + 398, + 470 + ], + "score": 0.88, + "content": "9 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 398, + 459, + 506, + 473 + ], + "score": 1.0, + "content": "True Positive Rate (TPR)", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "(Liang et al., 2018) indicates the probability for a negative sample to be misclassified as positive", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 482, + 506, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 234, + 495 + ], + "score": 1.0, + "content": "when the true positive rate is at", + "type": "text" + }, + { + "bbox": [ + 234, + 482, + 253, + 492 + ], + "score": 0.87, + "content": "9 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 253, + 482, + 506, + 495 + ], + "score": 1.0, + "content": ". We also include in-distribution testing accuracy (ID-ACC) to", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 492, + 482, + 505 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 482, + 505 + ], + "score": 1.0, + "content": "reflect the preservation level of the performance for the original classification task on ID data.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 27, + "bbox_fs": [ + 105, + 404, + 506, + 505 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 518, + 505, + 628 + ], + "lines": [ + { + "bbox": [ + 106, + 518, + 505, + 531 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 505, + 531 + ], + "score": 1.0, + "content": "OOD detection baselines. We compare the proposed method with several competitive baselines", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 529, + 506, + 542 + ], + "spans": [ + { + "bbox": [ + 105, + 529, + 506, + 542 + ], + "score": 1.0, + "content": "in the two directions. Specifically, we adopt Maximum Softmax Probability (MSP) (Hendrycks &", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 540, + 505, + 553 + ], + "spans": [ + { + "bbox": [ + 105, + 540, + 505, + 553 + ], + "score": 1.0, + "content": "Gimpel, 2017a), ODIN (Liang et al., 2018), Mahalanobis score (Lee et al., 2018b), and Energy score", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 550, + 507, + 565 + ], + "spans": [ + { + "bbox": [ + 105, + 550, + 507, + 565 + ], + "score": 1.0, + "content": "(Liu et al., 2020) as scoring function baselines; We adopt OE (Hendrycks et al., 2019b), Energy-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 561, + 506, + 575 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 506, + 575 + ], + "score": 1.0, + "content": "bounded learning (Liu et al., 2020), and POEM (Ming et al., 2022) as baselines with outliers. For", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 573, + 505, + 586 + ], + "spans": [ + { + "bbox": [ + 106, + 573, + 505, + 586 + ], + "score": 1.0, + "content": "all scoring function methods, we assume the accessibility of well-trained models. For all methods", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 583, + 505, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 583, + 505, + 597 + ], + "score": 1.0, + "content": "involving outliers, we constrain all major experiments to a fine-tuning scenario, which is more", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 594, + 505, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 594, + 505, + 609 + ], + "score": 1.0, + "content": "practical in real cases. Different from training a dual-task model at the very beginning, equipping", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 606, + 505, + 619 + ], + "spans": [ + { + "bbox": [ + 105, + 606, + 505, + 619 + ], + "score": 1.0, + "content": "deployed models with OOD Detection ability is a much more common circumstance, considering the", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 616, + 485, + 630 + ], + "spans": [ + { + "bbox": [ + 105, + 616, + 485, + 630 + ], + "score": 1.0, + "content": "millions of existing and running deep learning systems. We leave more details in Appendix A.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 36.5, + "bbox_fs": [ + 105, + 518, + 507, + 630 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 644, + 257, + 655 + ], + "lines": [ + { + "bbox": [ + 105, + 644, + 259, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 259, + 657 + ], + "score": 1.0, + "content": "4.2 PERFORMANCE COMPARISON", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 42 + }, + { + "type": "text", + "bbox": [ + 107, + 666, + 505, + 731 + ], + "lines": [ + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "score": 1.0, + "content": "In this part, we present the performance comparison with some representative baseline methods to", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "demonstrate the effectiveness of our UM and UMAP. Our proposed UM is designed for excavating the", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 104, + 687, + 506, + 701 + ], + "spans": [ + { + "bbox": [ + 104, + 687, + 506, + 701 + ], + "score": 1.0, + "content": "potential OOD detection capability of the given model. Here we consider several scoring functions to", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "score": 1.0, + "content": "compare the detection performance, and also some outlier exposure methods to further regularize", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "the given model and boost the OOD detection ability. In each category, we choose one with the best", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 720, + 481, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 481, + 733 + ], + "score": 1.0, + "content": "detection performance to adopt UM/UMAP and check the detection results with the ID-ACC.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 45.5, + "bbox_fs": [ + 104, + 665, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 99, + 505, + 304 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 135, + 80, + 472, + 92 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 137, + 79, + 474, + 93 + ], + "spans": [ + { + "bbox": [ + 137, + 79, + 228, + 93 + ], + "score": 1.0, + "content": "Table 1: Main Results", + "type": "text" + }, + { + "bbox": [ + 228, + 80, + 243, + 91 + ], + "score": 0.77, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 244, + 79, + 474, + 93 + ], + "score": 1.0, + "content": ". Comparison with competitive OOD detection baselines.", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table_body", + "bbox": [ + 108, + 99, + 505, + 304 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 99, + 505, + 304 + ], + "spans": [ + { + "bbox": [ + 108, + 99, + 505, + 304 + ], + "score": 0.984, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑w./w.o Daux
CIFAR-10MSP(Hendrycks & Gimpel,2017a) ODIN(Liang et al.,2018)89.90 ±0.30 91.46 ± 0.5691.48 ± 0.43 91.67 ± 0.5860.08 ±0.76 42.31 ± 1.3894.01 ±0.08 94.01 ± 0.08
Mahalanobis(Lee et al.,2018b)75.10 ± 1.0472.32 ± 1.9261.35 ± 1.2594.01±0.08
Energy(Liu et al.,2020)92.07 ± 0.2292.72 ± 0.3942.69 ± 1.3194.01± 0.08
Energy+UM(ours)93.73 ±0.3694.27 ± 0.6033.29 ± 1.7092.80±0.47
Energy+UMAP(ours)93.97 ± 0.1194.38 ± 0.0630.71 ± 1.9494.01± 0.08
OE(Hendrycks et al., 2019b)97.07 ± 0.0197.31± 0.0513.80 ± 0.2892.59 ± 0.32
Energy (w.Daux)(Liu et al.,2020)94.58 ± 0.6494.69 ± 0.6518.79 ± 2.3180.91 ± 3.13
POEM(Ming et al.,2022) OE+UM(ours)94.37 ± 0.07 97.60 ± 0.0394.51± 0.06 97.87 ± 0.0218.50 ± 0.33 11.22 ± 0.1677.24 ± 2.22 93.66 ±0.12>>>>
OE+UMAP(ours)97.48 ± 0.0197.74 ± 0.0012.21± 0.0994.01 ± 0.08√
MSP(Hendrycks & Gimpel,2017a) ODIN(Liang et al., 2018)74.06 ± 0.6975.37 ± 0.7383.14± 0.8774.86 ± 0.21
Mahalanobis(Lee et al.,2018b)76.18 ± 0.14 63.90 ± 1.9176.49 ± 0.20 64.31 ± 0.9178.93 ± 0.31 78.79± 0.5074.86 ± 0.21
Energy(Liu et al.,2020)76.29 ±0.2477.06± 0.5578.46± 0.0674.86± 0.21 74.86 ± 0.21
Energy+UM(ours)76.22 ± 0.4276.39 ± 1.0374.05 ± 0.5564.55± 0.24
Energy+UMAP(ours)75.57 ± 0.5975.66 ±0.0772.21 ± 1.4674.86 ± 0.21
OE(Hendrycks et al., 2019b)
90.55 ± 0.87 88.92 ± 0.5790.34 ± 0.94 89.13 ± 0.5634.73 ± 3.85 37.90 ± 2.5973.59 ± 0.30
Energy (w.Daux)(Liu et al.,2020)57.85 ± 2.65广
POEM(Ming et al.,2022)88.95 ± 0.5488.94 ± 0.3138.10 ± 1.3056.18 ± 1.92√
OE+UM(ours)91.04±0.1191.13 ± 0.2434.71 ± 0.8175.15 ± 0.18√
OE+UMAP (ours)91.10 ± 0.1690.99 ±0.2333.62 ±0.2674.76 ± 0.11√
", + "type": "table", + "image_path": "dbf3cde50a959ea56cfbc57c0fe25efe3f04e4f79e781821b4936a380e3c5ee4.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 108, + 99, + 505, + 167.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 167.33333333333331, + 505, + 235.66666666666663 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 108, + 235.66666666666663, + 505, + 303.99999999999994 + ], + "spans": [], + "index": 3 + } + ] + } + ], + "index": 1.0 + }, + { + "type": "table", + "bbox": [ + 108, + 335, + 505, + 465 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 131, + 316, + 478, + 328 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 131, + 315, + 480, + 330 + ], + "spans": [ + { + "bbox": [ + 131, + 315, + 252, + 330 + ], + "score": 1.0, + "content": "Table 2: Fine-grained Results", + "type": "text" + }, + { + "bbox": [ + 252, + 316, + 267, + 328 + ], + "score": 0.78, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 267, + 315, + 480, + 330 + ], + "score": 1.0, + "content": ". Comparison on different OOD benchmark datasets.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "table_body", + "bbox": [ + 108, + 335, + 505, + 465 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 108, + 335, + 505, + 465 + ], + "spans": [ + { + "bbox": [ + 108, + 335, + 505, + 465 + ], + "score": 0.98, + "html": "
ID datasetMethodOOD dataset
CIFAR-100TexturesPlaces365
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-10MSP66.43 ± 1.2587.73± 0.0265.20±1.3388.06 ±0.6161.34 ± 0.6089.63±0.15
ODIN55.31± 0.8587.75 ± 0.3753.11 ± 4.8487.13 ± 2.0443.77 ±0.2091.70 ± 0.30
Mahalanobis81.61 ± 0.9664.52 ±0..7320.04 ± 1.4394.38 ± 0.7886.21±1.3664.00 ± 1.21
Energy54.65 ± 1.2489.01 ± 1.1857.09 ±3.5287.51 ± 1.4338.62 ± 1.6493.03±0.20
Energy+UM(ours)54.62 ± 1.1688.30±0.3041.61 ± 3.6791.31 ± 0.0130.85 ±0.5894.27 ±0.16
MethodSUN FPR95↓LSUNiNaturalist
AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
MSP60.27 ± 0.6690.00±0.2436.43 ± 1.9495.17 ± 0.3267.53 ± 1.6488.01±0.82
ODIN41.14 ± 1.2992.34±0.625.16 ±0.7698.96 ±0.0954.41 ± 0.9190.17 ± 0.19
Mahalanobis84.56 ± 1.5166.41 ± 4.5769.18 ± 3.5266.41 ± 4.5780.76± 2.4871.77 ± 1.12
Energy36.73 ± 1.7293.63 ± 0.34 94.83 ± 0.116.25 ± 0.4398.77±0.07 99.22 ± 0.1159.11 ± 1.1889.71 ± 0.06 92.75 ± 0.80
Energy+UM(ours)27.88 ±0.732.91 ± 0.5346.27 ± 2.74
", + "type": "table", + "image_path": "8fb71e78539bf2c0d61786777ad61aa9dee9b855b204b5fc5b22ef8d595be04e.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 108, + 335, + 505, + 378.3333333333333 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 108, + 378.3333333333333, + 505, + 421.66666666666663 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 108, + 421.66666666666663, + 505, + 464.99999999999994 + ], + "spans": [], + "index": 7 + } + ] + } + ], + "index": 5.0 + }, + { + "type": "text", + "bbox": [ + 106, + 477, + 505, + 663 + ], + "lines": [ + { + "bbox": [ + 105, + 476, + 507, + 489 + ], + "spans": [ + { + "bbox": [ + 105, + 476, + 507, + 489 + ], + "score": 1.0, + "content": "In Table 1, we summarize the results of different OOD test sets using different methods. Note that,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 487, + 506, + 500 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 506, + 500 + ], + "score": 1.0, + "content": "here the evaluation results are obtained by averaging several OOD test datasets across multiple", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 499, + 505, + 510 + ], + "spans": [ + { + "bbox": [ + 106, + 499, + 505, + 510 + ], + "score": 1.0, + "content": "independent trials. For the scoring-based methods, our UM can further improve the overall detection", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 509, + 505, + 522 + ], + "spans": [ + { + "bbox": [ + 105, + 509, + 505, + 522 + ], + "score": 1.0, + "content": "performance by alleviating the memorization of atypical ID data, when the ID-ACC keeps comparable", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 519, + 505, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 519, + 505, + 534 + ], + "score": 1.0, + "content": "with the baseline. For the complex CIFAR-100 dataset, our UMAP can be adopted as a practical way", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 531, + 505, + 544 + ], + "spans": [ + { + "bbox": [ + 105, + 531, + 505, + 544 + ], + "score": 1.0, + "content": "to empower the detection performance and simultaneously avoid affecting the original performance", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 104, + 540, + 504, + 558 + ], + "spans": [ + { + "bbox": [ + 104, + 540, + 480, + 558 + ], + "score": 1.0, + "content": "on ID data. As for those methods of the second category (i.e., involving auxiliary outlier", + "type": "text" + }, + { + "bbox": [ + 480, + 543, + 504, + 554 + ], + "score": 0.9, + "content": "D _ { a u x }", + "type": "inline_equation" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 553, + 505, + 566 + ], + "spans": [ + { + "bbox": [ + 106, + 553, + 505, + 566 + ], + "score": 1.0, + "content": "sampling from ImageNet), since we consider a practical workflow, i.e., fine-tuning, on the given", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 564, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 506, + 577 + ], + "score": 1.0, + "content": "model, OE achieves the best performance on the task. Due to the special optimization characteristic,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 575, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 575, + 139, + 588 + ], + "score": 1.0, + "content": "Energy", + "type": "text" + }, + { + "bbox": [ + 140, + 576, + 178, + 587 + ], + "score": 0.88, + "content": "( \\mathbf { w . } D _ { a u x } )", + "type": "inline_equation" + }, + { + "bbox": [ + 179, + 575, + 505, + 588 + ], + "score": 1.0, + "content": "and POEM focus more on the energy loss on differentiating OOD data while", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 586, + 505, + 599 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 505, + 599 + ], + "score": 1.0, + "content": "performing not well on the preservation of ID-ACC. Without sacrificing much performance on ID", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 597, + 506, + 610 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 506, + 610 + ], + "score": 1.0, + "content": "data, OE with our UM can still achieve better detection performance. In Table 3, the fine-grained", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 608, + 506, + 620 + ], + "spans": [ + { + "bbox": [ + 105, + 608, + 506, + 620 + ], + "score": 1.0, + "content": "detection performance on each OOD testing set demonstrates the general effectiveness of UM. We", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 618, + 505, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 618, + 368, + 631 + ], + "score": 1.0, + "content": "have comprehensively verified the significant improvement (up to", + "type": "text" + }, + { + "bbox": [ + 368, + 619, + 388, + 630 + ], + "score": 0.87, + "content": "1 8 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 388, + 618, + 505, + 631 + ], + "score": 1.0, + "content": "reduced on averaged FPR95)", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 630, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 506, + 642 + ], + "score": 1.0, + "content": "in OOD detection of our methods across different setups in Appendix F, the complete results can refer", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 641, + 506, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 506, + 654 + ], + "score": 1.0, + "content": "to Tables 8 to 20. More fine-grained results of the experiments on CIFAR-100 is provided in Table 16", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 652, + 482, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 652, + 482, + 664 + ], + "score": 1.0, + "content": "and 19. In addition, we also provide similar results using another model structure in Table 18.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 16 + }, + { + "type": "title", + "bbox": [ + 108, + 678, + 296, + 689 + ], + "lines": [ + { + "bbox": [ + 105, + 677, + 298, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 298, + 690 + ], + "score": 1.0, + "content": "4.3 ABLATION AND FURTHER DISCUSSION", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "In this part, we conduct various explorations to provide a thorough understanding of our presented", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 708, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 506, + 723 + ], + "score": 1.0, + "content": "Unleashing Mask from different perspectives. To be specific, we first present the general existence of", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 720, + 506, + 734 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 506, + 734 + ], + "score": 1.0, + "content": "the once-covered intrinsic OOD detection capability across different settings. Second, we present", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27 + } + ], + "page_idx": 7, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 309, + 761 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 309, + 761 + ], + "score": 1.0, + "content": "8", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 99, + 505, + 304 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 135, + 80, + 472, + 92 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 137, + 79, + 474, + 93 + ], + "spans": [ + { + "bbox": [ + 137, + 79, + 228, + 93 + ], + "score": 1.0, + "content": "Table 1: Main Results", + "type": "text" + }, + { + "bbox": [ + 228, + 80, + 243, + 91 + ], + "score": 0.77, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 244, + 79, + 474, + 93 + ], + "score": 1.0, + "content": ". Comparison with competitive OOD detection baselines.", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table_body", + "bbox": [ + 108, + 99, + 505, + 304 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 99, + 505, + 304 + ], + "spans": [ + { + "bbox": [ + 108, + 99, + 505, + 304 + ], + "score": 0.984, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑w./w.o Daux
CIFAR-10MSP(Hendrycks & Gimpel,2017a) ODIN(Liang et al.,2018)89.90 ±0.30 91.46 ± 0.5691.48 ± 0.43 91.67 ± 0.5860.08 ±0.76 42.31 ± 1.3894.01 ±0.08 94.01 ± 0.08
Mahalanobis(Lee et al.,2018b)75.10 ± 1.0472.32 ± 1.9261.35 ± 1.2594.01±0.08
Energy(Liu et al.,2020)92.07 ± 0.2292.72 ± 0.3942.69 ± 1.3194.01± 0.08
Energy+UM(ours)93.73 ±0.3694.27 ± 0.6033.29 ± 1.7092.80±0.47
Energy+UMAP(ours)93.97 ± 0.1194.38 ± 0.0630.71 ± 1.9494.01± 0.08
OE(Hendrycks et al., 2019b)97.07 ± 0.0197.31± 0.0513.80 ± 0.2892.59 ± 0.32
Energy (w.Daux)(Liu et al.,2020)94.58 ± 0.6494.69 ± 0.6518.79 ± 2.3180.91 ± 3.13
POEM(Ming et al.,2022) OE+UM(ours)94.37 ± 0.07 97.60 ± 0.0394.51± 0.06 97.87 ± 0.0218.50 ± 0.33 11.22 ± 0.1677.24 ± 2.22 93.66 ±0.12>>>>
OE+UMAP(ours)97.48 ± 0.0197.74 ± 0.0012.21± 0.0994.01 ± 0.08√
MSP(Hendrycks & Gimpel,2017a) ODIN(Liang et al., 2018)74.06 ± 0.6975.37 ± 0.7383.14± 0.8774.86 ± 0.21
Mahalanobis(Lee et al.,2018b)76.18 ± 0.14 63.90 ± 1.9176.49 ± 0.20 64.31 ± 0.9178.93 ± 0.31 78.79± 0.5074.86 ± 0.21
Energy(Liu et al.,2020)76.29 ±0.2477.06± 0.5578.46± 0.0674.86± 0.21 74.86 ± 0.21
Energy+UM(ours)76.22 ± 0.4276.39 ± 1.0374.05 ± 0.5564.55± 0.24
Energy+UMAP(ours)75.57 ± 0.5975.66 ±0.0772.21 ± 1.4674.86 ± 0.21
OE(Hendrycks et al., 2019b)
90.55 ± 0.87 88.92 ± 0.5790.34 ± 0.94 89.13 ± 0.5634.73 ± 3.85 37.90 ± 2.5973.59 ± 0.30
Energy (w.Daux)(Liu et al.,2020)57.85 ± 2.65广
POEM(Ming et al.,2022)88.95 ± 0.5488.94 ± 0.3138.10 ± 1.3056.18 ± 1.92√
OE+UM(ours)91.04±0.1191.13 ± 0.2434.71 ± 0.8175.15 ± 0.18√
OE+UMAP (ours)91.10 ± 0.1690.99 ±0.2333.62 ±0.2674.76 ± 0.11√
", + "type": "table", + "image_path": "dbf3cde50a959ea56cfbc57c0fe25efe3f04e4f79e781821b4936a380e3c5ee4.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 108, + 99, + 505, + 167.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 167.33333333333331, + 505, + 235.66666666666663 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 108, + 235.66666666666663, + 505, + 303.99999999999994 + ], + "spans": [], + "index": 3 + } + ] + } + ], + "index": 1.0 + }, + { + "type": "table", + "bbox": [ + 108, + 335, + 505, + 465 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 131, + 316, + 478, + 328 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 131, + 315, + 480, + 330 + ], + "spans": [ + { + "bbox": [ + 131, + 315, + 252, + 330 + ], + "score": 1.0, + "content": "Table 2: Fine-grained Results", + "type": "text" + }, + { + "bbox": [ + 252, + 316, + 267, + 328 + ], + "score": 0.78, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 267, + 315, + 480, + 330 + ], + "score": 1.0, + "content": ". Comparison on different OOD benchmark datasets.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "table_body", + "bbox": [ + 108, + 335, + 505, + 465 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 108, + 335, + 505, + 465 + ], + "spans": [ + { + "bbox": [ + 108, + 335, + 505, + 465 + ], + "score": 0.98, + "html": "
ID datasetMethodOOD dataset
CIFAR-100TexturesPlaces365
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-10MSP66.43 ± 1.2587.73± 0.0265.20±1.3388.06 ±0.6161.34 ± 0.6089.63±0.15
ODIN55.31± 0.8587.75 ± 0.3753.11 ± 4.8487.13 ± 2.0443.77 ±0.2091.70 ± 0.30
Mahalanobis81.61 ± 0.9664.52 ±0..7320.04 ± 1.4394.38 ± 0.7886.21±1.3664.00 ± 1.21
Energy54.65 ± 1.2489.01 ± 1.1857.09 ±3.5287.51 ± 1.4338.62 ± 1.6493.03±0.20
Energy+UM(ours)54.62 ± 1.1688.30±0.3041.61 ± 3.6791.31 ± 0.0130.85 ±0.5894.27 ±0.16
MethodSUN FPR95↓LSUNiNaturalist
AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
MSP60.27 ± 0.6690.00±0.2436.43 ± 1.9495.17 ± 0.3267.53 ± 1.6488.01±0.82
ODIN41.14 ± 1.2992.34±0.625.16 ±0.7698.96 ±0.0954.41 ± 0.9190.17 ± 0.19
Mahalanobis84.56 ± 1.5166.41 ± 4.5769.18 ± 3.5266.41 ± 4.5780.76± 2.4871.77 ± 1.12
Energy36.73 ± 1.7293.63 ± 0.34 94.83 ± 0.116.25 ± 0.4398.77±0.07 99.22 ± 0.1159.11 ± 1.1889.71 ± 0.06 92.75 ± 0.80
Energy+UM(ours)27.88 ±0.732.91 ± 0.5346.27 ± 2.74
", + "type": "table", + "image_path": "8fb71e78539bf2c0d61786777ad61aa9dee9b855b204b5fc5b22ef8d595be04e.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 108, + 335, + 505, + 378.3333333333333 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 108, + 378.3333333333333, + 505, + 421.66666666666663 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 108, + 421.66666666666663, + 505, + 464.99999999999994 + ], + "spans": [], + "index": 7 + } + ] + } + ], + "index": 5.0 + }, + { + "type": "text", + "bbox": [ + 106, + 477, + 505, + 663 + ], + "lines": [ + { + "bbox": [ + 105, + 476, + 507, + 489 + ], + "spans": [ + { + "bbox": [ + 105, + 476, + 507, + 489 + ], + "score": 1.0, + "content": "In Table 1, we summarize the results of different OOD test sets using different methods. Note that,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 487, + 506, + 500 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 506, + 500 + ], + "score": 1.0, + "content": "here the evaluation results are obtained by averaging several OOD test datasets across multiple", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 499, + 505, + 510 + ], + "spans": [ + { + "bbox": [ + 106, + 499, + 505, + 510 + ], + "score": 1.0, + "content": "independent trials. For the scoring-based methods, our UM can further improve the overall detection", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 509, + 505, + 522 + ], + "spans": [ + { + "bbox": [ + 105, + 509, + 505, + 522 + ], + "score": 1.0, + "content": "performance by alleviating the memorization of atypical ID data, when the ID-ACC keeps comparable", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 519, + 505, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 519, + 505, + 534 + ], + "score": 1.0, + "content": "with the baseline. For the complex CIFAR-100 dataset, our UMAP can be adopted as a practical way", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 531, + 505, + 544 + ], + "spans": [ + { + "bbox": [ + 105, + 531, + 505, + 544 + ], + "score": 1.0, + "content": "to empower the detection performance and simultaneously avoid affecting the original performance", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 104, + 540, + 504, + 558 + ], + "spans": [ + { + "bbox": [ + 104, + 540, + 480, + 558 + ], + "score": 1.0, + "content": "on ID data. As for those methods of the second category (i.e., involving auxiliary outlier", + "type": "text" + }, + { + "bbox": [ + 480, + 543, + 504, + 554 + ], + "score": 0.9, + "content": "D _ { a u x }", + "type": "inline_equation" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 553, + 505, + 566 + ], + "spans": [ + { + "bbox": [ + 106, + 553, + 505, + 566 + ], + "score": 1.0, + "content": "sampling from ImageNet), since we consider a practical workflow, i.e., fine-tuning, on the given", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 564, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 506, + 577 + ], + "score": 1.0, + "content": "model, OE achieves the best performance on the task. Due to the special optimization characteristic,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 575, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 575, + 139, + 588 + ], + "score": 1.0, + "content": "Energy", + "type": "text" + }, + { + "bbox": [ + 140, + 576, + 178, + 587 + ], + "score": 0.88, + "content": "( \\mathbf { w . } D _ { a u x } )", + "type": "inline_equation" + }, + { + "bbox": [ + 179, + 575, + 505, + 588 + ], + "score": 1.0, + "content": "and POEM focus more on the energy loss on differentiating OOD data while", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 586, + 505, + 599 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 505, + 599 + ], + "score": 1.0, + "content": "performing not well on the preservation of ID-ACC. Without sacrificing much performance on ID", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 597, + 506, + 610 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 506, + 610 + ], + "score": 1.0, + "content": "data, OE with our UM can still achieve better detection performance. In Table 3, the fine-grained", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 608, + 506, + 620 + ], + "spans": [ + { + "bbox": [ + 105, + 608, + 506, + 620 + ], + "score": 1.0, + "content": "detection performance on each OOD testing set demonstrates the general effectiveness of UM. We", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 618, + 505, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 618, + 368, + 631 + ], + "score": 1.0, + "content": "have comprehensively verified the significant improvement (up to", + "type": "text" + }, + { + "bbox": [ + 368, + 619, + 388, + 630 + ], + "score": 0.87, + "content": "1 8 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 388, + 618, + 505, + 631 + ], + "score": 1.0, + "content": "reduced on averaged FPR95)", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 630, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 506, + 642 + ], + "score": 1.0, + "content": "in OOD detection of our methods across different setups in Appendix F, the complete results can refer", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 641, + 506, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 506, + 654 + ], + "score": 1.0, + "content": "to Tables 8 to 20. More fine-grained results of the experiments on CIFAR-100 is provided in Table 16", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 652, + 482, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 652, + 482, + 664 + ], + "score": 1.0, + "content": "and 19. In addition, we also provide similar results using another model structure in Table 18.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 16, + "bbox_fs": [ + 104, + 476, + 507, + 664 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 678, + 296, + 689 + ], + "lines": [ + { + "bbox": [ + 105, + 677, + 298, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 298, + 690 + ], + "score": 1.0, + "content": "4.3 ABLATION AND FURTHER DISCUSSION", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "In this part, we conduct various explorations to provide a thorough understanding of our presented", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 708, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 506, + 723 + ], + "score": 1.0, + "content": "Unleashing Mask from different perspectives. To be specific, we first present the general existence of", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 720, + 506, + 734 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 506, + 734 + ], + "score": 1.0, + "content": "the once-covered intrinsic OOD detection capability across different settings. Second, we present", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 234, + 505, + 247 + ], + "spans": [ + { + "bbox": [ + 106, + 234, + 505, + 247 + ], + "score": 1.0, + "content": "the exploration about the mask, which helps us to characterize its effect on figuring out the atypical", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 244, + 485, + 260 + ], + "spans": [ + { + "bbox": [ + 105, + 244, + 485, + 260 + ], + "score": 1.0, + "content": "samples. Third, we provide further exploration on excavating the detection power via pruning.", + "type": "text", + "cross_page": true + } + ], + "index": 8 + } + ], + "index": 27, + "bbox_fs": [ + 105, + 698, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 113, + 86, + 499, + 167 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 113, + 86, + 499, + 167 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 113, + 86, + 499, + 167 + ], + "spans": [ + { + "bbox": [ + 113, + 86, + 499, + 167 + ], + "score": 0.961, + "type": "image", + "image_path": "ddbd0f1c22b91a329ea8bf02dce6b6e526822b5945af1d952567d8f633227704.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 113, + 86, + 499, + 113.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 113, + 113.0, + 499, + 140.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 113, + 140.0, + 499, + 167.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 174, + 505, + 219 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 174, + 505, + 186 + ], + "spans": [ + { + "bbox": [ + 106, + 174, + 505, + 186 + ], + "score": 1.0, + "content": "Figure 4: Ablation studies. (a)-(b) exploring the existence of potential OOD detection capability with", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 184, + 506, + 198 + ], + "spans": [ + { + "bbox": [ + 105, + 184, + 506, + 198 + ], + "score": 1.0, + "content": "different learning rate schedules; (c) comparison of loss value using original output with masked", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 196, + 505, + 208 + ], + "spans": [ + { + "bbox": [ + 106, + 196, + 152, + 208 + ], + "score": 1.0, + "content": "output (the", + "type": "text" + }, + { + "bbox": [ + 152, + 198, + 158, + 206 + ], + "score": 0.38, + "content": "\\mathbf { X }", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 196, + 505, + 208 + ], + "score": 1.0, + "content": "-axis represents the index of samples within a mini-batch); (d) effects of using different", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 207, + 471, + 219 + ], + "spans": [ + { + "bbox": [ + 105, + 207, + 471, + 219 + ], + "score": 1.0, + "content": "masking ratios in UM; (e) comparison of using original pruning with our proposed UMAP.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "text", + "bbox": [ + 107, + 234, + 503, + 257 + ], + "lines": [ + { + "bbox": [ + 106, + 234, + 505, + 247 + ], + "spans": [ + { + "bbox": [ + 106, + 234, + 505, + 247 + ], + "score": 1.0, + "content": "the exploration about the mask, which helps us to characterize its effect on figuring out the atypical", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 244, + 485, + 260 + ], + "spans": [ + { + "bbox": [ + 105, + 244, + 485, + 260 + ], + "score": 1.0, + "content": "samples. Third, we provide further exploration on excavating the detection power via pruning.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 107, + 268, + 505, + 357 + ], + "lines": [ + { + "bbox": [ + 105, + 268, + 505, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 268, + 505, + 282 + ], + "score": 1.0, + "content": "General existence of once-covered OOD detection capability. In Figure 4(a) and Figure 4(b), we", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 280, + 505, + 293 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 505, + 293 + ], + "score": 1.0, + "content": "explore 4 learning rate schedules to demonstrate the general existence of once-covered OOD detection", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 291, + 505, + 303 + ], + "spans": [ + { + "bbox": [ + 105, + 291, + 505, + 303 + ], + "score": 1.0, + "content": "capability. To be specific, the OOD performance (indicated by FPR95) is evaluated along with the", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 302, + 506, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 302, + 506, + 314 + ], + "score": 1.0, + "content": "training in every 5 epoch, in which the model takes CIFAR-10 as ID data and SVHN as OOD data.", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 313, + 506, + 325 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 506, + 325 + ], + "score": 1.0, + "content": "As shown by the curves, a middle stage exists with better OOD performance than that of the final", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 324, + 506, + 336 + ], + "spans": [ + { + "bbox": [ + 105, + 324, + 506, + 336 + ], + "score": 1.0, + "content": "stage across different schedules. We empirically verify the existence of this phenomenon without", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 336, + 505, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 336, + 505, + 347 + ], + "score": 1.0, + "content": "schedule specificity. More explorations on the other ID dataset and model structure reveal similar", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 345, + 490, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 345, + 490, + 358 + ], + "score": 1.0, + "content": "results in Figures 5, 6, and 7. The detailed information and discussion can refer to Appendix F.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 12.5 + }, + { + "type": "text", + "bbox": [ + 107, + 369, + 505, + 458 + ], + "lines": [ + { + "bbox": [ + 105, + 369, + 506, + 383 + ], + "spans": [ + { + "bbox": [ + 105, + 369, + 506, + 383 + ], + "score": 1.0, + "content": "Effects of the mask on mining atypical samples. Following our previous illustration of Figure 3(c),", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 381, + 505, + 393 + ], + "spans": [ + { + "bbox": [ + 105, + 381, + 505, + 393 + ], + "score": 1.0, + "content": "we scrutinize the change of training loss on a random batch of the training set in Figure 4(c). The", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 392, + 505, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 392, + 505, + 403 + ], + "score": 1.0, + "content": "results further explain why the loss value estimated by the UM can be used to force the model to", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 402, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 105, + 402, + 505, + 416 + ], + "score": 1.0, + "content": "forget atypical samples. It can be seen from Figure 4(c) that the loss is proportionally increased by", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 413, + 505, + 426 + ], + "spans": [ + { + "bbox": [ + 105, + 414, + 204, + 426 + ], + "score": 1.0, + "content": "randomly knocking out", + "type": "text" + }, + { + "bbox": [ + 204, + 413, + 226, + 424 + ], + "score": 0.87, + "content": "2 . 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 227, + 414, + 505, + 426 + ], + "score": 1.0, + "content": "weights. In this case, the estimated loss is more influenced by those", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 423, + 505, + 438 + ], + "spans": [ + { + "bbox": [ + 105, + 423, + 505, + 438 + ], + "score": 1.0, + "content": "who have a higher initial loss and are what we termed as atypical samples. By controlling the training", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 435, + 505, + 448 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 505, + 448 + ], + "score": 1.0, + "content": "loss to the estimated value, the model is encouraged to backtrack to a middle training stage where", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 446, + 496, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 496, + 459 + ], + "score": 1.0, + "content": "samples with high loss value have little influence on the forgetting process of the gradient ascent.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 20.5 + }, + { + "type": "text", + "bbox": [ + 107, + 470, + 505, + 591 + ], + "lines": [ + { + "bbox": [ + 106, + 470, + 506, + 483 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 506, + 483 + ], + "score": 1.0, + "content": "Exploration on revealing detection capability with model pruning. Although the large constrain", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 481, + 505, + 493 + ], + "spans": [ + { + "bbox": [ + 106, + 481, + 505, + 493 + ], + "score": 1.0, + "content": "on training loss can help reveal the model’s OOD performance, the ID-ACC is undermined under", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 491, + 506, + 505 + ], + "spans": [ + { + "bbox": [ + 105, + 491, + 506, + 505 + ], + "score": 1.0, + "content": "such circumstances. Generally speaking, the proposed UM forces the model to forget the atypical", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 502, + 506, + 517 + ], + "spans": [ + { + "bbox": [ + 105, + 502, + 506, + 517 + ], + "score": 1.0, + "content": "samples and may result in lower test performance. To mitigate this issue, we further adopt pruning as", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 104, + 513, + 506, + 527 + ], + "spans": [ + { + "bbox": [ + 104, + 513, + 506, + 527 + ], + "score": 1.0, + "content": "a countermeasure to learn a mask instead of tuning the model parameters directly. In Figure 4(e), we", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 525, + 505, + 537 + ], + "spans": [ + { + "bbox": [ + 105, + 525, + 251, + 537 + ], + "score": 1.0, + "content": "experiment with various prune rates", + "type": "text" + }, + { + "bbox": [ + 252, + 526, + 258, + 536 + ], + "score": 0.8, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 259, + 525, + 505, + 537 + ], + "score": 1.0, + "content": "and demonstrate that we can achieve the same or better OOD", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 535, + 506, + 549 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 506, + 549 + ], + "score": 1.0, + "content": "performance also by pruning. Specifically, our UMAP can achieve a lower FPR95 than pure pruning", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 546, + 505, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 546, + 436, + 560 + ], + "score": 1.0, + "content": "with the original objective. The prune rate can be selected from a wide range (e.g.", + "type": "text" + }, + { + "bbox": [ + 437, + 547, + 492, + 558 + ], + "score": 0.91, + "content": "p \\in [ 0 . 3 , 0 . 9 ] )", + "type": "inline_equation" + }, + { + "bbox": [ + 492, + 546, + 505, + 560 + ], + "score": 1.0, + "content": ") to", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 557, + 506, + 570 + ], + "spans": [ + { + "bbox": [ + 105, + 557, + 506, + 570 + ], + "score": 1.0, + "content": "guarantee a fast convergence and effectiveness. Since pruning doesn’t change the well-trained model", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 569, + 506, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 569, + 506, + 581 + ], + "score": 1.0, + "content": "parameters, it can preserve the performance of the original task. We also provide additional empirical", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 579, + 469, + 592 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 469, + 592 + ], + "score": 1.0, + "content": "results and corresponding discussion about the effectiveness of our UMAP in Appendix F.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 30 + }, + { + "type": "title", + "bbox": [ + 108, + 607, + 195, + 620 + ], + "lines": [ + { + "bbox": [ + 104, + 605, + 198, + 623 + ], + "spans": [ + { + "bbox": [ + 104, + 605, + 198, + 623 + ], + "score": 1.0, + "content": "5 CONCLUSION", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 36 + }, + { + "type": "text", + "bbox": [ + 107, + 631, + 505, + 720 + ], + "lines": [ + { + "bbox": [ + 106, + 632, + 506, + 644 + ], + "spans": [ + { + "bbox": [ + 106, + 632, + 506, + 644 + ], + "score": 1.0, + "content": "In this work, we explore the intrinsic OOD detection capability of a well-trained model. Without", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 642, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 506, + 657 + ], + "score": 1.0, + "content": "involving any auxiliary outliers in training, we reveal the inconsistent trend between minimizing", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "score": 1.0, + "content": "original training loss and gaining OOD detection capability. We further attribute it to the memorization", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 665, + 506, + 677 + ], + "spans": [ + { + "bbox": [ + 106, + 665, + 506, + 677 + ], + "score": 1.0, + "content": "behavior of atypical samples. To excavate the once-covered capability, we propose a new method,", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 676, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 689 + ], + "score": 1.0, + "content": "namely, Unleashing Mask (UM). Through UM, we construct model-level discrepancy that figures out", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 104, + 685, + 507, + 701 + ], + "spans": [ + { + "bbox": [ + 104, + 685, + 507, + 701 + ], + "score": 1.0, + "content": "the memorized atypical samples and utilizes the constrained gradient ascent to encourage forgetting.", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 697, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 697, + 505, + 711 + ], + "score": 1.0, + "content": "It better utilizes the given model for OOD detection via backtracking or sub-structure pruning. We", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 708, + 503, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 503, + 722 + ], + "score": 1.0, + "content": "hope our work could provide new insights for revisiting the model development in OOD detection.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 40.5 + } + ], + "page_idx": 8, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 106, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "score": 1.0, + "content": "9", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 113, + 86, + 499, + 167 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 113, + 86, + 499, + 167 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 113, + 86, + 499, + 167 + ], + "spans": [ + { + "bbox": [ + 113, + 86, + 499, + 167 + ], + "score": 0.961, + "type": "image", + "image_path": "ddbd0f1c22b91a329ea8bf02dce6b6e526822b5945af1d952567d8f633227704.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 113, + 86, + 499, + 113.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 113, + 113.0, + 499, + 140.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 113, + 140.0, + 499, + 167.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 174, + 505, + 219 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 174, + 505, + 186 + ], + "spans": [ + { + "bbox": [ + 106, + 174, + 505, + 186 + ], + "score": 1.0, + "content": "Figure 4: Ablation studies. (a)-(b) exploring the existence of potential OOD detection capability with", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 184, + 506, + 198 + ], + "spans": [ + { + "bbox": [ + 105, + 184, + 506, + 198 + ], + "score": 1.0, + "content": "different learning rate schedules; (c) comparison of loss value using original output with masked", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 196, + 505, + 208 + ], + "spans": [ + { + "bbox": [ + 106, + 196, + 152, + 208 + ], + "score": 1.0, + "content": "output (the", + "type": "text" + }, + { + "bbox": [ + 152, + 198, + 158, + 206 + ], + "score": 0.38, + "content": "\\mathbf { X }", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 196, + 505, + 208 + ], + "score": 1.0, + "content": "-axis represents the index of samples within a mini-batch); (d) effects of using different", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 207, + 471, + 219 + ], + "spans": [ + { + "bbox": [ + 105, + 207, + 471, + 219 + ], + "score": 1.0, + "content": "masking ratios in UM; (e) comparison of using original pruning with our proposed UMAP.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "text", + "bbox": [ + 107, + 234, + 503, + 257 + ], + "lines": [], + "index": 7.5, + "bbox_fs": [ + 105, + 234, + 505, + 260 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 268, + 505, + 357 + ], + "lines": [ + { + "bbox": [ + 105, + 268, + 505, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 268, + 505, + 282 + ], + "score": 1.0, + "content": "General existence of once-covered OOD detection capability. In Figure 4(a) and Figure 4(b), we", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 280, + 505, + 293 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 505, + 293 + ], + "score": 1.0, + "content": "explore 4 learning rate schedules to demonstrate the general existence of once-covered OOD detection", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 291, + 505, + 303 + ], + "spans": [ + { + "bbox": [ + 105, + 291, + 505, + 303 + ], + "score": 1.0, + "content": "capability. To be specific, the OOD performance (indicated by FPR95) is evaluated along with the", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 302, + 506, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 302, + 506, + 314 + ], + "score": 1.0, + "content": "training in every 5 epoch, in which the model takes CIFAR-10 as ID data and SVHN as OOD data.", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 313, + 506, + 325 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 506, + 325 + ], + "score": 1.0, + "content": "As shown by the curves, a middle stage exists with better OOD performance than that of the final", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 324, + 506, + 336 + ], + "spans": [ + { + "bbox": [ + 105, + 324, + 506, + 336 + ], + "score": 1.0, + "content": "stage across different schedules. We empirically verify the existence of this phenomenon without", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 336, + 505, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 336, + 505, + 347 + ], + "score": 1.0, + "content": "schedule specificity. More explorations on the other ID dataset and model structure reveal similar", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 345, + 490, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 345, + 490, + 358 + ], + "score": 1.0, + "content": "results in Figures 5, 6, and 7. The detailed information and discussion can refer to Appendix F.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 12.5, + "bbox_fs": [ + 105, + 268, + 506, + 358 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 369, + 505, + 458 + ], + "lines": [ + { + "bbox": [ + 105, + 369, + 506, + 383 + ], + "spans": [ + { + "bbox": [ + 105, + 369, + 506, + 383 + ], + "score": 1.0, + "content": "Effects of the mask on mining atypical samples. Following our previous illustration of Figure 3(c),", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 381, + 505, + 393 + ], + "spans": [ + { + "bbox": [ + 105, + 381, + 505, + 393 + ], + "score": 1.0, + "content": "we scrutinize the change of training loss on a random batch of the training set in Figure 4(c). The", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 392, + 505, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 392, + 505, + 403 + ], + "score": 1.0, + "content": "results further explain why the loss value estimated by the UM can be used to force the model to", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 402, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 105, + 402, + 505, + 416 + ], + "score": 1.0, + "content": "forget atypical samples. It can be seen from Figure 4(c) that the loss is proportionally increased by", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 413, + 505, + 426 + ], + "spans": [ + { + "bbox": [ + 105, + 414, + 204, + 426 + ], + "score": 1.0, + "content": "randomly knocking out", + "type": "text" + }, + { + "bbox": [ + 204, + 413, + 226, + 424 + ], + "score": 0.87, + "content": "2 . 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 227, + 414, + 505, + 426 + ], + "score": 1.0, + "content": "weights. In this case, the estimated loss is more influenced by those", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 423, + 505, + 438 + ], + "spans": [ + { + "bbox": [ + 105, + 423, + 505, + 438 + ], + "score": 1.0, + "content": "who have a higher initial loss and are what we termed as atypical samples. By controlling the training", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 435, + 505, + 448 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 505, + 448 + ], + "score": 1.0, + "content": "loss to the estimated value, the model is encouraged to backtrack to a middle training stage where", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 446, + 496, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 496, + 459 + ], + "score": 1.0, + "content": "samples with high loss value have little influence on the forgetting process of the gradient ascent.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 20.5, + "bbox_fs": [ + 105, + 369, + 506, + 459 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 470, + 505, + 591 + ], + "lines": [ + { + "bbox": [ + 106, + 470, + 506, + 483 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 506, + 483 + ], + "score": 1.0, + "content": "Exploration on revealing detection capability with model pruning. Although the large constrain", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 481, + 505, + 493 + ], + "spans": [ + { + "bbox": [ + 106, + 481, + 505, + 493 + ], + "score": 1.0, + "content": "on training loss can help reveal the model’s OOD performance, the ID-ACC is undermined under", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 491, + 506, + 505 + ], + "spans": [ + { + "bbox": [ + 105, + 491, + 506, + 505 + ], + "score": 1.0, + "content": "such circumstances. Generally speaking, the proposed UM forces the model to forget the atypical", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 502, + 506, + 517 + ], + "spans": [ + { + "bbox": [ + 105, + 502, + 506, + 517 + ], + "score": 1.0, + "content": "samples and may result in lower test performance. To mitigate this issue, we further adopt pruning as", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 104, + 513, + 506, + 527 + ], + "spans": [ + { + "bbox": [ + 104, + 513, + 506, + 527 + ], + "score": 1.0, + "content": "a countermeasure to learn a mask instead of tuning the model parameters directly. In Figure 4(e), we", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 525, + 505, + 537 + ], + "spans": [ + { + "bbox": [ + 105, + 525, + 251, + 537 + ], + "score": 1.0, + "content": "experiment with various prune rates", + "type": "text" + }, + { + "bbox": [ + 252, + 526, + 258, + 536 + ], + "score": 0.8, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 259, + 525, + 505, + 537 + ], + "score": 1.0, + "content": "and demonstrate that we can achieve the same or better OOD", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 535, + 506, + 549 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 506, + 549 + ], + "score": 1.0, + "content": "performance also by pruning. Specifically, our UMAP can achieve a lower FPR95 than pure pruning", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 546, + 505, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 546, + 436, + 560 + ], + "score": 1.0, + "content": "with the original objective. The prune rate can be selected from a wide range (e.g.", + "type": "text" + }, + { + "bbox": [ + 437, + 547, + 492, + 558 + ], + "score": 0.91, + "content": "p \\in [ 0 . 3 , 0 . 9 ] )", + "type": "inline_equation" + }, + { + "bbox": [ + 492, + 546, + 505, + 560 + ], + "score": 1.0, + "content": ") to", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 557, + 506, + 570 + ], + "spans": [ + { + "bbox": [ + 105, + 557, + 506, + 570 + ], + "score": 1.0, + "content": "guarantee a fast convergence and effectiveness. Since pruning doesn’t change the well-trained model", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 569, + 506, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 569, + 506, + 581 + ], + "score": 1.0, + "content": "parameters, it can preserve the performance of the original task. We also provide additional empirical", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 579, + 469, + 592 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 469, + 592 + ], + "score": 1.0, + "content": "results and corresponding discussion about the effectiveness of our UMAP in Appendix F.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 30, + "bbox_fs": [ + 104, + 470, + 506, + 592 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 607, + 195, + 620 + ], + "lines": [ + { + "bbox": [ + 104, + 605, + 198, + 623 + ], + "spans": [ + { + "bbox": [ + 104, + 605, + 198, + 623 + ], + "score": 1.0, + "content": "5 CONCLUSION", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 36 + }, + { + "type": "text", + "bbox": [ + 107, + 631, + 505, + 720 + ], + "lines": [ + { + "bbox": [ + 106, + 632, + 506, + 644 + ], + "spans": [ + { + "bbox": [ + 106, + 632, + 506, + 644 + ], + "score": 1.0, + "content": "In this work, we explore the intrinsic OOD detection capability of a well-trained model. Without", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 642, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 506, + 657 + ], + "score": 1.0, + "content": "involving any auxiliary outliers in training, we reveal the inconsistent trend between minimizing", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "score": 1.0, + "content": "original training loss and gaining OOD detection capability. We further attribute it to the memorization", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 665, + 506, + 677 + ], + "spans": [ + { + "bbox": [ + 106, + 665, + 506, + 677 + ], + "score": 1.0, + "content": "behavior of atypical samples. To excavate the once-covered capability, we propose a new method,", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 676, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 689 + ], + "score": 1.0, + "content": "namely, Unleashing Mask (UM). Through UM, we construct model-level discrepancy that figures out", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 104, + 685, + 507, + 701 + ], + "spans": [ + { + "bbox": [ + 104, + 685, + 507, + 701 + ], + "score": 1.0, + "content": "the memorized atypical samples and utilizes the constrained gradient ascent to encourage forgetting.", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 697, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 697, + 505, + 711 + ], + "score": 1.0, + "content": "It better utilizes the given model for OOD detection via backtracking or sub-structure pruning. We", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 708, + 503, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 503, + 722 + ], + "score": 1.0, + "content": "hope our work could provide new insights for revisiting the model development in OOD detection.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 40.5, + "bbox_fs": [ + 104, + 632, + 507, + 722 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 81, + 268, + 94 + ], + "lines": [ + { + "bbox": [ + 106, + 81, + 269, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 269, + 95 + ], + "score": 1.0, + "content": "REPRODUCIBILITY STATEMENT", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 106, + 505, + 140 + ], + "lines": [ + { + "bbox": [ + 105, + 106, + 507, + 119 + ], + "spans": [ + { + "bbox": [ + 105, + 106, + 507, + 119 + ], + "score": 1.0, + "content": "For the experimental setups, we have provided the details in Section 4.1 and Appendixes A and E.", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 117, + 506, + 131 + ], + "spans": [ + { + "bbox": [ + 105, + 117, + 506, + 131 + ], + "score": 1.0, + "content": "We will also provide the anonymous repository about our source codes in the discussion phase for", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 128, + 413, + 141 + ], + "spans": [ + { + "bbox": [ + 105, + 128, + 413, + 141 + ], + "score": 1.0, + "content": "reviewing purposes to ensure the reproducibility of our experimental results.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2 + }, + { + "type": "title", + "bbox": [ + 107, + 156, + 176, + 168 + ], + "lines": [ + { + "bbox": [ + 106, + 156, + 176, + 170 + ], + "spans": [ + { + "bbox": [ + 106, + 156, + 176, + 170 + ], + "score": 1.0, + "content": "REFERENCES", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 108, + 175, + 506, + 208 + ], + "lines": [ + { + "bbox": [ + 106, + 175, + 506, + 187 + ], + "spans": [ + { + "bbox": [ + 106, + 175, + 506, + 187 + ], + "score": 1.0, + "content": "Devansh Arpit, Stanislaw Jastrzebski, Nicolas Ballas, David Krueger, Emmanuel Bengio, Maxinder S.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 116, + 186, + 507, + 199 + ], + "spans": [ + { + "bbox": [ + 116, + 186, + 507, + 199 + ], + "score": 1.0, + "content": "Kanwal, Tegan Maharaj, Asja Fischer, Aaron C. Courville, Yoshua Bengio, and Simon Lacoste-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 115, + 197, + 409, + 209 + ], + "spans": [ + { + "bbox": [ + 115, + 197, + 409, + 209 + ], + "score": 1.0, + "content": "Julien. A closer look at memorization in deep networks. In ICML, 2017.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 107, + 216, + 506, + 250 + ], + "lines": [ + { + "bbox": [ + 105, + 215, + 506, + 230 + ], + "spans": [ + { + "bbox": [ + 105, + 215, + 506, + 230 + ], + "score": 1.0, + "content": "Paul Barham, Aakanksha Chowdhery, Jeff Dean, Sanjay Ghemawat, Steven Hand, Daniel Hurt,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 115, + 227, + 505, + 241 + ], + "spans": [ + { + "bbox": [ + 115, + 227, + 505, + 241 + ], + "score": 1.0, + "content": "Michael Isard, Hyeontaek Lim, Ruoming Pang, Sudip Roy, et al. Pathways: Asynchronous", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 117, + 239, + 297, + 251 + ], + "spans": [ + { + "bbox": [ + 117, + 239, + 297, + 251 + ], + "score": 1.0, + "content": "distributed dataflow for ml. In MLSys, 2022.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 106, + 258, + 505, + 291 + ], + "lines": [ + { + "bbox": [ + 105, + 256, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 105, + 256, + 505, + 271 + ], + "score": 1.0, + "content": "Mikhail Belkin, Daniel Hsu, Siyuan Ma, and Soumik Mandal. Reconciling modern machine-learning", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 115, + 269, + 506, + 282 + ], + "spans": [ + { + "bbox": [ + 115, + 269, + 506, + 282 + ], + "score": 1.0, + "content": "practice and the classical bias–variance trade-off. In Proceedings of the National Academy of", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 115, + 280, + 180, + 291 + ], + "spans": [ + { + "bbox": [ + 115, + 280, + 180, + 291 + ], + "score": 1.0, + "content": "Sciences, 2019.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12 + }, + { + "type": "text", + "bbox": [ + 104, + 299, + 505, + 322 + ], + "lines": [ + { + "bbox": [ + 106, + 300, + 506, + 313 + ], + "spans": [ + { + "bbox": [ + 106, + 300, + 506, + 313 + ], + "score": 1.0, + "content": "Jiefeng Chen, Yixuan Li, Xi Wu, Yingyu Liang, and Somesh Jha. Atom: Robustifying out-of-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 116, + 310, + 387, + 323 + ], + "spans": [ + { + "bbox": [ + 116, + 310, + 387, + 323 + ], + "score": 1.0, + "content": "distribution detection using outlier mining. In ECML PKDD, 2021.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5 + }, + { + "type": "text", + "bbox": [ + 106, + 329, + 505, + 353 + ], + "lines": [ + { + "bbox": [ + 106, + 330, + 506, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 330, + 506, + 343 + ], + "score": 1.0, + "content": "Mircea Cimpoi, Subhransu Maji, Iasonas Kokkinos, Sammy Mohamed, and Andrea Vedaldi. De-", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 116, + 341, + 299, + 353 + ], + "spans": [ + { + "bbox": [ + 116, + 341, + 299, + 353 + ], + "score": 1.0, + "content": "scribing textures in the wild. In CVPR, 2014.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5 + }, + { + "type": "text", + "bbox": [ + 105, + 360, + 505, + 383 + ], + "lines": [ + { + "bbox": [ + 105, + 360, + 506, + 373 + ], + "spans": [ + { + "bbox": [ + 105, + 360, + 506, + 373 + ], + "score": 1.0, + "content": "Jesse Davis and Mark Goadrich. The relationship between precision-recall and roc curves. In ICML,", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 114, + 370, + 143, + 384 + ], + "spans": [ + { + "bbox": [ + 114, + 370, + 143, + 384 + ], + "score": 1.0, + "content": "2006.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18.5 + }, + { + "type": "text", + "bbox": [ + 105, + 390, + 504, + 414 + ], + "lines": [ + { + "bbox": [ + 106, + 391, + 505, + 404 + ], + "spans": [ + { + "bbox": [ + 106, + 391, + 505, + 404 + ], + "score": 1.0, + "content": "Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 115, + 402, + 299, + 414 + ], + "spans": [ + { + "bbox": [ + 115, + 402, + 299, + 414 + ], + "score": 1.0, + "content": "hierarchical image database. In CVPR, 2009.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20.5 + }, + { + "type": "text", + "bbox": [ + 106, + 421, + 504, + 444 + ], + "lines": [ + { + "bbox": [ + 106, + 421, + 505, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 421, + 505, + 434 + ], + "score": 1.0, + "content": "Xuefeng Du, Zhaoning Wang, Mu Cai, and Sharon Li. Towards unknown-aware learning with virtual", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 116, + 433, + 253, + 444 + ], + "spans": [ + { + "bbox": [ + 116, + 433, + 253, + 444 + ], + "score": 1.0, + "content": "outlier synthesis. In ICLR, 2022a.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22.5 + }, + { + "type": "text", + "bbox": [ + 106, + 451, + 504, + 475 + ], + "lines": [ + { + "bbox": [ + 106, + 451, + 505, + 465 + ], + "spans": [ + { + "bbox": [ + 106, + 451, + 505, + 465 + ], + "score": 1.0, + "content": "Xuefeng Du, Zhaoning Wang, Mu Cai, and Yixuan Li. VOS: learning what you don’t know by virtual", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 116, + 464, + 255, + 475 + ], + "spans": [ + { + "bbox": [ + 116, + 464, + 255, + 475 + ], + "score": 1.0, + "content": "outlier synthesis. In ICLR, 2022b.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24.5 + }, + { + "type": "text", + "bbox": [ + 106, + 482, + 504, + 505 + ], + "lines": [ + { + "bbox": [ + 106, + 482, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 106, + 482, + 505, + 495 + ], + "score": 1.0, + "content": "John Duchi, Elad Hazan, and Yoram Singer. Adaptive subgradient methods for online learning and", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 116, + 494, + 280, + 505 + ], + "spans": [ + { + "bbox": [ + 116, + 494, + 280, + 505 + ], + "score": 1.0, + "content": "stochastic optimization. In JMLR, 2011.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26.5 + }, + { + "type": "text", + "bbox": [ + 107, + 513, + 444, + 525 + ], + "lines": [ + { + "bbox": [ + 105, + 512, + 443, + 527 + ], + "spans": [ + { + "bbox": [ + 105, + 512, + 443, + 527 + ], + "score": 1.0, + "content": "Tom Fawcett. An introduction to roc analysis. In Pattern Recognition Letters, 2006.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 108, + 532, + 505, + 555 + ], + "lines": [ + { + "bbox": [ + 106, + 532, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 532, + 505, + 546 + ], + "score": 1.0, + "content": "Stanislav Fort, Jie Ren, and Balaji Lakshminarayanan. Exploring the limits of out-of-distribution", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 544, + 234, + 555 + ], + "spans": [ + { + "bbox": [ + 116, + 544, + 234, + 555 + ], + "score": 1.0, + "content": "detection. In NeurIPS, 2021.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5 + }, + { + "type": "text", + "bbox": [ + 106, + 563, + 504, + 586 + ], + "lines": [ + { + "bbox": [ + 106, + 563, + 506, + 576 + ], + "spans": [ + { + "bbox": [ + 106, + 563, + 506, + 576 + ], + "score": 1.0, + "content": "Jonathan Frankle and Michael Carbin. The lottery ticket hypothesis: Training pruned neural networks.", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 116, + 574, + 178, + 586 + ], + "spans": [ + { + "bbox": [ + 116, + 574, + 178, + 586 + ], + "score": 1.0, + "content": "In ICLR, 2019.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 106, + 593, + 504, + 616 + ], + "lines": [ + { + "bbox": [ + 104, + 592, + 506, + 607 + ], + "spans": [ + { + "bbox": [ + 104, + 592, + 506, + 607 + ], + "score": 1.0, + "content": "Ian Goodfellow, Yoshua Bengio, Aaron Courville, and Yoshua Bengio. Deep learning. MIT Press,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 114, + 603, + 143, + 617 + ], + "spans": [ + { + "bbox": [ + 114, + 603, + 143, + 617 + ], + "score": 1.0, + "content": "2016.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5 + }, + { + "type": "text", + "bbox": [ + 107, + 624, + 506, + 702 + ], + "lines": [ + { + "bbox": [ + 105, + 623, + 506, + 637 + ], + "spans": [ + { + "bbox": [ + 105, + 623, + 506, + 637 + ], + "score": 1.0, + "content": "Ian J. Goodfellow, Dumitru Erhan, Pierre Luc Carrier, Aaron Courville, Mehdi Mirza, Ben Hamner,", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 635, + 506, + 648 + ], + "spans": [ + { + "bbox": [ + 115, + 635, + 506, + 648 + ], + "score": 1.0, + "content": "Will Cukierski, Yichuan Tang, David Thaler, Dong-Hyun Lee, Yingbo Zhou, Chetan Ramaiah,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 646, + 506, + 659 + ], + "spans": [ + { + "bbox": [ + 115, + 646, + 506, + 659 + ], + "score": 1.0, + "content": "Fangxiang Feng, Ruifan Li, Xiaojie Wang, Dimitris Athanasakis, John Shawe-Taylor, Maxim", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 656, + 506, + 671 + ], + "spans": [ + { + "bbox": [ + 115, + 656, + 506, + 671 + ], + "score": 1.0, + "content": "Milakov, John Park, Radu Ionescu, Marius Popescu, Cristian Grozea, James Bergstra, Jingjing Xie,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 668, + 505, + 681 + ], + "spans": [ + { + "bbox": [ + 115, + 668, + 505, + 681 + ], + "score": 1.0, + "content": "Lukasz Romaszko, Bing Xu, Zhang Chuang, and Yoshua Bengio. Challenges in representation", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 114, + 678, + 506, + 693 + ], + "spans": [ + { + "bbox": [ + 114, + 678, + 506, + 693 + ], + "score": 1.0, + "content": "learning: A report on three machine learning contests. In Minho Lee, Akira Hirose, Zeng-Guang", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 690, + 419, + 703 + ], + "spans": [ + { + "bbox": [ + 115, + 690, + 419, + 703 + ], + "score": 1.0, + "content": "Hou, and Rhee Man Kil (eds.), NeurIPS. Springer Berlin Heidelberg, 2013.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 109, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 107, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 107, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "Song Han, Huizi Mao, and William J Dally. Deep compression: Compressing deep neural networks", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 116, + 720, + 403, + 733 + ], + "spans": [ + { + "bbox": [ + 116, + 720, + 403, + 733 + ], + "score": 1.0, + "content": "with pruning, trained quantization and huffman coding. In ICLR, 2016.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42.5 + } + ], + "page_idx": 9, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "10", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 81, + 268, + 94 + ], + "lines": [ + { + "bbox": [ + 106, + 81, + 269, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 269, + 95 + ], + "score": 1.0, + "content": "REPRODUCIBILITY STATEMENT", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 106, + 505, + 140 + ], + "lines": [ + { + "bbox": [ + 105, + 106, + 507, + 119 + ], + "spans": [ + { + "bbox": [ + 105, + 106, + 507, + 119 + ], + "score": 1.0, + "content": "For the experimental setups, we have provided the details in Section 4.1 and Appendixes A and E.", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 117, + 506, + 131 + ], + "spans": [ + { + "bbox": [ + 105, + 117, + 506, + 131 + ], + "score": 1.0, + "content": "We will also provide the anonymous repository about our source codes in the discussion phase for", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 128, + 413, + 141 + ], + "spans": [ + { + "bbox": [ + 105, + 128, + 413, + 141 + ], + "score": 1.0, + "content": "reviewing purposes to ensure the reproducibility of our experimental results.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2, + "bbox_fs": [ + 105, + 106, + 507, + 141 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 156, + 176, + 168 + ], + "lines": [ + { + "bbox": [ + 106, + 156, + 176, + 170 + ], + "spans": [ + { + "bbox": [ + 106, + 156, + 176, + 170 + ], + "score": 1.0, + "content": "REFERENCES", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 108, + 175, + 506, + 208 + ], + "lines": [ + { + "bbox": [ + 106, + 175, + 506, + 187 + ], + "spans": [ + { + "bbox": [ + 106, + 175, + 506, + 187 + ], + "score": 1.0, + "content": "Devansh Arpit, Stanislaw Jastrzebski, Nicolas Ballas, David Krueger, Emmanuel Bengio, Maxinder S.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 116, + 186, + 507, + 199 + ], + "spans": [ + { + "bbox": [ + 116, + 186, + 507, + 199 + ], + "score": 1.0, + "content": "Kanwal, Tegan Maharaj, Asja Fischer, Aaron C. Courville, Yoshua Bengio, and Simon Lacoste-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 115, + 197, + 409, + 209 + ], + "spans": [ + { + "bbox": [ + 115, + 197, + 409, + 209 + ], + "score": 1.0, + "content": "Julien. A closer look at memorization in deep networks. In ICML, 2017.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6, + "bbox_fs": [ + 106, + 175, + 507, + 209 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 216, + 506, + 250 + ], + "lines": [ + { + "bbox": [ + 105, + 215, + 506, + 230 + ], + "spans": [ + { + "bbox": [ + 105, + 215, + 506, + 230 + ], + "score": 1.0, + "content": "Paul Barham, Aakanksha Chowdhery, Jeff Dean, Sanjay Ghemawat, Steven Hand, Daniel Hurt,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 115, + 227, + 505, + 241 + ], + "spans": [ + { + "bbox": [ + 115, + 227, + 505, + 241 + ], + "score": 1.0, + "content": "Michael Isard, Hyeontaek Lim, Ruoming Pang, Sudip Roy, et al. Pathways: Asynchronous", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 117, + 239, + 297, + 251 + ], + "spans": [ + { + "bbox": [ + 117, + 239, + 297, + 251 + ], + "score": 1.0, + "content": "distributed dataflow for ml. In MLSys, 2022.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9, + "bbox_fs": [ + 105, + 215, + 506, + 251 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 258, + 505, + 291 + ], + "lines": [ + { + "bbox": [ + 105, + 256, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 105, + 256, + 505, + 271 + ], + "score": 1.0, + "content": "Mikhail Belkin, Daniel Hsu, Siyuan Ma, and Soumik Mandal. Reconciling modern machine-learning", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 115, + 269, + 506, + 282 + ], + "spans": [ + { + "bbox": [ + 115, + 269, + 506, + 282 + ], + "score": 1.0, + "content": "practice and the classical bias–variance trade-off. In Proceedings of the National Academy of", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 115, + 280, + 180, + 291 + ], + "spans": [ + { + "bbox": [ + 115, + 280, + 180, + 291 + ], + "score": 1.0, + "content": "Sciences, 2019.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12, + "bbox_fs": [ + 105, + 256, + 506, + 291 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 299, + 505, + 322 + ], + "lines": [ + { + "bbox": [ + 106, + 300, + 506, + 313 + ], + "spans": [ + { + "bbox": [ + 106, + 300, + 506, + 313 + ], + "score": 1.0, + "content": "Jiefeng Chen, Yixuan Li, Xi Wu, Yingyu Liang, and Somesh Jha. Atom: Robustifying out-of-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 116, + 310, + 387, + 323 + ], + "spans": [ + { + "bbox": [ + 116, + 310, + 387, + 323 + ], + "score": 1.0, + "content": "distribution detection using outlier mining. In ECML PKDD, 2021.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5, + "bbox_fs": [ + 106, + 300, + 506, + 323 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 329, + 505, + 353 + ], + "lines": [ + { + "bbox": [ + 106, + 330, + 506, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 330, + 506, + 343 + ], + "score": 1.0, + "content": "Mircea Cimpoi, Subhransu Maji, Iasonas Kokkinos, Sammy Mohamed, and Andrea Vedaldi. De-", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 116, + 341, + 299, + 353 + ], + "spans": [ + { + "bbox": [ + 116, + 341, + 299, + 353 + ], + "score": 1.0, + "content": "scribing textures in the wild. In CVPR, 2014.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5, + "bbox_fs": [ + 106, + 330, + 506, + 353 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 360, + 505, + 383 + ], + "lines": [ + { + "bbox": [ + 105, + 360, + 506, + 373 + ], + "spans": [ + { + "bbox": [ + 105, + 360, + 506, + 373 + ], + "score": 1.0, + "content": "Jesse Davis and Mark Goadrich. The relationship between precision-recall and roc curves. In ICML,", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 114, + 370, + 143, + 384 + ], + "spans": [ + { + "bbox": [ + 114, + 370, + 143, + 384 + ], + "score": 1.0, + "content": "2006.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18.5, + "bbox_fs": [ + 105, + 360, + 506, + 384 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 390, + 504, + 414 + ], + "lines": [ + { + "bbox": [ + 106, + 391, + 505, + 404 + ], + "spans": [ + { + "bbox": [ + 106, + 391, + 505, + 404 + ], + "score": 1.0, + "content": "Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 115, + 402, + 299, + 414 + ], + "spans": [ + { + "bbox": [ + 115, + 402, + 299, + 414 + ], + "score": 1.0, + "content": "hierarchical image database. In CVPR, 2009.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20.5, + "bbox_fs": [ + 106, + 391, + 505, + 414 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 421, + 504, + 444 + ], + "lines": [ + { + "bbox": [ + 106, + 421, + 505, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 421, + 505, + 434 + ], + "score": 1.0, + "content": "Xuefeng Du, Zhaoning Wang, Mu Cai, and Sharon Li. Towards unknown-aware learning with virtual", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 116, + 433, + 253, + 444 + ], + "spans": [ + { + "bbox": [ + 116, + 433, + 253, + 444 + ], + "score": 1.0, + "content": "outlier synthesis. In ICLR, 2022a.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22.5, + "bbox_fs": [ + 106, + 421, + 505, + 444 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 451, + 504, + 475 + ], + "lines": [ + { + "bbox": [ + 106, + 451, + 505, + 465 + ], + "spans": [ + { + "bbox": [ + 106, + 451, + 505, + 465 + ], + "score": 1.0, + "content": "Xuefeng Du, Zhaoning Wang, Mu Cai, and Yixuan Li. VOS: learning what you don’t know by virtual", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 116, + 464, + 255, + 475 + ], + "spans": [ + { + "bbox": [ + 116, + 464, + 255, + 475 + ], + "score": 1.0, + "content": "outlier synthesis. In ICLR, 2022b.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24.5, + "bbox_fs": [ + 106, + 451, + 505, + 475 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 482, + 504, + 505 + ], + "lines": [ + { + "bbox": [ + 106, + 482, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 106, + 482, + 505, + 495 + ], + "score": 1.0, + "content": "John Duchi, Elad Hazan, and Yoram Singer. Adaptive subgradient methods for online learning and", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 116, + 494, + 280, + 505 + ], + "spans": [ + { + "bbox": [ + 116, + 494, + 280, + 505 + ], + "score": 1.0, + "content": "stochastic optimization. In JMLR, 2011.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26.5, + "bbox_fs": [ + 106, + 482, + 505, + 505 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 513, + 444, + 525 + ], + "lines": [ + { + "bbox": [ + 105, + 512, + 443, + 527 + ], + "spans": [ + { + "bbox": [ + 105, + 512, + 443, + 527 + ], + "score": 1.0, + "content": "Tom Fawcett. An introduction to roc analysis. In Pattern Recognition Letters, 2006.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28, + "bbox_fs": [ + 105, + 512, + 443, + 527 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 532, + 505, + 555 + ], + "lines": [ + { + "bbox": [ + 106, + 532, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 532, + 505, + 546 + ], + "score": 1.0, + "content": "Stanislav Fort, Jie Ren, and Balaji Lakshminarayanan. Exploring the limits of out-of-distribution", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 544, + 234, + 555 + ], + "spans": [ + { + "bbox": [ + 116, + 544, + 234, + 555 + ], + "score": 1.0, + "content": "detection. In NeurIPS, 2021.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5, + "bbox_fs": [ + 106, + 532, + 505, + 555 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 563, + 504, + 586 + ], + "lines": [ + { + "bbox": [ + 106, + 563, + 506, + 576 + ], + "spans": [ + { + "bbox": [ + 106, + 563, + 506, + 576 + ], + "score": 1.0, + "content": "Jonathan Frankle and Michael Carbin. The lottery ticket hypothesis: Training pruned neural networks.", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 116, + 574, + 178, + 586 + ], + "spans": [ + { + "bbox": [ + 116, + 574, + 178, + 586 + ], + "score": 1.0, + "content": "In ICLR, 2019.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5, + "bbox_fs": [ + 106, + 563, + 506, + 586 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 593, + 504, + 616 + ], + "lines": [ + { + "bbox": [ + 104, + 592, + 506, + 607 + ], + "spans": [ + { + "bbox": [ + 104, + 592, + 506, + 607 + ], + "score": 1.0, + "content": "Ian Goodfellow, Yoshua Bengio, Aaron Courville, and Yoshua Bengio. Deep learning. MIT Press,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 114, + 603, + 143, + 617 + ], + "spans": [ + { + "bbox": [ + 114, + 603, + 143, + 617 + ], + "score": 1.0, + "content": "2016.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5, + "bbox_fs": [ + 104, + 592, + 506, + 617 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 624, + 506, + 702 + ], + "lines": [ + { + "bbox": [ + 105, + 623, + 506, + 637 + ], + "spans": [ + { + "bbox": [ + 105, + 623, + 506, + 637 + ], + "score": 1.0, + "content": "Ian J. Goodfellow, Dumitru Erhan, Pierre Luc Carrier, Aaron Courville, Mehdi Mirza, Ben Hamner,", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 635, + 506, + 648 + ], + "spans": [ + { + "bbox": [ + 115, + 635, + 506, + 648 + ], + "score": 1.0, + "content": "Will Cukierski, Yichuan Tang, David Thaler, Dong-Hyun Lee, Yingbo Zhou, Chetan Ramaiah,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 646, + 506, + 659 + ], + "spans": [ + { + "bbox": [ + 115, + 646, + 506, + 659 + ], + "score": 1.0, + "content": "Fangxiang Feng, Ruifan Li, Xiaojie Wang, Dimitris Athanasakis, John Shawe-Taylor, Maxim", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 656, + 506, + 671 + ], + "spans": [ + { + "bbox": [ + 115, + 656, + 506, + 671 + ], + "score": 1.0, + "content": "Milakov, John Park, Radu Ionescu, Marius Popescu, Cristian Grozea, James Bergstra, Jingjing Xie,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 668, + 505, + 681 + ], + "spans": [ + { + "bbox": [ + 115, + 668, + 505, + 681 + ], + "score": 1.0, + "content": "Lukasz Romaszko, Bing Xu, Zhang Chuang, and Yoshua Bengio. Challenges in representation", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 114, + 678, + 506, + 693 + ], + "spans": [ + { + "bbox": [ + 114, + 678, + 506, + 693 + ], + "score": 1.0, + "content": "learning: A report on three machine learning contests. In Minho Lee, Akira Hirose, Zeng-Guang", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 690, + 419, + 703 + ], + "spans": [ + { + "bbox": [ + 115, + 690, + 419, + 703 + ], + "score": 1.0, + "content": "Hou, and Rhee Man Kil (eds.), NeurIPS. Springer Berlin Heidelberg, 2013.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 38, + "bbox_fs": [ + 105, + 623, + 506, + 703 + ] + }, + { + "type": "text", + "bbox": [ + 109, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 107, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 107, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "Song Han, Huizi Mao, and William J Dally. Deep compression: Compressing deep neural networks", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 116, + 720, + 403, + 733 + ], + "spans": [ + { + "bbox": [ + 116, + 720, + 403, + 733 + ], + "score": 1.0, + "content": "with pruning, trained quantization and huffman coding. In ICLR, 2016.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42.5, + "bbox_fs": [ + 107, + 709, + 505, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 82, + 504, + 105 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 505, + 95 + ], + "score": 1.0, + "content": "Trevor Hastie, Robert Tibshirani, and Jerome Friedman. The elements of statistical learning: Data", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 443, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 443, + 106 + ], + "score": 1.0, + "content": "mining, inference, and prediction, second edition. Springer New York, NY, 2009.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 106, + 112, + 504, + 136 + ], + "lines": [ + { + "bbox": [ + 106, + 112, + 505, + 125 + ], + "spans": [ + { + "bbox": [ + 106, + 112, + 505, + 125 + ], + "score": 1.0, + "content": "Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 116, + 124, + 303, + 135 + ], + "spans": [ + { + "bbox": [ + 116, + 124, + 303, + 135 + ], + "score": 1.0, + "content": "examples in neural networks. In ICLR, 2017a.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 106, + 142, + 504, + 166 + ], + "lines": [ + { + "bbox": [ + 106, + 142, + 505, + 155 + ], + "spans": [ + { + "bbox": [ + 106, + 142, + 505, + 155 + ], + "score": 1.0, + "content": "Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 155, + 304, + 165 + ], + "spans": [ + { + "bbox": [ + 115, + 155, + 304, + 165 + ], + "score": 1.0, + "content": "examples in neural networks. In ICLR, 2017b.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4.5 + }, + { + "type": "text", + "bbox": [ + 106, + 173, + 504, + 196 + ], + "lines": [ + { + "bbox": [ + 105, + 172, + 506, + 186 + ], + "spans": [ + { + "bbox": [ + 105, + 172, + 506, + 186 + ], + "score": 1.0, + "content": "Dan Hendrycks, Kimin Lee, and Mantas Mazeika. Using pre-training can improve model robustness", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 116, + 185, + 252, + 197 + ], + "spans": [ + { + "bbox": [ + 116, + 185, + 252, + 197 + ], + "score": 1.0, + "content": "and uncertainty. In ICML, 2019a.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 107, + 203, + 504, + 227 + ], + "lines": [ + { + "bbox": [ + 105, + 203, + 505, + 217 + ], + "spans": [ + { + "bbox": [ + 105, + 203, + 505, + 217 + ], + "score": 1.0, + "content": "Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 116, + 216, + 225, + 226 + ], + "spans": [ + { + "bbox": [ + 116, + 216, + 225, + 226 + ], + "score": 1.0, + "content": "exposure. In ICLR, 2019b.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 107, + 234, + 504, + 257 + ], + "lines": [ + { + "bbox": [ + 105, + 233, + 505, + 247 + ], + "spans": [ + { + "bbox": [ + 105, + 233, + 505, + 247 + ], + "score": 1.0, + "content": "Dan Hendrycks, Mantas Mazeika, and Thomas G. Dietterich. Deep anomaly detection with outlier", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 116, + 246, + 224, + 256 + ], + "spans": [ + { + "bbox": [ + 116, + 246, + 224, + 256 + ], + "score": 1.0, + "content": "exposure. In ICLR, 2019c.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5 + }, + { + "type": "text", + "bbox": [ + 107, + 264, + 504, + 288 + ], + "lines": [ + { + "bbox": [ + 106, + 264, + 505, + 277 + ], + "spans": [ + { + "bbox": [ + 106, + 264, + 505, + 277 + ], + "score": 1.0, + "content": "Gao Huang, Zhuang Liu, Laurens Van Der Maaten, and Kilian Q Weinberger. Densely connected", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 116, + 276, + 281, + 287 + ], + "spans": [ + { + "bbox": [ + 116, + 276, + 281, + 287 + ], + "score": 1.0, + "content": "convolutional networks. In CVPR, 2017.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12.5 + }, + { + "type": "text", + "bbox": [ + 106, + 294, + 504, + 318 + ], + "lines": [ + { + "bbox": [ + 106, + 295, + 504, + 308 + ], + "spans": [ + { + "bbox": [ + 106, + 295, + 504, + 308 + ], + "score": 1.0, + "content": "Rui Huang, Andrew Geng, and Yixuan Li. On the importance of gradients for detecting distributional", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 116, + 306, + 264, + 317 + ], + "spans": [ + { + "bbox": [ + 116, + 306, + 264, + 317 + ], + "score": 1.0, + "content": "shifts in the wild. In NeurIPS, 2021.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5 + }, + { + "type": "text", + "bbox": [ + 106, + 325, + 504, + 349 + ], + "lines": [ + { + "bbox": [ + 106, + 324, + 505, + 339 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 505, + 339 + ], + "score": 1.0, + "content": "Takashi Ishida, Ikko Yamane, Tomoya Sakai, Gang Niu, and Masashi Sugiyama. Do we need zero", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 116, + 336, + 375, + 349 + ], + "spans": [ + { + "bbox": [ + 116, + 336, + 375, + 349 + ], + "score": 1.0, + "content": "training loss after achieving zero training error? In ICML, 2020.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5 + }, + { + "type": "text", + "bbox": [ + 106, + 355, + 505, + 379 + ], + "lines": [ + { + "bbox": [ + 105, + 354, + 506, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 354, + 506, + 369 + ], + "score": 1.0, + "content": "J. Kiefer and J. Wolfowitz. Stochastic estimation of the maximum of a regression function. In Ann.", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 366, + 199, + 378 + ], + "spans": [ + { + "bbox": [ + 115, + 366, + 199, + 378 + ], + "score": 1.0, + "content": "Math. Statist., 1952.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18.5 + }, + { + "type": "text", + "bbox": [ + 104, + 385, + 462, + 398 + ], + "lines": [ + { + "bbox": [ + 106, + 385, + 461, + 399 + ], + "spans": [ + { + "bbox": [ + 106, + 385, + 461, + 399 + ], + "score": 1.0, + "content": "Alex Krizhevsky. Learning multiple layers of features from tiny images. In arXiv, 2009.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 106, + 405, + 504, + 429 + ], + "lines": [ + { + "bbox": [ + 106, + 405, + 505, + 418 + ], + "spans": [ + { + "bbox": [ + 106, + 405, + 505, + 418 + ], + "score": 1.0, + "content": "Kimin Lee, Honglak Lee, Kibok Lee, and Jinwoo Shin. Training confidence-calibrated classifiers for", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 117, + 417, + 336, + 429 + ], + "spans": [ + { + "bbox": [ + 117, + 417, + 336, + 429 + ], + "score": 1.0, + "content": "detecting out-of-distribution samples. In ICLR, 2018a.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5 + }, + { + "type": "text", + "bbox": [ + 107, + 435, + 504, + 459 + ], + "lines": [ + { + "bbox": [ + 105, + 434, + 506, + 450 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 450 + ], + "score": 1.0, + "content": "Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. A simple unified framework for detecting", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 116, + 447, + 404, + 459 + ], + "spans": [ + { + "bbox": [ + 116, + 447, + 404, + 459 + ], + "score": 1.0, + "content": "out-of-distribution samples and adversarial attacks. In NeurIPS, 2018b.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 23.5 + }, + { + "type": "text", + "bbox": [ + 107, + 466, + 504, + 489 + ], + "lines": [ + { + "bbox": [ + 105, + 464, + 506, + 480 + ], + "spans": [ + { + "bbox": [ + 105, + 464, + 506, + 480 + ], + "score": 1.0, + "content": "Shiyu Liang, Yixuan Li, and R. Srikant. Enhancing the reliability of out-of-distribution image", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 116, + 477, + 298, + 489 + ], + "spans": [ + { + "bbox": [ + 116, + 477, + 298, + 489 + ], + "score": 1.0, + "content": "detection in neural networks. In ICLR, 2018.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25.5 + }, + { + "type": "text", + "bbox": [ + 107, + 496, + 505, + 520 + ], + "lines": [ + { + "bbox": [ + 106, + 496, + 505, + 509 + ], + "spans": [ + { + "bbox": [ + 106, + 496, + 505, + 509 + ], + "score": 1.0, + "content": "Ziqian Lin, Sreya Dutta Roy, and Yixuan Li. Mood: Multi-level out-of-distribution detection. In", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 508, + 171, + 520 + ], + "spans": [ + { + "bbox": [ + 115, + 508, + 171, + 520 + ], + "score": 1.0, + "content": "CVPR, 2021.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 106, + 526, + 505, + 550 + ], + "lines": [ + { + "bbox": [ + 105, + 526, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 505, + 541 + ], + "score": 1.0, + "content": "Weitang Liu, Xiaoyun Wang, John D Owens, and Yixuan Li. Energy-based out-of-distribution", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 538, + 234, + 550 + ], + "spans": [ + { + "bbox": [ + 116, + 538, + 234, + 550 + ], + "score": 1.0, + "content": "detection. In NeurIPS, 2020.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5 + }, + { + "type": "text", + "bbox": [ + 105, + 557, + 506, + 581 + ], + "lines": [ + { + "bbox": [ + 106, + 557, + 507, + 570 + ], + "spans": [ + { + "bbox": [ + 106, + 557, + 507, + 570 + ], + "score": 1.0, + "content": "Ilya Loshchilov and Frank Hutter. SGDR: stochastic gradient descent with warm restarts. In ICLR,", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 114, + 567, + 143, + 581 + ], + "spans": [ + { + "bbox": [ + 114, + 567, + 143, + 581 + ], + "score": 1.0, + "content": "2017.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 106, + 587, + 506, + 611 + ], + "lines": [ + { + "bbox": [ + 106, + 586, + 506, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 586, + 506, + 601 + ], + "score": 1.0, + "content": "Christopher D. Manning and Hinrich Schütze. Foundations of Statistical Natural Language Process-", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 116, + 599, + 336, + 611 + ], + "spans": [ + { + "bbox": [ + 116, + 599, + 336, + 611 + ], + "score": 1.0, + "content": "ing. The MIT Press, Cambridge, Massachusetts, 1999.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5 + }, + { + "type": "text", + "bbox": [ + 106, + 617, + 505, + 641 + ], + "lines": [ + { + "bbox": [ + 105, + 615, + 507, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 507, + 633 + ], + "score": 1.0, + "content": "Yifei Ming, Ying Fan, and Yixuan Li. Poem: Out-of-distribution detection with posterior sampling.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 628, + 180, + 640 + ], + "spans": [ + { + "bbox": [ + 115, + 628, + 180, + 640 + ], + "score": 1.0, + "content": "In ICML, 2022.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5 + }, + { + "type": "text", + "bbox": [ + 105, + 648, + 505, + 672 + ], + "lines": [ + { + "bbox": [ + 106, + 648, + 506, + 662 + ], + "spans": [ + { + "bbox": [ + 106, + 648, + 506, + 662 + ], + "score": 1.0, + "content": "Sina Mohseni, Mandar Pitale, J. B. S. Yadawa, and Zhangyang Wang. Self-supervised learning for", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 660, + 353, + 672 + ], + "spans": [ + { + "bbox": [ + 115, + 660, + 353, + 672 + ], + "score": 1.0, + "content": "generalizable out-of-distribution detection. In AAAI, 2020.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 37.5 + }, + { + "type": "text", + "bbox": [ + 107, + 678, + 503, + 702 + ], + "lines": [ + { + "bbox": [ + 106, + 677, + 505, + 693 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 505, + 693 + ], + "score": 1.0, + "content": "Yuval Netzer, Tao Wang, Adam Coates, Alessandro Bissacco, Bo Wu, and Andrew Y Ng. Reading", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 116, + 690, + 473, + 702 + ], + "spans": [ + { + "bbox": [ + 116, + 690, + 473, + 702 + ], + "score": 1.0, + "content": "digits in natural images with unsupervised feature learning. In NeurIPS Workshop, 2011.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5 + }, + { + "type": "text", + "bbox": [ + 108, + 709, + 502, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 708, + 504, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 708, + 504, + 723 + ], + "score": 1.0, + "content": "Anh Nguyen, Jason Yosinski, and Jeff Clune. Deep neural networks are easily fooled: High confidence", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 720, + 340, + 733 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 340, + 733 + ], + "score": 1.0, + "content": "predictions for unrecognizable images. In CVPR, 2015.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41.5 + } + ], + "page_idx": 10, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 761 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 765 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 765 + ], + "score": 1.0, + "content": "11", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 82, + 504, + 105 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 505, + 95 + ], + "score": 1.0, + "content": "Trevor Hastie, Robert Tibshirani, and Jerome Friedman. The elements of statistical learning: Data", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 443, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 443, + 106 + ], + "score": 1.0, + "content": "mining, inference, and prediction, second edition. Springer New York, NY, 2009.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5, + "bbox_fs": [ + 105, + 81, + 505, + 106 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 112, + 504, + 136 + ], + "lines": [ + { + "bbox": [ + 106, + 112, + 505, + 125 + ], + "spans": [ + { + "bbox": [ + 106, + 112, + 505, + 125 + ], + "score": 1.0, + "content": "Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 116, + 124, + 303, + 135 + ], + "spans": [ + { + "bbox": [ + 116, + 124, + 303, + 135 + ], + "score": 1.0, + "content": "examples in neural networks. In ICLR, 2017a.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5, + "bbox_fs": [ + 106, + 112, + 505, + 135 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 142, + 504, + 166 + ], + "lines": [ + { + "bbox": [ + 106, + 142, + 505, + 155 + ], + "spans": [ + { + "bbox": [ + 106, + 142, + 505, + 155 + ], + "score": 1.0, + "content": "Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 155, + 304, + 165 + ], + "spans": [ + { + "bbox": [ + 115, + 155, + 304, + 165 + ], + "score": 1.0, + "content": "examples in neural networks. In ICLR, 2017b.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4.5, + "bbox_fs": [ + 106, + 142, + 505, + 165 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 173, + 504, + 196 + ], + "lines": [ + { + "bbox": [ + 105, + 172, + 506, + 186 + ], + "spans": [ + { + "bbox": [ + 105, + 172, + 506, + 186 + ], + "score": 1.0, + "content": "Dan Hendrycks, Kimin Lee, and Mantas Mazeika. Using pre-training can improve model robustness", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 116, + 185, + 252, + 197 + ], + "spans": [ + { + "bbox": [ + 116, + 185, + 252, + 197 + ], + "score": 1.0, + "content": "and uncertainty. In ICML, 2019a.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5, + "bbox_fs": [ + 105, + 172, + 506, + 197 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 203, + 504, + 227 + ], + "lines": [ + { + "bbox": [ + 105, + 203, + 505, + 217 + ], + "spans": [ + { + "bbox": [ + 105, + 203, + 505, + 217 + ], + "score": 1.0, + "content": "Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 116, + 216, + 225, + 226 + ], + "spans": [ + { + "bbox": [ + 116, + 216, + 225, + 226 + ], + "score": 1.0, + "content": "exposure. In ICLR, 2019b.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8.5, + "bbox_fs": [ + 105, + 203, + 505, + 226 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 234, + 504, + 257 + ], + "lines": [ + { + "bbox": [ + 105, + 233, + 505, + 247 + ], + "spans": [ + { + "bbox": [ + 105, + 233, + 505, + 247 + ], + "score": 1.0, + "content": "Dan Hendrycks, Mantas Mazeika, and Thomas G. Dietterich. Deep anomaly detection with outlier", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 116, + 246, + 224, + 256 + ], + "spans": [ + { + "bbox": [ + 116, + 246, + 224, + 256 + ], + "score": 1.0, + "content": "exposure. In ICLR, 2019c.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5, + "bbox_fs": [ + 105, + 233, + 505, + 256 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 264, + 504, + 288 + ], + "lines": [ + { + "bbox": [ + 106, + 264, + 505, + 277 + ], + "spans": [ + { + "bbox": [ + 106, + 264, + 505, + 277 + ], + "score": 1.0, + "content": "Gao Huang, Zhuang Liu, Laurens Van Der Maaten, and Kilian Q Weinberger. Densely connected", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 116, + 276, + 281, + 287 + ], + "spans": [ + { + "bbox": [ + 116, + 276, + 281, + 287 + ], + "score": 1.0, + "content": "convolutional networks. In CVPR, 2017.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12.5, + "bbox_fs": [ + 106, + 264, + 505, + 287 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 294, + 504, + 318 + ], + "lines": [ + { + "bbox": [ + 106, + 295, + 504, + 308 + ], + "spans": [ + { + "bbox": [ + 106, + 295, + 504, + 308 + ], + "score": 1.0, + "content": "Rui Huang, Andrew Geng, and Yixuan Li. On the importance of gradients for detecting distributional", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 116, + 306, + 264, + 317 + ], + "spans": [ + { + "bbox": [ + 116, + 306, + 264, + 317 + ], + "score": 1.0, + "content": "shifts in the wild. In NeurIPS, 2021.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5, + "bbox_fs": [ + 106, + 295, + 504, + 317 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 325, + 504, + 349 + ], + "lines": [ + { + "bbox": [ + 106, + 324, + 505, + 339 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 505, + 339 + ], + "score": 1.0, + "content": "Takashi Ishida, Ikko Yamane, Tomoya Sakai, Gang Niu, and Masashi Sugiyama. Do we need zero", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 116, + 336, + 375, + 349 + ], + "spans": [ + { + "bbox": [ + 116, + 336, + 375, + 349 + ], + "score": 1.0, + "content": "training loss after achieving zero training error? In ICML, 2020.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5, + "bbox_fs": [ + 106, + 324, + 505, + 349 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 355, + 505, + 379 + ], + "lines": [ + { + "bbox": [ + 105, + 354, + 506, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 354, + 506, + 369 + ], + "score": 1.0, + "content": "J. Kiefer and J. Wolfowitz. Stochastic estimation of the maximum of a regression function. In Ann.", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 366, + 199, + 378 + ], + "spans": [ + { + "bbox": [ + 115, + 366, + 199, + 378 + ], + "score": 1.0, + "content": "Math. Statist., 1952.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18.5, + "bbox_fs": [ + 105, + 354, + 506, + 378 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 385, + 462, + 398 + ], + "lines": [ + { + "bbox": [ + 106, + 385, + 461, + 399 + ], + "spans": [ + { + "bbox": [ + 106, + 385, + 461, + 399 + ], + "score": 1.0, + "content": "Alex Krizhevsky. Learning multiple layers of features from tiny images. In arXiv, 2009.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20, + "bbox_fs": [ + 106, + 385, + 461, + 399 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 405, + 504, + 429 + ], + "lines": [ + { + "bbox": [ + 106, + 405, + 505, + 418 + ], + "spans": [ + { + "bbox": [ + 106, + 405, + 505, + 418 + ], + "score": 1.0, + "content": "Kimin Lee, Honglak Lee, Kibok Lee, and Jinwoo Shin. Training confidence-calibrated classifiers for", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 117, + 417, + 336, + 429 + ], + "spans": [ + { + "bbox": [ + 117, + 417, + 336, + 429 + ], + "score": 1.0, + "content": "detecting out-of-distribution samples. In ICLR, 2018a.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5, + "bbox_fs": [ + 106, + 405, + 505, + 429 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 435, + 504, + 459 + ], + "lines": [ + { + "bbox": [ + 105, + 434, + 506, + 450 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 450 + ], + "score": 1.0, + "content": "Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. A simple unified framework for detecting", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 116, + 447, + 404, + 459 + ], + "spans": [ + { + "bbox": [ + 116, + 447, + 404, + 459 + ], + "score": 1.0, + "content": "out-of-distribution samples and adversarial attacks. In NeurIPS, 2018b.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 23.5, + "bbox_fs": [ + 105, + 434, + 506, + 459 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 466, + 504, + 489 + ], + "lines": [ + { + "bbox": [ + 105, + 464, + 506, + 480 + ], + "spans": [ + { + "bbox": [ + 105, + 464, + 506, + 480 + ], + "score": 1.0, + "content": "Shiyu Liang, Yixuan Li, and R. Srikant. Enhancing the reliability of out-of-distribution image", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 116, + 477, + 298, + 489 + ], + "spans": [ + { + "bbox": [ + 116, + 477, + 298, + 489 + ], + "score": 1.0, + "content": "detection in neural networks. In ICLR, 2018.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25.5, + "bbox_fs": [ + 105, + 464, + 506, + 489 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 496, + 505, + 520 + ], + "lines": [ + { + "bbox": [ + 106, + 496, + 505, + 509 + ], + "spans": [ + { + "bbox": [ + 106, + 496, + 505, + 509 + ], + "score": 1.0, + "content": "Ziqian Lin, Sreya Dutta Roy, and Yixuan Li. Mood: Multi-level out-of-distribution detection. In", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 508, + 171, + 520 + ], + "spans": [ + { + "bbox": [ + 115, + 508, + 171, + 520 + ], + "score": 1.0, + "content": "CVPR, 2021.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27.5, + "bbox_fs": [ + 106, + 496, + 505, + 520 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 526, + 505, + 550 + ], + "lines": [ + { + "bbox": [ + 105, + 526, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 505, + 541 + ], + "score": 1.0, + "content": "Weitang Liu, Xiaoyun Wang, John D Owens, and Yixuan Li. Energy-based out-of-distribution", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 538, + 234, + 550 + ], + "spans": [ + { + "bbox": [ + 116, + 538, + 234, + 550 + ], + "score": 1.0, + "content": "detection. In NeurIPS, 2020.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5, + "bbox_fs": [ + 105, + 526, + 505, + 550 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 557, + 506, + 581 + ], + "lines": [ + { + "bbox": [ + 106, + 557, + 507, + 570 + ], + "spans": [ + { + "bbox": [ + 106, + 557, + 507, + 570 + ], + "score": 1.0, + "content": "Ilya Loshchilov and Frank Hutter. SGDR: stochastic gradient descent with warm restarts. In ICLR,", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 114, + 567, + 143, + 581 + ], + "spans": [ + { + "bbox": [ + 114, + 567, + 143, + 581 + ], + "score": 1.0, + "content": "2017.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5, + "bbox_fs": [ + 106, + 557, + 507, + 581 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 587, + 506, + 611 + ], + "lines": [ + { + "bbox": [ + 106, + 586, + 506, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 586, + 506, + 601 + ], + "score": 1.0, + "content": "Christopher D. Manning and Hinrich Schütze. Foundations of Statistical Natural Language Process-", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 116, + 599, + 336, + 611 + ], + "spans": [ + { + "bbox": [ + 116, + 599, + 336, + 611 + ], + "score": 1.0, + "content": "ing. The MIT Press, Cambridge, Massachusetts, 1999.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5, + "bbox_fs": [ + 106, + 586, + 506, + 611 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 617, + 505, + 641 + ], + "lines": [ + { + "bbox": [ + 105, + 615, + 507, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 507, + 633 + ], + "score": 1.0, + "content": "Yifei Ming, Ying Fan, and Yixuan Li. Poem: Out-of-distribution detection with posterior sampling.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 628, + 180, + 640 + ], + "spans": [ + { + "bbox": [ + 115, + 628, + 180, + 640 + ], + "score": 1.0, + "content": "In ICML, 2022.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5, + "bbox_fs": [ + 105, + 615, + 507, + 640 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 648, + 505, + 672 + ], + "lines": [ + { + "bbox": [ + 106, + 648, + 506, + 662 + ], + "spans": [ + { + "bbox": [ + 106, + 648, + 506, + 662 + ], + "score": 1.0, + "content": "Sina Mohseni, Mandar Pitale, J. B. S. Yadawa, and Zhangyang Wang. Self-supervised learning for", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 660, + 353, + 672 + ], + "spans": [ + { + "bbox": [ + 115, + 660, + 353, + 672 + ], + "score": 1.0, + "content": "generalizable out-of-distribution detection. In AAAI, 2020.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 37.5, + "bbox_fs": [ + 106, + 648, + 506, + 672 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 678, + 503, + 702 + ], + "lines": [ + { + "bbox": [ + 106, + 677, + 505, + 693 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 505, + 693 + ], + "score": 1.0, + "content": "Yuval Netzer, Tao Wang, Adam Coates, Alessandro Bissacco, Bo Wu, and Andrew Y Ng. Reading", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 116, + 690, + 473, + 702 + ], + "spans": [ + { + "bbox": [ + 116, + 690, + 473, + 702 + ], + "score": 1.0, + "content": "digits in natural images with unsupervised feature learning. In NeurIPS Workshop, 2011.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5, + "bbox_fs": [ + 106, + 677, + 505, + 702 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 709, + 502, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 708, + 504, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 708, + 504, + 723 + ], + "score": 1.0, + "content": "Anh Nguyen, Jason Yosinski, and Jeff Clune. Deep neural networks are easily fooled: High confidence", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 720, + 340, + 733 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 340, + 733 + ], + "score": 1.0, + "content": "predictions for unrecognizable images. In CVPR, 2015.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41.5, + "bbox_fs": [ + 106, + 708, + 504, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 82, + 505, + 105 + ], + "lines": [ + { + "bbox": [ + 106, + 81, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 506, + 95 + ], + "score": 1.0, + "content": "Vivek Ramanujan, Mitchell Wortsman, Aniruddha Kembhavi, Ali Farhadi, and Mohammad Rastegari.", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 408, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 408, + 106 + ], + "score": 1.0, + "content": "What’s hidden in a randomly weighted neural network? In CVPR, 2020.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 109, + 111, + 504, + 146 + ], + "lines": [ + { + "bbox": [ + 106, + 111, + 506, + 125 + ], + "spans": [ + { + "bbox": [ + 106, + 111, + 506, + 125 + ], + "score": 1.0, + "content": "Mohammadreza Salehi, Hossein Mirzaei, Dan Hendrycks, Yixuan Li, Mohammad Hossein Rohban,", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 116, + 123, + 505, + 136 + ], + "spans": [ + { + "bbox": [ + 116, + 123, + 505, + 136 + ], + "score": 1.0, + "content": "and Mohammad Sabokrou. A unified survey on anomaly, novelty, open-set, and out-of-distribution", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 116, + 133, + 353, + 146 + ], + "spans": [ + { + "bbox": [ + 116, + 133, + 353, + 146 + ], + "score": 1.0, + "content": "detection: Solutions and future challenges. In arXiv, 2021.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 105, + 153, + 504, + 176 + ], + "lines": [ + { + "bbox": [ + 106, + 153, + 505, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 153, + 505, + 165 + ], + "score": 1.0, + "content": "Vikash Sehwag, Mung Chiang, and Prateek Mittal. SSD: A unified framework for self-supervised", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 164, + 250, + 175 + ], + "spans": [ + { + "bbox": [ + 115, + 164, + 250, + 175 + ], + "score": 1.0, + "content": "outlier detection. In ICLR, 2021.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + }, + { + "type": "text", + "bbox": [ + 106, + 182, + 505, + 205 + ], + "lines": [ + { + "bbox": [ + 105, + 182, + 505, + 196 + ], + "spans": [ + { + "bbox": [ + 105, + 182, + 505, + 196 + ], + "score": 1.0, + "content": "Jonathan Sorg, Richard L Lewis, and Satinder Singh. Reward design via online gradient ascent. In", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 115, + 194, + 180, + 205 + ], + "spans": [ + { + "bbox": [ + 115, + 194, + 180, + 205 + ], + "score": 1.0, + "content": "NeurIPS, 2010.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 106, + 212, + 505, + 236 + ], + "lines": [ + { + "bbox": [ + 106, + 213, + 506, + 225 + ], + "spans": [ + { + "bbox": [ + 106, + 213, + 506, + 225 + ], + "score": 1.0, + "content": "Nitish Srivastava, Geoffrey E. Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 116, + 223, + 506, + 236 + ], + "spans": [ + { + "bbox": [ + 116, + 223, + 506, + 236 + ], + "score": 1.0, + "content": "Dropout: a simple way to prevent neural networks from overfitting. In J. Mach. Learn. Res., 2014.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "text", + "bbox": [ + 107, + 243, + 505, + 266 + ], + "lines": [ + { + "bbox": [ + 106, + 241, + 507, + 256 + ], + "spans": [ + { + "bbox": [ + 106, + 241, + 507, + 256 + ], + "score": 1.0, + "content": "Yiyou Sun and Yixuan Li. Dice: Leveraging sparsification for out-of-distribution detection. In ECCV,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 114, + 252, + 143, + 267 + ], + "spans": [ + { + "bbox": [ + 114, + 252, + 143, + 267 + ], + "score": 1.0, + "content": "2022.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5 + }, + { + "type": "text", + "bbox": [ + 107, + 272, + 505, + 295 + ], + "lines": [ + { + "bbox": [ + 106, + 272, + 506, + 285 + ], + "spans": [ + { + "bbox": [ + 106, + 272, + 506, + 285 + ], + "score": 1.0, + "content": "Yiyou Sun, Chuan Guo, and Yixuan Li. React: Out-of-distribution detection with rectified activations.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 116, + 283, + 192, + 295 + ], + "spans": [ + { + "bbox": [ + 116, + 283, + 192, + 295 + ], + "score": 1.0, + "content": "In NeurIPS, 2021.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5 + }, + { + "type": "text", + "bbox": [ + 108, + 302, + 504, + 325 + ], + "lines": [ + { + "bbox": [ + 106, + 301, + 506, + 316 + ], + "spans": [ + { + "bbox": [ + 106, + 301, + 506, + 316 + ], + "score": 1.0, + "content": "Yiyou Sun, Yifei Ming, Xiaojin Zhu, and Yixuan Li. Out-of-distribution detection with deep nearest", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 314, + 226, + 326 + ], + "spans": [ + { + "bbox": [ + 115, + 314, + 226, + 326 + ], + "score": 1.0, + "content": "neighbors. In ICML, 2022.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5 + }, + { + "type": "text", + "bbox": [ + 106, + 332, + 504, + 355 + ], + "lines": [ + { + "bbox": [ + 105, + 332, + 505, + 345 + ], + "spans": [ + { + "bbox": [ + 105, + 332, + 505, + 345 + ], + "score": 1.0, + "content": "Jihoon Tack, Sangwoo Mo, Jongheon Jeong, and Jinwoo Shin. CSI: novelty detection via contrastive", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 116, + 344, + 374, + 355 + ], + "spans": [ + { + "bbox": [ + 116, + 344, + 374, + 355 + ], + "score": 1.0, + "content": "learning on distributionally shifted instances. In NeurIPS, 2020.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5 + }, + { + "type": "text", + "bbox": [ + 108, + 362, + 504, + 385 + ], + "lines": [ + { + "bbox": [ + 106, + 362, + 505, + 375 + ], + "spans": [ + { + "bbox": [ + 106, + 362, + 505, + 375 + ], + "score": 1.0, + "content": "Amirhossein Tavanaei. Embedded encoder-decoder in convolutional networks towards explainable", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 116, + 373, + 195, + 385 + ], + "spans": [ + { + "bbox": [ + 116, + 373, + 195, + 385 + ], + "score": 1.0, + "content": "AI. In arXiv, 2020.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5 + }, + { + "type": "text", + "bbox": [ + 107, + 392, + 503, + 415 + ], + "lines": [ + { + "bbox": [ + 106, + 393, + 505, + 404 + ], + "spans": [ + { + "bbox": [ + 106, + 393, + 505, + 404 + ], + "score": 1.0, + "content": "William R Thompson. On the likelihood that one unknown probability exceeds another in view of", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 117, + 403, + 316, + 415 + ], + "spans": [ + { + "bbox": [ + 117, + 403, + 316, + 415 + ], + "score": 1.0, + "content": "the evidence of two sample. In Biometrika, 1933.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5 + }, + { + "type": "text", + "bbox": [ + 107, + 421, + 504, + 456 + ], + "lines": [ + { + "bbox": [ + 105, + 421, + 506, + 435 + ], + "spans": [ + { + "bbox": [ + 105, + 421, + 506, + 435 + ], + "score": 1.0, + "content": "Grant Van Horn, Oisin Mac Aodha, Yang Song, Yin Cui, Chen Sun, Alex Shepard, Hartwig Adam,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 116, + 434, + 505, + 445 + ], + "spans": [ + { + "bbox": [ + 116, + 434, + 505, + 445 + ], + "score": 1.0, + "content": "Pietro Perona, and Serge Belongie. The inaturalist species classification and detection dataset. In", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 116, + 443, + 172, + 456 + ], + "spans": [ + { + "bbox": [ + 116, + 443, + 172, + 456 + ], + "score": 1.0, + "content": "CVPR, 2018.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 106, + 462, + 503, + 486 + ], + "lines": [ + { + "bbox": [ + 105, + 462, + 505, + 476 + ], + "spans": [ + { + "bbox": [ + 105, + 462, + 505, + 476 + ], + "score": 1.0, + "content": "Hongxin Wei, Renchunzi Xie, Hao Cheng, Lei Feng, Bo An, and Yixuan Li. Mitigating neural", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 115, + 474, + 381, + 486 + ], + "spans": [ + { + "bbox": [ + 115, + 474, + 381, + 486 + ], + "score": 1.0, + "content": "network overconfidence with logit normalization. In ICML, 2022.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26.5 + }, + { + "type": "text", + "bbox": [ + 106, + 492, + 504, + 527 + ], + "lines": [ + { + "bbox": [ + 106, + 492, + 506, + 506 + ], + "spans": [ + { + "bbox": [ + 106, + 492, + 506, + 506 + ], + "score": 1.0, + "content": "Jianxiong Xiao, James Hays, Krista A. Ehinger, Aude Oliva, and Antonio Torralba. Sun database:", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 116, + 505, + 505, + 517 + ], + "spans": [ + { + "bbox": [ + 116, + 505, + 505, + 517 + ], + "score": 1.0, + "content": "Large-scale scene recognition from abbey to zoo. In IEEE Computer Society Conference on", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 515, + 313, + 528 + ], + "spans": [ + { + "bbox": [ + 116, + 515, + 313, + 528 + ], + "score": 1.0, + "content": "Computer Vision and Pattern Recognition, 2010.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 106, + 533, + 505, + 556 + ], + "lines": [ + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "score": 1.0, + "content": "Jingkang Yang, Kaiyang Zhou, Yixuan Li, and Ziwei Liu. Generalized out-of-distribution detection:", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 544, + 221, + 557 + ], + "spans": [ + { + "bbox": [ + 115, + 544, + 221, + 557 + ], + "score": 1.0, + "content": "A survey. In arXiv, 2021.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 107, + 563, + 505, + 597 + ], + "lines": [ + { + "bbox": [ + 105, + 562, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 562, + 506, + 577 + ], + "score": 1.0, + "content": "Fisher Yu, Ari Seff, Yinda Zhang, Shuran Song, Thomas Funkhouser, and Jianxiong Xiao. Lsun:", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 574, + 506, + 587 + ], + "spans": [ + { + "bbox": [ + 115, + 574, + 506, + 587 + ], + "score": 1.0, + "content": "Construction of a large-scale image dataset using deep learning with humans in the loop. In arXiv,", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 115, + 585, + 142, + 597 + ], + "spans": [ + { + "bbox": [ + 115, + 585, + 142, + 597 + ], + "score": 1.0, + "content": "2015.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 34 + }, + { + "type": "text", + "bbox": [ + 106, + 604, + 504, + 627 + ], + "lines": [ + { + "bbox": [ + 106, + 604, + 505, + 617 + ], + "spans": [ + { + "bbox": [ + 106, + 604, + 505, + 617 + ], + "score": 1.0, + "content": "Qing Yu and Kiyoharu Aizawa. Unsupervised out-of-distribution detection by maximum classifier", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 615, + 233, + 628 + ], + "spans": [ + { + "bbox": [ + 115, + 615, + 233, + 628 + ], + "score": 1.0, + "content": "discrepancy. In ICCV, 2019.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5 + }, + { + "type": "text", + "bbox": [ + 106, + 633, + 445, + 646 + ], + "lines": [ + { + "bbox": [ + 105, + 633, + 446, + 647 + ], + "spans": [ + { + "bbox": [ + 105, + 633, + 446, + 647 + ], + "score": 1.0, + "content": "Sergey Zagoruyko and Nikos Komodakis. Wide residual networks. In BMVC, 2016.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 107, + 653, + 505, + 676 + ], + "lines": [ + { + "bbox": [ + 106, + 653, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 653, + 505, + 666 + ], + "score": 1.0, + "content": "Bolei Zhou, Agata Lapedriza, Antonio Torralba, and Aude Oliva. Places: An image database for", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 116, + 664, + 334, + 676 + ], + "spans": [ + { + "bbox": [ + 116, + 664, + 334, + 676 + ], + "score": 1.0, + "content": "deep scene understanding. In Journal of Vision, 2017.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5 + } + ], + "page_idx": 11, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "12", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 82, + 505, + 105 + ], + "lines": [ + { + "bbox": [ + 106, + 81, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 506, + 95 + ], + "score": 1.0, + "content": "Vivek Ramanujan, Mitchell Wortsman, Aniruddha Kembhavi, Ali Farhadi, and Mohammad Rastegari.", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 408, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 408, + 106 + ], + "score": 1.0, + "content": "What’s hidden in a randomly weighted neural network? In CVPR, 2020.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5, + "bbox_fs": [ + 106, + 81, + 506, + 106 + ] + }, + { + "type": "text", + "bbox": [ + 109, + 111, + 504, + 146 + ], + "lines": [ + { + "bbox": [ + 106, + 111, + 506, + 125 + ], + "spans": [ + { + "bbox": [ + 106, + 111, + 506, + 125 + ], + "score": 1.0, + "content": "Mohammadreza Salehi, Hossein Mirzaei, Dan Hendrycks, Yixuan Li, Mohammad Hossein Rohban,", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 116, + 123, + 505, + 136 + ], + "spans": [ + { + "bbox": [ + 116, + 123, + 505, + 136 + ], + "score": 1.0, + "content": "and Mohammad Sabokrou. A unified survey on anomaly, novelty, open-set, and out-of-distribution", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 116, + 133, + 353, + 146 + ], + "spans": [ + { + "bbox": [ + 116, + 133, + 353, + 146 + ], + "score": 1.0, + "content": "detection: Solutions and future challenges. In arXiv, 2021.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3, + "bbox_fs": [ + 106, + 111, + 506, + 146 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 153, + 504, + 176 + ], + "lines": [ + { + "bbox": [ + 106, + 153, + 505, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 153, + 505, + 165 + ], + "score": 1.0, + "content": "Vikash Sehwag, Mung Chiang, and Prateek Mittal. SSD: A unified framework for self-supervised", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 164, + 250, + 175 + ], + "spans": [ + { + "bbox": [ + 115, + 164, + 250, + 175 + ], + "score": 1.0, + "content": "outlier detection. In ICLR, 2021.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5, + "bbox_fs": [ + 106, + 153, + 505, + 175 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 182, + 505, + 205 + ], + "lines": [ + { + "bbox": [ + 105, + 182, + 505, + 196 + ], + "spans": [ + { + "bbox": [ + 105, + 182, + 505, + 196 + ], + "score": 1.0, + "content": "Jonathan Sorg, Richard L Lewis, and Satinder Singh. Reward design via online gradient ascent. In", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 115, + 194, + 180, + 205 + ], + "spans": [ + { + "bbox": [ + 115, + 194, + 180, + 205 + ], + "score": 1.0, + "content": "NeurIPS, 2010.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5, + "bbox_fs": [ + 105, + 182, + 505, + 205 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 212, + 505, + 236 + ], + "lines": [ + { + "bbox": [ + 106, + 213, + 506, + 225 + ], + "spans": [ + { + "bbox": [ + 106, + 213, + 506, + 225 + ], + "score": 1.0, + "content": "Nitish Srivastava, Geoffrey E. Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 116, + 223, + 506, + 236 + ], + "spans": [ + { + "bbox": [ + 116, + 223, + 506, + 236 + ], + "score": 1.0, + "content": "Dropout: a simple way to prevent neural networks from overfitting. In J. Mach. Learn. Res., 2014.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5, + "bbox_fs": [ + 106, + 213, + 506, + 236 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 243, + 505, + 266 + ], + "lines": [ + { + "bbox": [ + 106, + 241, + 507, + 256 + ], + "spans": [ + { + "bbox": [ + 106, + 241, + 507, + 256 + ], + "score": 1.0, + "content": "Yiyou Sun and Yixuan Li. Dice: Leveraging sparsification for out-of-distribution detection. In ECCV,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 114, + 252, + 143, + 267 + ], + "spans": [ + { + "bbox": [ + 114, + 252, + 143, + 267 + ], + "score": 1.0, + "content": "2022.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5, + "bbox_fs": [ + 106, + 241, + 507, + 267 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 272, + 505, + 295 + ], + "lines": [ + { + "bbox": [ + 106, + 272, + 506, + 285 + ], + "spans": [ + { + "bbox": [ + 106, + 272, + 506, + 285 + ], + "score": 1.0, + "content": "Yiyou Sun, Chuan Guo, and Yixuan Li. React: Out-of-distribution detection with rectified activations.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 116, + 283, + 192, + 295 + ], + "spans": [ + { + "bbox": [ + 116, + 283, + 192, + 295 + ], + "score": 1.0, + "content": "In NeurIPS, 2021.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5, + "bbox_fs": [ + 106, + 272, + 506, + 295 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 302, + 504, + 325 + ], + "lines": [ + { + "bbox": [ + 106, + 301, + 506, + 316 + ], + "spans": [ + { + "bbox": [ + 106, + 301, + 506, + 316 + ], + "score": 1.0, + "content": "Yiyou Sun, Yifei Ming, Xiaojin Zhu, and Yixuan Li. Out-of-distribution detection with deep nearest", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 314, + 226, + 326 + ], + "spans": [ + { + "bbox": [ + 115, + 314, + 226, + 326 + ], + "score": 1.0, + "content": "neighbors. In ICML, 2022.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5, + "bbox_fs": [ + 106, + 301, + 506, + 326 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 332, + 504, + 355 + ], + "lines": [ + { + "bbox": [ + 105, + 332, + 505, + 345 + ], + "spans": [ + { + "bbox": [ + 105, + 332, + 505, + 345 + ], + "score": 1.0, + "content": "Jihoon Tack, Sangwoo Mo, Jongheon Jeong, and Jinwoo Shin. CSI: novelty detection via contrastive", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 116, + 344, + 374, + 355 + ], + "spans": [ + { + "bbox": [ + 116, + 344, + 374, + 355 + ], + "score": 1.0, + "content": "learning on distributionally shifted instances. In NeurIPS, 2020.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5, + "bbox_fs": [ + 105, + 332, + 505, + 355 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 362, + 504, + 385 + ], + "lines": [ + { + "bbox": [ + 106, + 362, + 505, + 375 + ], + "spans": [ + { + "bbox": [ + 106, + 362, + 505, + 375 + ], + "score": 1.0, + "content": "Amirhossein Tavanaei. Embedded encoder-decoder in convolutional networks towards explainable", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 116, + 373, + 195, + 385 + ], + "spans": [ + { + "bbox": [ + 116, + 373, + 195, + 385 + ], + "score": 1.0, + "content": "AI. In arXiv, 2020.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5, + "bbox_fs": [ + 106, + 362, + 505, + 385 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 392, + 503, + 415 + ], + "lines": [ + { + "bbox": [ + 106, + 393, + 505, + 404 + ], + "spans": [ + { + "bbox": [ + 106, + 393, + 505, + 404 + ], + "score": 1.0, + "content": "William R Thompson. On the likelihood that one unknown probability exceeds another in view of", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 117, + 403, + 316, + 415 + ], + "spans": [ + { + "bbox": [ + 117, + 403, + 316, + 415 + ], + "score": 1.0, + "content": "the evidence of two sample. In Biometrika, 1933.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5, + "bbox_fs": [ + 106, + 393, + 505, + 415 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 421, + 504, + 456 + ], + "lines": [ + { + "bbox": [ + 105, + 421, + 506, + 435 + ], + "spans": [ + { + "bbox": [ + 105, + 421, + 506, + 435 + ], + "score": 1.0, + "content": "Grant Van Horn, Oisin Mac Aodha, Yang Song, Yin Cui, Chen Sun, Alex Shepard, Hartwig Adam,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 116, + 434, + 505, + 445 + ], + "spans": [ + { + "bbox": [ + 116, + 434, + 505, + 445 + ], + "score": 1.0, + "content": "Pietro Perona, and Serge Belongie. The inaturalist species classification and detection dataset. In", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 116, + 443, + 172, + 456 + ], + "spans": [ + { + "bbox": [ + 116, + 443, + 172, + 456 + ], + "score": 1.0, + "content": "CVPR, 2018.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24, + "bbox_fs": [ + 105, + 421, + 506, + 456 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 462, + 503, + 486 + ], + "lines": [ + { + "bbox": [ + 105, + 462, + 505, + 476 + ], + "spans": [ + { + "bbox": [ + 105, + 462, + 505, + 476 + ], + "score": 1.0, + "content": "Hongxin Wei, Renchunzi Xie, Hao Cheng, Lei Feng, Bo An, and Yixuan Li. Mitigating neural", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 115, + 474, + 381, + 486 + ], + "spans": [ + { + "bbox": [ + 115, + 474, + 381, + 486 + ], + "score": 1.0, + "content": "network overconfidence with logit normalization. In ICML, 2022.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26.5, + "bbox_fs": [ + 105, + 462, + 505, + 486 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 492, + 504, + 527 + ], + "lines": [ + { + "bbox": [ + 106, + 492, + 506, + 506 + ], + "spans": [ + { + "bbox": [ + 106, + 492, + 506, + 506 + ], + "score": 1.0, + "content": "Jianxiong Xiao, James Hays, Krista A. Ehinger, Aude Oliva, and Antonio Torralba. Sun database:", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 116, + 505, + 505, + 517 + ], + "spans": [ + { + "bbox": [ + 116, + 505, + 505, + 517 + ], + "score": 1.0, + "content": "Large-scale scene recognition from abbey to zoo. In IEEE Computer Society Conference on", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 515, + 313, + 528 + ], + "spans": [ + { + "bbox": [ + 116, + 515, + 313, + 528 + ], + "score": 1.0, + "content": "Computer Vision and Pattern Recognition, 2010.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29, + "bbox_fs": [ + 106, + 492, + 506, + 528 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 533, + 505, + 556 + ], + "lines": [ + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "score": 1.0, + "content": "Jingkang Yang, Kaiyang Zhou, Yixuan Li, and Ziwei Liu. Generalized out-of-distribution detection:", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 544, + 221, + 557 + ], + "spans": [ + { + "bbox": [ + 115, + 544, + 221, + 557 + ], + "score": 1.0, + "content": "A survey. In arXiv, 2021.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5, + "bbox_fs": [ + 106, + 533, + 506, + 557 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 563, + 505, + 597 + ], + "lines": [ + { + "bbox": [ + 105, + 562, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 562, + 506, + 577 + ], + "score": 1.0, + "content": "Fisher Yu, Ari Seff, Yinda Zhang, Shuran Song, Thomas Funkhouser, and Jianxiong Xiao. Lsun:", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 574, + 506, + 587 + ], + "spans": [ + { + "bbox": [ + 115, + 574, + 506, + 587 + ], + "score": 1.0, + "content": "Construction of a large-scale image dataset using deep learning with humans in the loop. In arXiv,", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 115, + 585, + 142, + 597 + ], + "spans": [ + { + "bbox": [ + 115, + 585, + 142, + 597 + ], + "score": 1.0, + "content": "2015.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 34, + "bbox_fs": [ + 105, + 562, + 506, + 597 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 604, + 504, + 627 + ], + "lines": [ + { + "bbox": [ + 106, + 604, + 505, + 617 + ], + "spans": [ + { + "bbox": [ + 106, + 604, + 505, + 617 + ], + "score": 1.0, + "content": "Qing Yu and Kiyoharu Aizawa. Unsupervised out-of-distribution detection by maximum classifier", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 615, + 233, + 628 + ], + "spans": [ + { + "bbox": [ + 115, + 615, + 233, + 628 + ], + "score": 1.0, + "content": "discrepancy. In ICCV, 2019.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5, + "bbox_fs": [ + 106, + 604, + 505, + 628 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 633, + 445, + 646 + ], + "lines": [ + { + "bbox": [ + 105, + 633, + 446, + 647 + ], + "spans": [ + { + "bbox": [ + 105, + 633, + 446, + 647 + ], + "score": 1.0, + "content": "Sergey Zagoruyko and Nikos Komodakis. Wide residual networks. In BMVC, 2016.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38, + "bbox_fs": [ + 105, + 633, + 446, + 647 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 653, + 505, + 676 + ], + "lines": [ + { + "bbox": [ + 106, + 653, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 653, + 505, + 666 + ], + "score": 1.0, + "content": "Bolei Zhou, Agata Lapedriza, Antonio Torralba, and Aude Oliva. Places: An image database for", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 116, + 664, + 334, + 676 + ], + "spans": [ + { + "bbox": [ + 116, + 664, + 334, + 676 + ], + "score": 1.0, + "content": "deep scene understanding. In Journal of Vision, 2017.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5, + "bbox_fs": [ + 106, + 653, + 505, + 676 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 82, + 161, + 93 + ], + "lines": [ + { + "bbox": [ + 106, + 81, + 163, + 96 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 163, + 96 + ], + "score": 1.0, + "content": "APPENDIX", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "title", + "bbox": [ + 108, + 107, + 268, + 120 + ], + "lines": [ + { + "bbox": [ + 106, + 108, + 269, + 121 + ], + "spans": [ + { + "bbox": [ + 106, + 108, + 269, + 121 + ], + "score": 1.0, + "content": "REPRODUCIBILITY STATEMENT", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 107, + 132, + 505, + 165 + ], + "lines": [ + { + "bbox": [ + 105, + 132, + 505, + 145 + ], + "spans": [ + { + "bbox": [ + 105, + 132, + 505, + 145 + ], + "score": 1.0, + "content": "We will provide the anonymous repository about our source codes in the discussion phase for", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 143, + 505, + 156 + ], + "spans": [ + { + "bbox": [ + 105, + 143, + 505, + 156 + ], + "score": 1.0, + "content": "reviewing purposes to ensure the reproducibility of our experimental results. Below we summarize", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 154, + 338, + 166 + ], + "spans": [ + { + "bbox": [ + 105, + 154, + 338, + 166 + ], + "score": 1.0, + "content": "some critical aspects to facilitate the reproducible results:", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 133, + 174, + 505, + 307 + ], + "lines": [ + { + "bbox": [ + 132, + 174, + 507, + 188 + ], + "spans": [ + { + "bbox": [ + 132, + 174, + 507, + 188 + ], + "score": 1.0, + "content": "• Datasets. The datasets we used are all publicly accessible, which is introduced in Section 4.1.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 141, + 186, + 506, + 198 + ], + "spans": [ + { + "bbox": [ + 141, + 186, + 506, + 198 + ], + "score": 1.0, + "content": "For methods involving auxiliary outliers, we strictly follow previous works (Sun et al., 2021;", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 141, + 196, + 506, + 209 + ], + "spans": [ + { + "bbox": [ + 141, + 196, + 506, + 209 + ], + "score": 1.0, + "content": "Du et al., 2022b) to avoid overlap between the auxiliary dataset (ImageNet-1k) (Deng et al.,", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 141, + 207, + 285, + 221 + ], + "spans": [ + { + "bbox": [ + 141, + 207, + 285, + 221 + ], + "score": 1.0, + "content": "2009) and any other OOD datasets.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 135, + 222, + 506, + 235 + ], + "spans": [ + { + "bbox": [ + 135, + 222, + 506, + 235 + ], + "score": 1.0, + "content": "• Assumption. We set our experiments to a post-hoc scenario where a well-trained model is", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 140, + 232, + 505, + 247 + ], + "spans": [ + { + "bbox": [ + 140, + 232, + 505, + 247 + ], + "score": 1.0, + "content": "available, and some parts of training samples are also available for subsequent fine-tuning.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 135, + 248, + 505, + 261 + ], + "spans": [ + { + "bbox": [ + 135, + 248, + 505, + 261 + ], + "score": 1.0, + "content": "• Open source. The code repository will be available in an anonymous repository for", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 258, + 505, + 272 + ], + "spans": [ + { + "bbox": [ + 141, + 258, + 505, + 272 + ], + "score": 1.0, + "content": "reviewing purposes. We provide a backbone of our experiments as well as several auxiliary", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 142, + 271, + 296, + 282 + ], + "spans": [ + { + "bbox": [ + 142, + 271, + 296, + 282 + ], + "score": 1.0, + "content": "components, such as score estimation.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 132, + 283, + 506, + 297 + ], + "spans": [ + { + "bbox": [ + 132, + 283, + 506, + 297 + ], + "score": 1.0, + "content": "• Environment. All experiments are conducted multiple runs on NVIDIA Tesla V100-SXM2-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 295, + 331, + 307 + ], + "spans": [ + { + "bbox": [ + 141, + 295, + 331, + 307 + ], + "score": 1.0, + "content": "32GB GPUs with Python 3.6 and PyTorch 1.8.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 10 + }, + { + "type": "title", + "bbox": [ + 108, + 323, + 482, + 335 + ], + "lines": [ + { + "bbox": [ + 105, + 323, + 483, + 337 + ], + "spans": [ + { + "bbox": [ + 105, + 323, + 483, + 337 + ], + "score": 1.0, + "content": "A DETAILED INFORMATION ABOUT THE USED BASELINES AND METRICS", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 106, + 346, + 505, + 370 + ], + "lines": [ + { + "bbox": [ + 105, + 344, + 505, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 505, + 361 + ], + "score": 1.0, + "content": "In this section, we provide the details about the baselines for the scoring functions and finetuning", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 358, + 490, + 370 + ], + "spans": [ + { + "bbox": [ + 106, + 358, + 490, + 370 + ], + "score": 1.0, + "content": "with auxiliary outliers, and the corresponding hyper-parameters that are considered in our work.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5 + }, + { + "type": "text", + "bbox": [ + 106, + 380, + 504, + 403 + ], + "lines": [ + { + "bbox": [ + 105, + 381, + 505, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 381, + 505, + 394 + ], + "score": 1.0, + "content": "Maximum Softmax Probability (MSP). (Hendrycks & Gimpel, 2017a) proposes to use maximum", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 392, + 469, + 405 + ], + "spans": [ + { + "bbox": [ + 106, + 392, + 469, + 405 + ], + "score": 1.0, + "content": "softmax probability to discriminate ID and OOD samples. The score is defined as follows,", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5 + }, + { + "type": "interline_equation", + "bbox": [ + 184, + 406, + 427, + 424 + ], + "lines": [ + { + "bbox": [ + 184, + 406, + 427, + 424 + ], + "spans": [ + { + "bbox": [ + 184, + 406, + 427, + 424 + ], + "score": 0.9, + "content": "S _ { \\mathrm { M S P } } ( x ; f ) = \\operatorname* { m a x } _ { c } P ( y = c | x ; f ) = \\operatorname* { m a x } { \\ s o f \\ t m a x } ( f ( x ) )", + "type": "interline_equation", + "image_path": "eb647d90be010a2a4bc2264b9e3f27009ae5a64b4cbb043708a540be775f9884.jpg" + } + ] + } + ], + "index": 21, + "virtual_lines": [ + { + "bbox": [ + 184, + 406, + 427, + 424 + ], + "spans": [], + "index": 21 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 428, + 504, + 461 + ], + "lines": [ + { + "bbox": [ + 105, + 427, + 506, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 427, + 133, + 441 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 134, + 429, + 141, + 440 + ], + "score": 0.85, + "content": "f", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 427, + 323, + 441 + ], + "score": 1.0, + "content": "represents the given well-trained model and", + "type": "text" + }, + { + "bbox": [ + 323, + 430, + 329, + 438 + ], + "score": 0.74, + "content": "c", + "type": "inline_equation" + }, + { + "bbox": [ + 330, + 427, + 414, + 441 + ], + "score": 1.0, + "content": "is one of the classes", + "type": "text" + }, + { + "bbox": [ + 414, + 428, + 482, + 440 + ], + "score": 0.93, + "content": "\\mathcal { V } = \\{ 1 , \\ldots , C \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 482, + 427, + 506, + 441 + ], + "score": 1.0, + "content": ". The", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 439, + 506, + 451 + ], + "spans": [ + { + "bbox": [ + 106, + 439, + 506, + 451 + ], + "score": 1.0, + "content": "larger softmax score indicates the larger probability for a sample to be ID data, reflecting the model’s", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 450, + 212, + 462 + ], + "spans": [ + { + "bbox": [ + 106, + 450, + 212, + 462 + ], + "score": 1.0, + "content": "confidence on the sample.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 23 + }, + { + "type": "text", + "bbox": [ + 106, + 473, + 505, + 506 + ], + "lines": [ + { + "bbox": [ + 105, + 472, + 505, + 487 + ], + "spans": [ + { + "bbox": [ + 105, + 472, + 505, + 487 + ], + "score": 1.0, + "content": "ODIN. (Liang et al., 2018) designed the ODIN score, leveraging the temperature scaling and tiny", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 484, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 484, + 505, + 496 + ], + "score": 1.0, + "content": "perturbations to widen the gap between the distributions of ID and OOD samples. The ODIN score is", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 494, + 185, + 508 + ], + "spans": [ + { + "bbox": [ + 105, + 494, + 185, + 508 + ], + "score": 1.0, + "content": "defined as follows,", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26 + }, + { + "type": "interline_equation", + "bbox": [ + 181, + 509, + 430, + 533 + ], + "lines": [ + { + "bbox": [ + 181, + 509, + 430, + 533 + ], + "spans": [ + { + "bbox": [ + 181, + 509, + 430, + 533 + ], + "score": 0.91, + "content": "S _ { \\mathrm { O D I N } } ( x ; f ) = \\operatorname* { m a x } _ { c } P ( y = c | \\tilde { x } ; f ) = \\operatorname* { m a x } \\ s \\ o f \\tan \\alpha ( \\frac { f ( \\tilde { x } ) } { T } )", + "type": "interline_equation", + "image_path": "58226108e1cc1addb3f70429e71b26864132fcdcf209f7b1ce63d32ddfcb0abd.jpg" + } + ] + } + ], + "index": 28, + "virtual_lines": [ + { + "bbox": [ + 181, + 509, + 430, + 533 + ], + "spans": [], + "index": 28 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 536, + 504, + 570 + ], + "lines": [ + { + "bbox": [ + 105, + 535, + 506, + 549 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 134, + 549 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 135, + 537, + 142, + 546 + ], + "score": 0.8, + "content": "\\tilde { x }", + "type": "inline_equation" + }, + { + "bbox": [ + 142, + 535, + 344, + 549 + ], + "score": 1.0, + "content": "represents the perturbed samples (controled by", + "type": "text" + }, + { + "bbox": [ + 344, + 538, + 350, + 547 + ], + "score": 0.61, + "content": "\\dot { \\epsilon }", + "type": "inline_equation" + }, + { + "bbox": [ + 351, + 535, + 358, + 549 + ], + "score": 1.0, + "content": "),", + "type": "text" + }, + { + "bbox": [ + 358, + 537, + 367, + 546 + ], + "score": 0.75, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 367, + 535, + 506, + 549 + ], + "score": 1.0, + "content": "represents the temperature. For", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 104, + 545, + 502, + 560 + ], + "spans": [ + { + "bbox": [ + 104, + 545, + 435, + 560 + ], + "score": 1.0, + "content": "fair comparison, we adopt the suggested hyperparameters (Liang et al., 2018):", + "type": "text" + }, + { + "bbox": [ + 435, + 547, + 502, + 558 + ], + "score": 0.93, + "content": "\\epsilon \\doteq 1 . 4 \\times 1 0 ^ { - 3 }", + "type": "inline_equation" + } + ], + "index": 30 + }, + { + "bbox": [ + 107, + 558, + 172, + 569 + ], + "spans": [ + { + "bbox": [ + 107, + 558, + 167, + 569 + ], + "score": 0.91, + "content": "T = 1 . 0 { \\dot { \\times } } 1 0 ^ { 4 }", + "type": "inline_equation" + }, + { + "bbox": [ + 168, + 558, + 172, + 569 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 106, + 581, + 505, + 614 + ], + "lines": [ + { + "bbox": [ + 105, + 580, + 506, + 595 + ], + "spans": [ + { + "bbox": [ + 105, + 580, + 506, + 595 + ], + "score": 1.0, + "content": "Mahalanobis. (Lee et al., 2018b) introduces a Mahalanobis distance-based confidence score,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 593, + 505, + 604 + ], + "spans": [ + { + "bbox": [ + 105, + 593, + 505, + 604 + ], + "score": 1.0, + "content": "exploiting the feature space of the neural networks by inspecting the class conditional Gaussian", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 603, + 379, + 615 + ], + "spans": [ + { + "bbox": [ + 106, + 603, + 379, + 615 + ], + "score": 1.0, + "content": "distributions. The Mahalanobis distance score is defined as follows,", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33 + }, + { + "type": "interline_equation", + "bbox": [ + 189, + 616, + 422, + 636 + ], + "lines": [ + { + "bbox": [ + 189, + 616, + 422, + 636 + ], + "spans": [ + { + "bbox": [ + 189, + 616, + 422, + 636 + ], + "score": 0.91, + "content": "S _ { \\mathrm { M a h a l a n o b i s } } ( x ; f ) = \\operatorname* { m a x } _ { c } - ( f ( x ) - \\hat { \\mu } _ { c } ) ^ { T } \\hat { \\Sigma } ^ { - 1 } ( f ( x ) - \\hat { \\mu } _ { c } )", + "type": "interline_equation", + "image_path": "891b3f112444c2d329b35cc3e1e45abd0703a13b9d9b5a9a91c0a457b3486865.jpg" + } + ] + } + ], + "index": 35, + "virtual_lines": [ + { + "bbox": [ + 189, + 616, + 422, + 636 + ], + "spans": [], + "index": 35 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 641, + 505, + 664 + ], + "lines": [ + { + "bbox": [ + 105, + 640, + 505, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 640, + 132, + 654 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 132, + 642, + 144, + 653 + ], + "score": 0.89, + "content": "\\hat { \\mu } _ { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 144, + 640, + 444, + 654 + ], + "score": 1.0, + "content": "represents the estimated mean of multivariate Gaussian distribution of class", + "type": "text" + }, + { + "bbox": [ + 444, + 644, + 450, + 652 + ], + "score": 0.51, + "content": "c", + "type": "inline_equation" + }, + { + "bbox": [ + 450, + 640, + 453, + 654 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 453, + 640, + 462, + 651 + ], + "score": 0.75, + "content": "\\hat { \\Sigma }", + "type": "inline_equation" + }, + { + "bbox": [ + 462, + 640, + 505, + 654 + ], + "score": 1.0, + "content": "represents", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 653, + 423, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 653, + 250, + 664 + ], + "score": 1.0, + "content": "the estimated tied covariance of the", + "type": "text" + }, + { + "bbox": [ + 250, + 653, + 259, + 662 + ], + "score": 0.82, + "content": "C", + "type": "inline_equation" + }, + { + "bbox": [ + 259, + 653, + 423, + 664 + ], + "score": 1.0, + "content": "class-conditional Gaussian distributions.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5 + }, + { + "type": "text", + "bbox": [ + 105, + 675, + 504, + 698 + ], + "lines": [ + { + "bbox": [ + 106, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "Energy. (Liu et al., 2020) proposes to use the Energy of the predicted logits to distinguish the ID", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 686, + 344, + 699 + ], + "spans": [ + { + "bbox": [ + 106, + 686, + 344, + 699 + ], + "score": 1.0, + "content": "and OOD samples. The Energy score is defined as follows,", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5 + }, + { + "type": "interline_equation", + "bbox": [ + 232, + 702, + 379, + 735 + ], + "lines": [ + { + "bbox": [ + 232, + 702, + 379, + 735 + ], + "spans": [ + { + "bbox": [ + 232, + 702, + 379, + 735 + ], + "score": 0.95, + "content": "S _ { \\mathrm { E n e r g y } } ( x ; f ) = - T \\log \\sum _ { c = 1 } ^ { C } e ^ { f ( x ) _ { c } / T }", + "type": "interline_equation", + "image_path": "943af2b0f9767b994b1104b7822273e96ffa5715dea3e0fd413a2e99977a8a73.jpg" + } + ] + } + ], + "index": 40.5, + "virtual_lines": [ + { + "bbox": [ + 232, + 702, + 379, + 718.5 + ], + "spans": [], + "index": 40 + }, + { + "bbox": [ + 232, + 718.5, + 379, + 735.0 + ], + "spans": [], + "index": 41 + } + ] + } + ], + "page_idx": 12, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 26, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "13", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 82, + 161, + 93 + ], + "lines": [ + { + "bbox": [ + 106, + 81, + 163, + 96 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 163, + 96 + ], + "score": 1.0, + "content": "APPENDIX", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "title", + "bbox": [ + 108, + 107, + 268, + 120 + ], + "lines": [ + { + "bbox": [ + 106, + 108, + 269, + 121 + ], + "spans": [ + { + "bbox": [ + 106, + 108, + 269, + 121 + ], + "score": 1.0, + "content": "REPRODUCIBILITY STATEMENT", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 107, + 132, + 505, + 165 + ], + "lines": [ + { + "bbox": [ + 105, + 132, + 505, + 145 + ], + "spans": [ + { + "bbox": [ + 105, + 132, + 505, + 145 + ], + "score": 1.0, + "content": "We will provide the anonymous repository about our source codes in the discussion phase for", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 143, + 505, + 156 + ], + "spans": [ + { + "bbox": [ + 105, + 143, + 505, + 156 + ], + "score": 1.0, + "content": "reviewing purposes to ensure the reproducibility of our experimental results. Below we summarize", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 154, + 338, + 166 + ], + "spans": [ + { + "bbox": [ + 105, + 154, + 338, + 166 + ], + "score": 1.0, + "content": "some critical aspects to facilitate the reproducible results:", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3, + "bbox_fs": [ + 105, + 132, + 505, + 166 + ] + }, + { + "type": "list", + "bbox": [ + 133, + 174, + 505, + 307 + ], + "lines": [ + { + "bbox": [ + 132, + 174, + 507, + 188 + ], + "spans": [ + { + "bbox": [ + 132, + 174, + 507, + 188 + ], + "score": 1.0, + "content": "• Datasets. The datasets we used are all publicly accessible, which is introduced in Section 4.1.", + "type": "text" + } + ], + "index": 5, + "is_list_start_line": true + }, + { + "bbox": [ + 141, + 186, + 506, + 198 + ], + "spans": [ + { + "bbox": [ + 141, + 186, + 506, + 198 + ], + "score": 1.0, + "content": "For methods involving auxiliary outliers, we strictly follow previous works (Sun et al., 2021;", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 141, + 196, + 506, + 209 + ], + "spans": [ + { + "bbox": [ + 141, + 196, + 506, + 209 + ], + "score": 1.0, + "content": "Du et al., 2022b) to avoid overlap between the auxiliary dataset (ImageNet-1k) (Deng et al.,", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 141, + 207, + 285, + 221 + ], + "spans": [ + { + "bbox": [ + 141, + 207, + 285, + 221 + ], + "score": 1.0, + "content": "2009) and any other OOD datasets.", + "type": "text" + } + ], + "index": 8, + "is_list_end_line": true + }, + { + "bbox": [ + 135, + 222, + 506, + 235 + ], + "spans": [ + { + "bbox": [ + 135, + 222, + 506, + 235 + ], + "score": 1.0, + "content": "• Assumption. We set our experiments to a post-hoc scenario where a well-trained model is", + "type": "text" + } + ], + "index": 9, + "is_list_start_line": true + }, + { + "bbox": [ + 140, + 232, + 505, + 247 + ], + "spans": [ + { + "bbox": [ + 140, + 232, + 505, + 247 + ], + "score": 1.0, + "content": "available, and some parts of training samples are also available for subsequent fine-tuning.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 135, + 248, + 505, + 261 + ], + "spans": [ + { + "bbox": [ + 135, + 248, + 505, + 261 + ], + "score": 1.0, + "content": "• Open source. The code repository will be available in an anonymous repository for", + "type": "text" + } + ], + "index": 11, + "is_list_start_line": true + }, + { + "bbox": [ + 141, + 258, + 505, + 272 + ], + "spans": [ + { + "bbox": [ + 141, + 258, + 505, + 272 + ], + "score": 1.0, + "content": "reviewing purposes. We provide a backbone of our experiments as well as several auxiliary", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 142, + 271, + 296, + 282 + ], + "spans": [ + { + "bbox": [ + 142, + 271, + 296, + 282 + ], + "score": 1.0, + "content": "components, such as score estimation.", + "type": "text" + } + ], + "index": 13, + "is_list_end_line": true + }, + { + "bbox": [ + 132, + 283, + 506, + 297 + ], + "spans": [ + { + "bbox": [ + 132, + 283, + 506, + 297 + ], + "score": 1.0, + "content": "• Environment. All experiments are conducted multiple runs on NVIDIA Tesla V100-SXM2-", + "type": "text" + } + ], + "index": 14, + "is_list_start_line": true + }, + { + "bbox": [ + 141, + 295, + 331, + 307 + ], + "spans": [ + { + "bbox": [ + 141, + 295, + 331, + 307 + ], + "score": 1.0, + "content": "32GB GPUs with Python 3.6 and PyTorch 1.8.", + "type": "text" + } + ], + "index": 15, + "is_list_end_line": true + } + ], + "index": 10, + "bbox_fs": [ + 132, + 174, + 507, + 307 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 323, + 482, + 335 + ], + "lines": [ + { + "bbox": [ + 105, + 323, + 483, + 337 + ], + "spans": [ + { + "bbox": [ + 105, + 323, + 483, + 337 + ], + "score": 1.0, + "content": "A DETAILED INFORMATION ABOUT THE USED BASELINES AND METRICS", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 106, + 346, + 505, + 370 + ], + "lines": [ + { + "bbox": [ + 105, + 344, + 505, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 505, + 361 + ], + "score": 1.0, + "content": "In this section, we provide the details about the baselines for the scoring functions and finetuning", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 358, + 490, + 370 + ], + "spans": [ + { + "bbox": [ + 106, + 358, + 490, + 370 + ], + "score": 1.0, + "content": "with auxiliary outliers, and the corresponding hyper-parameters that are considered in our work.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5, + "bbox_fs": [ + 105, + 344, + 505, + 370 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 380, + 504, + 403 + ], + "lines": [ + { + "bbox": [ + 105, + 381, + 505, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 381, + 505, + 394 + ], + "score": 1.0, + "content": "Maximum Softmax Probability (MSP). (Hendrycks & Gimpel, 2017a) proposes to use maximum", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 392, + 469, + 405 + ], + "spans": [ + { + "bbox": [ + 106, + 392, + 469, + 405 + ], + "score": 1.0, + "content": "softmax probability to discriminate ID and OOD samples. The score is defined as follows,", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5, + "bbox_fs": [ + 105, + 381, + 505, + 405 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 184, + 406, + 427, + 424 + ], + "lines": [ + { + "bbox": [ + 184, + 406, + 427, + 424 + ], + "spans": [ + { + "bbox": [ + 184, + 406, + 427, + 424 + ], + "score": 0.9, + "content": "S _ { \\mathrm { M S P } } ( x ; f ) = \\operatorname* { m a x } _ { c } P ( y = c | x ; f ) = \\operatorname* { m a x } { \\ s o f \\ t m a x } ( f ( x ) )", + "type": "interline_equation", + "image_path": "eb647d90be010a2a4bc2264b9e3f27009ae5a64b4cbb043708a540be775f9884.jpg" + } + ] + } + ], + "index": 21, + "virtual_lines": [ + { + "bbox": [ + 184, + 406, + 427, + 424 + ], + "spans": [], + "index": 21 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 428, + 504, + 461 + ], + "lines": [ + { + "bbox": [ + 105, + 427, + 506, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 427, + 133, + 441 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 134, + 429, + 141, + 440 + ], + "score": 0.85, + "content": "f", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 427, + 323, + 441 + ], + "score": 1.0, + "content": "represents the given well-trained model and", + "type": "text" + }, + { + "bbox": [ + 323, + 430, + 329, + 438 + ], + "score": 0.74, + "content": "c", + "type": "inline_equation" + }, + { + "bbox": [ + 330, + 427, + 414, + 441 + ], + "score": 1.0, + "content": "is one of the classes", + "type": "text" + }, + { + "bbox": [ + 414, + 428, + 482, + 440 + ], + "score": 0.93, + "content": "\\mathcal { V } = \\{ 1 , \\ldots , C \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 482, + 427, + 506, + 441 + ], + "score": 1.0, + "content": ". The", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 439, + 506, + 451 + ], + "spans": [ + { + "bbox": [ + 106, + 439, + 506, + 451 + ], + "score": 1.0, + "content": "larger softmax score indicates the larger probability for a sample to be ID data, reflecting the model’s", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 450, + 212, + 462 + ], + "spans": [ + { + "bbox": [ + 106, + 450, + 212, + 462 + ], + "score": 1.0, + "content": "confidence on the sample.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 23, + "bbox_fs": [ + 105, + 427, + 506, + 462 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 473, + 505, + 506 + ], + "lines": [ + { + "bbox": [ + 105, + 472, + 505, + 487 + ], + "spans": [ + { + "bbox": [ + 105, + 472, + 505, + 487 + ], + "score": 1.0, + "content": "ODIN. (Liang et al., 2018) designed the ODIN score, leveraging the temperature scaling and tiny", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 484, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 484, + 505, + 496 + ], + "score": 1.0, + "content": "perturbations to widen the gap between the distributions of ID and OOD samples. The ODIN score is", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 494, + 185, + 508 + ], + "spans": [ + { + "bbox": [ + 105, + 494, + 185, + 508 + ], + "score": 1.0, + "content": "defined as follows,", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26, + "bbox_fs": [ + 105, + 472, + 505, + 508 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 181, + 509, + 430, + 533 + ], + "lines": [ + { + "bbox": [ + 181, + 509, + 430, + 533 + ], + "spans": [ + { + "bbox": [ + 181, + 509, + 430, + 533 + ], + "score": 0.91, + "content": "S _ { \\mathrm { O D I N } } ( x ; f ) = \\operatorname* { m a x } _ { c } P ( y = c | \\tilde { x } ; f ) = \\operatorname* { m a x } \\ s \\ o f \\tan \\alpha ( \\frac { f ( \\tilde { x } ) } { T } )", + "type": "interline_equation", + "image_path": "58226108e1cc1addb3f70429e71b26864132fcdcf209f7b1ce63d32ddfcb0abd.jpg" + } + ] + } + ], + "index": 28, + "virtual_lines": [ + { + "bbox": [ + 181, + 509, + 430, + 533 + ], + "spans": [], + "index": 28 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 536, + 504, + 570 + ], + "lines": [ + { + "bbox": [ + 105, + 535, + 506, + 549 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 134, + 549 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 135, + 537, + 142, + 546 + ], + "score": 0.8, + "content": "\\tilde { x }", + "type": "inline_equation" + }, + { + "bbox": [ + 142, + 535, + 344, + 549 + ], + "score": 1.0, + "content": "represents the perturbed samples (controled by", + "type": "text" + }, + { + "bbox": [ + 344, + 538, + 350, + 547 + ], + "score": 0.61, + "content": "\\dot { \\epsilon }", + "type": "inline_equation" + }, + { + "bbox": [ + 351, + 535, + 358, + 549 + ], + "score": 1.0, + "content": "),", + "type": "text" + }, + { + "bbox": [ + 358, + 537, + 367, + 546 + ], + "score": 0.75, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 367, + 535, + 506, + 549 + ], + "score": 1.0, + "content": "represents the temperature. For", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 104, + 545, + 502, + 560 + ], + "spans": [ + { + "bbox": [ + 104, + 545, + 435, + 560 + ], + "score": 1.0, + "content": "fair comparison, we adopt the suggested hyperparameters (Liang et al., 2018):", + "type": "text" + }, + { + "bbox": [ + 435, + 547, + 502, + 558 + ], + "score": 0.93, + "content": "\\epsilon \\doteq 1 . 4 \\times 1 0 ^ { - 3 }", + "type": "inline_equation" + } + ], + "index": 30 + }, + { + "bbox": [ + 107, + 558, + 172, + 569 + ], + "spans": [ + { + "bbox": [ + 107, + 558, + 167, + 569 + ], + "score": 0.91, + "content": "T = 1 . 0 { \\dot { \\times } } 1 0 ^ { 4 }", + "type": "inline_equation" + }, + { + "bbox": [ + 168, + 558, + 172, + 569 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30, + "bbox_fs": [ + 104, + 535, + 506, + 569 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 581, + 505, + 614 + ], + "lines": [ + { + "bbox": [ + 105, + 580, + 506, + 595 + ], + "spans": [ + { + "bbox": [ + 105, + 580, + 506, + 595 + ], + "score": 1.0, + "content": "Mahalanobis. (Lee et al., 2018b) introduces a Mahalanobis distance-based confidence score,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 593, + 505, + 604 + ], + "spans": [ + { + "bbox": [ + 105, + 593, + 505, + 604 + ], + "score": 1.0, + "content": "exploiting the feature space of the neural networks by inspecting the class conditional Gaussian", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 603, + 379, + 615 + ], + "spans": [ + { + "bbox": [ + 106, + 603, + 379, + 615 + ], + "score": 1.0, + "content": "distributions. The Mahalanobis distance score is defined as follows,", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33, + "bbox_fs": [ + 105, + 580, + 506, + 615 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 189, + 616, + 422, + 636 + ], + "lines": [ + { + "bbox": [ + 189, + 616, + 422, + 636 + ], + "spans": [ + { + "bbox": [ + 189, + 616, + 422, + 636 + ], + "score": 0.91, + "content": "S _ { \\mathrm { M a h a l a n o b i s } } ( x ; f ) = \\operatorname* { m a x } _ { c } - ( f ( x ) - \\hat { \\mu } _ { c } ) ^ { T } \\hat { \\Sigma } ^ { - 1 } ( f ( x ) - \\hat { \\mu } _ { c } )", + "type": "interline_equation", + "image_path": "891b3f112444c2d329b35cc3e1e45abd0703a13b9d9b5a9a91c0a457b3486865.jpg" + } + ] + } + ], + "index": 35, + "virtual_lines": [ + { + "bbox": [ + 189, + 616, + 422, + 636 + ], + "spans": [], + "index": 35 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 641, + 505, + 664 + ], + "lines": [ + { + "bbox": [ + 105, + 640, + 505, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 640, + 132, + 654 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 132, + 642, + 144, + 653 + ], + "score": 0.89, + "content": "\\hat { \\mu } _ { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 144, + 640, + 444, + 654 + ], + "score": 1.0, + "content": "represents the estimated mean of multivariate Gaussian distribution of class", + "type": "text" + }, + { + "bbox": [ + 444, + 644, + 450, + 652 + ], + "score": 0.51, + "content": "c", + "type": "inline_equation" + }, + { + "bbox": [ + 450, + 640, + 453, + 654 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 453, + 640, + 462, + 651 + ], + "score": 0.75, + "content": "\\hat { \\Sigma }", + "type": "inline_equation" + }, + { + "bbox": [ + 462, + 640, + 505, + 654 + ], + "score": 1.0, + "content": "represents", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 653, + 423, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 653, + 250, + 664 + ], + "score": 1.0, + "content": "the estimated tied covariance of the", + "type": "text" + }, + { + "bbox": [ + 250, + 653, + 259, + 662 + ], + "score": 0.82, + "content": "C", + "type": "inline_equation" + }, + { + "bbox": [ + 259, + 653, + 423, + 664 + ], + "score": 1.0, + "content": "class-conditional Gaussian distributions.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5, + "bbox_fs": [ + 105, + 640, + 505, + 664 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 675, + 504, + 698 + ], + "lines": [ + { + "bbox": [ + 106, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "Energy. (Liu et al., 2020) proposes to use the Energy of the predicted logits to distinguish the ID", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 686, + 344, + 699 + ], + "spans": [ + { + "bbox": [ + 106, + 686, + 344, + 699 + ], + "score": 1.0, + "content": "and OOD samples. The Energy score is defined as follows,", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5, + "bbox_fs": [ + 106, + 676, + 505, + 699 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 232, + 702, + 379, + 735 + ], + "lines": [ + { + "bbox": [ + 232, + 702, + 379, + 735 + ], + "spans": [ + { + "bbox": [ + 232, + 702, + 379, + 735 + ], + "score": 0.95, + "content": "S _ { \\mathrm { E n e r g y } } ( x ; f ) = - T \\log \\sum _ { c = 1 } ^ { C } e ^ { f ( x ) _ { c } / T }", + "type": "interline_equation", + "image_path": "943af2b0f9767b994b1104b7822273e96ffa5715dea3e0fd413a2e99977a8a73.jpg" + } + ] + } + ], + "index": 40.5, + "virtual_lines": [ + { + "bbox": [ + 232, + 702, + 379, + 718.5 + ], + "spans": [], + "index": 40 + }, + { + "bbox": [ + 232, + 718.5, + 379, + 735.0 + ], + "spans": [], + "index": 41 + } + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 132, + 95 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 132, + 83, + 141, + 92 + ], + "score": 0.74, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 82, + 506, + 95 + ], + "score": 1.0, + "content": "represents the temperature parameter. As theoretically illustrated in Liu et al. (2020), a lower", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 94, + 506, + 106 + ], + "spans": [ + { + "bbox": [ + 106, + 94, + 506, + 106 + ], + "score": 1.0, + "content": "Energy score indicates a higher probability for a sample to be ID. Following (Liu et al., 2020), we fix", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 104, + 269, + 117 + ], + "spans": [ + { + "bbox": [ + 106, + 104, + 121, + 117 + ], + "score": 1.0, + "content": "the", + "type": "text" + }, + { + "bbox": [ + 121, + 105, + 130, + 114 + ], + "score": 0.83, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 130, + 104, + 269, + 117 + ], + "score": 1.0, + "content": "to 1.0 throughout all experiments.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 107, + 128, + 505, + 172 + ], + "lines": [ + { + "bbox": [ + 106, + 128, + 505, + 142 + ], + "spans": [ + { + "bbox": [ + 106, + 128, + 505, + 142 + ], + "score": 1.0, + "content": "Outlier Exposure (OE). (Hendrycks et al., 2019b) initiates a promising approach towards OOD", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 140, + 505, + 152 + ], + "spans": [ + { + "bbox": [ + 105, + 140, + 505, + 152 + ], + "score": 1.0, + "content": "detections by involving outliers to force apart the distributions of ID and OOD samples. In the", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 104, + 151, + 507, + 163 + ], + "spans": [ + { + "bbox": [ + 104, + 151, + 281, + 163 + ], + "score": 1.0, + "content": "experiments, we use the cross-entropy from", + "type": "text" + }, + { + "bbox": [ + 282, + 151, + 311, + 163 + ], + "score": 0.89, + "content": "f ( x _ { \\mathrm { o u t } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 311, + 151, + 442, + 163 + ], + "score": 1.0, + "content": "to the uniform distribution as the", + "type": "text" + }, + { + "bbox": [ + 443, + 151, + 460, + 162 + ], + "score": 0.86, + "content": "\\mathcal { L } _ { \\mathrm { O E } }", + "type": "inline_equation" + }, + { + "bbox": [ + 461, + 151, + 507, + 163 + ], + "score": 1.0, + "content": "(Lee et al.,", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 159, + 141, + 175 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 141, + 175 + ], + "score": 1.0, + "content": "2018a),", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + }, + { + "type": "interline_equation", + "bbox": [ + 170, + 172, + 441, + 207 + ], + "lines": [ + { + "bbox": [ + 170, + 172, + 441, + 207 + ], + "spans": [ + { + "bbox": [ + 170, + 172, + 441, + 207 + ], + "score": 0.94, + "content": "\\mathcal { L } _ { f } = \\mathbb { E } _ { D _ { \\mathrm { i n } } } \\left[ \\ell _ { \\mathrm { C E } } ( f ( x ) , y ) \\right] + \\lambda \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } \\left[ \\log \\sum _ { c = 1 } ^ { C } e ^ { f ( x ) _ { c } } - \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } ( f ( x ) ) \\right]", + "type": "interline_equation", + "image_path": "e370174786b5ffabd541014636396d5283038b86fad99a11cfbc378251857514.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 170, + 172, + 441, + 183.66666666666666 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 170, + 183.66666666666666, + 441, + 195.33333333333331 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 170, + 195.33333333333331, + 441, + 206.99999999999997 + ], + "spans": [], + "index": 9 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 218, + 505, + 251 + ], + "lines": [ + { + "bbox": [ + 105, + 217, + 506, + 231 + ], + "spans": [ + { + "bbox": [ + 105, + 217, + 160, + 231 + ], + "score": 1.0, + "content": "Energy (w.", + "type": "text" + }, + { + "bbox": [ + 160, + 218, + 183, + 230 + ], + "score": 0.87, + "content": "D _ { \\mathbf { a u x } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 183, + 217, + 473, + 231 + ], + "score": 1.0, + "content": "). In addition to using the Energy as a post-hoc score to distinguish", + "type": "text" + }, + { + "bbox": [ + 473, + 218, + 486, + 228 + ], + "score": 0.33, + "content": "\\mathrm { I D }", + "type": "inline_equation" + }, + { + "bbox": [ + 487, + 217, + 506, + 231 + ], + "score": 1.0, + "content": "and", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 228, + 505, + 242 + ], + "spans": [ + { + "bbox": [ + 105, + 228, + 505, + 242 + ], + "score": 1.0, + "content": "OOD samples, (Liu et al., 2020) proposes an Energy-bounded objective to further separate the two", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 239, + 289, + 252 + ], + "spans": [ + { + "bbox": [ + 106, + 239, + 289, + 252 + ], + "score": 1.0, + "content": "distributions. The OE objective is as follows,", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11 + }, + { + "type": "interline_equation", + "bbox": [ + 128, + 256, + 465, + 272 + ], + "lines": [ + { + "bbox": [ + 128, + 256, + 465, + 272 + ], + "spans": [ + { + "bbox": [ + 128, + 256, + 465, + 272 + ], + "score": 0.86, + "content": "\\mathcal { L } _ { \\mathrm { O E } } = \\mathbb { E } _ { D _ { \\mathrm { i n } } ^ { s } } ( \\operatorname* { m a x } ( 0 , S _ { \\mathrm { E n e r g y } } ( x , f ) - m _ { \\mathrm { i n } } ) ) ^ { 2 } + \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } ( \\operatorname* { m a x } ( 0 , m _ { \\mathrm { o u t } } - S _ { \\mathrm { E n e r g y } } ( x , f ) ) ) ^ { 2 }", + "type": "interline_equation", + "image_path": "54a9f407cbd94084751fcdf26413676f1b8a4a35835da9c373810c40b9ce2039.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 128, + 256, + 465, + 272 + ], + "spans": [], + "index": 13 + } + ] + }, + { + "type": "text", + "bbox": [ + 108, + 276, + 423, + 289 + ], + "lines": [ + { + "bbox": [ + 106, + 276, + 422, + 291 + ], + "spans": [ + { + "bbox": [ + 106, + 276, + 307, + 291 + ], + "score": 1.0, + "content": "We keep the thresholds same to (Liu et al., 2020):", + "type": "text" + }, + { + "bbox": [ + 307, + 277, + 362, + 289 + ], + "score": 0.76, + "content": "m _ { \\mathrm { i n } } = - 2 5 . 0", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 276, + 365, + 291 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 365, + 277, + 419, + 289 + ], + "score": 0.81, + "content": "m _ { \\mathrm { o u t } } = - 7 . 0", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 276, + 422, + 291 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14 + }, + { + "type": "text", + "bbox": [ + 106, + 301, + 504, + 347 + ], + "lines": [ + { + "bbox": [ + 104, + 301, + 506, + 315 + ], + "spans": [ + { + "bbox": [ + 104, + 301, + 506, + 315 + ], + "score": 1.0, + "content": "POEM. (Ming et al., 2022) explores the Thompson sampling strategy (Thompson, 1933) to make", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 313, + 505, + 325 + ], + "spans": [ + { + "bbox": [ + 106, + 313, + 505, + 325 + ], + "score": 1.0, + "content": "the most use of outliers to learn a tight decision boundary. Though given the POEM’s nature to be", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 324, + 506, + 337 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 327, + 337 + ], + "score": 1.0, + "content": "orthogonal to other OE methods, we use the Energy(w.", + "type": "text" + }, + { + "bbox": [ + 328, + 324, + 348, + 335 + ], + "score": 0.88, + "content": "D _ { \\mathrm { a u x . } }", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 324, + 506, + 337 + ], + "score": 1.0, + "content": ") as the backbone, which is the same as", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 335, + 488, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 335, + 488, + 347 + ], + "score": 1.0, + "content": "Eq.( 10) in Liu et al. (2020). The details of Thompson sampling can refer to Ming et al. (2022).", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 16.5 + }, + { + "type": "text", + "bbox": [ + 107, + 358, + 505, + 415 + ], + "lines": [ + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "score": 1.0, + "content": "Detailed formulations of FPR and TPR. Suppose we have a binary classification task (to predict", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 370, + 505, + 382 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 505, + 382 + ], + "score": 1.0, + "content": "an image to be an ID or OOD sample in this paper). There are two possible outputs: a positive result", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 380, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 106, + 380, + 506, + 394 + ], + "score": 1.0, + "content": "(the model predicts an image to be an ID sample); a negative result (the model predicts an image to", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 392, + 506, + 404 + ], + "spans": [ + { + "bbox": [ + 106, + 392, + 506, + 404 + ], + "score": 1.0, + "content": "be an OOD sample). Since we have two possible labels and two possible outputs, we can form a", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 402, + 322, + 416 + ], + "spans": [ + { + "bbox": [ + 105, + 402, + 322, + 416 + ], + "score": 1.0, + "content": "confusion matrix with all possible outputs as follows.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 21 + }, + { + "type": "table", + "bbox": [], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 250, + 425, + 360, + 437 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 249, + 424, + 361, + 438 + ], + "spans": [ + { + "bbox": [ + 249, + 424, + 361, + 438 + ], + "score": 1.0, + "content": "Table 3: Confusion Matrix.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "interline_equation", + "bbox": [ + 185, + 441, + 422, + 488 + ], + "lines": [ + { + "bbox": [ + 185, + 441, + 422, + 488 + ], + "spans": [ + { + "bbox": [ + 185, + 441, + 422, + 488 + ], + "score": 0.36, + "content": "\\frac { | \\mathrm { ~ { T r u t h : I D } ~ } | \\mathrm { { T r u t h : O O D } } } { \\mathrm { { P r e d i c t : ~ O O D ~ | \\mathrm { ~ { ~ T r u e ~ P o s i t i v e ~ ( T P ) ~ } | { \\mathrm { ~ F a l s e ~ P o s i t i v e ~ ( F P ) ~ } } } } } }", + "type": "interline_equation", + "image_path": "ca2f2a2fa96503d7c6f7f24da26223a214941e13ae665bd918c6f60ee89e3b68.jpg" + } + ] + } + ], + "index": 26, + "virtual_lines": [ + { + "bbox": [ + 185, + 441, + 422, + 456.6666666666667 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 185, + 456.6666666666667, + 422, + 472.33333333333337 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 185, + 472.33333333333337, + 422, + 488.00000000000006 + ], + "spans": [], + "index": 27 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 498, + 330, + 510 + ], + "lines": [ + { + "bbox": [ + 106, + 497, + 333, + 510 + ], + "spans": [ + { + "bbox": [ + 106, + 497, + 333, + 510 + ], + "score": 1.0, + "content": "Therefore, the false positive rate (FPR) is calculated as :", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "interline_equation", + "bbox": [ + 262, + 515, + 348, + 540 + ], + "lines": [ + { + "bbox": [ + 262, + 515, + 348, + 540 + ], + "spans": [ + { + "bbox": [ + 262, + 515, + 348, + 540 + ], + "score": 0.95, + "content": "F P R = { \\frac { F P } { F P + T N } }", + "type": "interline_equation", + "image_path": "6de9bd865babdf43ef851a8d2f51cca43a0e5ddbbc75bd8c1eb583ef265e7786.jpg" + } + ] + } + ], + "index": 29, + "virtual_lines": [ + { + "bbox": [ + 262, + 515, + 348, + 540 + ], + "spans": [], + "index": 29 + } + ] + }, + { + "type": "text", + "bbox": [ + 108, + 551, + 285, + 563 + ], + "lines": [ + { + "bbox": [ + 106, + 552, + 285, + 563 + ], + "spans": [ + { + "bbox": [ + 106, + 552, + 285, + 563 + ], + "score": 1.0, + "content": "The true positive rate (TPR) is calculated as:", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "interline_equation", + "bbox": [ + 263, + 568, + 348, + 594 + ], + "lines": [ + { + "bbox": [ + 263, + 568, + 348, + 594 + ], + "spans": [ + { + "bbox": [ + 263, + 568, + 348, + 594 + ], + "score": 0.95, + "content": "T P R = { \\frac { T P } { T P + F N } }", + "type": "interline_equation", + "image_path": "1175e2c392fa210f94055024b476db4a2fb3df5a12672ac7bfe9beb3d5388cde.jpg" + } + ] + } + ], + "index": 31, + "virtual_lines": [ + { + "bbox": [ + 263, + 568, + 348, + 594 + ], + "spans": [], + "index": 31 + } + ] + }, + { + "type": "title", + "bbox": [ + 105, + 608, + 492, + 620 + ], + "lines": [ + { + "bbox": [ + 106, + 608, + 493, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 117, + 619 + ], + "score": 1.0, + "content": "B", + "type": "text" + }, + { + "bbox": [ + 128, + 609, + 493, + 619 + ], + "score": 1.0, + "content": "ADDITIONAL EXPLANATION TOWARDS MINING THE ATYPICAL SAMPLES", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 32 + }, + { + "type": "text", + "bbox": [ + 106, + 632, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 632, + 506, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 506, + 645 + ], + "score": 1.0, + "content": "First, for identifying those atypical samples using a layer-wise mask with the well-pre-trained model,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 644, + 506, + 655 + ], + "spans": [ + { + "bbox": [ + 106, + 644, + 506, + 655 + ], + "score": 1.0, + "content": "the core intuition behind is constructing the parameter-level discrepancy to mine the atypical samples.", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 654, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 505, + 667 + ], + "score": 1.0, + "content": "It is inspired by and based on the evidence drawn from previous literature about learning behaviors", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 664, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 664, + 505, + 678 + ], + "score": 1.0, + "content": "(Arpit et al., 2017; Goodfellow et al., 2016) of deep neural networks (DNNs) and sparse representation", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 677, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 505, + 689 + ], + "score": 1.0, + "content": "(Frankle & Carbin, 2019; Goodfellow et al., 2013; Barham et al., 2022). To be specific, the atypical", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "samples tend to be learned by the DNNs later than those typical samples (Arpit et al., 2017), and are", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 699, + 505, + 710 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 505, + 710 + ], + "score": 1.0, + "content": "relatively more sensitive to the changes of the model parameter as the model does not generalize well", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "on that. By the layer-wise mask, the constructed discrepancy can make the model misclassify the", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "atypical samples and estimate loss constraint for the forgetting objective, as visualized in Figure 3(c).", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 37 + } + ], + "page_idx": 13, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 26, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 132, + 95 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 132, + 83, + 141, + 92 + ], + "score": 0.74, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 82, + 506, + 95 + ], + "score": 1.0, + "content": "represents the temperature parameter. As theoretically illustrated in Liu et al. (2020), a lower", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 94, + 506, + 106 + ], + "spans": [ + { + "bbox": [ + 106, + 94, + 506, + 106 + ], + "score": 1.0, + "content": "Energy score indicates a higher probability for a sample to be ID. Following (Liu et al., 2020), we fix", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 104, + 269, + 117 + ], + "spans": [ + { + "bbox": [ + 106, + 104, + 121, + 117 + ], + "score": 1.0, + "content": "the", + "type": "text" + }, + { + "bbox": [ + 121, + 105, + 130, + 114 + ], + "score": 0.83, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 130, + 104, + 269, + 117 + ], + "score": 1.0, + "content": "to 1.0 throughout all experiments.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1, + "bbox_fs": [ + 106, + 82, + 506, + 117 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 128, + 505, + 172 + ], + "lines": [ + { + "bbox": [ + 106, + 128, + 505, + 142 + ], + "spans": [ + { + "bbox": [ + 106, + 128, + 505, + 142 + ], + "score": 1.0, + "content": "Outlier Exposure (OE). (Hendrycks et al., 2019b) initiates a promising approach towards OOD", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 140, + 505, + 152 + ], + "spans": [ + { + "bbox": [ + 105, + 140, + 505, + 152 + ], + "score": 1.0, + "content": "detections by involving outliers to force apart the distributions of ID and OOD samples. In the", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 104, + 151, + 507, + 163 + ], + "spans": [ + { + "bbox": [ + 104, + 151, + 281, + 163 + ], + "score": 1.0, + "content": "experiments, we use the cross-entropy from", + "type": "text" + }, + { + "bbox": [ + 282, + 151, + 311, + 163 + ], + "score": 0.89, + "content": "f ( x _ { \\mathrm { o u t } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 311, + 151, + 442, + 163 + ], + "score": 1.0, + "content": "to the uniform distribution as the", + "type": "text" + }, + { + "bbox": [ + 443, + 151, + 460, + 162 + ], + "score": 0.86, + "content": "\\mathcal { L } _ { \\mathrm { O E } }", + "type": "inline_equation" + }, + { + "bbox": [ + 461, + 151, + 507, + 163 + ], + "score": 1.0, + "content": "(Lee et al.,", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 159, + 141, + 175 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 141, + 175 + ], + "score": 1.0, + "content": "2018a),", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5, + "bbox_fs": [ + 104, + 128, + 507, + 175 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 170, + 172, + 441, + 207 + ], + "lines": [ + { + "bbox": [ + 170, + 172, + 441, + 207 + ], + "spans": [ + { + "bbox": [ + 170, + 172, + 441, + 207 + ], + "score": 0.94, + "content": "\\mathcal { L } _ { f } = \\mathbb { E } _ { D _ { \\mathrm { i n } } } \\left[ \\ell _ { \\mathrm { C E } } ( f ( x ) , y ) \\right] + \\lambda \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } \\left[ \\log \\sum _ { c = 1 } ^ { C } e ^ { f ( x ) _ { c } } - \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } ( f ( x ) ) \\right]", + "type": "interline_equation", + "image_path": "e370174786b5ffabd541014636396d5283038b86fad99a11cfbc378251857514.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 170, + 172, + 441, + 183.66666666666666 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 170, + 183.66666666666666, + 441, + 195.33333333333331 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 170, + 195.33333333333331, + 441, + 206.99999999999997 + ], + "spans": [], + "index": 9 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 218, + 505, + 251 + ], + "lines": [ + { + "bbox": [ + 105, + 217, + 506, + 231 + ], + "spans": [ + { + "bbox": [ + 105, + 217, + 160, + 231 + ], + "score": 1.0, + "content": "Energy (w.", + "type": "text" + }, + { + "bbox": [ + 160, + 218, + 183, + 230 + ], + "score": 0.87, + "content": "D _ { \\mathbf { a u x } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 183, + 217, + 473, + 231 + ], + "score": 1.0, + "content": "). In addition to using the Energy as a post-hoc score to distinguish", + "type": "text" + }, + { + "bbox": [ + 473, + 218, + 486, + 228 + ], + "score": 0.33, + "content": "\\mathrm { I D }", + "type": "inline_equation" + }, + { + "bbox": [ + 487, + 217, + 506, + 231 + ], + "score": 1.0, + "content": "and", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 228, + 505, + 242 + ], + "spans": [ + { + "bbox": [ + 105, + 228, + 505, + 242 + ], + "score": 1.0, + "content": "OOD samples, (Liu et al., 2020) proposes an Energy-bounded objective to further separate the two", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 239, + 289, + 252 + ], + "spans": [ + { + "bbox": [ + 106, + 239, + 289, + 252 + ], + "score": 1.0, + "content": "distributions. The OE objective is as follows,", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 217, + 506, + 252 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 128, + 256, + 465, + 272 + ], + "lines": [ + { + "bbox": [ + 128, + 256, + 465, + 272 + ], + "spans": [ + { + "bbox": [ + 128, + 256, + 465, + 272 + ], + "score": 0.86, + "content": "\\mathcal { L } _ { \\mathrm { O E } } = \\mathbb { E } _ { D _ { \\mathrm { i n } } ^ { s } } ( \\operatorname* { m a x } ( 0 , S _ { \\mathrm { E n e r g y } } ( x , f ) - m _ { \\mathrm { i n } } ) ) ^ { 2 } + \\mathbb { E } _ { D _ { \\mathrm { o u t } } ^ { s } } ( \\operatorname* { m a x } ( 0 , m _ { \\mathrm { o u t } } - S _ { \\mathrm { E n e r g y } } ( x , f ) ) ) ^ { 2 }", + "type": "interline_equation", + "image_path": "54a9f407cbd94084751fcdf26413676f1b8a4a35835da9c373810c40b9ce2039.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 128, + 256, + 465, + 272 + ], + "spans": [], + "index": 13 + } + ] + }, + { + "type": "text", + "bbox": [ + 108, + 276, + 423, + 289 + ], + "lines": [ + { + "bbox": [ + 106, + 276, + 422, + 291 + ], + "spans": [ + { + "bbox": [ + 106, + 276, + 307, + 291 + ], + "score": 1.0, + "content": "We keep the thresholds same to (Liu et al., 2020):", + "type": "text" + }, + { + "bbox": [ + 307, + 277, + 362, + 289 + ], + "score": 0.76, + "content": "m _ { \\mathrm { i n } } = - 2 5 . 0", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 276, + 365, + 291 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 365, + 277, + 419, + 289 + ], + "score": 0.81, + "content": "m _ { \\mathrm { o u t } } = - 7 . 0", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 276, + 422, + 291 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14, + "bbox_fs": [ + 106, + 276, + 422, + 291 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 301, + 504, + 347 + ], + "lines": [ + { + "bbox": [ + 104, + 301, + 506, + 315 + ], + "spans": [ + { + "bbox": [ + 104, + 301, + 506, + 315 + ], + "score": 1.0, + "content": "POEM. (Ming et al., 2022) explores the Thompson sampling strategy (Thompson, 1933) to make", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 313, + 505, + 325 + ], + "spans": [ + { + "bbox": [ + 106, + 313, + 505, + 325 + ], + "score": 1.0, + "content": "the most use of outliers to learn a tight decision boundary. Though given the POEM’s nature to be", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 324, + 506, + 337 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 327, + 337 + ], + "score": 1.0, + "content": "orthogonal to other OE methods, we use the Energy(w.", + "type": "text" + }, + { + "bbox": [ + 328, + 324, + 348, + 335 + ], + "score": 0.88, + "content": "D _ { \\mathrm { a u x . } }", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 324, + 506, + 337 + ], + "score": 1.0, + "content": ") as the backbone, which is the same as", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 335, + 488, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 335, + 488, + 347 + ], + "score": 1.0, + "content": "Eq.( 10) in Liu et al. (2020). The details of Thompson sampling can refer to Ming et al. (2022).", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 16.5, + "bbox_fs": [ + 104, + 301, + 506, + 347 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 358, + 505, + 415 + ], + "lines": [ + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "score": 1.0, + "content": "Detailed formulations of FPR and TPR. Suppose we have a binary classification task (to predict", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 370, + 505, + 382 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 505, + 382 + ], + "score": 1.0, + "content": "an image to be an ID or OOD sample in this paper). There are two possible outputs: a positive result", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 380, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 106, + 380, + 506, + 394 + ], + "score": 1.0, + "content": "(the model predicts an image to be an ID sample); a negative result (the model predicts an image to", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 392, + 506, + 404 + ], + "spans": [ + { + "bbox": [ + 106, + 392, + 506, + 404 + ], + "score": 1.0, + "content": "be an OOD sample). Since we have two possible labels and two possible outputs, we can form a", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 402, + 322, + 416 + ], + "spans": [ + { + "bbox": [ + 105, + 402, + 322, + 416 + ], + "score": 1.0, + "content": "confusion matrix with all possible outputs as follows.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 21, + "bbox_fs": [ + 105, + 358, + 506, + 416 + ] + }, + { + "type": "table", + "bbox": [], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 250, + 425, + 360, + 437 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 249, + 424, + 361, + 438 + ], + "spans": [ + { + "bbox": [ + 249, + 424, + 361, + 438 + ], + "score": 1.0, + "content": "Table 3: Confusion Matrix.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "interline_equation", + "bbox": [ + 185, + 441, + 422, + 488 + ], + "lines": [ + { + "bbox": [ + 185, + 441, + 422, + 488 + ], + "spans": [ + { + "bbox": [ + 185, + 441, + 422, + 488 + ], + "score": 0.36, + "content": "\\frac { | \\mathrm { ~ { T r u t h : I D } ~ } | \\mathrm { { T r u t h : O O D } } } { \\mathrm { { P r e d i c t : ~ O O D ~ | \\mathrm { ~ { ~ T r u e ~ P o s i t i v e ~ ( T P ) ~ } | { \\mathrm { ~ F a l s e ~ P o s i t i v e ~ ( F P ) ~ } } } } } }", + "type": "interline_equation", + "image_path": "ca2f2a2fa96503d7c6f7f24da26223a214941e13ae665bd918c6f60ee89e3b68.jpg" + } + ] + } + ], + "index": 26, + "virtual_lines": [ + { + "bbox": [ + 185, + 441, + 422, + 456.6666666666667 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 185, + 456.6666666666667, + 422, + 472.33333333333337 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 185, + 472.33333333333337, + 422, + 488.00000000000006 + ], + "spans": [], + "index": 27 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 498, + 330, + 510 + ], + "lines": [ + { + "bbox": [ + 106, + 497, + 333, + 510 + ], + "spans": [ + { + "bbox": [ + 106, + 497, + 333, + 510 + ], + "score": 1.0, + "content": "Therefore, the false positive rate (FPR) is calculated as :", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28, + "bbox_fs": [ + 106, + 497, + 333, + 510 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 262, + 515, + 348, + 540 + ], + "lines": [ + { + "bbox": [ + 262, + 515, + 348, + 540 + ], + "spans": [ + { + "bbox": [ + 262, + 515, + 348, + 540 + ], + "score": 0.95, + "content": "F P R = { \\frac { F P } { F P + T N } }", + "type": "interline_equation", + "image_path": "6de9bd865babdf43ef851a8d2f51cca43a0e5ddbbc75bd8c1eb583ef265e7786.jpg" + } + ] + } + ], + "index": 29, + "virtual_lines": [ + { + "bbox": [ + 262, + 515, + 348, + 540 + ], + "spans": [], + "index": 29 + } + ] + }, + { + "type": "text", + "bbox": [ + 108, + 551, + 285, + 563 + ], + "lines": [ + { + "bbox": [ + 106, + 552, + 285, + 563 + ], + "spans": [ + { + "bbox": [ + 106, + 552, + 285, + 563 + ], + "score": 1.0, + "content": "The true positive rate (TPR) is calculated as:", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30, + "bbox_fs": [ + 106, + 552, + 285, + 563 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 263, + 568, + 348, + 594 + ], + "lines": [ + { + "bbox": [ + 263, + 568, + 348, + 594 + ], + "spans": [ + { + "bbox": [ + 263, + 568, + 348, + 594 + ], + "score": 0.95, + "content": "T P R = { \\frac { T P } { T P + F N } }", + "type": "interline_equation", + "image_path": "1175e2c392fa210f94055024b476db4a2fb3df5a12672ac7bfe9beb3d5388cde.jpg" + } + ] + } + ], + "index": 31, + "virtual_lines": [ + { + "bbox": [ + 263, + 568, + 348, + 594 + ], + "spans": [], + "index": 31 + } + ] + }, + { + "type": "title", + "bbox": [ + 105, + 608, + 492, + 620 + ], + "lines": [ + { + "bbox": [ + 106, + 608, + 493, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 117, + 619 + ], + "score": 1.0, + "content": "B", + "type": "text" + }, + { + "bbox": [ + 128, + 609, + 493, + 619 + ], + "score": 1.0, + "content": "ADDITIONAL EXPLANATION TOWARDS MINING THE ATYPICAL SAMPLES", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 32 + }, + { + "type": "text", + "bbox": [ + 106, + 632, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 632, + 506, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 506, + 645 + ], + "score": 1.0, + "content": "First, for identifying those atypical samples using a layer-wise mask with the well-pre-trained model,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 644, + 506, + 655 + ], + "spans": [ + { + "bbox": [ + 106, + 644, + 506, + 655 + ], + "score": 1.0, + "content": "the core intuition behind is constructing the parameter-level discrepancy to mine the atypical samples.", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 654, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 505, + 667 + ], + "score": 1.0, + "content": "It is inspired by and based on the evidence drawn from previous literature about learning behaviors", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 664, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 664, + 505, + 678 + ], + "score": 1.0, + "content": "(Arpit et al., 2017; Goodfellow et al., 2016) of deep neural networks (DNNs) and sparse representation", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 677, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 505, + 689 + ], + "score": 1.0, + "content": "(Frankle & Carbin, 2019; Goodfellow et al., 2013; Barham et al., 2022). To be specific, the atypical", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "samples tend to be learned by the DNNs later than those typical samples (Arpit et al., 2017), and are", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 699, + 505, + 710 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 505, + 710 + ], + "score": 1.0, + "content": "relatively more sensitive to the changes of the model parameter as the model does not generalize well", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "on that. By the layer-wise mask, the constructed discrepancy can make the model misclassify the", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "atypical samples and estimate loss constraint for the forgetting objective, as visualized in Figure 3(c).", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 37, + "bbox_fs": [ + 105, + 632, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 171 + ], + "lines": [ + { + "bbox": [ + 106, + 83, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 106, + 83, + 505, + 93 + ], + "score": 1.0, + "content": "Second, introducing the layer-wise mask has several advantages for achieving the staged target of", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 93, + 506, + 105 + ], + "spans": [ + { + "bbox": [ + 105, + 93, + 506, + 105 + ], + "score": 1.0, + "content": "mining atypical samples in our proposed method, while we would also admit that the layer-wise mask", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 505, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 505, + 117 + ], + "score": 1.0, + "content": "is not an irreplaceable option or maybe not optimal. On the one hand, considering that the model has", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 116, + 505, + 126 + ], + "spans": [ + { + "bbox": [ + 106, + 116, + 505, + 126 + ], + "score": 1.0, + "content": "been trained to approach the zero error on training data, utilizing the layer-wise mask is an integrated", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 127, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 105, + 127, + 505, + 138 + ], + "score": 1.0, + "content": "strategy to 1) figure out the atypical samples and 2) obtain the loss value computed by the masked", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 137, + 505, + 149 + ], + "spans": [ + { + "bbox": [ + 106, + 137, + 505, + 149 + ], + "score": 1.0, + "content": "output that misclassifies them. The loss constraint is later used in the forgetting objective to fine-tune", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 149, + 505, + 160 + ], + "spans": [ + { + "bbox": [ + 106, + 149, + 505, + 160 + ], + "score": 1.0, + "content": "the model. On the other hand, the layer-wise mask is also compatible with the proposed UMAP to", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 159, + 441, + 171 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 441, + 171 + ], + "score": 1.0, + "content": "generate a flexible mask for restoring the detection capability of the original model.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 107, + 176, + 505, + 253 + ], + "lines": [ + { + "bbox": [ + 105, + 175, + 506, + 187 + ], + "spans": [ + { + "bbox": [ + 105, + 175, + 506, + 187 + ], + "score": 1.0, + "content": "Third, we also adopt the unit/weight mask and visualize the misclassified samples in Figure 12. We", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 187, + 506, + 199 + ], + "spans": [ + { + "bbox": [ + 106, + 187, + 506, + 199 + ], + "score": 1.0, + "content": "think they can also be used to mine the atypical samples and can be extended or improved to be a", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 198, + 505, + 209 + ], + "spans": [ + { + "bbox": [ + 106, + 198, + 505, + 209 + ], + "score": 1.0, + "content": "more flexible choice. Further investigating the specific effect of different methods that construct the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 209, + 505, + 220 + ], + "spans": [ + { + "bbox": [ + 105, + 209, + 505, + 220 + ], + "score": 1.0, + "content": "parameter-level discrepancy would be an interesting sub-topic in future work. For the value of CE", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 220, + 504, + 231 + ], + "spans": [ + { + "bbox": [ + 106, + 220, + 504, + 231 + ], + "score": 1.0, + "content": "loss, although the atypical samples tend to have high CE loss value, they are already memorized", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 232, + 505, + 242 + ], + "spans": [ + { + "bbox": [ + 106, + 232, + 505, + 242 + ], + "score": 1.0, + "content": "and correctly classified as indicated by the zero training error. Only using the high CE error can not", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 241, + 449, + 254 + ], + "spans": [ + { + "bbox": [ + 105, + 241, + 449, + 254 + ], + "score": 1.0, + "content": "provide the loss estimation when the model does not correctly classify those samples.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 11 + }, + { + "type": "title", + "bbox": [ + 106, + 275, + 458, + 286 + ], + "lines": [ + { + "bbox": [ + 105, + 272, + 461, + 288 + ], + "spans": [ + { + "bbox": [ + 105, + 272, + 461, + 288 + ], + "score": 1.0, + "content": "C CONCEPTUAL AND EMPIRICAL COMPARISON WITH OVERFITTING", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 107, + 302, + 505, + 390 + ], + "lines": [ + { + "bbox": [ + 105, + 302, + 507, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 302, + 507, + 314 + ], + "score": 1.0, + "content": "First of all, we would refer to the concept of the conventional overfitting (Goodfellow et al., 2016;", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 312, + 505, + 325 + ], + "spans": [ + { + "bbox": [ + 106, + 312, + 505, + 325 + ], + "score": 1.0, + "content": "Belkin et al., 2019), i.e., the model \"overfit\" the training data but fail to generalize and perform well", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 324, + 505, + 336 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 505, + 336 + ], + "score": 1.0, + "content": "on the test data that is unseen during training. The common empirical reflection of overfitting is that", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 107, + 336, + 504, + 345 + ], + "spans": [ + { + "bbox": [ + 107, + 336, + 504, + 345 + ], + "score": 1.0, + "content": "the training error is decreasing while the test error is increasing at the same time, which enlarges the", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 346, + 505, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 346, + 505, + 358 + ], + "score": 1.0, + "content": "generalization gap of the model. It has been empirically confirmed not the case in our observation", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 357, + 505, + 368 + ], + "spans": [ + { + "bbox": [ + 106, + 357, + 505, + 368 + ], + "score": 1.0, + "content": "as observed in Figure 2(a) and 2(b). To be specific, for the original classification task, there is no", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 367, + 507, + 381 + ], + "spans": [ + { + "bbox": [ + 105, + 367, + 507, + 381 + ], + "score": 1.0, + "content": "conventional overfitting observed as the test performance is still improved at the later training stage,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 378, + 428, + 390 + ], + "spans": [ + { + "bbox": [ + 106, + 378, + 428, + 390 + ], + "score": 1.0, + "content": "which is a general pursuit of the model development phase on the original tasks.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 19.5 + }, + { + "type": "text", + "bbox": [ + 107, + 395, + 505, + 484 + ], + "lines": [ + { + "bbox": [ + 105, + 394, + 505, + 408 + ], + "spans": [ + { + "bbox": [ + 105, + 394, + 505, + 408 + ], + "score": 1.0, + "content": "Then, when we consider the OOD detection performance of the well-pretrained model, our unique", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 406, + 506, + 419 + ], + "spans": [ + { + "bbox": [ + 106, + 406, + 506, + 419 + ], + "score": 1.0, + "content": "observation is about the inconsistency between gaining better OOD detection capability and pursuing", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 416, + 506, + 430 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 506, + 430 + ], + "score": 1.0, + "content": "better performance on the original classification task for the in-distribution (ID) data. It is worth noting", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 429, + 505, + 440 + ], + "spans": [ + { + "bbox": [ + 106, + 429, + 505, + 440 + ], + "score": 1.0, + "content": "that here the training task is not the binary classification of OOD detection, but the classification task", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 439, + 505, + 451 + ], + "spans": [ + { + "bbox": [ + 105, + 439, + 505, + 451 + ], + "score": 1.0, + "content": "on ID data. It is out of the rigorous concept of the conventional overfitting and has received limited", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 450, + 506, + 463 + ], + "spans": [ + { + "bbox": [ + 105, + 450, + 506, + 463 + ], + "score": 1.0, + "content": "focus and discussion in the previous literature about OOD detection to the best of our knowledge.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 461, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 106, + 461, + 505, + 474 + ], + "score": 1.0, + "content": "Considering the practical scenario that exists target-level discrepancy, our revealed observation may", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 472, + 399, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 472, + 399, + 484 + ], + "score": 1.0, + "content": "encourage us to revisit the detection capability of the well-trained model.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 107, + 489, + 505, + 599 + ], + "lines": [ + { + "bbox": [ + 106, + 489, + 505, + 501 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 505, + 501 + ], + "score": 1.0, + "content": "Third, through empirical observation, those strategies designed for preventing the conventional", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 500, + 506, + 512 + ], + "spans": [ + { + "bbox": [ + 105, + 500, + 506, + 512 + ], + "score": 1.0, + "content": "overfitting may need to change the target to the OOD detection based on the important observation.", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 511, + 505, + 523 + ], + "spans": [ + { + "bbox": [ + 106, + 511, + 505, + 523 + ], + "score": 1.0, + "content": "In our experiments, for all the baseline models including that used in Figure 1, we have adopted", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 522, + 505, + 535 + ], + "spans": [ + { + "bbox": [ + 106, + 522, + 505, + 535 + ], + "score": 1.0, + "content": "those strategies (Srivastava et al., 2014; Hastie et al., 2009) (e.g., drop-out, weight decay) to reduce", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 533, + 505, + 545 + ], + "spans": [ + { + "bbox": [ + 105, + 533, + 505, + 545 + ], + "score": 1.0, + "content": "overfitting. It is found to be not enough to restore the OOD detection performance. For another shared", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 545, + 504, + 556 + ], + "spans": [ + { + "bbox": [ + 106, + 545, + 504, + 556 + ], + "score": 1.0, + "content": "issue, on the CIFAR-100 dataset, our UM restore the OOD detection capability of the well-trained", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 554, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 106, + 554, + 505, + 567 + ], + "score": 1.0, + "content": "model with a significant sacrifice on \"ID-ACC\". Using those strategies for reducing overfitting in", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 565, + 506, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 506, + 578 + ], + "score": 1.0, + "content": "the model development phase maybe not be acceptable to the users that it achieves such a lower", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 577, + 505, + 590 + ], + "spans": [ + { + "bbox": [ + 106, + 577, + 505, + 590 + ], + "score": 1.0, + "content": "performance on the original task. In contrast, our proposed UMAP can be a more practical and", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 588, + 291, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 291, + 600 + ], + "score": 1.0, + "content": "flexible way to restore detection performance.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 36.5 + }, + { + "type": "text", + "bbox": [ + 107, + 605, + 505, + 682 + ], + "lines": [ + { + "bbox": [ + 105, + 603, + 506, + 618 + ], + "spans": [ + { + "bbox": [ + 105, + 603, + 506, + 618 + ], + "score": 1.0, + "content": "We conduct the extra comparisons between our UM and UMAP with those methods for reducing", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 615, + 506, + 628 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 506, + 628 + ], + "score": 1.0, + "content": "overfitting. The results are summarized in the following Tables 4, 5, 6 and 7. According to our extra", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 627, + 505, + 639 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 505, + 639 + ], + "score": 1.0, + "content": "experiments, most conventional methods proposed to prevent conventional overfitting show limited", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 639, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 106, + 639, + 505, + 650 + ], + "score": 1.0, + "content": "benefits on gaining better OOD detection performance. Based on our important observation, the", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 648, + 506, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 506, + 662 + ], + "score": 1.0, + "content": "effective criterion, i.e., early stopping, also need to change its validation target to be the OOD data.", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 106, + 660, + 505, + 671 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 505, + 671 + ], + "score": 1.0, + "content": "However, most of them suffer from higher sacrifice on the performance of the original task and maybe", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 671, + 500, + 682 + ], + "spans": [ + { + "bbox": [ + 105, + 671, + 500, + 682 + ], + "score": 1.0, + "content": "not compatible and practical in the current general setting, i.e., starting from a well-trained model.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 45 + }, + { + "type": "text", + "bbox": [ + 108, + 688, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "Given the concept discrepancy aforementioned, one conclusive message is that \"memorization of", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "the atypical samples\" are not \"memorization in overfitting\". Those atypical samples are empirically", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "beneficial in improving the performance on the original classification task as shown in Figure 2.", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 106, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "However, this part of knowledge is not very necessary and even harmful to the OOD detection task as", + "type": "text" + } + ], + "index": 52 + } + ], + "index": 50.5 + } + ], + "page_idx": 14, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 171 + ], + "lines": [ + { + "bbox": [ + 106, + 83, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 106, + 83, + 505, + 93 + ], + "score": 1.0, + "content": "Second, introducing the layer-wise mask has several advantages for achieving the staged target of", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 93, + 506, + 105 + ], + "spans": [ + { + "bbox": [ + 105, + 93, + 506, + 105 + ], + "score": 1.0, + "content": "mining atypical samples in our proposed method, while we would also admit that the layer-wise mask", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 505, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 505, + 117 + ], + "score": 1.0, + "content": "is not an irreplaceable option or maybe not optimal. On the one hand, considering that the model has", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 116, + 505, + 126 + ], + "spans": [ + { + "bbox": [ + 106, + 116, + 505, + 126 + ], + "score": 1.0, + "content": "been trained to approach the zero error on training data, utilizing the layer-wise mask is an integrated", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 127, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 105, + 127, + 505, + 138 + ], + "score": 1.0, + "content": "strategy to 1) figure out the atypical samples and 2) obtain the loss value computed by the masked", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 137, + 505, + 149 + ], + "spans": [ + { + "bbox": [ + 106, + 137, + 505, + 149 + ], + "score": 1.0, + "content": "output that misclassifies them. The loss constraint is later used in the forgetting objective to fine-tune", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 149, + 505, + 160 + ], + "spans": [ + { + "bbox": [ + 106, + 149, + 505, + 160 + ], + "score": 1.0, + "content": "the model. On the other hand, the layer-wise mask is also compatible with the proposed UMAP to", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 159, + 441, + 171 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 441, + 171 + ], + "score": 1.0, + "content": "generate a flexible mask for restoring the detection capability of the original model.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 3.5, + "bbox_fs": [ + 105, + 83, + 506, + 171 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 176, + 505, + 253 + ], + "lines": [ + { + "bbox": [ + 105, + 175, + 506, + 187 + ], + "spans": [ + { + "bbox": [ + 105, + 175, + 506, + 187 + ], + "score": 1.0, + "content": "Third, we also adopt the unit/weight mask and visualize the misclassified samples in Figure 12. We", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 187, + 506, + 199 + ], + "spans": [ + { + "bbox": [ + 106, + 187, + 506, + 199 + ], + "score": 1.0, + "content": "think they can also be used to mine the atypical samples and can be extended or improved to be a", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 198, + 505, + 209 + ], + "spans": [ + { + "bbox": [ + 106, + 198, + 505, + 209 + ], + "score": 1.0, + "content": "more flexible choice. Further investigating the specific effect of different methods that construct the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 209, + 505, + 220 + ], + "spans": [ + { + "bbox": [ + 105, + 209, + 505, + 220 + ], + "score": 1.0, + "content": "parameter-level discrepancy would be an interesting sub-topic in future work. For the value of CE", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 220, + 504, + 231 + ], + "spans": [ + { + "bbox": [ + 106, + 220, + 504, + 231 + ], + "score": 1.0, + "content": "loss, although the atypical samples tend to have high CE loss value, they are already memorized", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 232, + 505, + 242 + ], + "spans": [ + { + "bbox": [ + 106, + 232, + 505, + 242 + ], + "score": 1.0, + "content": "and correctly classified as indicated by the zero training error. Only using the high CE error can not", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 241, + 449, + 254 + ], + "spans": [ + { + "bbox": [ + 105, + 241, + 449, + 254 + ], + "score": 1.0, + "content": "provide the loss estimation when the model does not correctly classify those samples.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 175, + 506, + 254 + ] + }, + { + "type": "title", + "bbox": [ + 106, + 275, + 458, + 286 + ], + "lines": [ + { + "bbox": [ + 105, + 272, + 461, + 288 + ], + "spans": [ + { + "bbox": [ + 105, + 272, + 461, + 288 + ], + "score": 1.0, + "content": "C CONCEPTUAL AND EMPIRICAL COMPARISON WITH OVERFITTING", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 107, + 302, + 505, + 390 + ], + "lines": [ + { + "bbox": [ + 105, + 302, + 507, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 302, + 507, + 314 + ], + "score": 1.0, + "content": "First of all, we would refer to the concept of the conventional overfitting (Goodfellow et al., 2016;", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 312, + 505, + 325 + ], + "spans": [ + { + "bbox": [ + 106, + 312, + 505, + 325 + ], + "score": 1.0, + "content": "Belkin et al., 2019), i.e., the model \"overfit\" the training data but fail to generalize and perform well", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 324, + 505, + 336 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 505, + 336 + ], + "score": 1.0, + "content": "on the test data that is unseen during training. The common empirical reflection of overfitting is that", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 107, + 336, + 504, + 345 + ], + "spans": [ + { + "bbox": [ + 107, + 336, + 504, + 345 + ], + "score": 1.0, + "content": "the training error is decreasing while the test error is increasing at the same time, which enlarges the", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 346, + 505, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 346, + 505, + 358 + ], + "score": 1.0, + "content": "generalization gap of the model. It has been empirically confirmed not the case in our observation", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 357, + 505, + 368 + ], + "spans": [ + { + "bbox": [ + 106, + 357, + 505, + 368 + ], + "score": 1.0, + "content": "as observed in Figure 2(a) and 2(b). To be specific, for the original classification task, there is no", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 367, + 507, + 381 + ], + "spans": [ + { + "bbox": [ + 105, + 367, + 507, + 381 + ], + "score": 1.0, + "content": "conventional overfitting observed as the test performance is still improved at the later training stage,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 378, + 428, + 390 + ], + "spans": [ + { + "bbox": [ + 106, + 378, + 428, + 390 + ], + "score": 1.0, + "content": "which is a general pursuit of the model development phase on the original tasks.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 19.5, + "bbox_fs": [ + 105, + 302, + 507, + 390 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 395, + 505, + 484 + ], + "lines": [ + { + "bbox": [ + 105, + 394, + 505, + 408 + ], + "spans": [ + { + "bbox": [ + 105, + 394, + 505, + 408 + ], + "score": 1.0, + "content": "Then, when we consider the OOD detection performance of the well-pretrained model, our unique", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 406, + 506, + 419 + ], + "spans": [ + { + "bbox": [ + 106, + 406, + 506, + 419 + ], + "score": 1.0, + "content": "observation is about the inconsistency between gaining better OOD detection capability and pursuing", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 416, + 506, + 430 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 506, + 430 + ], + "score": 1.0, + "content": "better performance on the original classification task for the in-distribution (ID) data. It is worth noting", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 429, + 505, + 440 + ], + "spans": [ + { + "bbox": [ + 106, + 429, + 505, + 440 + ], + "score": 1.0, + "content": "that here the training task is not the binary classification of OOD detection, but the classification task", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 439, + 505, + 451 + ], + "spans": [ + { + "bbox": [ + 105, + 439, + 505, + 451 + ], + "score": 1.0, + "content": "on ID data. It is out of the rigorous concept of the conventional overfitting and has received limited", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 450, + 506, + 463 + ], + "spans": [ + { + "bbox": [ + 105, + 450, + 506, + 463 + ], + "score": 1.0, + "content": "focus and discussion in the previous literature about OOD detection to the best of our knowledge.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 461, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 106, + 461, + 505, + 474 + ], + "score": 1.0, + "content": "Considering the practical scenario that exists target-level discrepancy, our revealed observation may", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 472, + 399, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 472, + 399, + 484 + ], + "score": 1.0, + "content": "encourage us to revisit the detection capability of the well-trained model.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 27.5, + "bbox_fs": [ + 105, + 394, + 506, + 484 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 489, + 505, + 599 + ], + "lines": [ + { + "bbox": [ + 106, + 489, + 505, + 501 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 505, + 501 + ], + "score": 1.0, + "content": "Third, through empirical observation, those strategies designed for preventing the conventional", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 500, + 506, + 512 + ], + "spans": [ + { + "bbox": [ + 105, + 500, + 506, + 512 + ], + "score": 1.0, + "content": "overfitting may need to change the target to the OOD detection based on the important observation.", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 511, + 505, + 523 + ], + "spans": [ + { + "bbox": [ + 106, + 511, + 505, + 523 + ], + "score": 1.0, + "content": "In our experiments, for all the baseline models including that used in Figure 1, we have adopted", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 522, + 505, + 535 + ], + "spans": [ + { + "bbox": [ + 106, + 522, + 505, + 535 + ], + "score": 1.0, + "content": "those strategies (Srivastava et al., 2014; Hastie et al., 2009) (e.g., drop-out, weight decay) to reduce", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 533, + 505, + 545 + ], + "spans": [ + { + "bbox": [ + 105, + 533, + 505, + 545 + ], + "score": 1.0, + "content": "overfitting. It is found to be not enough to restore the OOD detection performance. For another shared", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 545, + 504, + 556 + ], + "spans": [ + { + "bbox": [ + 106, + 545, + 504, + 556 + ], + "score": 1.0, + "content": "issue, on the CIFAR-100 dataset, our UM restore the OOD detection capability of the well-trained", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 554, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 106, + 554, + 505, + 567 + ], + "score": 1.0, + "content": "model with a significant sacrifice on \"ID-ACC\". Using those strategies for reducing overfitting in", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 565, + 506, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 506, + 578 + ], + "score": 1.0, + "content": "the model development phase maybe not be acceptable to the users that it achieves such a lower", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 577, + 505, + 590 + ], + "spans": [ + { + "bbox": [ + 106, + 577, + 505, + 590 + ], + "score": 1.0, + "content": "performance on the original task. In contrast, our proposed UMAP can be a more practical and", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 588, + 291, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 291, + 600 + ], + "score": 1.0, + "content": "flexible way to restore detection performance.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 36.5, + "bbox_fs": [ + 105, + 489, + 506, + 600 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 605, + 505, + 682 + ], + "lines": [ + { + "bbox": [ + 105, + 603, + 506, + 618 + ], + "spans": [ + { + "bbox": [ + 105, + 603, + 506, + 618 + ], + "score": 1.0, + "content": "We conduct the extra comparisons between our UM and UMAP with those methods for reducing", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 615, + 506, + 628 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 506, + 628 + ], + "score": 1.0, + "content": "overfitting. The results are summarized in the following Tables 4, 5, 6 and 7. According to our extra", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 627, + 505, + 639 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 505, + 639 + ], + "score": 1.0, + "content": "experiments, most conventional methods proposed to prevent conventional overfitting show limited", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 639, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 106, + 639, + 505, + 650 + ], + "score": 1.0, + "content": "benefits on gaining better OOD detection performance. Based on our important observation, the", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 648, + 506, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 506, + 662 + ], + "score": 1.0, + "content": "effective criterion, i.e., early stopping, also need to change its validation target to be the OOD data.", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 106, + 660, + 505, + 671 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 505, + 671 + ], + "score": 1.0, + "content": "However, most of them suffer from higher sacrifice on the performance of the original task and maybe", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 671, + 500, + 682 + ], + "spans": [ + { + "bbox": [ + 105, + 671, + 500, + 682 + ], + "score": 1.0, + "content": "not compatible and practical in the current general setting, i.e., starting from a well-trained model.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 45, + "bbox_fs": [ + 105, + 603, + 506, + 682 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 688, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "Given the concept discrepancy aforementioned, one conclusive message is that \"memorization of", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "the atypical samples\" are not \"memorization in overfitting\". Those atypical samples are empirically", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "beneficial in improving the performance on the original classification task as shown in Figure 2.", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 106, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "However, this part of knowledge is not very necessary and even harmful to the OOD detection task as", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 106, + 631, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 631, + 505, + 645 + ], + "score": 1.0, + "content": "the detection performance of the model is drop significantly. Based on the training and test curves", + "type": "text", + "cross_page": true + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 642, + 505, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 505, + 657 + ], + "score": 1.0, + "content": "in our observation, the memorization in overfitting is expected to happen later than the final stage", + "type": "text", + "cross_page": true + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "score": 1.0, + "content": "in which the test performance would drop. Since we have already used some strategies to prevent", + "type": "text", + "cross_page": true + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 666, + 505, + 677 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 505, + 677 + ], + "score": 1.0, + "content": "overfitting, it does not exist. Intuitively, the \"atypical samples\" identified in our work are relative to", + "type": "text", + "cross_page": true + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 677, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 505, + 689 + ], + "score": 1.0, + "content": "the OOD detection task. The memorization of \"atypical samples\" indicates that the model may not be", + "type": "text", + "cross_page": true + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 686, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 686, + 505, + 700 + ], + "score": 1.0, + "content": "able to draw the general information of the ID distribution through further learning on those atypical", + "type": "text", + "cross_page": true + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "samples through the original classification task Since we mainly provide the empirical observation", + "type": "text", + "cross_page": true + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "and understanding of the proposed algorithm in this work, further analysis from other views or", + "type": "text", + "cross_page": true + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 721, + 384, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 384, + 732 + ], + "score": 1.0, + "content": "theoretically would be an interesting and a major part of future work.", + "type": "text", + "cross_page": true + } + ], + "index": 23 + } + ], + "index": 50.5, + "bbox_fs": [ + 106, + 687, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 140, + 109, + 468, + 244 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 503, + 103 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 437, + 93 + ], + "score": 1.0, + "content": "Table 4: Comparison among overfitting methods and ODIN with DenseNet-101", + "type": "text" + }, + { + "bbox": [ + 437, + 80, + 451, + 92 + ], + "score": 0.56, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 80, + 457, + 93 + ], + "score": 1.0, + "content": ").", + "type": "text" + }, + { + "bbox": [ + 458, + 80, + 466, + 92 + ], + "score": 0.44, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 80, + 505, + 93 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 90, + 361, + 103 + ], + "spans": [ + { + "bbox": [ + 106, + 90, + 220, + 103 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 92, + 228, + 102 + ], + "score": 0.56, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 90, + 361, + 103 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 140, + 109, + 468, + 244 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 140, + 109, + 468, + 244 + ], + "spans": [ + { + "bbox": [ + 140, + 109, + 468, + 244 + ], + "score": 0.984, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline91.6791.8940.7493.67
Early Stopping w. ACC92.1392.4638.8693.69
Early Stopping w. OOD92.9593.2635.2393.18
Weight Decay 0.186.6486.6760.0788.53
Weight Decay 0.0190.7691.2544.2092.07
Weight Decay 0.00188.9388.2548.9594.26
Drop Rate 0.391.1492.2146.5890.05
Drop Rate 0.484.9586.6262.5282.55
Drop Rate 0.583.7585.1762.1775.31
UM (ours)92.4593.0637.1392.76
UMAP (ours)91.9292.8837.6993.69
", + "type": "table", + "image_path": "f430d358f55a03ab246fefbad236a7c78ca3462d7a9d197666430a7d1e0c2f71.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 140, + 109, + 468, + 154.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 140, + 154.0, + 468, + 199.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 140, + 199.0, + 468, + 244.0 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "table", + "bbox": [ + 140, + 290, + 468, + 424 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 104, + 261, + 504, + 284 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 104, + 259, + 506, + 274 + ], + "spans": [ + { + "bbox": [ + 104, + 259, + 438, + 274 + ], + "score": 1.0, + "content": "Table 5: Comparison among overfitting methods and Energy with DenseNet-101", + "type": "text" + }, + { + "bbox": [ + 438, + 261, + 453, + 272 + ], + "score": 0.63, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 453, + 259, + 457, + 274 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 458, + 261, + 466, + 273 + ], + "score": 0.3, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 259, + 506, + 274 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 272, + 361, + 284 + ], + "spans": [ + { + "bbox": [ + 106, + 272, + 220, + 284 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 272, + 228, + 283 + ], + "score": 0.58, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 272, + 361, + 284 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + }, + { + "type": "table_body", + "bbox": [ + 140, + 290, + 468, + 424 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 140, + 290, + 468, + 424 + ], + "spans": [ + { + "bbox": [ + 140, + 290, + 468, + 424 + ], + "score": 0.985, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline92.7293.4838.3093.67
Early Stopping w. ACC92.7593.5437.8493.69
Early Stopping w. OOD93.0493.7836.5693.18
Weight Decay 0.186.7888.0465.0888.53
Weight Decay 0.0190.8691.7747.6492.07
Weight Decay 0.00190.6890.9047.3894.26
Drop Rate 0.390.5291.7951.2390.05
Drop Rate 0.484.2986.4368.1782.55
Drop Rate 0.583.2985.1468.1775.31
UM (ours)93.5894.1433.6692.76
UMAP (ours)93.1793.8736.1193.69
", + "type": "table", + "image_path": "1d838572164e14370718419838c8f7295c19a674645d6d63804cae4c92da9386.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 140, + 290, + 468, + 334.6666666666667 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 140, + 334.6666666666667, + 468, + 379.33333333333337 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 140, + 379.33333333333337, + 468, + 424.00000000000006 + ], + "spans": [], + "index": 9 + } + ] + } + ], + "index": 6.75 + }, + { + "type": "table", + "bbox": [ + 140, + 471, + 468, + 605 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 105, + 441, + 505, + 464 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 105, + 441, + 505, + 454 + ], + "spans": [ + { + "bbox": [ + 105, + 441, + 412, + 454 + ], + "score": 1.0, + "content": "Table 6: Comparison among overfitting methods and ODIN with WRN-40-4", + "type": "text" + }, + { + "bbox": [ + 412, + 442, + 427, + 453 + ], + "score": 0.65, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 427, + 441, + 431, + 454 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 432, + 442, + 439, + 453 + ], + "score": 0.38, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 439, + 441, + 505, + 454 + ], + "score": 1.0, + "content": "indicates higher", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 453, + 332, + 465 + ], + "spans": [ + { + "bbox": [ + 106, + 453, + 192, + 465 + ], + "score": 1.0, + "content": "values are better, and", + "type": "text" + }, + { + "bbox": [ + 192, + 453, + 200, + 464 + ], + "score": 0.47, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 453, + 332, + 465 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5 + }, + { + "type": "table_body", + "bbox": [ + 140, + 471, + 468, + 605 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 140, + 471, + 468, + 605 + ], + "spans": [ + { + "bbox": [ + 140, + 471, + 468, + 605 + ], + "score": 0.982, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline86.2485.9060.1393.86
Early Stopping w. ACC83.8083.3065.1393.99
Early Stopping w. OOD89.9290.9449.8791.47
Weight Decay 0.184.3884.7565.7589.88
Weight Decay 0.0188.0888.4555.1693.16
Weight Decay 0.00186.3486.3857.4294.91
Drop Rate 0.387.5387.2556.1294.22
Drop Rate 0.488.2488.4154.6294.20
Drop Rate 0.589.1389.9953.0793.91
UM (ours)89.6191.1350.9792.68
UMAP (ours)90.4391.7346.9693.86
", + "type": "table", + "image_path": "fb5b51d30e40d5894a6888f9d323b424b639886b2073340516d18dddee62178d.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 140, + 471, + 468, + 515.6666666666666 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 140, + 515.6666666666666, + 468, + 560.3333333333333 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 140, + 560.3333333333333, + 468, + 604.9999999999999 + ], + "spans": [], + "index": 14 + } + ] + } + ], + "index": 11.75 + }, + { + "type": "text", + "bbox": [ + 106, + 632, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 631, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 631, + 505, + 645 + ], + "score": 1.0, + "content": "the detection performance of the model is drop significantly. Based on the training and test curves", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 642, + 505, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 505, + 657 + ], + "score": 1.0, + "content": "in our observation, the memorization in overfitting is expected to happen later than the final stage", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "score": 1.0, + "content": "in which the test performance would drop. Since we have already used some strategies to prevent", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 666, + 505, + 677 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 505, + 677 + ], + "score": 1.0, + "content": "overfitting, it does not exist. Intuitively, the \"atypical samples\" identified in our work are relative to", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 677, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 505, + 689 + ], + "score": 1.0, + "content": "the OOD detection task. The memorization of \"atypical samples\" indicates that the model may not be", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 686, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 686, + 505, + 700 + ], + "score": 1.0, + "content": "able to draw the general information of the ID distribution through further learning on those atypical", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "samples through the original classification task Since we mainly provide the empirical observation", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "and understanding of the proposed algorithm in this work, further analysis from other views or", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 721, + 384, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 384, + 732 + ], + "score": 1.0, + "content": "theoretically would be an interesting and a major part of future work.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 19 + } + ], + "page_idx": 15, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 752, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 140, + 109, + 468, + 244 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 503, + 103 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 437, + 93 + ], + "score": 1.0, + "content": "Table 4: Comparison among overfitting methods and ODIN with DenseNet-101", + "type": "text" + }, + { + "bbox": [ + 437, + 80, + 451, + 92 + ], + "score": 0.56, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 80, + 457, + 93 + ], + "score": 1.0, + "content": ").", + "type": "text" + }, + { + "bbox": [ + 458, + 80, + 466, + 92 + ], + "score": 0.44, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 80, + 505, + 93 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 90, + 361, + 103 + ], + "spans": [ + { + "bbox": [ + 106, + 90, + 220, + 103 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 92, + 228, + 102 + ], + "score": 0.56, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 90, + 361, + 103 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 140, + 109, + 468, + 244 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 140, + 109, + 468, + 244 + ], + "spans": [ + { + "bbox": [ + 140, + 109, + 468, + 244 + ], + "score": 0.984, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline91.6791.8940.7493.67
Early Stopping w. ACC92.1392.4638.8693.69
Early Stopping w. OOD92.9593.2635.2393.18
Weight Decay 0.186.6486.6760.0788.53
Weight Decay 0.0190.7691.2544.2092.07
Weight Decay 0.00188.9388.2548.9594.26
Drop Rate 0.391.1492.2146.5890.05
Drop Rate 0.484.9586.6262.5282.55
Drop Rate 0.583.7585.1762.1775.31
UM (ours)92.4593.0637.1392.76
UMAP (ours)91.9292.8837.6993.69
", + "type": "table", + "image_path": "f430d358f55a03ab246fefbad236a7c78ca3462d7a9d197666430a7d1e0c2f71.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 140, + 109, + 468, + 154.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 140, + 154.0, + 468, + 199.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 140, + 199.0, + 468, + 244.0 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "table", + "bbox": [ + 140, + 290, + 468, + 424 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 104, + 261, + 504, + 284 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 104, + 259, + 506, + 274 + ], + "spans": [ + { + "bbox": [ + 104, + 259, + 438, + 274 + ], + "score": 1.0, + "content": "Table 5: Comparison among overfitting methods and Energy with DenseNet-101", + "type": "text" + }, + { + "bbox": [ + 438, + 261, + 453, + 272 + ], + "score": 0.63, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 453, + 259, + 457, + 274 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 458, + 261, + 466, + 273 + ], + "score": 0.3, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 259, + 506, + 274 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 272, + 361, + 284 + ], + "spans": [ + { + "bbox": [ + 106, + 272, + 220, + 284 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 272, + 228, + 283 + ], + "score": 0.58, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 272, + 361, + 284 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + }, + { + "type": "table_body", + "bbox": [ + 140, + 290, + 468, + 424 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 140, + 290, + 468, + 424 + ], + "spans": [ + { + "bbox": [ + 140, + 290, + 468, + 424 + ], + "score": 0.985, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline92.7293.4838.3093.67
Early Stopping w. ACC92.7593.5437.8493.69
Early Stopping w. OOD93.0493.7836.5693.18
Weight Decay 0.186.7888.0465.0888.53
Weight Decay 0.0190.8691.7747.6492.07
Weight Decay 0.00190.6890.9047.3894.26
Drop Rate 0.390.5291.7951.2390.05
Drop Rate 0.484.2986.4368.1782.55
Drop Rate 0.583.2985.1468.1775.31
UM (ours)93.5894.1433.6692.76
UMAP (ours)93.1793.8736.1193.69
", + "type": "table", + "image_path": "1d838572164e14370718419838c8f7295c19a674645d6d63804cae4c92da9386.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 140, + 290, + 468, + 334.6666666666667 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 140, + 334.6666666666667, + 468, + 379.33333333333337 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 140, + 379.33333333333337, + 468, + 424.00000000000006 + ], + "spans": [], + "index": 9 + } + ] + } + ], + "index": 6.75 + }, + { + "type": "table", + "bbox": [ + 140, + 471, + 468, + 605 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 105, + 441, + 505, + 464 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 105, + 441, + 505, + 454 + ], + "spans": [ + { + "bbox": [ + 105, + 441, + 412, + 454 + ], + "score": 1.0, + "content": "Table 6: Comparison among overfitting methods and ODIN with WRN-40-4", + "type": "text" + }, + { + "bbox": [ + 412, + 442, + 427, + 453 + ], + "score": 0.65, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 427, + 441, + 431, + 454 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 432, + 442, + 439, + 453 + ], + "score": 0.38, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 439, + 441, + 505, + 454 + ], + "score": 1.0, + "content": "indicates higher", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 453, + 332, + 465 + ], + "spans": [ + { + "bbox": [ + 106, + 453, + 192, + 465 + ], + "score": 1.0, + "content": "values are better, and", + "type": "text" + }, + { + "bbox": [ + 192, + 453, + 200, + 464 + ], + "score": 0.47, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 453, + 332, + 465 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5 + }, + { + "type": "table_body", + "bbox": [ + 140, + 471, + 468, + 605 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 140, + 471, + 468, + 605 + ], + "spans": [ + { + "bbox": [ + 140, + 471, + 468, + 605 + ], + "score": 0.982, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline86.2485.9060.1393.86
Early Stopping w. ACC83.8083.3065.1393.99
Early Stopping w. OOD89.9290.9449.8791.47
Weight Decay 0.184.3884.7565.7589.88
Weight Decay 0.0188.0888.4555.1693.16
Weight Decay 0.00186.3486.3857.4294.91
Drop Rate 0.387.5387.2556.1294.22
Drop Rate 0.488.2488.4154.6294.20
Drop Rate 0.589.1389.9953.0793.91
UM (ours)89.6191.1350.9792.68
UMAP (ours)90.4391.7346.9693.86
", + "type": "table", + "image_path": "fb5b51d30e40d5894a6888f9d323b424b639886b2073340516d18dddee62178d.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 140, + 471, + 468, + 515.6666666666666 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 140, + 515.6666666666666, + 468, + 560.3333333333333 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 140, + 560.3333333333333, + 468, + 604.9999999999999 + ], + "spans": [], + "index": 14 + } + ] + } + ], + "index": 11.75 + }, + { + "type": "text", + "bbox": [ + 106, + 632, + 505, + 732 + ], + "lines": [], + "index": 19, + "bbox_fs": [ + 105, + 631, + 506, + 732 + ], + "lines_deleted": true + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 140, + 109, + 468, + 244 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 504, + 102 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 413, + 93 + ], + "score": 1.0, + "content": "Table 7: Comparison among overfitting methods and Energy with WRN-40-4", + "type": "text" + }, + { + "bbox": [ + 413, + 80, + 428, + 92 + ], + "score": 0.68, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 80, + 432, + 93 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 432, + 81, + 440, + 92 + ], + "score": 0.55, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 440, + 80, + 505, + 93 + ], + "score": 1.0, + "content": "indicates higher", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 90, + 333, + 104 + ], + "spans": [ + { + "bbox": [ + 105, + 90, + 192, + 104 + ], + "score": 1.0, + "content": "values are better, and", + "type": "text" + }, + { + "bbox": [ + 192, + 92, + 200, + 102 + ], + "score": 0.47, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 90, + 333, + 104 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 140, + 109, + 468, + 244 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 140, + 109, + 468, + 244 + ], + "spans": [ + { + "bbox": [ + 140, + 109, + 468, + 244 + ], + "score": 0.984, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline87.6988.1658.4793.86
Early Stopping w. ACC88.0788.6567.6193.99
Early Stopping w. OOD90.9291.9446.6391.47
Weight Decay 0.186.9788.5163.5489.88
Weight Decay 0.0189.7789.8250.2393.16
Weight Decay 0.00189.2589.8450.9593.91
Drop Rate 0.389.7490.0752.1693.22
Drop Rate 0.489.9490.5351.1394.20
Drop Rate 0.590.0991.0452.7693.91
UM (ours)91.7492.6740.4092.68
UMAP (ours)88.8489.3150.2393.86
", + "type": "table", + "image_path": "6ddca4b36c8e630d21ddb3e64e00889c3562f9ae052ec1b20ad4eb06c9b47932.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 140, + 109, + 468, + 154.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 140, + 154.0, + 468, + 199.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 140, + 199.0, + 468, + 244.0 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "title", + "bbox": [ + 106, + 264, + 421, + 276 + ], + "lines": [ + { + "bbox": [ + 105, + 263, + 423, + 280 + ], + "spans": [ + { + "bbox": [ + 105, + 263, + 423, + 280 + ], + "score": 1.0, + "content": "D DETAILED REALIZATION OF THE PROPOSED ALGORITHMS", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 106, + 289, + 504, + 312 + ], + "lines": [ + { + "bbox": [ + 106, + 289, + 505, + 302 + ], + "spans": [ + { + "bbox": [ + 106, + 289, + 505, + 302 + ], + "score": 1.0, + "content": "In this section, we provide the detailed realization of our proposed Unleashing Mask (UM) (i.e., in", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 300, + 437, + 313 + ], + "spans": [ + { + "bbox": [ + 106, + 300, + 437, + 313 + ], + "score": 1.0, + "content": "Algorithm 1) and Unleashing Mask Adopt Pruning (UMAP) (i.e., in Algorithm 2).", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 107, + 318, + 505, + 407 + ], + "lines": [ + { + "bbox": [ + 105, + 317, + 506, + 332 + ], + "spans": [ + { + "bbox": [ + 105, + 317, + 233, + 332 + ], + "score": 1.0, + "content": "To estimate the loss constrain", + "type": "text" + }, + { + "bbox": [ + 234, + 319, + 241, + 331 + ], + "score": 0.79, + "content": "\\zeta", + "type": "inline_equation" + }, + { + "bbox": [ + 241, + 317, + 259, + 332 + ], + "score": 1.0, + "content": "(i.e.", + "type": "text" + }, + { + "bbox": [ + 259, + 317, + 320, + 331 + ], + "score": 0.88, + "content": "\\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } )", + "type": "inline_equation" + }, + { + "bbox": [ + 321, + 317, + 473, + 332 + ], + "score": 1.0, + "content": "in Eq 3 with the fixed given model", + "type": "text" + }, + { + "bbox": [ + 473, + 320, + 485, + 331 + ], + "score": 0.83, + "content": "f ^ { * }", + "type": "inline_equation" + }, + { + "bbox": [ + 486, + 317, + 506, + 332 + ], + "score": 1.0, + "content": ") for", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 330, + 505, + 342 + ], + "spans": [ + { + "bbox": [ + 105, + 330, + 482, + 342 + ], + "score": 1.0, + "content": "forgetting, we need to randomly knock out parts of weights according to the given mask ratio", + "type": "text" + }, + { + "bbox": [ + 482, + 331, + 488, + 340 + ], + "score": 0.68, + "content": "\\delta", + "type": "inline_equation" + }, + { + "bbox": [ + 488, + 330, + 505, + 342 + ], + "score": 1.0, + "content": ". To", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 342, + 506, + 353 + ], + "spans": [ + { + "bbox": [ + 106, + 342, + 506, + 353 + ], + "score": 1.0, + "content": "be specific, we sample a score from a Gaussian distribution for every weight. Then we initialize a", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 351, + 504, + 365 + ], + "spans": [ + { + "bbox": [ + 105, + 351, + 489, + 365 + ], + "score": 1.0, + "content": "unit matrix for every layer of the model concerning the size of the layer. We formulate the mask", + "type": "text" + }, + { + "bbox": [ + 489, + 354, + 504, + 363 + ], + "score": 0.82, + "content": "m _ { \\delta }", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 363, + 505, + 375 + ], + "spans": [ + { + "bbox": [ + 105, + 363, + 505, + 375 + ], + "score": 1.0, + "content": "according to the sampled scores. Find the threshold for each layer that is smaller than the score of the", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 374, + 506, + 386 + ], + "spans": [ + { + "bbox": [ + 105, + 374, + 506, + 386 + ], + "score": 1.0, + "content": "given mask ratio in that layer (termed as quantile). Then set all the ones, whose corresponding scores", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 385, + 504, + 397 + ], + "spans": [ + { + "bbox": [ + 105, + 385, + 497, + 397 + ], + "score": 1.0, + "content": "are more significant than the layers’ thresholds, to zeros. In our algorithms, the fine-tuning epochs", + "type": "text" + }, + { + "bbox": [ + 497, + 386, + 504, + 395 + ], + "score": 0.75, + "content": "k", + "type": "inline_equation" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 396, + 357, + 408 + ], + "spans": [ + { + "bbox": [ + 105, + 396, + 357, + 408 + ], + "score": 1.0, + "content": "is the epochs we fine-tune after we get the well-trained model.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 11.5 + }, + { + "type": "text", + "bbox": [ + 107, + 413, + 505, + 457 + ], + "lines": [ + { + "bbox": [ + 105, + 412, + 505, + 425 + ], + "spans": [ + { + "bbox": [ + 105, + 412, + 505, + 425 + ], + "score": 1.0, + "content": "We dot-multiply every layer’s weights with the formulated binary matrix as if we delete some parts", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 424, + 506, + 437 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 506, + 437 + ], + "score": 1.0, + "content": "of the weights. We input a batch of training samples to the masked model and treat the mean value of", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 435, + 505, + 447 + ], + "spans": [ + { + "bbox": [ + 106, + 435, + 505, + 447 + ], + "score": 1.0, + "content": "the outputs’ cross-entropy loss as the loss constraint. After all of these have been done, we begin to", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 446, + 455, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 455, + 459 + ], + "score": 1.0, + "content": "fine-tune the model’s weights with the loss constraint applied to the cross-entropy loss.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 17.5 + }, + { + "type": "text", + "bbox": [ + 107, + 462, + 505, + 518 + ], + "lines": [ + { + "bbox": [ + 105, + 463, + 506, + 475 + ], + "spans": [ + { + "bbox": [ + 105, + 463, + 506, + 475 + ], + "score": 1.0, + "content": "For UMAP, the only difference from UM is that, instead of fine-tuning the weights, we generate a", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 474, + 506, + 487 + ], + "spans": [ + { + "bbox": [ + 105, + 474, + 506, + 487 + ], + "score": 1.0, + "content": "popup score for every weight, and force the gradients to pass through the scores. In every iteration,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 485, + 506, + 498 + ], + "spans": [ + { + "bbox": [ + 105, + 485, + 378, + 498 + ], + "score": 1.0, + "content": "we need to formulate a binary mask according to the given prune rate", + "type": "text" + }, + { + "bbox": [ + 379, + 487, + 385, + 497 + ], + "score": 0.76, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 385, + 485, + 506, + 498 + ], + "score": 1.0, + "content": ". This is just what we do when", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 496, + 507, + 508 + ], + "spans": [ + { + "bbox": [ + 106, + 496, + 507, + 508 + ], + "score": 1.0, + "content": "estimating the loss constraint. For more details, it can refer to (Ramanujan et al., 2020). In Table 8,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 506, + 476, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 476, + 518 + ], + "score": 1.0, + "content": "we summarize the overall comparison results of UM and UMAP to show their effectiveness.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 22 + }, + { + "type": "title", + "bbox": [ + 107, + 536, + 349, + 548 + ], + "lines": [ + { + "bbox": [ + 105, + 536, + 350, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 536, + 350, + 550 + ], + "score": 1.0, + "content": "E ADDITIONAL SETUPS OF THE EXPERIMENTS", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 105, + 561, + 468, + 573 + ], + "lines": [ + { + "bbox": [ + 104, + 559, + 470, + 576 + ], + "spans": [ + { + "bbox": [ + 104, + 559, + 470, + 576 + ], + "score": 1.0, + "content": "In this section, we describe more details about the experimental setups for our exploration.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 26 + }, + { + "type": "text", + "bbox": [ + 107, + 586, + 505, + 664 + ], + "lines": [ + { + "bbox": [ + 105, + 586, + 505, + 598 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 505, + 598 + ], + "score": 1.0, + "content": "Model setups. For DenseNet-101, we fix the growth rate and reduce rate to 12 and 0.5 respectively", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 597, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 506, + 609 + ], + "score": 1.0, + "content": "with the bottleneck block included in the backbone. We also explore the proposed UM on WideResNet", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 608, + 506, + 620 + ], + "spans": [ + { + "bbox": [ + 105, + 608, + 506, + 620 + ], + "score": 1.0, + "content": "(Zagoruyko & Komodakis, 2016) with 40 depth and 4 widen factor, which is termed as WRN-40-4.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 618, + 506, + 632 + ], + "spans": [ + { + "bbox": [ + 105, + 618, + 506, + 632 + ], + "score": 1.0, + "content": "The batch size for both ID and OOD testing samples is 64, and the batch size of auxiliary samples is", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 630, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 150, + 642 + ], + "score": 1.0, + "content": "2000. The", + "type": "text" + }, + { + "bbox": [ + 151, + 630, + 158, + 640 + ], + "score": 0.71, + "content": "\\lambda", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 630, + 506, + 642 + ], + "score": 1.0, + "content": "in Eq.(9) is 0.5 to keep the OE loss comparable to the CE loss. As for the strategy of", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 641, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 505, + 653 + ], + "score": 1.0, + "content": "sampling outliers, we randomly retrieve 50000 samples from ImageNet-1k (Deng et al., 2009) for OE", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 651, + 506, + 665 + ], + "spans": [ + { + "bbox": [ + 105, + 651, + 170, + 665 + ], + "score": 1.0, + "content": "and Energy (w.", + "type": "text" + }, + { + "bbox": [ + 170, + 652, + 190, + 663 + ], + "score": 0.86, + "content": "D _ { \\mathrm { a u x . } }", + "type": "inline_equation" + }, + { + "bbox": [ + 191, + 651, + 506, + 665 + ], + "score": 1.0, + "content": ") and 50000 samples using Thompson sampling (Ming et al., 2022) for POEM.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 107, + 677, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "Learning rate schedules. We use 4 different learning rate schedules to demonstrate the existence", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "score": 1.0, + "content": "of the once-covered OOD detection capability. For cosine annealing, we follow the common setups", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "in Loshchilov & Hutter (2017); for linear schedule, the learning rate remains the same in the first", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "one-third epochs, decreases linearly to the tenth of the initial rate in the middle one-third epochs,", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 207, + 734 + ], + "score": 1.0, + "content": "and decrease linearly to", + "type": "text" + }, + { + "bbox": [ + 207, + 720, + 222, + 731 + ], + "score": 0.86, + "content": "1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 222, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "of the initial rate in the last one-third epochs; for the multiple decay", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 36 + } + ], + "page_idx": 16, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 140, + 109, + 468, + 244 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 504, + 102 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 413, + 93 + ], + "score": 1.0, + "content": "Table 7: Comparison among overfitting methods and Energy with WRN-40-4", + "type": "text" + }, + { + "bbox": [ + 413, + 80, + 428, + 92 + ], + "score": 0.68, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 80, + 432, + 93 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 432, + 81, + 440, + 92 + ], + "score": 0.55, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 440, + 80, + 505, + 93 + ], + "score": 1.0, + "content": "indicates higher", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 90, + 333, + 104 + ], + "spans": [ + { + "bbox": [ + 105, + 90, + 192, + 104 + ], + "score": 1.0, + "content": "values are better, and", + "type": "text" + }, + { + "bbox": [ + 192, + 92, + 200, + 102 + ], + "score": 0.47, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 90, + 333, + 104 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 140, + 109, + 468, + 244 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 140, + 109, + 468, + 244 + ], + "spans": [ + { + "bbox": [ + 140, + 109, + 468, + 244 + ], + "score": 0.984, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10Baseline87.6988.1658.4793.86
Early Stopping w. ACC88.0788.6567.6193.99
Early Stopping w. OOD90.9291.9446.6391.47
Weight Decay 0.186.9788.5163.5489.88
Weight Decay 0.0189.7789.8250.2393.16
Weight Decay 0.00189.2589.8450.9593.91
Drop Rate 0.389.7490.0752.1693.22
Drop Rate 0.489.9490.5351.1394.20
Drop Rate 0.590.0991.0452.7693.91
UM (ours)91.7492.6740.4092.68
UMAP (ours)88.8489.3150.2393.86
", + "type": "table", + "image_path": "6ddca4b36c8e630d21ddb3e64e00889c3562f9ae052ec1b20ad4eb06c9b47932.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 140, + 109, + 468, + 154.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 140, + 154.0, + 468, + 199.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 140, + 199.0, + 468, + 244.0 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "title", + "bbox": [ + 106, + 264, + 421, + 276 + ], + "lines": [ + { + "bbox": [ + 105, + 263, + 423, + 280 + ], + "spans": [ + { + "bbox": [ + 105, + 263, + 423, + 280 + ], + "score": 1.0, + "content": "D DETAILED REALIZATION OF THE PROPOSED ALGORITHMS", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 106, + 289, + 504, + 312 + ], + "lines": [ + { + "bbox": [ + 106, + 289, + 505, + 302 + ], + "spans": [ + { + "bbox": [ + 106, + 289, + 505, + 302 + ], + "score": 1.0, + "content": "In this section, we provide the detailed realization of our proposed Unleashing Mask (UM) (i.e., in", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 300, + 437, + 313 + ], + "spans": [ + { + "bbox": [ + 106, + 300, + 437, + 313 + ], + "score": 1.0, + "content": "Algorithm 1) and Unleashing Mask Adopt Pruning (UMAP) (i.e., in Algorithm 2).", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5, + "bbox_fs": [ + 106, + 289, + 505, + 313 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 318, + 505, + 407 + ], + "lines": [ + { + "bbox": [ + 105, + 317, + 506, + 332 + ], + "spans": [ + { + "bbox": [ + 105, + 317, + 233, + 332 + ], + "score": 1.0, + "content": "To estimate the loss constrain", + "type": "text" + }, + { + "bbox": [ + 234, + 319, + 241, + 331 + ], + "score": 0.79, + "content": "\\zeta", + "type": "inline_equation" + }, + { + "bbox": [ + 241, + 317, + 259, + 332 + ], + "score": 1.0, + "content": "(i.e.", + "type": "text" + }, + { + "bbox": [ + 259, + 317, + 320, + 331 + ], + "score": 0.88, + "content": "\\widehat { \\ell } _ { \\mathrm { C E } } ( m _ { \\delta } \\odot f ^ { * } )", + "type": "inline_equation" + }, + { + "bbox": [ + 321, + 317, + 473, + 332 + ], + "score": 1.0, + "content": "in Eq 3 with the fixed given model", + "type": "text" + }, + { + "bbox": [ + 473, + 320, + 485, + 331 + ], + "score": 0.83, + "content": "f ^ { * }", + "type": "inline_equation" + }, + { + "bbox": [ + 486, + 317, + 506, + 332 + ], + "score": 1.0, + "content": ") for", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 330, + 505, + 342 + ], + "spans": [ + { + "bbox": [ + 105, + 330, + 482, + 342 + ], + "score": 1.0, + "content": "forgetting, we need to randomly knock out parts of weights according to the given mask ratio", + "type": "text" + }, + { + "bbox": [ + 482, + 331, + 488, + 340 + ], + "score": 0.68, + "content": "\\delta", + "type": "inline_equation" + }, + { + "bbox": [ + 488, + 330, + 505, + 342 + ], + "score": 1.0, + "content": ". To", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 342, + 506, + 353 + ], + "spans": [ + { + "bbox": [ + 106, + 342, + 506, + 353 + ], + "score": 1.0, + "content": "be specific, we sample a score from a Gaussian distribution for every weight. Then we initialize a", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 351, + 504, + 365 + ], + "spans": [ + { + "bbox": [ + 105, + 351, + 489, + 365 + ], + "score": 1.0, + "content": "unit matrix for every layer of the model concerning the size of the layer. We formulate the mask", + "type": "text" + }, + { + "bbox": [ + 489, + 354, + 504, + 363 + ], + "score": 0.82, + "content": "m _ { \\delta }", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 363, + 505, + 375 + ], + "spans": [ + { + "bbox": [ + 105, + 363, + 505, + 375 + ], + "score": 1.0, + "content": "according to the sampled scores. Find the threshold for each layer that is smaller than the score of the", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 374, + 506, + 386 + ], + "spans": [ + { + "bbox": [ + 105, + 374, + 506, + 386 + ], + "score": 1.0, + "content": "given mask ratio in that layer (termed as quantile). Then set all the ones, whose corresponding scores", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 385, + 504, + 397 + ], + "spans": [ + { + "bbox": [ + 105, + 385, + 497, + 397 + ], + "score": 1.0, + "content": "are more significant than the layers’ thresholds, to zeros. In our algorithms, the fine-tuning epochs", + "type": "text" + }, + { + "bbox": [ + 497, + 386, + 504, + 395 + ], + "score": 0.75, + "content": "k", + "type": "inline_equation" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 396, + 357, + 408 + ], + "spans": [ + { + "bbox": [ + 105, + 396, + 357, + 408 + ], + "score": 1.0, + "content": "is the epochs we fine-tune after we get the well-trained model.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 11.5, + "bbox_fs": [ + 105, + 317, + 506, + 408 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 413, + 505, + 457 + ], + "lines": [ + { + "bbox": [ + 105, + 412, + 505, + 425 + ], + "spans": [ + { + "bbox": [ + 105, + 412, + 505, + 425 + ], + "score": 1.0, + "content": "We dot-multiply every layer’s weights with the formulated binary matrix as if we delete some parts", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 424, + 506, + 437 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 506, + 437 + ], + "score": 1.0, + "content": "of the weights. We input a batch of training samples to the masked model and treat the mean value of", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 435, + 505, + 447 + ], + "spans": [ + { + "bbox": [ + 106, + 435, + 505, + 447 + ], + "score": 1.0, + "content": "the outputs’ cross-entropy loss as the loss constraint. After all of these have been done, we begin to", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 446, + 455, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 455, + 459 + ], + "score": 1.0, + "content": "fine-tune the model’s weights with the loss constraint applied to the cross-entropy loss.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 17.5, + "bbox_fs": [ + 105, + 412, + 506, + 459 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 462, + 505, + 518 + ], + "lines": [ + { + "bbox": [ + 105, + 463, + 506, + 475 + ], + "spans": [ + { + "bbox": [ + 105, + 463, + 506, + 475 + ], + "score": 1.0, + "content": "For UMAP, the only difference from UM is that, instead of fine-tuning the weights, we generate a", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 474, + 506, + 487 + ], + "spans": [ + { + "bbox": [ + 105, + 474, + 506, + 487 + ], + "score": 1.0, + "content": "popup score for every weight, and force the gradients to pass through the scores. In every iteration,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 485, + 506, + 498 + ], + "spans": [ + { + "bbox": [ + 105, + 485, + 378, + 498 + ], + "score": 1.0, + "content": "we need to formulate a binary mask according to the given prune rate", + "type": "text" + }, + { + "bbox": [ + 379, + 487, + 385, + 497 + ], + "score": 0.76, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 385, + 485, + 506, + 498 + ], + "score": 1.0, + "content": ". This is just what we do when", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 496, + 507, + 508 + ], + "spans": [ + { + "bbox": [ + 106, + 496, + 507, + 508 + ], + "score": 1.0, + "content": "estimating the loss constraint. For more details, it can refer to (Ramanujan et al., 2020). In Table 8,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 506, + 476, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 476, + 518 + ], + "score": 1.0, + "content": "we summarize the overall comparison results of UM and UMAP to show their effectiveness.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 22, + "bbox_fs": [ + 105, + 463, + 507, + 518 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 536, + 349, + 548 + ], + "lines": [ + { + "bbox": [ + 105, + 536, + 350, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 536, + 350, + 550 + ], + "score": 1.0, + "content": "E ADDITIONAL SETUPS OF THE EXPERIMENTS", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 105, + 561, + 468, + 573 + ], + "lines": [ + { + "bbox": [ + 104, + 559, + 470, + 576 + ], + "spans": [ + { + "bbox": [ + 104, + 559, + 470, + 576 + ], + "score": 1.0, + "content": "In this section, we describe more details about the experimental setups for our exploration.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 26, + "bbox_fs": [ + 104, + 559, + 470, + 576 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 586, + 505, + 664 + ], + "lines": [ + { + "bbox": [ + 105, + 586, + 505, + 598 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 505, + 598 + ], + "score": 1.0, + "content": "Model setups. For DenseNet-101, we fix the growth rate and reduce rate to 12 and 0.5 respectively", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 597, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 506, + 609 + ], + "score": 1.0, + "content": "with the bottleneck block included in the backbone. We also explore the proposed UM on WideResNet", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 608, + 506, + 620 + ], + "spans": [ + { + "bbox": [ + 105, + 608, + 506, + 620 + ], + "score": 1.0, + "content": "(Zagoruyko & Komodakis, 2016) with 40 depth and 4 widen factor, which is termed as WRN-40-4.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 618, + 506, + 632 + ], + "spans": [ + { + "bbox": [ + 105, + 618, + 506, + 632 + ], + "score": 1.0, + "content": "The batch size for both ID and OOD testing samples is 64, and the batch size of auxiliary samples is", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 630, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 150, + 642 + ], + "score": 1.0, + "content": "2000. The", + "type": "text" + }, + { + "bbox": [ + 151, + 630, + 158, + 640 + ], + "score": 0.71, + "content": "\\lambda", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 630, + 506, + 642 + ], + "score": 1.0, + "content": "in Eq.(9) is 0.5 to keep the OE loss comparable to the CE loss. As for the strategy of", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 641, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 505, + 653 + ], + "score": 1.0, + "content": "sampling outliers, we randomly retrieve 50000 samples from ImageNet-1k (Deng et al., 2009) for OE", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 651, + 506, + 665 + ], + "spans": [ + { + "bbox": [ + 105, + 651, + 170, + 665 + ], + "score": 1.0, + "content": "and Energy (w.", + "type": "text" + }, + { + "bbox": [ + 170, + 652, + 190, + 663 + ], + "score": 0.86, + "content": "D _ { \\mathrm { a u x . } }", + "type": "inline_equation" + }, + { + "bbox": [ + 191, + 651, + 506, + 665 + ], + "score": 1.0, + "content": ") and 50000 samples using Thompson sampling (Ming et al., 2022) for POEM.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 30, + "bbox_fs": [ + 105, + 586, + 506, + 665 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 677, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "Learning rate schedules. We use 4 different learning rate schedules to demonstrate the existence", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "score": 1.0, + "content": "of the once-covered OOD detection capability. For cosine annealing, we follow the common setups", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "in Loshchilov & Hutter (2017); for linear schedule, the learning rate remains the same in the first", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "one-third epochs, decreases linearly to the tenth of the initial rate in the middle one-third epochs,", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 207, + 734 + ], + "score": 1.0, + "content": "and decrease linearly to", + "type": "text" + }, + { + "bbox": [ + 207, + 720, + 222, + 731 + ], + "score": 0.86, + "content": "1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 222, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "of the initial rate in the last one-third epochs; for the multiple decay", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 36, + "bbox_fs": [ + 105, + 676, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 82, + 257, + 94 + ], + "lines": [ + { + "bbox": [ + 105, + 80, + 259, + 97 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 259, + 97 + ], + "score": 1.0, + "content": "Algorithm 1 Unleashing Mask (UM)", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 105, + 96, + 504, + 119 + ], + "lines": [ + { + "bbox": [ + 105, + 94, + 505, + 109 + ], + "spans": [ + { + "bbox": [ + 105, + 94, + 219, + 109 + ], + "score": 1.0, + "content": "Input: well-trained model :", + "type": "text" + }, + { + "bbox": [ + 219, + 97, + 226, + 106 + ], + "score": 0.35, + "content": "\\theta", + "type": "inline_equation" + }, + { + "bbox": [ + 226, + 94, + 276, + 109 + ], + "score": 1.0, + "content": ", mask ratio:", + "type": "text" + }, + { + "bbox": [ + 277, + 96, + 315, + 108 + ], + "score": 0.92, + "content": "\\delta \\in [ 0 , 1 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 94, + 425, + 109 + ], + "score": 1.0, + "content": ", fine-tuning epochs of UM:", + "type": "text" + }, + { + "bbox": [ + 426, + 97, + 433, + 107 + ], + "score": 0.37, + "content": "k", + "type": "inline_equation" + }, + { + "bbox": [ + 433, + 94, + 505, + 109 + ], + "score": 1.0, + "content": ", training samples:", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 105, + 149, + 122 + ], + "spans": [ + { + "bbox": [ + 106, + 108, + 141, + 120 + ], + "score": 0.88, + "content": "x \\sim D _ { \\mathrm { i n } } ^ { \\mathrm { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 105, + 149, + 122 + ], + "score": 1.0, + "content": ";", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1.5 + }, + { + "type": "text", + "bbox": [ + 107, + 120, + 223, + 132 + ], + "lines": [ + { + "bbox": [ + 105, + 118, + 221, + 133 + ], + "spans": [ + { + "bbox": [ + 105, + 119, + 214, + 133 + ], + "score": 1.0, + "content": "Output: fine-tuned model", + "type": "text" + }, + { + "bbox": [ + 214, + 118, + 221, + 131 + ], + "score": 0.77, + "content": "\\hat { \\theta }", + "type": "inline_equation" + } + ], + "index": 3 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 110, + 134, + 292, + 307 + ], + "lines": [ + { + "bbox": [ + 111, + 133, + 292, + 150 + ], + "spans": [ + { + "bbox": [ + 111, + 136, + 123, + 146 + ], + "score": 1.0, + "content": "1:", + "type": "text" + }, + { + "bbox": [ + 123, + 133, + 292, + 150 + ], + "score": 1.0, + "content": "Initialize a popup score for every weight", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 111, + 148, + 179, + 160 + ], + "spans": [ + { + "bbox": [ + 111, + 148, + 138, + 160 + ], + "score": 1.0, + "content": "2: for", + "type": "text" + }, + { + "bbox": [ + 138, + 148, + 164, + 159 + ], + "score": 0.84, + "content": "w \\in { \\boldsymbol { \\theta } }", + "type": "inline_equation" + }, + { + "bbox": [ + 165, + 148, + 179, + 160 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 110, + 159, + 197, + 171 + ], + "spans": [ + { + "bbox": [ + 110, + 159, + 122, + 171 + ], + "score": 1.0, + "content": "3:", + "type": "text" + }, + { + "bbox": [ + 133, + 159, + 197, + 171 + ], + "score": 0.76, + "content": "s _ { w } \\sim N ( \\mu , \\sigma ^ { 2 } )", + "type": "inline_equation" + } + ], + "index": 6 + }, + { + "bbox": [ + 111, + 170, + 158, + 182 + ], + "spans": [ + { + "bbox": [ + 111, + 171, + 122, + 181 + ], + "score": 1.0, + "content": "4:", + "type": "text" + }, + { + "bbox": [ + 121, + 170, + 158, + 182 + ], + "score": 1.0, + "content": "end for", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 110, + 179, + 272, + 195 + ], + "spans": [ + { + "bbox": [ + 110, + 181, + 123, + 192 + ], + "score": 1.0, + "content": "5:", + "type": "text" + }, + { + "bbox": [ + 123, + 179, + 272, + 195 + ], + "score": 1.0, + "content": "Generate mask by the popup scores", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 111, + 192, + 192, + 208 + ], + "spans": [ + { + "bbox": [ + 111, + 195, + 122, + 205 + ], + "score": 1.0, + "content": "6:", + "type": "text" + }, + { + "bbox": [ + 121, + 192, + 179, + 208 + ], + "score": 1.0, + "content": "for l ∈ θlayers", + "type": "text" + }, + { + "bbox": [ + 177, + 195, + 192, + 205 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 111, + 205, + 239, + 221 + ], + "spans": [ + { + "bbox": [ + 111, + 207, + 122, + 218 + ], + "score": 1.0, + "content": "7:", + "type": "text" + }, + { + "bbox": [ + 133, + 206, + 180, + 219 + ], + "score": 0.44, + "content": "m _ { \\delta } ^ { l } = \\dot { s } ^ { l } >", + "type": "inline_equation" + }, + { + "bbox": [ + 180, + 205, + 213, + 221 + ], + "score": 1.0, + "content": "quantile", + "type": "text" + }, + { + "bbox": [ + 214, + 206, + 239, + 219 + ], + "score": 0.85, + "content": "( s ^ { l } , \\delta )", + "type": "inline_equation" + } + ], + "index": 10 + }, + { + "bbox": [ + 111, + 217, + 157, + 228 + ], + "spans": [ + { + "bbox": [ + 111, + 217, + 157, + 228 + ], + "score": 1.0, + "content": "8: end for", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 110, + 227, + 223, + 241 + ], + "spans": [ + { + "bbox": [ + 110, + 227, + 223, + 241 + ], + "score": 1.0, + "content": "9: Estimate loss constrain", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 108, + 240, + 243, + 257 + ], + "spans": [ + { + "bbox": [ + 108, + 240, + 123, + 257 + ], + "score": 1.0, + "content": "10:", + "type": "text" + }, + { + "bbox": [ + 124, + 240, + 243, + 255 + ], + "score": 0.89, + "content": "\\zeta = \\mathbb { E } _ { x \\sim D _ { \\mathrm { i n } } ^ { s } } ( \\hat { \\mathcal { L } } _ { \\mathrm { C E } } ( x , m _ { \\delta } \\odot \\theta ) )", + "type": "inline_equation" + } + ], + "index": 13 + }, + { + "bbox": [ + 107, + 253, + 252, + 271 + ], + "spans": [ + { + "bbox": [ + 107, + 253, + 252, + 271 + ], + "score": 1.0, + "content": "11: Unleashing Mask: fine-tuning", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 270, + 211, + 283 + ], + "spans": [ + { + "bbox": [ + 106, + 270, + 138, + 283 + ], + "score": 1.0, + "content": "12: for", + "type": "text" + }, + { + "bbox": [ + 138, + 270, + 196, + 282 + ], + "score": 0.85, + "content": "t \\in ( 1 , \\ldots , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 270, + 211, + 283 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 104, + 278, + 276, + 297 + ], + "spans": [ + { + "bbox": [ + 104, + 278, + 276, + 297 + ], + "score": 1.0, + "content": "13: θ(t+1) = θ(t) − η ∂(|LCE(x,θ)−ζ|+ζ)", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 294, + 158, + 307 + ], + "spans": [ + { + "bbox": [ + 106, + 294, + 158, + 307 + ], + "score": 1.0, + "content": "14: end for", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 10.5 + }, + { + "type": "title", + "bbox": [ + 107, + 321, + 334, + 333 + ], + "lines": [ + { + "bbox": [ + 105, + 318, + 333, + 337 + ], + "spans": [ + { + "bbox": [ + 105, + 318, + 333, + 337 + ], + "score": 1.0, + "content": "Algorithm 2 Unleashing Mask Adopt Pruning (UMAP)", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 106, + 335, + 504, + 358 + ], + "lines": [ + { + "bbox": [ + 105, + 334, + 506, + 349 + ], + "spans": [ + { + "bbox": [ + 105, + 334, + 219, + 349 + ], + "score": 1.0, + "content": "Input: well-trained model :", + "type": "text" + }, + { + "bbox": [ + 219, + 336, + 226, + 345 + ], + "score": 0.51, + "content": "\\theta", + "type": "inline_equation" + }, + { + "bbox": [ + 226, + 334, + 276, + 349 + ], + "score": 1.0, + "content": ", mask ratio:", + "type": "text" + }, + { + "bbox": [ + 277, + 335, + 315, + 348 + ], + "score": 0.92, + "content": "\\delta \\in [ 0 , 1 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 334, + 425, + 349 + ], + "score": 1.0, + "content": ", fine-tuning epochs of UM:", + "type": "text" + }, + { + "bbox": [ + 426, + 336, + 433, + 346 + ], + "score": 0.59, + "content": "k", + "type": "inline_equation" + }, + { + "bbox": [ + 433, + 334, + 506, + 349 + ], + "score": 1.0, + "content": ", training samples:", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 345, + 205, + 361 + ], + "spans": [ + { + "bbox": [ + 106, + 347, + 141, + 359 + ], + "score": 0.9, + "content": "x \\sim D _ { \\mathrm { i n } } ^ { \\mathrm { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 345, + 190, + 361 + ], + "score": 1.0, + "content": ", prune rate:", + "type": "text" + }, + { + "bbox": [ + 190, + 348, + 198, + 358 + ], + "score": 0.7, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 198, + 345, + 205, + 361 + ], + "score": 1.0, + "content": ";", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5 + }, + { + "type": "text", + "bbox": [ + 107, + 358, + 239, + 369 + ], + "lines": [ + { + "bbox": [ + 106, + 356, + 240, + 371 + ], + "spans": [ + { + "bbox": [ + 106, + 356, + 220, + 371 + ], + "score": 1.0, + "content": "Output: learnt binary mask", + "type": "text" + }, + { + "bbox": [ + 221, + 358, + 235, + 370 + ], + "score": 0.89, + "content": "\\hat { m } _ { p }", + "type": "inline_equation" + }, + { + "bbox": [ + 235, + 356, + 240, + 371 + ], + "score": 1.0, + "content": ";", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 21 + }, + { + "type": "text", + "bbox": [ + 109, + 371, + 313, + 589 + ], + "lines": [ + { + "bbox": [ + 111, + 371, + 292, + 387 + ], + "spans": [ + { + "bbox": [ + 111, + 371, + 292, + 387 + ], + "score": 1.0, + "content": "1: Initialize a popup score for every weight", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 110, + 385, + 179, + 398 + ], + "spans": [ + { + "bbox": [ + 110, + 385, + 138, + 398 + ], + "score": 1.0, + "content": "2: for", + "type": "text" + }, + { + "bbox": [ + 138, + 386, + 164, + 396 + ], + "score": 0.84, + "content": "w \\in { \\boldsymbol { \\theta } }", + "type": "inline_equation" + }, + { + "bbox": [ + 165, + 385, + 179, + 398 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 111, + 397, + 197, + 409 + ], + "spans": [ + { + "bbox": [ + 111, + 398, + 121, + 408 + ], + "score": 1.0, + "content": "3:", + "type": "text" + }, + { + "bbox": [ + 133, + 397, + 197, + 409 + ], + "score": 0.83, + "content": "s _ { w } \\sim N ( \\mu , \\sigma ^ { 2 } )", + "type": "inline_equation" + } + ], + "index": 24 + }, + { + "bbox": [ + 111, + 407, + 158, + 420 + ], + "spans": [ + { + "bbox": [ + 111, + 409, + 121, + 418 + ], + "score": 1.0, + "content": "4:", + "type": "text" + }, + { + "bbox": [ + 121, + 407, + 158, + 420 + ], + "score": 1.0, + "content": "end for", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 111, + 417, + 273, + 433 + ], + "spans": [ + { + "bbox": [ + 111, + 420, + 120, + 429 + ], + "score": 1.0, + "content": "5:", + "type": "text" + }, + { + "bbox": [ + 123, + 417, + 273, + 433 + ], + "score": 1.0, + "content": "Generate mask by the popup scores", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 111, + 431, + 192, + 445 + ], + "spans": [ + { + "bbox": [ + 111, + 431, + 138, + 445 + ], + "score": 1.0, + "content": "6: for", + "type": "text" + }, + { + "bbox": [ + 138, + 432, + 177, + 443 + ], + "score": 0.68, + "content": "l \\in \\theta _ { \\mathrm { l a y e r s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 177, + 431, + 192, + 445 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 111, + 444, + 239, + 457 + ], + "spans": [ + { + "bbox": [ + 111, + 445, + 122, + 455 + ], + "score": 1.0, + "content": "7:", + "type": "text" + }, + { + "bbox": [ + 133, + 444, + 239, + 457 + ], + "score": 0.4, + "content": "m _ { \\delta } ^ { l } = s ^ { l } > \\mathrm { q u a n t i l e } ( s ^ { l } , \\delta )", + "type": "inline_equation" + } + ], + "index": 28 + }, + { + "bbox": [ + 111, + 455, + 157, + 466 + ], + "spans": [ + { + "bbox": [ + 111, + 456, + 121, + 465 + ], + "score": 1.0, + "content": "8:", + "type": "text" + }, + { + "bbox": [ + 122, + 455, + 157, + 466 + ], + "score": 1.0, + "content": "end for", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 111, + 465, + 223, + 478 + ], + "spans": [ + { + "bbox": [ + 111, + 467, + 122, + 477 + ], + "score": 1.0, + "content": "9:", + "type": "text" + }, + { + "bbox": [ + 124, + 465, + 223, + 478 + ], + "score": 1.0, + "content": "Estimate loss constrain", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 478, + 244, + 494 + ], + "spans": [ + { + "bbox": [ + 106, + 478, + 123, + 494 + ], + "score": 1.0, + "content": "10:", + "type": "text" + }, + { + "bbox": [ + 124, + 478, + 244, + 493 + ], + "score": 0.88, + "content": "\\zeta = \\mathbb { E } _ { x \\sim D _ { \\mathrm { i n } } ^ { s } } ( \\hat { \\mathcal { L } } _ { \\mathrm { C E } } ( x , m _ { \\delta } \\odot \\theta ) )", + "type": "inline_equation" + } + ], + "index": 31 + }, + { + "bbox": [ + 107, + 491, + 313, + 509 + ], + "spans": [ + { + "bbox": [ + 107, + 491, + 313, + 509 + ], + "score": 1.0, + "content": "11: Unleashing Mask Adopt Pruning: fine-tuning", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 108, + 507, + 191, + 522 + ], + "spans": [ + { + "bbox": [ + 108, + 507, + 123, + 522 + ], + "score": 1.0, + "content": "12:", + "type": "text" + }, + { + "bbox": [ + 124, + 508, + 191, + 521 + ], + "score": 0.84, + "content": "s ^ { ( 1 ) } \\sim N ( \\mu , \\sigma ^ { 2 } )", + "type": "inline_equation" + } + ], + "index": 33 + }, + { + "bbox": [ + 107, + 519, + 212, + 533 + ], + "spans": [ + { + "bbox": [ + 107, + 519, + 138, + 533 + ], + "score": 1.0, + "content": "13: for", + "type": "text" + }, + { + "bbox": [ + 138, + 521, + 195, + 532 + ], + "score": 0.81, + "content": "t \\in ( 1 , \\ldots , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 519, + 212, + 533 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 530, + 203, + 545 + ], + "spans": [ + { + "bbox": [ + 106, + 531, + 122, + 543 + ], + "score": 1.0, + "content": "14:", + "type": "text" + }, + { + "bbox": [ + 131, + 530, + 148, + 545 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 148, + 532, + 187, + 543 + ], + "score": 0.55, + "content": "l \\in \\theta _ { \\mathrm { l a y e r s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 187, + 530, + 203, + 545 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 539, + 271, + 559 + ], + "spans": [ + { + "bbox": [ + 106, + 544, + 122, + 556 + ], + "score": 1.0, + "content": "15:", + "type": "text" + }, + { + "bbox": [ + 140, + 539, + 142, + 559 + ], + "score": 0.0, + "content": "", + "type": "text" + }, + { + "bbox": [ + 142, + 543, + 271, + 558 + ], + "score": 0.33, + "content": "\\hat { m } _ { p } ^ { l } = \\dot { s ^ { l ( t ) } } > \\mathrm { q u a n t i l e } ( s ^ { l ( t ) } , p )", + "type": "inline_equation", + "image_path": "654b43dbe7313e661d44c6af014d15ad8a172fd16e9aa8358e75b3dbaf056d34.jpg" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 554, + 168, + 567 + ], + "spans": [ + { + "bbox": [ + 106, + 555, + 122, + 567 + ], + "score": 1.0, + "content": "16:", + "type": "text" + }, + { + "bbox": [ + 132, + 554, + 168, + 567 + ], + "score": 1.0, + "content": "end for", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 563, + 290, + 579 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 123, + 579 + ], + "score": 1.0, + "content": "17:", + "type": "text" + }, + { + "bbox": [ + 133, + 563, + 290, + 579 + ], + "score": 0.56, + "content": "\\begin{array} { r } { s ^ { ( t + 1 ) } = s ^ { ( t ) } - \\eta \\frac { \\partial ( | \\mathcal { L } _ { \\mathrm { C E } } ( x , \\hat { m } _ { p } \\odot \\theta ) - \\zeta | + \\zeta ) } { \\partial \\theta } } \\end{array}", + "type": "inline_equation" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 576, + 157, + 589 + ], + "spans": [ + { + "bbox": [ + 105, + 576, + 157, + 589 + ], + "score": 1.0, + "content": "18: end for", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 30.5 + }, + { + "type": "text", + "bbox": [ + 107, + 612, + 505, + 645 + ], + "lines": [ + { + "bbox": [ + 106, + 611, + 506, + 625 + ], + "spans": [ + { + "bbox": [ + 106, + 611, + 251, + 625 + ], + "score": 1.0, + "content": "schedule, the learning rate decreases", + "type": "text" + }, + { + "bbox": [ + 252, + 612, + 271, + 623 + ], + "score": 0.89, + "content": "1 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 611, + 389, + 625 + ], + "score": 1.0, + "content": "of the initial rate (0.01) every", + "type": "text" + }, + { + "bbox": [ + 390, + 612, + 410, + 623 + ], + "score": 0.88, + "content": "1 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 410, + 611, + 506, + 625 + ], + "score": 1.0, + "content": "epochs (10 epochs); for", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 623, + 506, + 636 + ], + "spans": [ + { + "bbox": [ + 106, + 623, + 328, + 636 + ], + "score": 1.0, + "content": "the multiple step schedule, the learning rate decreases to", + "type": "text" + }, + { + "bbox": [ + 329, + 623, + 348, + 634 + ], + "score": 0.89, + "content": "1 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 623, + 506, + 636 + ], + "score": 1.0, + "content": "of the current rate every 30 epochs. All", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 634, + 465, + 646 + ], + "spans": [ + { + "bbox": [ + 106, + 634, + 465, + 646 + ], + "score": 1.0, + "content": "those learning rate schedules for our experiments are intuitively illustrated in Figure 4(a).", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41 + }, + { + "type": "title", + "bbox": [ + 107, + 663, + 427, + 675 + ], + "lines": [ + { + "bbox": [ + 104, + 662, + 429, + 678 + ], + "spans": [ + { + "bbox": [ + 104, + 662, + 429, + 678 + ], + "score": 1.0, + "content": "F ADDITIONAL EXPERIMENT RESULTS AND ABLATION STUDY", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 43 + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 506, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 688, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 506, + 700 + ], + "score": 1.0, + "content": "In this section, we provide more experiment results. We first show the fine-grained results on CIFAR-", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 698, + 507, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 507, + 711 + ], + "score": 1.0, + "content": "10 and CIFAR-100, then conduct the experiments under a different model structure (i.e., WRN-40-4),", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 710, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 722 + ], + "score": 1.0, + "content": "and finally apply an additional ablation study on the proposed UM and UMAP. The mean and variance", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 720, + 506, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 732 + ], + "score": 1.0, + "content": "of all metrics (ID-ACC, AUROC, AUPR, FPR95) are reported based on multiple independent trials.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 45.5 + } + ], + "page_idx": 17, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 763 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 763 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 82, + 257, + 94 + ], + "lines": [ + { + "bbox": [ + 105, + 80, + 259, + 97 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 259, + 97 + ], + "score": 1.0, + "content": "Algorithm 1 Unleashing Mask (UM)", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 105, + 96, + 504, + 119 + ], + "lines": [ + { + "bbox": [ + 105, + 94, + 505, + 109 + ], + "spans": [ + { + "bbox": [ + 105, + 94, + 219, + 109 + ], + "score": 1.0, + "content": "Input: well-trained model :", + "type": "text" + }, + { + "bbox": [ + 219, + 97, + 226, + 106 + ], + "score": 0.35, + "content": "\\theta", + "type": "inline_equation" + }, + { + "bbox": [ + 226, + 94, + 276, + 109 + ], + "score": 1.0, + "content": ", mask ratio:", + "type": "text" + }, + { + "bbox": [ + 277, + 96, + 315, + 108 + ], + "score": 0.92, + "content": "\\delta \\in [ 0 , 1 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 94, + 425, + 109 + ], + "score": 1.0, + "content": ", fine-tuning epochs of UM:", + "type": "text" + }, + { + "bbox": [ + 426, + 97, + 433, + 107 + ], + "score": 0.37, + "content": "k", + "type": "inline_equation" + }, + { + "bbox": [ + 433, + 94, + 505, + 109 + ], + "score": 1.0, + "content": ", training samples:", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 105, + 149, + 122 + ], + "spans": [ + { + "bbox": [ + 106, + 108, + 141, + 120 + ], + "score": 0.88, + "content": "x \\sim D _ { \\mathrm { i n } } ^ { \\mathrm { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 105, + 149, + 122 + ], + "score": 1.0, + "content": ";", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1.5, + "bbox_fs": [ + 105, + 94, + 505, + 122 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 120, + 223, + 132 + ], + "lines": [ + { + "bbox": [ + 105, + 118, + 221, + 133 + ], + "spans": [ + { + "bbox": [ + 105, + 119, + 214, + 133 + ], + "score": 1.0, + "content": "Output: fine-tuned model", + "type": "text" + }, + { + "bbox": [ + 214, + 118, + 221, + 131 + ], + "score": 0.77, + "content": "\\hat { \\theta }", + "type": "inline_equation" + } + ], + "index": 3 + } + ], + "index": 3, + "bbox_fs": [ + 105, + 118, + 221, + 133 + ] + }, + { + "type": "index", + "bbox": [ + 110, + 134, + 292, + 307 + ], + "lines": [ + { + "bbox": [ + 111, + 133, + 292, + 150 + ], + "spans": [ + { + "bbox": [ + 111, + 136, + 123, + 146 + ], + "score": 1.0, + "content": "1:", + "type": "text" + }, + { + "bbox": [ + 123, + 133, + 292, + 150 + ], + "score": 1.0, + "content": "Initialize a popup score for every weight", + "type": "text" + } + ], + "index": 4, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 148, + 179, + 160 + ], + "spans": [ + { + "bbox": [ + 111, + 148, + 138, + 160 + ], + "score": 1.0, + "content": "2: for", + "type": "text" + }, + { + "bbox": [ + 138, + 148, + 164, + 159 + ], + "score": 0.84, + "content": "w \\in { \\boldsymbol { \\theta } }", + "type": "inline_equation" + }, + { + "bbox": [ + 165, + 148, + 179, + 160 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 5, + "is_list_start_line": true + }, + { + "bbox": [ + 110, + 159, + 197, + 171 + ], + "spans": [ + { + "bbox": [ + 110, + 159, + 122, + 171 + ], + "score": 1.0, + "content": "3:", + "type": "text" + }, + { + "bbox": [ + 133, + 159, + 197, + 171 + ], + "score": 0.76, + "content": "s _ { w } \\sim N ( \\mu , \\sigma ^ { 2 } )", + "type": "inline_equation" + } + ], + "index": 6, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 170, + 158, + 182 + ], + "spans": [ + { + "bbox": [ + 111, + 171, + 122, + 181 + ], + "score": 1.0, + "content": "4:", + "type": "text" + }, + { + "bbox": [ + 121, + 170, + 158, + 182 + ], + "score": 1.0, + "content": "end for", + "type": "text" + } + ], + "index": 7, + "is_list_start_line": true + }, + { + "bbox": [ + 110, + 179, + 272, + 195 + ], + "spans": [ + { + "bbox": [ + 110, + 181, + 123, + 192 + ], + "score": 1.0, + "content": "5:", + "type": "text" + }, + { + "bbox": [ + 123, + 179, + 272, + 195 + ], + "score": 1.0, + "content": "Generate mask by the popup scores", + "type": "text" + } + ], + "index": 8, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 192, + 192, + 208 + ], + "spans": [ + { + "bbox": [ + 111, + 195, + 122, + 205 + ], + "score": 1.0, + "content": "6:", + "type": "text" + }, + { + "bbox": [ + 121, + 192, + 179, + 208 + ], + "score": 1.0, + "content": "for l ∈ θlayers", + "type": "text" + }, + { + "bbox": [ + 177, + 195, + 192, + 205 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 9, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 205, + 239, + 221 + ], + "spans": [ + { + "bbox": [ + 111, + 207, + 122, + 218 + ], + "score": 1.0, + "content": "7:", + "type": "text" + }, + { + "bbox": [ + 133, + 206, + 180, + 219 + ], + "score": 0.44, + "content": "m _ { \\delta } ^ { l } = \\dot { s } ^ { l } >", + "type": "inline_equation" + }, + { + "bbox": [ + 180, + 205, + 213, + 221 + ], + "score": 1.0, + "content": "quantile", + "type": "text" + }, + { + "bbox": [ + 214, + 206, + 239, + 219 + ], + "score": 0.85, + "content": "( s ^ { l } , \\delta )", + "type": "inline_equation" + } + ], + "index": 10, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 217, + 157, + 228 + ], + "spans": [ + { + "bbox": [ + 111, + 217, + 157, + 228 + ], + "score": 1.0, + "content": "8: end for", + "type": "text" + } + ], + "index": 11, + "is_list_start_line": true + }, + { + "bbox": [ + 110, + 227, + 223, + 241 + ], + "spans": [ + { + "bbox": [ + 110, + 227, + 223, + 241 + ], + "score": 1.0, + "content": "9: Estimate loss constrain", + "type": "text" + } + ], + "index": 12, + "is_list_start_line": true + }, + { + "bbox": [ + 108, + 240, + 243, + 257 + ], + "spans": [ + { + "bbox": [ + 108, + 240, + 123, + 257 + ], + "score": 1.0, + "content": "10:", + "type": "text" + }, + { + "bbox": [ + 124, + 240, + 243, + 255 + ], + "score": 0.89, + "content": "\\zeta = \\mathbb { E } _ { x \\sim D _ { \\mathrm { i n } } ^ { s } } ( \\hat { \\mathcal { L } } _ { \\mathrm { C E } } ( x , m _ { \\delta } \\odot \\theta ) )", + "type": "inline_equation" + } + ], + "index": 13, + "is_list_start_line": true + }, + { + "bbox": [ + 107, + 253, + 252, + 271 + ], + "spans": [ + { + "bbox": [ + 107, + 253, + 252, + 271 + ], + "score": 1.0, + "content": "11: Unleashing Mask: fine-tuning", + "type": "text" + } + ], + "index": 14, + "is_list_start_line": true + }, + { + "bbox": [ + 106, + 270, + 211, + 283 + ], + "spans": [ + { + "bbox": [ + 106, + 270, + 138, + 283 + ], + "score": 1.0, + "content": "12: for", + "type": "text" + }, + { + "bbox": [ + 138, + 270, + 196, + 282 + ], + "score": 0.85, + "content": "t \\in ( 1 , \\ldots , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 270, + 211, + 283 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 15, + "is_list_start_line": true + }, + { + "bbox": [ + 104, + 278, + 276, + 297 + ], + "spans": [ + { + "bbox": [ + 104, + 278, + 276, + 297 + ], + "score": 1.0, + "content": "13: θ(t+1) = θ(t) − η ∂(|LCE(x,θ)−ζ|+ζ)", + "type": "text" + } + ], + "index": 16, + "is_list_start_line": true + }, + { + "bbox": [ + 106, + 294, + 158, + 307 + ], + "spans": [ + { + "bbox": [ + 106, + 294, + 158, + 307 + ], + "score": 1.0, + "content": "14: end for", + "type": "text" + } + ], + "index": 17, + "is_list_start_line": true + } + ], + "index": 10.5, + "bbox_fs": [ + 104, + 133, + 292, + 307 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 321, + 334, + 333 + ], + "lines": [ + { + "bbox": [ + 105, + 318, + 333, + 337 + ], + "spans": [ + { + "bbox": [ + 105, + 318, + 333, + 337 + ], + "score": 1.0, + "content": "Algorithm 2 Unleashing Mask Adopt Pruning (UMAP)", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 106, + 335, + 504, + 358 + ], + "lines": [ + { + "bbox": [ + 105, + 334, + 506, + 349 + ], + "spans": [ + { + "bbox": [ + 105, + 334, + 219, + 349 + ], + "score": 1.0, + "content": "Input: well-trained model :", + "type": "text" + }, + { + "bbox": [ + 219, + 336, + 226, + 345 + ], + "score": 0.51, + "content": "\\theta", + "type": "inline_equation" + }, + { + "bbox": [ + 226, + 334, + 276, + 349 + ], + "score": 1.0, + "content": ", mask ratio:", + "type": "text" + }, + { + "bbox": [ + 277, + 335, + 315, + 348 + ], + "score": 0.92, + "content": "\\delta \\in [ 0 , 1 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 334, + 425, + 349 + ], + "score": 1.0, + "content": ", fine-tuning epochs of UM:", + "type": "text" + }, + { + "bbox": [ + 426, + 336, + 433, + 346 + ], + "score": 0.59, + "content": "k", + "type": "inline_equation" + }, + { + "bbox": [ + 433, + 334, + 506, + 349 + ], + "score": 1.0, + "content": ", training samples:", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 345, + 205, + 361 + ], + "spans": [ + { + "bbox": [ + 106, + 347, + 141, + 359 + ], + "score": 0.9, + "content": "x \\sim D _ { \\mathrm { i n } } ^ { \\mathrm { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 345, + 190, + 361 + ], + "score": 1.0, + "content": ", prune rate:", + "type": "text" + }, + { + "bbox": [ + 190, + 348, + 198, + 358 + ], + "score": 0.7, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 198, + 345, + 205, + 361 + ], + "score": 1.0, + "content": ";", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5, + "bbox_fs": [ + 105, + 334, + 506, + 361 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 358, + 239, + 369 + ], + "lines": [ + { + "bbox": [ + 106, + 356, + 240, + 371 + ], + "spans": [ + { + "bbox": [ + 106, + 356, + 220, + 371 + ], + "score": 1.0, + "content": "Output: learnt binary mask", + "type": "text" + }, + { + "bbox": [ + 221, + 358, + 235, + 370 + ], + "score": 0.89, + "content": "\\hat { m } _ { p }", + "type": "inline_equation" + }, + { + "bbox": [ + 235, + 356, + 240, + 371 + ], + "score": 1.0, + "content": ";", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 21, + "bbox_fs": [ + 106, + 356, + 240, + 371 + ] + }, + { + "type": "index", + "bbox": [ + 109, + 371, + 313, + 589 + ], + "lines": [ + { + "bbox": [ + 111, + 371, + 292, + 387 + ], + "spans": [ + { + "bbox": [ + 111, + 371, + 292, + 387 + ], + "score": 1.0, + "content": "1: Initialize a popup score for every weight", + "type": "text" + } + ], + "index": 22, + "is_list_start_line": true + }, + { + "bbox": [ + 110, + 385, + 179, + 398 + ], + "spans": [ + { + "bbox": [ + 110, + 385, + 138, + 398 + ], + "score": 1.0, + "content": "2: for", + "type": "text" + }, + { + "bbox": [ + 138, + 386, + 164, + 396 + ], + "score": 0.84, + "content": "w \\in { \\boldsymbol { \\theta } }", + "type": "inline_equation" + }, + { + "bbox": [ + 165, + 385, + 179, + 398 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 23, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 397, + 197, + 409 + ], + "spans": [ + { + "bbox": [ + 111, + 398, + 121, + 408 + ], + "score": 1.0, + "content": "3:", + "type": "text" + }, + { + "bbox": [ + 133, + 397, + 197, + 409 + ], + "score": 0.83, + "content": "s _ { w } \\sim N ( \\mu , \\sigma ^ { 2 } )", + "type": "inline_equation" + } + ], + "index": 24, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 407, + 158, + 420 + ], + "spans": [ + { + "bbox": [ + 111, + 409, + 121, + 418 + ], + "score": 1.0, + "content": "4:", + "type": "text" + }, + { + "bbox": [ + 121, + 407, + 158, + 420 + ], + "score": 1.0, + "content": "end for", + "type": "text" + } + ], + "index": 25, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 417, + 273, + 433 + ], + "spans": [ + { + "bbox": [ + 111, + 420, + 120, + 429 + ], + "score": 1.0, + "content": "5:", + "type": "text" + }, + { + "bbox": [ + 123, + 417, + 273, + 433 + ], + "score": 1.0, + "content": "Generate mask by the popup scores", + "type": "text" + } + ], + "index": 26, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 431, + 192, + 445 + ], + "spans": [ + { + "bbox": [ + 111, + 431, + 138, + 445 + ], + "score": 1.0, + "content": "6: for", + "type": "text" + }, + { + "bbox": [ + 138, + 432, + 177, + 443 + ], + "score": 0.68, + "content": "l \\in \\theta _ { \\mathrm { l a y e r s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 177, + 431, + 192, + 445 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 27, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 444, + 239, + 457 + ], + "spans": [ + { + "bbox": [ + 111, + 445, + 122, + 455 + ], + "score": 1.0, + "content": "7:", + "type": "text" + }, + { + "bbox": [ + 133, + 444, + 239, + 457 + ], + "score": 0.4, + "content": "m _ { \\delta } ^ { l } = s ^ { l } > \\mathrm { q u a n t i l e } ( s ^ { l } , \\delta )", + "type": "inline_equation" + } + ], + "index": 28, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 455, + 157, + 466 + ], + "spans": [ + { + "bbox": [ + 111, + 456, + 121, + 465 + ], + "score": 1.0, + "content": "8:", + "type": "text" + }, + { + "bbox": [ + 122, + 455, + 157, + 466 + ], + "score": 1.0, + "content": "end for", + "type": "text" + } + ], + "index": 29, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 465, + 223, + 478 + ], + "spans": [ + { + "bbox": [ + 111, + 467, + 122, + 477 + ], + "score": 1.0, + "content": "9:", + "type": "text" + }, + { + "bbox": [ + 124, + 465, + 223, + 478 + ], + "score": 1.0, + "content": "Estimate loss constrain", + "type": "text" + } + ], + "index": 30, + "is_list_start_line": true + }, + { + "bbox": [ + 106, + 478, + 244, + 494 + ], + "spans": [ + { + "bbox": [ + 106, + 478, + 123, + 494 + ], + "score": 1.0, + "content": "10:", + "type": "text" + }, + { + "bbox": [ + 124, + 478, + 244, + 493 + ], + "score": 0.88, + "content": "\\zeta = \\mathbb { E } _ { x \\sim D _ { \\mathrm { i n } } ^ { s } } ( \\hat { \\mathcal { L } } _ { \\mathrm { C E } } ( x , m _ { \\delta } \\odot \\theta ) )", + "type": "inline_equation" + } + ], + "index": 31, + "is_list_start_line": true + }, + { + "bbox": [ + 107, + 491, + 313, + 509 + ], + "spans": [ + { + "bbox": [ + 107, + 491, + 313, + 509 + ], + "score": 1.0, + "content": "11: Unleashing Mask Adopt Pruning: fine-tuning", + "type": "text" + } + ], + "index": 32, + "is_list_start_line": true + }, + { + "bbox": [ + 108, + 507, + 191, + 522 + ], + "spans": [ + { + "bbox": [ + 108, + 507, + 123, + 522 + ], + "score": 1.0, + "content": "12:", + "type": "text" + }, + { + "bbox": [ + 124, + 508, + 191, + 521 + ], + "score": 0.84, + "content": "s ^ { ( 1 ) } \\sim N ( \\mu , \\sigma ^ { 2 } )", + "type": "inline_equation" + } + ], + "index": 33, + "is_list_start_line": true + }, + { + "bbox": [ + 107, + 519, + 212, + 533 + ], + "spans": [ + { + "bbox": [ + 107, + 519, + 138, + 533 + ], + "score": 1.0, + "content": "13: for", + "type": "text" + }, + { + "bbox": [ + 138, + 521, + 195, + 532 + ], + "score": 0.81, + "content": "t \\in ( 1 , \\ldots , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 519, + 212, + 533 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 34, + "is_list_start_line": true + }, + { + "bbox": [ + 106, + 530, + 203, + 545 + ], + "spans": [ + { + "bbox": [ + 106, + 531, + 122, + 543 + ], + "score": 1.0, + "content": "14:", + "type": "text" + }, + { + "bbox": [ + 131, + 530, + 148, + 545 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 148, + 532, + 187, + 543 + ], + "score": 0.55, + "content": "l \\in \\theta _ { \\mathrm { l a y e r s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 187, + 530, + 203, + 545 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 35, + "is_list_start_line": true + }, + { + "bbox": [ + 106, + 539, + 271, + 559 + ], + "spans": [ + { + "bbox": [ + 106, + 544, + 122, + 556 + ], + "score": 1.0, + "content": "15:", + "type": "text" + }, + { + "bbox": [ + 140, + 539, + 142, + 559 + ], + "score": 0.0, + "content": "", + "type": "text" + }, + { + "bbox": [ + 142, + 543, + 271, + 558 + ], + "score": 0.33, + "content": "\\hat { m } _ { p } ^ { l } = \\dot { s ^ { l ( t ) } } > \\mathrm { q u a n t i l e } ( s ^ { l ( t ) } , p )", + "type": "inline_equation", + "image_path": "654b43dbe7313e661d44c6af014d15ad8a172fd16e9aa8358e75b3dbaf056d34.jpg" + } + ], + "index": 36, + "is_list_start_line": true + }, + { + "bbox": [ + 106, + 554, + 168, + 567 + ], + "spans": [ + { + "bbox": [ + 106, + 555, + 122, + 567 + ], + "score": 1.0, + "content": "16:", + "type": "text" + }, + { + "bbox": [ + 132, + 554, + 168, + 567 + ], + "score": 1.0, + "content": "end for", + "type": "text" + } + ], + "index": 37, + "is_list_start_line": true + }, + { + "bbox": [ + 106, + 563, + 290, + 579 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 123, + 579 + ], + "score": 1.0, + "content": "17:", + "type": "text" + }, + { + "bbox": [ + 133, + 563, + 290, + 579 + ], + "score": 0.56, + "content": "\\begin{array} { r } { s ^ { ( t + 1 ) } = s ^ { ( t ) } - \\eta \\frac { \\partial ( | \\mathcal { L } _ { \\mathrm { C E } } ( x , \\hat { m } _ { p } \\odot \\theta ) - \\zeta | + \\zeta ) } { \\partial \\theta } } \\end{array}", + "type": "inline_equation" + } + ], + "index": 38, + "is_list_start_line": true + }, + { + "bbox": [ + 105, + 576, + 157, + 589 + ], + "spans": [ + { + "bbox": [ + 105, + 576, + 157, + 589 + ], + "score": 1.0, + "content": "18: end for", + "type": "text" + } + ], + "index": 39, + "is_list_start_line": true + } + ], + "index": 30.5, + "bbox_fs": [ + 105, + 371, + 313, + 589 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 612, + 505, + 645 + ], + "lines": [ + { + "bbox": [ + 106, + 611, + 506, + 625 + ], + "spans": [ + { + "bbox": [ + 106, + 611, + 251, + 625 + ], + "score": 1.0, + "content": "schedule, the learning rate decreases", + "type": "text" + }, + { + "bbox": [ + 252, + 612, + 271, + 623 + ], + "score": 0.89, + "content": "1 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 611, + 389, + 625 + ], + "score": 1.0, + "content": "of the initial rate (0.01) every", + "type": "text" + }, + { + "bbox": [ + 390, + 612, + 410, + 623 + ], + "score": 0.88, + "content": "1 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 410, + 611, + 506, + 625 + ], + "score": 1.0, + "content": "epochs (10 epochs); for", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 623, + 506, + 636 + ], + "spans": [ + { + "bbox": [ + 106, + 623, + 328, + 636 + ], + "score": 1.0, + "content": "the multiple step schedule, the learning rate decreases to", + "type": "text" + }, + { + "bbox": [ + 329, + 623, + 348, + 634 + ], + "score": 0.89, + "content": "1 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 623, + 506, + 636 + ], + "score": 1.0, + "content": "of the current rate every 30 epochs. All", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 634, + 465, + 646 + ], + "spans": [ + { + "bbox": [ + 106, + 634, + 465, + 646 + ], + "score": 1.0, + "content": "those learning rate schedules for our experiments are intuitively illustrated in Figure 4(a).", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41, + "bbox_fs": [ + 106, + 611, + 506, + 646 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 663, + 427, + 675 + ], + "lines": [ + { + "bbox": [ + 104, + 662, + 429, + 678 + ], + "spans": [ + { + "bbox": [ + 104, + 662, + 429, + 678 + ], + "score": 1.0, + "content": "F ADDITIONAL EXPERIMENT RESULTS AND ABLATION STUDY", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 43 + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 506, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 688, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 506, + 700 + ], + "score": 1.0, + "content": "In this section, we provide more experiment results. We first show the fine-grained results on CIFAR-", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 698, + 507, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 507, + 711 + ], + "score": 1.0, + "content": "10 and CIFAR-100, then conduct the experiments under a different model structure (i.e., WRN-40-4),", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 710, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 722 + ], + "score": 1.0, + "content": "and finally apply an additional ablation study on the proposed UM and UMAP. The mean and variance", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 720, + 506, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 732 + ], + "score": 1.0, + "content": "of all metrics (ID-ACC, AUROC, AUPR, FPR95) are reported based on multiple independent trials.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 45.5, + "bbox_fs": [ + 105, + 688, + 507, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 107, + 109, + 505, + 379 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 107, + 80, + 503, + 102 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 79, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 79, + 218, + 93 + ], + "score": 1.0, + "content": "Table 8: Completed Results", + "type": "text" + }, + { + "bbox": [ + 218, + 80, + 233, + 91 + ], + "score": 0.81, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 233, + 79, + 459, + 93 + ], + "score": 1.0, + "content": ". Comparison with competitive OOD detection baselines.", + "type": "text" + }, + { + "bbox": [ + 460, + 81, + 467, + 92 + ], + "score": 0.69, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 468, + 79, + 505, + 93 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 91, + 361, + 104 + ], + "spans": [ + { + "bbox": [ + 105, + 91, + 220, + 104 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 92, + 228, + 102 + ], + "score": 0.52, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 91, + 361, + 104 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 107, + 109, + 505, + 379 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 109, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 107, + 109, + 505, + 379 + ], + "score": 0.986, + "html": "
DinMethodAUROC↑AUPR个FPR95↓ID-ACC↑w./w.o Daux
CIFAR-10MSP(Hendrycks & Gimpel, 2017a)89.90 ±0.30 91.46 ± 0.5691.48 ± 0.43 91.67 ± 0.5860.08 ±0.76 42.31 ± 1.3894.01 ± 0.08 94.01 ± 0.08
ODIN(Liang et al.,2018)
Mahalanobis(Lee et al.,2018b)75.10 ± 1.0472.32 ± 1.9261.35 ± 1.2594.01 ± 0.08
Energy(Liu et al.,2020)92.07±0.2292.72 ± 0.3942.69 ± 1.3194.01 ± 0.08
Energy+UM(ours)93.73±0.3694.27 ± 0.6033.29 ± 1.7092.80±0.47
Energy+UMAP(ours)93.97 ± 0.1194.38 ± 0.0630.71 ± 1.9494.01 ± 0.08
OE(Hendrycks et al.,2019b)97.07 ± 0.0197.31± 0.0513.80 ± 0.2892.59 ± 0.32
Energy (w.Daux)(Liu et al.,2020)94.58 ± 0.6494.69 ± 0.6518.79 ± 2.3180.91 ± 3.13>>>>
POEM(Ming et al.,2022) OE+UM(ours)94.37 ± 0.0794.51± 0.0618.50 ± 0.3377.24 ± 2.22
Energy+UM(ours)97.60 ± 0.03 93.02±0.4297.87 ± 0.02 92.36 ± 0.3811.22 ± 0.16 24.41 ± 1.6593.66 ± 0.12 71.97 ± 0.92
POEM+UM(ours)
OE+UMAP(ours)93.04± 0.0292.99 ± 0.0223.52 ±0.1667.41 ± 0.27
Energy+UMAP(ours)95.63 ± 1.1597.48 ± 0.0197.74±0.00 95.92 ± 1.1712.21 ± 0.09 17.51 ± 2.5993.44± 0.21>>>>
POEM+UMAP (ours)94.18 ± 2.9894.15 ± 3.4620.55± 8.7088.12 ± 4.22 76.62 ± 17.95√
MSP(Hendrycks & Gimpel,2017a)74.06 ±0.6975.37 ± 0.7383.14 ± 0.8774.86 ± 0.21
ODIN(Liang et al.,2018)76.18 ± 0.1476.49 ± 0.2078.93 ± 0.3174.86 ± 0.21
Mahalanobis(Lee et al.,2018b)63.90 ± 1.9164.31 ± 0.9178.79 ± 0.5074.86 ± 0.21
Energy(Liu et al.,2020)76.29 ±0.2477.06± 0.5578.46 ± 0.0674.86 ± 0.21
Energy+UM(ours)76.22 ± 0.4276.39 ±1.0374.05 ± 0.5564.55 ± 0.24
Energy+UMAP(ours)75.57 ± 0.5975.66 ± 0.0772.21 ± 1.4674.86 ± 0.21
OE(Hendrycks et al.,2019b)90.55 ± 0.8790.34± 0.9434.73 ±3.8573.59 ± 0.30
Energy(w.Daux)(Liu et al.,2020)88.92±0.5789.13± 0.5637.90 ± 2.5957.85± 2.65>>>>
POEM(Ming et al.,2022)88.95 ±0.5488.94± 0.3138.10 ±1.3056.18 ± 1.92
91.04± 0.1191.13 ± 0.2434.71 ± 0.8175.15 ± 0.18
OE+UM(ours)90.14± 0.4532.65 ± 3.13
Energy+UM(ours)90.39 ± 0.40 91.18 ± 0.3591.45 ± 0.2730.78 ± 1.7671.95 ± 0.23 70.17 ± 0.01
POEM+UM(ours) OE+UMAP (ours)91.10 ± 0.1690.99±0.2333.62±0.2674.76 ± 0.11
Energy+UMAP(ours)90.52 ±0.2690.46 ±0.5032.17 ± 0.3072.76 ± 0.18√
POEM+UMAP (ours)91.10 ±0.2991.41 ± 0.2831.02 ± 1.7071.05 ± 0.04√
", + "type": "table", + "image_path": "062ca41fb7ee42b648df09bd68a14fa3e91b58cf9f5c49dbf4812e0737189050.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 107, + 109, + 505, + 199.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 107, + 199.0, + 505, + 289.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 107, + 289.0, + 505, + 379.0 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "text", + "bbox": [ + 107, + 408, + 505, + 452 + ], + "lines": [ + { + "bbox": [ + 105, + 407, + 506, + 420 + ], + "spans": [ + { + "bbox": [ + 105, + 407, + 506, + 420 + ], + "score": 1.0, + "content": "Empirical verification on typical/atypical data. In the following Tables 9, 10, 11, 12, and 13, we", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 418, + 505, + 432 + ], + "spans": [ + { + "bbox": [ + 105, + 418, + 505, + 432 + ], + "score": 1.0, + "content": "further conduct the experiments to identify the negative effect of learning on those atypical samples", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 430, + 505, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 430, + 505, + 441 + ], + "score": 1.0, + "content": "by comparing with a counterpart that learning only with the typical samples. The results confirm that", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 441, + 498, + 453 + ], + "spans": [ + { + "bbox": [ + 106, + 441, + 498, + 453 + ], + "score": 1.0, + "content": "the degeneration on detection performance is more likely to come from learning atypical samples.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 107, + 458, + 505, + 600 + ], + "lines": [ + { + "bbox": [ + 106, + 457, + 506, + 470 + ], + "spans": [ + { + "bbox": [ + 106, + 457, + 506, + 470 + ], + "score": 1.0, + "content": "In Table 9, we provide the main results for verification on typical/atypical samples. Intuitively, we", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 468, + 505, + 482 + ], + "spans": [ + { + "bbox": [ + 105, + 468, + 505, + 482 + ], + "score": 1.0, + "content": "intend to separate the training dataset into a typical set and an atypical set and train respectively", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 480, + 505, + 492 + ], + "spans": [ + { + "bbox": [ + 106, + 480, + 505, + 492 + ], + "score": 1.0, + "content": "on these two sets to see whether it is learning atypical samples that causes OOD performance to", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 491, + 504, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 491, + 504, + 502 + ], + "score": 1.0, + "content": "decrease during the latter part of the training phase. We force training samples through the model", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 501, + 505, + 514 + ], + "spans": [ + { + "bbox": [ + 106, + 501, + 505, + 514 + ], + "score": 1.0, + "content": "(DenseNet-101) of the 60th epoch and get the CE loss for separation. We provide the ACC of the", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 512, + 505, + 525 + ], + "spans": [ + { + "bbox": [ + 105, + 512, + 505, + 525 + ], + "score": 1.0, + "content": "generated sets on the model of the 60th epoch (ACC in the tables). The extremely low ACCs of the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 523, + 506, + 536 + ], + "spans": [ + { + "bbox": [ + 105, + 523, + 506, + 536 + ], + "score": 1.0, + "content": "atypical sets show that the model of the 60th epoch can hardly predict the samples, which meets our", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 534, + 505, + 547 + ], + "spans": [ + { + "bbox": [ + 105, + 534, + 505, + 547 + ], + "score": 1.0, + "content": "definition of the atypical sample. We then fine-tune the model of the 60th epoch with the generated", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 546, + 505, + 557 + ], + "spans": [ + { + "bbox": [ + 106, + 546, + 505, + 557 + ], + "score": 1.0, + "content": "dataset and report the OOD performance. The results show learning from only those atypical data", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 556, + 505, + 569 + ], + "spans": [ + { + "bbox": [ + 105, + 556, + 505, + 569 + ], + "score": 1.0, + "content": "fail to gain better detection performance than its counterpart, i.e., learning from only those typical", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 567, + 505, + 580 + ], + "spans": [ + { + "bbox": [ + 106, + 567, + 505, + 580 + ], + "score": 1.0, + "content": "data. Learning on those atypical samples fails to draw the suitable features for the OOD detection", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 578, + 505, + 591 + ], + "spans": [ + { + "bbox": [ + 105, + 578, + 505, + 591 + ], + "score": 1.0, + "content": "task, though it still can improve the original task performance. The experiments provide a conceptual", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 588, + 442, + 601 + ], + "spans": [ + { + "bbox": [ + 105, + 588, + 442, + 601 + ], + "score": 1.0, + "content": "verification of our conjecture which links our observation and the proposed method.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 107, + 622, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 621, + 506, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 506, + 634 + ], + "score": 1.0, + "content": "Results of fine-tuning for less epochs. UM adopts finetuning on the proposed objective for", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 632, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 632, + 505, + 645 + ], + "score": 1.0, + "content": "forgetting has shown the advantages of being cost-effective compared with train-from-scratch. For", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 643, + 505, + 656 + ], + "spans": [ + { + "bbox": [ + 106, + 643, + 505, + 656 + ], + "score": 1.0, + "content": "the tuning epochs, we show in Figures 9 and 10 that fine-tuning using UM can converge within about", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "score": 1.0, + "content": "20 epochs, indicating that we can apply our UM/UMAP for far less than 100 epochs (compared with", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 666, + 506, + 677 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 506, + 677 + ], + "score": 1.0, + "content": "train-from-scratch) to restore the better detection performance of the original well-trained model.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "It is intuitively reasonable that finetuning with the newly designed objective would benefit from", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "the well-trained model, allowing a faster convergence since the two phases consider the same task", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "with the same training data. As for the major experiments conducted in our work, finetuning adopts", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "100 epochs for better exploring and understanding its learning dynamics for research purposes, this", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 720, + 358, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 358, + 732 + ], + "score": 1.0, + "content": "configuration is indicated in the training details of Section 4.1.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 26.5 + } + ], + "page_idx": 18, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 107, + 109, + 505, + 379 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 107, + 80, + 503, + 102 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 79, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 79, + 218, + 93 + ], + "score": 1.0, + "content": "Table 8: Completed Results", + "type": "text" + }, + { + "bbox": [ + 218, + 80, + 233, + 91 + ], + "score": 0.81, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 233, + 79, + 459, + 93 + ], + "score": 1.0, + "content": ". Comparison with competitive OOD detection baselines.", + "type": "text" + }, + { + "bbox": [ + 460, + 81, + 467, + 92 + ], + "score": 0.69, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 468, + 79, + 505, + 93 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 91, + 361, + 104 + ], + "spans": [ + { + "bbox": [ + 105, + 91, + 220, + 104 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 92, + 228, + 102 + ], + "score": 0.52, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 91, + 361, + 104 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 107, + 109, + 505, + 379 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 109, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 107, + 109, + 505, + 379 + ], + "score": 0.986, + "html": "
DinMethodAUROC↑AUPR个FPR95↓ID-ACC↑w./w.o Daux
CIFAR-10MSP(Hendrycks & Gimpel, 2017a)89.90 ±0.30 91.46 ± 0.5691.48 ± 0.43 91.67 ± 0.5860.08 ±0.76 42.31 ± 1.3894.01 ± 0.08 94.01 ± 0.08
ODIN(Liang et al.,2018)
Mahalanobis(Lee et al.,2018b)75.10 ± 1.0472.32 ± 1.9261.35 ± 1.2594.01 ± 0.08
Energy(Liu et al.,2020)92.07±0.2292.72 ± 0.3942.69 ± 1.3194.01 ± 0.08
Energy+UM(ours)93.73±0.3694.27 ± 0.6033.29 ± 1.7092.80±0.47
Energy+UMAP(ours)93.97 ± 0.1194.38 ± 0.0630.71 ± 1.9494.01 ± 0.08
OE(Hendrycks et al.,2019b)97.07 ± 0.0197.31± 0.0513.80 ± 0.2892.59 ± 0.32
Energy (w.Daux)(Liu et al.,2020)94.58 ± 0.6494.69 ± 0.6518.79 ± 2.3180.91 ± 3.13>>>>
POEM(Ming et al.,2022) OE+UM(ours)94.37 ± 0.0794.51± 0.0618.50 ± 0.3377.24 ± 2.22
Energy+UM(ours)97.60 ± 0.03 93.02±0.4297.87 ± 0.02 92.36 ± 0.3811.22 ± 0.16 24.41 ± 1.6593.66 ± 0.12 71.97 ± 0.92
POEM+UM(ours)
OE+UMAP(ours)93.04± 0.0292.99 ± 0.0223.52 ±0.1667.41 ± 0.27
Energy+UMAP(ours)95.63 ± 1.1597.48 ± 0.0197.74±0.00 95.92 ± 1.1712.21 ± 0.09 17.51 ± 2.5993.44± 0.21>>>>
POEM+UMAP (ours)94.18 ± 2.9894.15 ± 3.4620.55± 8.7088.12 ± 4.22 76.62 ± 17.95√
MSP(Hendrycks & Gimpel,2017a)74.06 ±0.6975.37 ± 0.7383.14 ± 0.8774.86 ± 0.21
ODIN(Liang et al.,2018)76.18 ± 0.1476.49 ± 0.2078.93 ± 0.3174.86 ± 0.21
Mahalanobis(Lee et al.,2018b)63.90 ± 1.9164.31 ± 0.9178.79 ± 0.5074.86 ± 0.21
Energy(Liu et al.,2020)76.29 ±0.2477.06± 0.5578.46 ± 0.0674.86 ± 0.21
Energy+UM(ours)76.22 ± 0.4276.39 ±1.0374.05 ± 0.5564.55 ± 0.24
Energy+UMAP(ours)75.57 ± 0.5975.66 ± 0.0772.21 ± 1.4674.86 ± 0.21
OE(Hendrycks et al.,2019b)90.55 ± 0.8790.34± 0.9434.73 ±3.8573.59 ± 0.30
Energy(w.Daux)(Liu et al.,2020)88.92±0.5789.13± 0.5637.90 ± 2.5957.85± 2.65>>>>
POEM(Ming et al.,2022)88.95 ±0.5488.94± 0.3138.10 ±1.3056.18 ± 1.92
91.04± 0.1191.13 ± 0.2434.71 ± 0.8175.15 ± 0.18
OE+UM(ours)90.14± 0.4532.65 ± 3.13
Energy+UM(ours)90.39 ± 0.40 91.18 ± 0.3591.45 ± 0.2730.78 ± 1.7671.95 ± 0.23 70.17 ± 0.01
POEM+UM(ours) OE+UMAP (ours)91.10 ± 0.1690.99±0.2333.62±0.2674.76 ± 0.11
Energy+UMAP(ours)90.52 ±0.2690.46 ±0.5032.17 ± 0.3072.76 ± 0.18√
POEM+UMAP (ours)91.10 ±0.2991.41 ± 0.2831.02 ± 1.7071.05 ± 0.04√
", + "type": "table", + "image_path": "062ca41fb7ee42b648df09bd68a14fa3e91b58cf9f5c49dbf4812e0737189050.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 107, + 109, + 505, + 199.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 107, + 199.0, + 505, + 289.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 107, + 289.0, + 505, + 379.0 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "text", + "bbox": [ + 107, + 408, + 505, + 452 + ], + "lines": [ + { + "bbox": [ + 105, + 407, + 506, + 420 + ], + "spans": [ + { + "bbox": [ + 105, + 407, + 506, + 420 + ], + "score": 1.0, + "content": "Empirical verification on typical/atypical data. In the following Tables 9, 10, 11, 12, and 13, we", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 418, + 505, + 432 + ], + "spans": [ + { + "bbox": [ + 105, + 418, + 505, + 432 + ], + "score": 1.0, + "content": "further conduct the experiments to identify the negative effect of learning on those atypical samples", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 430, + 505, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 430, + 505, + 441 + ], + "score": 1.0, + "content": "by comparing with a counterpart that learning only with the typical samples. The results confirm that", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 441, + 498, + 453 + ], + "spans": [ + { + "bbox": [ + 106, + 441, + 498, + 453 + ], + "score": 1.0, + "content": "the degeneration on detection performance is more likely to come from learning atypical samples.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 6.5, + "bbox_fs": [ + 105, + 407, + 506, + 453 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 458, + 505, + 600 + ], + "lines": [ + { + "bbox": [ + 106, + 457, + 506, + 470 + ], + "spans": [ + { + "bbox": [ + 106, + 457, + 506, + 470 + ], + "score": 1.0, + "content": "In Table 9, we provide the main results for verification on typical/atypical samples. Intuitively, we", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 468, + 505, + 482 + ], + "spans": [ + { + "bbox": [ + 105, + 468, + 505, + 482 + ], + "score": 1.0, + "content": "intend to separate the training dataset into a typical set and an atypical set and train respectively", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 480, + 505, + 492 + ], + "spans": [ + { + "bbox": [ + 106, + 480, + 505, + 492 + ], + "score": 1.0, + "content": "on these two sets to see whether it is learning atypical samples that causes OOD performance to", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 491, + 504, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 491, + 504, + 502 + ], + "score": 1.0, + "content": "decrease during the latter part of the training phase. We force training samples through the model", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 501, + 505, + 514 + ], + "spans": [ + { + "bbox": [ + 106, + 501, + 505, + 514 + ], + "score": 1.0, + "content": "(DenseNet-101) of the 60th epoch and get the CE loss for separation. We provide the ACC of the", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 512, + 505, + 525 + ], + "spans": [ + { + "bbox": [ + 105, + 512, + 505, + 525 + ], + "score": 1.0, + "content": "generated sets on the model of the 60th epoch (ACC in the tables). The extremely low ACCs of the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 523, + 506, + 536 + ], + "spans": [ + { + "bbox": [ + 105, + 523, + 506, + 536 + ], + "score": 1.0, + "content": "atypical sets show that the model of the 60th epoch can hardly predict the samples, which meets our", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 534, + 505, + 547 + ], + "spans": [ + { + "bbox": [ + 105, + 534, + 505, + 547 + ], + "score": 1.0, + "content": "definition of the atypical sample. We then fine-tune the model of the 60th epoch with the generated", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 546, + 505, + 557 + ], + "spans": [ + { + "bbox": [ + 106, + 546, + 505, + 557 + ], + "score": 1.0, + "content": "dataset and report the OOD performance. The results show learning from only those atypical data", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 556, + 505, + 569 + ], + "spans": [ + { + "bbox": [ + 105, + 556, + 505, + 569 + ], + "score": 1.0, + "content": "fail to gain better detection performance than its counterpart, i.e., learning from only those typical", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 567, + 505, + 580 + ], + "spans": [ + { + "bbox": [ + 106, + 567, + 505, + 580 + ], + "score": 1.0, + "content": "data. Learning on those atypical samples fails to draw the suitable features for the OOD detection", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 578, + 505, + 591 + ], + "spans": [ + { + "bbox": [ + 105, + 578, + 505, + 591 + ], + "score": 1.0, + "content": "task, though it still can improve the original task performance. The experiments provide a conceptual", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 588, + 442, + 601 + ], + "spans": [ + { + "bbox": [ + 105, + 588, + 442, + 601 + ], + "score": 1.0, + "content": "verification of our conjecture which links our observation and the proposed method.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 15, + "bbox_fs": [ + 105, + 457, + 506, + 601 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 622, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 621, + 506, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 506, + 634 + ], + "score": 1.0, + "content": "Results of fine-tuning for less epochs. UM adopts finetuning on the proposed objective for", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 632, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 632, + 505, + 645 + ], + "score": 1.0, + "content": "forgetting has shown the advantages of being cost-effective compared with train-from-scratch. For", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 643, + 505, + 656 + ], + "spans": [ + { + "bbox": [ + 106, + 643, + 505, + 656 + ], + "score": 1.0, + "content": "the tuning epochs, we show in Figures 9 and 10 that fine-tuning using UM can converge within about", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 667 + ], + "score": 1.0, + "content": "20 epochs, indicating that we can apply our UM/UMAP for far less than 100 epochs (compared with", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 666, + 506, + 677 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 506, + 677 + ], + "score": 1.0, + "content": "train-from-scratch) to restore the better detection performance of the original well-trained model.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "It is intuitively reasonable that finetuning with the newly designed objective would benefit from", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "the well-trained model, allowing a faster convergence since the two phases consider the same task", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "with the same training data. As for the major experiments conducted in our work, finetuning adopts", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "100 epochs for better exploring and understanding its learning dynamics for research purposes, this", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 720, + 358, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 358, + 732 + ], + "score": 1.0, + "content": "configuration is indicated in the training details of Section 4.1.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 26.5, + "bbox_fs": [ + 105, + 621, + 506, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 109, + 505, + 234 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 504, + 103 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 437, + 93 + ], + "score": 1.0, + "content": "Table 9: Fine-tuning on typical/atypical samples with different model structures", + "type": "text" + }, + { + "bbox": [ + 437, + 80, + 451, + 92 + ], + "score": 0.43, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 452, + 80, + 457, + 93 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 457, + 81, + 466, + 92 + ], + "score": 0.52, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 80, + 505, + 93 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 91, + 361, + 104 + ], + "spans": [ + { + "bbox": [ + 106, + 91, + 220, + 104 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 92, + 228, + 102 + ], + "score": 0.48, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 91, + 361, + 104 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 108, + 109, + 505, + 234 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 109, + 505, + 234 + ], + "spans": [ + { + "bbox": [ + 108, + 109, + 505, + 234 + ], + "score": 0.984, + "html": "
DinDataset SizeStructureAtypical/TypicalAUROC↑AUPR↑FPR95↓
CIFAR-10200DenseNet-101Atypical Typical81.45 82.8682.40 84.3862.10 60.01
WRN-40-4Atypical Typical85.13 86.2686.57 86.8966.41
DenseNet-101Atypical Typical71.96 74.7973.16 75.8359.93 85.57
CIFAR-1001000WRN-40-4Atypical Typical66.64 71.9567.41 72.0280.97 86.92 80.00
", + "type": "table", + "image_path": "5aed1946c9645f06bcccbbf0abdca6c6cad363a926c926ad7e6d449c9e74e6fc.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 108, + 109, + 505, + 150.66666666666666 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 108, + 150.66666666666666, + 505, + 192.33333333333331 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 108, + 192.33333333333331, + 505, + 233.99999999999997 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "table", + "bbox": [ + 109, + 274, + 505, + 381 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 245, + 503, + 269 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 244, + 505, + 259 + ], + "spans": [ + { + "bbox": [ + 105, + 244, + 438, + 259 + ], + "score": 1.0, + "content": "Table 10: Fine-tuning on typical/atypical CIFAR-10 samples with DenseNet-101", + "type": "text" + }, + { + "bbox": [ + 438, + 245, + 453, + 257 + ], + "score": 0.65, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 453, + 244, + 457, + 259 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 458, + 246, + 466, + 257 + ], + "score": 0.38, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 244, + 505, + 259 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 257, + 361, + 270 + ], + "spans": [ + { + "bbox": [ + 106, + 257, + 220, + 270 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 257, + 228, + 268 + ], + "score": 0.68, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 257, + 361, + 270 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + }, + { + "type": "table_body", + "bbox": [ + 109, + 274, + 505, + 381 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 109, + 274, + 505, + 381 + ], + "spans": [ + { + "bbox": [ + 109, + 274, + 505, + 381 + ], + "score": 0.981, + "html": "
DinDataset SizeAtypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-10200Atypical Typical3.50 100.0081.45 82.8682.40 83.4862.10 60.01
350Atypical11.14 100.0085.9086.0155.10
500Typical Atypical16.8085.90 84.9486.16 85.3352.81 59.27
Typical100.0085.5386.1058.74
", + "type": "table", + "image_path": "41341d0a19719a641a9201de0fb28a0d33c42b7e5f9abb12cd8ce42a17e3219c.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 109, + 274, + 505, + 309.6666666666667 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 109, + 309.6666666666667, + 505, + 345.33333333333337 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 109, + 345.33333333333337, + 505, + 381.00000000000006 + ], + "spans": [], + "index": 9 + } + ] + } + ], + "index": 6.75 + }, + { + "type": "table", + "bbox": [ + 109, + 423, + 505, + 529 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 393, + 504, + 416 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 106, + 393, + 505, + 406 + ], + "spans": [ + { + "bbox": [ + 106, + 393, + 413, + 406 + ], + "score": 1.0, + "content": "Table 11: Fine-tuning on typical/atypical CIFAR-10 samples with WRN-40-4", + "type": "text" + }, + { + "bbox": [ + 413, + 393, + 428, + 405 + ], + "score": 0.73, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 393, + 432, + 406 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 433, + 394, + 440, + 405 + ], + "score": 0.57, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 440, + 393, + 505, + 406 + ], + "score": 1.0, + "content": "indicates higher", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 404, + 333, + 417 + ], + "spans": [ + { + "bbox": [ + 106, + 404, + 192, + 417 + ], + "score": 1.0, + "content": "values are better, and", + "type": "text" + }, + { + "bbox": [ + 192, + 405, + 200, + 416 + ], + "score": 0.74, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 404, + 333, + 417 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5 + }, + { + "type": "table_body", + "bbox": [ + 109, + 423, + 505, + 529 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 109, + 423, + 505, + 529 + ], + "spans": [ + { + "bbox": [ + 109, + 423, + 505, + 529 + ], + "score": 0.981, + "html": "
DinDataset SizeAtypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-10200Atypical Typical3.50 100.0085.13 86.2686.57 86.8966.41 59.93
350Atypical11.1482.9284.2468.57
Typical Atypical100.00 16.8085.8287.8465.54
500Typical100.0082.88 87.3883.22 87.9366.75 52.27
", + "type": "table", + "image_path": "04631571d0fae549eb9e4a58dda2f7a5e481b5a024a6e1c3734b8ec291637d1c.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 109, + 423, + 505, + 458.3333333333333 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 109, + 458.3333333333333, + 505, + 493.66666666666663 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 109, + 493.66666666666663, + 505, + 529.0 + ], + "spans": [], + "index": 14 + } + ] + } + ], + "index": 11.75 + }, + { + "type": "table", + "bbox": [ + 109, + 570, + 505, + 676 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 542, + 502, + 564 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 105, + 542, + 505, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 439, + 554 + ], + "score": 1.0, + "content": "Table 12: Fine-tuning on typical/atypical CIFAR-100 samples with DenseNet-101", + "type": "text" + }, + { + "bbox": [ + 439, + 542, + 453, + 553 + ], + "score": 0.75, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 454, + 542, + 505, + 554 + ], + "score": 1.0, + "content": ". ↑ indicates", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 552, + 361, + 565 + ], + "spans": [ + { + "bbox": [ + 105, + 552, + 220, + 565 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 553, + 228, + 564 + ], + "score": 0.62, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 552, + 361, + 565 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5 + }, + { + "type": "table_body", + "bbox": [ + 109, + 570, + 505, + 676 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 109, + 570, + 505, + 676 + ], + "spans": [ + { + "bbox": [ + 109, + 570, + 505, + 676 + ], + "score": 0.981, + "html": "
DinDataset Size Atypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-100500Atypical Typical1.00 100.0072.69 74.0773.28 75.2080.71 80.19
800Atypical2.8869.7471.1585.46
Typical Atypical100.00 3.5072.49 71.9673.17 73.1681.97
1000Typical100.0074.7975.8385.57 80.97
", + "type": "table", + "image_path": "3b7899f69d285eacdfc1b81f218d626770641869566f8065f253fcbd065face8.jpg" + } + ] + } + ], + "index": 18, + "virtual_lines": [ + { + "bbox": [ + 109, + 570, + 505, + 605.3333333333334 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 109, + 605.3333333333334, + 505, + 640.6666666666667 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 109, + 640.6666666666667, + 505, + 676.0000000000001 + ], + "spans": [], + "index": 19 + } + ] + } + ], + "index": 16.75 + }, + { + "type": "text", + "bbox": [ + 107, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 697, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 697, + 506, + 711 + ], + "score": 1.0, + "content": "We also provide an extra comparison to show the relative efficiency of our proposed UM/UMAP in", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 708, + 505, + 721 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 505, + 721 + ], + "score": 1.0, + "content": "the following Table 14 and Table 15. The results show that UM and UMAP can efficiently restore", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 720, + 505, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 505, + 732 + ], + "score": 1.0, + "content": "detection performance compared with the baseline. It is intuitively reasonable that fine-tuning would", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21 + } + ], + "page_idx": 19, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 109, + 505, + 234 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 504, + 103 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 437, + 93 + ], + "score": 1.0, + "content": "Table 9: Fine-tuning on typical/atypical samples with different model structures", + "type": "text" + }, + { + "bbox": [ + 437, + 80, + 451, + 92 + ], + "score": 0.43, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 452, + 80, + 457, + 93 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 457, + 81, + 466, + 92 + ], + "score": 0.52, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 80, + 505, + 93 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 91, + 361, + 104 + ], + "spans": [ + { + "bbox": [ + 106, + 91, + 220, + 104 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 92, + 228, + 102 + ], + "score": 0.48, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 91, + 361, + 104 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 108, + 109, + 505, + 234 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 109, + 505, + 234 + ], + "spans": [ + { + "bbox": [ + 108, + 109, + 505, + 234 + ], + "score": 0.984, + "html": "
DinDataset SizeStructureAtypical/TypicalAUROC↑AUPR↑FPR95↓
CIFAR-10200DenseNet-101Atypical Typical81.45 82.8682.40 84.3862.10 60.01
WRN-40-4Atypical Typical85.13 86.2686.57 86.8966.41
DenseNet-101Atypical Typical71.96 74.7973.16 75.8359.93 85.57
CIFAR-1001000WRN-40-4Atypical Typical66.64 71.9567.41 72.0280.97 86.92 80.00
", + "type": "table", + "image_path": "5aed1946c9645f06bcccbbf0abdca6c6cad363a926c926ad7e6d449c9e74e6fc.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 108, + 109, + 505, + 150.66666666666666 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 108, + 150.66666666666666, + 505, + 192.33333333333331 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 108, + 192.33333333333331, + 505, + 233.99999999999997 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "table", + "bbox": [ + 109, + 274, + 505, + 381 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 245, + 503, + 269 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 244, + 505, + 259 + ], + "spans": [ + { + "bbox": [ + 105, + 244, + 438, + 259 + ], + "score": 1.0, + "content": "Table 10: Fine-tuning on typical/atypical CIFAR-10 samples with DenseNet-101", + "type": "text" + }, + { + "bbox": [ + 438, + 245, + 453, + 257 + ], + "score": 0.65, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 453, + 244, + 457, + 259 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 458, + 246, + 466, + 257 + ], + "score": 0.38, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 244, + 505, + 259 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 257, + 361, + 270 + ], + "spans": [ + { + "bbox": [ + 106, + 257, + 220, + 270 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 257, + 228, + 268 + ], + "score": 0.68, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 257, + 361, + 270 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + }, + { + "type": "table_body", + "bbox": [ + 109, + 274, + 505, + 381 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 109, + 274, + 505, + 381 + ], + "spans": [ + { + "bbox": [ + 109, + 274, + 505, + 381 + ], + "score": 0.981, + "html": "
DinDataset SizeAtypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-10200Atypical Typical3.50 100.0081.45 82.8682.40 83.4862.10 60.01
350Atypical11.14 100.0085.9086.0155.10
500Typical Atypical16.8085.90 84.9486.16 85.3352.81 59.27
Typical100.0085.5386.1058.74
", + "type": "table", + "image_path": "41341d0a19719a641a9201de0fb28a0d33c42b7e5f9abb12cd8ce42a17e3219c.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 109, + 274, + 505, + 309.6666666666667 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 109, + 309.6666666666667, + 505, + 345.33333333333337 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 109, + 345.33333333333337, + 505, + 381.00000000000006 + ], + "spans": [], + "index": 9 + } + ] + } + ], + "index": 6.75 + }, + { + "type": "table", + "bbox": [ + 109, + 423, + 505, + 529 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 393, + 504, + 416 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 106, + 393, + 505, + 406 + ], + "spans": [ + { + "bbox": [ + 106, + 393, + 413, + 406 + ], + "score": 1.0, + "content": "Table 11: Fine-tuning on typical/atypical CIFAR-10 samples with WRN-40-4", + "type": "text" + }, + { + "bbox": [ + 413, + 393, + 428, + 405 + ], + "score": 0.73, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 393, + 432, + 406 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 433, + 394, + 440, + 405 + ], + "score": 0.57, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 440, + 393, + 505, + 406 + ], + "score": 1.0, + "content": "indicates higher", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 404, + 333, + 417 + ], + "spans": [ + { + "bbox": [ + 106, + 404, + 192, + 417 + ], + "score": 1.0, + "content": "values are better, and", + "type": "text" + }, + { + "bbox": [ + 192, + 405, + 200, + 416 + ], + "score": 0.74, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 404, + 333, + 417 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5 + }, + { + "type": "table_body", + "bbox": [ + 109, + 423, + 505, + 529 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 109, + 423, + 505, + 529 + ], + "spans": [ + { + "bbox": [ + 109, + 423, + 505, + 529 + ], + "score": 0.981, + "html": "
DinDataset SizeAtypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-10200Atypical Typical3.50 100.0085.13 86.2686.57 86.8966.41 59.93
350Atypical11.1482.9284.2468.57
Typical Atypical100.00 16.8085.8287.8465.54
500Typical100.0082.88 87.3883.22 87.9366.75 52.27
", + "type": "table", + "image_path": "04631571d0fae549eb9e4a58dda2f7a5e481b5a024a6e1c3734b8ec291637d1c.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 109, + 423, + 505, + 458.3333333333333 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 109, + 458.3333333333333, + 505, + 493.66666666666663 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 109, + 493.66666666666663, + 505, + 529.0 + ], + "spans": [], + "index": 14 + } + ] + } + ], + "index": 11.75 + }, + { + "type": "table", + "bbox": [ + 109, + 570, + 505, + 676 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 542, + 502, + 564 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 105, + 542, + 505, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 439, + 554 + ], + "score": 1.0, + "content": "Table 12: Fine-tuning on typical/atypical CIFAR-100 samples with DenseNet-101", + "type": "text" + }, + { + "bbox": [ + 439, + 542, + 453, + 553 + ], + "score": 0.75, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 454, + 542, + 505, + 554 + ], + "score": 1.0, + "content": ". ↑ indicates", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 552, + 361, + 565 + ], + "spans": [ + { + "bbox": [ + 105, + 552, + 220, + 565 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 553, + 228, + 564 + ], + "score": 0.62, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 552, + 361, + 565 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5 + }, + { + "type": "table_body", + "bbox": [ + 109, + 570, + 505, + 676 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 109, + 570, + 505, + 676 + ], + "spans": [ + { + "bbox": [ + 109, + 570, + 505, + 676 + ], + "score": 0.981, + "html": "
DinDataset Size Atypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-100500Atypical Typical1.00 100.0072.69 74.0773.28 75.2080.71 80.19
800Atypical2.8869.7471.1585.46
Typical Atypical100.00 3.5072.49 71.9673.17 73.1681.97
1000Typical100.0074.7975.8385.57 80.97
", + "type": "table", + "image_path": "3b7899f69d285eacdfc1b81f218d626770641869566f8065f253fcbd065face8.jpg" + } + ] + } + ], + "index": 18, + "virtual_lines": [ + { + "bbox": [ + 109, + 570, + 505, + 605.3333333333334 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 109, + 605.3333333333334, + 505, + 640.6666666666667 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 109, + 640.6666666666667, + 505, + 676.0000000000001 + ], + "spans": [], + "index": 19 + } + ] + } + ], + "index": 16.75 + }, + { + "type": "text", + "bbox": [ + 107, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 697, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 697, + 506, + 711 + ], + "score": 1.0, + "content": "We also provide an extra comparison to show the relative efficiency of our proposed UM/UMAP in", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 708, + 505, + 721 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 505, + 721 + ], + "score": 1.0, + "content": "the following Table 14 and Table 15. The results show that UM and UMAP can efficiently restore", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 720, + 505, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 505, + 732 + ], + "score": 1.0, + "content": "detection performance compared with the baseline. It is intuitively reasonable that fine-tuning would", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 236, + 506, + 249 + ], + "spans": [ + { + "bbox": [ + 105, + 236, + 506, + 249 + ], + "score": 1.0, + "content": "benefit from the well-trained model, allowing a faster convergence as the two phases consider the", + "type": "text", + "cross_page": true + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 247, + 506, + 261 + ], + "spans": [ + { + "bbox": [ + 105, + 247, + 506, + 261 + ], + "score": 1.0, + "content": "same task and training data. Considering the significance of the OOD awareness for those safety-", + "type": "text", + "cross_page": true + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 258, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 106, + 258, + 505, + 271 + ], + "score": 1.0, + "content": "critical areas, it is worthwhile to further excavate the OOD detection capability of the deployed", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 269, + 292, + 281 + ], + "spans": [ + { + "bbox": [ + 106, + 269, + 292, + 281 + ], + "score": 1.0, + "content": "well-trained model using our UM and UMAP.", + "type": "text", + "cross_page": true + } + ], + "index": 8 + } + ], + "index": 21, + "bbox_fs": [ + 105, + 697, + 506, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 109, + 505, + 215 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 504, + 103 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 435, + 93 + ], + "score": 1.0, + "content": "Table 13: Fine-tuning on typical/atypical CIFAR-100 samples with WRN-40-4", + "type": "text" + }, + { + "bbox": [ + 436, + 80, + 451, + 92 + ], + "score": 0.71, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 80, + 457, + 93 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 457, + 80, + 465, + 92 + ], + "score": 0.5, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 80, + 505, + 93 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 91, + 361, + 104 + ], + "spans": [ + { + "bbox": [ + 106, + 91, + 220, + 104 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 92, + 228, + 102 + ], + "score": 0.41, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 91, + 361, + 104 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 108, + 109, + 505, + 215 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 109, + 505, + 215 + ], + "spans": [ + { + "bbox": [ + 108, + 109, + 505, + 215 + ], + "score": 0.984, + "html": "
DinDataset Size Atypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-100500Atypical Typical1.00 100.0066.03 68.6066.17 69.9389.56 86.53
800Atypical2.8867.5968.6685.61
1000Typical Atypical100.00 3.5070.25 66.6468.95 67.4179.66 86.92
Typical100.0071.9572.0280.00
", + "type": "table", + "image_path": "0e79c48e21bf60c61eb8bca60e53b9117c40a5168d1316ab43e92c5ae5305b26.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 108, + 109, + 505, + 144.33333333333334 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 108, + 144.33333333333334, + 505, + 179.66666666666669 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 108, + 179.66666666666669, + 505, + 215.00000000000003 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "text", + "bbox": [ + 106, + 237, + 505, + 281 + ], + "lines": [ + { + "bbox": [ + 105, + 236, + 506, + 249 + ], + "spans": [ + { + "bbox": [ + 105, + 236, + 506, + 249 + ], + "score": 1.0, + "content": "benefit from the well-trained model, allowing a faster convergence as the two phases consider the", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 247, + 506, + 261 + ], + "spans": [ + { + "bbox": [ + 105, + 247, + 506, + 261 + ], + "score": 1.0, + "content": "same task and training data. Considering the significance of the OOD awareness for those safety-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 258, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 106, + 258, + 505, + 271 + ], + "score": 1.0, + "content": "critical areas, it is worthwhile to further excavate the OOD detection capability of the deployed", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 269, + 292, + 281 + ], + "spans": [ + { + "bbox": [ + 106, + 269, + 292, + 281 + ], + "score": 1.0, + "content": "well-trained model using our UM and UMAP.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 6.5 + }, + { + "type": "table", + "bbox": [ + 135, + 321, + 474, + 460 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 108, + 292, + 504, + 315 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 292, + 505, + 305 + ], + "spans": [ + { + "bbox": [ + 106, + 292, + 327, + 305 + ], + "score": 1.0, + "content": "Table 14: Fine-tuning for 20 epochs with DenseNet-101", + "type": "text" + }, + { + "bbox": [ + 328, + 292, + 342, + 304 + ], + "score": 0.72, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 343, + 292, + 347, + 305 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 348, + 293, + 355, + 304 + ], + "score": 0.52, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 356, + 292, + 505, + 305 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 303, + 246, + 315 + ], + "spans": [ + { + "bbox": [ + 106, + 304, + 113, + 315 + ], + "score": 0.53, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 114, + 303, + 246, + 315 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "table_body", + "bbox": [ + 135, + 321, + 474, + 460 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 135, + 321, + 474, + 460 + ], + "spans": [ + { + "bbox": [ + 135, + 321, + 474, + 460 + ], + "score": 0.984, + "html": "
DinEpochMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10100MSP89.9091.4860.0894.01
ODIN91.4691.6742.3194.01
Energy92.0792.7242.6994.01
Energy +UM93.7394.2733.2992.80
Energy + UMAP93.9794.3830.7194.01
MSP +UM90.3191.9953.6191.70
20ODIN +UM94.0894.6731.0191.70
Energy +UM93.6094.3233.0391.70
MSP+UMAP88.7090.3957.6994.01
ODIN+UMAP92.8893.3335.19
Energy + UMAP92.8893.3935.6094.01
94.01
", + "type": "table", + "image_path": "809c900b629a6946687476f96bee55ce8ee51dbace08c9351de633a026c6f197.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 135, + 321, + 474, + 367.3333333333333 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 135, + 367.3333333333333, + 474, + 413.66666666666663 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 135, + 413.66666666666663, + 474, + 459.99999999999994 + ], + "spans": [], + "index": 13 + } + ] + } + ], + "index": 10.75 + }, + { + "type": "table", + "bbox": [ + 135, + 507, + 474, + 646 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 105, + 478, + 505, + 500 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 104, + 476, + 506, + 491 + ], + "spans": [ + { + "bbox": [ + 104, + 476, + 318, + 491 + ], + "score": 1.0, + "content": "Table 15: Fine-tuning for 20 epochs with WRN-40-4", + "type": "text" + }, + { + "bbox": [ + 318, + 478, + 334, + 489 + ], + "score": 0.72, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 334, + 476, + 338, + 491 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 338, + 478, + 345, + 490 + ], + "score": 0.41, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 346, + 476, + 506, + 491 + ], + "score": 1.0, + "content": "indicates higher values are better, and ↓", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 489, + 239, + 500 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 239, + 500 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5 + }, + { + "type": "table_body", + "bbox": [ + 135, + 507, + 474, + 646 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 135, + 507, + 474, + 646 + ], + "spans": [ + { + "bbox": [ + 135, + 507, + 474, + 646 + ], + "score": 0.983, + "html": "
DinEpochMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10100MSP87.1287.8468.2993.86
ODIN83.2982.7465.6893.86
Energy87.6988.1658.4793.86
Energy + UM91.7492.6740.4092.68
Energy + UMAP88.8489.3150.2393.86
MSP +UM89.8691.3251.6291.96
20ODIN +UM91.9792.5841.7891.96
Energy +UM92.9593.6436.2191.96
MSP+UMAP88.7790.6161.6093.86
ODIN + UMAP90.8591.8945.7093.86
Energy + UMAP91.6692.4942.9493.86
", + "type": "table", + "image_path": "7aa54b6d164b551fcc233f0d1a80763b50c75dd589734ab7a9c8593a303ca70f.jpg" + } + ] + } + ], + "index": 17, + "virtual_lines": [ + { + "bbox": [ + 135, + 507, + 474, + 553.3333333333334 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 135, + 553.3333333333334, + 474, + 599.6666666666667 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 135, + 599.6666666666667, + 474, + 646.0000000000001 + ], + "spans": [], + "index": 18 + } + ] + } + ], + "index": 15.75 + }, + { + "type": "text", + "bbox": [ + 106, + 665, + 506, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 665, + 507, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 507, + 678 + ], + "score": 1.0, + "content": "Fine-grained results on OOD data. In order to further figure out the effectiveness of the pro-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 104, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 104, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "posed UM and UMAP on different OOD datasets, we further report the fine-grained results of our", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 104, + 686, + 507, + 702 + ], + "spans": [ + { + "bbox": [ + 104, + 686, + 507, + 702 + ], + "score": 1.0, + "content": "experiments on CIFAR-10 and CIFAR-100 with 6 OOD datasets (CIFAR-10/CIFAR-100, textures,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 698, + 507, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 507, + 712 + ], + "score": 1.0, + "content": "Places365, SUN, LSUN, iNaturalist). The results on the 6 OOD datasets show the general effective-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 341, + 722 + ], + "score": 1.0, + "content": "ness of the proposed UM as well as UMAP. In Table 16,", + "type": "text" + }, + { + "bbox": [ + 341, + 710, + 385, + 720 + ], + "score": 0.68, + "content": "\\mathbf { O E + U M }", + "type": "inline_equation" + }, + { + "bbox": [ + 386, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "can outperform all the OOD", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "baselines, and further improve the OOD performance even though the original detection performance", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21.5 + } + ], + "page_idx": 20, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 14 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 109, + 505, + 215 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 504, + 103 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 435, + 93 + ], + "score": 1.0, + "content": "Table 13: Fine-tuning on typical/atypical CIFAR-100 samples with WRN-40-4", + "type": "text" + }, + { + "bbox": [ + 436, + 80, + 451, + 92 + ], + "score": 0.71, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 80, + 457, + 93 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 457, + 80, + 465, + 92 + ], + "score": 0.5, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 80, + 505, + 93 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 91, + 361, + 104 + ], + "spans": [ + { + "bbox": [ + 106, + 91, + 220, + 104 + ], + "score": 1.0, + "content": "higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 221, + 92, + 228, + 102 + ], + "score": 0.41, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 91, + 361, + 104 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 108, + 109, + 505, + 215 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 109, + 505, + 215 + ], + "spans": [ + { + "bbox": [ + 108, + 109, + 505, + 215 + ], + "score": 0.984, + "html": "
DinDataset Size Atypical/TypicalACCAUROC↑AUPR↑FPR95↓
CIFAR-100500Atypical Typical1.00 100.0066.03 68.6066.17 69.9389.56 86.53
800Atypical2.8867.5968.6685.61
1000Typical Atypical100.00 3.5070.25 66.6468.95 67.4179.66 86.92
Typical100.0071.9572.0280.00
", + "type": "table", + "image_path": "0e79c48e21bf60c61eb8bca60e53b9117c40a5168d1316ab43e92c5ae5305b26.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 108, + 109, + 505, + 144.33333333333334 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 108, + 144.33333333333334, + 505, + 179.66666666666669 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 108, + 179.66666666666669, + 505, + 215.00000000000003 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "text", + "bbox": [ + 106, + 237, + 505, + 281 + ], + "lines": [], + "index": 6.5, + "bbox_fs": [ + 105, + 236, + 506, + 281 + ], + "lines_deleted": true + }, + { + "type": "table", + "bbox": [ + 135, + 321, + 474, + 460 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 108, + 292, + 504, + 315 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 292, + 505, + 305 + ], + "spans": [ + { + "bbox": [ + 106, + 292, + 327, + 305 + ], + "score": 1.0, + "content": "Table 14: Fine-tuning for 20 epochs with DenseNet-101", + "type": "text" + }, + { + "bbox": [ + 328, + 292, + 342, + 304 + ], + "score": 0.72, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 343, + 292, + 347, + 305 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 348, + 293, + 355, + 304 + ], + "score": 0.52, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 356, + 292, + 505, + 305 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 303, + 246, + 315 + ], + "spans": [ + { + "bbox": [ + 106, + 304, + 113, + 315 + ], + "score": 0.53, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 114, + 303, + 246, + 315 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "table_body", + "bbox": [ + 135, + 321, + 474, + 460 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 135, + 321, + 474, + 460 + ], + "spans": [ + { + "bbox": [ + 135, + 321, + 474, + 460 + ], + "score": 0.984, + "html": "
DinEpochMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10100MSP89.9091.4860.0894.01
ODIN91.4691.6742.3194.01
Energy92.0792.7242.6994.01
Energy +UM93.7394.2733.2992.80
Energy + UMAP93.9794.3830.7194.01
MSP +UM90.3191.9953.6191.70
20ODIN +UM94.0894.6731.0191.70
Energy +UM93.6094.3233.0391.70
MSP+UMAP88.7090.3957.6994.01
ODIN+UMAP92.8893.3335.19
Energy + UMAP92.8893.3935.6094.01
94.01
", + "type": "table", + "image_path": "809c900b629a6946687476f96bee55ce8ee51dbace08c9351de633a026c6f197.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 135, + 321, + 474, + 367.3333333333333 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 135, + 367.3333333333333, + 474, + 413.66666666666663 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 135, + 413.66666666666663, + 474, + 459.99999999999994 + ], + "spans": [], + "index": 13 + } + ] + } + ], + "index": 10.75 + }, + { + "type": "table", + "bbox": [ + 135, + 507, + 474, + 646 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 105, + 478, + 505, + 500 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 104, + 476, + 506, + 491 + ], + "spans": [ + { + "bbox": [ + 104, + 476, + 318, + 491 + ], + "score": 1.0, + "content": "Table 15: Fine-tuning for 20 epochs with WRN-40-4", + "type": "text" + }, + { + "bbox": [ + 318, + 478, + 334, + 489 + ], + "score": 0.72, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 334, + 476, + 338, + 491 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 338, + 478, + 345, + 490 + ], + "score": 0.41, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 346, + 476, + 506, + 491 + ], + "score": 1.0, + "content": "indicates higher values are better, and ↓", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 489, + 239, + 500 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 239, + 500 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5 + }, + { + "type": "table_body", + "bbox": [ + 135, + 507, + 474, + 646 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 135, + 507, + 474, + 646 + ], + "spans": [ + { + "bbox": [ + 135, + 507, + 474, + 646 + ], + "score": 0.983, + "html": "
DinEpochMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10100MSP87.1287.8468.2993.86
ODIN83.2982.7465.6893.86
Energy87.6988.1658.4793.86
Energy + UM91.7492.6740.4092.68
Energy + UMAP88.8489.3150.2393.86
MSP +UM89.8691.3251.6291.96
20ODIN +UM91.9792.5841.7891.96
Energy +UM92.9593.6436.2191.96
MSP+UMAP88.7790.6161.6093.86
ODIN + UMAP90.8591.8945.7093.86
Energy + UMAP91.6692.4942.9493.86
", + "type": "table", + "image_path": "7aa54b6d164b551fcc233f0d1a80763b50c75dd589734ab7a9c8593a303ca70f.jpg" + } + ] + } + ], + "index": 17, + "virtual_lines": [ + { + "bbox": [ + 135, + 507, + 474, + 553.3333333333334 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 135, + 553.3333333333334, + 474, + 599.6666666666667 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 135, + 599.6666666666667, + 474, + 646.0000000000001 + ], + "spans": [], + "index": 18 + } + ] + } + ], + "index": 15.75 + }, + { + "type": "text", + "bbox": [ + 106, + 665, + 506, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 665, + 507, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 507, + 678 + ], + "score": 1.0, + "content": "Fine-grained results on OOD data. In order to further figure out the effectiveness of the pro-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 104, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 104, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "posed UM and UMAP on different OOD datasets, we further report the fine-grained results of our", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 104, + 686, + 507, + 702 + ], + "spans": [ + { + "bbox": [ + 104, + 686, + 507, + 702 + ], + "score": 1.0, + "content": "experiments on CIFAR-10 and CIFAR-100 with 6 OOD datasets (CIFAR-10/CIFAR-100, textures,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 698, + 507, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 507, + 712 + ], + "score": 1.0, + "content": "Places365, SUN, LSUN, iNaturalist). The results on the 6 OOD datasets show the general effective-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 341, + 722 + ], + "score": 1.0, + "content": "ness of the proposed UM as well as UMAP. In Table 16,", + "type": "text" + }, + { + "bbox": [ + 341, + 710, + 385, + 720 + ], + "score": 0.68, + "content": "\\mathbf { O E + U M }", + "type": "inline_equation" + }, + { + "bbox": [ + 386, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "can outperform all the OOD", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "baselines, and further improve the OOD performance even though the original detection performance", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "score": 1.0, + "content": "is already well. By equipping with our proposed UM and UMAP, the baselines can outperform their", + "type": "text", + "cross_page": true + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 655, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 506, + 667 + ], + "score": 1.0, + "content": "counterparts on most of the OOD datasets. For instance, the FPR95 can decrease from 1.91 to 1.42.", + "type": "text", + "cross_page": true + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "score": 1.0, + "content": "In Table 17, we also take a closer check about results on CIFAR-100 with 6 OOD datasets. Our", + "type": "text", + "cross_page": true + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 677, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 506, + 690 + ], + "score": 1.0, + "content": "proposed method can almost improve all competitive baselines (either the scoring functions or the", + "type": "text", + "cross_page": true + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 687, + 507, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 379, + 702 + ], + "score": 1.0, + "content": "finetuning with auxiliary outliers) on the 6 OOD datasets. In both w.", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 379, + 688, + 402, + 699 + ], + "score": 0.91, + "content": "D _ { a u x }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 403, + 687, + 439, + 702 + ], + "score": 1.0, + "content": "and w.o.", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 440, + 688, + 463, + 699 + ], + "score": 0.9, + "content": "D _ { a u x }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 464, + 687, + 507, + 702 + ], + "score": 1.0, + "content": "scenarios,", + "type": "text", + "cross_page": true + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "Unleashing Mask can significantly excavate the intrinsic OOD detection capability of the model. In", + "type": "text", + "cross_page": true + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "score": 1.0, + "content": "addition to unleashing the excellent OOD performance, UMAP can also maintain the high ID-ACC", + "type": "text", + "cross_page": true + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 721, + 461, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 461, + 734 + ], + "score": 1.0, + "content": "by learning a binary mask instead of tuning the well-trained original parameters directly.", + "type": "text", + "cross_page": true + } + ], + "index": 19 + } + ], + "index": 21.5, + "bbox_fs": [ + 104, + 665, + 507, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 123, + 505, + 304 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 505, + 113 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 92 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 232, + 92 + ], + "score": 1.0, + "content": "Table 16: Fine-grained Results", + "type": "text" + }, + { + "bbox": [ + 232, + 80, + 248, + 91 + ], + "score": 0.78, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 248, + 80, + 505, + 92 + ], + "score": 1.0, + "content": "of DenseNet-101 on CIFAR-10. Comparison on different OOD", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 104, + 90, + 505, + 104 + ], + "spans": [ + { + "bbox": [ + 104, + 90, + 236, + 104 + ], + "score": 1.0, + "content": "benchmark datasets respectively.", + "type": "text" + }, + { + "bbox": [ + 237, + 92, + 244, + 103 + ], + "score": 0.74, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 245, + 90, + 394, + 104 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 394, + 92, + 401, + 102 + ], + "score": 0.56, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 402, + 90, + 505, + 104 + ], + "score": 1.0, + "content": "indicates lower values are", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 101, + 135, + 115 + ], + "spans": [ + { + "bbox": [ + 105, + 101, + 135, + 115 + ], + "score": 1.0, + "content": "better.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "table_body", + "bbox": [ + 108, + 123, + 505, + 304 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 123, + 505, + 304 + ], + "spans": [ + { + "bbox": [ + 108, + 123, + 505, + 304 + ], + "score": 0.983, + "html": "
ID datasetMethodOOD dataset TexturesPlaces365
CIFAR-100 FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
Energy+UMAP54.95 ± 2.6187.72 ±1.0533.59 ±1.3292.67 ±0.2332.80 ± 4.1493.57 ± 1.12
OE59.29 ± 1.3088.51 ±0.222.89 ± 0.3099.16 ±0.0511.14 ± 1.1197.50 ±0.19
Energy (w. Dauz)79.88 ± 2.4774.99 ± 2.404.27 ± 0.5798.80 ±0.1914.22 ± 3.9997.07 ±0.72
POEM82.30 ± 1.5772.74 ± 1.421.91 ± 0.4199.40 ±0.1011.24 ± 2.7096.67 ±0.48
OE+UM(ours)55.74 ± 1.4789.53 ± 0.181.42 ± 0.1599.49± 0.047.77 ±0.6998.15 ± 0.08
Energy (w.Daux)+ UM(ours)84.52 ± 0.0170.09 ± 0.478.30 ±0.8897.76 ±0.0520.27 ±1.3096.06 ±0.31
POEM+UM(ours) OE+UMAP(ours)84.87 ± 1.5668.97 ±0.394.73 ± 0.5298.88 ±0.1319.83 ± 0.3496.35 ±0.09
Energy(w.Daur)+UMAP(ours)59.05 ± 1.4189.14 ± 0.141.86 ± 0.0799.35 ±0.008.21±0.1298.07 ±0.03
POEM+UMAP(ours)75.18 ± 4.96 79.33 ± 4.1480.93 ± 4.49 76.89 ± 4.862.24 ± 1.34 2.10 ± 1.3799.25 ±0.29 99.34 ±0.309.30 ± 2.12 9.94 ± 6.9297.90 ±0.40 98.01 ± 1.11
OEMethodSUNLSUNiNaturalist
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
Energy+UMAP29.05 ±2.7894.41 ± 0.732.31± 0.8899.42 ± 0.0447.22 ± 14.0392.63 ± 2.14
8.38±0.7198.00 ±0.145.90 ± 1.4398.60± 0.215.09 ± 0.6498.76 ± 0.11
Energy (w. Dauz)10.30 ±3.8297.77 ± 0.6412.80 ± 4.6796.08 ± 1.386.93 ± 1.8698.40 ±0.32
POEM8.39±2.4298.16 ±0.439.69 ± 1.8997.25±0.583.78 ±0.9098.99 ±0.17
OE+UM(ours)5.51 ± 0.4498.55 ± 0.073.51 ± 0.4398.93 ± 0.09 92.11 ± 0.942.87 ± 0.4999.14 ± 0.09
Energy(w.Daux)+ UM(ours) POEM+UM(ours)16.13 ± 1.86 16.16 ± 0.5796.84±0.30 97.01±0.0823.27 ± 2.40 25.69 ± 0.1593.38 ± 0.2711.20 ± 2.35 9.30 ± 1.6097.54±0.42 98.05 ±0.24
OE+UMAP(ours)6.16 ±0.0298.49 ± 0.014.53 ± 0.1698.86±0.063.40 ± 0.7498.96 ±0.09
Energy(w.Dauz)+UMAP(ours)6.67 ± 1.5098.40±0.3223.50 ± 5.6194.78 ± 2.043.77 ± 2.1498.93 ±0.46
94.74 ±3.67
POEM+UMAP(ours)7.00±5.8098.46±0.9621.17 ± 12.843.63±2.7898.99 ±0.54
", + "type": "table", + "image_path": "51c49194022a3f450f05525beb0e80c2781701dee1b944e1944f81d7233e6b6e.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 108, + 123, + 505, + 183.33333333333334 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 108, + 183.33333333333334, + 505, + 243.66666666666669 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 108, + 243.66666666666669, + 505, + 304.0 + ], + "spans": [], + "index": 5 + } + ] + } + ], + "index": 2.5 + }, + { + "type": "table", + "bbox": [ + 109, + 375, + 505, + 606 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 108, + 331, + 504, + 365 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 331, + 505, + 344 + ], + "spans": [ + { + "bbox": [ + 106, + 331, + 230, + 344 + ], + "score": 1.0, + "content": "Table 17: Fine-grained Results", + "type": "text" + }, + { + "bbox": [ + 231, + 332, + 246, + 343 + ], + "score": 0.84, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 246, + 331, + 505, + 344 + ], + "score": 1.0, + "content": "of DenseNet-101 on CIFAR-100. Comparison on different OOD", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 343, + 506, + 355 + ], + "spans": [ + { + "bbox": [ + 106, + 343, + 236, + 355 + ], + "score": 1.0, + "content": "benchmark datasets respectively.", + "type": "text" + }, + { + "bbox": [ + 237, + 344, + 244, + 354 + ], + "score": 0.78, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 245, + 343, + 394, + 355 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 394, + 343, + 401, + 354 + ], + "score": 0.64, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 402, + 343, + 506, + 355 + ], + "score": 1.0, + "content": "indicates lower values are", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 353, + 135, + 366 + ], + "spans": [ + { + "bbox": [ + 105, + 353, + 135, + 366 + ], + "score": 1.0, + "content": "better.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "table_body", + "bbox": [ + 109, + 375, + 505, + 606 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 109, + 375, + 505, + 606 + ], + "spans": [ + { + "bbox": [ + 109, + 375, + 505, + 606 + ], + "score": 0.982, + "html": "
ID datasetMethodOOD datasetPlaces365
CIFAR-10 FPR95↓AUROC↑Textures FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-100MSP ODIN Mahalanobis Energy83.53± 0.33 85.29 ± 0.1775.11 ± 0.27 73.31 ± 0.24 49.60 ± 1.5186.90 ±0.18 86.45 ± 1.27 33.06± 3.7671.45 ± 0.40 71.91 ± 0.2785.83 ±0.48 84.35 ± 0.6470.54 ± 0.42 73.58 ± 0.51
98.25 ±0.0590.19 ± 1.2195.20 ±0.4953.69 ± 1.55
Energy+UM(ours)82.16 ± 0.5975.31 ± 0.2190.20 ±0.3068.98 ±0.34 65.39 ± 1.4482.39 ±0.97 77.30 ± 2.0873.78 ±0.66 76.06 ± 1.05
OE89.62 ±0.0766.12 ± 0.9386.99 ± 1.22
Energy (w.Daux) POEM90.97 ±0.46 96.14 ±0.0669.02 ± 0.39 57.52 ± 0.8814.36 ± 0.25 9.02 ±0.0695.92 ±0.12 97.16 ±0.2640.19 ± 6.97 35.18 ± 4.7390.70 ± 2.01 93.29 ± 1.14
OE+UM (ours)96.19 ±0.1655.82 ± 1.057.63 ± 1.4097.69 ±0.0632.67 ± 3.7393.94±0.68
89.61±0.0871.24 ± 0.0816.78 ± 0.2595.60 ±0.0839.77 ± 0.3491.07 ± 0.13
Energy(w.Daux)+UM(ours)95.38 ± 0.4563.41 ± 0.146.41 ± 0.8397.77 ± 0.3330.96 ± 3.6192.85 ±0.69
POEM+UM(ours)95.78 ± 0.1460.23 ± 0.705.17 ±0.1898.53± 0.0323.90 ±0.8495.45 ± 0.11
OE+UMAP(ours) Energy(w.Daux)+UMAP(ours)90.72±0.3569.76 ± 0.2515.32 ± 0.2395.72 ± 0.0136.42 ± 1.9192.08 ±0.49
POEM+UMAP(ours)95.39±0.10 95.69 ± 0.1763.26 ±0.18 61.62 ± 0.246.52 ±0.44 5.23 ± 0.5897.83 ±0.18 98.52 ± 0.0131.18 ± 0.43 26.06 ± 1.1693.13±0.41 94.91 ± 0.25
MethodSUNLSUN
FPR95↓ AUROC↑FPR95↓AUROC↑iNaturalist FPR95↓AUROC↑
MSP88.75 ±0.2366.75 ± 0.2567.83 ±1.3782.94± 0.3285.00 ± 0.7376.62 ± 0.25
ODIN88.49 ± 0.9969.64± 0.61 54.37 ± 1.3534.80 ± 2.55 89.31 ± 4.8393.92 ±0.75 43.19 ± 16.3681.67 ± 2.77 93.63 ± 1.1978.36 ± 1.57
Mahalanobis95.53±0.3749.60 ± 1.51
Energy Energy+UM(ours)97.17 ± 0.92 69.04 ± 0.8335.09 ± 3.1793.49 ±0.8785.70 ± 2.1475.82 ± 1.72
81.96 ± 2.2671.47 ± 1.8822.54± 5.9394.98 ± 1.7574.28 ± 3.7280.72 ± 3.75
OE44.47 ± 9.10 90.70 ± 2.015.75 ± 1.1898.57 ± 0.1325.51 ± 4.1294.46 ± 0.88
Energy (w. Daux)32.69 ± 5.6993.63 ± 1.4855.75 ± 4.3187.96 ± 1.0317.34 ± 4.5496.50 ±0.81
POEM30.45 ± 5.1194.26 ± 0.9046.68 ± 3.5990.30 ± 2.1716.50 ± 2.0996.63±0.23
OE+UM(ours)44.23±0.20 90.28±0.035.80 ±0.3398.63 ±0.0326.72 ± 1.9594.51 ± 0.43
93.18 ± 0.6991.98 ± 0.8310.83 ± 2.0697.09 ±0.69
Energy(w.Daux)+UM(ours) POEM+UM(ours)28.98 ±3.1537.56 ± 4.8194.43 ± 1.18
21.34 ± 1.0795.76 ±0.2233.74 ±6.2298.64 ± 0.128.85±0.2397.93 ± 0.03 95.08 ±0.22
OE+UMAP(ours) Energy(w.Daux)+UMAP(ours)39.58 ± 2.0291.37 ± 0.52 93.38 ± 0.195.77 ± 0.7192.08 ±0.3923.33 ± 1.24 13.96 ± 2.4896.87 ± 0.15
29.65 ± 1.0635.94 ± 0.7593.57 ±0.88
POEM+UMAP(ours)23.73 ± 0.7195.25 ± 0.0633.09 ± 5.949.76 ± 1.0997.77 ± 0.24
", + "type": "table", + "image_path": "bbcdb0726730fb08f597974785e84bf8ec7f9af7e92c5725afbf29197e39ed92.jpg" + } + ] + } + ], + "index": 10, + "virtual_lines": [ + { + "bbox": [ + 109, + 375, + 505, + 452.0 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 109, + 452.0, + 505, + 529.0 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 109, + 529.0, + 505, + 606.0 + ], + "spans": [], + "index": 11 + } + ] + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 107, + 643, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "score": 1.0, + "content": "is already well. By equipping with our proposed UM and UMAP, the baselines can outperform their", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 655, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 506, + 667 + ], + "score": 1.0, + "content": "counterparts on most of the OOD datasets. For instance, the FPR95 can decrease from 1.91 to 1.42.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "score": 1.0, + "content": "In Table 17, we also take a closer check about results on CIFAR-100 with 6 OOD datasets. Our", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 677, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 506, + 690 + ], + "score": 1.0, + "content": "proposed method can almost improve all competitive baselines (either the scoring functions or the", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 687, + 507, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 379, + 702 + ], + "score": 1.0, + "content": "finetuning with auxiliary outliers) on the 6 OOD datasets. In both w.", + "type": "text" + }, + { + "bbox": [ + 379, + 688, + 402, + 699 + ], + "score": 0.91, + "content": "D _ { a u x }", + "type": "inline_equation" + }, + { + "bbox": [ + 403, + 687, + 439, + 702 + ], + "score": 1.0, + "content": "and w.o.", + "type": "text" + }, + { + "bbox": [ + 440, + 688, + 463, + 699 + ], + "score": 0.9, + "content": "D _ { a u x }", + "type": "inline_equation" + }, + { + "bbox": [ + 464, + 687, + 507, + 702 + ], + "score": 1.0, + "content": "scenarios,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "Unleashing Mask can significantly excavate the intrinsic OOD detection capability of the model. In", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "score": 1.0, + "content": "addition to unleashing the excellent OOD performance, UMAP can also maintain the high ID-ACC", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 721, + 461, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 461, + 734 + ], + "score": 1.0, + "content": "by learning a binary mask instead of tuning the well-trained original parameters directly.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 15.5 + } + ], + "page_idx": 21, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 25, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 25, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "22", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 123, + 505, + 304 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 505, + 113 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 92 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 232, + 92 + ], + "score": 1.0, + "content": "Table 16: Fine-grained Results", + "type": "text" + }, + { + "bbox": [ + 232, + 80, + 248, + 91 + ], + "score": 0.78, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 248, + 80, + 505, + 92 + ], + "score": 1.0, + "content": "of DenseNet-101 on CIFAR-10. Comparison on different OOD", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 104, + 90, + 505, + 104 + ], + "spans": [ + { + "bbox": [ + 104, + 90, + 236, + 104 + ], + "score": 1.0, + "content": "benchmark datasets respectively.", + "type": "text" + }, + { + "bbox": [ + 237, + 92, + 244, + 103 + ], + "score": 0.74, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 245, + 90, + 394, + 104 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 394, + 92, + 401, + 102 + ], + "score": 0.56, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 402, + 90, + 505, + 104 + ], + "score": 1.0, + "content": "indicates lower values are", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 101, + 135, + 115 + ], + "spans": [ + { + "bbox": [ + 105, + 101, + 135, + 115 + ], + "score": 1.0, + "content": "better.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "table_body", + "bbox": [ + 108, + 123, + 505, + 304 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 123, + 505, + 304 + ], + "spans": [ + { + "bbox": [ + 108, + 123, + 505, + 304 + ], + "score": 0.983, + "html": "
ID datasetMethodOOD dataset TexturesPlaces365
CIFAR-100 FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
Energy+UMAP54.95 ± 2.6187.72 ±1.0533.59 ±1.3292.67 ±0.2332.80 ± 4.1493.57 ± 1.12
OE59.29 ± 1.3088.51 ±0.222.89 ± 0.3099.16 ±0.0511.14 ± 1.1197.50 ±0.19
Energy (w. Dauz)79.88 ± 2.4774.99 ± 2.404.27 ± 0.5798.80 ±0.1914.22 ± 3.9997.07 ±0.72
POEM82.30 ± 1.5772.74 ± 1.421.91 ± 0.4199.40 ±0.1011.24 ± 2.7096.67 ±0.48
OE+UM(ours)55.74 ± 1.4789.53 ± 0.181.42 ± 0.1599.49± 0.047.77 ±0.6998.15 ± 0.08
Energy (w.Daux)+ UM(ours)84.52 ± 0.0170.09 ± 0.478.30 ±0.8897.76 ±0.0520.27 ±1.3096.06 ±0.31
POEM+UM(ours) OE+UMAP(ours)84.87 ± 1.5668.97 ±0.394.73 ± 0.5298.88 ±0.1319.83 ± 0.3496.35 ±0.09
Energy(w.Daur)+UMAP(ours)59.05 ± 1.4189.14 ± 0.141.86 ± 0.0799.35 ±0.008.21±0.1298.07 ±0.03
POEM+UMAP(ours)75.18 ± 4.96 79.33 ± 4.1480.93 ± 4.49 76.89 ± 4.862.24 ± 1.34 2.10 ± 1.3799.25 ±0.29 99.34 ±0.309.30 ± 2.12 9.94 ± 6.9297.90 ±0.40 98.01 ± 1.11
OEMethodSUNLSUNiNaturalist
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
Energy+UMAP29.05 ±2.7894.41 ± 0.732.31± 0.8899.42 ± 0.0447.22 ± 14.0392.63 ± 2.14
8.38±0.7198.00 ±0.145.90 ± 1.4398.60± 0.215.09 ± 0.6498.76 ± 0.11
Energy (w. Dauz)10.30 ±3.8297.77 ± 0.6412.80 ± 4.6796.08 ± 1.386.93 ± 1.8698.40 ±0.32
POEM8.39±2.4298.16 ±0.439.69 ± 1.8997.25±0.583.78 ±0.9098.99 ±0.17
OE+UM(ours)5.51 ± 0.4498.55 ± 0.073.51 ± 0.4398.93 ± 0.09 92.11 ± 0.942.87 ± 0.4999.14 ± 0.09
Energy(w.Daux)+ UM(ours) POEM+UM(ours)16.13 ± 1.86 16.16 ± 0.5796.84±0.30 97.01±0.0823.27 ± 2.40 25.69 ± 0.1593.38 ± 0.2711.20 ± 2.35 9.30 ± 1.6097.54±0.42 98.05 ±0.24
OE+UMAP(ours)6.16 ±0.0298.49 ± 0.014.53 ± 0.1698.86±0.063.40 ± 0.7498.96 ±0.09
Energy(w.Dauz)+UMAP(ours)6.67 ± 1.5098.40±0.3223.50 ± 5.6194.78 ± 2.043.77 ± 2.1498.93 ±0.46
94.74 ±3.67
POEM+UMAP(ours)7.00±5.8098.46±0.9621.17 ± 12.843.63±2.7898.99 ±0.54
", + "type": "table", + "image_path": "51c49194022a3f450f05525beb0e80c2781701dee1b944e1944f81d7233e6b6e.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 108, + 123, + 505, + 183.33333333333334 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 108, + 183.33333333333334, + 505, + 243.66666666666669 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 108, + 243.66666666666669, + 505, + 304.0 + ], + "spans": [], + "index": 5 + } + ] + } + ], + "index": 2.5 + }, + { + "type": "table", + "bbox": [ + 109, + 375, + 505, + 606 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 108, + 331, + 504, + 365 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 331, + 505, + 344 + ], + "spans": [ + { + "bbox": [ + 106, + 331, + 230, + 344 + ], + "score": 1.0, + "content": "Table 17: Fine-grained Results", + "type": "text" + }, + { + "bbox": [ + 231, + 332, + 246, + 343 + ], + "score": 0.84, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 246, + 331, + 505, + 344 + ], + "score": 1.0, + "content": "of DenseNet-101 on CIFAR-100. Comparison on different OOD", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 343, + 506, + 355 + ], + "spans": [ + { + "bbox": [ + 106, + 343, + 236, + 355 + ], + "score": 1.0, + "content": "benchmark datasets respectively.", + "type": "text" + }, + { + "bbox": [ + 237, + 344, + 244, + 354 + ], + "score": 0.78, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 245, + 343, + 394, + 355 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 394, + 343, + 401, + 354 + ], + "score": 0.64, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 402, + 343, + 506, + 355 + ], + "score": 1.0, + "content": "indicates lower values are", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 353, + 135, + 366 + ], + "spans": [ + { + "bbox": [ + 105, + 353, + 135, + 366 + ], + "score": 1.0, + "content": "better.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "table_body", + "bbox": [ + 109, + 375, + 505, + 606 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 109, + 375, + 505, + 606 + ], + "spans": [ + { + "bbox": [ + 109, + 375, + 505, + 606 + ], + "score": 0.982, + "html": "
ID datasetMethodOOD datasetPlaces365
CIFAR-10 FPR95↓AUROC↑Textures FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-100MSP ODIN Mahalanobis Energy83.53± 0.33 85.29 ± 0.1775.11 ± 0.27 73.31 ± 0.24 49.60 ± 1.5186.90 ±0.18 86.45 ± 1.27 33.06± 3.7671.45 ± 0.40 71.91 ± 0.2785.83 ±0.48 84.35 ± 0.6470.54 ± 0.42 73.58 ± 0.51
98.25 ±0.0590.19 ± 1.2195.20 ±0.4953.69 ± 1.55
Energy+UM(ours)82.16 ± 0.5975.31 ± 0.2190.20 ±0.3068.98 ±0.34 65.39 ± 1.4482.39 ±0.97 77.30 ± 2.0873.78 ±0.66 76.06 ± 1.05
OE89.62 ±0.0766.12 ± 0.9386.99 ± 1.22
Energy (w.Daux) POEM90.97 ±0.46 96.14 ±0.0669.02 ± 0.39 57.52 ± 0.8814.36 ± 0.25 9.02 ±0.0695.92 ±0.12 97.16 ±0.2640.19 ± 6.97 35.18 ± 4.7390.70 ± 2.01 93.29 ± 1.14
OE+UM (ours)96.19 ±0.1655.82 ± 1.057.63 ± 1.4097.69 ±0.0632.67 ± 3.7393.94±0.68
89.61±0.0871.24 ± 0.0816.78 ± 0.2595.60 ±0.0839.77 ± 0.3491.07 ± 0.13
Energy(w.Daux)+UM(ours)95.38 ± 0.4563.41 ± 0.146.41 ± 0.8397.77 ± 0.3330.96 ± 3.6192.85 ±0.69
POEM+UM(ours)95.78 ± 0.1460.23 ± 0.705.17 ±0.1898.53± 0.0323.90 ±0.8495.45 ± 0.11
OE+UMAP(ours) Energy(w.Daux)+UMAP(ours)90.72±0.3569.76 ± 0.2515.32 ± 0.2395.72 ± 0.0136.42 ± 1.9192.08 ±0.49
POEM+UMAP(ours)95.39±0.10 95.69 ± 0.1763.26 ±0.18 61.62 ± 0.246.52 ±0.44 5.23 ± 0.5897.83 ±0.18 98.52 ± 0.0131.18 ± 0.43 26.06 ± 1.1693.13±0.41 94.91 ± 0.25
MethodSUNLSUN
FPR95↓ AUROC↑FPR95↓AUROC↑iNaturalist FPR95↓AUROC↑
MSP88.75 ±0.2366.75 ± 0.2567.83 ±1.3782.94± 0.3285.00 ± 0.7376.62 ± 0.25
ODIN88.49 ± 0.9969.64± 0.61 54.37 ± 1.3534.80 ± 2.55 89.31 ± 4.8393.92 ±0.75 43.19 ± 16.3681.67 ± 2.77 93.63 ± 1.1978.36 ± 1.57
Mahalanobis95.53±0.3749.60 ± 1.51
Energy Energy+UM(ours)97.17 ± 0.92 69.04 ± 0.8335.09 ± 3.1793.49 ±0.8785.70 ± 2.1475.82 ± 1.72
81.96 ± 2.2671.47 ± 1.8822.54± 5.9394.98 ± 1.7574.28 ± 3.7280.72 ± 3.75
OE44.47 ± 9.10 90.70 ± 2.015.75 ± 1.1898.57 ± 0.1325.51 ± 4.1294.46 ± 0.88
Energy (w. Daux)32.69 ± 5.6993.63 ± 1.4855.75 ± 4.3187.96 ± 1.0317.34 ± 4.5496.50 ±0.81
POEM30.45 ± 5.1194.26 ± 0.9046.68 ± 3.5990.30 ± 2.1716.50 ± 2.0996.63±0.23
OE+UM(ours)44.23±0.20 90.28±0.035.80 ±0.3398.63 ±0.0326.72 ± 1.9594.51 ± 0.43
93.18 ± 0.6991.98 ± 0.8310.83 ± 2.0697.09 ±0.69
Energy(w.Daux)+UM(ours) POEM+UM(ours)28.98 ±3.1537.56 ± 4.8194.43 ± 1.18
21.34 ± 1.0795.76 ±0.2233.74 ±6.2298.64 ± 0.128.85±0.2397.93 ± 0.03 95.08 ±0.22
OE+UMAP(ours) Energy(w.Daux)+UMAP(ours)39.58 ± 2.0291.37 ± 0.52 93.38 ± 0.195.77 ± 0.7192.08 ±0.3923.33 ± 1.24 13.96 ± 2.4896.87 ± 0.15
29.65 ± 1.0635.94 ± 0.7593.57 ±0.88
POEM+UMAP(ours)23.73 ± 0.7195.25 ± 0.0633.09 ± 5.949.76 ± 1.0997.77 ± 0.24
", + "type": "table", + "image_path": "bbcdb0726730fb08f597974785e84bf8ec7f9af7e92c5725afbf29197e39ed92.jpg" + } + ] + } + ], + "index": 10, + "virtual_lines": [ + { + "bbox": [ + 109, + 375, + 505, + 452.0 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 109, + 452.0, + 505, + 529.0 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 109, + 529.0, + 505, + 606.0 + ], + "spans": [], + "index": 11 + } + ] + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 107, + 643, + 505, + 732 + ], + "lines": [], + "index": 15.5, + "bbox_fs": [ + 105, + 644, + 507, + 734 + ], + "lines_deleted": true + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 131, + 505, + 266 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 506, + 124 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 104, + 79, + 506, + 92 + ], + "spans": [ + { + "bbox": [ + 104, + 79, + 180, + 92 + ], + "score": 1.0, + "content": "Table 18: Results", + "type": "text" + }, + { + "bbox": [ + 180, + 80, + 195, + 91 + ], + "score": 0.82, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 79, + 506, + 92 + ], + "score": 1.0, + "content": "of WRN-40-4. Comparison with competitive OOD detection baselines. We", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 91, + 507, + 104 + ], + "spans": [ + { + "bbox": [ + 105, + 91, + 507, + 104 + ], + "score": 1.0, + "content": "respectively train WRN-40-4 on CIFAR-10 and CIFAR-100. For those methods involving outliers,", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 102, + 506, + 115 + ], + "spans": [ + { + "bbox": [ + 105, + 102, + 293, + 115 + ], + "score": 1.0, + "content": "we retrieve 5000 samples from ImageNet-1k.", + "type": "text" + }, + { + "bbox": [ + 294, + 103, + 302, + 114 + ], + "score": 0.38, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 302, + 102, + 458, + 115 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 458, + 103, + 465, + 114 + ], + "score": 0.75, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 102, + 506, + 115 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 113, + 201, + 125 + ], + "spans": [ + { + "bbox": [ + 105, + 113, + 201, + 125 + ], + "score": 1.0, + "content": "lower values are better.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 1.5 + }, + { + "type": "table_body", + "bbox": [ + 108, + 131, + 505, + 266 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 131, + 505, + 266 + ], + "spans": [ + { + "bbox": [ + 108, + 131, + 505, + 266 + ], + "score": 0.985, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10MSP(Hendrycks & Gimpel,2017a)87.12 ± 0.2587.84± 0.3068.29 ±0.9693.86 ± 0.19
ODIN(Liang et al.,2018)83.29 ±0.7282.74±0.7965.68 ± 0.7793.86 ± 0.19
Mahalanobis(Lee et al., 2018b)77.57 ± 0.2876.11 ±0.1061.18 ±0.1093.86 ± 0.19
Energy(Liu et al.,2020)87.69 ± 0.5488.16 ±0.6958.47 ± 1.9493.86 ± 0.19
Energy+UM(ours)91.74 ± 0.4392.67 ± 0.5240.40±1.3292.68±0.23
Energy+UMAP(ours)88.84± 1.0289.31 ± 1.4450.23 ±2.2593.86 ± 0.19
CIFAR-100MSP(Hendrycks & Gimpel,2017a)72.34± 0.6372.69 ± 0.4485.40 ±0.5975.01 ±0.07
ODIN(Liang et al., 2018)68.78±0.6766.92 ± 0.7285.28 ± 0.6475.01 ± 0.07
Mahalanobis(Lee et al., 2018b)68.20±0.9968.30 ± 1.1576.46 ± 2.0275.01 ± 0.07
Energy(Liu et al.,2020)74.00 ± 0.4173.02 ± 0.4781.37 ±0.0875.01 ± 0.07
Energy+UM(ours)76.07 ± 0.0476.94 ± 0.0674.29 ± 1.6659.08 ± 2.75
Energy+UMAP(ours)77.35 ± 0.7877.43 ± 0.9168.20 ± 0.0675.01 ±0.07
", + "type": "table", + "image_path": "4bcf6ecf1f82bd834d5a6547f5e9543455c4aec84f74493ab4a1e5067ca38c27.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 108, + 131, + 505, + 176.0 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 108, + 176.0, + 505, + 221.0 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 108, + 221.0, + 505, + 266.0 + ], + "spans": [], + "index": 6 + } + ] + } + ], + "index": 3.25 + }, + { + "type": "table", + "bbox": [ + 108, + 312, + 505, + 455 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 105, + 279, + 504, + 303 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 280, + 505, + 291 + ], + "spans": [ + { + "bbox": [ + 106, + 280, + 236, + 291 + ], + "score": 1.0, + "content": "Table 19: Fine-grained Results", + "type": "text" + }, + { + "bbox": [ + 236, + 280, + 252, + 291 + ], + "score": 0.81, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 253, + 280, + 505, + 291 + ], + "score": 1.0, + "content": "of WRN-40-4 on CIFAR-10. Comparison on different OOD", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 290, + 490, + 304 + ], + "spans": [ + { + "bbox": [ + 105, + 290, + 190, + 304 + ], + "score": 1.0, + "content": "benchmark datasets.", + "type": "text" + }, + { + "bbox": [ + 190, + 291, + 197, + 302 + ], + "score": 0.75, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 198, + 290, + 349, + 304 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 350, + 291, + 357, + 302 + ], + "score": 0.31, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 358, + 290, + 490, + 304 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "table_body", + "bbox": [ + 108, + 312, + 505, + 455 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 108, + 312, + 505, + 455 + ], + "spans": [ + { + "bbox": [ + 108, + 312, + 505, + 455 + ], + "score": 0.984, + "html": "
ID datasetMethodOOD dataset
CIFAR-100TexturesPlaces365
MSPFPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-10ODIN70.96 ±0.7086.08 ±0.0868.81 ± 1.2986.53 ±0.8368.31 ± 0.2586.71 ±0.13
64.97 ±0.0883.36 ±0.1166.86 ±2.2481.34 ± 0.8166.49 ±1.1683.47 ±0.93
Mahalanobis79.84±0.5570.33 ± 0.2422.56 ± 0.0894.07 ± 0.0485.09 ± 0.5967.90 ± 0.37
Energy61.09±0.5886.66 ±0.0464.29 ± 1.7285.56 ±0.5355.32 ± 0.1388.29 ±0.26
Energy+UM(ours)57.21 ± 1.4187.56± 0.1546.49 ± 1.0389.74±0.4540.68±4.4692.51± 0.97 90.43± 0.47
Energy+UMAP(ours)65.45 ± 1.1084.65 ±0.9559.14 ± 1.6485.27 ± 1.7448.16 ± 1.89
MethodSUNLSUNiNaturalist
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
MSP68.62 ±0.5086.95 ±0.2352.97± 3.0792.41 ±0.1876.05 ± 0.0183.44±0.36
ODIN65.47 ±0.7883.79 ± 1.1031.89 ± 3.4494.34 ±0.8979.28 ±0.1879.80 ±0.35
Mahalanobis82.92 ±0.2870.52 ± 0.4764.31 ± 0.5767.75 ±0.5581.50 ± 2.9174.97 ± 2.91
Energy54.88 ±0.1888.67±0.3024.99 ± 1.3895.98± 0.3775.89 ± 0.8582.40±0.22
Energy+UM(ours)38.92 ± 3.4692.98 ± 0.958.38 ±0.7798.18 ± 0.1666.02 ± 6.7085.22 ±3.42
Energy+UMAP(ours)45.94 ± 2.6491.27±0.5614.10 ± 0.0497.46 ± 0.1474.69±0.1581.13 ± 1.12
", + "type": "table", + "image_path": "5ce55c54c2423aee8aef57faccdc3d8b2906c7115943d3777e81a97a06580fd9.jpg" + } + ] + } + ], + "index": 10, + "virtual_lines": [ + { + "bbox": [ + 108, + 312, + 505, + 359.6666666666667 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 108, + 359.6666666666667, + 505, + 407.33333333333337 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 108, + 407.33333333333337, + 505, + 455.00000000000006 + ], + "spans": [], + "index": 11 + } + ] + } + ], + "index": 8.75 + }, + { + "type": "text", + "bbox": [ + 106, + 479, + 505, + 557 + ], + "lines": [ + { + "bbox": [ + 106, + 479, + 506, + 492 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 506, + 492 + ], + "score": 1.0, + "content": "Experiment on different model structure. Following 4.2, we additionally conduct critical experi-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 490, + 506, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 490, + 506, + 504 + ], + "score": 1.0, + "content": "ments on the WRN-40-4 (Lin et al., 2021) backbone to demonstrate the effectiveness of the proposed", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 501, + 505, + 514 + ], + "spans": [ + { + "bbox": [ + 106, + 501, + 505, + 514 + ], + "score": 1.0, + "content": "UM and UMAP. In Figure 7, we can find during the model training phase on ID data, there also exists", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 513, + 506, + 525 + ], + "spans": [ + { + "bbox": [ + 106, + 513, + 506, + 525 + ], + "score": 1.0, + "content": "the once-covered OOD detection capability can be explored in later development. In Table 18, we", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 524, + 506, + 535 + ], + "spans": [ + { + "bbox": [ + 106, + 524, + 506, + 535 + ], + "score": 1.0, + "content": "show the comparison of multiple OOD detection baselines, evaluating the OOD performance on the 7", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 534, + 506, + 547 + ], + "spans": [ + { + "bbox": [ + 106, + 534, + 506, + 547 + ], + "score": 1.0, + "content": "OOD datasets mentioned in Section 4.1. The results again demonstrate that our proposed method", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 544, + 433, + 559 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 433, + 559 + ], + "score": 1.0, + "content": "indeed excavates the intrinsic detection capability and improves the performance.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 107, + 562, + 505, + 628 + ], + "lines": [ + { + "bbox": [ + 105, + 561, + 505, + 575 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 505, + 575 + ], + "score": 1.0, + "content": "As for the fine-grained results of WRN-40-4, we report results on 6 OOD datasets respectively. When", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 573, + 505, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 573, + 505, + 586 + ], + "score": 1.0, + "content": "trained on CIFAR-10, UM can outstrip all the scoring function baselines on 5 OOD datasets except", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 584, + 505, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 584, + 505, + 597 + ], + "score": 1.0, + "content": "Textures on which Mahalanobis performs better while UMAP still has excellent OOD performance", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 596, + 505, + 607 + ], + "spans": [ + { + "bbox": [ + 106, + 596, + 505, + 607 + ], + "score": 1.0, + "content": "ranking only second to UM. When trained on CIFAR-100, UM and UMAP can also outperform the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 607, + 505, + 618 + ], + "spans": [ + { + "bbox": [ + 106, + 607, + 505, + 618 + ], + "score": 1.0, + "content": "baselines on most OOD datasets. The fine-grained results of WRN-40-4 further demonstrate the", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 618, + 366, + 629 + ], + "spans": [ + { + "bbox": [ + 106, + 618, + 366, + 629 + ], + "score": 1.0, + "content": "effectiveness of the proposed UM/UMAP on other architectures.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21.5 + }, + { + "type": "text", + "bbox": [ + 107, + 643, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 642, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 506, + 657 + ], + "score": 1.0, + "content": "Additional results about the general existence of once-covered OOD detection capability. In", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 655, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 505, + 667 + ], + "score": 1.0, + "content": "Section 4.3, we display the once-covered OOD detection capability on CIFAR-10 using SVHN as", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "score": 1.0, + "content": "the OOD dataset. Here, we additionally verify the previously observed trend during training when", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "training DenseNet-101 on CIFAR-100 using iNaturalist as an OOD dataset. In Figure 5, we trace", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 687, + 507, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 507, + 701 + ], + "score": 1.0, + "content": "the three evaluation metrics during training on CIFAR-100 using 4 different learning rate schedules.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "Consistent with the original experiment, we still use iNaturalist as the OOD dataset. It can be seen", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 711, + 505, + 721 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 505, + 721 + ], + "score": 1.0, + "content": "for all the three metrics that exists a middle stage where the model has the better OOD detection", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "capability (For FPR95, it is smaller (better) in the middle stage; for AUROC and AUPR, they are", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 28.5 + } + ], + "page_idx": 22, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 25, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 25, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 131, + 505, + 266 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 506, + 124 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 104, + 79, + 506, + 92 + ], + "spans": [ + { + "bbox": [ + 104, + 79, + 180, + 92 + ], + "score": 1.0, + "content": "Table 18: Results", + "type": "text" + }, + { + "bbox": [ + 180, + 80, + 195, + 91 + ], + "score": 0.82, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 79, + 506, + 92 + ], + "score": 1.0, + "content": "of WRN-40-4. Comparison with competitive OOD detection baselines. We", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 91, + 507, + 104 + ], + "spans": [ + { + "bbox": [ + 105, + 91, + 507, + 104 + ], + "score": 1.0, + "content": "respectively train WRN-40-4 on CIFAR-10 and CIFAR-100. For those methods involving outliers,", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 102, + 506, + 115 + ], + "spans": [ + { + "bbox": [ + 105, + 102, + 293, + 115 + ], + "score": 1.0, + "content": "we retrieve 5000 samples from ImageNet-1k.", + "type": "text" + }, + { + "bbox": [ + 294, + 103, + 302, + 114 + ], + "score": 0.38, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 302, + 102, + 458, + 115 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 458, + 103, + 465, + 114 + ], + "score": 0.75, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 102, + 506, + 115 + ], + "score": 1.0, + "content": "indicates", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 113, + 201, + 125 + ], + "spans": [ + { + "bbox": [ + 105, + 113, + 201, + 125 + ], + "score": 1.0, + "content": "lower values are better.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 1.5 + }, + { + "type": "table_body", + "bbox": [ + 108, + 131, + 505, + 266 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 131, + 505, + 266 + ], + "spans": [ + { + "bbox": [ + 108, + 131, + 505, + 266 + ], + "score": 0.985, + "html": "
DinMethodAUROC↑AUPR↑FPR95↓ID-ACC↑
CIFAR-10MSP(Hendrycks & Gimpel,2017a)87.12 ± 0.2587.84± 0.3068.29 ±0.9693.86 ± 0.19
ODIN(Liang et al.,2018)83.29 ±0.7282.74±0.7965.68 ± 0.7793.86 ± 0.19
Mahalanobis(Lee et al., 2018b)77.57 ± 0.2876.11 ±0.1061.18 ±0.1093.86 ± 0.19
Energy(Liu et al.,2020)87.69 ± 0.5488.16 ±0.6958.47 ± 1.9493.86 ± 0.19
Energy+UM(ours)91.74 ± 0.4392.67 ± 0.5240.40±1.3292.68±0.23
Energy+UMAP(ours)88.84± 1.0289.31 ± 1.4450.23 ±2.2593.86 ± 0.19
CIFAR-100MSP(Hendrycks & Gimpel,2017a)72.34± 0.6372.69 ± 0.4485.40 ±0.5975.01 ±0.07
ODIN(Liang et al., 2018)68.78±0.6766.92 ± 0.7285.28 ± 0.6475.01 ± 0.07
Mahalanobis(Lee et al., 2018b)68.20±0.9968.30 ± 1.1576.46 ± 2.0275.01 ± 0.07
Energy(Liu et al.,2020)74.00 ± 0.4173.02 ± 0.4781.37 ±0.0875.01 ± 0.07
Energy+UM(ours)76.07 ± 0.0476.94 ± 0.0674.29 ± 1.6659.08 ± 2.75
Energy+UMAP(ours)77.35 ± 0.7877.43 ± 0.9168.20 ± 0.0675.01 ±0.07
", + "type": "table", + "image_path": "4bcf6ecf1f82bd834d5a6547f5e9543455c4aec84f74493ab4a1e5067ca38c27.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 108, + 131, + 505, + 176.0 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 108, + 176.0, + 505, + 221.0 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 108, + 221.0, + 505, + 266.0 + ], + "spans": [], + "index": 6 + } + ] + } + ], + "index": 3.25 + }, + { + "type": "table", + "bbox": [ + 108, + 312, + 505, + 455 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 105, + 279, + 504, + 303 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 280, + 505, + 291 + ], + "spans": [ + { + "bbox": [ + 106, + 280, + 236, + 291 + ], + "score": 1.0, + "content": "Table 19: Fine-grained Results", + "type": "text" + }, + { + "bbox": [ + 236, + 280, + 252, + 291 + ], + "score": 0.81, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 253, + 280, + 505, + 291 + ], + "score": 1.0, + "content": "of WRN-40-4 on CIFAR-10. Comparison on different OOD", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 290, + 490, + 304 + ], + "spans": [ + { + "bbox": [ + 105, + 290, + 190, + 304 + ], + "score": 1.0, + "content": "benchmark datasets.", + "type": "text" + }, + { + "bbox": [ + 190, + 291, + 197, + 302 + ], + "score": 0.75, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 198, + 290, + 349, + 304 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 350, + 291, + 357, + 302 + ], + "score": 0.31, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 358, + 290, + 490, + 304 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "table_body", + "bbox": [ + 108, + 312, + 505, + 455 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 108, + 312, + 505, + 455 + ], + "spans": [ + { + "bbox": [ + 108, + 312, + 505, + 455 + ], + "score": 0.984, + "html": "
ID datasetMethodOOD dataset
CIFAR-100TexturesPlaces365
MSPFPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-10ODIN70.96 ±0.7086.08 ±0.0868.81 ± 1.2986.53 ±0.8368.31 ± 0.2586.71 ±0.13
64.97 ±0.0883.36 ±0.1166.86 ±2.2481.34 ± 0.8166.49 ±1.1683.47 ±0.93
Mahalanobis79.84±0.5570.33 ± 0.2422.56 ± 0.0894.07 ± 0.0485.09 ± 0.5967.90 ± 0.37
Energy61.09±0.5886.66 ±0.0464.29 ± 1.7285.56 ±0.5355.32 ± 0.1388.29 ±0.26
Energy+UM(ours)57.21 ± 1.4187.56± 0.1546.49 ± 1.0389.74±0.4540.68±4.4692.51± 0.97 90.43± 0.47
Energy+UMAP(ours)65.45 ± 1.1084.65 ±0.9559.14 ± 1.6485.27 ± 1.7448.16 ± 1.89
MethodSUNLSUNiNaturalist
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
MSP68.62 ±0.5086.95 ±0.2352.97± 3.0792.41 ±0.1876.05 ± 0.0183.44±0.36
ODIN65.47 ±0.7883.79 ± 1.1031.89 ± 3.4494.34 ±0.8979.28 ±0.1879.80 ±0.35
Mahalanobis82.92 ±0.2870.52 ± 0.4764.31 ± 0.5767.75 ±0.5581.50 ± 2.9174.97 ± 2.91
Energy54.88 ±0.1888.67±0.3024.99 ± 1.3895.98± 0.3775.89 ± 0.8582.40±0.22
Energy+UM(ours)38.92 ± 3.4692.98 ± 0.958.38 ±0.7798.18 ± 0.1666.02 ± 6.7085.22 ±3.42
Energy+UMAP(ours)45.94 ± 2.6491.27±0.5614.10 ± 0.0497.46 ± 0.1474.69±0.1581.13 ± 1.12
", + "type": "table", + "image_path": "5ce55c54c2423aee8aef57faccdc3d8b2906c7115943d3777e81a97a06580fd9.jpg" + } + ] + } + ], + "index": 10, + "virtual_lines": [ + { + "bbox": [ + 108, + 312, + 505, + 359.6666666666667 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 108, + 359.6666666666667, + 505, + 407.33333333333337 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 108, + 407.33333333333337, + 505, + 455.00000000000006 + ], + "spans": [], + "index": 11 + } + ] + } + ], + "index": 8.75 + }, + { + "type": "text", + "bbox": [ + 106, + 479, + 505, + 557 + ], + "lines": [ + { + "bbox": [ + 106, + 479, + 506, + 492 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 506, + 492 + ], + "score": 1.0, + "content": "Experiment on different model structure. Following 4.2, we additionally conduct critical experi-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 490, + 506, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 490, + 506, + 504 + ], + "score": 1.0, + "content": "ments on the WRN-40-4 (Lin et al., 2021) backbone to demonstrate the effectiveness of the proposed", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 501, + 505, + 514 + ], + "spans": [ + { + "bbox": [ + 106, + 501, + 505, + 514 + ], + "score": 1.0, + "content": "UM and UMAP. In Figure 7, we can find during the model training phase on ID data, there also exists", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 513, + 506, + 525 + ], + "spans": [ + { + "bbox": [ + 106, + 513, + 506, + 525 + ], + "score": 1.0, + "content": "the once-covered OOD detection capability can be explored in later development. In Table 18, we", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 524, + 506, + 535 + ], + "spans": [ + { + "bbox": [ + 106, + 524, + 506, + 535 + ], + "score": 1.0, + "content": "show the comparison of multiple OOD detection baselines, evaluating the OOD performance on the 7", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 534, + 506, + 547 + ], + "spans": [ + { + "bbox": [ + 106, + 534, + 506, + 547 + ], + "score": 1.0, + "content": "OOD datasets mentioned in Section 4.1. The results again demonstrate that our proposed method", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 544, + 433, + 559 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 433, + 559 + ], + "score": 1.0, + "content": "indeed excavates the intrinsic detection capability and improves the performance.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 15, + "bbox_fs": [ + 105, + 479, + 506, + 559 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 562, + 505, + 628 + ], + "lines": [ + { + "bbox": [ + 105, + 561, + 505, + 575 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 505, + 575 + ], + "score": 1.0, + "content": "As for the fine-grained results of WRN-40-4, we report results on 6 OOD datasets respectively. When", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 573, + 505, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 573, + 505, + 586 + ], + "score": 1.0, + "content": "trained on CIFAR-10, UM can outstrip all the scoring function baselines on 5 OOD datasets except", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 584, + 505, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 584, + 505, + 597 + ], + "score": 1.0, + "content": "Textures on which Mahalanobis performs better while UMAP still has excellent OOD performance", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 596, + 505, + 607 + ], + "spans": [ + { + "bbox": [ + 106, + 596, + 505, + 607 + ], + "score": 1.0, + "content": "ranking only second to UM. When trained on CIFAR-100, UM and UMAP can also outperform the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 607, + 505, + 618 + ], + "spans": [ + { + "bbox": [ + 106, + 607, + 505, + 618 + ], + "score": 1.0, + "content": "baselines on most OOD datasets. The fine-grained results of WRN-40-4 further demonstrate the", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 618, + 366, + 629 + ], + "spans": [ + { + "bbox": [ + 106, + 618, + 366, + 629 + ], + "score": 1.0, + "content": "effectiveness of the proposed UM/UMAP on other architectures.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21.5, + "bbox_fs": [ + 105, + 561, + 505, + 629 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 643, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 642, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 506, + 657 + ], + "score": 1.0, + "content": "Additional results about the general existence of once-covered OOD detection capability. In", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 655, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 505, + 667 + ], + "score": 1.0, + "content": "Section 4.3, we display the once-covered OOD detection capability on CIFAR-10 using SVHN as", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "score": 1.0, + "content": "the OOD dataset. Here, we additionally verify the previously observed trend during training when", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "training DenseNet-101 on CIFAR-100 using iNaturalist as an OOD dataset. In Figure 5, we trace", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 687, + 507, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 507, + 701 + ], + "score": 1.0, + "content": "the three evaluation metrics during training on CIFAR-100 using 4 different learning rate schedules.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "Consistent with the original experiment, we still use iNaturalist as the OOD dataset. It can be seen", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 711, + 505, + 721 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 505, + 721 + ], + "score": 1.0, + "content": "for all the three metrics that exists a middle stage where the model has the better OOD detection", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "capability (For FPR95, it is smaller (better) in the middle stage; for AUROC and AUPR, they are", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 496, + 505, + 510 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 505, + 510 + ], + "score": 1.0, + "content": "higher (better) in the middle stage). Besides that, we also look into the change of OOD performance", + "type": "text", + "cross_page": true + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 507, + 506, + 522 + ], + "spans": [ + { + "bbox": [ + 105, + 507, + 506, + 522 + ], + "score": 1.0, + "content": "on other architecture (e.g., WRN-40-4) in Figure 6 and Figure 7. In Figure 6, we display the curves", + "type": "text", + "cross_page": true + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 518, + 506, + 532 + ], + "spans": [ + { + "bbox": [ + 105, + 518, + 506, + 532 + ], + "score": 1.0, + "content": "of three metrics of WRN-40-4 when trained on CIFAR-10 with SVHN and Textures as OOD datasets.", + "type": "text", + "cross_page": true + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 530, + 506, + 543 + ], + "spans": [ + { + "bbox": [ + 106, + 530, + 506, + 543 + ], + "score": 1.0, + "content": "The trend that the OOD performance first goes better and then converges to worse OOD performance", + "type": "text", + "cross_page": true + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 540, + 505, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 540, + 505, + 554 + ], + "score": 1.0, + "content": "can be reflected. In Figure 7, we continually provide curves of the three metrics of WRN-40-4 during", + "type": "text", + "cross_page": true + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 552, + 505, + 564 + ], + "spans": [ + { + "bbox": [ + 105, + 552, + 505, + 564 + ], + "score": 1.0, + "content": "training on CIFAR-100 with iNaturalist, Places365, and SUN as OOD datasets. A clear better middle", + "type": "text", + "cross_page": true + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 564, + 281, + 575 + ], + "spans": [ + { + "bbox": [ + 106, + 564, + 281, + 575 + ], + "score": 1.0, + "content": "stage can still be excavated in this scenario.", + "type": "text", + "cross_page": true + } + ], + "index": 19 + } + ], + "index": 28.5, + "bbox_fs": [ + 105, + 642, + 507, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 112, + 505, + 253 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 504, + 103 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 79, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 79, + 234, + 93 + ], + "score": 1.0, + "content": "Table 20: Fine-grained Results", + "type": "text" + }, + { + "bbox": [ + 235, + 80, + 250, + 91 + ], + "score": 0.81, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 251, + 79, + 505, + 93 + ], + "score": 1.0, + "content": "of WRN-40-4 on CIFAR-100. Comparison on different OOD", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 90, + 490, + 104 + ], + "spans": [ + { + "bbox": [ + 105, + 90, + 189, + 104 + ], + "score": 1.0, + "content": "benchmark datasets.", + "type": "text" + }, + { + "bbox": [ + 190, + 92, + 197, + 103 + ], + "score": 0.73, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 198, + 90, + 349, + 104 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 350, + 92, + 357, + 102 + ], + "score": 0.26, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 358, + 90, + 490, + 104 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 108, + 112, + 505, + 253 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 112, + 505, + 253 + ], + "spans": [ + { + "bbox": [ + 108, + 112, + 505, + 253 + ], + "score": 0.975, + "html": "
ID datasetMethodOOD dataset
CIFAR-10TexturesPlaces365
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-100MSP83.83 ±0.2975.50±0.2186.15 ± 0.2372.36 ± 0.4086.72 ±0.2969.60± 0.27
ODIN83.70 ±0.3074.32 ±0.0381.57 ± 1.7471.67±0.2888.07 ± 0.1164.83 ± 1.36
Mahalanobis96.89 ± 0.1168.78 ±0.6731.02 ± 2.0491.85 ± 0.9193.34 ± 0.4461.28 ± 1.62
Energy81.32 ± 0.4777.49 ± 0.2686.38 ±0.4973.50±0.4584.45 ± 0.3869.82 ± 0.63
Energy+UM(ours)89.23 ± 1.5163.85 ± 1.7378.90 ±0.0772.58 ± 1.3380.46 ± 1.9970.49 ± 1.01
Energy+UMAP(ours)94.11 ± 0.7260.77 ± 0.9666.94 ± 4.4975.82 ± 5.3382.59±0.2771.92 ± 3.58
MethodSUNLSUNiNaturalist
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
MSP88.88±0.8365.22 ± 0.8578.56±0.6679.10 ± 0.5386.72 ± 0.2973,75 ±0.56
ODIN91.00 ±0.1059.06 ± 1.8170.14 ± 2.4284.03±0.8687.86±1.1364.52 ± 1.51
Mahalanobis94.22 ± 0.0160.09 ± 1.5689.73± 2.8740.81± 3.0787.25±3.2874.98 ± 2.85
Energy88.35±0.5264.04 ±0.7659.84 ±0.0687.91 ± 0.5388.91±0.7867.81 ± 0.91
Energy+UM(ours)84.04 ±0.0967.19 ± 0.1433.87 ± 1.2192.29 ± 0.0276.91 ± 6.0779.28 ± 4.17
Energy+UMAP(ours)80.53 ± 1.3172.68± 4.5027.79 ± 2.1993.39 ± 0.5755.53± 4.6185.65± 0.83
", + "type": "table", + "image_path": "db8ab563d2f7b2f21e22411a59359b785cf8ebd0be133a80a03550c6303ee097.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 108, + 112, + 505, + 159.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 108, + 159.0, + 505, + 206.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 108, + 206.0, + 505, + 253.0 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "image", + "bbox": [ + 118, + 273, + 493, + 401 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 118, + 273, + 493, + 401 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 118, + 273, + 493, + 401 + ], + "spans": [ + { + "bbox": [ + 118, + 273, + 493, + 401 + ], + "score": 0.971, + "type": "image", + "image_path": "1674497873272ebcef525d5b25ffb65e99d34866ce9d3e4d699f013f99848ca6.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 118, + 273, + 493, + 315.6666666666667 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 118, + 315.6666666666667, + 493, + 358.33333333333337 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 118, + 358.33333333333337, + 493, + 401.00000000000006 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 418, + 505, + 474 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 419, + 505, + 430 + ], + "spans": [ + { + "bbox": [ + 106, + 419, + 505, + 430 + ], + "score": 1.0, + "content": "Figure 5: Ablation studies on three metrics with 4 different learning rate schedules. The model is", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 429, + 505, + 442 + ], + "spans": [ + { + "bbox": [ + 105, + 429, + 505, + 442 + ], + "score": 1.0, + "content": "DenseNet-101 trained on CIFAR-100 with iNaturalist as the OOD dataset. (a) change of FPR95", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 440, + 505, + 453 + ], + "spans": [ + { + "bbox": [ + 106, + 440, + 505, + 453 + ], + "score": 1.0, + "content": "throughout the pruning phase when training on CIFAR-100; (b) change of AUROC throughout the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 451, + 505, + 465 + ], + "spans": [ + { + "bbox": [ + 105, + 451, + 505, + 465 + ], + "score": 1.0, + "content": "pruning phase when training on CIFAR-100; (c) change of AUPR throughout the pruning phase when", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 462, + 502, + 475 + ], + "spans": [ + { + "bbox": [ + 106, + 462, + 502, + 475 + ], + "score": 1.0, + "content": "training on CIFAR-100. It demonstrates a better middle stage exists according to the three metrics.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 10 + } + ], + "index": 8.0 + }, + { + "type": "text", + "bbox": [ + 106, + 496, + 505, + 574 + ], + "lines": [ + { + "bbox": [ + 105, + 496, + 505, + 510 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 505, + 510 + ], + "score": 1.0, + "content": "higher (better) in the middle stage). Besides that, we also look into the change of OOD performance", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 507, + 506, + 522 + ], + "spans": [ + { + "bbox": [ + 105, + 507, + 506, + 522 + ], + "score": 1.0, + "content": "on other architecture (e.g., WRN-40-4) in Figure 6 and Figure 7. In Figure 6, we display the curves", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 518, + 506, + 532 + ], + "spans": [ + { + "bbox": [ + 105, + 518, + 506, + 532 + ], + "score": 1.0, + "content": "of three metrics of WRN-40-4 when trained on CIFAR-10 with SVHN and Textures as OOD datasets.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 530, + 506, + 543 + ], + "spans": [ + { + "bbox": [ + 106, + 530, + 506, + 543 + ], + "score": 1.0, + "content": "The trend that the OOD performance first goes better and then converges to worse OOD performance", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 540, + 505, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 540, + 505, + 554 + ], + "score": 1.0, + "content": "can be reflected. In Figure 7, we continually provide curves of the three metrics of WRN-40-4 during", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 552, + 505, + 564 + ], + "spans": [ + { + "bbox": [ + 105, + 552, + 505, + 564 + ], + "score": 1.0, + "content": "training on CIFAR-100 with iNaturalist, Places365, and SUN as OOD datasets. A clear better middle", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 564, + 281, + 575 + ], + "spans": [ + { + "bbox": [ + 106, + 564, + 281, + 575 + ], + "score": 1.0, + "content": "stage can still be excavated in this scenario.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 106, + 589, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 588, + 506, + 602 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 506, + 602 + ], + "score": 1.0, + "content": "UMAP: adopting pruning on UM. We conduct various experiments to see whether pruning has", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 600, + 506, + 613 + ], + "spans": [ + { + "bbox": [ + 106, + 600, + 506, + 613 + ], + "score": 1.0, + "content": "an impact on Unleashing Mask itself. To be specific, we expect the pruning to learn a mask on the", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "score": 1.0, + "content": "given model while not impairing the excellent OOD performance that UM brings. In Figure 8, it", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 621, + 506, + 636 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 297, + 636 + ], + "score": 1.0, + "content": "presents that pruning from a wide range (e.g.", + "type": "text" + }, + { + "bbox": [ + 298, + 622, + 355, + 634 + ], + "score": 0.89, + "content": "p \\in [ 0 . 3 , 0 . 9 ] )", + "type": "inline_equation" + }, + { + "bbox": [ + 355, + 621, + 506, + 636 + ], + "score": 1.0, + "content": ") can well maintain the effectiveness", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 632, + 506, + 648 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 506, + 648 + ], + "score": 1.0, + "content": "of UM while possessing a terrific convergence trend. For simplicity, we use prune to indicate the", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 104, + 643, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 104, + 643, + 506, + 657 + ], + "score": 1.0, + "content": "original pruning approach and UMAP indicate UM with pruning on the mask with our newly designed", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "score": 1.0, + "content": "forgetting objective in Figure 8. In Figure 8(a), the solid lines represent the proposed UMAP and the", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "score": 1.0, + "content": "dashed lines represent only pruning the well-trained model at prune rates 0.2, 0.5, and 0.8. While the", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 104, + 676, + 507, + 691 + ], + "spans": [ + { + "bbox": [ + 104, + 676, + 507, + 691 + ], + "score": 1.0, + "content": "model’s OOD performance can’t be improved (not better than the baseline) through only pruning,", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "using our proposed forgetting objective for the loss constrain can significantly bring out better OOD", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 298, + 712 + ], + "score": 1.0, + "content": "performance at a wide range of mask rates (e.g.", + "type": "text" + }, + { + "bbox": [ + 299, + 699, + 354, + 711 + ], + "score": 0.84, + "content": "p \\in [ 0 . 5 , 0 . 8 ] \\rangle", + "type": "inline_equation" + }, + { + "bbox": [ + 354, + 699, + 506, + 712 + ], + "score": 1.0, + "content": "). In Figure 8(b), we intuitively reflect", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "the effect of the estimated loss constraint by the initialized mask which redirects the gradients when", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 720, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 505, + 733 + ], + "score": 1.0, + "content": "the loss reaches the value, while the loss will just approach 0 when pruning only. In Figure 8(c), we", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 26 + } + ], + "page_idx": 23, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 108, + 112, + 505, + 253 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 80, + 504, + 103 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 79, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 79, + 234, + 93 + ], + "score": 1.0, + "content": "Table 20: Fine-grained Results", + "type": "text" + }, + { + "bbox": [ + 235, + 80, + 250, + 91 + ], + "score": 0.81, + "content": "( \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 251, + 79, + 505, + 93 + ], + "score": 1.0, + "content": "of WRN-40-4 on CIFAR-100. Comparison on different OOD", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 90, + 490, + 104 + ], + "spans": [ + { + "bbox": [ + 105, + 90, + 189, + 104 + ], + "score": 1.0, + "content": "benchmark datasets.", + "type": "text" + }, + { + "bbox": [ + 190, + 92, + 197, + 103 + ], + "score": 0.73, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 198, + 90, + 349, + 104 + ], + "score": 1.0, + "content": "indicates higher values are better, and", + "type": "text" + }, + { + "bbox": [ + 350, + 92, + 357, + 102 + ], + "score": 0.26, + "content": "\\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 358, + 90, + 490, + 104 + ], + "score": 1.0, + "content": "indicates lower values are better.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 108, + 112, + 505, + 253 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 112, + 505, + 253 + ], + "spans": [ + { + "bbox": [ + 108, + 112, + 505, + 253 + ], + "score": 0.975, + "html": "
ID datasetMethodOOD dataset
CIFAR-10TexturesPlaces365
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
CIFAR-100MSP83.83 ±0.2975.50±0.2186.15 ± 0.2372.36 ± 0.4086.72 ±0.2969.60± 0.27
ODIN83.70 ±0.3074.32 ±0.0381.57 ± 1.7471.67±0.2888.07 ± 0.1164.83 ± 1.36
Mahalanobis96.89 ± 0.1168.78 ±0.6731.02 ± 2.0491.85 ± 0.9193.34 ± 0.4461.28 ± 1.62
Energy81.32 ± 0.4777.49 ± 0.2686.38 ±0.4973.50±0.4584.45 ± 0.3869.82 ± 0.63
Energy+UM(ours)89.23 ± 1.5163.85 ± 1.7378.90 ±0.0772.58 ± 1.3380.46 ± 1.9970.49 ± 1.01
Energy+UMAP(ours)94.11 ± 0.7260.77 ± 0.9666.94 ± 4.4975.82 ± 5.3382.59±0.2771.92 ± 3.58
MethodSUNLSUNiNaturalist
FPR95↓AUROC↑FPR95↓AUROC↑FPR95↓AUROC↑
MSP88.88±0.8365.22 ± 0.8578.56±0.6679.10 ± 0.5386.72 ± 0.2973,75 ±0.56
ODIN91.00 ±0.1059.06 ± 1.8170.14 ± 2.4284.03±0.8687.86±1.1364.52 ± 1.51
Mahalanobis94.22 ± 0.0160.09 ± 1.5689.73± 2.8740.81± 3.0787.25±3.2874.98 ± 2.85
Energy88.35±0.5264.04 ±0.7659.84 ±0.0687.91 ± 0.5388.91±0.7867.81 ± 0.91
Energy+UM(ours)84.04 ±0.0967.19 ± 0.1433.87 ± 1.2192.29 ± 0.0276.91 ± 6.0779.28 ± 4.17
Energy+UMAP(ours)80.53 ± 1.3172.68± 4.5027.79 ± 2.1993.39 ± 0.5755.53± 4.6185.65± 0.83
", + "type": "table", + "image_path": "db8ab563d2f7b2f21e22411a59359b785cf8ebd0be133a80a03550c6303ee097.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 108, + 112, + 505, + 159.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 108, + 159.0, + 505, + 206.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 108, + 206.0, + 505, + 253.0 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "image", + "bbox": [ + 118, + 273, + 493, + 401 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 118, + 273, + 493, + 401 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 118, + 273, + 493, + 401 + ], + "spans": [ + { + "bbox": [ + 118, + 273, + 493, + 401 + ], + "score": 0.971, + "type": "image", + "image_path": "1674497873272ebcef525d5b25ffb65e99d34866ce9d3e4d699f013f99848ca6.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 118, + 273, + 493, + 315.6666666666667 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 118, + 315.6666666666667, + 493, + 358.33333333333337 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 118, + 358.33333333333337, + 493, + 401.00000000000006 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 418, + 505, + 474 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 419, + 505, + 430 + ], + "spans": [ + { + "bbox": [ + 106, + 419, + 505, + 430 + ], + "score": 1.0, + "content": "Figure 5: Ablation studies on three metrics with 4 different learning rate schedules. The model is", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 429, + 505, + 442 + ], + "spans": [ + { + "bbox": [ + 105, + 429, + 505, + 442 + ], + "score": 1.0, + "content": "DenseNet-101 trained on CIFAR-100 with iNaturalist as the OOD dataset. (a) change of FPR95", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 440, + 505, + 453 + ], + "spans": [ + { + "bbox": [ + 106, + 440, + 505, + 453 + ], + "score": 1.0, + "content": "throughout the pruning phase when training on CIFAR-100; (b) change of AUROC throughout the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 451, + 505, + 465 + ], + "spans": [ + { + "bbox": [ + 105, + 451, + 505, + 465 + ], + "score": 1.0, + "content": "pruning phase when training on CIFAR-100; (c) change of AUPR throughout the pruning phase when", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 462, + 502, + 475 + ], + "spans": [ + { + "bbox": [ + 106, + 462, + 502, + 475 + ], + "score": 1.0, + "content": "training on CIFAR-100. It demonstrates a better middle stage exists according to the three metrics.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 10 + } + ], + "index": 8.0 + }, + { + "type": "text", + "bbox": [ + 106, + 496, + 505, + 574 + ], + "lines": [], + "index": 16, + "bbox_fs": [ + 105, + 496, + 506, + 575 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 106, + 589, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 588, + 506, + 602 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 506, + 602 + ], + "score": 1.0, + "content": "UMAP: adopting pruning on UM. We conduct various experiments to see whether pruning has", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 600, + 506, + 613 + ], + "spans": [ + { + "bbox": [ + 106, + 600, + 506, + 613 + ], + "score": 1.0, + "content": "an impact on Unleashing Mask itself. To be specific, we expect the pruning to learn a mask on the", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "score": 1.0, + "content": "given model while not impairing the excellent OOD performance that UM brings. In Figure 8, it", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 621, + 506, + 636 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 297, + 636 + ], + "score": 1.0, + "content": "presents that pruning from a wide range (e.g.", + "type": "text" + }, + { + "bbox": [ + 298, + 622, + 355, + 634 + ], + "score": 0.89, + "content": "p \\in [ 0 . 3 , 0 . 9 ] )", + "type": "inline_equation" + }, + { + "bbox": [ + 355, + 621, + 506, + 636 + ], + "score": 1.0, + "content": ") can well maintain the effectiveness", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 632, + 506, + 648 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 506, + 648 + ], + "score": 1.0, + "content": "of UM while possessing a terrific convergence trend. For simplicity, we use prune to indicate the", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 104, + 643, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 104, + 643, + 506, + 657 + ], + "score": 1.0, + "content": "original pruning approach and UMAP indicate UM with pruning on the mask with our newly designed", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "score": 1.0, + "content": "forgetting objective in Figure 8. In Figure 8(a), the solid lines represent the proposed UMAP and the", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 505, + 678 + ], + "score": 1.0, + "content": "dashed lines represent only pruning the well-trained model at prune rates 0.2, 0.5, and 0.8. While the", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 104, + 676, + 507, + 691 + ], + "spans": [ + { + "bbox": [ + 104, + 676, + 507, + 691 + ], + "score": 1.0, + "content": "model’s OOD performance can’t be improved (not better than the baseline) through only pruning,", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "using our proposed forgetting objective for the loss constrain can significantly bring out better OOD", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 298, + 712 + ], + "score": 1.0, + "content": "performance at a wide range of mask rates (e.g.", + "type": "text" + }, + { + "bbox": [ + 299, + 699, + 354, + 711 + ], + "score": 0.84, + "content": "p \\in [ 0 . 5 , 0 . 8 ] \\rangle", + "type": "inline_equation" + }, + { + "bbox": [ + 354, + 699, + 506, + 712 + ], + "score": 1.0, + "content": "). In Figure 8(b), we intuitively reflect", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "the effect of the estimated loss constraint by the initialized mask which redirects the gradients when", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 720, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 505, + 733 + ], + "score": 1.0, + "content": "the loss reaches the value, while the loss will just approach 0 when pruning only. In Figure 8(c), we", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 525, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 525, + 506, + 540 + ], + "score": 1.0, + "content": "can see that ID-ACC for both UMAP and Prune can converge to approximately the same high level", + "type": "text", + "cross_page": true + } + ], + "index": 16 + }, + { + "bbox": [ + 109, + 537, + 472, + 550 + ], + "spans": [ + { + "bbox": [ + 109, + 537, + 145, + 548 + ], + "score": 0.82, + "content": "( 9 2 \\sim 9 4 ", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 145, + 537, + 472, + 550 + ], + "score": 1.0, + "content": "), though we can simply remove the learned mask to recover the original ID-ACC.", + "type": "text", + "cross_page": true + } + ], + "index": 17 + } + ], + "index": 26, + "bbox_fs": [ + 104, + 588, + 507, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 115, + 84, + 496, + 214 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 84, + 496, + 214 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 115, + 84, + 496, + 214 + ], + "spans": [ + { + "bbox": [ + 115, + 84, + 496, + 214 + ], + "score": 0.967, + "type": "image", + "image_path": "7f407edbb5ffbadd5e84855897933477d5ac0f0382566b843728bbc0b1f1e5c4.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 115, + 84, + 496, + 127.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 115, + 127.33333333333334, + 496, + 170.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 115, + 170.66666666666669, + 496, + 214.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 231, + 506, + 286 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 231, + 507, + 243 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 507, + 243 + ], + "score": 1.0, + "content": "Figure 6: Ablation studies on three metrics of WRN-40-4 with CIFAR-10 as ID dataset, SVHN,", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 241, + 506, + 256 + ], + "spans": [ + { + "bbox": [ + 105, + 241, + 506, + 256 + ], + "score": 1.0, + "content": "and Textures as OOD datasets. (a) change of FPR95 throughout the pruning phase when training on", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 252, + 506, + 267 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 506, + 267 + ], + "score": 1.0, + "content": "CIFAR-10; (b) change of AUROC throughout the pruning phase when training on CIFAR-10; (c)", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 263, + 506, + 278 + ], + "spans": [ + { + "bbox": [ + 105, + 263, + 506, + 278 + ], + "score": 1.0, + "content": "change of AUPR throughout the pruning phase when training on CIFAR-10. It demonstrates a better", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 274, + 308, + 289 + ], + "spans": [ + { + "bbox": [ + 105, + 274, + 308, + 289 + ], + "score": 1.0, + "content": "middle stage exists according to the three metrics.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5 + } + ], + "index": 3.0 + }, + { + "type": "image", + "bbox": [ + 118, + 303, + 492, + 432 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 118, + 303, + 492, + 432 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 118, + 303, + 492, + 432 + ], + "spans": [ + { + "bbox": [ + 118, + 303, + 492, + 432 + ], + "score": 0.973, + "type": "image", + "image_path": "24ec436949fcf37fd12376ca16998e59459938abcbe0fd87b4d16de8c5362499.jpg" + } + ] + } + ], + "index": 9, + "virtual_lines": [ + { + "bbox": [ + 118, + 303, + 492, + 346.0 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 118, + 346.0, + 492, + 389.0 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 118, + 389.0, + 492, + 432.0 + ], + "spans": [], + "index": 10 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 449, + 506, + 504 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 448, + 506, + 462 + ], + "spans": [ + { + "bbox": [ + 105, + 448, + 506, + 462 + ], + "score": 1.0, + "content": "Figure 7: Ablation studies on three metrics of WRN-40-4 with CIFAR-100 as ID dataset, iNaturalist,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 459, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 505, + 473 + ], + "score": 1.0, + "content": "Places365, and SUN as OOD datasets. (a) change of FPR95 throughout the pruning phase when", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 471, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 471, + 505, + 484 + ], + "score": 1.0, + "content": "training on CIFAR-100; (b) change of AUROC throughout the pruning phase when training on", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 482, + 506, + 495 + ], + "spans": [ + { + "bbox": [ + 106, + 482, + 506, + 495 + ], + "score": 1.0, + "content": "CIFAR-100; (c) change of AUPR throughout the pruning phase when training on CIFAR-100. It", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 492, + 395, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 395, + 506 + ], + "score": 1.0, + "content": "demonstrates a better middle stage exists according to the three metrics.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 13 + } + ], + "index": 11.0 + }, + { + "type": "text", + "bbox": [ + 106, + 526, + 504, + 549 + ], + "lines": [ + { + "bbox": [ + 105, + 525, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 525, + 506, + 540 + ], + "score": 1.0, + "content": "can see that ID-ACC for both UMAP and Prune can converge to approximately the same high level", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 109, + 537, + 472, + 550 + ], + "spans": [ + { + "bbox": [ + 109, + 537, + 145, + 548 + ], + "score": 0.82, + "content": "( 9 2 \\sim 9 4 ", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 537, + 472, + 550 + ], + "score": 1.0, + "content": "), though we can simply remove the learned mask to recover the original ID-ACC.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5 + }, + { + "type": "text", + "bbox": [ + 107, + 561, + 505, + 704 + ], + "lines": [ + { + "bbox": [ + 106, + 560, + 505, + 574 + ], + "spans": [ + { + "bbox": [ + 106, + 560, + 505, + 574 + ], + "score": 1.0, + "content": "The effectiveness of UM. In Figure 9, we present the FPR95, AUROC, and AUPR curves during", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 572, + 505, + 584 + ], + "spans": [ + { + "bbox": [ + 106, + 572, + 505, + 584 + ], + "score": 1.0, + "content": "training to show the comparison of the original training and our proposed UM on ID data. We observe", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 583, + 505, + 595 + ], + "spans": [ + { + "bbox": [ + 106, + 583, + 505, + 595 + ], + "score": 1.0, + "content": "that training using UM can consistently outperform than the vanilla model training, either for the", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 594, + 504, + 606 + ], + "spans": [ + { + "bbox": [ + 106, + 594, + 504, + 606 + ], + "score": 1.0, + "content": "final stage or the middle stage with the best OOD detection performance indicated by the FPR95", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 605, + 505, + 617 + ], + "spans": [ + { + "bbox": [ + 105, + 605, + 505, + 617 + ], + "score": 1.0, + "content": "curve. In Figure 10, we also adopt different mask rates for the initialized loss constraint estimation", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 616, + 504, + 628 + ], + "spans": [ + { + "bbox": [ + 106, + 616, + 504, + 628 + ], + "score": 1.0, + "content": "for forgetting the atypical samples. The results show that a wide range of mask ratios (i.e., from", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 626, + 505, + 639 + ], + "spans": [ + { + "bbox": [ + 106, + 627, + 126, + 637 + ], + "score": 0.87, + "content": "96 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 626, + 137, + 639 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 137, + 627, + 156, + 637 + ], + "score": 0.85, + "content": "9 9 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 157, + 626, + 505, + 639 + ], + "score": 1.0, + "content": ") to estimate the loss constraint used in Eq.3 can gain better OOD detection performance", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 637, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 106, + 637, + 505, + 650 + ], + "score": 1.0, + "content": "than the baseline. It shows the mask ratio would be robust to hyper-parameter selection under a", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 649, + 505, + 660 + ], + "spans": [ + { + "bbox": [ + 106, + 649, + 505, + 660 + ], + "score": 1.0, + "content": "certain value. The principle intuition behind this is our revealed important observation as indicated", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 658, + 505, + 673 + ], + "spans": [ + { + "bbox": [ + 105, + 658, + 505, + 673 + ], + "score": 1.0, + "content": "in Figures 1(a), 2(b), and 2(c). With the guidance of the general mechanism, empirically choosing", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 670, + 505, + 683 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 505, + 683 + ], + "score": 1.0, + "content": "the hyper-parameter using the validation set is supportable and valuable for excavating better OOD", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 681, + 505, + 694 + ], + "spans": [ + { + "bbox": [ + 106, + 681, + 505, + 694 + ], + "score": 1.0, + "content": "detection capability of the model as conducted by previous literature (Hendrycks et al., 2019b; Liu", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 691, + 226, + 704 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 226, + 704 + ], + "score": 1.0, + "content": "et al., 2020; Sun et al., 2021).", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 106, + 710, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 104, + 708, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 104, + 708, + 506, + 723 + ], + "score": 1.0, + "content": "In our experiments, we empirically determine the value of our proposed UM and UMAP by examining", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 720, + 505, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 477, + 732 + ], + "score": 1.0, + "content": "the training loss on the masked output. For CIFAR-10 as ID datasets, the value of mask rate is", + "type": "text" + }, + { + "bbox": [ + 477, + 720, + 505, + 731 + ], + "score": 0.85, + "content": "9 7 . 5 \\%", + "type": "inline_equation" + } + ], + "index": 32 + } + ], + "index": 31.5 + } + ], + "page_idx": 24, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 115, + 84, + 496, + 214 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 84, + 496, + 214 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 115, + 84, + 496, + 214 + ], + "spans": [ + { + "bbox": [ + 115, + 84, + 496, + 214 + ], + "score": 0.967, + "type": "image", + "image_path": "7f407edbb5ffbadd5e84855897933477d5ac0f0382566b843728bbc0b1f1e5c4.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 115, + 84, + 496, + 127.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 115, + 127.33333333333334, + 496, + 170.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 115, + 170.66666666666669, + 496, + 214.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 231, + 506, + 286 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 231, + 507, + 243 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 507, + 243 + ], + "score": 1.0, + "content": "Figure 6: Ablation studies on three metrics of WRN-40-4 with CIFAR-10 as ID dataset, SVHN,", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 241, + 506, + 256 + ], + "spans": [ + { + "bbox": [ + 105, + 241, + 506, + 256 + ], + "score": 1.0, + "content": "and Textures as OOD datasets. (a) change of FPR95 throughout the pruning phase when training on", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 252, + 506, + 267 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 506, + 267 + ], + "score": 1.0, + "content": "CIFAR-10; (b) change of AUROC throughout the pruning phase when training on CIFAR-10; (c)", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 263, + 506, + 278 + ], + "spans": [ + { + "bbox": [ + 105, + 263, + 506, + 278 + ], + "score": 1.0, + "content": "change of AUPR throughout the pruning phase when training on CIFAR-10. It demonstrates a better", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 274, + 308, + 289 + ], + "spans": [ + { + "bbox": [ + 105, + 274, + 308, + 289 + ], + "score": 1.0, + "content": "middle stage exists according to the three metrics.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5 + } + ], + "index": 3.0 + }, + { + "type": "image", + "bbox": [ + 118, + 303, + 492, + 432 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 118, + 303, + 492, + 432 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 118, + 303, + 492, + 432 + ], + "spans": [ + { + "bbox": [ + 118, + 303, + 492, + 432 + ], + "score": 0.973, + "type": "image", + "image_path": "24ec436949fcf37fd12376ca16998e59459938abcbe0fd87b4d16de8c5362499.jpg" + } + ] + } + ], + "index": 9, + "virtual_lines": [ + { + "bbox": [ + 118, + 303, + 492, + 346.0 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 118, + 346.0, + 492, + 389.0 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 118, + 389.0, + 492, + 432.0 + ], + "spans": [], + "index": 10 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 449, + 506, + 504 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 448, + 506, + 462 + ], + "spans": [ + { + "bbox": [ + 105, + 448, + 506, + 462 + ], + "score": 1.0, + "content": "Figure 7: Ablation studies on three metrics of WRN-40-4 with CIFAR-100 as ID dataset, iNaturalist,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 459, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 505, + 473 + ], + "score": 1.0, + "content": "Places365, and SUN as OOD datasets. (a) change of FPR95 throughout the pruning phase when", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 471, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 471, + 505, + 484 + ], + "score": 1.0, + "content": "training on CIFAR-100; (b) change of AUROC throughout the pruning phase when training on", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 482, + 506, + 495 + ], + "spans": [ + { + "bbox": [ + 106, + 482, + 506, + 495 + ], + "score": 1.0, + "content": "CIFAR-100; (c) change of AUPR throughout the pruning phase when training on CIFAR-100. It", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 492, + 395, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 395, + 506 + ], + "score": 1.0, + "content": "demonstrates a better middle stage exists according to the three metrics.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 13 + } + ], + "index": 11.0 + }, + { + "type": "text", + "bbox": [ + 106, + 526, + 504, + 549 + ], + "lines": [], + "index": 16.5, + "bbox_fs": [ + 105, + 525, + 506, + 550 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 561, + 505, + 704 + ], + "lines": [ + { + "bbox": [ + 106, + 560, + 505, + 574 + ], + "spans": [ + { + "bbox": [ + 106, + 560, + 505, + 574 + ], + "score": 1.0, + "content": "The effectiveness of UM. In Figure 9, we present the FPR95, AUROC, and AUPR curves during", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 572, + 505, + 584 + ], + "spans": [ + { + "bbox": [ + 106, + 572, + 505, + 584 + ], + "score": 1.0, + "content": "training to show the comparison of the original training and our proposed UM on ID data. We observe", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 583, + 505, + 595 + ], + "spans": [ + { + "bbox": [ + 106, + 583, + 505, + 595 + ], + "score": 1.0, + "content": "that training using UM can consistently outperform than the vanilla model training, either for the", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 594, + 504, + 606 + ], + "spans": [ + { + "bbox": [ + 106, + 594, + 504, + 606 + ], + "score": 1.0, + "content": "final stage or the middle stage with the best OOD detection performance indicated by the FPR95", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 605, + 505, + 617 + ], + "spans": [ + { + "bbox": [ + 105, + 605, + 505, + 617 + ], + "score": 1.0, + "content": "curve. In Figure 10, we also adopt different mask rates for the initialized loss constraint estimation", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 616, + 504, + 628 + ], + "spans": [ + { + "bbox": [ + 106, + 616, + 504, + 628 + ], + "score": 1.0, + "content": "for forgetting the atypical samples. The results show that a wide range of mask ratios (i.e., from", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 626, + 505, + 639 + ], + "spans": [ + { + "bbox": [ + 106, + 627, + 126, + 637 + ], + "score": 0.87, + "content": "96 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 626, + 137, + 639 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 137, + 627, + 156, + 637 + ], + "score": 0.85, + "content": "9 9 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 157, + 626, + 505, + 639 + ], + "score": 1.0, + "content": ") to estimate the loss constraint used in Eq.3 can gain better OOD detection performance", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 637, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 106, + 637, + 505, + 650 + ], + "score": 1.0, + "content": "than the baseline. It shows the mask ratio would be robust to hyper-parameter selection under a", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 649, + 505, + 660 + ], + "spans": [ + { + "bbox": [ + 106, + 649, + 505, + 660 + ], + "score": 1.0, + "content": "certain value. The principle intuition behind this is our revealed important observation as indicated", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 658, + 505, + 673 + ], + "spans": [ + { + "bbox": [ + 105, + 658, + 505, + 673 + ], + "score": 1.0, + "content": "in Figures 1(a), 2(b), and 2(c). With the guidance of the general mechanism, empirically choosing", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 670, + 505, + 683 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 505, + 683 + ], + "score": 1.0, + "content": "the hyper-parameter using the validation set is supportable and valuable for excavating better OOD", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 681, + 505, + 694 + ], + "spans": [ + { + "bbox": [ + 106, + 681, + 505, + 694 + ], + "score": 1.0, + "content": "detection capability of the model as conducted by previous literature (Hendrycks et al., 2019b; Liu", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 691, + 226, + 704 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 226, + 704 + ], + "score": 1.0, + "content": "et al., 2020; Sun et al., 2021).", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 24, + "bbox_fs": [ + 105, + 560, + 505, + 704 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 710, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 104, + 708, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 104, + 708, + 506, + 723 + ], + "score": 1.0, + "content": "In our experiments, we empirically determine the value of our proposed UM and UMAP by examining", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 720, + 505, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 477, + 732 + ], + "score": 1.0, + "content": "the training loss on the masked output. For CIFAR-10 as ID datasets, the value of mask rate is", + "type": "text" + }, + { + "bbox": [ + 477, + 720, + 505, + 731 + ], + "score": 0.85, + "content": "9 7 . 5 \\%", + "type": "inline_equation" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 482, + 506, + 494 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 506, + 494 + ], + "score": 1.0, + "content": "and the estimated loss constraint for forgetting is 0.10 for our tuning until the convergence; For", + "type": "text", + "cross_page": true + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 492, + 506, + 505 + ], + "spans": [ + { + "bbox": [ + 106, + 492, + 255, + 505 + ], + "score": 1.0, + "content": "CIFAR-100, the value of mask rate is", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 256, + 493, + 276, + 504 + ], + "score": 0.88, + "content": "9 7 \\%", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 276, + 492, + 506, + 505 + ], + "score": 1.0, + "content": "and the estimated loss constraint for forgetting is 1.20 for", + "type": "text", + "cross_page": true + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 504, + 240, + 516 + ], + "spans": [ + { + "bbox": [ + 105, + 504, + 240, + 516 + ], + "score": 1.0, + "content": "our tuning until the convergence.", + "type": "text", + "cross_page": true + } + ], + "index": 14 + } + ], + "index": 31.5, + "bbox_fs": [ + 104, + 708, + 506, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 119, + 85, + 492, + 214 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 119, + 85, + 492, + 214 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 119, + 85, + 492, + 214 + ], + "spans": [ + { + "bbox": [ + 119, + 85, + 492, + 214 + ], + "score": 0.97, + "type": "image", + "image_path": "be945de072e51170925a45b431f1d3d1c0191ad541bc3cedc09a89a54d7509db.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 119, + 85, + 492, + 128.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 119, + 128.0, + 492, + 171.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 119, + 171.0, + 492, + 214.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 231, + 506, + 265 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 231, + 506, + 244 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 506, + 244 + ], + "score": 1.0, + "content": "Figure 8: Ablation studies on Prune Rate of UMAP. (a) change of OOD performance throughout the", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 243, + 505, + 255 + ], + "spans": [ + { + "bbox": [ + 106, + 243, + 505, + 255 + ], + "score": 1.0, + "content": "pruning phase; (b) training loss converges to estimated loss constraint properly; (c) though ID-ACC", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 253, + 460, + 266 + ], + "spans": [ + { + "bbox": [ + 105, + 253, + 460, + 266 + ], + "score": 1.0, + "content": "is not taken into consideration for UMAP, it still raise high after training for 100 epochs.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 120, + 281, + 490, + 410 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 120, + 281, + 490, + 410 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 120, + 281, + 490, + 410 + ], + "spans": [ + { + "bbox": [ + 120, + 281, + 490, + 410 + ], + "score": 0.971, + "type": "image", + "image_path": "70757eca6d628ecb94634d260d5782c79827c5486109a38181ba093662388d29.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 120, + 281, + 490, + 324.0 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 120, + 324.0, + 490, + 367.0 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 120, + 367.0, + 490, + 410.0 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 426, + 505, + 461 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 426, + 506, + 440 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 460, + 440 + ], + "score": 1.0, + "content": "Figure 9: Ablation studies to reflect the effectiveness of UM. The mask ratio of UM is", + "type": "text" + }, + { + "bbox": [ + 460, + 426, + 487, + 438 + ], + "score": 0.87, + "content": "9 9 . 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 488, + 426, + 506, + 440 + ], + "score": 1.0, + "content": ". (a)", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 438, + 505, + 451 + ], + "spans": [ + { + "bbox": [ + 106, + 438, + 505, + 451 + ], + "score": 1.0, + "content": "change of FPR95 throughout the training phase on CIFAR-10; (b) change of AUROC throughout the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 449, + 487, + 461 + ], + "spans": [ + { + "bbox": [ + 105, + 449, + 487, + 461 + ], + "score": 1.0, + "content": "training phase on CIFAR-10; (c) change of AUPR throughout the training phase on CIFAR-10.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 107, + 482, + 505, + 515 + ], + "lines": [ + { + "bbox": [ + 105, + 482, + 506, + 494 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 506, + 494 + ], + "score": 1.0, + "content": "and the estimated loss constraint for forgetting is 0.10 for our tuning until the convergence; For", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 492, + 506, + 505 + ], + "spans": [ + { + "bbox": [ + 106, + 492, + 255, + 505 + ], + "score": 1.0, + "content": "CIFAR-100, the value of mask rate is", + "type": "text" + }, + { + "bbox": [ + 256, + 493, + 276, + 504 + ], + "score": 0.88, + "content": "9 7 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 276, + 492, + 506, + 505 + ], + "score": 1.0, + "content": "and the estimated loss constraint for forgetting is 1.20 for", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 504, + 240, + 516 + ], + "spans": [ + { + "bbox": [ + 105, + 504, + 240, + 516 + ], + "score": 1.0, + "content": "our tuning until the convergence.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 106, + 521, + 505, + 598 + ], + "lines": [ + { + "bbox": [ + 106, + 521, + 505, + 532 + ], + "spans": [ + { + "bbox": [ + 106, + 521, + 505, + 532 + ], + "score": 1.0, + "content": "To choose the parameters of the estimated loss constraint, we use the TinyImageNet (Tavanaei, 2020)", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 531, + 506, + 544 + ], + "spans": [ + { + "bbox": [ + 105, + 531, + 506, + 544 + ], + "score": 1.0, + "content": "dataset as the validation set, which is not seen during training and is not considered in our evaluation", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 542, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 106, + 542, + 505, + 555 + ], + "score": 1.0, + "content": "of OOD detection performance. Since the core intuition behind our method is to restore the OOD", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 554, + 504, + 566 + ], + "spans": [ + { + "bbox": [ + 106, + 554, + 504, + 566 + ], + "score": 1.0, + "content": "detection performance starting from the well-trained model stage, forgetting a relatively small portion", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 564, + 505, + 576 + ], + "spans": [ + { + "bbox": [ + 106, + 564, + 211, + 576 + ], + "score": 1.0, + "content": "(empirically found around", + "type": "text" + }, + { + "bbox": [ + 211, + 565, + 231, + 576 + ], + "score": 0.86, + "content": "9 7 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 231, + 564, + 505, + 576 + ], + "score": 1.0, + "content": "mask ratio) of atypical samples can be beneficial. To find the optimal", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 576, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 106, + 576, + 505, + 588 + ], + "score": 1.0, + "content": "parameter for tuning, more advanced searching techniques like AutoML or validation design based", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 587, + 429, + 599 + ], + "spans": [ + { + "bbox": [ + 106, + 587, + 429, + 599 + ], + "score": 1.0, + "content": "on the important observation in our work may be further employed in the future.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 106, + 610, + 505, + 699 + ], + "lines": [ + { + "bbox": [ + 105, + 610, + 506, + 623 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 506, + 623 + ], + "score": 1.0, + "content": "Fine-grained comparison of the model weights. We display the weights of the original model,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 622, + 505, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 622, + 505, + 634 + ], + "score": 1.0, + "content": "pruned model, and the UMAP model respectively in Figure 11. The histograms show that the", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "adopted pruning algorithm tends to choose weights far from 0 for the first convolution layer, shown in", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 644, + 505, + 655 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 505, + 655 + ], + "score": 1.0, + "content": "Figure 11(a). However, for almost all layers (from the 2nd to the 98th), the pruning chooses weights", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "score": 1.0, + "content": "with no respect to the value of weights, shown in Figure 11(b). For the \"head\" of the model (the fully", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "score": 1.0, + "content": "connected layer), the pruning algorithm itself still keeps its behavior on the first layer, while UMAP", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "forces the prune algorithm to choose weights near 0, shown in Figure 11(c), indicating that forgetting", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 687, + 480, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 480, + 700 + ], + "score": 1.0, + "content": "learned atypical samples doesn’t necessarily correspond to larger weights or smaller weights.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 25.5 + } + ], + "page_idx": 25, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 119, + 85, + 492, + 214 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 119, + 85, + 492, + 214 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 119, + 85, + 492, + 214 + ], + "spans": [ + { + "bbox": [ + 119, + 85, + 492, + 214 + ], + "score": 0.97, + "type": "image", + "image_path": "be945de072e51170925a45b431f1d3d1c0191ad541bc3cedc09a89a54d7509db.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 119, + 85, + 492, + 128.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 119, + 128.0, + 492, + 171.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 119, + 171.0, + 492, + 214.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 231, + 506, + 265 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 231, + 506, + 244 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 506, + 244 + ], + "score": 1.0, + "content": "Figure 8: Ablation studies on Prune Rate of UMAP. (a) change of OOD performance throughout the", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 243, + 505, + 255 + ], + "spans": [ + { + "bbox": [ + 106, + 243, + 505, + 255 + ], + "score": 1.0, + "content": "pruning phase; (b) training loss converges to estimated loss constraint properly; (c) though ID-ACC", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 253, + 460, + 266 + ], + "spans": [ + { + "bbox": [ + 105, + 253, + 460, + 266 + ], + "score": 1.0, + "content": "is not taken into consideration for UMAP, it still raise high after training for 100 epochs.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 120, + 281, + 490, + 410 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 120, + 281, + 490, + 410 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 120, + 281, + 490, + 410 + ], + "spans": [ + { + "bbox": [ + 120, + 281, + 490, + 410 + ], + "score": 0.971, + "type": "image", + "image_path": "70757eca6d628ecb94634d260d5782c79827c5486109a38181ba093662388d29.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 120, + 281, + 490, + 324.0 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 120, + 324.0, + 490, + 367.0 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 120, + 367.0, + 490, + 410.0 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 426, + 505, + 461 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 426, + 506, + 440 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 460, + 440 + ], + "score": 1.0, + "content": "Figure 9: Ablation studies to reflect the effectiveness of UM. The mask ratio of UM is", + "type": "text" + }, + { + "bbox": [ + 460, + 426, + 487, + 438 + ], + "score": 0.87, + "content": "9 9 . 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 488, + 426, + 506, + 440 + ], + "score": 1.0, + "content": ". (a)", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 438, + 505, + 451 + ], + "spans": [ + { + "bbox": [ + 106, + 438, + 505, + 451 + ], + "score": 1.0, + "content": "change of FPR95 throughout the training phase on CIFAR-10; (b) change of AUROC throughout the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 449, + 487, + 461 + ], + "spans": [ + { + "bbox": [ + 105, + 449, + 487, + 461 + ], + "score": 1.0, + "content": "training phase on CIFAR-10; (c) change of AUPR throughout the training phase on CIFAR-10.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 107, + 482, + 505, + 515 + ], + "lines": [], + "index": 13, + "bbox_fs": [ + 105, + 482, + 506, + 516 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 106, + 521, + 505, + 598 + ], + "lines": [ + { + "bbox": [ + 106, + 521, + 505, + 532 + ], + "spans": [ + { + "bbox": [ + 106, + 521, + 505, + 532 + ], + "score": 1.0, + "content": "To choose the parameters of the estimated loss constraint, we use the TinyImageNet (Tavanaei, 2020)", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 531, + 506, + 544 + ], + "spans": [ + { + "bbox": [ + 105, + 531, + 506, + 544 + ], + "score": 1.0, + "content": "dataset as the validation set, which is not seen during training and is not considered in our evaluation", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 542, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 106, + 542, + 505, + 555 + ], + "score": 1.0, + "content": "of OOD detection performance. Since the core intuition behind our method is to restore the OOD", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 554, + 504, + 566 + ], + "spans": [ + { + "bbox": [ + 106, + 554, + 504, + 566 + ], + "score": 1.0, + "content": "detection performance starting from the well-trained model stage, forgetting a relatively small portion", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 564, + 505, + 576 + ], + "spans": [ + { + "bbox": [ + 106, + 564, + 211, + 576 + ], + "score": 1.0, + "content": "(empirically found around", + "type": "text" + }, + { + "bbox": [ + 211, + 565, + 231, + 576 + ], + "score": 0.86, + "content": "9 7 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 231, + 564, + 505, + 576 + ], + "score": 1.0, + "content": "mask ratio) of atypical samples can be beneficial. To find the optimal", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 576, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 106, + 576, + 505, + 588 + ], + "score": 1.0, + "content": "parameter for tuning, more advanced searching techniques like AutoML or validation design based", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 587, + 429, + 599 + ], + "spans": [ + { + "bbox": [ + 106, + 587, + 429, + 599 + ], + "score": 1.0, + "content": "on the important observation in our work may be further employed in the future.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 18, + "bbox_fs": [ + 105, + 521, + 506, + 599 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 610, + 505, + 699 + ], + "lines": [ + { + "bbox": [ + 105, + 610, + 506, + 623 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 506, + 623 + ], + "score": 1.0, + "content": "Fine-grained comparison of the model weights. We display the weights of the original model,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 622, + 505, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 622, + 505, + 634 + ], + "score": 1.0, + "content": "pruned model, and the UMAP model respectively in Figure 11. The histograms show that the", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "adopted pruning algorithm tends to choose weights far from 0 for the first convolution layer, shown in", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 644, + 505, + 655 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 505, + 655 + ], + "score": 1.0, + "content": "Figure 11(a). However, for almost all layers (from the 2nd to the 98th), the pruning chooses weights", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "score": 1.0, + "content": "with no respect to the value of weights, shown in Figure 11(b). For the \"head\" of the model (the fully", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "score": 1.0, + "content": "connected layer), the pruning algorithm itself still keeps its behavior on the first layer, while UMAP", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "forces the prune algorithm to choose weights near 0, shown in Figure 11(c), indicating that forgetting", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 687, + 480, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 480, + 700 + ], + "score": 1.0, + "content": "learned atypical samples doesn’t necessarily correspond to larger weights or smaller weights.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 25.5, + "bbox_fs": [ + 105, + 610, + 506, + 700 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 114, + 68, + 495, + 488 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 114, + 68, + 495, + 488 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 114, + 68, + 495, + 488 + ], + "spans": [ + { + "bbox": [ + 114, + 68, + 495, + 488 + ], + "score": 0.952, + "type": "image", + "image_path": "88729bef6a9f4ab2516da4709ab0cf6ff01bd447fe513e8f4350607f94bf5553.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 114, + 68, + 495, + 208.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 114, + 208.0, + 495, + 348.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 114, + 348.0, + 495, + 488.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 502, + 506, + 537 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 502, + 505, + 514 + ], + "spans": [ + { + "bbox": [ + 106, + 503, + 466, + 514 + ], + "score": 1.0, + "content": "Figure 10: Ablation studies of WRN-40-4 on various Mask Ratios. The mask rate is from", + "type": "text" + }, + { + "bbox": [ + 466, + 502, + 494, + 513 + ], + "score": 0.87, + "content": "9 5 . 8 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 494, + 503, + 505, + 514 + ], + "score": 1.0, + "content": "to", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 513, + 505, + 526 + ], + "spans": [ + { + "bbox": [ + 106, + 514, + 134, + 524 + ], + "score": 0.87, + "content": "9 9 . 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 134, + 513, + 505, + 526 + ], + "score": 1.0, + "content": ". (a) change of FPR95 throughout the training on CIFAR-10; (b) change of AUROC throughout", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 525, + 452, + 537 + ], + "spans": [ + { + "bbox": [ + 106, + 525, + 452, + 537 + ], + "score": 1.0, + "content": "the training on CIFAR-10; (c) change of AUPR throughout the training on CIFAR-10.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 115, + 548, + 495, + 678 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 548, + 495, + 678 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 115, + 548, + 495, + 678 + ], + "spans": [ + { + "bbox": [ + 115, + 548, + 495, + 678 + ], + "score": 0.97, + "type": "image", + "image_path": "6123471821ec415b46dc126c81c28e0604d37b4633d536b6328c02e27cb00670.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 115, + 548, + 495, + 591.3333333333334 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 115, + 591.3333333333334, + 495, + 634.6666666666667 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 115, + 634.6666666666667, + 495, + 678.0000000000001 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 695, + 506, + 729 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 694, + 507, + 708 + ], + "spans": [ + { + "bbox": [ + 105, + 694, + 507, + 708 + ], + "score": 1.0, + "content": "Figure 11: Histograms of different layers for the original model, pruned model, and UMAP model.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 704, + 506, + 720 + ], + "spans": [ + { + "bbox": [ + 105, + 704, + 297, + 720 + ], + "score": 1.0, + "content": "The model is DenseNet-101 with a prune rate of", + "type": "text" + }, + { + "bbox": [ + 298, + 706, + 317, + 717 + ], + "score": 0.88, + "content": "5 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 317, + 704, + 506, + 720 + ], + "score": 1.0, + "content": ". (a) the histogram of the first convolution layer;", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 716, + 505, + 730 + ], + "spans": [ + { + "bbox": [ + 105, + 716, + 505, + 730 + ], + "score": 1.0, + "content": "(b) the histogram of the 50th convolution layer; (c) the histogram of the last (fully connected) layer.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + } + ], + "index": 8.5 + } + ], + "page_idx": 26, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 106, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "27", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 114, + 68, + 495, + 488 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 114, + 68, + 495, + 488 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 114, + 68, + 495, + 488 + ], + "spans": [ + { + "bbox": [ + 114, + 68, + 495, + 488 + ], + "score": 0.952, + "type": "image", + "image_path": "88729bef6a9f4ab2516da4709ab0cf6ff01bd447fe513e8f4350607f94bf5553.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 114, + 68, + 495, + 208.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 114, + 208.0, + 495, + 348.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 114, + 348.0, + 495, + 488.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 502, + 506, + 537 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 502, + 505, + 514 + ], + "spans": [ + { + "bbox": [ + 106, + 503, + 466, + 514 + ], + "score": 1.0, + "content": "Figure 10: Ablation studies of WRN-40-4 on various Mask Ratios. The mask rate is from", + "type": "text" + }, + { + "bbox": [ + 466, + 502, + 494, + 513 + ], + "score": 0.87, + "content": "9 5 . 8 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 494, + 503, + 505, + 514 + ], + "score": 1.0, + "content": "to", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 513, + 505, + 526 + ], + "spans": [ + { + "bbox": [ + 106, + 514, + 134, + 524 + ], + "score": 0.87, + "content": "9 9 . 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 134, + 513, + 505, + 526 + ], + "score": 1.0, + "content": ". (a) change of FPR95 throughout the training on CIFAR-10; (b) change of AUROC throughout", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 525, + 452, + 537 + ], + "spans": [ + { + "bbox": [ + 106, + 525, + 452, + 537 + ], + "score": 1.0, + "content": "the training on CIFAR-10; (c) change of AUPR throughout the training on CIFAR-10.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 115, + 548, + 495, + 678 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 548, + 495, + 678 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 115, + 548, + 495, + 678 + ], + "spans": [ + { + "bbox": [ + 115, + 548, + 495, + 678 + ], + "score": 0.97, + "type": "image", + "image_path": "6123471821ec415b46dc126c81c28e0604d37b4633d536b6328c02e27cb00670.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 115, + 548, + 495, + 591.3333333333334 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 115, + 591.3333333333334, + 495, + 634.6666666666667 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 115, + 634.6666666666667, + 495, + 678.0000000000001 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 695, + 506, + 729 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 694, + 507, + 708 + ], + "spans": [ + { + "bbox": [ + 105, + 694, + 507, + 708 + ], + "score": 1.0, + "content": "Figure 11: Histograms of different layers for the original model, pruned model, and UMAP model.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 704, + 506, + 720 + ], + "spans": [ + { + "bbox": [ + 105, + 704, + 297, + 720 + ], + "score": 1.0, + "content": "The model is DenseNet-101 with a prune rate of", + "type": "text" + }, + { + "bbox": [ + 298, + 706, + 317, + 717 + ], + "score": 0.88, + "content": "5 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 317, + 704, + 506, + 720 + ], + "score": 1.0, + "content": ". (a) the histogram of the first convolution layer;", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 716, + 505, + 730 + ], + "spans": [ + { + "bbox": [ + 105, + 716, + 505, + 730 + ], + "score": 1.0, + "content": "(b) the histogram of the 50th convolution layer; (c) the histogram of the last (fully connected) layer.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + } + ], + "index": 8.5 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 145, + 55, + 464, + 700 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 145, + 55, + 464, + 700 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 145, + 55, + 464, + 700 + ], + "spans": [ + { + "bbox": [ + 145, + 55, + 464, + 700 + ], + "score": 0.938, + "type": "image", + "image_path": "a2e38ac79e2f9708309fa1f31e758448e2d469ae77bb8ce0661afd67144d1919.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 145, + 55, + 464, + 270.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 145, + 270.0, + 464, + 485.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 145, + 485.0, + 464, + 700.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 704, + 505, + 738 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 704, + 506, + 717 + ], + "spans": [ + { + "bbox": [ + 106, + 704, + 506, + 717 + ], + "score": 1.0, + "content": "Figure 12: Examples of misclassified samples after masking the original well-trained model. The", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 714, + 505, + 728 + ], + "spans": [ + { + "bbox": [ + 105, + 714, + 505, + 728 + ], + "score": 1.0, + "content": "scores are estimated according to uniform distribution. (a) layer-wise masking scores; (b) layer-wise", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 726, + 505, + 740 + ], + "spans": [ + { + "bbox": [ + 106, + 726, + 505, + 740 + ], + "score": 1.0, + "content": "masking weights directly; (c) model-wise masking scores; (d) model-wise masking weights directly.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + } + ], + "page_idx": 27, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 106, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 25, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 25, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 312, + 763 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 312, + 763 + ], + "score": 1.0, + "content": "28", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 145, + 55, + 464, + 700 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 145, + 55, + 464, + 700 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 145, + 55, + 464, + 700 + ], + "spans": [ + { + "bbox": [ + 145, + 55, + 464, + 700 + ], + "score": 0.938, + "type": "image", + "image_path": "a2e38ac79e2f9708309fa1f31e758448e2d469ae77bb8ce0661afd67144d1919.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 145, + 55, + 464, + 270.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 145, + 270.0, + 464, + 485.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 145, + 485.0, + 464, + 700.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 704, + 505, + 738 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 704, + 506, + 717 + ], + "spans": [ + { + "bbox": [ + 106, + 704, + 506, + 717 + ], + "score": 1.0, + "content": "Figure 12: Examples of misclassified samples after masking the original well-trained model. The", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 714, + 505, + 728 + ], + "spans": [ + { + "bbox": [ + 105, + 714, + 505, + 728 + ], + "score": 1.0, + "content": "scores are estimated according to uniform distribution. (a) layer-wise masking scores; (b) layer-wise", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 726, + 505, + 740 + ], + "spans": [ + { + "bbox": [ + 106, + 726, + 505, + 740 + ], + "score": 1.0, + "content": "masking weights directly; (c) model-wise masking scores; (d) model-wise masking weights directly.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + } + ] + } + ], + "_backend": "pipeline", + "_version_name": "2.2.2" +} \ No newline at end of file diff --git a/parse/dev/KmtVD97J43e/images/4ee72afdcd58f9a50a74f22fc9f6ae18a713ba18e2359636b3087340548267df.jpg b/parse/dev/KmtVD97J43e/images/4ee72afdcd58f9a50a74f22fc9f6ae18a713ba18e2359636b3087340548267df.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6aeeb930d72f9954c9704abfc36c3e0bd4fb1e4b --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/4ee72afdcd58f9a50a74f22fc9f6ae18a713ba18e2359636b3087340548267df.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8127e74fc51c1c4c0e72cc6c7a41236f363ba99299c9767f8624c698e63c9bf4 +size 40838 diff --git a/parse/dev/KmtVD97J43e/images/51ca099f0a353ef73980d7a9dcb647c3f688cea4fa6d61a24687d05d750f7a7b.jpg b/parse/dev/KmtVD97J43e/images/51ca099f0a353ef73980d7a9dcb647c3f688cea4fa6d61a24687d05d750f7a7b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ef0da15483e63cce753d08bed4184bd28d960753 --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/51ca099f0a353ef73980d7a9dcb647c3f688cea4fa6d61a24687d05d750f7a7b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:61f93248d5e70701b08b924d27e3a43f05abfa7f1eb99ce710698808bc6bc113 +size 94279 diff --git a/parse/dev/KmtVD97J43e/images/5cca456b93b4c6f0e60897b4cd8020353385ebb79c961a4b21bf987792fa8e04.jpg b/parse/dev/KmtVD97J43e/images/5cca456b93b4c6f0e60897b4cd8020353385ebb79c961a4b21bf987792fa8e04.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5fbaada4f529359549aa708f8d6f50c65945677d --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/5cca456b93b4c6f0e60897b4cd8020353385ebb79c961a4b21bf987792fa8e04.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:36a144ee53f947d290a3c8af2696adab698f7c7d11810eb96737bc73eff4127f +size 37496 diff --git a/parse/dev/KmtVD97J43e/images/6216e907c51723bcfce7b4a56b98ff27a67f1ff3656e3f47127c1f106bb3d1ca.jpg b/parse/dev/KmtVD97J43e/images/6216e907c51723bcfce7b4a56b98ff27a67f1ff3656e3f47127c1f106bb3d1ca.jpg new file mode 100644 index 0000000000000000000000000000000000000000..97d8eab724a0d6693a632408fe2c6da3b93dab1d --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/6216e907c51723bcfce7b4a56b98ff27a67f1ff3656e3f47127c1f106bb3d1ca.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8cc11a8caef2941ab935053c1eafa208563f371bf16c3ffd12e454c636427012 +size 42007 diff --git a/parse/dev/KmtVD97J43e/images/70637a4840bc21fbf682e1992488a4d992e36a997e3242438ea090ce0971c051.jpg b/parse/dev/KmtVD97J43e/images/70637a4840bc21fbf682e1992488a4d992e36a997e3242438ea090ce0971c051.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ee7be7be5e6d4c8ecd9e4aa58e1adbbbbc7c8c20 --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/70637a4840bc21fbf682e1992488a4d992e36a997e3242438ea090ce0971c051.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c68bf2ec6bb8db6e445b935c35df2eca5b6fcbc7012e64928b10a9d883154fb +size 36367 diff --git a/parse/dev/KmtVD97J43e/images/9873c080d7592acd9bfc6a38f0eecacbc0de33b5d218091e9ac2fee44f891e77.jpg b/parse/dev/KmtVD97J43e/images/9873c080d7592acd9bfc6a38f0eecacbc0de33b5d218091e9ac2fee44f891e77.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b06650b1173ebe8a87443354ddb909e8800a2400 --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/9873c080d7592acd9bfc6a38f0eecacbc0de33b5d218091e9ac2fee44f891e77.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4237ed18e369002f93e4a3a29593b16aaa6aae1db942ac9c4b4527d61bc50e5 +size 5607 diff --git a/parse/dev/KmtVD97J43e/images/a99f2329098200b4f7756607e6cdbb9eed436c0006f99ccaa8dd103d000ab796.jpg b/parse/dev/KmtVD97J43e/images/a99f2329098200b4f7756607e6cdbb9eed436c0006f99ccaa8dd103d000ab796.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c9a55c4ca2256e3feb5e92c6786ef82710fc32cb --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/a99f2329098200b4f7756607e6cdbb9eed436c0006f99ccaa8dd103d000ab796.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e49f8bc4b18a25577042145f00cbecec9c8d47317c9eb4d231c993e2eaeb37a +size 92096 diff --git a/parse/dev/KmtVD97J43e/images/b72a43b98590942d3ac76f2a3028bc547f395f3ba5c9cef80d20255039b456eb.jpg b/parse/dev/KmtVD97J43e/images/b72a43b98590942d3ac76f2a3028bc547f395f3ba5c9cef80d20255039b456eb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..17e2b09fb524cfb24b1f3915ba43ec129ca84eb4 --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/b72a43b98590942d3ac76f2a3028bc547f395f3ba5c9cef80d20255039b456eb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:100aa0a07cef674c5e2ad01586deafe2379316641019385d51d1f699cc914137 +size 29522 diff --git a/parse/dev/KmtVD97J43e/images/c223b33a3bbb8b6a86ee9a5edcf0a622ca3a59bedbab90927725987b00ff77d5.jpg b/parse/dev/KmtVD97J43e/images/c223b33a3bbb8b6a86ee9a5edcf0a622ca3a59bedbab90927725987b00ff77d5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0f3f25a4f842b1fa0fd64004e04d10a8d5b88b7a --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/c223b33a3bbb8b6a86ee9a5edcf0a622ca3a59bedbab90927725987b00ff77d5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d682fcbcd8c2d56bd6ac03388afea232cc4d116d9fea35becb30fd996d085273 +size 36956 diff --git a/parse/dev/KmtVD97J43e/images/c591d21468f0c24bc5a777cd1caf3ce90ac7abfae426daf444df560839ef66ce.jpg b/parse/dev/KmtVD97J43e/images/c591d21468f0c24bc5a777cd1caf3ce90ac7abfae426daf444df560839ef66ce.jpg new file mode 100644 index 0000000000000000000000000000000000000000..50112291cfe852f9c755876a22b4112d64b33735 --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/c591d21468f0c24bc5a777cd1caf3ce90ac7abfae426daf444df560839ef66ce.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5be7ab10871c3cdc2f17eb036362e5b0815b69902cb23a60ab8c3d813df16267 +size 51799 diff --git a/parse/dev/KmtVD97J43e/images/e57b32d5be1fcd2ed9713ac93293f454c54dbf5f806c77b7ee054957c307ac88.jpg b/parse/dev/KmtVD97J43e/images/e57b32d5be1fcd2ed9713ac93293f454c54dbf5f806c77b7ee054957c307ac88.jpg new file mode 100644 index 0000000000000000000000000000000000000000..929b943ff22eb9776f8c7e7199d1c10aed95590f --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/e57b32d5be1fcd2ed9713ac93293f454c54dbf5f806c77b7ee054957c307ac88.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2709cbd82585304f0bbc23283afdb4ef6515a5231460f0cf9a6db919dbe535d5 +size 43998 diff --git a/parse/dev/KmtVD97J43e/images/ebec9a1e25eb9ffc4b92a1647e3d5fb91361ae4d818eef8b3c4e2bb409240a7a.jpg b/parse/dev/KmtVD97J43e/images/ebec9a1e25eb9ffc4b92a1647e3d5fb91361ae4d818eef8b3c4e2bb409240a7a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f47ed14e72c7f85604bd2f302624bd8b5f9103cf --- /dev/null +++ b/parse/dev/KmtVD97J43e/images/ebec9a1e25eb9ffc4b92a1647e3d5fb91361ae4d818eef8b3c4e2bb409240a7a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f3bc04f98d514f93f9dab48c1fe1f7ad60fb4e0dccc79df151c6e15c3bc8d56f +size 108964 diff --git a/parse/dev/LPB2BFZvncQ/images/104c3455809388dc48b4b7600bc47c5caae535c8475f2bb22250941a1b03720f.jpg b/parse/dev/LPB2BFZvncQ/images/104c3455809388dc48b4b7600bc47c5caae535c8475f2bb22250941a1b03720f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bd3545ede0a403f92d95548fb868e459c963dc87 --- /dev/null +++ b/parse/dev/LPB2BFZvncQ/images/104c3455809388dc48b4b7600bc47c5caae535c8475f2bb22250941a1b03720f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b20fee4687e261a01e4bd06604ba3c786198f9850119c4cd33b25fd31b96405 +size 3892 diff --git a/parse/dev/LPB2BFZvncQ/images/3a1fa4e53f0211633d44ae9bf0c5420e114e924b7bb7ba4c8d7d736f0a864718.jpg b/parse/dev/LPB2BFZvncQ/images/3a1fa4e53f0211633d44ae9bf0c5420e114e924b7bb7ba4c8d7d736f0a864718.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f0b5a0655a61f236a99356269fcd15798f57e876 --- /dev/null +++ b/parse/dev/LPB2BFZvncQ/images/3a1fa4e53f0211633d44ae9bf0c5420e114e924b7bb7ba4c8d7d736f0a864718.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:787e82bd67998edd8af08804c03200367ae90056e31379785d912d2f150212dc +size 10361 diff --git a/parse/dev/LPB2BFZvncQ/images/48b8fb0ce160fc82aa361cca703ec122c79d748d8c6d9903ebdebc73363a9987.jpg b/parse/dev/LPB2BFZvncQ/images/48b8fb0ce160fc82aa361cca703ec122c79d748d8c6d9903ebdebc73363a9987.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0c495818fc43856fd008dc2ddf551753d01f2a86 --- /dev/null +++ b/parse/dev/LPB2BFZvncQ/images/48b8fb0ce160fc82aa361cca703ec122c79d748d8c6d9903ebdebc73363a9987.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e366b143dd92df2fe31df67ef8797830a65e15d9948e4e5eb3599f5623d2429 +size 24520 diff --git a/parse/dev/LPB2BFZvncQ/images/4e3ed913335175c591c8c0b34ac958d9be0ae4ee0ef95ae2cc3543a1345868cf.jpg b/parse/dev/LPB2BFZvncQ/images/4e3ed913335175c591c8c0b34ac958d9be0ae4ee0ef95ae2cc3543a1345868cf.jpg new file mode 100644 index 0000000000000000000000000000000000000000..929c3e75fca499492669ab9d756c38e6dd97eda3 --- /dev/null +++ b/parse/dev/LPB2BFZvncQ/images/4e3ed913335175c591c8c0b34ac958d9be0ae4ee0ef95ae2cc3543a1345868cf.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a68958b171dbc14aba1fa06dd85ceb6664187acb8ad68d80b3dbac4101496214 +size 22161 diff --git a/parse/dev/LPB2BFZvncQ/images/7ca3fc2a2a10aab521320d0807aeedfce5ea2be7e3c84aac99efaad0b9f311cb.jpg b/parse/dev/LPB2BFZvncQ/images/7ca3fc2a2a10aab521320d0807aeedfce5ea2be7e3c84aac99efaad0b9f311cb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..aa7743bc22f009078aaf0d45918146a66d6f2d5d --- /dev/null +++ b/parse/dev/LPB2BFZvncQ/images/7ca3fc2a2a10aab521320d0807aeedfce5ea2be7e3c84aac99efaad0b9f311cb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c962eb6ee361b7af693b4b45339685fd8addff38be51f37ad5a175ccfc718fdd +size 8811 diff --git a/parse/dev/LPB2BFZvncQ/images/8d64805f1950d5782a44442c0372ce9fa1a9e550af3d592f374933fed4400b20.jpg b/parse/dev/LPB2BFZvncQ/images/8d64805f1950d5782a44442c0372ce9fa1a9e550af3d592f374933fed4400b20.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7f937f9af7dec30f6fe2726a9c744ba0f305d7ec --- /dev/null +++ b/parse/dev/LPB2BFZvncQ/images/8d64805f1950d5782a44442c0372ce9fa1a9e550af3d592f374933fed4400b20.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8356b4e4f31828cef25f257e2a6a7e87f4016a35b1144d4920bc3452369f7a10 +size 9044 diff --git a/parse/dev/LPB2BFZvncQ/images/dbc2588cc14f97bb030cc036099b48d5bcea429a883318e318f0718e8c2afb25.jpg b/parse/dev/LPB2BFZvncQ/images/dbc2588cc14f97bb030cc036099b48d5bcea429a883318e318f0718e8c2afb25.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5c2fbdbd902589728e8275f511c64a6d23e3b744 --- /dev/null +++ b/parse/dev/LPB2BFZvncQ/images/dbc2588cc14f97bb030cc036099b48d5bcea429a883318e318f0718e8c2afb25.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:716f4818e5b2e48dd86757b01c28df7f4453620e3ac9ba8f852da845281989bb +size 3948 diff --git a/parse/dev/LPB2BFZvncQ/images/eeb361f91698ad6334c7a4361ba3dd103b04cc61fd6a1f2820ea9dec7caf216b.jpg b/parse/dev/LPB2BFZvncQ/images/eeb361f91698ad6334c7a4361ba3dd103b04cc61fd6a1f2820ea9dec7caf216b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c74962da8d2d53e507596377131bce822c107e0b --- /dev/null +++ b/parse/dev/LPB2BFZvncQ/images/eeb361f91698ad6334c7a4361ba3dd103b04cc61fd6a1f2820ea9dec7caf216b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e439ff9f0fe05509603004965f84a55044a6b214dde6baf1a88127e85c298390 +size 14095 diff --git a/parse/dev/LQnyIk5dUA/images/08c769d5c3e4e25175d8858955d5fa1ccaa2ab322ebf32aef6e4353ff984e25c.jpg b/parse/dev/LQnyIk5dUA/images/08c769d5c3e4e25175d8858955d5fa1ccaa2ab322ebf32aef6e4353ff984e25c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..77786286ea32975a558e46d9499bfad4122ceede --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/08c769d5c3e4e25175d8858955d5fa1ccaa2ab322ebf32aef6e4353ff984e25c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6aea04426d66f83f14a9a74c5cedecd095de1987f1b37ea3cb4e22e1661cc28 +size 16686 diff --git a/parse/dev/LQnyIk5dUA/images/0aa0deb18087b00ef43211082b1d062786c500a05712f6a7261f30bcde4e6f1b.jpg b/parse/dev/LQnyIk5dUA/images/0aa0deb18087b00ef43211082b1d062786c500a05712f6a7261f30bcde4e6f1b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9f11d2810e18571ec35df6fd5b868015c4d2ba94 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/0aa0deb18087b00ef43211082b1d062786c500a05712f6a7261f30bcde4e6f1b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f51910cdabd3131d9addda98ceb69fb227b23a8c992a0c39a3572d691285e504 +size 16403 diff --git a/parse/dev/LQnyIk5dUA/images/0d70cb80a608a5482bfe7f7cedce750b7a0be5535139c348a9f5f8cf10fa3686.jpg b/parse/dev/LQnyIk5dUA/images/0d70cb80a608a5482bfe7f7cedce750b7a0be5535139c348a9f5f8cf10fa3686.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c1550a57f622966f5160e8431f63312af0c89919 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/0d70cb80a608a5482bfe7f7cedce750b7a0be5535139c348a9f5f8cf10fa3686.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0837a3bc457e0f73fe1d45aa1d459eb92e54f7125bcb3758208a23f680e5b167 +size 3101 diff --git a/parse/dev/LQnyIk5dUA/images/164c0924dea60f7ab0870f03f1694765adec8436f5c64d1841cc07608923f78d.jpg b/parse/dev/LQnyIk5dUA/images/164c0924dea60f7ab0870f03f1694765adec8436f5c64d1841cc07608923f78d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7a79f8d08429b086a63a6a7b20f45bb0f3bddaaa --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/164c0924dea60f7ab0870f03f1694765adec8436f5c64d1841cc07608923f78d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d47c60713437f9eab74290cbf5f22f0cd48359fd9e3aacb97723969c1e0c24b +size 58863 diff --git a/parse/dev/LQnyIk5dUA/images/1aac1f87ffa84ece8cdf586df64cd62620cc071130e150a925b0530a3028e9eb.jpg b/parse/dev/LQnyIk5dUA/images/1aac1f87ffa84ece8cdf586df64cd62620cc071130e150a925b0530a3028e9eb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f3bd8f102babcf8a6a51a298b624141020f9253a --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/1aac1f87ffa84ece8cdf586df64cd62620cc071130e150a925b0530a3028e9eb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:21308211580eb100a978e0bdb79d5a34ece6d2c5260338a4d702ba2d701d93eb +size 10390 diff --git a/parse/dev/LQnyIk5dUA/images/1c39c4f1e861b4dc40fddb3f600409a03025e9f2bac16d9b137b418ac455fb62.jpg b/parse/dev/LQnyIk5dUA/images/1c39c4f1e861b4dc40fddb3f600409a03025e9f2bac16d9b137b418ac455fb62.jpg new file mode 100644 index 0000000000000000000000000000000000000000..437ceb2db51b1ab3ff805419270e200c17218cc3 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/1c39c4f1e861b4dc40fddb3f600409a03025e9f2bac16d9b137b418ac455fb62.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2cb40e3992517a9b3f18abbeae2a2f6ea15b59ea41cf7ca562f605fe044c00a8 +size 158230 diff --git a/parse/dev/LQnyIk5dUA/images/1e01f841e08ed52886f2cf7faf6216a65b7df0e8ef8e17d87b511958c9afc13e.jpg b/parse/dev/LQnyIk5dUA/images/1e01f841e08ed52886f2cf7faf6216a65b7df0e8ef8e17d87b511958c9afc13e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f7e72e21904fd5a1696c72d91b1c8bf44747a99f --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/1e01f841e08ed52886f2cf7faf6216a65b7df0e8ef8e17d87b511958c9afc13e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a837699817224081cf7e8e97444c9cd94d48cc4cb60f262ad38653618052e00 +size 5567 diff --git a/parse/dev/LQnyIk5dUA/images/1e4f78c1a09e7482ff156cf62e9cd0f0ba75ce8272db5e8bbd5af1083fa9e52c.jpg b/parse/dev/LQnyIk5dUA/images/1e4f78c1a09e7482ff156cf62e9cd0f0ba75ce8272db5e8bbd5af1083fa9e52c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e6b5c26c47625a68d3a8aeb6a7f8ab7430a4d3ad --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/1e4f78c1a09e7482ff156cf62e9cd0f0ba75ce8272db5e8bbd5af1083fa9e52c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c8139b49622e54f877c06e7c72b4418ee5e31b69b65ea52dbd2931b9f0929762 +size 12236 diff --git a/parse/dev/LQnyIk5dUA/images/223cd9693cd4c22cfa62466403639f3d2573259f8730e12f1b2db503a456a1d8.jpg b/parse/dev/LQnyIk5dUA/images/223cd9693cd4c22cfa62466403639f3d2573259f8730e12f1b2db503a456a1d8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a87a0bf333c8827106263e1f81438831b2f88ef1 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/223cd9693cd4c22cfa62466403639f3d2573259f8730e12f1b2db503a456a1d8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:903080ad02ec281e667182961368900261ad5a430f531854b330ca27aab4cbaf +size 14324 diff --git a/parse/dev/LQnyIk5dUA/images/251df1c2cf8694eeb416499f62dd6a6a7ef357aae7cdf9e288020262f2aa3688.jpg b/parse/dev/LQnyIk5dUA/images/251df1c2cf8694eeb416499f62dd6a6a7ef357aae7cdf9e288020262f2aa3688.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4b784d1ebb2a447409d0a635507d0ddd93fe60e0 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/251df1c2cf8694eeb416499f62dd6a6a7ef357aae7cdf9e288020262f2aa3688.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a0188292e6044752db2bab05c8b249a39b5f8218b649557bc63f48e173887d0 +size 10275 diff --git a/parse/dev/LQnyIk5dUA/images/29c53e7702c6ccf47f9e4e0e697e4ac8c2f722561714879982bb183c55f08f40.jpg b/parse/dev/LQnyIk5dUA/images/29c53e7702c6ccf47f9e4e0e697e4ac8c2f722561714879982bb183c55f08f40.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fc92a0e70b712ee90445af8ff60f7a752188675f --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/29c53e7702c6ccf47f9e4e0e697e4ac8c2f722561714879982bb183c55f08f40.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d94215af9fe9ed7aa411fd2423b74dfd56e15ecd216509265908d51fe8ddc73f +size 11532 diff --git a/parse/dev/LQnyIk5dUA/images/2bb326e534fb3e0c2b9efa7b28d457d765f4e5292a04b28f2f5711faf0606874.jpg b/parse/dev/LQnyIk5dUA/images/2bb326e534fb3e0c2b9efa7b28d457d765f4e5292a04b28f2f5711faf0606874.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f8b9ee6a2c0935a7f7216ce2ddb33141288b7cf6 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/2bb326e534fb3e0c2b9efa7b28d457d765f4e5292a04b28f2f5711faf0606874.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b56288dcca380ab7b268197a59c8450bb91a0e4b0680942d4cf8b4258a6c88d9 +size 121580 diff --git a/parse/dev/LQnyIk5dUA/images/2ffe92efbdc1be61a767009410f74cfdeac5635d9600b768fe9c69287297121b.jpg b/parse/dev/LQnyIk5dUA/images/2ffe92efbdc1be61a767009410f74cfdeac5635d9600b768fe9c69287297121b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8a4730010e2e48444d98e52a3be795358e9f33b2 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/2ffe92efbdc1be61a767009410f74cfdeac5635d9600b768fe9c69287297121b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75c7a917de33669d933fa120c751b9c6138aa8dca2fbe8bdda3e941c9e419263 +size 13666 diff --git a/parse/dev/LQnyIk5dUA/images/3636b62f6ec294cd3470e68b648e38c74565996d60cf1a23fb78ce5ecb1865b4.jpg b/parse/dev/LQnyIk5dUA/images/3636b62f6ec294cd3470e68b648e38c74565996d60cf1a23fb78ce5ecb1865b4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d5eb3eb2affb81ce5cf85b1cf57dd09b7acdfad9 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/3636b62f6ec294cd3470e68b648e38c74565996d60cf1a23fb78ce5ecb1865b4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ddb536e3f94d56970c2fa6a2f6cc417181676628fb1b6f8327fe7499ba0f1331 +size 75475 diff --git a/parse/dev/LQnyIk5dUA/images/389757cd4fbdce55690ea7ebba36988c56e82becf66e1d71e1e3e7ce458027b1.jpg b/parse/dev/LQnyIk5dUA/images/389757cd4fbdce55690ea7ebba36988c56e82becf66e1d71e1e3e7ce458027b1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..70f5b7aecd5f13722ede6a1084145bc7d60c7f90 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/389757cd4fbdce55690ea7ebba36988c56e82becf66e1d71e1e3e7ce458027b1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d23a4946ba7df8665195b41d7d7b1c96faf011318804ee22a1a89f4ab2fe3243 +size 34722 diff --git a/parse/dev/LQnyIk5dUA/images/39078fb1972e975828883012737505da03f6a728859142b60f7403c6b7c9e073.jpg b/parse/dev/LQnyIk5dUA/images/39078fb1972e975828883012737505da03f6a728859142b60f7403c6b7c9e073.jpg new file mode 100644 index 0000000000000000000000000000000000000000..07da2e95c5a603f4d513707cd050d43aa9039d01 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/39078fb1972e975828883012737505da03f6a728859142b60f7403c6b7c9e073.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:273e4ba2f726c15e975c051cbf6e5c679fce1cd5c2561642568bfce339277bc5 +size 36088 diff --git a/parse/dev/LQnyIk5dUA/images/397aefd68fc3ee61c4b832e2d8b0a661feae38d9a3128f51f6dffda5d97ec04a.jpg b/parse/dev/LQnyIk5dUA/images/397aefd68fc3ee61c4b832e2d8b0a661feae38d9a3128f51f6dffda5d97ec04a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8785c34522560fd206da2ed51dd37c53bb19324e --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/397aefd68fc3ee61c4b832e2d8b0a661feae38d9a3128f51f6dffda5d97ec04a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a195cc79c302438603820a20a49743517850569332dc1fc98603c233572bdd0a +size 11640 diff --git a/parse/dev/LQnyIk5dUA/images/39bba2863cf99c9e5328b96363d2476531a733b5dac6b49d4f07e7a78a1d0dc1.jpg b/parse/dev/LQnyIk5dUA/images/39bba2863cf99c9e5328b96363d2476531a733b5dac6b49d4f07e7a78a1d0dc1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9067938e2e7877b19690d9d7d1349e239c1a3653 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/39bba2863cf99c9e5328b96363d2476531a733b5dac6b49d4f07e7a78a1d0dc1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:365c6c667a962c002c9e06912a4b3dcf3dd60128c4c4a22857334ed1c09d39b4 +size 12825 diff --git a/parse/dev/LQnyIk5dUA/images/3f2b65358b19279f4b8a4b87fe7cc4f26cf9949d18950c53b18069e1bf0bef51.jpg b/parse/dev/LQnyIk5dUA/images/3f2b65358b19279f4b8a4b87fe7cc4f26cf9949d18950c53b18069e1bf0bef51.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ecd6c5df59b4269780625fd5759dd39757f746f2 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/3f2b65358b19279f4b8a4b87fe7cc4f26cf9949d18950c53b18069e1bf0bef51.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82129b973334491907f63b8a48ffb0ac6a6449cf3cea20f31f0d8d76b2083dc8 +size 5876 diff --git a/parse/dev/LQnyIk5dUA/images/403394957383b71b8d2a034451ea8f38ba58a8c4a241d22b1e56e4761df4deaa.jpg b/parse/dev/LQnyIk5dUA/images/403394957383b71b8d2a034451ea8f38ba58a8c4a241d22b1e56e4761df4deaa.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9acf02d0608cb22681ff849759734e9d663e1b73 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/403394957383b71b8d2a034451ea8f38ba58a8c4a241d22b1e56e4761df4deaa.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc51d1985dd8be99624b39ff00edcd9da63bd087213047222e520ca4a1570a08 +size 22539 diff --git a/parse/dev/LQnyIk5dUA/images/4066b5d6b745d2ff9575e127a3ebb731dc81e1020753ecd07cdf5e279f5e52b1.jpg b/parse/dev/LQnyIk5dUA/images/4066b5d6b745d2ff9575e127a3ebb731dc81e1020753ecd07cdf5e279f5e52b1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a792776958a97bb2160fbd747dd2e73c2c415a4e --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/4066b5d6b745d2ff9575e127a3ebb731dc81e1020753ecd07cdf5e279f5e52b1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:108b8c3fc8349e83ae1c2f6ad19ffc7e976e82ef6e733e3fc2d5693492aa36f6 +size 36340 diff --git a/parse/dev/LQnyIk5dUA/images/4716da73f61abd1d0fa91e712a07cee64efde5074b9b361c45ccb893fe405e50.jpg b/parse/dev/LQnyIk5dUA/images/4716da73f61abd1d0fa91e712a07cee64efde5074b9b361c45ccb893fe405e50.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1bc4a5599c330a9b53c733890bb5dcdc0a6bb307 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/4716da73f61abd1d0fa91e712a07cee64efde5074b9b361c45ccb893fe405e50.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cebf715c8be8db4fe1da4dbb086c384034219fb0b317be9be8d788fbd4225f44 +size 30148 diff --git a/parse/dev/LQnyIk5dUA/images/4c5ea7871eb1dbed96f9746f3489964b61031ceb90687f6f87482fdd496159f2.jpg b/parse/dev/LQnyIk5dUA/images/4c5ea7871eb1dbed96f9746f3489964b61031ceb90687f6f87482fdd496159f2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6f0a7d67dd6dd2f61ec153186b4f3c3887fc32a7 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/4c5ea7871eb1dbed96f9746f3489964b61031ceb90687f6f87482fdd496159f2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f92af22d8eab6e2d38c781585ca715f7652394eb8c5ffdc558b1fff7d2575dcb +size 7961 diff --git a/parse/dev/LQnyIk5dUA/images/4ca0e6e0e5c04282cc2e96e3b3cb1eb1985e6e109c16cc1e953925b49cad0f52.jpg b/parse/dev/LQnyIk5dUA/images/4ca0e6e0e5c04282cc2e96e3b3cb1eb1985e6e109c16cc1e953925b49cad0f52.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2fff7be57f56133b1d32efa95ab50bc7eb607fb6 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/4ca0e6e0e5c04282cc2e96e3b3cb1eb1985e6e109c16cc1e953925b49cad0f52.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:36193e9a5a95344d6342d06c48b18c3fe2ea7ae835611e78622e71c51ceca5b0 +size 24446 diff --git a/parse/dev/LQnyIk5dUA/images/4de87cf819e5642928e45b0db51fd2a551b8ba45636d57bf2d3a458a70837763.jpg b/parse/dev/LQnyIk5dUA/images/4de87cf819e5642928e45b0db51fd2a551b8ba45636d57bf2d3a458a70837763.jpg new file mode 100644 index 0000000000000000000000000000000000000000..02be3447e81a29abe68c8e7d783a70f5e225ccaa --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/4de87cf819e5642928e45b0db51fd2a551b8ba45636d57bf2d3a458a70837763.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1a59e30499d92ad7538e10f7554fe22e521c86c46f6df20524288124eac57a5 +size 5296 diff --git a/parse/dev/LQnyIk5dUA/images/6185cdb95ed14484d35c1adf30a463c7a3e731740ed3f1008bcef8aa9640f873.jpg b/parse/dev/LQnyIk5dUA/images/6185cdb95ed14484d35c1adf30a463c7a3e731740ed3f1008bcef8aa9640f873.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b8c65ed50175268df5ec882c33129b637ee331d0 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/6185cdb95ed14484d35c1adf30a463c7a3e731740ed3f1008bcef8aa9640f873.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6bc0503a23c36e3504c832b684fa2958bb243eff515264e0fce03dec86ef2ab8 +size 6470 diff --git a/parse/dev/LQnyIk5dUA/images/6c47a32becb32575e6ef3a7e39125ea18ab189f881a667fe5535b2a8bf885b37.jpg b/parse/dev/LQnyIk5dUA/images/6c47a32becb32575e6ef3a7e39125ea18ab189f881a667fe5535b2a8bf885b37.jpg new file mode 100644 index 0000000000000000000000000000000000000000..aefa2abfc0bd4cd9dd643479a9c149c557df62bd --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/6c47a32becb32575e6ef3a7e39125ea18ab189f881a667fe5535b2a8bf885b37.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b925bf13d4375f5dd075931d844f3b744da8987d19d8bd796248e969181b4c1c +size 3191 diff --git a/parse/dev/LQnyIk5dUA/images/70436b1fd9418a6a4e68a2a0da9aea28a8068f78ea438fabcf2f5acc80cc9e7d.jpg b/parse/dev/LQnyIk5dUA/images/70436b1fd9418a6a4e68a2a0da9aea28a8068f78ea438fabcf2f5acc80cc9e7d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e5da1f094b57544e1a836b9e329309ff10ce290e --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/70436b1fd9418a6a4e68a2a0da9aea28a8068f78ea438fabcf2f5acc80cc9e7d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a7b20792e79a554956a5bdd391d389382bcc8adaf74b6db37ec7f0f3a41d6e00 +size 5706 diff --git a/parse/dev/LQnyIk5dUA/images/7554ee90dd61cc62864a7974651c8f93e2699cb219870e6b20ecf30e2d0b4f6b.jpg b/parse/dev/LQnyIk5dUA/images/7554ee90dd61cc62864a7974651c8f93e2699cb219870e6b20ecf30e2d0b4f6b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7a134aad5cc9a0b445ebdf6d30a275d4d34d81d8 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/7554ee90dd61cc62864a7974651c8f93e2699cb219870e6b20ecf30e2d0b4f6b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ac38ffa777e10744cf2070d64495227436a0acbe8075f3efc17854d84c50767 +size 3082 diff --git a/parse/dev/LQnyIk5dUA/images/76f68ae788eae7354faba959066a76b1c5e09e6ff501340b18ea02093c70d3fd.jpg b/parse/dev/LQnyIk5dUA/images/76f68ae788eae7354faba959066a76b1c5e09e6ff501340b18ea02093c70d3fd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0ffa41db21eb38695fe88d3986c09f3505b4063b --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/76f68ae788eae7354faba959066a76b1c5e09e6ff501340b18ea02093c70d3fd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df2f5fbe3c5f6970d184e7cbb58deb570ca2aebd4f6d4bd2ca630cdd9c4603bd +size 7462 diff --git a/parse/dev/LQnyIk5dUA/images/796ace5e15ea150faf79e097d9eac7404c51eb8756f6b67b01d5ce56d85c644f.jpg b/parse/dev/LQnyIk5dUA/images/796ace5e15ea150faf79e097d9eac7404c51eb8756f6b67b01d5ce56d85c644f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0d38e1095d246067bebaccd6b2994bb563ba23b3 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/796ace5e15ea150faf79e097d9eac7404c51eb8756f6b67b01d5ce56d85c644f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1cc6223cb909de4fa9d9cfa4b0f3dad1f69abbfc40fff4ddf2fda60d4f61dec +size 5572 diff --git a/parse/dev/LQnyIk5dUA/images/7c3af118f67073825cfd1b6be40cce284508327b7214e0aefdb27b01f205dd96.jpg b/parse/dev/LQnyIk5dUA/images/7c3af118f67073825cfd1b6be40cce284508327b7214e0aefdb27b01f205dd96.jpg new file mode 100644 index 0000000000000000000000000000000000000000..00cae2ad2558a7cffabe79c8d15b00e4da827fdd --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/7c3af118f67073825cfd1b6be40cce284508327b7214e0aefdb27b01f205dd96.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:293e72567d23a703a14152189a346ee7fc3fefa39a8e03b360c499878cc5ca26 +size 20012 diff --git a/parse/dev/LQnyIk5dUA/images/8898c7d2794eb8e47e1e5a75f501a2acb27c9f1065a3b83baf3d0308f1c53833.jpg b/parse/dev/LQnyIk5dUA/images/8898c7d2794eb8e47e1e5a75f501a2acb27c9f1065a3b83baf3d0308f1c53833.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d33b533f131f98bb1c387ec2621452cacfe4dac0 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/8898c7d2794eb8e47e1e5a75f501a2acb27c9f1065a3b83baf3d0308f1c53833.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df91bc75caba8f995e35860262674bda079c8ef6842d1c4b0a79f830badb5a29 +size 18814 diff --git a/parse/dev/LQnyIk5dUA/images/8d680272239d06dffd9edc69a58e6e7e20aaaf358c0454b53c2756eb5cc47b09.jpg b/parse/dev/LQnyIk5dUA/images/8d680272239d06dffd9edc69a58e6e7e20aaaf358c0454b53c2756eb5cc47b09.jpg new file mode 100644 index 0000000000000000000000000000000000000000..00450e55da1a03bb44cf7f4b34c85f34039a505a --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/8d680272239d06dffd9edc69a58e6e7e20aaaf358c0454b53c2756eb5cc47b09.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e7c8dfb6e73fdbc1dd492b905aaa0126f19897008536360ddd9d49c6b449172 +size 16700 diff --git a/parse/dev/LQnyIk5dUA/images/8eb8d35541718722baa2771b01b4a636371508696b11b66b537aa55d210634c6.jpg b/parse/dev/LQnyIk5dUA/images/8eb8d35541718722baa2771b01b4a636371508696b11b66b537aa55d210634c6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a843a1048e01cd4b7b7eb2fd28f8ab257cb49a9b --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/8eb8d35541718722baa2771b01b4a636371508696b11b66b537aa55d210634c6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aeeb4198cc001c1fc04afd75d5e9e3292a7f23b9a00cff5f937324e6f3e49347 +size 20420 diff --git a/parse/dev/LQnyIk5dUA/images/8ed2db92a5953326e4a412c6ecc085a4024c2b8a90fafc25d2ba0e87ae262143.jpg b/parse/dev/LQnyIk5dUA/images/8ed2db92a5953326e4a412c6ecc085a4024c2b8a90fafc25d2ba0e87ae262143.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d19ed9d10ad7a22350a387b9d9660e3e22e276cb --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/8ed2db92a5953326e4a412c6ecc085a4024c2b8a90fafc25d2ba0e87ae262143.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6ebd4ac16978584c64e89b266a89044e7460f44cb80c2fcd29377eddfd6e2c7 +size 33455 diff --git a/parse/dev/LQnyIk5dUA/images/948f44466566a0bb7ce10c7485d29bc72dd552ad7e648f2b750bf2660732b195.jpg b/parse/dev/LQnyIk5dUA/images/948f44466566a0bb7ce10c7485d29bc72dd552ad7e648f2b750bf2660732b195.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d6eb5ecb789d0fd80765198428830c15b6323b9c --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/948f44466566a0bb7ce10c7485d29bc72dd552ad7e648f2b750bf2660732b195.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2ae6f9d0f42ea9a7ac2597908968e964a4d7e37519bd62040991d20c5edfe26f +size 117371 diff --git a/parse/dev/LQnyIk5dUA/images/960af1c88a741271974cd61d7910ab3517c1d1de4b6110fe48c6981b23b20125.jpg b/parse/dev/LQnyIk5dUA/images/960af1c88a741271974cd61d7910ab3517c1d1de4b6110fe48c6981b23b20125.jpg new file mode 100644 index 0000000000000000000000000000000000000000..03c4fc1bf8f64221c10c1771a218ff2884952164 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/960af1c88a741271974cd61d7910ab3517c1d1de4b6110fe48c6981b23b20125.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5f264abfeca24830d5f221083ea1d622e5038228a574b5a6b293eaeb0dbf260 +size 20963 diff --git a/parse/dev/LQnyIk5dUA/images/97156199e2d57c76ecea4e60b5015b0587132e9cc1da2b8c99cec60ad2ea4926.jpg b/parse/dev/LQnyIk5dUA/images/97156199e2d57c76ecea4e60b5015b0587132e9cc1da2b8c99cec60ad2ea4926.jpg new file mode 100644 index 0000000000000000000000000000000000000000..111b8eefe2beb4b5b47fdf79a6f30d814498dadf --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/97156199e2d57c76ecea4e60b5015b0587132e9cc1da2b8c99cec60ad2ea4926.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0fc364ea80a8d97c25c1743d3158db4f4096f9eedb6be8cb72842d3b401e3822 +size 26577 diff --git a/parse/dev/LQnyIk5dUA/images/9e26e41a9faa05cb706e504f046ace736e07c3f15efa9ee54b625ffd408bed0b.jpg b/parse/dev/LQnyIk5dUA/images/9e26e41a9faa05cb706e504f046ace736e07c3f15efa9ee54b625ffd408bed0b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3501340efb3aa008d7b14266f73878f96e6de376 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/9e26e41a9faa05cb706e504f046ace736e07c3f15efa9ee54b625ffd408bed0b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a39fad5a820af344577e16c9c423d9bf87fd55c15aa16ec5cedfc74046cb3a33 +size 14754 diff --git a/parse/dev/LQnyIk5dUA/images/a0489fa0b0eb21c094d6884b8ee5e23a04844e26a7e785f05c8e224f8555fc98.jpg b/parse/dev/LQnyIk5dUA/images/a0489fa0b0eb21c094d6884b8ee5e23a04844e26a7e785f05c8e224f8555fc98.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1bbc8e91edc062d7dda5ced7f69dfd9e55b91ba8 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/a0489fa0b0eb21c094d6884b8ee5e23a04844e26a7e785f05c8e224f8555fc98.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d148dde261341f480c81c71838a3d50da13663eff795c8b70fc190f0cf2eedd2 +size 5850 diff --git a/parse/dev/LQnyIk5dUA/images/a0f59fc5039f870822da8c02bff34bf7c65367c4dd8c6a235db00cca7b5a63d4.jpg b/parse/dev/LQnyIk5dUA/images/a0f59fc5039f870822da8c02bff34bf7c65367c4dd8c6a235db00cca7b5a63d4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ebb1af24a97820913c0b663e56a6217f2adb560d --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/a0f59fc5039f870822da8c02bff34bf7c65367c4dd8c6a235db00cca7b5a63d4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:919edba048a1e66af02163bc409cc48221bd6065836072f94f15f23910286fae +size 16318 diff --git a/parse/dev/LQnyIk5dUA/images/a136a95d8e79308a42f9e6137d729e64bfaaf5eb86baf0cdd82451fdd3e21534.jpg b/parse/dev/LQnyIk5dUA/images/a136a95d8e79308a42f9e6137d729e64bfaaf5eb86baf0cdd82451fdd3e21534.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ebb8bfa5850a5bcc44a08362555648327770b695 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/a136a95d8e79308a42f9e6137d729e64bfaaf5eb86baf0cdd82451fdd3e21534.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d80571b9fb807d3c4c99b954e6e024786b7d26be0d9fb01e079a336f217d516 +size 7308 diff --git a/parse/dev/LQnyIk5dUA/images/a1ca7bcbf90b8a8996a0d5f6722d9fab9e3f145a3fa42d2ec2c6f3b260011c6b.jpg b/parse/dev/LQnyIk5dUA/images/a1ca7bcbf90b8a8996a0d5f6722d9fab9e3f145a3fa42d2ec2c6f3b260011c6b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2b0ec2a26ed61f9f04589d80b9700cd643d4476f --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/a1ca7bcbf90b8a8996a0d5f6722d9fab9e3f145a3fa42d2ec2c6f3b260011c6b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:95fa387d81be521af9af0a1d210a1c3c57c5d9f2ed457c109d7d73d8b3e55ac6 +size 7552 diff --git a/parse/dev/LQnyIk5dUA/images/a42bbf66e6650c3d64d9f28886ef5779f5f3d5c2795d14b0cb20a6785b42b4d0.jpg b/parse/dev/LQnyIk5dUA/images/a42bbf66e6650c3d64d9f28886ef5779f5f3d5c2795d14b0cb20a6785b42b4d0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..86a567fe4f69520dadd68e15e536459d5f3764c4 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/a42bbf66e6650c3d64d9f28886ef5779f5f3d5c2795d14b0cb20a6785b42b4d0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:63641c50c60fe6e611cd4bfd84665068c068bddbab87cb681de9db7893c72832 +size 17186 diff --git a/parse/dev/LQnyIk5dUA/images/a47bb1df66864a8a14d2cd5498e8c641aa73dd95c0de6cdee6c4f6aadad70621.jpg b/parse/dev/LQnyIk5dUA/images/a47bb1df66864a8a14d2cd5498e8c641aa73dd95c0de6cdee6c4f6aadad70621.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a40c52b7270a02116dbb668b38cc72ed76bba938 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/a47bb1df66864a8a14d2cd5498e8c641aa73dd95c0de6cdee6c4f6aadad70621.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b38fa84e349c72adae0bbf8843c366d49b7b87556f6647a5556c3c9e77457c5 +size 41652 diff --git a/parse/dev/LQnyIk5dUA/images/a8bb86b7ede95e9d76b146a8cc113dad41a8c8bae5c2b723c61fb526655c5557.jpg b/parse/dev/LQnyIk5dUA/images/a8bb86b7ede95e9d76b146a8cc113dad41a8c8bae5c2b723c61fb526655c5557.jpg new file mode 100644 index 0000000000000000000000000000000000000000..af7f4c6b66c9b742a25825a890fafe8e448e7a4c --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/a8bb86b7ede95e9d76b146a8cc113dad41a8c8bae5c2b723c61fb526655c5557.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b507fd055e55b42c204f697a875d2bb24c26f72e5d1a02c36b06e829fe1fe40a +size 39254 diff --git a/parse/dev/LQnyIk5dUA/images/ab0cba7b0b3c7fef1359b78ee3092393cc4b4685c18ecb821694d937242aace8.jpg b/parse/dev/LQnyIk5dUA/images/ab0cba7b0b3c7fef1359b78ee3092393cc4b4685c18ecb821694d937242aace8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..60e25597484e2d42757f70fb09bac13d6e314825 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/ab0cba7b0b3c7fef1359b78ee3092393cc4b4685c18ecb821694d937242aace8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1f53975e604f1c9624d9973bef6fb3eb4fb92004c9d883ff599f69e16ef1ef1 +size 4969 diff --git a/parse/dev/LQnyIk5dUA/images/b2ea66bf22c33b0ecad117a25cc56da678badb076ee25e462a58e2dfcaa607d3.jpg b/parse/dev/LQnyIk5dUA/images/b2ea66bf22c33b0ecad117a25cc56da678badb076ee25e462a58e2dfcaa607d3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..350f0678ab76aa51859a4194caa807a676b40965 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/b2ea66bf22c33b0ecad117a25cc56da678badb076ee25e462a58e2dfcaa607d3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16cfc613197840b5b1d1eead7f366af6dee68825f41c9b548a1048f7bc70a3a1 +size 5083 diff --git a/parse/dev/LQnyIk5dUA/images/b568b7c129850fef19c39a3232d40411ec0221dd589e2bff4af81755ba283407.jpg b/parse/dev/LQnyIk5dUA/images/b568b7c129850fef19c39a3232d40411ec0221dd589e2bff4af81755ba283407.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a5706d9c376c3587f1e2b22f1a25e3b36a3d5540 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/b568b7c129850fef19c39a3232d40411ec0221dd589e2bff4af81755ba283407.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b4026073e4d6af7291c1ab3b3e3e2284cd9e1577080ef65eaa207024306e57b +size 6453 diff --git a/parse/dev/LQnyIk5dUA/images/b626cdf9497ec754c1215cb56aa8c10f3a26dc2165e659357138c3cce2ebc992.jpg b/parse/dev/LQnyIk5dUA/images/b626cdf9497ec754c1215cb56aa8c10f3a26dc2165e659357138c3cce2ebc992.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2f0ea29faaab17ce6bb90e96aad93a9004efd6bf --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/b626cdf9497ec754c1215cb56aa8c10f3a26dc2165e659357138c3cce2ebc992.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:523b0467ca1875bb83346ba81654ca3095c3fc37d1089a606e26493df85dff66 +size 16749 diff --git a/parse/dev/LQnyIk5dUA/images/ba389d0248097d205d6d1152de40cbf0746d8340b89ea8bfa6ab001ca9cd7e15.jpg b/parse/dev/LQnyIk5dUA/images/ba389d0248097d205d6d1152de40cbf0746d8340b89ea8bfa6ab001ca9cd7e15.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6c1c76139b652c4fe37a0c2a65b252fdafe36cf1 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/ba389d0248097d205d6d1152de40cbf0746d8340b89ea8bfa6ab001ca9cd7e15.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e02bd2dca1f7cc2cc680bb27ebb4888992f3edf6d502f9c8c9d3a18ad8948fa5 +size 67184 diff --git a/parse/dev/LQnyIk5dUA/images/c10619b27f2424bb49255913f0aa6322b96a3c848ae276a4a7a017307684d7d7.jpg b/parse/dev/LQnyIk5dUA/images/c10619b27f2424bb49255913f0aa6322b96a3c848ae276a4a7a017307684d7d7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..64ba9ee07d83440368396fbd9c69dd65c8b75656 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/c10619b27f2424bb49255913f0aa6322b96a3c848ae276a4a7a017307684d7d7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39c219124963238f944679daa469a51a0711ad23204430176dc6903a6f86649b +size 36685 diff --git a/parse/dev/LQnyIk5dUA/images/c2e88fb738e5ad642beda99170adcd929c08310c0b9e3cb0f9aef858c4a0cb77.jpg b/parse/dev/LQnyIk5dUA/images/c2e88fb738e5ad642beda99170adcd929c08310c0b9e3cb0f9aef858c4a0cb77.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4b371fa7fefa0ac1698aef4ab160a53af30f344b --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/c2e88fb738e5ad642beda99170adcd929c08310c0b9e3cb0f9aef858c4a0cb77.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b164128c32bd5ae289697f3a428bc1d68a1908d2ffe1e38995d0f4c7329c4c39 +size 9376 diff --git a/parse/dev/LQnyIk5dUA/images/ca53b9fc32750bfdd57fe3ebab08fa9ea8653f9714c12fd8690fc57779e96e6a.jpg b/parse/dev/LQnyIk5dUA/images/ca53b9fc32750bfdd57fe3ebab08fa9ea8653f9714c12fd8690fc57779e96e6a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5b4b19d0142d310135a337980b7bc059b740fdbe --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/ca53b9fc32750bfdd57fe3ebab08fa9ea8653f9714c12fd8690fc57779e96e6a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:731eb501da60eabdfc3630dcda0876d643827a3a9c69af68c7395bd5b45eb48f +size 3624 diff --git a/parse/dev/LQnyIk5dUA/images/ca594cbac14a6dadc0a0356814ce5f53ca9086402865e579d0265954d746dda1.jpg b/parse/dev/LQnyIk5dUA/images/ca594cbac14a6dadc0a0356814ce5f53ca9086402865e579d0265954d746dda1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5142a62eb669d695380ce1859dbe72048fed4070 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/ca594cbac14a6dadc0a0356814ce5f53ca9086402865e579d0265954d746dda1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bc0a3218548cf6ec0541d123de037aa75d4ed71aedb8a525353ff1fd825e0f73 +size 25354 diff --git a/parse/dev/LQnyIk5dUA/images/cb8777ef1ac8012d35e6cf8fb5346d4ebe99690f5af6f43875b3421c0cd5f49a.jpg b/parse/dev/LQnyIk5dUA/images/cb8777ef1ac8012d35e6cf8fb5346d4ebe99690f5af6f43875b3421c0cd5f49a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0ffb7e53f01e4855fd03ce30e4fb57995f43cd23 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/cb8777ef1ac8012d35e6cf8fb5346d4ebe99690f5af6f43875b3421c0cd5f49a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a250ddc6476005014e9ffeaffd7f5fdc89e487364fe19a2b60ed6ab7316b113 +size 11764 diff --git a/parse/dev/LQnyIk5dUA/images/d38b31c5630090223290d649358e2582390ba72e2cd35ecddcda21f9e149a498.jpg b/parse/dev/LQnyIk5dUA/images/d38b31c5630090223290d649358e2582390ba72e2cd35ecddcda21f9e149a498.jpg new file mode 100644 index 0000000000000000000000000000000000000000..92b65dfb5bbac61ec1f63b1cfb7d489af74abf9f --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/d38b31c5630090223290d649358e2582390ba72e2cd35ecddcda21f9e149a498.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86cea25de95acec71e6c78ab75ad5bf19d8aac7157945123b9d0f58dff6f5c1a +size 2306 diff --git a/parse/dev/LQnyIk5dUA/images/d8ad63311c02f52320a336fbdd7d44a7d0eb6d589ef03ca286c1c88d671379d3.jpg b/parse/dev/LQnyIk5dUA/images/d8ad63311c02f52320a336fbdd7d44a7d0eb6d589ef03ca286c1c88d671379d3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..44a9f2f7493a71703e04b556a3a10de1b1a3f538 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/d8ad63311c02f52320a336fbdd7d44a7d0eb6d589ef03ca286c1c88d671379d3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:296b89daa238b2023210e0041b8f8e0f03f37069581bf34498486ce7def04848 +size 7603 diff --git a/parse/dev/LQnyIk5dUA/images/d98835f6000c667c4b34dbb03972a5228120c77b6e16408bd485de39faf129c1.jpg b/parse/dev/LQnyIk5dUA/images/d98835f6000c667c4b34dbb03972a5228120c77b6e16408bd485de39faf129c1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..59d26678af1ac8aa97dc786b2224853cdc119390 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/d98835f6000c667c4b34dbb03972a5228120c77b6e16408bd485de39faf129c1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a31c415151bf0325d3c49dc76f2e83d3c1e34ba1d7c2d251d170fda1f9d4f8dd +size 30762 diff --git a/parse/dev/LQnyIk5dUA/images/f7e96f86a6539ec634fae583ac621b346c4e8d10185016c22c278203e271152f.jpg b/parse/dev/LQnyIk5dUA/images/f7e96f86a6539ec634fae583ac621b346c4e8d10185016c22c278203e271152f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..53a2dd5b2524d561cfd8aefc61e5b587fe7a7e16 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/f7e96f86a6539ec634fae583ac621b346c4e8d10185016c22c278203e271152f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e1aa3e854ebeb2301afaeee92071b7a10ee518bb06a2b6a761dd4d066a1ecff +size 14924 diff --git a/parse/dev/LQnyIk5dUA/images/fa339171cdf616df43322407a19bba7f6f5c9235bf6f78e36ff3e612b5c35272.jpg b/parse/dev/LQnyIk5dUA/images/fa339171cdf616df43322407a19bba7f6f5c9235bf6f78e36ff3e612b5c35272.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2ec725d8ecb472262167c35f78e9896ab21c1b62 --- /dev/null +++ b/parse/dev/LQnyIk5dUA/images/fa339171cdf616df43322407a19bba7f6f5c9235bf6f78e36ff3e612b5c35272.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89af99add7bbd957c989a05e5e00f3512d1f45771a62f04bd1f427fa4e17e469 +size 28126 diff --git a/parse/dev/Nayau9fwXU/images/13d2edd64d22266e53f75d6ec4786d90cfca2499da45f63fe4f9879be7344106.jpg b/parse/dev/Nayau9fwXU/images/13d2edd64d22266e53f75d6ec4786d90cfca2499da45f63fe4f9879be7344106.jpg new file mode 100644 index 0000000000000000000000000000000000000000..caaa49f51fb45ff245560c0ef7ec49b24bff6e4a --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/13d2edd64d22266e53f75d6ec4786d90cfca2499da45f63fe4f9879be7344106.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6dd6b240c4e4fb776797ab704f6c1211f9e4e09483a6fca7f042c5e13191aff6 +size 8245 diff --git a/parse/dev/Nayau9fwXU/images/322f063de18e5e1a1f15a68e8e35bc489133b6b7f256dd7ad1ff2b593c8b7187.jpg b/parse/dev/Nayau9fwXU/images/322f063de18e5e1a1f15a68e8e35bc489133b6b7f256dd7ad1ff2b593c8b7187.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f67a0ce5cf8e39b7ab4c4a89e0cd49a9cda2a10a --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/322f063de18e5e1a1f15a68e8e35bc489133b6b7f256dd7ad1ff2b593c8b7187.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5becf334b561f4068131337f2488bf51c32f1d2e2563cd4a0fd0423ac66e26ee +size 10127 diff --git a/parse/dev/Nayau9fwXU/images/4157a888c44a1d4ff5b6a67f8bf295987cb71bb2f262f64fde306c2dc1486c75.jpg b/parse/dev/Nayau9fwXU/images/4157a888c44a1d4ff5b6a67f8bf295987cb71bb2f262f64fde306c2dc1486c75.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1c864aa41fba85033bfc3b4ea3cf801bc9f715c6 --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/4157a888c44a1d4ff5b6a67f8bf295987cb71bb2f262f64fde306c2dc1486c75.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22dcec8cd844c26d367952a3adccc8876afece7ede68bea77e4a344e2d8ddee1 +size 5254 diff --git a/parse/dev/Nayau9fwXU/images/451034d747982fc238b6126f37297de04099cb9a6abee6457ef83feac91c56f6.jpg b/parse/dev/Nayau9fwXU/images/451034d747982fc238b6126f37297de04099cb9a6abee6457ef83feac91c56f6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a2f52c1e834afb4eb56b90e91d5a9e9797bab02e --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/451034d747982fc238b6126f37297de04099cb9a6abee6457ef83feac91c56f6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0bb8efa072e32fa0e03bd9a378bece743d235686f47219431827b546f5e2c95c +size 6724 diff --git a/parse/dev/Nayau9fwXU/images/54272da6b758f38fcfcc3f8e240d18b2cd04fa874c35cb4c8ffe103888a39048.jpg b/parse/dev/Nayau9fwXU/images/54272da6b758f38fcfcc3f8e240d18b2cd04fa874c35cb4c8ffe103888a39048.jpg new file mode 100644 index 0000000000000000000000000000000000000000..341b4ee7f8e8096ff4db8f2314fd6cb676538848 --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/54272da6b758f38fcfcc3f8e240d18b2cd04fa874c35cb4c8ffe103888a39048.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:529937b1adc896bfe5f335016b83893942b364a4aa0574b4ae083649047c9e03 +size 11637 diff --git a/parse/dev/Nayau9fwXU/images/5c9684f7f01650fa7135dd1ba49537701dd7541fc40ebc78eec3ffcf6e6c8b35.jpg b/parse/dev/Nayau9fwXU/images/5c9684f7f01650fa7135dd1ba49537701dd7541fc40ebc78eec3ffcf6e6c8b35.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c3575b399d3642b249a4905987d8b29dd65737d7 --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/5c9684f7f01650fa7135dd1ba49537701dd7541fc40ebc78eec3ffcf6e6c8b35.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:704801b7a1dcf0c7bb55f39b8d20a2440ae2aacea51501920ec75ec15e569458 +size 6392 diff --git a/parse/dev/Nayau9fwXU/images/5dfe65e9e6e935f7cedb0ccd2d0520898b7eb10bddf2427f46ca53467be759fb.jpg b/parse/dev/Nayau9fwXU/images/5dfe65e9e6e935f7cedb0ccd2d0520898b7eb10bddf2427f46ca53467be759fb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..54048836f1f9f937d5c235499de263d74d95fbba --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/5dfe65e9e6e935f7cedb0ccd2d0520898b7eb10bddf2427f46ca53467be759fb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3ae0279a4b4ce47a51f104696ac1f392a1b0f3890f5e6773426998e191a8828 +size 10304 diff --git a/parse/dev/Nayau9fwXU/images/6d46b2f5bee59937008c1a6ed15e5f0ebc60f1f9756a2940dbd63e6c3268265a.jpg b/parse/dev/Nayau9fwXU/images/6d46b2f5bee59937008c1a6ed15e5f0ebc60f1f9756a2940dbd63e6c3268265a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d6f71ee312998d9378e0341616136e4b352a2cc5 --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/6d46b2f5bee59937008c1a6ed15e5f0ebc60f1f9756a2940dbd63e6c3268265a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ff9607bbf85e6ee166af418a6191569a69ddf541575bdacae33765667bdfc12b +size 8021 diff --git a/parse/dev/Nayau9fwXU/images/7ad2f0cdfe2bc49808028e1bdf12269009517c2d91281c4bcb30ed70a0fd9f33.jpg b/parse/dev/Nayau9fwXU/images/7ad2f0cdfe2bc49808028e1bdf12269009517c2d91281c4bcb30ed70a0fd9f33.jpg new file mode 100644 index 0000000000000000000000000000000000000000..57ce5fab15a4b401d5f10ec42e939b48ea3e3546 --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/7ad2f0cdfe2bc49808028e1bdf12269009517c2d91281c4bcb30ed70a0fd9f33.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14a729758865e3f05c4f29d13c696a73e6b6f422f11d1f6b93a30dc361b75a4b +size 7982 diff --git a/parse/dev/Nayau9fwXU/images/7ce23fbb90b8852e63bfb56a4dd4c73b80f4652a6ebdd06272efac219b9b0859.jpg b/parse/dev/Nayau9fwXU/images/7ce23fbb90b8852e63bfb56a4dd4c73b80f4652a6ebdd06272efac219b9b0859.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cac481ba52e621b77a09da55767205c6ea9c90c5 --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/7ce23fbb90b8852e63bfb56a4dd4c73b80f4652a6ebdd06272efac219b9b0859.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:83c339fce9c4e00af7647bae3a9016e6d9ce80e1ca8db1f53ae3e81727c7c471 +size 42305 diff --git a/parse/dev/Nayau9fwXU/images/85cdf41346f04f48e50036c08b903bbec891a34ac015532f02bc6e9d2e62d8e9.jpg b/parse/dev/Nayau9fwXU/images/85cdf41346f04f48e50036c08b903bbec891a34ac015532f02bc6e9d2e62d8e9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..77527ecc0ffd6ae622af0ef677a245c54b17a55c --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/85cdf41346f04f48e50036c08b903bbec891a34ac015532f02bc6e9d2e62d8e9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a975a6d65d9899b1303259faab2e589eec546b4d5049c8ce828a15c854afca36 +size 6016 diff --git a/parse/dev/Nayau9fwXU/images/936b6fba4101d4ca8bbdbe7dde5b23c57161a091bab935864a599aef6983d958.jpg b/parse/dev/Nayau9fwXU/images/936b6fba4101d4ca8bbdbe7dde5b23c57161a091bab935864a599aef6983d958.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6d9ac4ffd73fa1def263668e7e93e521e20a4c35 --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/936b6fba4101d4ca8bbdbe7dde5b23c57161a091bab935864a599aef6983d958.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:63de9bbec1666b53f05138f3c8c22bf958578cf0cb14eef8fbfcbd4e5932f8f9 +size 266328 diff --git a/parse/dev/Nayau9fwXU/images/9cd3f4687ea51fdbfb0bf9967d443e408ca5470fd5c3315cbc9bb851370f6ce3.jpg b/parse/dev/Nayau9fwXU/images/9cd3f4687ea51fdbfb0bf9967d443e408ca5470fd5c3315cbc9bb851370f6ce3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..20b235a5d11ea8b4966372fccc9853aaadbcfbff --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/9cd3f4687ea51fdbfb0bf9967d443e408ca5470fd5c3315cbc9bb851370f6ce3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3690e97cdd625c6a2bad4fc01659113dade5e63d625c708654614df21676db02 +size 117725 diff --git a/parse/dev/Nayau9fwXU/images/c74cbbaa0eb1d0061d3e1b799d8c764a5dfa0c5d6f9cc11372d8717d7266f2b5.jpg b/parse/dev/Nayau9fwXU/images/c74cbbaa0eb1d0061d3e1b799d8c764a5dfa0c5d6f9cc11372d8717d7266f2b5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9f160f701d7546cd497f1bc4195ea050491af0c1 --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/c74cbbaa0eb1d0061d3e1b799d8c764a5dfa0c5d6f9cc11372d8717d7266f2b5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4449b07432006c1d9d8a22a9ce80f2589edd6c20474d02c01a5500794c7d4ca +size 104221 diff --git a/parse/dev/Nayau9fwXU/images/c7ec43ad26beab0208081211ebfdb6332ce118dc1e9646a417b56c812d9d66bb.jpg b/parse/dev/Nayau9fwXU/images/c7ec43ad26beab0208081211ebfdb6332ce118dc1e9646a417b56c812d9d66bb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ee0641e7a40955103f888dacd6a4a70363b9fd3d --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/c7ec43ad26beab0208081211ebfdb6332ce118dc1e9646a417b56c812d9d66bb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a74658b823dfe0ee51314aa4a4acdcc43e4c33f31aff348974a9cad643a0a91 +size 10334 diff --git a/parse/dev/Nayau9fwXU/images/fc398b6716868248f87a5520fdcdc3ba9c014aff4ecfc4a67e36d2a2e9b67014.jpg b/parse/dev/Nayau9fwXU/images/fc398b6716868248f87a5520fdcdc3ba9c014aff4ecfc4a67e36d2a2e9b67014.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7f6a80ebb484ee87348bf31b8d72e449ad8874ea --- /dev/null +++ b/parse/dev/Nayau9fwXU/images/fc398b6716868248f87a5520fdcdc3ba9c014aff4ecfc4a67e36d2a2e9b67014.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2304d6491e323813b823792a0b08a93973e1cc13fc7d32caa506f26add826454 +size 7497 diff --git a/parse/dev/OkQD6RMUK5/images/06938f58cd9baf24515f7d457b9d802d1a90dc95314f5ab3b6c12563cfebee7d.jpg b/parse/dev/OkQD6RMUK5/images/06938f58cd9baf24515f7d457b9d802d1a90dc95314f5ab3b6c12563cfebee7d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fe828f10124aecffd5c44246c37b0bac3775b5ca --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/06938f58cd9baf24515f7d457b9d802d1a90dc95314f5ab3b6c12563cfebee7d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8b06d481db0e1792d4101b0912bddee94c2077f92cf4e3b6dd83255ac337140f +size 28825 diff --git a/parse/dev/OkQD6RMUK5/images/0e160abacfb80722d115ef5bd8a61cf4aa7d6df30908f9d64ef4b53ca24018fc.jpg b/parse/dev/OkQD6RMUK5/images/0e160abacfb80722d115ef5bd8a61cf4aa7d6df30908f9d64ef4b53ca24018fc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4b93d73b3893d05d084c024e06deff4a7f0a1ad8 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/0e160abacfb80722d115ef5bd8a61cf4aa7d6df30908f9d64ef4b53ca24018fc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3609c78851a057a8e029dc89862b3deb90b33110efdf6d380a3b1c33d0b6e270 +size 48895 diff --git a/parse/dev/OkQD6RMUK5/images/0ed0bd057cbf5665be2a8a2e2eccbd30a726cbbe64cf2a3503faf9aa30a173e1.jpg b/parse/dev/OkQD6RMUK5/images/0ed0bd057cbf5665be2a8a2e2eccbd30a726cbbe64cf2a3503faf9aa30a173e1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c570dc51d1c51a3c52ece60af4ae6fe74d7068f2 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/0ed0bd057cbf5665be2a8a2e2eccbd30a726cbbe64cf2a3503faf9aa30a173e1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9127b09268a7155d53d5c61ba2fe9c1b89f20d35739bee17b40a4b999dd8d28b +size 51650 diff --git a/parse/dev/OkQD6RMUK5/images/1348d5a03207686f5ef409931e1ebb1e41413c288ad46e4df6735ff8266f0aa9.jpg b/parse/dev/OkQD6RMUK5/images/1348d5a03207686f5ef409931e1ebb1e41413c288ad46e4df6735ff8266f0aa9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f7c3ffb2168b711856e266cc13d9f8f9b8495512 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/1348d5a03207686f5ef409931e1ebb1e41413c288ad46e4df6735ff8266f0aa9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:504d49422da069753ab117b55a2e3ed516679d85598a5f1716df4aa0e19064b6 +size 33124 diff --git a/parse/dev/OkQD6RMUK5/images/250166857ceb5887e133811aba1e119a33c7f831b3503f8f26a583c8e51a3779.jpg b/parse/dev/OkQD6RMUK5/images/250166857ceb5887e133811aba1e119a33c7f831b3503f8f26a583c8e51a3779.jpg new file mode 100644 index 0000000000000000000000000000000000000000..53b60dd35bd02692382b36874ee094550d318cb8 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/250166857ceb5887e133811aba1e119a33c7f831b3503f8f26a583c8e51a3779.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d13ff4a5330ae30b741c3347e1aaa28926e10615e4721fae1b070de8f2fcbcc3 +size 29474 diff --git a/parse/dev/OkQD6RMUK5/images/27a54c9018261b7411252a0aea7eaa159692c520952619f2e8386d351de4e772.jpg b/parse/dev/OkQD6RMUK5/images/27a54c9018261b7411252a0aea7eaa159692c520952619f2e8386d351de4e772.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3bdc0f1601e552f283dca4175ac36928bad1537d --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/27a54c9018261b7411252a0aea7eaa159692c520952619f2e8386d351de4e772.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a46605474d84a27942be7a793100835df0cb7a149ab7d83c8a94139cca3fd35d +size 18075 diff --git a/parse/dev/OkQD6RMUK5/images/298da20585b9b56a69e5ad20e173977740e8a9b45729bfec2d9a7bc8b09078f5.jpg b/parse/dev/OkQD6RMUK5/images/298da20585b9b56a69e5ad20e173977740e8a9b45729bfec2d9a7bc8b09078f5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8f63c7112f7006d4ab8953ba73ee33a7deb367a2 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/298da20585b9b56a69e5ad20e173977740e8a9b45729bfec2d9a7bc8b09078f5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ec722202ba55f7bae9b4a92a934c0ec5ee0ff505f27aa2c59dda61a87ed122e5 +size 4335 diff --git a/parse/dev/OkQD6RMUK5/images/2af9bf05b12374d05402e3b50ac8d65e2b1e619247fdbbf52da30793f4495037.jpg b/parse/dev/OkQD6RMUK5/images/2af9bf05b12374d05402e3b50ac8d65e2b1e619247fdbbf52da30793f4495037.jpg new file mode 100644 index 0000000000000000000000000000000000000000..df20bae989ad0259e5da74682299bae1d550eadb --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/2af9bf05b12374d05402e3b50ac8d65e2b1e619247fdbbf52da30793f4495037.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:abffdbc5770c2b192e81a6b4132227a604e54b37f163b64564413d21c2015f30 +size 26498 diff --git a/parse/dev/OkQD6RMUK5/images/2c3bf673e79efc567fd114c1a9b06ed02a52e0920de35bceaadcdd0ef0292bc3.jpg b/parse/dev/OkQD6RMUK5/images/2c3bf673e79efc567fd114c1a9b06ed02a52e0920de35bceaadcdd0ef0292bc3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f01f879752dae542870601c5879fd27b35319c3a --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/2c3bf673e79efc567fd114c1a9b06ed02a52e0920de35bceaadcdd0ef0292bc3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:67b4a496efc6a99ef0a3231399c79a378f1c217a8054166277996cab75b6a0ed +size 28146 diff --git a/parse/dev/OkQD6RMUK5/images/338b9edc77315aff0583ef135817bfe3206446ac9dce3bcc8dd14dc0ec173b84.jpg b/parse/dev/OkQD6RMUK5/images/338b9edc77315aff0583ef135817bfe3206446ac9dce3bcc8dd14dc0ec173b84.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6dd474de62dc6784c891dde6f36dfad33c50eb06 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/338b9edc77315aff0583ef135817bfe3206446ac9dce3bcc8dd14dc0ec173b84.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07787e6774c28492d293c053b5949082aef02b5d7405d4af93bdb288b432bcb6 +size 54180 diff --git a/parse/dev/OkQD6RMUK5/images/35ae802dfb32b3844e12f89ca0f45149fad716bfb96fc03080cdf31763902a22.jpg b/parse/dev/OkQD6RMUK5/images/35ae802dfb32b3844e12f89ca0f45149fad716bfb96fc03080cdf31763902a22.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3a353e87a759cd20ba24a645508fccd930cb913a --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/35ae802dfb32b3844e12f89ca0f45149fad716bfb96fc03080cdf31763902a22.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f2328e6623d0fe114e507f3d450e7ee3cca64ce4a25e3a8dc3b394bdd67f4536 +size 26847 diff --git a/parse/dev/OkQD6RMUK5/images/40944a62dfd01cef80fe496f830b217ab1afbfa089a37cd615ffee7d3cf1714f.jpg b/parse/dev/OkQD6RMUK5/images/40944a62dfd01cef80fe496f830b217ab1afbfa089a37cd615ffee7d3cf1714f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8db226249a2d541af367eb6e10a9253931379154 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/40944a62dfd01cef80fe496f830b217ab1afbfa089a37cd615ffee7d3cf1714f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c56bd6c0843f672e88c45e0ee99acab3f4ec200dbfe3733ac0c4eca112a1b4a +size 8871 diff --git a/parse/dev/OkQD6RMUK5/images/45f5fb60b396c291a15216bff8c3401dc3ead2f43a877386913caa8773275815.jpg b/parse/dev/OkQD6RMUK5/images/45f5fb60b396c291a15216bff8c3401dc3ead2f43a877386913caa8773275815.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0b316e400645143a794bf84d0a315de436da15b4 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/45f5fb60b396c291a15216bff8c3401dc3ead2f43a877386913caa8773275815.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:379b19dac02dd0bff0470278eec399f61d685c8fb79417ad1737ec184c22cb11 +size 8153 diff --git a/parse/dev/OkQD6RMUK5/images/69b6bb6ed3d5fa0f39d9a64fd4488c6c18eb916732befa58b76bb19e7dd77c4e.jpg b/parse/dev/OkQD6RMUK5/images/69b6bb6ed3d5fa0f39d9a64fd4488c6c18eb916732befa58b76bb19e7dd77c4e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..21335be4514f2ac34b1e66f6afaad25b22004545 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/69b6bb6ed3d5fa0f39d9a64fd4488c6c18eb916732befa58b76bb19e7dd77c4e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5ab8edae36ce9b52e3de08e9adf18d6c7bc74633d786cf4736b43786567256c +size 6763 diff --git a/parse/dev/OkQD6RMUK5/images/6f242a6dc19d75c1692f890222573bcdd6d2c1dadeaf50101a62949694a19576.jpg b/parse/dev/OkQD6RMUK5/images/6f242a6dc19d75c1692f890222573bcdd6d2c1dadeaf50101a62949694a19576.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ec6fa449b369e090f9a604893dda53993822e2f2 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/6f242a6dc19d75c1692f890222573bcdd6d2c1dadeaf50101a62949694a19576.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a86bcdf97a8ea35f40dc09e6f31691fb216c17d7acdf331f95a70946ae1b052 +size 7445 diff --git a/parse/dev/OkQD6RMUK5/images/7ab4a19ec2049d39e0aafdeb82ac7e062ff32292b424dbe7ca3d0e593be95035.jpg b/parse/dev/OkQD6RMUK5/images/7ab4a19ec2049d39e0aafdeb82ac7e062ff32292b424dbe7ca3d0e593be95035.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e08ebbf9cdb1d4b7d7cfab3cf8e2bcbe1500ac4a --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/7ab4a19ec2049d39e0aafdeb82ac7e062ff32292b424dbe7ca3d0e593be95035.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f8e9a38c48c1d26ef039c844a246b234ae6cb9ec29a18adf05c367efa1822c14 +size 80704 diff --git a/parse/dev/OkQD6RMUK5/images/91bb9dacc15d85cd07ef56e948bc7ab3690cadb6c80d61ce0026488c6037a697.jpg b/parse/dev/OkQD6RMUK5/images/91bb9dacc15d85cd07ef56e948bc7ab3690cadb6c80d61ce0026488c6037a697.jpg new file mode 100644 index 0000000000000000000000000000000000000000..06ea3b59a3a2509b4d98e19726c1f8b0c692e8ca --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/91bb9dacc15d85cd07ef56e948bc7ab3690cadb6c80d61ce0026488c6037a697.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:206a22ca964c5c435b6d188265cfae6a128bed1e00780396fcda109edb55af4f +size 16661 diff --git a/parse/dev/OkQD6RMUK5/images/9760be076722c0dadc6a3acc822ccd8d78986e14421c723353d6940663e68749.jpg b/parse/dev/OkQD6RMUK5/images/9760be076722c0dadc6a3acc822ccd8d78986e14421c723353d6940663e68749.jpg new file mode 100644 index 0000000000000000000000000000000000000000..da55bf1ac14a7b8945485c4b9f2ae980557c1e9a --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/9760be076722c0dadc6a3acc822ccd8d78986e14421c723353d6940663e68749.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7453794bc56d196109d11986b27a3f18ac99e83842d41f4640297f7f7d76171 +size 170259 diff --git a/parse/dev/OkQD6RMUK5/images/9b3dda8f2cd783f15de23576c42f7c6fcc74d7b0861bc371d344dad8d8fbf03d.jpg b/parse/dev/OkQD6RMUK5/images/9b3dda8f2cd783f15de23576c42f7c6fcc74d7b0861bc371d344dad8d8fbf03d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ad0850e21b7d0f8fefe944eafbd63fe0b2b67b96 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/9b3dda8f2cd783f15de23576c42f7c6fcc74d7b0861bc371d344dad8d8fbf03d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:821133bc16ff48a701539c5fa042abad6e458814e4be495892d277341b6a0484 +size 53251 diff --git a/parse/dev/OkQD6RMUK5/images/9d20d5997e0309263bd6912ecb7f66b8ff95983abbf3f8f93de4b995298e4cd5.jpg b/parse/dev/OkQD6RMUK5/images/9d20d5997e0309263bd6912ecb7f66b8ff95983abbf3f8f93de4b995298e4cd5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..896a91cf01c6708fbd0ec419520d28ff1db6a359 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/9d20d5997e0309263bd6912ecb7f66b8ff95983abbf3f8f93de4b995298e4cd5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d41bdced5e54a134958b447943845277583a1527511aded1e45078f3ba2b85e +size 53812 diff --git a/parse/dev/OkQD6RMUK5/images/a36cc5cfb2b9d391e9512aa3f217b721c17b5bf0794ee19ac5547fc8dfcfc764.jpg b/parse/dev/OkQD6RMUK5/images/a36cc5cfb2b9d391e9512aa3f217b721c17b5bf0794ee19ac5547fc8dfcfc764.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2afdefa4a5d8656156347b50a30ccb053175c864 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/a36cc5cfb2b9d391e9512aa3f217b721c17b5bf0794ee19ac5547fc8dfcfc764.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d197e652dfdc212174ef8611a253e1157a500193bc9da49b085be5048404a89b +size 15581 diff --git a/parse/dev/OkQD6RMUK5/images/a815f2ceed6826c0128ce36d2e48a9e4cabd3b9ec8129e03237b4932294db235.jpg b/parse/dev/OkQD6RMUK5/images/a815f2ceed6826c0128ce36d2e48a9e4cabd3b9ec8129e03237b4932294db235.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bb36c3837d3e596f8dac896abf40241024b1152c --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/a815f2ceed6826c0128ce36d2e48a9e4cabd3b9ec8129e03237b4932294db235.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5d79c8219e7e78431ff844756b3ed8724e79c75a29b0d512d0987ed9a002caa +size 52587 diff --git a/parse/dev/OkQD6RMUK5/images/a8eca342476a542ed54bc76e5959783edb3901a86a0bfc32d41d2707ee2a2615.jpg b/parse/dev/OkQD6RMUK5/images/a8eca342476a542ed54bc76e5959783edb3901a86a0bfc32d41d2707ee2a2615.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a9795777b866c473b3d7afdf2f5d58b9e057664e --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/a8eca342476a542ed54bc76e5959783edb3901a86a0bfc32d41d2707ee2a2615.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:619c398e735bda5831b0264e7ebd0c57d80d7b919488ec27f039a2119c19b560 +size 7865 diff --git a/parse/dev/OkQD6RMUK5/images/aa119df19f1993178ce7da93db050823f3f64ff34fdb55fd9ad886e8f0380af1.jpg b/parse/dev/OkQD6RMUK5/images/aa119df19f1993178ce7da93db050823f3f64ff34fdb55fd9ad886e8f0380af1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1068f89a118647d6ac756c2a1fac81f3c222e5e9 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/aa119df19f1993178ce7da93db050823f3f64ff34fdb55fd9ad886e8f0380af1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27f3b87fac005742f4057ce7345ade733df658bf0bbbdfd60eb81e337aba8ec8 +size 24781 diff --git a/parse/dev/OkQD6RMUK5/images/aab81b4153e76a9e60150e44776274b92d38186201c5300b997a5eeb23db79d5.jpg b/parse/dev/OkQD6RMUK5/images/aab81b4153e76a9e60150e44776274b92d38186201c5300b997a5eeb23db79d5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ee540711a55d3b8052761bc31f77c9539c847e6e --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/aab81b4153e76a9e60150e44776274b92d38186201c5300b997a5eeb23db79d5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d548bac8b24619d953172cb91a4f713f7814e1aace126a31002ac5fb09c4209 +size 29457 diff --git a/parse/dev/OkQD6RMUK5/images/b2b1923beac377db0c881c03b21bd382b318f133115f39a0791b2c10b8b8dde6.jpg b/parse/dev/OkQD6RMUK5/images/b2b1923beac377db0c881c03b21bd382b318f133115f39a0791b2c10b8b8dde6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7326ce2a1f385b608d4d6f4f1b64376f2eb3a650 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/b2b1923beac377db0c881c03b21bd382b318f133115f39a0791b2c10b8b8dde6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c18718625ab64ffecd05559530b720faeeec99473f284f12450fab6e5e7142d +size 40181 diff --git a/parse/dev/OkQD6RMUK5/images/b7822d3fde5651b5d36fd88326039a3a90e11d6ad9c9337825334c556311b656.jpg b/parse/dev/OkQD6RMUK5/images/b7822d3fde5651b5d36fd88326039a3a90e11d6ad9c9337825334c556311b656.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1f443f52807a684999e41cc86a02c0dc969f9c9c --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/b7822d3fde5651b5d36fd88326039a3a90e11d6ad9c9337825334c556311b656.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6257c87a49f2b8707b37f5a7482c83ffe023b5131439ba990ad9d1b404a8d9df +size 26327 diff --git a/parse/dev/OkQD6RMUK5/images/bb6b4ddf8ee18d9b5ad9f2445851d108a5ea758c37e6d0d51609fbe456b87b2d.jpg b/parse/dev/OkQD6RMUK5/images/bb6b4ddf8ee18d9b5ad9f2445851d108a5ea758c37e6d0d51609fbe456b87b2d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1a525fa6038948fa28a5b7feaef3c1f55c341854 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/bb6b4ddf8ee18d9b5ad9f2445851d108a5ea758c37e6d0d51609fbe456b87b2d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:195de7c13cfe4ca1677be87f32e23cfb3de8229a2143f63eeae2d5d66ab24639 +size 48373 diff --git a/parse/dev/OkQD6RMUK5/images/bedfb0e70bdd77f43edd8ae1e88945810f9a341176280a0f43d292ecff486687.jpg b/parse/dev/OkQD6RMUK5/images/bedfb0e70bdd77f43edd8ae1e88945810f9a341176280a0f43d292ecff486687.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4e6dccbd06e1e7a5fa2994c6c7d169bfd0c04c87 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/bedfb0e70bdd77f43edd8ae1e88945810f9a341176280a0f43d292ecff486687.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e24d51e8fa446a8124a4a6436f23ce2a5868e3dee20ad862d56e64830946fa6 +size 4407 diff --git a/parse/dev/OkQD6RMUK5/images/c5800fc19b72e3f240082b75cf146da670e5580d62ab6cc98e7b7a5443645c13.jpg b/parse/dev/OkQD6RMUK5/images/c5800fc19b72e3f240082b75cf146da670e5580d62ab6cc98e7b7a5443645c13.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5d5b7344f8bbfc9a09faeb9a0004daea203fa03d --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/c5800fc19b72e3f240082b75cf146da670e5580d62ab6cc98e7b7a5443645c13.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:122667ffce8d560c79172dd9a2bd99dd92b98c2d72a917b42d882f6194fd8ea6 +size 15172 diff --git a/parse/dev/OkQD6RMUK5/images/d9e0c8418a9fed47fca0f926adad96abe2bd5c92603a10d7623daae30a52d9ae.jpg b/parse/dev/OkQD6RMUK5/images/d9e0c8418a9fed47fca0f926adad96abe2bd5c92603a10d7623daae30a52d9ae.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a45ae1e1a4f02018aa816b42a8ce688b47010dcd --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/d9e0c8418a9fed47fca0f926adad96abe2bd5c92603a10d7623daae30a52d9ae.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a3212295230b9b46beef18c07e03aaeb1f351eef26bd9191242abc3e3b05ce9 +size 35248 diff --git a/parse/dev/OkQD6RMUK5/images/dcf59c6c737bc50f6a35ab4b730fea66cba4af9976e2502c10b76f97dae03f40.jpg b/parse/dev/OkQD6RMUK5/images/dcf59c6c737bc50f6a35ab4b730fea66cba4af9976e2502c10b76f97dae03f40.jpg new file mode 100644 index 0000000000000000000000000000000000000000..39c6cc9a7e5b99981afb62a84c564d02f7411ca9 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/dcf59c6c737bc50f6a35ab4b730fea66cba4af9976e2502c10b76f97dae03f40.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6a18548158aaf26463559c7fd577ed65d2cd8fc7546e9cfe20d1f8cfe6d3cd7 +size 46343 diff --git a/parse/dev/OkQD6RMUK5/images/ee98817bf6e66c96fb70f818b1349a3986bb789b3845d1b6d097c411f6af88ba.jpg b/parse/dev/OkQD6RMUK5/images/ee98817bf6e66c96fb70f818b1349a3986bb789b3845d1b6d097c411f6af88ba.jpg new file mode 100644 index 0000000000000000000000000000000000000000..820456bf4c71d571a7ef495c0d1f52dbb713acfa --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/ee98817bf6e66c96fb70f818b1349a3986bb789b3845d1b6d097c411f6af88ba.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c838d47de6732c746e74cbf8cf89b3cc0172487143816290ec5203f2b36d032c +size 6609 diff --git a/parse/dev/OkQD6RMUK5/images/f1641bc506962d25bef4bf8fe8cdec894f5a85f235be6d49a21c0141fef0f889.jpg b/parse/dev/OkQD6RMUK5/images/f1641bc506962d25bef4bf8fe8cdec894f5a85f235be6d49a21c0141fef0f889.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c23368453e533cf0ddfb0a4932ef6cc99f03b589 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/f1641bc506962d25bef4bf8fe8cdec894f5a85f235be6d49a21c0141fef0f889.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25fb5428cd155482de69b91c04fab69e40acb156c06ebc802c97c0b7671b8b84 +size 6487 diff --git a/parse/dev/OkQD6RMUK5/images/f6b0921a653f98748ed45d262e560d328f5a211756f2afa9d7700f1428a72000.jpg b/parse/dev/OkQD6RMUK5/images/f6b0921a653f98748ed45d262e560d328f5a211756f2afa9d7700f1428a72000.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fe744219d90446ac836cb7031199fdf8aa3ce233 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/f6b0921a653f98748ed45d262e560d328f5a211756f2afa9d7700f1428a72000.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5afb570b9c8f90654f047aa65df21d88321420cd338aab13e4ca685fdda685b +size 5060 diff --git a/parse/dev/OkQD6RMUK5/images/f8e4848e47b19dcc8f979ea2a204e41473173a704862df66f31c362f07acd040.jpg b/parse/dev/OkQD6RMUK5/images/f8e4848e47b19dcc8f979ea2a204e41473173a704862df66f31c362f07acd040.jpg new file mode 100644 index 0000000000000000000000000000000000000000..43ec010a2e93af4af05c398d3f99e6f3d65702d8 --- /dev/null +++ b/parse/dev/OkQD6RMUK5/images/f8e4848e47b19dcc8f979ea2a204e41473173a704862df66f31c362f07acd040.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4628d6318cc4ec3611bd25c838f2cdf71142303c201402af53276868d39224c2 +size 5727 diff --git a/parse/dev/RzXb6a3H3rs/images/099581c7bafe70e14b202bdacd3b566df6f0b7cee4f789a8aedc2189bde09adf.jpg b/parse/dev/RzXb6a3H3rs/images/099581c7bafe70e14b202bdacd3b566df6f0b7cee4f789a8aedc2189bde09adf.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4223f5587704fcc6944c77209e1c331f97e5f1a0 --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/099581c7bafe70e14b202bdacd3b566df6f0b7cee4f789a8aedc2189bde09adf.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43a276881605484c1bd07193561e0024d0f9f2f77a1cef876255e8ee0c7b4bad +size 89755 diff --git a/parse/dev/RzXb6a3H3rs/images/2bc959636378fc23584d34ac0ff712fe11be9d11936b03094222f1eb92a5c5d1.jpg b/parse/dev/RzXb6a3H3rs/images/2bc959636378fc23584d34ac0ff712fe11be9d11936b03094222f1eb92a5c5d1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c20d15e4858e462db7587c7aee19eb7e0b18aa44 --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/2bc959636378fc23584d34ac0ff712fe11be9d11936b03094222f1eb92a5c5d1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a92488563f963fe04a1790aa0d2b2c287de16d020fa6cf3349013eed981ee82c +size 29388 diff --git a/parse/dev/RzXb6a3H3rs/images/39d54beb042a611a07554aa87244f0da1d9dc74cf43abbe18941e2c79ac89b14.jpg b/parse/dev/RzXb6a3H3rs/images/39d54beb042a611a07554aa87244f0da1d9dc74cf43abbe18941e2c79ac89b14.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a77a6a176bd3c54f1c6789c93dbba94defcdba33 --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/39d54beb042a611a07554aa87244f0da1d9dc74cf43abbe18941e2c79ac89b14.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6a0cfed7865b2c92b25631511358040f693dbe14e444cb5a0c07a75a76af20e2 +size 4324 diff --git a/parse/dev/RzXb6a3H3rs/images/4bada4c7fb75fc17f5d671b20d278b882ba3f46caedbd0ca7570db1b6c34be84.jpg b/parse/dev/RzXb6a3H3rs/images/4bada4c7fb75fc17f5d671b20d278b882ba3f46caedbd0ca7570db1b6c34be84.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f077d8a8d8a33f53b21ce96a40ac8600b5e70cf7 --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/4bada4c7fb75fc17f5d671b20d278b882ba3f46caedbd0ca7570db1b6c34be84.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:17b638731c7ae14d8e117e6f166c2e6d79e7657410e22fde15a441633fc220fe +size 36357 diff --git a/parse/dev/RzXb6a3H3rs/images/69632bb40b2896abb78ee3472dd39d3ad423ee9f487c2188e4b2ab40959ff310.jpg b/parse/dev/RzXb6a3H3rs/images/69632bb40b2896abb78ee3472dd39d3ad423ee9f487c2188e4b2ab40959ff310.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d508c34464552e6c83d515760730a2c569ca6be5 --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/69632bb40b2896abb78ee3472dd39d3ad423ee9f487c2188e4b2ab40959ff310.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a9b176cb18f2afc4978be4b9b6092697cdb4bc0b0f08e7467bdb8439452a0d5a +size 6096 diff --git a/parse/dev/RzXb6a3H3rs/images/791ba5f7100cc72c11bc599dc9c7f189726cce930a5b6b7ffeabcddbd4201067.jpg b/parse/dev/RzXb6a3H3rs/images/791ba5f7100cc72c11bc599dc9c7f189726cce930a5b6b7ffeabcddbd4201067.jpg new file mode 100644 index 0000000000000000000000000000000000000000..118935ccaf6b2b2e20978e0bb95e8bbb5f1f3189 --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/791ba5f7100cc72c11bc599dc9c7f189726cce930a5b6b7ffeabcddbd4201067.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6410579c01947dfdb77ad825b836c9c20c0cc1f0c9be34caf32ff3e1afe264a1 +size 46312 diff --git a/parse/dev/RzXb6a3H3rs/images/8713d6ae02e37be2e274e4f7057eb2f33cdd4da84b11a9a6d6fb6126964b32ec.jpg b/parse/dev/RzXb6a3H3rs/images/8713d6ae02e37be2e274e4f7057eb2f33cdd4da84b11a9a6d6fb6126964b32ec.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e90b740740e7bbc47843ad073d1f73c66c81605a --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/8713d6ae02e37be2e274e4f7057eb2f33cdd4da84b11a9a6d6fb6126964b32ec.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e2e03f258e6b5115277cacc472e7f4faf6658066b56acc0b66759ccd1cab0f76 +size 8267 diff --git a/parse/dev/RzXb6a3H3rs/images/88ccd56ac23367bbd981ba1cc80c57351dc78cfbd89415bee8fd6e5f80c7bf30.jpg b/parse/dev/RzXb6a3H3rs/images/88ccd56ac23367bbd981ba1cc80c57351dc78cfbd89415bee8fd6e5f80c7bf30.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fcf67e90aef639340ac609e0570c0b74d8acdc39 --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/88ccd56ac23367bbd981ba1cc80c57351dc78cfbd89415bee8fd6e5f80c7bf30.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d8111a30db8e6ee7f3392e0b0d6dfd8bba984f88668ede8b24de4a3582bc2b5c +size 11092 diff --git a/parse/dev/RzXb6a3H3rs/images/a5b18375003ad8be65525bc772163ea949f3a3366ea66a4212b70750bd154e79.jpg b/parse/dev/RzXb6a3H3rs/images/a5b18375003ad8be65525bc772163ea949f3a3366ea66a4212b70750bd154e79.jpg new file mode 100644 index 0000000000000000000000000000000000000000..373ad3e002bbc6b1155cf9235efe8838f7124d10 --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/a5b18375003ad8be65525bc772163ea949f3a3366ea66a4212b70750bd154e79.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4540fc4ee25d90bf949a729ffb99da2210b3e5088a1c8ee20904f79d9c4edd29 +size 30491 diff --git a/parse/dev/RzXb6a3H3rs/images/b2b885e90ab9fc48f9e4bc5fae2bc06fc57b7771b4a1d50e0dbb70dda6b510b5.jpg b/parse/dev/RzXb6a3H3rs/images/b2b885e90ab9fc48f9e4bc5fae2bc06fc57b7771b4a1d50e0dbb70dda6b510b5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..43742032966575027b35008a7fe2474abfe175fd --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/b2b885e90ab9fc48f9e4bc5fae2bc06fc57b7771b4a1d50e0dbb70dda6b510b5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:569532e8d36f5899ab8779730b51396cbf42674f907333ab838bdbf9c2c3c064 +size 35244 diff --git a/parse/dev/RzXb6a3H3rs/images/b70852603919816c86f2bdac7581a13e2bf5dc73c04df766cf727b1bbb48b666.jpg b/parse/dev/RzXb6a3H3rs/images/b70852603919816c86f2bdac7581a13e2bf5dc73c04df766cf727b1bbb48b666.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b3b747e0c80f22491475e6d79772a333f9dff480 --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/b70852603919816c86f2bdac7581a13e2bf5dc73c04df766cf727b1bbb48b666.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5168f71503210f12e225259ce6b9dfede7d9f3b0d0a9f87f60a6e66b7d741e9f +size 25665 diff --git a/parse/dev/RzXb6a3H3rs/images/b938d210ec55c6aaeec67c899910f486fadd0534007595d80a0b3c8944935d44.jpg b/parse/dev/RzXb6a3H3rs/images/b938d210ec55c6aaeec67c899910f486fadd0534007595d80a0b3c8944935d44.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2fdd1776d33bcd50f2c8ee924e50abf46fef6658 --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/b938d210ec55c6aaeec67c899910f486fadd0534007595d80a0b3c8944935d44.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:973dcd717cf32589068aa9f3254a4453c5a6aa239aa6419dd43b8de7241529f2 +size 7101 diff --git a/parse/dev/RzXb6a3H3rs/images/c3b5792f03bdf715958baa91139338a6d8ec5c0b077f8e0b6f67559ee8955cbf.jpg b/parse/dev/RzXb6a3H3rs/images/c3b5792f03bdf715958baa91139338a6d8ec5c0b077f8e0b6f67559ee8955cbf.jpg new file mode 100644 index 0000000000000000000000000000000000000000..59d0834dbe78c443381e6acc8dc60fb824531a5a --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/c3b5792f03bdf715958baa91139338a6d8ec5c0b077f8e0b6f67559ee8955cbf.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ded3f26e3b466c6a1f897a38135e8b14b2acd7ff5cab01ae6bc318b5b3ab9a4a +size 25725 diff --git a/parse/dev/RzXb6a3H3rs/images/cefb0a123923292d414b2c6e64ea16e81efc69f9b266e020811e3f0a0240ab06.jpg b/parse/dev/RzXb6a3H3rs/images/cefb0a123923292d414b2c6e64ea16e81efc69f9b266e020811e3f0a0240ab06.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d02d7c3b0a3ccdb40979693982f1b21dbfd7d69c --- /dev/null +++ b/parse/dev/RzXb6a3H3rs/images/cefb0a123923292d414b2c6e64ea16e81efc69f9b266e020811e3f0a0240ab06.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0c21454634b1cbeedd714b429a8611fe9a4cf215daac03946a6a89fd7e73a4d +size 6673 diff --git a/parse/dev/S5wmbQc1We/images/01bf6b5de01736cd06d52ccf64b92caac2e8f38abe74f3c56b9c06e02d77f1b0.jpg b/parse/dev/S5wmbQc1We/images/01bf6b5de01736cd06d52ccf64b92caac2e8f38abe74f3c56b9c06e02d77f1b0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c2ce4d3cc7a35da06e07cee68431d923d753518e --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/01bf6b5de01736cd06d52ccf64b92caac2e8f38abe74f3c56b9c06e02d77f1b0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d5b30a74497095290045e097aaaf13e5432a101f860060e3c46cccaba69cb371 +size 120225 diff --git a/parse/dev/S5wmbQc1We/images/04f9a22f324d3223bac22f363b658863254014c0d441d4dea386aa215fe7ae5f.jpg b/parse/dev/S5wmbQc1We/images/04f9a22f324d3223bac22f363b658863254014c0d441d4dea386aa215fe7ae5f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2275909138a6580bab3d5f698dd8c4a1fc590dba --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/04f9a22f324d3223bac22f363b658863254014c0d441d4dea386aa215fe7ae5f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:137c25a0ef85044e1069d9e2928ec5d80b4b6228e503d940cfab7674708db437 +size 4518 diff --git a/parse/dev/S5wmbQc1We/images/07ca1f822ae49d607b02114515653b44d472ba661eca33f7fce790e554e9baf3.jpg b/parse/dev/S5wmbQc1We/images/07ca1f822ae49d607b02114515653b44d472ba661eca33f7fce790e554e9baf3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..84038054cb89d77577000fa50e8e31197ecd2064 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/07ca1f822ae49d607b02114515653b44d472ba661eca33f7fce790e554e9baf3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c1c65e540375ff47f1c2c932dc8a50d3cd13fd97aa81ba92180d5e4bf3d0dd9 +size 8458 diff --git a/parse/dev/S5wmbQc1We/images/0ae2d95f43ce3cbd3b3900bde01a24648ad17822897c07fdc74a6cdf00b0bfb9.jpg b/parse/dev/S5wmbQc1We/images/0ae2d95f43ce3cbd3b3900bde01a24648ad17822897c07fdc74a6cdf00b0bfb9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c1779f044ebdb452e68a5112fdc155aa8b305e37 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/0ae2d95f43ce3cbd3b3900bde01a24648ad17822897c07fdc74a6cdf00b0bfb9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25134abd8ec7e0d3af695111bb3f402a8954fde5ca301e8e328bfe5179f4a277 +size 12390 diff --git a/parse/dev/S5wmbQc1We/images/0feb2db95d70a731bc0f9cfba96608247c2b687b80ec80cbd916a74cea8ae715.jpg b/parse/dev/S5wmbQc1We/images/0feb2db95d70a731bc0f9cfba96608247c2b687b80ec80cbd916a74cea8ae715.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dd9dd45ddd82f7feb3a78a8f626f16f293377ecc --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/0feb2db95d70a731bc0f9cfba96608247c2b687b80ec80cbd916a74cea8ae715.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f843a1e495add55d19fc9fa321e5b39d6d3794b1eec4e92d8adf3f49a6b6cff +size 144535 diff --git a/parse/dev/S5wmbQc1We/images/1d249505fc3f0761cb53e16ff7dd3894f83afd1bfb2d56a6e7abbf7be57a418b.jpg b/parse/dev/S5wmbQc1We/images/1d249505fc3f0761cb53e16ff7dd3894f83afd1bfb2d56a6e7abbf7be57a418b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8aec5d2ecaa55db0a9110ed0d7aac4ccbecb91e2 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/1d249505fc3f0761cb53e16ff7dd3894f83afd1bfb2d56a6e7abbf7be57a418b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8a5b9900c1b7d2a818710e64909c756c08df95e70a62fd765a5ef069c7cf1917 +size 3890 diff --git a/parse/dev/S5wmbQc1We/images/1e6cf8ed36fef9d92dcdbd371a70bbed922e5e824ad7df98c2b822c61f0b86af.jpg b/parse/dev/S5wmbQc1We/images/1e6cf8ed36fef9d92dcdbd371a70bbed922e5e824ad7df98c2b822c61f0b86af.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a3ec9dc811c32eb272ab0a96b658436decad2047 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/1e6cf8ed36fef9d92dcdbd371a70bbed922e5e824ad7df98c2b822c61f0b86af.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d99a3552eb9d78513cd0ac13819187b7dc84988f457ac27e0aed3fef79f124f0 +size 7618 diff --git a/parse/dev/S5wmbQc1We/images/208a9e146b1796dbfcc6aed685b316be60b30f70153c68d9352806b6d6967502.jpg b/parse/dev/S5wmbQc1We/images/208a9e146b1796dbfcc6aed685b316be60b30f70153c68d9352806b6d6967502.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2d63f04e2083ed8d90a0a8bcddfb2fd963dadb22 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/208a9e146b1796dbfcc6aed685b316be60b30f70153c68d9352806b6d6967502.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9250cca5f766e1088885abdc2f99152f831e545f22f88169c8f8fb5187bd1b60 +size 109355 diff --git a/parse/dev/S5wmbQc1We/images/21d102a425be8f7ecd1521e7f89086bb694334c9adbd6fc9fd7307aea8c45af7.jpg b/parse/dev/S5wmbQc1We/images/21d102a425be8f7ecd1521e7f89086bb694334c9adbd6fc9fd7307aea8c45af7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..29173570cfc2026cf5d4aa6d1382c65827e7df00 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/21d102a425be8f7ecd1521e7f89086bb694334c9adbd6fc9fd7307aea8c45af7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b8424f4d71003e604b79687c36e7bb227ed6c7c89dd255cef019b46581974ce +size 105566 diff --git a/parse/dev/S5wmbQc1We/images/22396d7ed4ac01a8ba85838d60d8ab7d94398fa581964c76ecc4244c63e47fd5.jpg b/parse/dev/S5wmbQc1We/images/22396d7ed4ac01a8ba85838d60d8ab7d94398fa581964c76ecc4244c63e47fd5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..016ef85299b1c2169ca3737829a7feda4cfc18d2 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/22396d7ed4ac01a8ba85838d60d8ab7d94398fa581964c76ecc4244c63e47fd5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89bd8eceae26269b62535604d596ca137be7592d5537d80737cedcf578cd1786 +size 42376 diff --git a/parse/dev/S5wmbQc1We/images/25459a229478da7afbc938b0a2d2ab789a18c3adb27d1be2413c5ae527e1ea20.jpg b/parse/dev/S5wmbQc1We/images/25459a229478da7afbc938b0a2d2ab789a18c3adb27d1be2413c5ae527e1ea20.jpg new file mode 100644 index 0000000000000000000000000000000000000000..47b8a4cfc8a841a4f544bdbca5807e6239cc3f38 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/25459a229478da7afbc938b0a2d2ab789a18c3adb27d1be2413c5ae527e1ea20.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5482aa9adf6831571ae85b2204a50a8165b26ca9464d62fb48a922215b933d62 +size 21588 diff --git a/parse/dev/S5wmbQc1We/images/262aacc85fad06c0956a5c94d9416eca3a007771e7088076425c9761096eab86.jpg b/parse/dev/S5wmbQc1We/images/262aacc85fad06c0956a5c94d9416eca3a007771e7088076425c9761096eab86.jpg new file mode 100644 index 0000000000000000000000000000000000000000..373dfbd43322d6be68d4e123cf49fbee3fbc9d5a --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/262aacc85fad06c0956a5c94d9416eca3a007771e7088076425c9761096eab86.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a95d2cf6b420bc9f9baaaa1c0a072dd89813bf18681f7f3d99ffd1185af1a180 +size 108516 diff --git a/parse/dev/S5wmbQc1We/images/27e9e673e2d2e485ab1dc14b4b52023a2ca88454d83f6c4353efa0239e77a651.jpg b/parse/dev/S5wmbQc1We/images/27e9e673e2d2e485ab1dc14b4b52023a2ca88454d83f6c4353efa0239e77a651.jpg new file mode 100644 index 0000000000000000000000000000000000000000..33e0a19643044823adf703ced27bcb32686250e7 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/27e9e673e2d2e485ab1dc14b4b52023a2ca88454d83f6c4353efa0239e77a651.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4cc3adac7b245d8fe3c92b5b182b21b264e4501202ab2698ec46094f5efd9718 +size 58266 diff --git a/parse/dev/S5wmbQc1We/images/2f25e792d71bb04f39c1bf014fa70aa28c6f7d41afa66156b41e29d261c38c52.jpg b/parse/dev/S5wmbQc1We/images/2f25e792d71bb04f39c1bf014fa70aa28c6f7d41afa66156b41e29d261c38c52.jpg new file mode 100644 index 0000000000000000000000000000000000000000..54a8a08494347d585922b5c3efb360b255ec35f3 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/2f25e792d71bb04f39c1bf014fa70aa28c6f7d41afa66156b41e29d261c38c52.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e5f0e4f0f5d20209a1e3bd88e5cb60454402a59f86a9e1f6e650e9817e3a4da +size 27373 diff --git a/parse/dev/S5wmbQc1We/images/2fa350e434aee390247637cb1bc350bce65583764bb6e7798a495579481a6360.jpg b/parse/dev/S5wmbQc1We/images/2fa350e434aee390247637cb1bc350bce65583764bb6e7798a495579481a6360.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0c6802e443beeedddca942509953bf263e42defe --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/2fa350e434aee390247637cb1bc350bce65583764bb6e7798a495579481a6360.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:589bacd1e40d101b95fcbc484ed5d8652d28c40115ca9004470784f7ab11e3c8 +size 7967 diff --git a/parse/dev/S5wmbQc1We/images/31ed86340a77685732e7e9363ad4e39c29c6ea847776269b7ccfee7e7580835a.jpg b/parse/dev/S5wmbQc1We/images/31ed86340a77685732e7e9363ad4e39c29c6ea847776269b7ccfee7e7580835a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..17de74918a86b56932ab643da62ee6964b09b0bb --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/31ed86340a77685732e7e9363ad4e39c29c6ea847776269b7ccfee7e7580835a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4fe9faaaae4a4a8ba3e895dfe9d5c13402286a641fb8af50f167d416548dd9ed +size 22308 diff --git a/parse/dev/S5wmbQc1We/images/39705dfeb893c4170e800d2a99e816eb82bb1d03503e16e63da1c1f86bab3291.jpg b/parse/dev/S5wmbQc1We/images/39705dfeb893c4170e800d2a99e816eb82bb1d03503e16e63da1c1f86bab3291.jpg new file mode 100644 index 0000000000000000000000000000000000000000..48258a2f1870c858f621899c5f36b267b6769c8e --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/39705dfeb893c4170e800d2a99e816eb82bb1d03503e16e63da1c1f86bab3291.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e62d6fe0ed3ccb45baae3cc28f661e7dfcb6cadc24b9a88d8a280358803cc44c +size 191645 diff --git a/parse/dev/S5wmbQc1We/images/3b5bcb2407867634c3a3a64e1ba5e3e12dd8262482af1732c6a562c71758e744.jpg b/parse/dev/S5wmbQc1We/images/3b5bcb2407867634c3a3a64e1ba5e3e12dd8262482af1732c6a562c71758e744.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dccb7faee0788749ddf3b48675424847095a136e --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/3b5bcb2407867634c3a3a64e1ba5e3e12dd8262482af1732c6a562c71758e744.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31c43741394981e0c9c68463e91a631c0ceb0025c785941ea768bcaf751aabd8 +size 39516 diff --git a/parse/dev/S5wmbQc1We/images/4518f1c3f7b0e97afc8bc880c00613ffc7fb0a21dbf71b429831d34673932c81.jpg b/parse/dev/S5wmbQc1We/images/4518f1c3f7b0e97afc8bc880c00613ffc7fb0a21dbf71b429831d34673932c81.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3b5d5d00145cf25c438ba7b14b733f8dc46a6f96 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/4518f1c3f7b0e97afc8bc880c00613ffc7fb0a21dbf71b429831d34673932c81.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a0a56a2cbbf50c969b849fc505ded9684f36478907b64d70d1871102be1970a7 +size 343594 diff --git a/parse/dev/S5wmbQc1We/images/47a0c498b5770d8bfa4c34124b0a6ec340016648a966d6625a5affe4d90c69f0.jpg b/parse/dev/S5wmbQc1We/images/47a0c498b5770d8bfa4c34124b0a6ec340016648a966d6625a5affe4d90c69f0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5f5bb15ea01fe719c781714664eb82d5a0ca73aa --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/47a0c498b5770d8bfa4c34124b0a6ec340016648a966d6625a5affe4d90c69f0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c1c8e09349b3c42a9174165399f5c3bd6ce97507ad1dd0c42bac32e6a47b1c9 +size 37924 diff --git a/parse/dev/S5wmbQc1We/images/58674c5336c13104d117005c72a37e55d1073de56a5820d831f5191397e18680.jpg b/parse/dev/S5wmbQc1We/images/58674c5336c13104d117005c72a37e55d1073de56a5820d831f5191397e18680.jpg new file mode 100644 index 0000000000000000000000000000000000000000..35f36ebef76254cf364d1076140a20610669f4ae --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/58674c5336c13104d117005c72a37e55d1073de56a5820d831f5191397e18680.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3afe153cafb59f3ec382a5227a4d7b166fad52bbecf301870d5fb65fa0c6cdd +size 124408 diff --git a/parse/dev/S5wmbQc1We/images/5ac4b2302c6e22ff00f14df38874e91cbd43407065c9487f049eda28ffe47326.jpg b/parse/dev/S5wmbQc1We/images/5ac4b2302c6e22ff00f14df38874e91cbd43407065c9487f049eda28ffe47326.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e05e46c467ecf617f6f9acab14188f1a1f6b50ba --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/5ac4b2302c6e22ff00f14df38874e91cbd43407065c9487f049eda28ffe47326.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3663203943bf2583a67ece27a90a5bcbf219a25ce34697671352b6b8b08abff3 +size 112000 diff --git a/parse/dev/S5wmbQc1We/images/5e37af99bd1fbc6dec5e7d6cf988a6e708ed329f036637611f2cc08df2710cb4.jpg b/parse/dev/S5wmbQc1We/images/5e37af99bd1fbc6dec5e7d6cf988a6e708ed329f036637611f2cc08df2710cb4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e2dc9381cdfd9d1e275cbad9adf7073e92e9a7c3 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/5e37af99bd1fbc6dec5e7d6cf988a6e708ed329f036637611f2cc08df2710cb4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba90674294cefa43fd660b5aad09f0e89baf6115141e2536133316af0f1e360b +size 21280 diff --git a/parse/dev/S5wmbQc1We/images/60dd0aff3a24dafd4477f18335d37e63462d10075ade3f02f05c85e428094e6e.jpg b/parse/dev/S5wmbQc1We/images/60dd0aff3a24dafd4477f18335d37e63462d10075ade3f02f05c85e428094e6e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..427a507acaed46c6ad3584cb12b70fe5c54955bf --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/60dd0aff3a24dafd4477f18335d37e63462d10075ade3f02f05c85e428094e6e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:620b0bff8de2d4f9aa54a29a43d00889cf06173607a49538d974aff21f987473 +size 58435 diff --git a/parse/dev/S5wmbQc1We/images/6a19a13ddcdd757cf6d8f5ef7ee8b2fb27edf3f827fe9c82e37625373f8993c0.jpg b/parse/dev/S5wmbQc1We/images/6a19a13ddcdd757cf6d8f5ef7ee8b2fb27edf3f827fe9c82e37625373f8993c0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..be092c877dc095ea44a9e666e0347a7c6cbfbce9 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/6a19a13ddcdd757cf6d8f5ef7ee8b2fb27edf3f827fe9c82e37625373f8993c0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:558959dab954aa715d4f74fd929fb939bfcb62b14baa12a24c77997e2f54de02 +size 94746 diff --git a/parse/dev/S5wmbQc1We/images/6ac774d02f9716d0114dd5cafb5d09c68f61d47dc99829d7d5cf0d7f2d73b526.jpg b/parse/dev/S5wmbQc1We/images/6ac774d02f9716d0114dd5cafb5d09c68f61d47dc99829d7d5cf0d7f2d73b526.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3d7aaa22a3e1de776bae621486c7a31f8e8312f1 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/6ac774d02f9716d0114dd5cafb5d09c68f61d47dc99829d7d5cf0d7f2d73b526.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9db8dca6dc0a6c48a1ab58caa072a66f1284d8d1e621da089fa9044247dc4e86 +size 10584 diff --git a/parse/dev/S5wmbQc1We/images/6fc3a1254cdc9bffbae427dacb3b40c15f5ab806783ec6a7a10163f3d911a8d6.jpg b/parse/dev/S5wmbQc1We/images/6fc3a1254cdc9bffbae427dacb3b40c15f5ab806783ec6a7a10163f3d911a8d6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9762811a83a529757e38cbc736b0ce2b8fe5abe2 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/6fc3a1254cdc9bffbae427dacb3b40c15f5ab806783ec6a7a10163f3d911a8d6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb0e554038c17595ce810c9b3073892f1b2ab89a747db248401a32136707ff6f +size 7126 diff --git a/parse/dev/S5wmbQc1We/images/6fed1a6631949119b31389b7a5afdee812b60f0e29cd7b57f05276bcd07b88e0.jpg b/parse/dev/S5wmbQc1We/images/6fed1a6631949119b31389b7a5afdee812b60f0e29cd7b57f05276bcd07b88e0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7453acc71dcb23a1ab2dd86585988291c8d4aeee --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/6fed1a6631949119b31389b7a5afdee812b60f0e29cd7b57f05276bcd07b88e0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89ef7ac6428eb6cb470734503320e20bd0b4d0ac494dfae61e1903848145b5da +size 8555 diff --git a/parse/dev/S5wmbQc1We/images/7c6e6391aac371a512032d8b6f9c884cb9bf068d1e4662987fbc2d0ac59f2372.jpg b/parse/dev/S5wmbQc1We/images/7c6e6391aac371a512032d8b6f9c884cb9bf068d1e4662987fbc2d0ac59f2372.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4f50d224dc59d44c2b92dc46f9039031a4671abd --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/7c6e6391aac371a512032d8b6f9c884cb9bf068d1e4662987fbc2d0ac59f2372.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85795d8142585a0278800c993c29496070cc2314215ee3c93c0326e55e7f2dfe +size 3023 diff --git a/parse/dev/S5wmbQc1We/images/8b77cd016ee8b0bb8159304cbcdbfe4da16e61f7d8d6ec157f44a9e89eb71835.jpg b/parse/dev/S5wmbQc1We/images/8b77cd016ee8b0bb8159304cbcdbfe4da16e61f7d8d6ec157f44a9e89eb71835.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f124feb646fd73da70743eed8ab4582642f6c0ce --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/8b77cd016ee8b0bb8159304cbcdbfe4da16e61f7d8d6ec157f44a9e89eb71835.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c8a745eca81e8ba180e1c8e6470594de9d9945e45a829cc909bad3e8aa4ee38d +size 73659 diff --git a/parse/dev/S5wmbQc1We/images/8e3aa8c8d51780d597724dbf1b2353c2c20309c869362e114321b4956eb28135.jpg b/parse/dev/S5wmbQc1We/images/8e3aa8c8d51780d597724dbf1b2353c2c20309c869362e114321b4956eb28135.jpg new file mode 100644 index 0000000000000000000000000000000000000000..eb14970b0961817f6662f17b9b700ba242d2ee08 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/8e3aa8c8d51780d597724dbf1b2353c2c20309c869362e114321b4956eb28135.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e9ad3545a88b6906242d1795f536351ae93b65d22d6a8b0ca3b3551b57f0b5a +size 120204 diff --git a/parse/dev/S5wmbQc1We/images/8fa5b7823ebbc0a6a9ba4028c30569536cb8d63f33ca30d7152bac0d7d0af27e.jpg b/parse/dev/S5wmbQc1We/images/8fa5b7823ebbc0a6a9ba4028c30569536cb8d63f33ca30d7152bac0d7d0af27e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2c47d31891db74f2253f2ae4521f8936200003b3 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/8fa5b7823ebbc0a6a9ba4028c30569536cb8d63f33ca30d7152bac0d7d0af27e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3dd21c4335b950a98f883c7aef29994de1fdc6dad082c8e40a0f8627cd37d4ec +size 12451 diff --git a/parse/dev/S5wmbQc1We/images/919fd114fce35c7ab580a67983c7d5977104770180d8b0cc953aa581423c6c44.jpg b/parse/dev/S5wmbQc1We/images/919fd114fce35c7ab580a67983c7d5977104770180d8b0cc953aa581423c6c44.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d08a7bba83e3e782288eb3934ddc8f625fcf4def --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/919fd114fce35c7ab580a67983c7d5977104770180d8b0cc953aa581423c6c44.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:09d169b7322af0da97cc0613eaaf2648453041c499203cd25c80ee5fc1915a9d +size 7850 diff --git a/parse/dev/S5wmbQc1We/images/99a9e9d2a7254857d61c6bfa24e92e9119b409ba6cc8e144dc9a6963113d1e31.jpg b/parse/dev/S5wmbQc1We/images/99a9e9d2a7254857d61c6bfa24e92e9119b409ba6cc8e144dc9a6963113d1e31.jpg new file mode 100644 index 0000000000000000000000000000000000000000..751dc08a7043cf79b2e2b0a0f65e3f216d7eb5a5 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/99a9e9d2a7254857d61c6bfa24e92e9119b409ba6cc8e144dc9a6963113d1e31.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ceda9be869638afb3ea3787d4a218704cd67fdbf94ffb34f46cd3ac262f9b6a7 +size 18068 diff --git a/parse/dev/S5wmbQc1We/images/9ab56455c63f42b084c7bb69e2ca2ce850030ee26667e696190020375d56223b.jpg b/parse/dev/S5wmbQc1We/images/9ab56455c63f42b084c7bb69e2ca2ce850030ee26667e696190020375d56223b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a87421b956b3f4cfe0cfdea15d0049f24aac6116 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/9ab56455c63f42b084c7bb69e2ca2ce850030ee26667e696190020375d56223b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:801e55f0ba28c5b6f0713eb823ecd51b9468f377ab995954f4bf262c145b5ed8 +size 1575 diff --git a/parse/dev/S5wmbQc1We/images/9d0d3575e176f1f2b5dddb06b63e304a3dfff3496d1d4fc543e627666cb32f3d.jpg b/parse/dev/S5wmbQc1We/images/9d0d3575e176f1f2b5dddb06b63e304a3dfff3496d1d4fc543e627666cb32f3d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0147eef8a915f7556be139d9eacc2a3fcb1c8dd2 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/9d0d3575e176f1f2b5dddb06b63e304a3dfff3496d1d4fc543e627666cb32f3d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1f437b85c54b710f52344aa29f27b07e44000abd38fbc3669d622bed03306e4 +size 15554 diff --git a/parse/dev/S5wmbQc1We/images/9da9ab0d9344b20583129c674adb2863f39dad4ec9b66dc3e3a17080209d2c9f.jpg b/parse/dev/S5wmbQc1We/images/9da9ab0d9344b20583129c674adb2863f39dad4ec9b66dc3e3a17080209d2c9f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..97d950c28378360cdd9407ba4a9d8fc90ba5b306 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/9da9ab0d9344b20583129c674adb2863f39dad4ec9b66dc3e3a17080209d2c9f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3011c56ed6ff70bf2506b499fe3a093f2450d7ddb4b097ebb5a7b9b94d9ca666 +size 13724 diff --git a/parse/dev/S5wmbQc1We/images/a4c16929c8d49f65231b30e3beaaf0cbf51a22e43b383cb260880db63d85c7c4.jpg b/parse/dev/S5wmbQc1We/images/a4c16929c8d49f65231b30e3beaaf0cbf51a22e43b383cb260880db63d85c7c4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..64c28a6c8a56aed70133f613a42e8fe633cb03cc --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/a4c16929c8d49f65231b30e3beaaf0cbf51a22e43b383cb260880db63d85c7c4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65cbc6373de1a457436efd3363040cf94b61c2ed35747bea7bebd63d889a6416 +size 95177 diff --git a/parse/dev/S5wmbQc1We/images/a4c73059a68ba2a9da455769c1f5731c615d66815fd2a58be9ffe4ca6d957dce.jpg b/parse/dev/S5wmbQc1We/images/a4c73059a68ba2a9da455769c1f5731c615d66815fd2a58be9ffe4ca6d957dce.jpg new file mode 100644 index 0000000000000000000000000000000000000000..664803c8dc9534d7d97610c8ed3de150f4a76ca2 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/a4c73059a68ba2a9da455769c1f5731c615d66815fd2a58be9ffe4ca6d957dce.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8b47da6ea9c979d2b26ddb27187bb011bcad47d21fbedcee47b6aa4e25cb8a1a +size 46829 diff --git a/parse/dev/S5wmbQc1We/images/a8e0022db1a64b8295bc2a10d5d9ef3d0796653ed202d0d3d29b73ded004df02.jpg b/parse/dev/S5wmbQc1We/images/a8e0022db1a64b8295bc2a10d5d9ef3d0796653ed202d0d3d29b73ded004df02.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a0f13ad3a01a06e565790b1acb2a5804b7969886 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/a8e0022db1a64b8295bc2a10d5d9ef3d0796653ed202d0d3d29b73ded004df02.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca58b8eb703d13b5f8418c0b54a6b4c8b2997f91352710ce4f95a4bdcfe4a7a7 +size 31669 diff --git a/parse/dev/S5wmbQc1We/images/ab8f6c5657ee7abe49da6fb6d9f1cfe7f621c533c464eb6e9838877566ef324e.jpg b/parse/dev/S5wmbQc1We/images/ab8f6c5657ee7abe49da6fb6d9f1cfe7f621c533c464eb6e9838877566ef324e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b26b351d2aebe60f661e28b68daedbf9ea29e8fe --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/ab8f6c5657ee7abe49da6fb6d9f1cfe7f621c533c464eb6e9838877566ef324e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e186726c9fa01434c3faf85fc95500a6a141456335d373aea1dc2755f8625cfc +size 203398 diff --git a/parse/dev/S5wmbQc1We/images/abdbd9fca314dc3e8e3bcc0f5d0e0c26da0990664e1e5c45f6144e8ec6ec83bd.jpg b/parse/dev/S5wmbQc1We/images/abdbd9fca314dc3e8e3bcc0f5d0e0c26da0990664e1e5c45f6144e8ec6ec83bd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fb70443d26ba84d892781bb3df2761d207bacbb3 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/abdbd9fca314dc3e8e3bcc0f5d0e0c26da0990664e1e5c45f6144e8ec6ec83bd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49053e1b9272fec205d9081bdc56c0fcba3ffe77f81872e8628c3a66b614e0f7 +size 106523 diff --git a/parse/dev/S5wmbQc1We/images/b35466e5619ef7d4f835e41cd63134bdb37213dc402449a21de091d2c5fd7fc8.jpg b/parse/dev/S5wmbQc1We/images/b35466e5619ef7d4f835e41cd63134bdb37213dc402449a21de091d2c5fd7fc8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3db06daae5c2d49b0baf49ee0b227b2c2b5f62a1 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/b35466e5619ef7d4f835e41cd63134bdb37213dc402449a21de091d2c5fd7fc8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d4b5414049d9fbc85fd68bd7a4a5020dbe8eaa47fe6c4f11df24be415e7eefc9 +size 12206 diff --git a/parse/dev/S5wmbQc1We/images/c7b66d3b3104beb6ffda565875d9345bb68581401d1ba011a649067b482cd8d0.jpg b/parse/dev/S5wmbQc1We/images/c7b66d3b3104beb6ffda565875d9345bb68581401d1ba011a649067b482cd8d0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cb8fff10648f5b93a0c0238b6b29e9254270960e --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/c7b66d3b3104beb6ffda565875d9345bb68581401d1ba011a649067b482cd8d0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b651ca1e316b7a3cdda4d3f450fb6c343d9ceafce1b363cdc02676eca346ac98 +size 7899 diff --git a/parse/dev/S5wmbQc1We/images/db427ec9128d569b9efb9358cb5819df97fffd4f84bd382d2986125be94cf974.jpg b/parse/dev/S5wmbQc1We/images/db427ec9128d569b9efb9358cb5819df97fffd4f84bd382d2986125be94cf974.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9f8ee008c7d73ff0f4e77dbfcca81732ee22fa34 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/db427ec9128d569b9efb9358cb5819df97fffd4f84bd382d2986125be94cf974.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cfea821c9f8592c82a700ccb0597d3165a57cc4d9ee04813128da0ea238dce35 +size 70570 diff --git a/parse/dev/S5wmbQc1We/images/dd864c968950275c4c817384eb7a4b3cd7f618e282f2c0dec6ec7cc0c8330af7.jpg b/parse/dev/S5wmbQc1We/images/dd864c968950275c4c817384eb7a4b3cd7f618e282f2c0dec6ec7cc0c8330af7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0b2bae0966de05a8fea13a9c12406bca4d5f6dde --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/dd864c968950275c4c817384eb7a4b3cd7f618e282f2c0dec6ec7cc0c8330af7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4246aa02907052b640c93c95e4511e7573bb1369aff74be104dad1d6e6550858 +size 6987 diff --git a/parse/dev/S5wmbQc1We/images/e855a035fa9f6b20eb042a7751bf3efaaa4efce15ea2db6d154cb1cc3673e02b.jpg b/parse/dev/S5wmbQc1We/images/e855a035fa9f6b20eb042a7751bf3efaaa4efce15ea2db6d154cb1cc3673e02b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f5d60f146b40a51f9195d069624cbc8cbcae4b1f --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/e855a035fa9f6b20eb042a7751bf3efaaa4efce15ea2db6d154cb1cc3673e02b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c6313c397c29c7689d9b2f17fea65f33bd6623cff805e5e0c3f22cec542bd58 +size 8749 diff --git a/parse/dev/S5wmbQc1We/images/e8fd35053c1f30d57d8e0fcbbd9dd4ce144784115afd24206efec7ddd16c126d.jpg b/parse/dev/S5wmbQc1We/images/e8fd35053c1f30d57d8e0fcbbd9dd4ce144784115afd24206efec7ddd16c126d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8a0b44713078445500d8d5a75e837f1eb4470f15 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/e8fd35053c1f30d57d8e0fcbbd9dd4ce144784115afd24206efec7ddd16c126d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a6ae71a7527dbf0bb4cfd0c25e181031fe32b2fcae547971cfa58c586da7ac4 +size 98266 diff --git a/parse/dev/S5wmbQc1We/images/ea4290352b0b65ad356165dd30ffc420464d1a19c9fb175bb23275c71e5bbc11.jpg b/parse/dev/S5wmbQc1We/images/ea4290352b0b65ad356165dd30ffc420464d1a19c9fb175bb23275c71e5bbc11.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c45e4de54e231b895e2a6d410045c099f7a297b2 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/ea4290352b0b65ad356165dd30ffc420464d1a19c9fb175bb23275c71e5bbc11.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:61cc7f2045c5f55319cc0030043968dc6c07b0b98c9f18fe6ea1f6b90fafcab3 +size 68053 diff --git a/parse/dev/S5wmbQc1We/images/f92e5c6e04a7cfb3d8a4380fbc8aa4fd803e2717c044113dc7e0f8e71a6f72d9.jpg b/parse/dev/S5wmbQc1We/images/f92e5c6e04a7cfb3d8a4380fbc8aa4fd803e2717c044113dc7e0f8e71a6f72d9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b8dd11dcd918c6936269bf16443f2c57bf8f535d --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/f92e5c6e04a7cfb3d8a4380fbc8aa4fd803e2717c044113dc7e0f8e71a6f72d9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:56e5e34d48672d77ccecd1ca2385dbddde6960633014876cf7c9dbd89c958d99 +size 2770 diff --git a/parse/dev/S5wmbQc1We/images/fdf88f847c21a28651926eb2c048c93e4eae3af679e27307acd08d82267b3ce1.jpg b/parse/dev/S5wmbQc1We/images/fdf88f847c21a28651926eb2c048c93e4eae3af679e27307acd08d82267b3ce1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3c40de5c487a31d69f26532b96d13ca162525882 --- /dev/null +++ b/parse/dev/S5wmbQc1We/images/fdf88f847c21a28651926eb2c048c93e4eae3af679e27307acd08d82267b3ce1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c53e11a63b7d955291e666238e1e77a60939b424b04135e1a0212e217e390f4b +size 3973 diff --git a/parse/dev/S9GpoS2TmN/images/242e80b9fd952d2ae0f581fe3540fac9d491cc3333c15df42155c80c59b571ce.jpg b/parse/dev/S9GpoS2TmN/images/242e80b9fd952d2ae0f581fe3540fac9d491cc3333c15df42155c80c59b571ce.jpg new file mode 100644 index 0000000000000000000000000000000000000000..46ee0a641ae6c34915c013976fb8eceab9532e56 --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/242e80b9fd952d2ae0f581fe3540fac9d491cc3333c15df42155c80c59b571ce.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:28a33d8ecb9f04a1879e6224b7bc6a2febcf6a00cf5c339a79e1665ad9c9e8ec +size 43950 diff --git a/parse/dev/S9GpoS2TmN/images/2f151772faa38208d3e4ba6e3b2b440b15c1714960fe8086d3729b2a327f937f.jpg b/parse/dev/S9GpoS2TmN/images/2f151772faa38208d3e4ba6e3b2b440b15c1714960fe8086d3729b2a327f937f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d3b392aaef71a26388d1e3784978b0d107093922 --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/2f151772faa38208d3e4ba6e3b2b440b15c1714960fe8086d3729b2a327f937f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f2c77d33a6f14d24ee8aafeae8c4f93e6e07255f2f09559777791687d11e0587 +size 31594 diff --git a/parse/dev/S9GpoS2TmN/images/4846ff8e9eea159678b9816f2f4ed4de783ed589f2ac615d30d502547b5fe577.jpg b/parse/dev/S9GpoS2TmN/images/4846ff8e9eea159678b9816f2f4ed4de783ed589f2ac615d30d502547b5fe577.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fc7877792286853b028eb25ca36f2f603df26034 --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/4846ff8e9eea159678b9816f2f4ed4de783ed589f2ac615d30d502547b5fe577.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5d9abc4363a5e14aee1567b72b61db66a94af8909bf946df3e5903f741a5a557 +size 17919 diff --git a/parse/dev/S9GpoS2TmN/images/4daadbb0ffc53802d4ccd62b2993c153941c868d1369ab4204e36103b733c3f1.jpg b/parse/dev/S9GpoS2TmN/images/4daadbb0ffc53802d4ccd62b2993c153941c868d1369ab4204e36103b733c3f1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3327ddec26c940e4c356d72d314515f26668ca90 --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/4daadbb0ffc53802d4ccd62b2993c153941c868d1369ab4204e36103b733c3f1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b70168c54080d8dbe556705d0a9d8b3c9deac714cea688b991d38a2d1073df63 +size 97031 diff --git a/parse/dev/S9GpoS2TmN/images/4f02598aa374f1e8a9498d9792f9e6d75c084cc0648fbeb54706e59430966364.jpg b/parse/dev/S9GpoS2TmN/images/4f02598aa374f1e8a9498d9792f9e6d75c084cc0648fbeb54706e59430966364.jpg new file mode 100644 index 0000000000000000000000000000000000000000..62d7f44ffe74c5451303b19f8a3ffce85c81f73b --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/4f02598aa374f1e8a9498d9792f9e6d75c084cc0648fbeb54706e59430966364.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:60e8326c8b6c3042b6ea200109f08a2373ab95c854e4f762466fecbb6d63aae6 +size 16656 diff --git a/parse/dev/S9GpoS2TmN/images/60c62ff0d16561fb0d023081d7b0d75f5172cbd18857f1ce3304ebd9e249b057.jpg b/parse/dev/S9GpoS2TmN/images/60c62ff0d16561fb0d023081d7b0d75f5172cbd18857f1ce3304ebd9e249b057.jpg new file mode 100644 index 0000000000000000000000000000000000000000..454acc914b5e119a09e42c78c8564672347dbf42 --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/60c62ff0d16561fb0d023081d7b0d75f5172cbd18857f1ce3304ebd9e249b057.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7551bedb4c16964c18fc794892d8b2fa0b3ccd88570b127737c8c5c8155e73a1 +size 20974 diff --git a/parse/dev/S9GpoS2TmN/images/8b6284c63e21685cbabb765656fc4594dc4cf975ef82cf4b1e4b1b6d051ee966.jpg b/parse/dev/S9GpoS2TmN/images/8b6284c63e21685cbabb765656fc4594dc4cf975ef82cf4b1e4b1b6d051ee966.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5a8d2b5c810cf5bb77f2cf8650b5bfa7049ff59d --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/8b6284c63e21685cbabb765656fc4594dc4cf975ef82cf4b1e4b1b6d051ee966.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:654bb8744578a0625c7ad69c3159b41c873ba0046019f134da4d8aa3056fc297 +size 17658 diff --git a/parse/dev/S9GpoS2TmN/images/98f6b11c0a548cf9602dc3957ee5a215f1afda8698f9e268b3726a5e72200d33.jpg b/parse/dev/S9GpoS2TmN/images/98f6b11c0a548cf9602dc3957ee5a215f1afda8698f9e268b3726a5e72200d33.jpg new file mode 100644 index 0000000000000000000000000000000000000000..593ab5e8867805b3e36b83a5806fb3cd0811d9ec --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/98f6b11c0a548cf9602dc3957ee5a215f1afda8698f9e268b3726a5e72200d33.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b906e0291355087b2cdb8418462523eedb0255f1f7f5e772a824f0f536ebfb1c +size 35749 diff --git a/parse/dev/S9GpoS2TmN/images/9fdf09489919b637df94c2562e27896b2b016f8018cee672a3ea3274fbe40a26.jpg b/parse/dev/S9GpoS2TmN/images/9fdf09489919b637df94c2562e27896b2b016f8018cee672a3ea3274fbe40a26.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2a27074f2193deb6fa03b000078867c4fa532f00 --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/9fdf09489919b637df94c2562e27896b2b016f8018cee672a3ea3274fbe40a26.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:11fdbb9644788d78e82c36b6441c8cd73ed4fc6fd87ed4d256d216f906a686ee +size 13161 diff --git a/parse/dev/S9GpoS2TmN/images/ae6f8477912fdf23fb4d54056340012a17e71a09760c7774515b4eac6b2510c1.jpg b/parse/dev/S9GpoS2TmN/images/ae6f8477912fdf23fb4d54056340012a17e71a09760c7774515b4eac6b2510c1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2904882f8956e0252b751bc78bf1f6301a09c2fb --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/ae6f8477912fdf23fb4d54056340012a17e71a09760c7774515b4eac6b2510c1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d24498cd0feefb0b3cc63ad43007b3126cb9e6d71fcf6926a59cd9471ac7ac7 +size 43608 diff --git a/parse/dev/S9GpoS2TmN/images/bf4ae3a90173e3f7a6d374a5473f1b308ea397bcc93262a6b455f4a8f4b46d13.jpg b/parse/dev/S9GpoS2TmN/images/bf4ae3a90173e3f7a6d374a5473f1b308ea397bcc93262a6b455f4a8f4b46d13.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b74820db1f596b3919ec3bea3f201d0d1273caf1 --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/bf4ae3a90173e3f7a6d374a5473f1b308ea397bcc93262a6b455f4a8f4b46d13.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59320297b434b63d4cb2ff90153649e207d45bd93bce89cfaede80a23a69d678 +size 129621 diff --git a/parse/dev/S9GpoS2TmN/images/d99a65d2dfbe63dad3230e2cb1d43f714655dc7f3724706562d79be5f7816096.jpg b/parse/dev/S9GpoS2TmN/images/d99a65d2dfbe63dad3230e2cb1d43f714655dc7f3724706562d79be5f7816096.jpg new file mode 100644 index 0000000000000000000000000000000000000000..99a4748ca3181f05aaa9939f04e684ccb4c2707b --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/d99a65d2dfbe63dad3230e2cb1d43f714655dc7f3724706562d79be5f7816096.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dfa2cb8c3e9c0a9bd7b39b97aeb9402077aff34bcffac72efeddaf956a0e2626 +size 23686 diff --git a/parse/dev/S9GpoS2TmN/images/df6f28ba81bad3d943985dd9bf006d28c398921cb8248030c712ba7ea236a018.jpg b/parse/dev/S9GpoS2TmN/images/df6f28ba81bad3d943985dd9bf006d28c398921cb8248030c712ba7ea236a018.jpg new file mode 100644 index 0000000000000000000000000000000000000000..26645e81182faba45d5597dfc50066bfa84fb89e --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/df6f28ba81bad3d943985dd9bf006d28c398921cb8248030c712ba7ea236a018.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1805b3fadfdcd71d3e01beed5475bce876f14a43a30651c1b6a292eb4c3cdcf +size 26331 diff --git a/parse/dev/S9GpoS2TmN/images/f9426fa8a28e23241ba25e2f6229222b1a278adc4e56968bb580f8316a46a808.jpg b/parse/dev/S9GpoS2TmN/images/f9426fa8a28e23241ba25e2f6229222b1a278adc4e56968bb580f8316a46a808.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9c99e9d282a20641310ea618a2d86439e5acbb69 --- /dev/null +++ b/parse/dev/S9GpoS2TmN/images/f9426fa8a28e23241ba25e2f6229222b1a278adc4e56968bb580f8316a46a808.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5f86d88aee8bd832b4a1e29fa5712f28f280ba5ccae43872bc78c4e92c11a056 +size 10867 diff --git a/parse/dev/SMa9EAovKMC/images/027b72acf9437d01f10744a3841881836cb9b16c4928ad7d09dba4b3221a05fe.jpg b/parse/dev/SMa9EAovKMC/images/027b72acf9437d01f10744a3841881836cb9b16c4928ad7d09dba4b3221a05fe.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ea093b30df8600b3cde39f8b39061aed304a7245 --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/027b72acf9437d01f10744a3841881836cb9b16c4928ad7d09dba4b3221a05fe.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8278c0db99c40759be538d035d8fc5f2a4fb4ea0b877b9bd4973aad02849840a +size 74476 diff --git a/parse/dev/SMa9EAovKMC/images/0618c1b53c8f61ad2c1ace44a644ca3c71dd722268b6931259a757c6a277fb43.jpg b/parse/dev/SMa9EAovKMC/images/0618c1b53c8f61ad2c1ace44a644ca3c71dd722268b6931259a757c6a277fb43.jpg new file mode 100644 index 0000000000000000000000000000000000000000..02536441313a1f15f367212292be111f4de86331 --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/0618c1b53c8f61ad2c1ace44a644ca3c71dd722268b6931259a757c6a277fb43.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:356fa8b1a4c3a9c651e61de81260513f769cc5e07e2c3c97712c1034521b5826 +size 69137 diff --git a/parse/dev/SMa9EAovKMC/images/237540f7093980291b4817da7f69f22cfd34648803a0159cbea5b01153532362.jpg b/parse/dev/SMa9EAovKMC/images/237540f7093980291b4817da7f69f22cfd34648803a0159cbea5b01153532362.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9787c289234eb506d8cc49235773fbb990e08e52 --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/237540f7093980291b4817da7f69f22cfd34648803a0159cbea5b01153532362.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e43ca21a1f4672739abacdab99b2c76ff7bfedbb6bc80f1d91a138a887208e92 +size 124287 diff --git a/parse/dev/SMa9EAovKMC/images/3851a79bddd181a8d136d7db1f8f66013f998acb1a0c5e97ed31bc96d19cd07f.jpg b/parse/dev/SMa9EAovKMC/images/3851a79bddd181a8d136d7db1f8f66013f998acb1a0c5e97ed31bc96d19cd07f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1a0087ee68736c3cac6d7a182c2048db463413a8 --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/3851a79bddd181a8d136d7db1f8f66013f998acb1a0c5e97ed31bc96d19cd07f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16b2d423ddf61bc9d2a5e0a5f89da83da69fac48b36f30b61c12003c264f7ad3 +size 79378 diff --git a/parse/dev/SMa9EAovKMC/images/4c5f9af1eec5ff0dbd3c26f59b1e423b93ac8e24f21200fbabbc41b1ed1a84de.jpg b/parse/dev/SMa9EAovKMC/images/4c5f9af1eec5ff0dbd3c26f59b1e423b93ac8e24f21200fbabbc41b1ed1a84de.jpg new file mode 100644 index 0000000000000000000000000000000000000000..47165a22d6fd13d262d18559ea4c7f8d2ecf17f6 --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/4c5f9af1eec5ff0dbd3c26f59b1e423b93ac8e24f21200fbabbc41b1ed1a84de.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6b0ec3a7575fd64c943c0ff4fff09a3cd4b57f8077d5123dcee3511f1b21aa6 +size 5884 diff --git a/parse/dev/SMa9EAovKMC/images/67573040b0afaac842bb1650e940a0f13e71ffaa603cf29431ea6b657708411d.jpg b/parse/dev/SMa9EAovKMC/images/67573040b0afaac842bb1650e940a0f13e71ffaa603cf29431ea6b657708411d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bea4d3ca1a47cc8826729f4c69354dea0bf95f9b --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/67573040b0afaac842bb1650e940a0f13e71ffaa603cf29431ea6b657708411d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef4ec0b6d14033b5e20f67d615294a02053a03e2481dd657da90a57303e50872 +size 85406 diff --git a/parse/dev/SMa9EAovKMC/images/82d8a4edfd8b3e05c239256e73e735c3e6d39fe93855d6ff3c573803f7701d6c.jpg b/parse/dev/SMa9EAovKMC/images/82d8a4edfd8b3e05c239256e73e735c3e6d39fe93855d6ff3c573803f7701d6c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c2f604cb92df454e6281af81e5974d056e1455b0 --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/82d8a4edfd8b3e05c239256e73e735c3e6d39fe93855d6ff3c573803f7701d6c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f0b808154644f771b1abdb4c5b9eafa10d05e3dd8df276758a016f5852d57b4 +size 3915 diff --git a/parse/dev/SMa9EAovKMC/images/a639aada5b0da5fa6bd0ea04c398c4150cac20b60874ca139a9460b2eabb4cf9.jpg b/parse/dev/SMa9EAovKMC/images/a639aada5b0da5fa6bd0ea04c398c4150cac20b60874ca139a9460b2eabb4cf9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a0e6efb032c53589727a2028e1c8de1465ce0b78 --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/a639aada5b0da5fa6bd0ea04c398c4150cac20b60874ca139a9460b2eabb4cf9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f3569cddfc593bdb098441143cbd0ace85f4091d6aaa74e199be9b1655ed4582 +size 6762 diff --git a/parse/dev/SMa9EAovKMC/images/aa460b3fb6dc8b7e32e9684370859e3189d97a411eb1fccc59693dc5150e9c9d.jpg b/parse/dev/SMa9EAovKMC/images/aa460b3fb6dc8b7e32e9684370859e3189d97a411eb1fccc59693dc5150e9c9d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0f4535ef268b62d8803cfe810459c6ae4b66b5a4 --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/aa460b3fb6dc8b7e32e9684370859e3189d97a411eb1fccc59693dc5150e9c9d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7127a7741f498c058dd7f0a30ddaa06baca78206679c58a882ec166854cf79ec +size 153593 diff --git a/parse/dev/SMa9EAovKMC/images/bb509eddceecc2747f0838e5b5ced68ff1a5bb31bd80def3541411c3847b77b3.jpg b/parse/dev/SMa9EAovKMC/images/bb509eddceecc2747f0838e5b5ced68ff1a5bb31bd80def3541411c3847b77b3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e185f312aeb335eccad8751af3560f226b5ab696 --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/bb509eddceecc2747f0838e5b5ced68ff1a5bb31bd80def3541411c3847b77b3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91ab456280ea5dd70430a115353dee84ac088fe3be6bb8a739b2096019acc39b +size 3867 diff --git a/parse/dev/SMa9EAovKMC/images/ce7d67bea877c9c19cbd7920c391c58a692f5c0e16fe0099748943b319c156bc.jpg b/parse/dev/SMa9EAovKMC/images/ce7d67bea877c9c19cbd7920c391c58a692f5c0e16fe0099748943b319c156bc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3614774248782bab46326e730fcbca402c0ea0b2 --- /dev/null +++ b/parse/dev/SMa9EAovKMC/images/ce7d67bea877c9c19cbd7920c391c58a692f5c0e16fe0099748943b319c156bc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3263383d7fead22567ea0194059725abe55750b58d9ae66bf0bb6567a90b26b7 +size 3767 diff --git a/parse/dev/SVBR6xBaMl/images/0315364352fa9997526daf93acc7fe13adfaf008a332dfad035212c45148a048.jpg b/parse/dev/SVBR6xBaMl/images/0315364352fa9997526daf93acc7fe13adfaf008a332dfad035212c45148a048.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e9fb4da3c6f6967bd2370f57341053ef96d4b625 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/0315364352fa9997526daf93acc7fe13adfaf008a332dfad035212c45148a048.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a39dddeb7611c138bb0e792796fb31c0d999fbbce292a17967482d130daa16e +size 21493 diff --git a/parse/dev/SVBR6xBaMl/images/11c7720c277528a91abebbbfa96fb755554c4eae4f41e9c3f4214355f5ce2f91.jpg b/parse/dev/SVBR6xBaMl/images/11c7720c277528a91abebbbfa96fb755554c4eae4f41e9c3f4214355f5ce2f91.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e51d37591660184a3a7b966756240eb11b4afc37 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/11c7720c277528a91abebbbfa96fb755554c4eae4f41e9c3f4214355f5ce2f91.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:445a7d67750588c005819a6a3bb0b7f12e3539798a503571dc0d8bf9dbd59826 +size 51156 diff --git a/parse/dev/SVBR6xBaMl/images/275216cb66957f954aa88126a2f15a0b1352f5c0e40efe590fd281148140eb01.jpg b/parse/dev/SVBR6xBaMl/images/275216cb66957f954aa88126a2f15a0b1352f5c0e40efe590fd281148140eb01.jpg new file mode 100644 index 0000000000000000000000000000000000000000..08d8322b7e42f9fa9e1310638eebc32729789a74 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/275216cb66957f954aa88126a2f15a0b1352f5c0e40efe590fd281148140eb01.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:81e3a44ba94090b00a7683e6382dcd843e88470c6f8c1293673b29666548fc98 +size 34924 diff --git a/parse/dev/SVBR6xBaMl/images/2f04714f07d0d49c9ff7c7e2c767cb5bf7cab3781c7ee57da7364533ec0e4756.jpg b/parse/dev/SVBR6xBaMl/images/2f04714f07d0d49c9ff7c7e2c767cb5bf7cab3781c7ee57da7364533ec0e4756.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f5b0aa0596940f90954027c3c3ae63b36b831ff7 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/2f04714f07d0d49c9ff7c7e2c767cb5bf7cab3781c7ee57da7364533ec0e4756.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:174854de2ef09bb147b437238dc50207cdef4aec10060a6992f1fb53dfd16afb +size 80220 diff --git a/parse/dev/SVBR6xBaMl/images/42eec7f27a34e051f92d074e20e6430c8ded7c1870db307ce1382f06517b2970.jpg b/parse/dev/SVBR6xBaMl/images/42eec7f27a34e051f92d074e20e6430c8ded7c1870db307ce1382f06517b2970.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b55958789eedb0a65e9af921551f479f3e8eed97 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/42eec7f27a34e051f92d074e20e6430c8ded7c1870db307ce1382f06517b2970.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13c562d0ce9e6cf8c6060664251393e7e8602bff2c1fdd66844462cf201b286f +size 51232 diff --git a/parse/dev/SVBR6xBaMl/images/48127db09078c138c5ee22cba8b52b99be6c353556c803e601e037cc49d26d6f.jpg b/parse/dev/SVBR6xBaMl/images/48127db09078c138c5ee22cba8b52b99be6c353556c803e601e037cc49d26d6f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b27e06cb5d56ed3b83a5c3d6b1f157377e5dbd5c --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/48127db09078c138c5ee22cba8b52b99be6c353556c803e601e037cc49d26d6f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8ff4b2829e6ba7901baa6ac110a583a08b55f1bfa92c678684a1672ad22c2ff +size 23999 diff --git a/parse/dev/SVBR6xBaMl/images/4b95a04184f63a2987345fc245ced2ff3c5f3d08a5d71423e9949fb762ca84fa.jpg b/parse/dev/SVBR6xBaMl/images/4b95a04184f63a2987345fc245ced2ff3c5f3d08a5d71423e9949fb762ca84fa.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7b265aedf4263cf828eadcb1a996e956c8cb50a2 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/4b95a04184f63a2987345fc245ced2ff3c5f3d08a5d71423e9949fb762ca84fa.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bf9bf8b35d76360eb014db01ce8094e69f4bf149a22347ae522e1ef9a63f9406 +size 4277 diff --git a/parse/dev/SVBR6xBaMl/images/4e0bead7ec141fc8b9e0c58f4e6b0ce9be8533f2eeda88dd312d9a2b32099255.jpg b/parse/dev/SVBR6xBaMl/images/4e0bead7ec141fc8b9e0c58f4e6b0ce9be8533f2eeda88dd312d9a2b32099255.jpg new file mode 100644 index 0000000000000000000000000000000000000000..867cd97f407ae0fc35eaa1e24a3a0d57fa973c92 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/4e0bead7ec141fc8b9e0c58f4e6b0ce9be8533f2eeda88dd312d9a2b32099255.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d02545e6fced77d036d8bb1b1eb508636548f5e2dcf352f578f9fb11087df87b +size 45902 diff --git a/parse/dev/SVBR6xBaMl/images/582adf351c1a177065f7095a77b9cd69eeb49193ccd0806ea1c411dacf3cd994.jpg b/parse/dev/SVBR6xBaMl/images/582adf351c1a177065f7095a77b9cd69eeb49193ccd0806ea1c411dacf3cd994.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9897cd775ad0f512caa10b5ef300a3a82f3804cf --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/582adf351c1a177065f7095a77b9cd69eeb49193ccd0806ea1c411dacf3cd994.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c169c463d5ab0068c5b9bab40cf5bb1e4e845b4d36796db1f12f075ea2c9f0b6 +size 5921 diff --git a/parse/dev/SVBR6xBaMl/images/651ef59e1b6bb61a2f3868d2cb3cdd0eca3ebd20411174075654384e803778ea.jpg b/parse/dev/SVBR6xBaMl/images/651ef59e1b6bb61a2f3868d2cb3cdd0eca3ebd20411174075654384e803778ea.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e9d62ab64a1ec6d11ea6fea7c5c776394b73d29b --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/651ef59e1b6bb61a2f3868d2cb3cdd0eca3ebd20411174075654384e803778ea.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:717b644dd37e877c24c345145b134482110abf1bec37b28c5213806fbe51099d +size 55510 diff --git a/parse/dev/SVBR6xBaMl/images/83016dd2e1f169de33aa09268523e0445aa43ebab5f2c67ca4d2de4e10e753c3.jpg b/parse/dev/SVBR6xBaMl/images/83016dd2e1f169de33aa09268523e0445aa43ebab5f2c67ca4d2de4e10e753c3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dd2f3ed366b2e219a3780923d64590fec2a01640 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/83016dd2e1f169de33aa09268523e0445aa43ebab5f2c67ca4d2de4e10e753c3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e5123aacd33c0642c9e535fab9daa70519f4ac80c462017609c2f7bafeef6f2 +size 87093 diff --git a/parse/dev/SVBR6xBaMl/images/8357e516c8a5e09ac825bec5e30339316281faf920f6f27dcf45244f9284cb62.jpg b/parse/dev/SVBR6xBaMl/images/8357e516c8a5e09ac825bec5e30339316281faf920f6f27dcf45244f9284cb62.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e5162a948ce32138671078a659369a247d25f0f0 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/8357e516c8a5e09ac825bec5e30339316281faf920f6f27dcf45244f9284cb62.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b8c450a68c3a0cca8e89d1de6b9371fafdbecb917cbf92c4157f9df98f97c59 +size 41042 diff --git a/parse/dev/SVBR6xBaMl/images/88a5b1529cd3512fab4520ce09e2e023f102b2afe5c85a3549f46f35b981416d.jpg b/parse/dev/SVBR6xBaMl/images/88a5b1529cd3512fab4520ce09e2e023f102b2afe5c85a3549f46f35b981416d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..38fe572e1c55b9158b3627454b3c43acc2eaf2fc --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/88a5b1529cd3512fab4520ce09e2e023f102b2afe5c85a3549f46f35b981416d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02248654c0783d280358d49632bb783c609a547676e15692f8f63dcf892dde83 +size 54608 diff --git a/parse/dev/SVBR6xBaMl/images/999d8fd25c2c9facc3d57d61e77984bec45e40471c60c5f02a557069d3ad06cd.jpg b/parse/dev/SVBR6xBaMl/images/999d8fd25c2c9facc3d57d61e77984bec45e40471c60c5f02a557069d3ad06cd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2cf27c143626ebc10a3a71d36220b8ff3751f693 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/999d8fd25c2c9facc3d57d61e77984bec45e40471c60c5f02a557069d3ad06cd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:da67b3d08e8de676953d74f00ef604186bf0e167b2023c28b66e2d16decd3aaf +size 15540 diff --git a/parse/dev/SVBR6xBaMl/images/9bb9b5a4704b8c43c1180e9d538cae1a109582244ffadecc158bbbfc53c3abbf.jpg b/parse/dev/SVBR6xBaMl/images/9bb9b5a4704b8c43c1180e9d538cae1a109582244ffadecc158bbbfc53c3abbf.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e24948d1163c105171cf2825114638d2dc72a48e --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/9bb9b5a4704b8c43c1180e9d538cae1a109582244ffadecc158bbbfc53c3abbf.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c438a4f7cd96627514644c2b486c4b3c345870b43c5b5f27cf35ab9bc051621 +size 23975 diff --git a/parse/dev/SVBR6xBaMl/images/a34d951ae285ff164258d6ef23e5ed573e3272e65f00f1ea7c4b1dd45bfe6e00.jpg b/parse/dev/SVBR6xBaMl/images/a34d951ae285ff164258d6ef23e5ed573e3272e65f00f1ea7c4b1dd45bfe6e00.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6d8c3ed8c0b3ad906745a54480c226fc9b1cb716 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/a34d951ae285ff164258d6ef23e5ed573e3272e65f00f1ea7c4b1dd45bfe6e00.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b368dcdb983e065a3646b6bdf0835fce5be8bf973cd82b3e38f0255e0d468e30 +size 31852 diff --git a/parse/dev/SVBR6xBaMl/images/a9626066bb330cf1bbf2c2cd19b8434b2d6dc5342d062e9244c7580b38c03b2c.jpg b/parse/dev/SVBR6xBaMl/images/a9626066bb330cf1bbf2c2cd19b8434b2d6dc5342d062e9244c7580b38c03b2c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1fab2aebb7413481f09dc126c98cd8e9cf9cec04 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/a9626066bb330cf1bbf2c2cd19b8434b2d6dc5342d062e9244c7580b38c03b2c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27157ff5f38636430418dd11dec2b574ba66e74b0f92a27dc4002b14e1658dff +size 158919 diff --git a/parse/dev/SVBR6xBaMl/images/b0851ebc0b5b21b242b340e237a3de6694b10d70730f2f90b0c2c7ba131cc795.jpg b/parse/dev/SVBR6xBaMl/images/b0851ebc0b5b21b242b340e237a3de6694b10d70730f2f90b0c2c7ba131cc795.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e5f15091627aa0a1d2f1fa7acb96a159ab474049 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/b0851ebc0b5b21b242b340e237a3de6694b10d70730f2f90b0c2c7ba131cc795.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f44acb38af3ca3aefc4504705a80ae912dacd911ab3f8175f394cfced366939e +size 10714 diff --git a/parse/dev/SVBR6xBaMl/images/b74ea754fb2c7536c54d95fa7817f8c06a373edaaa94620fc080e21da644a6cf.jpg b/parse/dev/SVBR6xBaMl/images/b74ea754fb2c7536c54d95fa7817f8c06a373edaaa94620fc080e21da644a6cf.jpg new file mode 100644 index 0000000000000000000000000000000000000000..600641e5d9a2a3e2148d9a11f1d0d3c31db9d140 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/b74ea754fb2c7536c54d95fa7817f8c06a373edaaa94620fc080e21da644a6cf.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b42dadf411e027e50f387df80fcda82f46b46fc70f368d6b2dd14dcbe1b99cd6 +size 19862 diff --git a/parse/dev/SVBR6xBaMl/images/cc468b069ceb7644721c432df78d495e02afe63b26d44f5633999786d4faacf5.jpg b/parse/dev/SVBR6xBaMl/images/cc468b069ceb7644721c432df78d495e02afe63b26d44f5633999786d4faacf5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..61f6d5b22f0f507e2ccb865d41037ec039f88a25 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/cc468b069ceb7644721c432df78d495e02afe63b26d44f5633999786d4faacf5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:71b1a79d1c441b3df7c081710c23ecaa6112897d317550c5df5446c475127dfb +size 44934 diff --git a/parse/dev/SVBR6xBaMl/images/da028bb58decb70bb73953ef2507216a4e235cc3ebd144f35477fd4435f5f78c.jpg b/parse/dev/SVBR6xBaMl/images/da028bb58decb70bb73953ef2507216a4e235cc3ebd144f35477fd4435f5f78c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..26eaad0928c1abff87969730cb911aabbbbe8e4c --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/da028bb58decb70bb73953ef2507216a4e235cc3ebd144f35477fd4435f5f78c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:30443e1e19e1c03cf1d4b0dc84b58d10b550401b1fcfb1c08177d6582b4a65f4 +size 93696 diff --git a/parse/dev/SVBR6xBaMl/images/e83ba1864bc0424a11c96f8387cfef1cca04f8b6c0716e0414adef3b8b8b6e91.jpg b/parse/dev/SVBR6xBaMl/images/e83ba1864bc0424a11c96f8387cfef1cca04f8b6c0716e0414adef3b8b8b6e91.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ec55b56a9b5fbb2d79de5ef9bab73a5090dde23b --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/e83ba1864bc0424a11c96f8387cfef1cca04f8b6c0716e0414adef3b8b8b6e91.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b55ef129ac210da3be9c4f2e676aec2749a087afb88548e028f1c27d50f2c604 +size 100433 diff --git a/parse/dev/SVBR6xBaMl/images/eb0fe736a63e3c6eaed20f7f7794264845222af292c4d1ef15494432809c66ed.jpg b/parse/dev/SVBR6xBaMl/images/eb0fe736a63e3c6eaed20f7f7794264845222af292c4d1ef15494432809c66ed.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ddf40f20ff23d444518c75cc0330a6b4afc531ae --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/eb0fe736a63e3c6eaed20f7f7794264845222af292c4d1ef15494432809c66ed.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8a5893160cc0bdb555c346d6a52b9b5b8a8a47c59e4917c662876a3c22cba17 +size 13337 diff --git a/parse/dev/SVBR6xBaMl/images/f82304a5fcb1f4f4644a875ee29983223996dba4786b0532a5de2c00bfa53748.jpg b/parse/dev/SVBR6xBaMl/images/f82304a5fcb1f4f4644a875ee29983223996dba4786b0532a5de2c00bfa53748.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fec69fe61f84a63e43375755eb42bf7e09866961 --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/f82304a5fcb1f4f4644a875ee29983223996dba4786b0532a5de2c00bfa53748.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:287a30a0a7a3182a1ad82563204274f95fdb2ebf44a6b3f17c566390d9c62a09 +size 24646 diff --git a/parse/dev/SVBR6xBaMl/images/f82ccd1b0716f29628152b96d724c7fc813e6d3f8e9c73cc37e68cae0220c848.jpg b/parse/dev/SVBR6xBaMl/images/f82ccd1b0716f29628152b96d724c7fc813e6d3f8e9c73cc37e68cae0220c848.jpg new file mode 100644 index 0000000000000000000000000000000000000000..69e88b44efb782199c17a3b89b725df9fc17701a --- /dev/null +++ b/parse/dev/SVBR6xBaMl/images/f82ccd1b0716f29628152b96d724c7fc813e6d3f8e9c73cc37e68cae0220c848.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87015a7067a4cf0b431669365131ed84cd5caef104524206ca1f13f3531b8cbb +size 58021 diff --git a/parse/dev/SfXjt1FtMQ/SfXjt1FtMQ_content_list.json b/parse/dev/SfXjt1FtMQ/SfXjt1FtMQ_content_list.json new file mode 100644 index 0000000000000000000000000000000000000000..5949d886b06d6204f0b0cc6696e50f0d2b756d59 --- /dev/null +++ b/parse/dev/SfXjt1FtMQ/SfXjt1FtMQ_content_list.json @@ -0,0 +1,892 @@ +[ + { + "type": "text", + "text": "GmGM: a fast Gaussian graphical model for multi-modal data ", + "text_level": 1, + "bbox": [ + 228, + 122, + 769, + 171 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Anonymous Author(s) \nAffiliation \nAddress \nemail ", + "bbox": [ + 423, + 222, + 578, + 276 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Abstract ", + "text_level": 1, + "bbox": [ + 462, + 313, + 535, + 329 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "1 This paper introduces the Gaussian multi-Graphical Model, a model to construct \n2 sparse graph representations of matrix- and tensor-variate data. We generalize \n3 prior work in this area by simultaneously learning this representation across several \n4 tensors that share axes, which is necessary to allow the analysis of multimodal \n5 datasets such as those encountered in multi-omics. Our algorithm uses only a \n6 single eigendecomposition per axis, achieving an order of magnitude speedup over \n7 prior work in the ungeneralized case. This allows the use of our methodology \n8 on large multi-modal datasets such as single-cell multi-omics data, which was \n9 challenging with previous approaches. We validate our model on synthetic data \n10 and five real-world datasets. ", + "bbox": [ + 148, + 343, + 766, + 482 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "11 1 Introduction ", + "text_level": 1, + "bbox": [ + 148, + 506, + 312, + 522 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "12 A number of modern applications require the estimation of networks (graphs) exploring the de \n13 pendency structures underlying the data. In this paper, we propose a new approach for estimating \n14 conditional dependency graphs. Two datapoints $x , y$ are conditionally independent (with respect to \n15 a dataset $\\mathcal { D }$ ) if knowing one provides no information about the other that is not already contained \n16 in the rest of the dataset: $\\mathbb { P } ( \\bar { x } | y , \\mathcal { D } _ { \\backslash x y } ) = \\mathbb { P } ( x | \\mathcal { D } _ { \\backslash x y } )$ . For normally distributed data, conditional \n17 dependencies are encoded in the inverse of the covariance matrix (the ‘precision’ matrix). Two \n18 datapoints are conditionally dependent on each other if and only if their corresponding element in the \n19 precision matrix is not zero. If our dataset were in the form of a vector $\\mathbf { d }$ , we could then model it as \n20 $\\mathbf { \\dot { d } } \\sim \\mathcal { N } ( \\mathbf { 0 } , \\Psi ^ { - 1 } )$ for precision matrix $\\Psi$ . This is a Gaussian Graphical Model (GGM); $\\Psi$ encodes \n21 the graph. \n22 However, datasets are often more structured than vectors. For example, single-cell RNA sequencing \n23 datasets (scRNA-seq) come in the form of a matrix of gene expression counts whose rows are cells \n24 and columns are genes. Video data naturally requires a third-order tensor of pixels to represent \n25 it - rows, columns, and frames. Furthermore multi-omics datasets such as those including both \n26 scRNA-seq and scATAC-seq may require two or more matrices to be properly represented; one for \n27 each modality. \n28 We could assume that each row of our matrix is an i.i.d. sample drawn from our model. However, \n29 independence is a strong and often incorrect assumption. If we wanted to make no independence \n30 assumptions, we could vectorize the dataset $\\mathcal { D }$ and estimate $\\Psi$ in $\\mathrm { v e c } [ \\mathcal { D } ] \\sim \\mathcal { N } ( \\mathbf { 0 } , \\Psi ^ { - 1 } )$ . However, \n31 this produces intractably large $\\Psi$ , whose number of elements is quadratic in the product of the lengths \n32 of our dataset’s axes. \n33 Thankfully, tensors are highly structured, and we are often interested in the dependency structure \n34 of each axis individually - i.e. the dependencies between samples or the dependencies between \n35 features - rather than the dependencies between the elements of the tensor themselves. To model this, \n36 we can represent $\\Psi$ as some deterministic combination of the axis-wise dependencies: $\\mathrm { v e c } [ \\mathbf { D } ] \\sim$ \n37 $\\mathcal { N } ( \\mathbf { 0 } , \\zeta ( \\Psi _ { \\mathrm { r o w } } , \\Psi _ { \\mathrm { c o l } } ) ^ { - 1 } )$ , for some function $\\zeta$ . The strategy is to estimate $\\Psi _ { \\mathrm { r o w } }$ , $\\Psi _ { \\mathrm { c o l } }$ directly, without \n38 computing the intractable $\\zeta ( \\Psi _ { \\mathrm { r o w } } , \\Psi _ { \\mathrm { c o l } } ) ^ { - \\mathrm { i } }$ . While there are multiple choices for $\\zeta$ , this paper \n39 considers only the Kronecker sum. ", + "bbox": [ + 147, + 536, + 825, + 675 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "", + "bbox": [ + 147, + 681, + 825, + 765 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "", + "bbox": [ + 147, + 771, + 825, + 840 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "", + "bbox": [ + 147, + 847, + 825, + 902 + ], + "page_idx": 0 + }, + { + "type": "image", + "img_path": "images/33ca9ec56b1460ae7897b2c4509fe84fab43453d8ccbfec3701b0b0a8e8e6934.jpg", + "image_caption": [ + "Figure 1: The two matrices of the LifeLines-DEEP dataset. As both matrices include data for the same people, the learned graph between people should be the same. " + ], + "image_footnote": [], + "bbox": [ + 266, + 84, + 727, + 252 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "", + "bbox": [ + 147, + 318, + 825, + 361 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "40 1.1 Prior work ", + "text_level": 1, + "bbox": [ + 147, + 381, + 289, + 395 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "41 The Kronecker sum BiGraphical Lasso (BiGLasso) model was first considered by Kalaitzis et al. \n42 [14]. BiGLasso is the multi-axis analog to graphical lasso methods [10], which are used to estimate \n43 covariance matrices of data drawn from a multivariate Gaussian distribution. The Kronecker sum \n44 of two matrices, $\\mathbf { A } \\oplus \\mathbf { B }$ , can be expressed in terms of Kronecker products: $\\mathbf { A } \\otimes \\mathbf { I } + \\mathbf { I } \\otimes \\mathbf { B }$ . When \n45 the matrices $\\mathbf { A } , \\mathbf { B }$ are adjacency matrices of graphs, the Kronecker sum has the interpretation as \n46 the Cartesian product of those graphs. This sum is one choice $\\zeta$ to combine the per-axis precision \n47 matrices into the precision matrix of the vectorized dataset, $\\mathrm { v e c } [ \\mathbf { \\bar { D } } ] \\sim \\mathcal { N } ( \\mathbf { 0 } , ( \\Psi _ { \\mathrm { r o w } } \\mathbf { \\hat { \\boldsymbol { \\Phi } } } \\boldsymbol { \\Psi } _ { \\mathrm { c o l } } \\mathbf { \\bar { \\rho } } ) ^ { - 1 } )$ . \n48 Other choices for $\\zeta$ have been considered, such as using the Kronecker product [23, 8], or the square \n49 of the Kronecker sum [24, 25]. Each method has its strengths; the benefits of a Kronecker sum \n50 structure are its interpretability as a graph product, stronger sparsity, and its allowance of inter-task \n51 transfer [14]. \n52 The original BiGLasso model was very slow to converge to a solution, in large part due to its non \n53 optimal space complexity of $O ( n ^ { 2 } p ^ { 2 } )$ . This prohibited its use on large datasets (measuring in a \n54 couple hundred samples and/or features). Numerous modifications have been made to the algorithm \n55 to improve its speed and achieve an optimal space complexity of $O ( n ^ { 2 } + p ^ { 2 } )$ , such as scBiGLasso \n56 [17], TeraLasso [12], and EiGLasso [27]. Of these, TeraLasso is notable in that it generalizes to an \n57 arbitrary number of axes, i.e. $\\zeta ( \\Psi _ { 1 } , . . . , \\Psi _ { k } ) = \\Psi _ { 1 } \\oplus . . . \\oplus \\Psi _ { k }$ . TeraLasso and EiGLasso, the fastest \n58 prior work, both rely on computing an eigendecomposition every iteration. \n59 All of these algorithms and models, including our own, rely on a normality assumption. We are most \n60 interested in the case of omics data, in which case a log-transform renders our dataset sufficiently \n61 Gaussian-like for our algorithm to achieve good performance. An overview of the use of GGMs in \n62 omics data is given by Altenbuchinger et al. [2]. ", + "bbox": [ + 147, + 407, + 825, + 506 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "", + "bbox": [ + 147, + 511, + 825, + 566 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "", + "bbox": [ + 145, + 573, + 825, + 671 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "", + "bbox": [ + 147, + 676, + 825, + 733 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "63 1.2 Unmet need ", + "text_level": 1, + "bbox": [ + 148, + 752, + 297, + 767 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "64 Many datasets, especially those in multi-omics, are representable as a collection of matrices or tensors. \n65 As a case study, we consider (a subset of) the Lifelines-DEEP dataset from Tigchelaar et al. [22], \n66 which is summarized graphically in Figure 1. \n67 In this dataset, two different modalities of data were gathered from the same people: counts of \n68 microbial species found in their stools (metagenomics) and counts of metabolites found in their \n69 blood plasma (metabolomics). While different matrices, each modality shares an axis. If we were to \n70 estimate a graph of people on each modality independently, they would likely yield different graphs. \n71 This is not ideal; if our aim is to estimate the true graph of conditional dependencies, there should be \n72 only one resultant graph. To estimate it, we should be considering both modalities simultaneously. \n73 One way to do this would be to concatenate the modalities, producing a matrix of people by \n74 \"species+metabolites\". This could yield interesting results, if one is interested in connections between \n75 individual species and a metabolite. However, it would increase the size of the output graph, which \n76 grows quadratically in the length of the axis. Furthermore, it is not always possible or feasible; some \n77 datasets may not be concatenatable. We visually demonstrate some cases where concatenation fails \n78 in Figure 2. ", + "bbox": [ + 147, + 779, + 826, + 821 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "", + "bbox": [ + 147, + 828, + 825, + 911 + ], + "page_idx": 1 + }, + { + "type": "image", + "img_path": "images/6cef0776eb7b7093f0d7fbe7a289d4f8e68f9a96bf213d83b77dd9502809aeb6.jpg", + "image_caption": [ + "Figure 2: (A) A hypothetical dataset whose structure cannot be reduced to a single tensor by concatenation. Concatenating would lead to a block of missing values for a hypothetical (and nonsensical) species by elements matrix. (B) A hypothetical single-cell RNA-sequencing dataset procured from multiple patients. Concatenation is possible, but would lead to a very large output graph for a modest number of patients. " + ], + "image_footnote": [], + "bbox": [ + 334, + 87, + 661, + 377 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "", + "bbox": [ + 147, + 487, + 825, + 570 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "79 1.3 Our contributions ", + "text_level": 1, + "bbox": [ + 148, + 590, + 338, + 606 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "80 We introduce a novel method to extend the use of Gaussian Graphical Models to multi-tensor datasets. \n81 This extension is essential to model conditional dependencies in multimodal datasets such as those \n82 frequently occurring in multi-omics. We present an efficient algorithm to estimate these conditional \n83 dependencies. When restricted to the single-tensor case, our algorithm is much faster than previous \n84 algorithms that estimated conditional dependency graphs for each axis, such as TeraLasso[12] and \n85 EiGLasso[27]. ", + "bbox": [ + 147, + 618, + 825, + 702 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "2 Methods ", + "text_level": 1, + "bbox": [ + 169, + 724, + 277, + 742 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "2.1 Notation ", + "text_level": 1, + "bbox": [ + 174, + 758, + 274, + 773 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "In prior work, a single-tensor dataset $\\mathcal { D }$ is modelled as vec $[ \\mathcal { D } ] \\sim \\mathcal { N } \\left( \\mathbf { 0 } , ( \\oplus _ { \\ell } \\Psi _ { \\ell } ) ^ { - 1 } \\right)$ , also written as $\\mathcal { D } \\sim \\mathcal { N } _ { K S } \\left( \\{ \\Psi _ { \\ell } \\} _ { \\ell } \\right)$ . ", + "bbox": [ + 171, + 785, + 823, + 823 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "90 Our model considers multiple tensors, each with their own (potentially shared) axes. We aim to \n91 estimate the precision matrices $\\Psi _ { \\ell }$ for each axis $\\ell$ of each tensor $\\mathcal { D } ^ { \\gamma }$ , indexed by $\\gamma \\in \\mathbb { N }$ . To describe \n92 that an axis $\\ell$ is one of the axes of a tensor $\\mathcal { D } ^ { \\gamma }$ , we will write $\\ell \\in \\gamma$ . Some values will be indexed \n93 by both an axis and a tensor; for consistency we will use subscripts to denote axes (typically $\\ell$ ) and \n94 superscripts to denote tensors (typically $\\gamma$ ). $d _ { \\breve { \\breve { \\breve { \\breve { \\tau } } } } } ^ { \\gamma }$ will represent the number of elements in $\\mathcal { D } ^ { \\gamma }$ , and \n95 $\\begin{array} { r } { d \\check { \\forall } = \\sum _ { \\gamma } \\dot { d } _ { \\forall } ^ { \\gamma } } \\end{array}$ . \n96 An important concept is the Gram matrix $\\mathbf { S } _ { \\ell } ^ { \\gamma }$ . In the single-tensor case, this is a sufficient statistic; all \n97 prior work first computes these matrices as the first step in their algorithm. Let $\\mathrm { m a t } _ { \\ell } \\left[ \\mathcal { D } ^ { \\gamma } \\right]$ represent \n98 the \"matricization\" of $\\mathcal { D } ^ { \\gamma }$ along axis $\\ell$ , then $\\mathbf { S } _ { \\ell } ^ { \\gamma } = \\operatorname* { m a t } _ { \\ell } \\left[ \\mathcal { D } ^ { \\gamma } \\right] \\operatorname* { m a t } _ { \\ell } \\left[ \\mathcal { D } ^ { \\gamma } \\right] ^ { T }$ . The matricization of a \n99 tensor picks one axis, $\\ell$ , to index the rows, and flattens the rest out into columns. Note that for \n100 a matrix $\\mathbf { M } , \\mathrm { m a t } _ { \\mathrm { c o l u m n s } } \\left[ \\mathbf { M } \\right] = \\mathbf { M } ^ { T }$ . Rather than $\\mathbf { S } _ { \\ell } ^ { \\gamma }$ , we consider the \"effective Gram matrices\" \n101 $\\begin{array} { r } { \\mathbf { S } _ { \\ell } = \\sum _ { \\gamma | \\ell \\in \\gamma } \\mathbf { S } _ { \\ell } ^ { \\gamma } } \\end{array}$ , as these fulfill the role of the Gram matrices in the multi-tensor case. ", + "bbox": [ + 147, + 828, + 825, + 914 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "", + "bbox": [ + 143, + 90, + 825, + 179 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "02 2.2 The model ", + "text_level": 1, + "bbox": [ + 153, + 195, + 287, + 210 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "103 To properly handle sets of tensors, we propose modelling each tensor as being drawn independently \n104 from a Kronecker-sum normal distribution. If the tensors share an axis $\\ell$ , then they will still be drawn \n105 independently - but their distributions will be parameterized by the same $\\Psi _ { \\ell }$ . For an arbitrary set of \n106 tensors, the model is: ", + "bbox": [ + 142, + 220, + 825, + 276 + ], + "page_idx": 3 + }, + { + "type": "equation", + "img_path": "images/025cfc2fc5b07b10dbdfad3d759e45fc70255dac7f268cb431bd870ced39ed61.jpg", + "text": "$$\n\\begin{array} { r } { \\mathcal { D } ^ { \\gamma } \\sim \\mathcal { N } _ { K S } \\left( \\{ \\Psi _ { \\ell } \\} _ { \\ell \\in \\gamma } \\right) } \\\\ { \\mathrm { f o r } \\mathcal { D } ^ { \\gamma } \\in \\{ \\mathcal { D } ^ { \\gamma } \\} _ { \\gamma } } \\end{array}\n$$", + "text_format": "latex", + "bbox": [ + 419, + 299, + 578, + 347 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "107 We call this model the \"Gaussian multi-Graphical Model\" (GmGM) as it extends Gaussian Graphical \n108 Models to estimate multiple graphs from a set of tensors. In this paper, we will make the assumption \n109 that no tensor in our set contains the same axis twice - notably, covariance matrices would violate \n110 this assumption. Any tensor with a repeated axis would naturally be interpretable as a graph - such \n111 datasets are rare, and if one already has a graph the need for an algorithm such as this is diminished. ", + "bbox": [ + 140, + 354, + 826, + 426 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "As an example, we model the LifeLines-DEEP dataset Dmetagenomics and Dmetabolomics 112 indepen113 dently as: ", + "bbox": [ + 140, + 430, + 826, + 460 + ], + "page_idx": 3 + }, + { + "type": "equation", + "img_path": "images/9478a0adf0d6334af50e508ee8fe57c312944a53e903060712bc3cd9c0200840.jpg", + "text": "$$\n\\begin{array} { r l } & { \\mathbf { D } ^ { \\mathrm { m e t a g e n o m i c s } } \\sim \\mathcal { N } _ { K S } \\left( \\Psi _ { \\mathrm { p e o p l e } } , \\Psi _ { \\mathrm { s p e c i e s } } \\right) } \\\\ & { \\mathbf { D } ^ { \\mathrm { m e t a b o l o m i c s } } \\sim \\mathcal { N } _ { K S } \\left( \\Psi _ { \\mathrm { p e o p l e } } , \\Psi _ { \\mathrm { m e t a b o l i t e s } } \\right) } \\end{array}\n$$", + "text_format": "latex", + "bbox": [ + 346, + 483, + 651, + 523 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "114 2.3 The algorithm ", + "text_level": 1, + "bbox": [ + 142, + 536, + 313, + 551 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "115 Here, we present an algorithm to compute the maximum likelihood estimate (MLE) jointly for all \n116 parameters $\\Psi _ { \\ell }$ of the GmGM. The general idea is to produce an analytic estimate for the eigenvectors \n117 of $\\Psi _ { \\ell }$ , and then iterate to solve for the eigenvalues; this is summed up graphically in Figure 3. ", + "bbox": [ + 142, + 561, + 825, + 604 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "118 In the supplementary material, we derive the following: ", + "bbox": [ + 140, + 609, + 539, + 626 + ], + "page_idx": 3 + }, + { + "type": "equation", + "img_path": "images/213d9e856d915f179848be7e6b33bede1e9291102cf081c28628a43e7d334da6.jpg", + "text": "$$\n\\begin{array} { r } { p ( \\{ \\mathcal { D } ^ { \\gamma } \\} ) = \\frac { \\displaystyle \\prod _ { \\gamma } \\sqrt { \\left| \\bigoplus _ { \\ell \\in \\gamma } \\Psi _ { \\ell } \\right| } } { \\displaystyle ( 2 \\pi ) ^ { \\frac { d _ { \\psi } } { 2 } } } e ^ { \\frac { - 1 } { 2 } \\sum _ { \\ell } \\mathrm { t r } [ \\Psi _ { \\ell } \\mathbf { S } _ { \\ell } ] } } \\\\ { \\mathrm { N L L } \\left[ \\{ \\mathcal { D } ^ { \\gamma } \\} \\right] \\propto \\displaystyle \\sum _ { \\ell } \\mathrm { t r } \\left[ \\Psi _ { \\ell } \\mathbf { S } _ { \\ell } \\right] - \\sum _ { \\gamma } \\log \\left| \\bigoplus _ { \\ell \\in \\gamma } \\Psi _ { \\ell } \\right| } \\end{array}\n$$", + "text_format": "latex", + "bbox": [ + 254, + 646, + 586, + 757 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "(pdf of GmGM) ", + "bbox": [ + 715, + 674, + 823, + 689 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "(negative log likelihood) ", + "bbox": [ + 661, + 722, + 825, + 737 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "119 From this, we can observe that the effective Gram matrices $\\mathbf { S } _ { \\ell }$ form a set of sufficient statistics for \n120 our distribution. Furthermore, the log-likelihood is the sum of log-likelihoods in the single-axis case, \n121 thus preserving convexity of the loss function. \n22 Theorem 1. Let $\\mathbf { V } _ { \\ell } \\mathbf { e } _ { \\ell } \\mathbf { V } _ { \\ell } ^ { T }$ be the eigendecomposition of $\\mathbf { S } _ { \\ell }$ (where $\\mathbf { V } _ { \\ell } \\in \\mathbb { R } ^ { d _ { \\ell } \\times d _ { \\ell } }$ and $\\mathbf { e } _ { \\ell } \\in \\mathbb { R } ^ { d _ { \\ell } \\times d _ { \\ell } }$ \n123 is a diagonal matrix). Then $\\mathbf { V } _ { \\ell }$ are the eigenvectors of the maximum likelihood estimate of $\\Psi _ { \\ell }$ . \n124 Theorem 1 is critical to allowing efficient estimation of $\\Psi _ { \\ell }$ , as it not only allows us to extract the \n125 computationally intensive eigendecomposition operation from the iterative portion of the algorithm, \n126 but also reduces the number of parameters to be linear in the length of an axis. ", + "bbox": [ + 140, + 765, + 826, + 808 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "", + "bbox": [ + 150, + 809, + 821, + 839 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "", + "bbox": [ + 142, + 848, + 826, + 891 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "127 To find the eigenvalues $\\mathbf { \\Delta } \\Lambda _ { \\ell }$ of $\\Psi _ { \\ell }$ , we produce the second theorem: ", + "bbox": [ + 142, + 896, + 611, + 912 + ], + "page_idx": 3 + }, + { + "type": "image", + "img_path": "images/c745f0a60bb4d1028aa6fe2aa662f9c81b432db1e0664152bfb36f4164dab7f9.jpg", + "image_caption": [ + "Figure 3: A graphical overview of how the GmGM algorithm works. We use $\\gamma$ to represent an arbitrary modality, and $\\ell$ to represent an arbitrary axis. Proofs are given in the supplementary material. " + ], + "image_footnote": [], + "bbox": [ + 204, + 83, + 772, + 602 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Theorem 2. Let $\\{ \\mathbf { G } _ { \\ell } ^ { \\gamma } \\}$ be matrices such that the expression $\\oplus _ { \\ell \\in \\gamma } \\mathbf { G } _ { \\ell } ^ { \\gamma }$ is the best Frobenius-norm approximation of $\\left( \\oplus _ { \\ell \\in \\gamma } \\mathbf { \\Lambda } \\mathbf { \\Lambda } \\mathbf { \\Lambda } \\mathbf { \\Lambda } _ { \\ell } ^ { t } \\right) ^ { - 1 }$ . Then, for a learning rate $\\mu _ { t }$ , gradient descent can be performed with the update equation $\\begin{array} { r } { \\pmb { \\Lambda } _ { \\ell } ^ { t + 1 } = \\pmb { \\Lambda } _ { \\ell } ^ { t } - \\mu _ { t } \\left[ \\mathbf { e } _ { \\ell } - \\sum _ { \\gamma | \\ell \\in \\gamma } \\mathbf { G } _ { \\ell } ^ { \\gamma } \\right] } \\end{array}$ . As $\\Psi _ { \\ell }$ is positive definite, $\\mu _ { t }$ must be chosen to prevent $\\Lambda _ { \\ell } ^ { t }$ from becoming negative. ", + "bbox": [ + 166, + 683, + 825, + 767 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "132 While the definition of $\\mathbf { G } _ { \\ell } ^ { \\gamma }$ is technical, it is analogous to the notion of the blockwise-trace from \n133 Kalaitzis et al. [14] and $\\mathrm { p r o j } _ { \\kappa }$ from Greenewald, Zhou, and Hero III [12]. Proofs of Theorems 1 \n134 and 2, along with a method to compute $\\mathbf { G } _ { \\ell } ^ { \\gamma }$ , are given in the supplementary material. Overall, our \n135 algorithm is described in the pseudocode at the top of the next page. \n136 For regularization, one can choose to either keep the top $p \\%$ of edges, or keep the top $k$ edges per \n137 vertex (for parameters $p , k )$ . The incorporation of more advanced regularizers, such as Lasso, would \n138 require an eigen-recomposition on each iteration, which would be much slower. As we demonstrate \n139 empirically in Section 3, it is not necessary to use advanced regularizers to recover the graph structure \n140 to the same precision as prior work. ", + "bbox": [ + 142, + 779, + 825, + 837 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "", + "bbox": [ + 142, + 842, + 825, + 912 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "The GmGM algorithm ", + "text_level": 1, + "bbox": [ + 174, + 90, + 334, + 106 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Input: $\\{ \\mathcal { D } _ { i } ^ { \\gamma } \\}$ , tolerance \nOutput: $\\{ \\Psi _ { \\ell } \\}$ \n1: for 1 ≤ ℓ ≤ K \n2: Sℓ ← Pγ |ℓ∈γ 1nγ Pnγi matℓ [Dγi ] matℓ [Dγi ]T \n3: Vℓ ← eigenvectors[Sℓ] \n4: eℓ ← eigenvalues[Sℓ] \n5: end for \n6 $\\vdots \\stackrel { \\wedge } { \\underset { \\mu 1 } { \\longrightarrow } } [ 1 \\quad \\cdots \\quad 1 ] ^ { T }$ \n7 \n8: while not converged \n9: for $1 \\leq \\ell \\leq \\bar { K }$ \n10: $\\begin{array} { r l } & { \\mathbf { \\mu } _ { \\mathbf { G } _ { \\ell } ^ { \\gamma } } ^ { \\mathbf { 1 } } \\mathrm { p r o j } _ { K S } [ ( \\bigoplus _ { \\ell ^ { \\prime } \\in \\gamma } \\mathbf { \\Lambda } _ { \\Lambda _ { \\ell } } ) ^ { - 1 } ] } \\\\ & { \\mathbf { \\Lambda } _ { \\mathbf { A } _ { \\ell } ^ { \\prime } } ^ { \\mathbf { 1 } } \\mathbf { \\Lambda } _ { \\mathbf { A } _ { \\ell } } - \\mu [ \\mathbf { e } _ { \\ell } - \\sum _ { \\gamma | \\ell \\in \\gamma } \\mathbf { G } _ { \\ell _ { \\ell } } ^ { \\gamma } ] } \\end{array}$ \n11: \n12: end for \n13: for 1 ≤ ℓ ≤ K \n14: Λℓ ← Λ′ℓ \n15: end for \n16: for $\\gamma$ \n17: if $\\sum _ { \\ell \\in \\gamma } \\operatorname* { m i n } \\pmb { \\Lambda } _ { \\ell } <$ < tolerance then \n18: decrease $\\mu$ so that this result is sufficiently far from zero \n19: end if \n20: end for \n21: end while \n22: for 1 ≤ ℓ ≤ K \n23: ${ \\Psi } _ { \\ell } \\gets { \\bf V } _ { \\ell } { \\bf A } _ { \\ell } { \\bf V } _ { \\ell } ^ { T }$ \n24: end for ", + "bbox": [ + 173, + 109, + 648, + 502 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "3 Results ", + "text_level": 1, + "bbox": [ + 169, + 526, + 266, + 542 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "We tested our algorithm on synthetic data and five real-world datasets. Explanations of data generation, collection, preprocessing, and regularization are given in the supplementary material. ", + "bbox": [ + 171, + 558, + 823, + 587 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "3.1 Synthetic Data ", + "text_level": 1, + "bbox": [ + 171, + 602, + 315, + 616 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "We verified that our algorithm was indeed faster on matrix-variate data compared to prior work (Figure 4) on our computer (Ubuntu 20.04 with Intel Core i7 Processor and 8GB RAM). Our results on matrix data are encouraging - extrapolating the runtimes, datasets up to size 16,000 by 16,000 could have their graphs estimated in less than an hour. Larger datasets would require more than 6GB of memory for our algorithm to run, pushing the limits of RAM. Our algorithm was not significantly faster on higher-order tensor data (see the supplementary material). This is due to the complexity of computing the Gram matrices, which grows exponentially with the number of axes. ", + "bbox": [ + 171, + 627, + 825, + 724 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "152 In addition to these speed improvements, we show that we perform equivalently to state-of-the-art \n153 on matrix data (Figure 5a). On higher-order tensor data, we are outperformed by TeraLasso, which \n154 is able to achieve near-perfect recovery of the graphs. We believe this is due to our algorithm’s use \n155 of thresholding rather than a more advanced regularization technique. Since our speed gains are \n156 not significant relative to TeraLasso, on higher-order tensor data without shared axes one should \n157 prefer TeraLasso to GmGM. Finally, we demonstrate that taking into account shared axes does indeed \n158 improve performance (see blue line, Figure 5b). Prior work could not take this into account. ", + "bbox": [ + 142, + 731, + 825, + 828 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "3.2 Real Data ", + "text_level": 1, + "bbox": [ + 173, + 844, + 281, + 858 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "160 We tested our method on various real datasets. These include two video datasets (COIL-20 [19] and \n161 EchoNet-Dynamic [20]), a transcriptomics dataset (E-MTAB-2805 [5]), and two multi-omics datasets \n162 (LifeLines-DEEP [22] and a $1 0 \\mathrm { x }$ Genomics dataset [1]). \n163 The E-MTAB-2805 dataset consists of transcriptomics data for individual cells split into three groups \n164 by their stage in the cell cycle (G, S, and G2/M). If our estimated precision matrices had a 3x3 \n165 block-diagonal structure, this would indicate that it had recreated this grouping. This is not what we \n166 see, but we do see a 3x3 block matrix structure (Figure 6a). We found that cells in the DNA synthesis \n167 stage (S) had few connections between them, and that there were many connections between the G1 \n168 and G2/M stages. This result is biologically plausible, as cells in the synthesis stage are the most \n169 variable. \n170 The results on EchoNet-Dynamic (Figure 6b) are much more encouraging, as we would expect a \n171 periodic structure due to the beating of the heart. A precision matrix with repeating diagonals is what \n172 we would expect to see in this case, which is what our algorithm produces. In the supplementary \n173 material, we further verify that this corresponds to a heartbeat by using the repetition to accurately \n174 predict the opening of the mitral valve in the video. \n175 The duck video in the COIL-20 dataset was considered in the original BiGLasso paper [14], in which \n176 they showed that their algorithm could recover the ordering of the frames of the video. To do this they \n177 had to heavily downsample the image (to a $9 \\mathrm { x } 9$ image with half the frames), and flatten the rows and \n178 frames into a single axis. Due to the speed improvements of our algorithm, and its ability to handle \n179 tensor-variate data, we were able to run our algorithm on the raw, unprocessed data and achieve a \n180 similar result in negligible time. Specifically, the reconstruction of the frames had an accuracy of \n181 $9 9 \\%$ . \n182 Prior work by Prost, Gazut, and Brüls [21] used assortativity to assess their validity of the species \n183 graph estimated by their model on the LifeLines-DEEP metagenomics dataset. Assortativity repre \n184 sents the tendency of related species to cluster together in the graph. A random graph would have \n185 an assortativity of zero, but we would expect moderate assortativity in the true network as similar \n186 species may fulfill similar roles in the gut microbiome. Our assortativity is comparable to prior work \n187 (Figure 7). We also found that our graphs were more robust to noise than prior work; we analyze this \n188 in the supplementary material. \n189 Finally, we tested our approach on a 10x Genomics single-cell (RNA $+$ ATAC) dataset taken from a \n190 B Cell lymphoma tumour. We demonstrate that the clusters we find (using Louvain clustering[3]) \n191 on the graph remain visually cohesive when projected into lower-dimensional space by UMAP[18] \n192 (Figure 8). In particular, the disconnected “islands” in UMAP correspond to their own cluster on the \n193 graph as well. As these island-clusters were arrived at independently through two methods, UMAP \n194 and our algorithm, it increases our confidence in the validity of the clustering. In the supplementary \n195 material, we verify that these clusters do represent distinct groups via a GO term enrichment analysis. \n196 Our overall approach has been implemented in Python. All of the code to run the algorithm and \n197 recreate the experiments has been made publicly available on GitHub; https://github.com/NeurIPS \n198 GmGM-Paper/GmGM. ", + "bbox": [ + 142, + 869, + 823, + 911 + ], + "page_idx": 5 + }, + { + "type": "image", + "img_path": "images/adf75c680df1627b4e583204f9c235d92619010c815d66d9cc9ac33e931b16d9.jpg", + "image_caption": [ + "Figure 4: A comparison of the runtimes of our algorithm against (a) bi-graphical and (b) tensorgraphical prior work. Runtimes were averaged over 5 runs. " + ], + "image_footnote": [], + "bbox": [ + 186, + 109, + 795, + 338 + ], + "page_idx": 6 + }, + { + "type": "image", + "img_path": "images/fe28e6abee872ced025d8f211a7d3a9623c3b39d9451928e3d9e90a6e829f38f.jpg", + "image_caption": [ + "Figure 5: (a) Precision-recall curves comparing various algorithms on synthetic $5 0 \\mathrm { x } 5 0$ matrix data. (b) Precision-recall curves comparing our algorithm on two $5 0 \\mathrm { x } 5 0$ matrices with one shared axis. We considered both modalities simultaneously (blue) and an individual modality (red, orange). In both subfigures, each edge of the true graphs was generated independently with probability $\\frac { 1 } { 5 }$ . " + ], + "image_footnote": [], + "bbox": [ + 204, + 396, + 807, + 579 + ], + "page_idx": 6 + }, + { + "type": "image", + "img_path": "images/fbc2cd5b1d52cd96c55ea92f177245ea6f124c0513023e7e67a96549ef2821d7.jpg", + "image_caption": [ + "Figure 6: The estimated precision matrices on the E-MTAB-2805 dataset (a) and the EchoNetDynamic dataset (b). Yellow represents an edge and purple represents the lack of an edge. The E-MTAB-2805 cells have been grouped together by cell cycle stage, in the order G, S, and G2/M. " + ], + "image_footnote": [], + "bbox": [ + 230, + 665, + 767, + 852 + ], + "page_idx": 6 + }, + { + "type": "image", + "img_path": "images/860cbfbc342cbb12b6484a8c57d4f69382fce3aa91bdbc76442be0f97d273bb1.jpg", + "image_caption": [ + "Figure 7: Assortativity with increasing regulatization in the LifeLines-DEEP dataset, comparing our method with the Zero-inflated Log-Normal (ZiLN) model. In one case we show the performance of our algorithm restricted to the metagenomics dataset (a) and when augmented with the metabolomics dataset (b). In both cases, ZiLN is only trained on the metagenomics dataset, as it is a single-axis model. " + ], + "image_footnote": [], + "bbox": [ + 174, + 88, + 821, + 333 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "", + "bbox": [ + 142, + 443, + 825, + 540 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "", + "bbox": [ + 142, + 545, + 825, + 614 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "", + "bbox": [ + 140, + 621, + 825, + 718 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "", + "bbox": [ + 140, + 724, + 825, + 821 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "", + "bbox": [ + 142, + 828, + 825, + 911 + ], + "page_idx": 7 + }, + { + "type": "image", + "img_path": "images/bc6bf907b9b5296b0c6fbfd071d89fc068a9ca33994ea71c4f3433b32778adf6.jpg", + "image_caption": [ + "Figure 8: Two plots of the same cells from the $1 0 \\mathrm { x }$ Genomics dataset, displayed via UMAP [18] (a) and the Fruchterman-Reingold layout algorithm[11] (b). Colors are based on Louvain clustering[3] of the graph, and represent the same clustering in both figures. " + ], + "image_footnote": [], + "bbox": [ + 181, + 92, + 816, + 334 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "", + "bbox": [ + 143, + 410, + 825, + 465 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "199 4 Limitations ", + "text_level": 1, + "bbox": [ + 150, + 484, + 302, + 502 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "200 Our method uses thresholding rather than more sophisticated regularizers. However, there is no \n201 fundamental barrier preventing our algorithm from allowing regularizers at the cost of an eigen \n202 recomposition per iteration. This would increase the asymptotic complexity of the iterative portion \n203 of our algorithm, making it questionable whether any gains in precision would be worth the loss in \n204 efficiency. \n205 Our method assumes that no tensor has a repeated axis (i.e. a matrix of people by people rather than \n206 people by species). If there is a repeated axis, one can no longer analytically find the eigenvectors of \n207 the MLE, at least by our methods. This is not a substantial issue, as such datasets are uncommon and \n208 already represent graphs. Rather than extending the algorithm to work with repeated-axis tensors, it \n209 would be more fruitful to extend it to work with priors. \n210 When considering multi-tensor datasets, it may be the case that two axes only partially overlap. For \n211 example, the full LifeLines-DEEP dataset contains a second (follow-up) metagenomics dataset for a \n212 third of the study participants; two thirds of the patients are missing from this dataset. We do not \n213 make an attempt to handle this type of missing data, even though missing data shows up in many \n214 applications. The lack of ability to handle missing data is a major limitation of our algorithm. It is \n215 nontrivial to extend the algorithm to handle this case, as it renders Theorem 1 ineffective and hence \n216 removes the speed advantage we attained. Prior work has not addressed this problem, as it only exists \n217 in multi-tensor datasets and we are the first to consider this case. ", + "bbox": [ + 140, + 515, + 825, + 585 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "", + "bbox": [ + 140, + 592, + 825, + 661 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "", + "bbox": [ + 140, + 667, + 825, + 779 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "218 5 Conclusion ", + "text_level": 1, + "bbox": [ + 147, + 796, + 299, + 814 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "We have created a novel model, GmGM, which successfully generalizes Gaussian graphical models to the common scenario of multi-tensor datasets. Furthermore, we demonstrated that our algorithm is significantly faster than prior work focusing on Gaussian tensor-graphical models such as EiGLasso and TeraLasso while still preserving state-of-the-art performance. These speed improvements allow tensor-graphical models to be applied to datasets with axes of length in the thousands. Finally, we demonstrated the application of our algorithm on five real-world datasets to prove its efficacy. ", + "bbox": [ + 168, + 827, + 823, + 911 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "225 References \n226 [1] 10x Genomics. Flash-Frozen Lymph Node with B Cell Lymphoma (14k sorted nuclei). en. May \n227 2021. URL: https://www.10xgenomics.com/resources/datasets/fresh-frozen \n228 lymph-node-with-b-cell-lymphoma-14-k-sorted-nuclei-1-standard-2-0-0 \n229 (visited on 05/10/2023). \n230 [2] Michael Altenbuchinger et al. “Gaussian and Mixed Graphical Models as (multi-)omics \n231 data analysis tools”. en. In: Biochimica et Biophysica Acta (BBA) - Gene Regulatory \n232 Mechanisms. Transcriptional Profiles and Regulatory Gene Networks 1863.6 (June 2020), \n233 p. 194418. ISSN: 1874-9399. DOI: 10 . 1016 / j . bbagrm . 2019 . 194418. URL: https : \n234 //www.sciencedirect.com/science/article/pii/S187493991930224X (visited on \n235 02/24/2023). \n236 [3] Vincent D. Blondel et al. “Fast unfolding of communities in large networks”. In: Journal of \n237 Statistical Mechanics: Theory and Experiment 2008.10 (Oct. 2008). arXiv:0803.0476 [cond \n238 mat, physics:physics], P10008. ISSN: 1742-5468. DOI: 10.1088/1742- 5468/2008/10/ \n239 P10008. URL: http://arxiv.org/abs/0803.0476 (visited on 05/16/2023). \n240 [4] Danila Bredikhin, Ilia Kats, and Oliver Stegle. “MUON: multimodal omics analysis frame \n241 work”. In: Genome Biology 23.1 (Feb. 2022), p. 42. ISSN: 1474-760X. DOI: 10.1186/s13059- \n242 021- 02577- 8. URL: https://doi.org/10.1186/s13059- 021- 02577- 8 (visited on \n243 05/16/2023). \n244 [5] Florian Buettner et al. “Computational analysis of cell-to-cell heterogeneity in single-cell \n245 RNA-sequencing data reveals hidden subpopulations of cells”. en. In: Nature Biotechnology \n246 33.2 (Feb. 2015). Number: 2 Publisher: Nature Publishing Group, pp. 155–160. ISSN: 1546- \n247 1696. DOI: 10.1038/nbt.3102. URL: https://www.nature.com/articles/nbt.3102 \n248 (visited on 05/10/2023). \n249 [6] Thomas A. Caswell et al. matplotlib/matplotlib: REL: v3.7.1. Mar. 2023. DOI: 10.5281/ \n250 zenodo.7697899. URL: https://zenodo.org/record/7697899 (visited on 05/16/2023). \n251 [7] Gabor Csardi and Tamas Nepusz. “The Igraph Software Package for Complex Network \n252 Research”. In: InterJournal Complex Systems (Nov. 2005), p. 1695. \n253 [8] Andy Dahl et al. Network inference in matrix-variate Gaussian models with non-independent \n254 noise. arXiv:1312.1622 [stat]. Dec. 2013. DOI: 10.48550/arXiv.1312.1622. URL: http: \n255 //arxiv.org/abs/1312.1622 (visited on 02/27/2023). \n256 [9] Pan Du, Warren A. Kibbe, and Simon M. Lin. “Improved peak detection in mass spectrum by \n257 incorporating continuous wavelet transform-based pattern matching”. In: Bioinformatics 22.17 \n258 (Sept. 2006), pp. 2059–2065. ISSN: 1367-4803. DOI: 10.1093/bioinformatics/btl355. \n259 URL: https://doi.org/10.1093/bioinformatics/btl355 (visited on 05/11/2023). \n260 [10] Jerome Friedman, Trevor Hastie, and Robert Tibshirani. “Sparse inverse covariance estima \n261 tion with the graphical lasso”. In: Biostatistics 9.3 (July 2008), pp. 432–441. ISSN: 1465- \n262 4644. DOI: 10 . 1093 / biostatistics / kxm045. URL: https : / / doi . org / 10 . 1093 / \n263 biostatistics/kxm045 (visited on 05/09/2023). \n264 [11] Thomas M. J. Fruchterman and Edward M. Reingold. “Graph drawing by force \n265 directed placement”. en. In: Software: Practice and Experience 21.11 (1991). _eprint: \n266 https://onlinelibrary.wiley.com/doi/pdf/10.1002/spe.4380211102, pp. 1129–1164. ISSN: 1097- \n267 024X. DOI: 10.1002/spe.4380211102. URL: https://onlinelibrary.wiley.com/ \n268 doi/abs/10.1002/spe.4380211102 (visited on 05/16/2023). \n269 [12] Kristjan Greenewald, Shuheng Zhou, and Alfred Hero III. Tensor Graphical Lasso (TeraLasso). \n270 arXiv:1705.03983 [stat]. Sept. 2019. URL: http://arxiv.org/abs/1705.03983 (visited \n271 on 02/24/2023). \n272 [13] Charles R. Harris et al. “Array programming with NumPy”. en. In: Nature 585.7825 (Sept. \n273 2020). Number: 7825 Publisher: Nature Publishing Group, pp. 357–362. ISSN: 1476-4687. DOI: \n274 10.1038/s41586-020-2649-2. URL: https://www.nature.com/articles/s41586- \n275 020-2649-2 (visited on 05/16/2023). \n276 [14] Alfredo Kalaitzis et al. “The Bigraphical Lasso”. en. In: Proceedings of the 30th International \n277 Conference on Machine Learning. ISSN: 1938-7228. PMLR, May 2013, pp. 1229–1237. URL: \n278 https://proceedings.mlr.press/v28/kalaitzis13.html (visited on 02/24/2023). \n279 [15] Tamara Kolda. Multilinear operators for higher-order decompositions. en. Tech. rep. \n280 SAND2006-2081, 923081. Apr. 2006, SAND2006–2081, 923081. DOI: 10.2172/923081. \n281 URL: https://www.osti.gov/servlets/purl/923081/ (visited on 05/05/2023). \n282 [16] Tamara G. Kolda and Brett W. Bader. “Tensor Decompositions and Applications”. en. In: \n283 SIAM Review 51.3 (Aug. 2009), pp. 455–500. ISSN: 0036-1445, 1095-7200. DOI: 10.1137/ \n284 07070111X. URL: http : / / epubs . siam . org / doi / 10 . 1137 / 07070111X (visited on \n285 02/26/2023). \n286 [17] Sijia Li et al. Scalable Bigraphical Lasso: Two-way Sparse Network Inference for Count \n287 Data. arXiv:2203.07912 [cs, stat]. Mar. 2022. URL: http://arxiv.org/abs/2203.07912 \n288 (visited on 02/24/2023). \n289 [18] Leland McInnes, John Healy, and James Melville. UMAP: Uniform Manifold Approximation \n290 and Projection for Dimension Reduction. arXiv:1802.03426 [cs, stat]. Sept. 2020. DOI: 10. \n291 48550/arXiv.1802.03426. URL: http://arxiv.org/abs/1802.03426 (visited on \n292 05/10/2023). \n293 [19] Sameer A Nene, Shree K Nayar, and Hiroshi Murase. “Columbia Object Image Library \n294 (COIL-20)”. en. In: (). \n295 [20] David Ouyang et al. “Video-based AI for beat-to-beat assessment of cardiac function”. en. In: \n296 Nature 580.7802 (Apr. 2020). Number: 7802 Publisher: Nature Publishing Group, pp. 252–256. \n297 ISSN: 1476-4687. DOI: 10.1038/s41586-020-2145-8. URL: https://www.nature.com/ \n298 articles/s41586-020-2145-8 (visited on 05/10/2023). \n299 [21] Vincent Prost, Stéphane Gazut, and Thomas Brüls. “A zero inflated log-normal model for \n300 inference of sparse microbial association networks”. en. In: PLOS Computational Biology \n301 17.6 (June 2021). Publisher: Public Library of Science, e1009089. ISSN: 1553-7358. DOI: 10. \n302 1371/journal.pcbi.1009089. URL: https://journals.plos.org/ploscompbiol/ \n303 article?id ${ \\tt = } 1 0$ .1371/journal.pcbi.1009089 (visited on 02/24/2023). \n304 [22] Ettje F. Tigchelaar et al. “Cohort profile: LifeLines DEEP, a prospective, general population \n305 cohort study in the northern Netherlands: study design and baseline characteristics”. en. In: \n306 BMJ Open 5.8 (Aug. 2015). Publisher: British Medical Journal Publishing Group Section: Epi \n307 demiology, e006772. ISSN: 2044-6055, 2044-6055. DOI: 10.1136/bmjopen-2014-006772. \n308 URL: https://bmjopen.bmj.com/content/5/8/e006772 (visited on 04/30/2023). \n309 [23] Theodoros Tsiligkaridis, Alfred O. Hero III, and Shuheng Zhou. “Convergence Properties \n310 of Kronecker Graphical Lasso Algorithms”. In: IEEE Transactions on Signal Processing \n311 61.7 (Apr. 2013). arXiv:1204.0585 [stat], pp. 1743–1755. ISSN: 1053-587X, 1941-0476. DOI: \n312 10.1109/TSP.2013.2240157. URL: http://arxiv.org/abs/1204.0585 (visited on \n313 02/27/2023). \n314 [24] Yu Wang and Alfred Hero. “SG-PALM: a Fast Physically Interpretable Tensor Graphical \n315 Model”. en. In: Proceedings of the 38th International Conference on Machine Learning. \n316 ISSN: 2640-3498. PMLR, July 2021, pp. 10783–10793. URL: https://proceedings.mlr. \n317 press/v139/wang21k.html (visited on 02/27/2023). \n318 [25] Yu Wang, Byoungwook Jang, and Alfred Hero. The Sylvester Graphical Lasso (SyGlasso). \n319 arXiv:2002.00288 [cs, stat]. Feb. 2020. DOI: 10.48550/arXiv.2002.00288. URL: http: \n320 //arxiv.org/abs/2002.00288 (visited on 02/27/2023). \n321 [26] F. Alexander Wolf, Philipp Angerer, and Fabian J. Theis. “SCANPY: large-scale single-cell \n322 gene expression data analysis”. In: Genome Biology 19.1 (Feb. 2018), p. 15. ISSN: 1474-760X. \n323 DOI: 10.1186/s13059-017-1382-0. URL: https://doi.org/10.1186/s13059-017- \n324 1382-0 (visited on 05/16/2023). \n325 [27] Jun Ho Yoon and Seyoung Kim. “EiGLasso: Scalable Estimation of Cartesian Product of Sparse \n326 Inverse Covariance Matrices”. en. In: Proceedings of the 36th Conference on Uncertainty \n327 in Artificial Intelligence (UAI). ISSN: 2640-3498. PMLR, Aug. 2020, pp. 1248–1257. URL: \n328 https://proceedings.mlr.press/v124/ho-yoon20a.html (visited on 02/24/2023). ", + "bbox": [ + 137, + 79, + 828, + 914 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "", + "bbox": [ + 135, + 73, + 830, + 845 + ], + "page_idx": 10 + } +] \ No newline at end of file diff --git a/parse/dev/SfXjt1FtMQ/SfXjt1FtMQ_model.json b/parse/dev/SfXjt1FtMQ/SfXjt1FtMQ_model.json new file mode 100644 index 0000000000000000000000000000000000000000..4ecf6b25aa6f71d392b9755fffb6a387132891e7 --- /dev/null +++ b/parse/dev/SfXjt1FtMQ/SfXjt1FtMQ_model.json @@ -0,0 +1,19210 @@ +[ + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 255, + 757, + 1304, + 757, + 1304, + 1063, + 255, + 1063 + ], + "score": 0.972 + }, + { + "category_id": 0, + "poly": [ + 391, + 271, + 1309, + 271, + 1309, + 380, + 391, + 380 + ], + "score": 0.961 + }, + { + "category_id": 1, + "poly": [ + 251, + 1183, + 1405, + 1183, + 1405, + 1487, + 251, + 1487 + ], + "score": 0.958 + }, + { + "category_id": 1, + "poly": [ + 252, + 1865, + 1403, + 1865, + 1403, + 1988, + 252, + 1988 + ], + "score": 0.941 + }, + { + "category_id": 1, + "poly": [ + 251, + 1698, + 1405, + 1698, + 1405, + 1850, + 251, + 1850 + ], + "score": 0.939 + }, + { + "category_id": 1, + "poly": [ + 250, + 1502, + 1404, + 1502, + 1404, + 1685, + 250, + 1685 + ], + "score": 0.935 + }, + { + "category_id": 0, + "poly": [ + 787, + 691, + 912, + 691, + 912, + 727, + 787, + 727 + ], + "score": 0.888 + }, + { + "category_id": 1, + "poly": [ + 720, + 490, + 986, + 490, + 986, + 611, + 720, + 611 + ], + "score": 0.887 + }, + { + "category_id": 0, + "poly": [ + 253, + 1115, + 531, + 1115, + 531, + 1152, + 253, + 1152 + ], + "score": 0.878 + }, + { + "category_id": 2, + "poly": [ + 285, + 2032, + 1388, + 2032, + 1388, + 2062, + 285, + 2062 + ], + "score": 0.514 + }, + { + "category_id": 1, + "poly": [ + 285, + 2032, + 1388, + 2032, + 1388, + 2062, + 285, + 2062 + ], + "score": 0.379 + }, + { + "category_id": 0, + "poly": [ + 253, + 1115, + 531, + 1115, + 531, + 1153, + 253, + 1153 + ], + "score": 0.104 + }, + { + "category_id": 13, + "poly": [ + 592, + 1304, + 905, + 1304, + 905, + 1340, + 592, + 1340 + ], + "score": 0.92, + "latex": "\\mathbb { P } ( \\bar { x } | y , \\mathcal { D } _ { \\backslash x y } ) = \\mathbb { P } ( x | \\mathcal { D } _ { \\backslash x y } )" + }, + { + "category_id": 13, + "poly": [ + 297, + 1424, + 487, + 1424, + 487, + 1458, + 297, + 1458 + ], + "score": 0.92, + "latex": "\\mathbf { \\dot { d } } \\sim \\mathcal { N } ( \\mathbf { 0 } , \\Psi ^ { - 1 } )" + }, + { + "category_id": 13, + "poly": [ + 1035, + 1758, + 1281, + 1758, + 1281, + 1793, + 1035, + 1793 + ], + "score": 0.9, + "latex": "\\mathrm { v e c } [ \\mathcal { D } ] \\sim \\mathcal { N } ( \\mathbf { 0 } , \\Psi ^ { - 1 } )" + }, + { + "category_id": 13, + "poly": [ + 836, + 1249, + 883, + 1249, + 883, + 1275, + 836, + 1275 + ], + "score": 0.87, + "latex": "x , y" + }, + { + "category_id": 13, + "poly": [ + 973, + 1760, + 1002, + 1760, + 1002, + 1788, + 973, + 1788 + ], + "score": 0.81, + "latex": "\\Psi" + }, + { + "category_id": 13, + "poly": [ + 722, + 1426, + 749, + 1426, + 749, + 1453, + 722, + 1453 + ], + "score": 0.81, + "latex": "\\Psi" + }, + { + "category_id": 13, + "poly": [ + 496, + 1958, + 523, + 1958, + 523, + 1984, + 496, + 1984 + ], + "score": 0.78, + "latex": "\\Psi" + }, + { + "category_id": 13, + "poly": [ + 629, + 1791, + 657, + 1791, + 657, + 1818, + 629, + 1818 + ], + "score": 0.78, + "latex": "\\Psi" + }, + { + "category_id": 13, + "poly": [ + 796, + 1761, + 822, + 1761, + 822, + 1787, + 796, + 1787 + ], + "score": 0.77, + "latex": "\\mathcal { D }" + }, + { + "category_id": 13, + "poly": [ + 403, + 1276, + 428, + 1276, + 428, + 1302, + 403, + 1302 + ], + "score": 0.75, + "latex": "\\mathcal { D }" + }, + { + "category_id": 13, + "poly": [ + 1276, + 1426, + 1303, + 1426, + 1303, + 1454, + 1276, + 1454 + ], + "score": 0.74, + "latex": "\\Psi" + }, + { + "category_id": 13, + "poly": [ + 1083, + 1397, + 1104, + 1397, + 1104, + 1424, + 1083, + 1424 + ], + "score": 0.59, + "latex": "\\mathbf { d }" + }, + { + "category_id": 13, + "poly": [ + 1288, + 1956, + 1404, + 1956, + 1404, + 1990, + 1288, + 1990 + ], + "score": 0.58, + "latex": "\\mathrm { v e c } [ \\mathbf { D } ] \\sim" + }, + { + "category_id": 15, + "poly": [ + 389.0, + 269.0, + 1313.0, + 269.0, + 1313.0, + 330.0, + 389.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 667.0, + 330.0, + 1032.0, + 330.0, + 1032.0, + 381.0, + 667.0, + 381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 782.0, + 688.0, + 920.0, + 688.0, + 920.0, + 732.0, + 782.0, + 732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 243.0, + 1112.0, + 535.0, + 1112.0, + 535.0, + 1158.0, + 243.0, + 1158.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2027.0, + 1389.0, + 2027.0, + 1389.0, + 2066.0, + 294.0, + 2066.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 244.0, + 1113.0, + 535.0, + 1113.0, + 535.0, + 1159.0, + 244.0, + 1159.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 260.0, + 768.0, + 274.0, + 768.0, + 274.0, + 787.0, + 260.0, + 787.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 758.0, + 1307.0, + 758.0, + 1307.0, + 794.0, + 393.0, + 794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 259.0, + 796.0, + 277.0, + 796.0, + 277.0, + 818.0, + 259.0, + 818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 790.0, + 1305.0, + 790.0, + 1305.0, + 821.0, + 394.0, + 821.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 259.0, + 827.0, + 277.0, + 827.0, + 277.0, + 850.0, + 259.0, + 850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 820.0, + 1306.0, + 820.0, + 1306.0, + 852.0, + 393.0, + 852.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 259.0, + 857.0, + 277.0, + 857.0, + 277.0, + 880.0, + 259.0, + 880.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 849.0, + 1304.0, + 849.0, + 1304.0, + 881.0, + 394.0, + 881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 259.0, + 887.0, + 278.0, + 887.0, + 278.0, + 910.0, + 259.0, + 910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 878.0, + 1307.0, + 878.0, + 1307.0, + 915.0, + 393.0, + 915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 259.0, + 918.0, + 277.0, + 918.0, + 277.0, + 942.0, + 259.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 911.0, + 1307.0, + 911.0, + 1307.0, + 947.0, + 394.0, + 947.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 259.0, + 948.0, + 278.0, + 948.0, + 278.0, + 971.0, + 259.0, + 971.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 938.0, + 1307.0, + 938.0, + 1307.0, + 978.0, + 392.0, + 978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 259.0, + 978.0, + 277.0, + 978.0, + 277.0, + 1002.0, + 259.0, + 1002.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 971.0, + 1305.0, + 971.0, + 1305.0, + 1003.0, + 394.0, + 1003.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 259.0, + 1009.0, + 277.0, + 1009.0, + 277.0, + 1032.0, + 259.0, + 1032.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 1001.0, + 1306.0, + 1001.0, + 1306.0, + 1037.0, + 395.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1037.0, + 279.0, + 1037.0, + 279.0, + 1065.0, + 248.0, + 1065.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 1031.0, + 715.0, + 1031.0, + 715.0, + 1065.0, + 395.0, + 1065.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 251.0, + 1189.0, + 279.0, + 1189.0, + 279.0, + 1214.0, + 251.0, + 1214.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1182.0, + 1408.0, + 1182.0, + 1408.0, + 1216.0, + 291.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1219.0, + 278.0, + 1219.0, + 278.0, + 1247.0, + 249.0, + 1247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1211.0, + 1407.0, + 1211.0, + 1407.0, + 1252.0, + 291.0, + 1252.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1249.0, + 279.0, + 1249.0, + 279.0, + 1276.0, + 249.0, + 1276.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1244.0, + 835.0, + 1244.0, + 835.0, + 1279.0, + 294.0, + 1279.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 884.0, + 1244.0, + 1407.0, + 1244.0, + 1407.0, + 1279.0, + 884.0, + 1279.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1281.0, + 278.0, + 1281.0, + 278.0, + 1307.0, + 249.0, + 1307.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1274.0, + 402.0, + 1274.0, + 402.0, + 1310.0, + 292.0, + 1310.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 429.0, + 1274.0, + 1405.0, + 1274.0, + 1405.0, + 1310.0, + 429.0, + 1310.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1311.0, + 278.0, + 1311.0, + 278.0, + 1336.0, + 248.0, + 1336.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 1301.0, + 591.0, + 1301.0, + 591.0, + 1344.0, + 288.0, + 1344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 906.0, + 1301.0, + 1408.0, + 1301.0, + 1408.0, + 1344.0, + 906.0, + 1344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1340.0, + 278.0, + 1340.0, + 278.0, + 1367.0, + 249.0, + 1367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1334.0, + 1405.0, + 1334.0, + 1405.0, + 1369.0, + 295.0, + 1369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 247.0, + 1370.0, + 281.0, + 1370.0, + 281.0, + 1399.0, + 247.0, + 1399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1367.0, + 1404.0, + 1367.0, + 1404.0, + 1398.0, + 295.0, + 1398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1402.0, + 278.0, + 1402.0, + 278.0, + 1428.0, + 248.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1394.0, + 1082.0, + 1394.0, + 1082.0, + 1430.0, + 292.0, + 1430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1105.0, + 1394.0, + 1407.0, + 1394.0, + 1407.0, + 1430.0, + 1105.0, + 1430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1431.0, + 278.0, + 1431.0, + 278.0, + 1457.0, + 248.0, + 1457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1421.0, + 296.0, + 1421.0, + 296.0, + 1461.0, + 290.0, + 1461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 488.0, + 1421.0, + 721.0, + 1421.0, + 721.0, + 1461.0, + 488.0, + 1461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 750.0, + 1421.0, + 1275.0, + 1421.0, + 1275.0, + 1461.0, + 750.0, + 1461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1304.0, + 1421.0, + 1407.0, + 1421.0, + 1407.0, + 1461.0, + 1304.0, + 1461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1461.0, + 278.0, + 1461.0, + 278.0, + 1488.0, + 248.0, + 1488.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1455.0, + 415.0, + 1455.0, + 415.0, + 1492.0, + 291.0, + 1492.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1871.0, + 279.0, + 1871.0, + 279.0, + 1899.0, + 249.0, + 1899.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1860.0, + 1406.0, + 1860.0, + 1406.0, + 1904.0, + 293.0, + 1904.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1901.0, + 279.0, + 1901.0, + 279.0, + 1928.0, + 248.0, + 1928.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1893.0, + 1405.0, + 1893.0, + 1405.0, + 1931.0, + 293.0, + 1931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1931.0, + 279.0, + 1931.0, + 279.0, + 1959.0, + 249.0, + 1959.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1924.0, + 1406.0, + 1924.0, + 1406.0, + 1960.0, + 295.0, + 1960.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1963.0, + 279.0, + 1963.0, + 279.0, + 1989.0, + 249.0, + 1989.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1953.0, + 495.0, + 1953.0, + 495.0, + 1993.0, + 291.0, + 1993.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 524.0, + 1953.0, + 1287.0, + 1953.0, + 1287.0, + 1993.0, + 524.0, + 1993.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1704.0, + 279.0, + 1704.0, + 279.0, + 1729.0, + 248.0, + 1729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1695.0, + 1409.0, + 1695.0, + 1409.0, + 1735.0, + 292.0, + 1735.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1735.0, + 279.0, + 1735.0, + 279.0, + 1760.0, + 248.0, + 1760.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1725.0, + 1407.0, + 1725.0, + 1407.0, + 1765.0, + 292.0, + 1765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1766.0, + 278.0, + 1766.0, + 278.0, + 1791.0, + 248.0, + 1791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1753.0, + 795.0, + 1753.0, + 795.0, + 1799.0, + 290.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 823.0, + 1753.0, + 972.0, + 1753.0, + 972.0, + 1799.0, + 823.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1003.0, + 1753.0, + 1034.0, + 1753.0, + 1034.0, + 1799.0, + 1003.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1282.0, + 1753.0, + 1412.0, + 1753.0, + 1412.0, + 1799.0, + 1282.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1795.0, + 278.0, + 1795.0, + 278.0, + 1821.0, + 248.0, + 1821.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1788.0, + 628.0, + 1788.0, + 628.0, + 1825.0, + 294.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 658.0, + 1788.0, + 1405.0, + 1788.0, + 1405.0, + 1825.0, + 658.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1825.0, + 281.0, + 1825.0, + 281.0, + 1850.0, + 248.0, + 1850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1817.0, + 536.0, + 1817.0, + 536.0, + 1854.0, + 292.0, + 1854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1506.0, + 277.0, + 1506.0, + 277.0, + 1533.0, + 248.0, + 1533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1497.0, + 1404.0, + 1497.0, + 1404.0, + 1537.0, + 292.0, + 1537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1537.0, + 277.0, + 1537.0, + 277.0, + 1563.0, + 248.0, + 1563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1533.0, + 1404.0, + 1533.0, + 1404.0, + 1564.0, + 295.0, + 1564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1568.0, + 277.0, + 1568.0, + 277.0, + 1594.0, + 248.0, + 1594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1562.0, + 1406.0, + 1562.0, + 1406.0, + 1596.0, + 293.0, + 1596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1597.0, + 277.0, + 1597.0, + 277.0, + 1624.0, + 248.0, + 1624.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1594.0, + 1404.0, + 1594.0, + 1404.0, + 1625.0, + 294.0, + 1625.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1628.0, + 278.0, + 1628.0, + 278.0, + 1654.0, + 248.0, + 1654.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1621.0, + 1404.0, + 1621.0, + 1404.0, + 1657.0, + 291.0, + 1657.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1659.0, + 278.0, + 1659.0, + 278.0, + 1684.0, + 248.0, + 1684.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1653.0, + 466.0, + 1653.0, + 466.0, + 1690.0, + 291.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 719.0, + 490.0, + 987.0, + 490.0, + 987.0, + 522.0, + 719.0, + 522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 517.0, + 913.0, + 517.0, + 913.0, + 553.0, + 787.0, + 553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 797.0, + 548.0, + 901.0, + 548.0, + 901.0, + 583.0, + 797.0, + 583.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 808.0, + 584.0, + 890.0, + 584.0, + 890.0, + 610.0, + 808.0, + 610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2027.0, + 1389.0, + 2027.0, + 1389.0, + 2066.0, + 294.0, + 2066.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 0, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 252, + 701, + 1405, + 701, + 1405, + 795, + 252, + 795 + ], + "score": 0.971 + }, + { + "category_id": 1, + "poly": [ + 249, + 1263, + 1404, + 1263, + 1404, + 1478, + 249, + 1478 + ], + "score": 0.967 + }, + { + "category_id": 1, + "poly": [ + 250, + 898, + 1404, + 898, + 1404, + 1115, + 250, + 1115 + ], + "score": 0.967 + }, + { + "category_id": 3, + "poly": [ + 453, + 187, + 1238, + 187, + 1238, + 558, + 453, + 558 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 251, + 1490, + 1403, + 1490, + 1403, + 1614, + 251, + 1614 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 251, + 1716, + 1406, + 1716, + 1406, + 1811, + 251, + 1811 + ], + "score": 0.965 + }, + { + "category_id": 1, + "poly": [ + 251, + 1823, + 1404, + 1823, + 1404, + 2008, + 251, + 2008 + ], + "score": 0.961 + }, + { + "category_id": 1, + "poly": [ + 251, + 1125, + 1404, + 1125, + 1404, + 1249, + 251, + 1249 + ], + "score": 0.96 + }, + { + "category_id": 4, + "poly": [ + 296, + 571, + 1401, + 571, + 1401, + 636, + 296, + 636 + ], + "score": 0.938 + }, + { + "category_id": 0, + "poly": [ + 252, + 839, + 494, + 839, + 494, + 872, + 252, + 872 + ], + "score": 0.898 + }, + { + "category_id": 0, + "poly": [ + 253, + 1658, + 507, + 1658, + 507, + 1690, + 253, + 1690 + ], + "score": 0.871 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.717 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.14 + }, + { + "category_id": 13, + "poly": [ + 632, + 1292, + 733, + 1292, + 733, + 1327, + 632, + 1327 + ], + "score": 0.93, + "latex": "O ( n ^ { 2 } p ^ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 1028, + 1353, + 1162, + 1353, + 1162, + 1388, + 1028, + 1388 + ], + "score": 0.93, + "latex": "O ( n ^ { 2 } + p ^ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 624, + 1415, + 988, + 1415, + 988, + 1448, + 624, + 1448 + ], + "score": 0.92, + "latex": "\\zeta ( \\Psi _ { 1 } , . . . , \\Psi _ { k } ) = \\Psi _ { 1 } \\oplus . . . \\oplus \\Psi _ { k }" + }, + { + "category_id": 13, + "poly": [ + 603, + 732, + 796, + 732, + 796, + 767, + 603, + 767 + ], + "score": 0.92, + "latex": "\\zeta ( \\Psi _ { \\mathrm { r o w } } , \\Psi _ { \\mathrm { c o l } } ) ^ { - \\mathrm { i } }" + }, + { + "category_id": 13, + "poly": [ + 1143, + 991, + 1317, + 991, + 1317, + 1020, + 1143, + 1020 + ], + "score": 0.9, + "latex": "\\mathbf { A } \\otimes \\mathbf { I } + \\mathbf { I } \\otimes \\mathbf { B }" + }, + { + "category_id": 13, + "poly": [ + 297, + 700, + 568, + 700, + 568, + 736, + 297, + 736 + ], + "score": 0.9, + "latex": "\\mathcal { N } ( \\mathbf { 0 } , \\zeta ( \\Psi _ { \\mathrm { r o w } } , \\Psi _ { \\mathrm { c o l } } ) ^ { - 1 } )" + }, + { + "category_id": 13, + "poly": [ + 960, + 1080, + 1352, + 1080, + 1352, + 1115, + 960, + 1115 + ], + "score": 0.88, + "latex": "\\mathrm { v e c } [ \\mathbf { \\bar { D } } ] \\sim \\mathcal { N } ( \\mathbf { 0 } , ( \\Psi _ { \\mathrm { r o w } } \\mathbf { \\hat { \\boldsymbol { \\Phi } } } \\boldsymbol { \\Psi } _ { \\mathrm { c o l } } \\mathbf { \\bar { \\rho } } ) ^ { - 1 } )" + }, + { + "category_id": 13, + "poly": [ + 485, + 991, + 569, + 991, + 569, + 1020, + 485, + 1020 + ], + "score": 0.88, + "latex": "\\mathbf { A } \\oplus \\mathbf { B }" + }, + { + "category_id": 13, + "poly": [ + 1089, + 703, + 1152, + 703, + 1152, + 734, + 1089, + 734 + ], + "score": 0.86, + "latex": "\\Psi _ { \\mathrm { r o w } }" + }, + { + "category_id": 13, + "poly": [ + 1163, + 703, + 1219, + 703, + 1219, + 734, + 1163, + 734 + ], + "score": 0.85, + "latex": "\\Psi _ { \\mathrm { c o l } }" + }, + { + "category_id": 13, + "poly": [ + 494, + 1127, + 513, + 1127, + 513, + 1159, + 494, + 1159 + ], + "score": 0.83, + "latex": "\\zeta" + }, + { + "category_id": 13, + "poly": [ + 1000, + 1053, + 1018, + 1053, + 1018, + 1083, + 1000, + 1083 + ], + "score": 0.82, + "latex": "\\zeta" + }, + { + "category_id": 13, + "poly": [ + 1252, + 735, + 1270, + 735, + 1270, + 766, + 1252, + 766 + ], + "score": 0.81, + "latex": "\\zeta" + }, + { + "category_id": 13, + "poly": [ + 775, + 704, + 793, + 704, + 793, + 733, + 775, + 733 + ], + "score": 0.78, + "latex": "\\zeta" + }, + { + "category_id": 13, + "poly": [ + 444, + 1021, + 506, + 1021, + 506, + 1052, + 444, + 1052 + ], + "score": 0.33, + "latex": "\\mathbf { A } , \\mathbf { B }" + }, + { + "category_id": 15, + "poly": [ + 682.0, + 346.0, + 780.0, + 346.0, + 780.0, + 385.0, + 682.0, + 385.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1065.0, + 343.0, + 1221.0, + 343.0, + 1221.0, + 382.0, + 1065.0, + 382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 613.0, + 421.0, + 649.0, + 421.0, + 649.0, + 513.0, + 613.0, + 513.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1027.0, + 421.0, + 1067.0, + 421.0, + 1067.0, + 515.0, + 1027.0, + 515.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 651.0, + 458.5, + 812.0, + 458.5, + 812.0, + 486.5, + 651.0, + 486.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1071.25, + 456.5, + 1223.25, + 456.5, + 1223.25, + 480.5, + 1071.25, + 480.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 571.0, + 1405.0, + 571.0, + 1405.0, + 606.0, + 294.0, + 606.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 602.0, + 1053.0, + 602.0, + 1053.0, + 638.0, + 295.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 244.0, + 835.0, + 499.0, + 835.0, + 499.0, + 876.0, + 244.0, + 876.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 243.0, + 1653.0, + 510.0, + 1653.0, + 510.0, + 1693.0, + 243.0, + 1693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2093.0, + 838.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2093.0, + 838.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 244.0, + 698.0, + 296.0, + 698.0, + 296.0, + 740.0, + 244.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 569.0, + 698.0, + 774.0, + 698.0, + 774.0, + 740.0, + 569.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 794.0, + 698.0, + 1088.0, + 698.0, + 1088.0, + 740.0, + 794.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1153.0, + 698.0, + 1162.0, + 698.0, + 1162.0, + 740.0, + 1153.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1220.0, + 698.0, + 1408.0, + 698.0, + 1408.0, + 740.0, + 1220.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 250.0, + 741.0, + 279.0, + 741.0, + 279.0, + 764.0, + 250.0, + 764.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 730.0, + 602.0, + 730.0, + 602.0, + 771.0, + 289.0, + 771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 797.0, + 730.0, + 1251.0, + 730.0, + 1251.0, + 771.0, + 797.0, + 771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1271.0, + 730.0, + 1408.0, + 730.0, + 1408.0, + 771.0, + 1271.0, + 771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 246.0, + 764.0, + 690.0, + 764.0, + 690.0, + 798.0, + 246.0, + 798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1268.0, + 277.0, + 1268.0, + 277.0, + 1295.0, + 249.0, + 1295.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1263.0, + 1407.0, + 1263.0, + 1407.0, + 1298.0, + 293.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1299.0, + 277.0, + 1299.0, + 277.0, + 1326.0, + 249.0, + 1326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1289.0, + 631.0, + 1289.0, + 631.0, + 1331.0, + 290.0, + 1331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 734.0, + 1289.0, + 1408.0, + 1289.0, + 1408.0, + 1331.0, + 734.0, + 1331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1331.0, + 279.0, + 1331.0, + 279.0, + 1357.0, + 249.0, + 1357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1325.0, + 1404.0, + 1325.0, + 1404.0, + 1359.0, + 294.0, + 1359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1361.0, + 279.0, + 1361.0, + 279.0, + 1386.0, + 249.0, + 1386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1355.0, + 1027.0, + 1355.0, + 1027.0, + 1389.0, + 293.0, + 1389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1163.0, + 1355.0, + 1406.0, + 1355.0, + 1406.0, + 1389.0, + 1163.0, + 1389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 246.0, + 1389.0, + 280.0, + 1389.0, + 280.0, + 1417.0, + 246.0, + 1417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1384.0, + 1406.0, + 1384.0, + 1406.0, + 1418.0, + 294.0, + 1418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1421.0, + 277.0, + 1421.0, + 277.0, + 1446.0, + 249.0, + 1446.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1415.0, + 623.0, + 1415.0, + 623.0, + 1449.0, + 293.0, + 1449.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 989.0, + 1415.0, + 1407.0, + 1415.0, + 1407.0, + 1449.0, + 989.0, + 1449.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1452.0, + 277.0, + 1452.0, + 277.0, + 1479.0, + 249.0, + 1479.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1446.0, + 1130.0, + 1446.0, + 1130.0, + 1480.0, + 293.0, + 1480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 905.0, + 276.0, + 905.0, + 276.0, + 930.0, + 248.0, + 930.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 899.0, + 1408.0, + 899.0, + 1408.0, + 933.0, + 293.0, + 933.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 935.0, + 277.0, + 935.0, + 277.0, + 961.0, + 248.0, + 961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 930.0, + 1406.0, + 930.0, + 1406.0, + 964.0, + 295.0, + 964.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 967.0, + 278.0, + 967.0, + 278.0, + 992.0, + 248.0, + 992.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 959.0, + 1404.0, + 959.0, + 1404.0, + 994.0, + 294.0, + 994.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 997.0, + 278.0, + 997.0, + 278.0, + 1021.0, + 248.0, + 1021.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 992.0, + 484.0, + 992.0, + 484.0, + 1023.0, + 294.0, + 1023.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 992.0, + 1142.0, + 992.0, + 1142.0, + 1023.0, + 570.0, + 1023.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1318.0, + 992.0, + 1404.0, + 992.0, + 1404.0, + 1023.0, + 1318.0, + 1023.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1026.0, + 278.0, + 1026.0, + 278.0, + 1052.0, + 248.0, + 1052.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1018.0, + 443.0, + 1018.0, + 443.0, + 1056.0, + 293.0, + 1056.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 507.0, + 1018.0, + 1406.0, + 1018.0, + 1406.0, + 1056.0, + 507.0, + 1056.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1057.0, + 278.0, + 1057.0, + 278.0, + 1082.0, + 248.0, + 1082.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1050.0, + 999.0, + 1050.0, + 999.0, + 1086.0, + 291.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1019.0, + 1050.0, + 1406.0, + 1050.0, + 1406.0, + 1086.0, + 1019.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1088.0, + 278.0, + 1088.0, + 278.0, + 1113.0, + 248.0, + 1113.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1080.0, + 959.0, + 1080.0, + 959.0, + 1118.0, + 291.0, + 1118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1353.0, + 1080.0, + 1364.0, + 1080.0, + 1364.0, + 1118.0, + 1353.0, + 1118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1496.0, + 278.0, + 1496.0, + 278.0, + 1523.0, + 249.0, + 1523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1489.0, + 1407.0, + 1489.0, + 1407.0, + 1528.0, + 294.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1527.0, + 279.0, + 1527.0, + 279.0, + 1553.0, + 248.0, + 1553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1517.0, + 1405.0, + 1517.0, + 1405.0, + 1559.0, + 292.0, + 1559.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1556.0, + 278.0, + 1556.0, + 278.0, + 1583.0, + 248.0, + 1583.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1551.0, + 1405.0, + 1551.0, + 1405.0, + 1586.0, + 296.0, + 1586.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1587.0, + 279.0, + 1587.0, + 279.0, + 1614.0, + 248.0, + 1614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1579.0, + 836.0, + 1579.0, + 836.0, + 1619.0, + 291.0, + 1619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 251.0, + 1723.0, + 278.0, + 1723.0, + 278.0, + 1748.0, + 251.0, + 1748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1717.0, + 1408.0, + 1717.0, + 1408.0, + 1751.0, + 291.0, + 1751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1756.0, + 279.0, + 1756.0, + 279.0, + 1779.0, + 249.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1745.0, + 1409.0, + 1745.0, + 1409.0, + 1785.0, + 290.0, + 1785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 244.0, + 1779.0, + 808.0, + 1779.0, + 808.0, + 1812.0, + 244.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1829.0, + 278.0, + 1829.0, + 278.0, + 1854.0, + 248.0, + 1854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1824.0, + 1406.0, + 1824.0, + 1406.0, + 1856.0, + 295.0, + 1856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1861.0, + 278.0, + 1861.0, + 278.0, + 1886.0, + 248.0, + 1886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1853.0, + 1407.0, + 1853.0, + 1407.0, + 1887.0, + 292.0, + 1887.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1890.0, + 278.0, + 1890.0, + 278.0, + 1915.0, + 248.0, + 1915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1882.0, + 1407.0, + 1882.0, + 1407.0, + 1919.0, + 292.0, + 1919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1921.0, + 279.0, + 1921.0, + 279.0, + 1946.0, + 248.0, + 1946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1914.0, + 1407.0, + 1914.0, + 1407.0, + 1950.0, + 295.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1950.0, + 278.0, + 1950.0, + 278.0, + 1977.0, + 248.0, + 1977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1943.0, + 1404.0, + 1943.0, + 1404.0, + 1979.0, + 294.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1982.0, + 278.0, + 1982.0, + 278.0, + 2008.0, + 249.0, + 2008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1975.0, + 1394.0, + 1975.0, + 1394.0, + 2010.0, + 292.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1132.0, + 278.0, + 1132.0, + 278.0, + 1159.0, + 249.0, + 1159.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1123.0, + 493.0, + 1123.0, + 493.0, + 1163.0, + 294.0, + 1163.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 514.0, + 1123.0, + 1406.0, + 1123.0, + 1406.0, + 1163.0, + 514.0, + 1163.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1163.0, + 278.0, + 1163.0, + 278.0, + 1190.0, + 248.0, + 1190.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1159.0, + 1403.0, + 1159.0, + 1403.0, + 1190.0, + 295.0, + 1190.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1193.0, + 279.0, + 1193.0, + 279.0, + 1221.0, + 248.0, + 1221.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1184.0, + 1406.0, + 1184.0, + 1406.0, + 1223.0, + 292.0, + 1223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1222.0, + 278.0, + 1222.0, + 278.0, + 1249.0, + 248.0, + 1249.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1213.0, + 449.0, + 1213.0, + 449.0, + 1253.0, + 293.0, + 1253.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 1, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 3, + "poly": [ + 572, + 193, + 1126, + 193, + 1126, + 831, + 572, + 831 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 251, + 1362, + 1405, + 1362, + 1405, + 1547, + 251, + 1547 + ], + "score": 0.967 + }, + { + "category_id": 4, + "poly": [ + 295, + 851, + 1405, + 851, + 1405, + 1005, + 295, + 1005 + ], + "score": 0.964 + }, + { + "category_id": 1, + "poly": [ + 251, + 1073, + 1404, + 1073, + 1404, + 1257, + 251, + 1257 + ], + "score": 0.964 + }, + { + "category_id": 1, + "poly": [ + 250, + 1823, + 1403, + 1823, + 1403, + 2013, + 250, + 2013 + ], + "score": 0.952 + }, + { + "category_id": 1, + "poly": [ + 293, + 1728, + 1402, + 1728, + 1402, + 1813, + 293, + 1813 + ], + "score": 0.944 + }, + { + "category_id": 0, + "poly": [ + 254, + 1302, + 577, + 1302, + 577, + 1335, + 254, + 1335 + ], + "score": 0.897 + }, + { + "category_id": 0, + "poly": [ + 291, + 1597, + 474, + 1597, + 474, + 1634, + 291, + 1634 + ], + "score": 0.884 + }, + { + "category_id": 0, + "poly": [ + 298, + 1672, + 467, + 1672, + 467, + 1704, + 298, + 1704 + ], + "score": 0.878 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.65 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.373 + }, + { + "category_id": 13, + "poly": [ + 954, + 1726, + 1256, + 1726, + 1256, + 1783, + 954, + 1783 + ], + "score": 0.93, + "latex": "[ \\mathcal { D } ] \\sim \\mathcal { N } \\left( \\mathbf { 0 } , ( \\oplus _ { \\ell } \\Psi _ { \\ell } ) ^ { - 1 } \\right)" + }, + { + "category_id": 13, + "poly": [ + 297, + 1974, + 442, + 1974, + 442, + 2014, + 297, + 2014 + ], + "score": 0.92, + "latex": "\\begin{array} { r } { d \\check { \\forall } = \\sum _ { \\gamma } \\dot { d } _ { \\forall } ^ { \\gamma } } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 1186, + 1855, + 1259, + 1855, + 1259, + 1887, + 1186, + 1887 + ], + "score": 0.91, + "latex": "\\gamma \\in \\mathbb { N }" + }, + { + "category_id": 13, + "poly": [ + 993, + 1886, + 1059, + 1886, + 1059, + 1917, + 993, + 1917 + ], + "score": 0.9, + "latex": "\\ell \\in \\gamma" + }, + { + "category_id": 13, + "poly": [ + 801, + 1946, + 832, + 1946, + 832, + 1980, + 801, + 1980 + ], + "score": 0.89, + "latex": "d _ { \\breve { \\breve { \\breve { \\breve { \\tau } } } } } ^ { \\gamma }" + }, + { + "category_id": 13, + "poly": [ + 328, + 1776, + 548, + 1776, + 548, + 1813, + 328, + 1813 + ], + "score": 0.89, + "latex": "\\mathcal { D } \\sim \\mathcal { N } _ { K S } \\left( \\{ \\Psi _ { \\ell } \\} _ { \\ell } \\right)" + }, + { + "category_id": 13, + "poly": [ + 642, + 1855, + 680, + 1855, + 680, + 1885, + 642, + 1885 + ], + "score": 0.88, + "latex": "\\Psi _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 1305, + 1947, + 1344, + 1947, + 1344, + 1973, + 1305, + 1973 + ], + "score": 0.87, + "latex": "\\mathcal { D } ^ { \\gamma }" + }, + { + "category_id": 13, + "poly": [ + 1010, + 1855, + 1048, + 1855, + 1048, + 1882, + 1010, + 1882 + ], + "score": 0.85, + "latex": "\\mathcal { D } ^ { \\gamma }" + }, + { + "category_id": 13, + "poly": [ + 785, + 1886, + 823, + 1886, + 823, + 1912, + 785, + 1912 + ], + "score": 0.84, + "latex": "\\mathcal { D } ^ { \\gamma }" + }, + { + "category_id": 13, + "poly": [ + 714, + 1738, + 740, + 1738, + 740, + 1764, + 714, + 1764 + ], + "score": 0.84, + "latex": "\\mathcal { D }" + }, + { + "category_id": 13, + "poly": [ + 831, + 1856, + 846, + 1856, + 846, + 1882, + 831, + 1882 + ], + "score": 0.79, + "latex": "\\ell" + }, + { + "category_id": 13, + "poly": [ + 434, + 1887, + 449, + 1887, + 449, + 1912, + 434, + 1912 + ], + "score": 0.75, + "latex": "\\ell" + }, + { + "category_id": 13, + "poly": [ + 757, + 1951, + 775, + 1951, + 775, + 1978, + 757, + 1978 + ], + "score": 0.72, + "latex": "\\gamma" + }, + { + "category_id": 13, + "poly": [ + 1330, + 1917, + 1346, + 1917, + 1346, + 1942, + 1330, + 1942 + ], + "score": 0.67, + "latex": "\\ell" + }, + { + "category_id": 15, + "poly": [ + 758.0, + 295.0, + 821.0, + 295.0, + 821.0, + 319.0, + 758.0, + 319.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 981.0, + 292.0, + 1072.0, + 292.0, + 1072.0, + 318.0, + 981.0, + 318.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 332.0, + 747.0, + 332.0, + 747.0, + 407.0, + 721.0, + 407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 959.0, + 336.0, + 983.0, + 336.0, + 983.0, + 395.0, + 959.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 980.0, + 436.0, + 1074.0, + 436.0, + 1074.0, + 459.0, + 980.0, + 459.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 959.0, + 470.0, + 982.0, + 470.0, + 982.0, + 555.0, + 959.0, + 555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 715.0, + 489.0, + 776.0, + 489.0, + 776.0, + 541.0, + 715.0, + 541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 825.0, + 597.0, + 882.0, + 597.0, + 882.0, + 644.0, + 825.0, + 644.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 655.0, + 661.0, + 733.0, + 661.0, + 733.0, + 688.0, + 655.0, + 688.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 990.0, + 658.0, + 1069.0, + 658.0, + 1069.0, + 687.0, + 990.0, + 687.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 661.0, + 678.0, + 730.0, + 678.0, + 730.0, + 705.0, + 661.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 995.0, + 675.0, + 1068.0, + 675.0, + 1068.0, + 702.0, + 995.0, + 702.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 606.0, + 724.0, + 615.0, + 724.0, + 615.0, + 733.0, + 606.0, + 733.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 626.0, + 717.0, + 650.0, + 717.0, + 650.0, + 774.0, + 626.0, + 774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 959.0, + 718.0, + 982.0, + 718.0, + 982.0, + 775.0, + 959.0, + 775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 738.0, + 357.0, + 842.0, + 357.0, + 842.0, + 379.0, + 738.0, + 379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 981.0, + 356.5, + 1074.0, + 356.5, + 1074.0, + 375.5, + 981.0, + 375.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 992.0, + 498.5, + 1064.0, + 498.5, + 1064.0, + 518.0, + 992.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 643.75, + 738.0, + 745.75, + 738.0, + 745.75, + 757.0, + 643.75, + 757.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 977.0, + 738.0, + 1080.0, + 738.0, + 1080.0, + 757.5, + 977.0, + 757.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 850.0, + 1405.0, + 850.0, + 1405.0, + 888.0, + 293.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 882.0, + 1405.0, + 882.0, + 1405.0, + 919.0, + 294.0, + 919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 913.0, + 1406.0, + 913.0, + 1406.0, + 949.0, + 293.0, + 949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 944.0, + 1406.0, + 944.0, + 1406.0, + 981.0, + 295.0, + 981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 974.0, + 733.0, + 974.0, + 733.0, + 1010.0, + 294.0, + 1010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 244.0, + 1299.0, + 582.0, + 1299.0, + 582.0, + 1339.0, + 244.0, + 1339.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1592.0, + 480.0, + 1592.0, + 480.0, + 1640.0, + 291.0, + 1640.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1668.0, + 470.0, + 1668.0, + 470.0, + 1708.0, + 294.0, + 1708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1368.0, + 278.0, + 1368.0, + 278.0, + 1393.0, + 248.0, + 1393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1363.0, + 1408.0, + 1363.0, + 1408.0, + 1395.0, + 295.0, + 1395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1399.0, + 277.0, + 1399.0, + 277.0, + 1425.0, + 248.0, + 1425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1392.0, + 1404.0, + 1392.0, + 1404.0, + 1425.0, + 294.0, + 1425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1429.0, + 278.0, + 1429.0, + 278.0, + 1454.0, + 248.0, + 1454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1423.0, + 1405.0, + 1423.0, + 1405.0, + 1458.0, + 294.0, + 1458.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1460.0, + 278.0, + 1460.0, + 278.0, + 1486.0, + 248.0, + 1486.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1454.0, + 1404.0, + 1454.0, + 1404.0, + 1486.0, + 296.0, + 1486.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1490.0, + 278.0, + 1490.0, + 278.0, + 1516.0, + 248.0, + 1516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1482.0, + 1404.0, + 1482.0, + 1404.0, + 1518.0, + 294.0, + 1518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 247.0, + 1518.0, + 281.0, + 1518.0, + 281.0, + 1547.0, + 247.0, + 1547.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1513.0, + 466.0, + 1513.0, + 466.0, + 1549.0, + 294.0, + 1549.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1080.0, + 278.0, + 1080.0, + 278.0, + 1105.0, + 249.0, + 1105.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1071.0, + 1406.0, + 1071.0, + 1406.0, + 1111.0, + 294.0, + 1111.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1110.0, + 278.0, + 1110.0, + 278.0, + 1136.0, + 248.0, + 1136.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1104.0, + 1407.0, + 1104.0, + 1407.0, + 1138.0, + 296.0, + 1138.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1141.0, + 278.0, + 1141.0, + 278.0, + 1166.0, + 249.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1133.0, + 1406.0, + 1133.0, + 1406.0, + 1168.0, + 292.0, + 1168.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1171.0, + 278.0, + 1171.0, + 278.0, + 1197.0, + 248.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1164.0, + 1407.0, + 1164.0, + 1407.0, + 1202.0, + 292.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1201.0, + 278.0, + 1201.0, + 278.0, + 1227.0, + 249.0, + 1227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1195.0, + 1406.0, + 1195.0, + 1406.0, + 1231.0, + 294.0, + 1231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 249.0, + 1231.0, + 278.0, + 1231.0, + 278.0, + 1258.0, + 249.0, + 1258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1225.0, + 437.0, + 1225.0, + 437.0, + 1261.0, + 292.0, + 1261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1829.0, + 278.0, + 1829.0, + 278.0, + 1855.0, + 248.0, + 1855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1821.0, + 1406.0, + 1821.0, + 1406.0, + 1858.0, + 293.0, + 1858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1860.0, + 277.0, + 1860.0, + 277.0, + 1886.0, + 248.0, + 1886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1855.0, + 641.0, + 1855.0, + 641.0, + 1887.0, + 295.0, + 1887.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 681.0, + 1855.0, + 830.0, + 1855.0, + 830.0, + 1887.0, + 681.0, + 1887.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 847.0, + 1855.0, + 1009.0, + 1855.0, + 1009.0, + 1887.0, + 847.0, + 1887.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1049.0, + 1855.0, + 1185.0, + 1855.0, + 1185.0, + 1887.0, + 1049.0, + 1887.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1260.0, + 1855.0, + 1405.0, + 1855.0, + 1405.0, + 1887.0, + 1260.0, + 1887.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1890.0, + 278.0, + 1890.0, + 278.0, + 1915.0, + 248.0, + 1915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1884.0, + 433.0, + 1884.0, + 433.0, + 1917.0, + 295.0, + 1917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 1884.0, + 784.0, + 1884.0, + 784.0, + 1917.0, + 450.0, + 1917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 824.0, + 1884.0, + 992.0, + 1884.0, + 992.0, + 1917.0, + 824.0, + 1917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1060.0, + 1884.0, + 1405.0, + 1884.0, + 1405.0, + 1917.0, + 1060.0, + 1917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1921.0, + 278.0, + 1921.0, + 278.0, + 1946.0, + 248.0, + 1946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1913.0, + 1329.0, + 1913.0, + 1329.0, + 1949.0, + 294.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1347.0, + 1913.0, + 1405.0, + 1913.0, + 1405.0, + 1949.0, + 1347.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 1952.0, + 278.0, + 1952.0, + 278.0, + 1976.0, + 248.0, + 1976.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1944.0, + 756.0, + 1944.0, + 756.0, + 1980.0, + 290.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 776.0, + 1944.0, + 800.0, + 1944.0, + 800.0, + 1980.0, + 776.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 833.0, + 1944.0, + 1304.0, + 1944.0, + 1304.0, + 1980.0, + 833.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1345.0, + 1944.0, + 1406.0, + 1944.0, + 1406.0, + 1980.0, + 1345.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 250.0, + 1980.0, + 278.0, + 1980.0, + 278.0, + 2007.0, + 250.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1972.0, + 296.0, + 1972.0, + 296.0, + 2015.0, + 289.0, + 2015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 443.0, + 1972.0, + 455.0, + 1972.0, + 455.0, + 2015.0, + 443.0, + 2015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 284.0, + 1724.0, + 713.0, + 1724.0, + 713.0, + 1781.0, + 284.0, + 1781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 741.0, + 1724.0, + 953.0, + 1724.0, + 953.0, + 1781.0, + 741.0, + 1781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1257.0, + 1724.0, + 1410.0, + 1724.0, + 1410.0, + 1781.0, + 1257.0, + 1781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1772.0, + 327.0, + 1772.0, + 327.0, + 1818.0, + 290.0, + 1818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 549.0, + 1772.0, + 561.0, + 1772.0, + 561.0, + 1818.0, + 549.0, + 1818.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 2, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 246, + 200, + 1405, + 200, + 1405, + 397, + 246, + 397 + ], + "score": 0.972 + }, + { + "category_id": 1, + "poly": [ + 242, + 487, + 1405, + 487, + 1405, + 611, + 242, + 611 + ], + "score": 0.971 + }, + { + "category_id": 1, + "poly": [ + 242, + 1238, + 1405, + 1238, + 1405, + 1333, + 242, + 1333 + ], + "score": 0.965 + }, + { + "category_id": 8, + "poly": [ + 590, + 1066, + 1108, + 1066, + 1108, + 1155, + 590, + 1155 + ], + "score": 0.942 + }, + { + "category_id": 1, + "poly": [ + 241, + 783, + 1406, + 783, + 1406, + 939, + 241, + 939 + ], + "score": 0.94 + }, + { + "category_id": 1, + "poly": [ + 242, + 1867, + 1406, + 1867, + 1406, + 1962, + 242, + 1962 + ], + "score": 0.934 + }, + { + "category_id": 0, + "poly": [ + 244, + 1182, + 535, + 1182, + 535, + 1215, + 244, + 1215 + ], + "score": 0.925 + }, + { + "category_id": 8, + "poly": [ + 714, + 659, + 983, + 659, + 983, + 761, + 714, + 761 + ], + "score": 0.922 + }, + { + "category_id": 1, + "poly": [ + 239, + 1684, + 1407, + 1684, + 1407, + 1778, + 239, + 1778 + ], + "score": 0.912 + }, + { + "category_id": 1, + "poly": [ + 241, + 1344, + 917, + 1344, + 917, + 1379, + 241, + 1379 + ], + "score": 0.906 + }, + { + "category_id": 1, + "poly": [ + 244, + 1974, + 1040, + 1974, + 1040, + 2009, + 244, + 2009 + ], + "score": 0.905 + }, + { + "category_id": 1, + "poly": [ + 241, + 949, + 1408, + 949, + 1408, + 1016, + 241, + 1016 + ], + "score": 0.905 + }, + { + "category_id": 0, + "poly": [ + 263, + 431, + 490, + 431, + 490, + 464, + 263, + 464 + ], + "score": 0.887 + }, + { + "category_id": 1, + "poly": [ + 258, + 1782, + 1399, + 1782, + 1399, + 1849, + 258, + 1849 + ], + "score": 0.866 + }, + { + "category_id": 1, + "poly": [ + 1127, + 1591, + 1403, + 1591, + 1403, + 1624, + 1127, + 1624 + ], + "score": 0.856 + }, + { + "category_id": 2, + "poly": [ + 840, + 2061, + 858, + 2061, + 858, + 2084, + 840, + 2084 + ], + "score": 0.779 + }, + { + "category_id": 1, + "poly": [ + 1219, + 1484, + 1402, + 1484, + 1402, + 1519, + 1219, + 1519 + ], + "score": 0.498 + }, + { + "category_id": 8, + "poly": [ + 438, + 1558, + 989, + 1558, + 989, + 1661, + 438, + 1661 + ], + "score": 0.455 + }, + { + "category_id": 9, + "poly": [ + 1219, + 1484, + 1402, + 1484, + 1402, + 1519, + 1219, + 1519 + ], + "score": 0.439 + }, + { + "category_id": 8, + "poly": [ + 471, + 1427, + 996, + 1427, + 996, + 1551, + 471, + 1551 + ], + "score": 0.418 + }, + { + "category_id": 8, + "poly": [ + 437, + 1425, + 997, + 1425, + 997, + 1663, + 437, + 1663 + ], + "score": 0.389 + }, + { + "category_id": 8, + "poly": [ + 428, + 1425, + 1406, + 1425, + 1406, + 1664, + 428, + 1664 + ], + "score": 0.142 + }, + { + "category_id": 8, + "poly": [ + 714, + 659, + 983, + 659, + 983, + 762, + 714, + 762 + ], + "score": 0.118 + }, + { + "category_id": 13, + "poly": [ + 298, + 360, + 488, + 360, + 488, + 400, + 298, + 400 + ], + "score": 0.94, + "latex": "\\begin{array} { r } { \\mathbf { S } _ { \\ell } = \\sum _ { \\gamma | \\ell \\in \\gamma } \\mathbf { S } _ { \\ell } ^ { \\gamma } } \\end{array}" + }, + { + "category_id": 14, + "poly": [ + 435, + 1424, + 999, + 1424, + 999, + 1667, + 435, + 1667 + ], + "score": 0.93, + "latex": "\\begin{array} { r } { p ( \\{ \\mathcal { D } ^ { \\gamma } \\} ) = \\frac { \\displaystyle \\prod _ { \\gamma } \\sqrt { \\left| \\bigoplus _ { \\ell \\in \\gamma } \\Psi _ { \\ell } \\right| } } { \\displaystyle ( 2 \\pi ) ^ { \\frac { d _ { \\psi } } { 2 } } } e ^ { \\frac { - 1 } { 2 } \\sum _ { \\ell } \\mathrm { t r } [ \\Psi _ { \\ell } \\mathbf { S } _ { \\ell } ] } } \\\\ { \\mathrm { N L L } \\left[ \\{ \\mathcal { D } ^ { \\gamma } \\} \\right] \\propto \\displaystyle \\sum _ { \\ell } \\mathrm { t r } \\left[ \\Psi _ { \\ell } \\mathbf { S } _ { \\ell } \\right] - \\sum _ { \\gamma } \\log \\left| \\bigoplus _ { \\ell \\in \\gamma } \\Psi _ { \\ell } \\right| } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 1053, + 1781, + 1203, + 1781, + 1203, + 1815, + 1053, + 1815 + ], + "score": 0.92, + "latex": "\\mathbf { V } _ { \\ell } \\in \\mathbb { R } ^ { d _ { \\ell } \\times d _ { \\ell } }" + }, + { + "category_id": 13, + "poly": [ + 490, + 1781, + 594, + 1781, + 594, + 1818, + 490, + 1818 + ], + "score": 0.92, + "latex": "\\mathbf { V } _ { \\ell } \\mathbf { e } _ { \\ell } \\mathbf { V } _ { \\ell } ^ { T }" + }, + { + "category_id": 13, + "poly": [ + 1257, + 1781, + 1399, + 1781, + 1399, + 1815, + 1257, + 1815 + ], + "score": 0.91, + "latex": "\\mathbf { e } _ { \\ell } \\in \\mathbb { R } ^ { d _ { \\ell } \\times d _ { \\ell } }" + }, + { + "category_id": 13, + "poly": [ + 423, + 1270, + 461, + 1270, + 461, + 1300, + 423, + 1300 + ], + "score": 0.9, + "latex": "\\Psi _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 802, + 265, + 1129, + 265, + 1129, + 304, + 802, + 304 + ], + "score": 0.9, + "latex": "\\mathbf { S } _ { \\ell } ^ { \\gamma } = \\operatorname* { m a t } _ { \\ell } \\left[ \\mathcal { D } ^ { \\gamma } \\right] \\operatorname* { m a t } _ { \\ell } \\left[ \\mathcal { D } ^ { \\gamma } \\right] ^ { T }" + }, + { + "category_id": 14, + "poly": [ + 715, + 659, + 984, + 659, + 984, + 765, + 715, + 765 + ], + "score": 0.89, + "latex": "\\begin{array} { r } { \\mathcal { D } ^ { \\gamma } \\sim \\mathcal { N } _ { K S } \\left( \\{ \\Psi _ { \\ell } \\} _ { \\ell \\in \\gamma } \\right) } \\\\ { \\mathrm { f o r } \\mathcal { D } ^ { \\gamma } \\in \\{ \\mathcal { D } ^ { \\gamma } \\} _ { \\gamma } } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 868, + 330, + 902, + 330, + 902, + 366, + 868, + 366 + ], + "score": 0.89, + "latex": "\\mathbf { S } _ { \\ell } ^ { \\gamma }" + }, + { + "category_id": 13, + "poly": [ + 328, + 1301, + 366, + 1301, + 366, + 1330, + 328, + 1330 + ], + "score": 0.89, + "latex": "\\Psi _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 1107, + 550, + 1144, + 550, + 1144, + 580, + 1107, + 580 + ], + "score": 0.89, + "latex": "\\Psi _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 751, + 202, + 786, + 202, + 786, + 237, + 751, + 237 + ], + "score": 0.88, + "latex": "\\mathbf { S } _ { \\ell } ^ { \\gamma }" + }, + { + "category_id": 13, + "poly": [ + 629, + 1977, + 666, + 1977, + 666, + 2006, + 629, + 2006 + ], + "score": 0.88, + "latex": "\\Psi _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 933, + 1871, + 967, + 1871, + 967, + 1900, + 933, + 1900 + ], + "score": 0.88, + "latex": "\\Psi _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 559, + 1976, + 595, + 1976, + 595, + 2006, + 559, + 2006 + ], + "score": 0.87, + "latex": "\\mathbf { \\Delta } \\Lambda _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 965, + 1686, + 997, + 1686, + 997, + 1716, + 965, + 1716 + ], + "score": 0.87, + "latex": "\\mathbf { S } _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 1308, + 1817, + 1344, + 1817, + 1344, + 1846, + 1308, + 1846 + ], + "score": 0.86, + "latex": "\\Psi _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 608, + 1817, + 645, + 1817, + 645, + 1846, + 608, + 1846 + ], + "score": 0.85, + "latex": "\\mathbf { V } _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 551, + 271, + 591, + 271, + 591, + 298, + 551, + 298 + ], + "score": 0.85, + "latex": "\\mathcal { D } ^ { \\gamma }" + }, + { + "category_id": 13, + "poly": [ + 933, + 1785, + 965, + 1785, + 965, + 1816, + 933, + 1816 + ], + "score": 0.84, + "latex": "\\mathbf { S } _ { \\ell }" + }, + { + "category_id": 14, + "poly": [ + 589, + 1065, + 1109, + 1065, + 1109, + 1155, + 589, + 1155 + ], + "score": 0.78, + "latex": "\\begin{array} { r l } & { \\mathbf { D } ^ { \\mathrm { m e t a g e n o m i c s } } \\sim \\mathcal { N } _ { K S } \\left( \\Psi _ { \\mathrm { p e o p l e } } , \\Psi _ { \\mathrm { s p e c i e s } } \\right) } \\\\ & { \\mathbf { D } ^ { \\mathrm { m e t a b o l o m i c s } } \\sim \\mathcal { N } _ { K S } \\left( \\Psi _ { \\mathrm { p e o p l e } } , \\Psi _ { \\mathrm { m e t a b o l i t e s } } \\right) } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 578, + 329, + 713, + 329, + 713, + 364, + 578, + 364 + ], + "score": 0.76, + "latex": "[ \\mathbf { M } ] = \\mathbf { M } ^ { T }" + }, + { + "category_id": 13, + "poly": [ + 1239, + 234, + 1292, + 234, + 1292, + 267, + 1239, + 267 + ], + "score": 0.76, + "latex": "[ \\mathcal { D } ^ { \\gamma } ]" + }, + { + "category_id": 13, + "poly": [ + 719, + 271, + 734, + 271, + 734, + 298, + 719, + 298 + ], + "score": 0.75, + "latex": "\\ell" + }, + { + "category_id": 13, + "poly": [ + 1069, + 521, + 1084, + 521, + 1084, + 546, + 1069, + 546 + ], + "score": 0.74, + "latex": "\\ell" + }, + { + "category_id": 13, + "poly": [ + 556, + 302, + 571, + 302, + 571, + 328, + 556, + 328 + ], + "score": 0.69, + "latex": "\\ell" + }, + { + "category_id": 13, + "poly": [ + 398, + 330, + 713, + 330, + 713, + 363, + 398, + 363 + ], + "score": 0.59, + "latex": "\\mathbf { M } , \\mathrm { m a t } _ { \\mathrm { c o l u m n s } } \\left[ \\mathbf { M } \\right] = \\mathbf { M } ^ { T }" + }, + { + "category_id": 13, + "poly": [ + 1174, + 233, + 1292, + 233, + 1292, + 267, + 1174, + 267 + ], + "score": 0.31, + "latex": "\\mathrm { m a t } _ { \\ell } \\left[ \\mathcal { D } ^ { \\gamma } \\right]" + }, + { + "category_id": 15, + "poly": [ + 236.0, + 1179.0, + 536.0, + 1179.0, + 536.0, + 1220.0, + 236.0, + 1220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 253.0, + 429.0, + 493.0, + 429.0, + 493.0, + 468.0, + 253.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 863.0, + 2058.0, + 863.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 208.0, + 279.0, + 208.0, + 279.0, + 235.0, + 248.0, + 235.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 202.0, + 750.0, + 202.0, + 750.0, + 239.0, + 292.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 202.0, + 1405.0, + 202.0, + 1405.0, + 239.0, + 787.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 238.0, + 277.0, + 238.0, + 277.0, + 265.0, + 248.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 232.0, + 1173.0, + 232.0, + 1173.0, + 269.0, + 293.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1293.0, + 232.0, + 1406.0, + 232.0, + 1406.0, + 269.0, + 1293.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 248.0, + 275.0, + 277.0, + 275.0, + 277.0, + 302.0, + 248.0, + 302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 265.0, + 550.0, + 265.0, + 550.0, + 305.0, + 290.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 265.0, + 718.0, + 265.0, + 718.0, + 305.0, + 592.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 735.0, + 265.0, + 801.0, + 265.0, + 801.0, + 305.0, + 735.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1130.0, + 265.0, + 1408.0, + 265.0, + 1408.0, + 305.0, + 1130.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 247.0, + 306.0, + 279.0, + 306.0, + 279.0, + 333.0, + 247.0, + 333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 299.0, + 555.0, + 299.0, + 555.0, + 336.0, + 292.0, + 336.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 572.0, + 299.0, + 1406.0, + 299.0, + 1406.0, + 336.0, + 572.0, + 336.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 334.0, + 279.0, + 334.0, + 279.0, + 363.0, + 239.0, + 363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 328.0, + 397.0, + 328.0, + 397.0, + 365.0, + 289.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 714.0, + 328.0, + 867.0, + 328.0, + 867.0, + 365.0, + 714.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 903.0, + 328.0, + 1406.0, + 328.0, + 1406.0, + 365.0, + 903.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 365.0, + 277.0, + 365.0, + 277.0, + 394.0, + 241.0, + 394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 357.0, + 297.0, + 357.0, + 297.0, + 403.0, + 290.0, + 403.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 489.0, + 357.0, + 1271.0, + 357.0, + 1271.0, + 403.0, + 489.0, + 403.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 492.0, + 279.0, + 492.0, + 279.0, + 522.0, + 239.0, + 522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 487.0, + 1405.0, + 487.0, + 1405.0, + 527.0, + 293.0, + 527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 524.0, + 279.0, + 524.0, + 279.0, + 550.0, + 241.0, + 550.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 518.0, + 1068.0, + 518.0, + 1068.0, + 553.0, + 295.0, + 553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1085.0, + 518.0, + 1405.0, + 518.0, + 1405.0, + 553.0, + 1085.0, + 553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 552.0, + 279.0, + 552.0, + 279.0, + 583.0, + 239.0, + 583.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 548.0, + 1106.0, + 548.0, + 1106.0, + 583.0, + 295.0, + 583.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1145.0, + 548.0, + 1405.0, + 548.0, + 1405.0, + 583.0, + 1145.0, + 583.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 585.0, + 279.0, + 585.0, + 279.0, + 611.0, + 241.0, + 611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 578.0, + 539.0, + 578.0, + 539.0, + 613.0, + 293.0, + 613.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1244.0, + 280.0, + 1244.0, + 280.0, + 1271.0, + 241.0, + 1271.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1239.0, + 1405.0, + 1239.0, + 1405.0, + 1273.0, + 292.0, + 1273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1277.0, + 279.0, + 1277.0, + 279.0, + 1300.0, + 241.0, + 1300.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1267.0, + 422.0, + 1267.0, + 422.0, + 1306.0, + 289.0, + 1306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 462.0, + 1267.0, + 1406.0, + 1267.0, + 1406.0, + 1306.0, + 462.0, + 1306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1305.0, + 280.0, + 1305.0, + 280.0, + 1332.0, + 241.0, + 1332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1299.0, + 327.0, + 1299.0, + 327.0, + 1337.0, + 291.0, + 1337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 1299.0, + 1347.0, + 1299.0, + 1347.0, + 1337.0, + 367.0, + 1337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 790.0, + 277.0, + 790.0, + 277.0, + 816.0, + 241.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 785.0, + 1403.0, + 785.0, + 1403.0, + 818.0, + 294.0, + 818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 821.0, + 278.0, + 821.0, + 278.0, + 846.0, + 241.0, + 846.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 816.0, + 1405.0, + 816.0, + 1405.0, + 849.0, + 295.0, + 849.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 853.0, + 277.0, + 853.0, + 277.0, + 878.0, + 240.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 849.0, + 1405.0, + 849.0, + 1405.0, + 878.0, + 296.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 882.0, + 278.0, + 882.0, + 278.0, + 908.0, + 240.0, + 908.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 877.0, + 1405.0, + 877.0, + 1405.0, + 910.0, + 296.0, + 910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 913.0, + 275.0, + 913.0, + 275.0, + 938.0, + 240.0, + 938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 905.0, + 1409.0, + 905.0, + 1409.0, + 941.0, + 291.0, + 941.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 242.0, + 1876.0, + 278.0, + 1876.0, + 278.0, + 1899.0, + 242.0, + 1899.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1869.0, + 932.0, + 1869.0, + 932.0, + 1904.0, + 292.0, + 1904.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 968.0, + 1869.0, + 1403.0, + 1869.0, + 1403.0, + 1904.0, + 968.0, + 1904.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 242.0, + 1907.0, + 278.0, + 1907.0, + 278.0, + 1929.0, + 242.0, + 1929.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1900.0, + 1407.0, + 1900.0, + 1407.0, + 1934.0, + 292.0, + 1934.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 242.0, + 1936.0, + 278.0, + 1936.0, + 278.0, + 1960.0, + 242.0, + 1960.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1932.0, + 1168.0, + 1932.0, + 1168.0, + 1962.0, + 293.0, + 1962.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 242.0, + 1692.0, + 277.0, + 1692.0, + 277.0, + 1715.0, + 242.0, + 1715.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1686.0, + 964.0, + 1686.0, + 964.0, + 1716.0, + 293.0, + 1716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 998.0, + 1686.0, + 1404.0, + 1686.0, + 1404.0, + 1716.0, + 998.0, + 1716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 242.0, + 1723.0, + 277.0, + 1723.0, + 277.0, + 1746.0, + 242.0, + 1746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1714.0, + 1409.0, + 1714.0, + 1409.0, + 1750.0, + 291.0, + 1750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1750.0, + 277.0, + 1750.0, + 277.0, + 1777.0, + 241.0, + 1777.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1746.0, + 815.0, + 1746.0, + 815.0, + 1779.0, + 295.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1341.0, + 920.0, + 1341.0, + 920.0, + 1385.0, + 237.0, + 1385.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1973.0, + 558.0, + 1973.0, + 558.0, + 2013.0, + 240.0, + 2013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 596.0, + 1973.0, + 628.0, + 1973.0, + 628.0, + 2013.0, + 596.0, + 2013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 667.0, + 1973.0, + 1043.0, + 1973.0, + 1043.0, + 2013.0, + 667.0, + 2013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 232.0, + 939.0, + 1413.0, + 939.0, + 1413.0, + 993.0, + 232.0, + 993.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 236.0, + 981.0, + 413.0, + 981.0, + 413.0, + 1019.0, + 236.0, + 1019.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 252.0, + 1789.0, + 280.0, + 1789.0, + 280.0, + 1816.0, + 252.0, + 1816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 1778.0, + 489.0, + 1778.0, + 489.0, + 1818.0, + 288.0, + 1818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 595.0, + 1778.0, + 932.0, + 1778.0, + 932.0, + 1818.0, + 595.0, + 1818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 966.0, + 1778.0, + 1052.0, + 1778.0, + 1052.0, + 1818.0, + 966.0, + 1818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1204.0, + 1778.0, + 1256.0, + 1778.0, + 1256.0, + 1818.0, + 1204.0, + 1818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1400.0, + 1778.0, + 1404.0, + 1778.0, + 1404.0, + 1818.0, + 1400.0, + 1818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 244.0, + 1813.0, + 607.0, + 1813.0, + 607.0, + 1852.0, + 244.0, + 1852.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 646.0, + 1813.0, + 1307.0, + 1813.0, + 1307.0, + 1852.0, + 646.0, + 1852.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1345.0, + 1813.0, + 1355.0, + 1813.0, + 1355.0, + 1852.0, + 1345.0, + 1852.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1125.0, + 1588.0, + 1405.0, + 1588.0, + 1405.0, + 1629.0, + 1125.0, + 1629.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1218.0, + 1482.0, + 1406.0, + 1482.0, + 1406.0, + 1523.0, + 1218.0, + 1523.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 3, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 3, + "poly": [ + 348, + 185, + 1315, + 185, + 1315, + 1326, + 348, + 1326 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 285, + 1503, + 1405, + 1503, + 1405, + 1691, + 285, + 1691 + ], + "score": 0.957 + }, + { + "category_id": 1, + "poly": [ + 242, + 1853, + 1404, + 1853, + 1404, + 2009, + 242, + 2009 + ], + "score": 0.951 + }, + { + "category_id": 4, + "poly": [ + 296, + 1341, + 1406, + 1341, + 1406, + 1437, + 296, + 1437 + ], + "score": 0.951 + }, + { + "category_id": 1, + "poly": [ + 243, + 1714, + 1405, + 1714, + 1405, + 1842, + 243, + 1842 + ], + "score": 0.945 + }, + { + "category_id": 2, + "poly": [ + 840, + 2061, + 859, + 2061, + 859, + 2085, + 840, + 2085 + ], + "score": 0.706 + }, + { + "category_id": 13, + "poly": [ + 587, + 1601, + 998, + 1601, + 998, + 1658, + 587, + 1658 + ], + "score": 0.94, + "latex": "\\begin{array} { r } { \\pmb { \\Lambda } _ { \\ell } ^ { t + 1 } = \\pmb { \\Lambda } _ { \\ell } ^ { t } - \\mu _ { t } \\left[ \\mathbf { e } _ { \\ell } - \\sum _ { \\gamma | \\ell \\in \\gamma } \\mathbf { G } _ { \\ell } ^ { \\gamma } \\right] } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 977, + 1506, + 1092, + 1506, + 1092, + 1547, + 977, + 1547 + ], + "score": 0.93, + "latex": "\\oplus _ { \\ell \\in \\gamma } \\mathbf { G } _ { \\ell } ^ { \\gamma }" + }, + { + "category_id": 13, + "poly": [ + 498, + 1543, + 668, + 1543, + 668, + 1605, + 498, + 1605 + ], + "score": 0.93, + "latex": "\\left( \\oplus _ { \\ell \\in \\gamma } \\mathbf { \\Lambda } \\mathbf { \\Lambda } \\mathbf { \\Lambda } \\mathbf { \\Lambda } _ { \\ell } ^ { t } \\right) ^ { - 1 }" + }, + { + "category_id": 13, + "poly": [ + 491, + 1506, + 560, + 1506, + 560, + 1543, + 491, + 1543 + ], + "score": 0.92, + "latex": "\\{ \\mathbf { G } _ { \\ell } ^ { \\gamma } \\}" + }, + { + "category_id": 13, + "poly": [ + 564, + 1717, + 606, + 1717, + 606, + 1752, + 564, + 1752 + ], + "score": 0.91, + "latex": "\\mathbf { G } _ { \\ell } ^ { \\gamma }" + }, + { + "category_id": 14, + "poly": [ + 695, + 1248, + 981, + 1248, + 981, + 1293, + 695, + 1293 + ], + "score": 0.91, + "latex": "\\begin{array} { r } { \\mathbf { A } _ { \\ell } ^ { t + 1 } = \\mathbf { A } _ { \\ell } ^ { t } - \\mu _ { t } \\left[ \\mathbf { e } _ { \\ell } - \\sum _ { \\gamma | \\ell \\in \\gamma } \\mathbf { G } _ { \\ell } ^ { \\gamma } \\right] } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 746, + 1778, + 787, + 1778, + 787, + 1812, + 746, + 1812 + ], + "score": 0.91, + "latex": "\\mathbf { G } _ { \\ell } ^ { \\gamma }" + }, + { + "category_id": 14, + "poly": [ + 691, + 1122, + 985, + 1122, + 985, + 1179, + 691, + 1179 + ], + "score": 0.91, + "latex": "\\begin{array} { r } { \\bigoplus _ { \\ell \\in \\gamma } \\mathbf { G } _ { \\ell } ^ { \\gamma } = \\mathrm { p r o j } _ { \\kappa } \\left[ \\left( \\bigoplus _ { \\ell \\in \\gamma } \\mathbf { A } _ { \\ell } ^ { t } \\right) ^ { - 1 } \\right] } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 531, + 1655, + 566, + 1655, + 566, + 1690, + 531, + 1690 + ], + "score": 0.88, + "latex": "\\Lambda _ { \\ell } ^ { t }" + }, + { + "category_id": 13, + "poly": [ + 939, + 1856, + 980, + 1856, + 980, + 1886, + 939, + 1886 + ], + "score": 0.88, + "latex": "p \\%" + }, + { + "category_id": 13, + "poly": [ + 547, + 1886, + 597, + 1886, + 597, + 1917, + 547, + 1917 + ], + "score": 0.84, + "latex": "p , k )" + }, + { + "category_id": 13, + "poly": [ + 1013, + 1085, + 1042, + 1085, + 1042, + 1112, + 1013, + 1112 + ], + "score": 0.84, + "latex": "\\mathbf { G } _ { \\ell } ^ { \\gamma }" + }, + { + "category_id": 13, + "poly": [ + 767, + 1090, + 785, + 1090, + 785, + 1111, + 767, + 1111 + ], + "score": 0.82, + "latex": "\\gamma ," + }, + { + "category_id": 13, + "poly": [ + 1201, + 1350, + 1223, + 1350, + 1223, + 1376, + 1201, + 1376 + ], + "score": 0.81, + "latex": "\\gamma" + }, + { + "category_id": 13, + "poly": [ + 1311, + 1619, + 1339, + 1619, + 1339, + 1645, + 1311, + 1645 + ], + "score": 0.81, + "latex": "\\mu _ { t }" + }, + { + "category_id": 13, + "poly": [ + 1035, + 1203, + 1062, + 1203, + 1062, + 1226, + 1035, + 1226 + ], + "score": 0.81, + "latex": "\\mathbf { \\Delta } \\Lambda _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 1049, + 1613, + 1088, + 1613, + 1088, + 1643, + 1049, + 1643 + ], + "score": 0.8, + "latex": "\\Psi _ { \\ell }" + }, + { + "category_id": 13, + "poly": [ + 967, + 1567, + 995, + 1567, + 995, + 1593, + 967, + 1593 + ], + "score": 0.8, + "latex": "\\mu _ { t }" + }, + { + "category_id": 13, + "poly": [ + 572, + 1376, + 588, + 1376, + 588, + 1402, + 572, + 1402 + ], + "score": 0.76, + "latex": "\\ell" + }, + { + "category_id": 13, + "poly": [ + 1268, + 1856, + 1287, + 1856, + 1287, + 1882, + 1268, + 1882 + ], + "score": 0.75, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 572, + 1753, + 644, + 1753, + 644, + 1782, + 572, + 1782 + ], + "score": 0.65, + "latex": "\\mathrm { p r o j } _ { \\kappa }" + }, + { + "category_id": 13, + "poly": [ + 624, + 1754, + 644, + 1754, + 644, + 1781, + 624, + 1781 + ], + "score": 0.33, + "latex": "\\kappa" + }, + { + "category_id": 15, + "poly": [ + 569.0, + 288.0, + 601.0, + 288.0, + 601.0, + 311.0, + 569.0, + 311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1050.0, + 289.0, + 1071.0, + 289.0, + 1071.0, + 308.0, + 1050.0, + 308.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1084.0, + 405.0, + 1096.0, + 405.0, + 1096.0, + 417.0, + 1084.0, + 417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 428.0, + 413.0, + 558.0, + 413.0, + 558.0, + 463.0, + 428.0, + 463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 565.0, + 443.0, + 580.0, + 443.0, + 580.0, + 459.0, + 565.0, + 459.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1102.0, + 421.0, + 1194.0, + 421.0, + 1194.0, + 454.0, + 1102.0, + 454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 560.0, + 569.0, + 588.0, + 569.0, + 588.0, + 596.0, + 560.0, + 596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1061.0, + 567.0, + 1094.0, + 567.0, + 1094.0, + 596.0, + 1061.0, + 596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 790.0, + 666.0, + 885.0, + 666.0, + 885.0, + 696.0, + 790.0, + 696.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 426.0, + 693.0, + 579.0, + 693.0, + 579.0, + 731.0, + 426.0, + 731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1092.0, + 700.0, + 1243.0, + 700.0, + 1243.0, + 736.0, + 1092.0, + 736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 575.0, + 787.0, + 601.0, + 787.0, + 601.0, + 807.0, + 575.0, + 807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 679.0, + 1037.0, + 997.0, + 1037.0, + 997.0, + 1067.0, + 679.0, + 1067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 626.0, + 1080.0, + 766.0, + 1080.0, + 766.0, + 1116.0, + 626.0, + 1116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 786.0, + 1080.0, + 1012.0, + 1080.0, + 1012.0, + 1116.0, + 786.0, + 1116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1043.0, + 1080.0, + 1048.0, + 1080.0, + 1048.0, + 1116.0, + 1043.0, + 1116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 684.0, + 1123.0, + 690.0, + 1123.0, + 690.0, + 1178.0, + 684.0, + 1178.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 608.0, + 1196.0, + 1034.0, + 1196.0, + 1034.0, + 1231.0, + 608.0, + 1231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1063.0, + 1196.0, + 1066.0, + 1196.0, + 1066.0, + 1231.0, + 1063.0, + 1231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 584.0, + 1243.0, + 599.0, + 1243.0, + 599.0, + 1258.0, + 584.0, + 1258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 689.0, + 1244.0, + 694.0, + 1244.0, + 694.0, + 1298.0, + 689.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1065.75, + 285.0, + 1107.75, + 285.0, + 1107.75, + 299.5, + 1065.75, + 299.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1033.0, + 796.5, + 1090.0, + 796.5, + 1090.0, + 814.0, + 1033.0, + 814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 548.0, + 821.5, + 617.0, + 821.5, + 617.0, + 834.5, + 548.0, + 834.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 819.0, + 1217.0, + 819.0, + 1217.0, + 840.5, + 1193.0, + 840.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1095.0, + 895.0, + 1114.0, + 895.0, + 1114.0, + 904.0, + 1095.0, + 904.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1341.0, + 1200.0, + 1341.0, + 1200.0, + 1379.0, + 294.0, + 1379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1224.0, + 1341.0, + 1406.0, + 1341.0, + 1406.0, + 1379.0, + 1224.0, + 1379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1371.0, + 571.0, + 1371.0, + 571.0, + 1412.0, + 293.0, + 1412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 589.0, + 1371.0, + 1406.0, + 1371.0, + 1406.0, + 1412.0, + 589.0, + 1412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1405.0, + 402.0, + 1405.0, + 402.0, + 1437.0, + 294.0, + 1437.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2059.0, + 861.0, + 2059.0, + 861.0, + 2091.0, + 839.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1501.0, + 490.0, + 1501.0, + 490.0, + 1548.0, + 293.0, + 1548.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 561.0, + 1501.0, + 976.0, + 1501.0, + 976.0, + 1548.0, + 561.0, + 1548.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1093.0, + 1501.0, + 1408.0, + 1501.0, + 1408.0, + 1548.0, + 1093.0, + 1548.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1547.0, + 497.0, + 1547.0, + 497.0, + 1607.0, + 291.0, + 1607.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 669.0, + 1547.0, + 966.0, + 1547.0, + 966.0, + 1607.0, + 669.0, + 1607.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 996.0, + 1547.0, + 1412.0, + 1547.0, + 1412.0, + 1607.0, + 996.0, + 1607.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1603.0, + 586.0, + 1603.0, + 586.0, + 1656.0, + 292.0, + 1656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 999.0, + 1603.0, + 1048.0, + 1603.0, + 1048.0, + 1656.0, + 999.0, + 1656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1089.0, + 1603.0, + 1310.0, + 1603.0, + 1310.0, + 1656.0, + 1089.0, + 1656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1340.0, + 1603.0, + 1409.0, + 1603.0, + 1409.0, + 1656.0, + 1340.0, + 1656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1652.0, + 530.0, + 1652.0, + 530.0, + 1692.0, + 296.0, + 1692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 567.0, + 1652.0, + 850.0, + 1652.0, + 850.0, + 1692.0, + 567.0, + 1692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1860.0, + 277.0, + 1860.0, + 277.0, + 1886.0, + 241.0, + 1886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1855.0, + 938.0, + 1855.0, + 938.0, + 1888.0, + 295.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 981.0, + 1855.0, + 1267.0, + 1855.0, + 1267.0, + 1888.0, + 981.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1288.0, + 1855.0, + 1405.0, + 1855.0, + 1405.0, + 1888.0, + 1288.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1891.0, + 277.0, + 1891.0, + 277.0, + 1916.0, + 241.0, + 1916.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1884.0, + 546.0, + 1884.0, + 546.0, + 1918.0, + 293.0, + 1918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 598.0, + 1884.0, + 1404.0, + 1884.0, + 1404.0, + 1918.0, + 598.0, + 1918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1920.0, + 277.0, + 1920.0, + 277.0, + 1946.0, + 241.0, + 1946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1914.0, + 1407.0, + 1914.0, + 1407.0, + 1948.0, + 292.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1950.0, + 279.0, + 1950.0, + 279.0, + 1979.0, + 239.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1944.0, + 1405.0, + 1944.0, + 1405.0, + 1982.0, + 295.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1982.0, + 277.0, + 1982.0, + 277.0, + 2007.0, + 241.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1976.0, + 698.0, + 1976.0, + 698.0, + 2010.0, + 293.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1722.0, + 277.0, + 1722.0, + 277.0, + 1750.0, + 240.0, + 1750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1715.0, + 563.0, + 1715.0, + 563.0, + 1751.0, + 294.0, + 1751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 607.0, + 1715.0, + 1404.0, + 1715.0, + 1404.0, + 1751.0, + 607.0, + 1751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1754.0, + 277.0, + 1754.0, + 277.0, + 1781.0, + 240.0, + 1781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1747.0, + 571.0, + 1747.0, + 571.0, + 1782.0, + 293.0, + 1782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 1747.0, + 1405.0, + 1747.0, + 1405.0, + 1782.0, + 645.0, + 1782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1784.0, + 277.0, + 1784.0, + 277.0, + 1811.0, + 240.0, + 1811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1778.0, + 745.0, + 1778.0, + 745.0, + 1814.0, + 294.0, + 1814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 788.0, + 1778.0, + 1405.0, + 1778.0, + 1405.0, + 1814.0, + 788.0, + 1814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1814.0, + 277.0, + 1814.0, + 277.0, + 1841.0, + 240.0, + 1841.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1804.0, + 1058.0, + 1804.0, + 1058.0, + 1848.0, + 293.0, + 1848.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 4, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 242, + 1609, + 1403, + 1609, + 1403, + 1823, + 242, + 1823 + ], + "score": 0.971 + }, + { + "category_id": 1, + "poly": [ + 243, + 1915, + 1402, + 1915, + 1402, + 2008, + 243, + 2008 + ], + "score": 0.968 + }, + { + "category_id": 1, + "poly": [ + 292, + 1382, + 1403, + 1382, + 1403, + 1596, + 292, + 1596 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 297, + 244, + 1104, + 244, + 1104, + 1106, + 297, + 1106 + ], + "score": 0.955 + }, + { + "category_id": 1, + "poly": [ + 293, + 1228, + 1402, + 1228, + 1402, + 1292, + 293, + 1292 + ], + "score": 0.947 + }, + { + "category_id": 0, + "poly": [ + 289, + 1159, + 454, + 1159, + 454, + 1196, + 289, + 1196 + ], + "score": 0.892 + }, + { + "category_id": 0, + "poly": [ + 293, + 1325, + 538, + 1325, + 538, + 1358, + 293, + 1358 + ], + "score": 0.892 + }, + { + "category_id": 0, + "poly": [ + 296, + 1859, + 480, + 1859, + 480, + 1890, + 296, + 1890 + ], + "score": 0.881 + }, + { + "category_id": 0, + "poly": [ + 299, + 202, + 571, + 202, + 571, + 234, + 299, + 234 + ], + "score": 0.796 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 859, + 2062, + 859, + 2085, + 840, + 2085 + ], + "score": 0.778 + }, + { + "category_id": 13, + "poly": [ + 451, + 853, + 635, + 853, + 635, + 893, + 451, + 893 + ], + "score": 0.86, + "latex": "\\sum _ { \\ell \\in \\gamma } \\operatorname* { m i n } \\pmb { \\Lambda } _ { \\ell } <" + }, + { + "category_id": 13, + "poly": [ + 387, + 242, + 453, + 242, + 453, + 275, + 387, + 275 + ], + "score": 0.85, + "latex": "\\{ \\mathcal { D } _ { i } ^ { \\gamma } \\}" + }, + { + "category_id": 14, + "poly": [ + 423, + 574, + 813, + 574, + 813, + 710, + 423, + 710 + ], + "score": 0.81, + "latex": "\\begin{array} { r l } & { \\mathbf { \\mu } _ { \\mathbf { G } _ { \\ell } ^ { \\gamma } } ^ { \\mathbf { 1 } } \\mathrm { p r o j } _ { K S } [ ( \\bigoplus _ { \\ell ^ { \\prime } \\in \\gamma } \\mathbf { \\Lambda } _ { \\Lambda _ { \\ell } } ) ^ { - 1 } ] } \\\\ & { \\mathbf { \\Lambda } _ { \\mathbf { A } _ { \\ell } ^ { \\prime } } ^ { \\mathbf { 1 } } \\mathbf { \\Lambda } _ { \\mathbf { A } _ { \\ell } } - \\mu [ \\mathbf { e } _ { \\ell } - \\sum _ { \\gamma | \\ell \\in \\gamma } \\mathbf { G } _ { \\ell _ { \\ell } } ^ { \\gamma } ] } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 571, + 895, + 592, + 895, + 592, + 920, + 571, + 920 + ], + "score": 0.79, + "latex": "\\mu" + }, + { + "category_id": 13, + "poly": [ + 407, + 274, + 473, + 274, + 473, + 304, + 407, + 304 + ], + "score": 0.66, + "latex": "\\{ \\Psi _ { \\ell } \\}" + }, + { + "category_id": 13, + "poly": [ + 433, + 830, + 455, + 830, + 455, + 856, + 433, + 856 + ], + "score": 0.61, + "latex": "\\gamma" + }, + { + "category_id": 13, + "poly": [ + 386, + 1038, + 578, + 1038, + 578, + 1074, + 386, + 1074 + ], + "score": 0.5, + "latex": "{ \\Psi } _ { \\ell } \\gets { \\bf V } _ { \\ell } { \\bf A } _ { \\ell } { \\bf V } _ { \\ell } ^ { T }" + }, + { + "category_id": 13, + "poly": [ + 618, + 1976, + 663, + 1976, + 663, + 2004, + 618, + 2004 + ], + "score": 0.42, + "latex": "1 0 \\mathrm { x }" + }, + { + "category_id": 13, + "poly": [ + 344, + 497, + 423, + 497, + 423, + 527, + 344, + 527 + ], + "score": 0.42, + "latex": "\\mu 1" + }, + { + "category_id": 13, + "poly": [ + 328, + 454, + 553, + 454, + 553, + 529, + 328, + 529 + ], + "score": 0.32, + "latex": "\\vdots \\stackrel { \\wedge } { \\underset { \\mu 1 } { \\longrightarrow } } [ 1 \\quad \\cdots \\quad 1 ] ^ { T }" + }, + { + "category_id": 13, + "poly": [ + 433, + 555, + 564, + 555, + 564, + 586, + 433, + 586 + ], + "score": 0.27, + "latex": "1 \\leq \\ell \\leq \\bar { K }" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1153.0, + 461.0, + 1153.0, + 461.0, + 1205.0, + 289.0, + 1205.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1322.0, + 543.0, + 1322.0, + 543.0, + 1364.0, + 293.0, + 1364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1855.0, + 486.0, + 1855.0, + 486.0, + 1895.0, + 293.0, + 1895.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 198.0, + 573.0, + 198.0, + 573.0, + 238.0, + 295.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2060.0, + 862.0, + 2060.0, + 862.0, + 2091.0, + 840.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1615.0, + 277.0, + 1615.0, + 277.0, + 1642.0, + 241.0, + 1642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1607.0, + 1406.0, + 1607.0, + 1406.0, + 1646.0, + 293.0, + 1646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1645.0, + 279.0, + 1645.0, + 279.0, + 1675.0, + 239.0, + 1675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1639.0, + 1406.0, + 1639.0, + 1406.0, + 1678.0, + 293.0, + 1678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1674.0, + 279.0, + 1674.0, + 279.0, + 1705.0, + 239.0, + 1705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1667.0, + 1404.0, + 1667.0, + 1404.0, + 1706.0, + 293.0, + 1706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1706.0, + 279.0, + 1706.0, + 279.0, + 1733.0, + 241.0, + 1733.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1699.0, + 1406.0, + 1699.0, + 1406.0, + 1738.0, + 293.0, + 1738.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1736.0, + 279.0, + 1736.0, + 279.0, + 1766.0, + 239.0, + 1766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1731.0, + 1404.0, + 1731.0, + 1404.0, + 1766.0, + 294.0, + 1766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1765.0, + 279.0, + 1765.0, + 279.0, + 1796.0, + 239.0, + 1796.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1761.0, + 1404.0, + 1761.0, + 1404.0, + 1796.0, + 296.0, + 1796.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1797.0, + 277.0, + 1797.0, + 277.0, + 1824.0, + 241.0, + 1824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1790.0, + 1320.0, + 1790.0, + 1320.0, + 1827.0, + 294.0, + 1827.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 242.0, + 1919.0, + 278.0, + 1919.0, + 278.0, + 1946.0, + 242.0, + 1946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1916.0, + 1403.0, + 1916.0, + 1403.0, + 1946.0, + 294.0, + 1946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1951.0, + 277.0, + 1951.0, + 277.0, + 1977.0, + 240.0, + 1977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1942.0, + 1405.0, + 1942.0, + 1405.0, + 1980.0, + 291.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 242.0, + 1981.0, + 280.0, + 1981.0, + 280.0, + 2007.0, + 242.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1976.0, + 617.0, + 1976.0, + 617.0, + 2010.0, + 294.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 664.0, + 1976.0, + 923.0, + 1976.0, + 923.0, + 2010.0, + 664.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1384.0, + 1403.0, + 1384.0, + 1403.0, + 1414.0, + 296.0, + 1414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1413.0, + 1405.0, + 1413.0, + 1405.0, + 1446.0, + 294.0, + 1446.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1444.0, + 1405.0, + 1444.0, + 1405.0, + 1477.0, + 292.0, + 1477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1473.0, + 1405.0, + 1473.0, + 1405.0, + 1506.0, + 294.0, + 1506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1504.0, + 1405.0, + 1504.0, + 1405.0, + 1538.0, + 295.0, + 1538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1534.0, + 1406.0, + 1534.0, + 1406.0, + 1569.0, + 295.0, + 1569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1566.0, + 1228.0, + 1566.0, + 1228.0, + 1601.0, + 295.0, + 1601.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 238.0, + 386.0, + 238.0, + 386.0, + 276.0, + 295.0, + 276.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 454.0, + 238.0, + 581.0, + 238.0, + 581.0, + 276.0, + 454.0, + 276.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 269.0, + 406.0, + 269.0, + 406.0, + 308.0, + 296.0, + 308.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 474.0, + 269.0, + 479.0, + 269.0, + 479.0, + 308.0, + 474.0, + 308.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 299.0, + 526.0, + 299.0, + 526.0, + 335.0, + 304.0, + 335.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 337.0, + 340.0, + 337.0, + 340.0, + 371.0, + 308.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 324.0, + 906.0, + 324.0, + 906.0, + 386.0, + 369.0, + 386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 372.0, + 339.0, + 372.0, + 339.0, + 406.0, + 307.0, + 406.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 382.0, + 370.0, + 660.0, + 370.0, + 660.0, + 408.0, + 382.0, + 408.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 406.0, + 339.0, + 406.0, + 339.0, + 437.0, + 308.0, + 437.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 379.0, + 398.0, + 643.0, + 398.0, + 643.0, + 443.0, + 379.0, + 443.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 336.0, + 431.0, + 441.0, + 431.0, + 441.0, + 467.0, + 336.0, + 467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 309.0, + 437.0, + 340.0, + 437.0, + 340.0, + 465.0, + 309.0, + 465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 453.0, + 327.0, + 453.0, + 327.0, + 506.0, + 293.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 554.0, + 453.0, + 564.0, + 453.0, + 564.0, + 506.0, + 554.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 494.0, + 327.0, + 494.0, + 327.0, + 527.0, + 304.0, + 527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 522.0, + 585.0, + 522.0, + 585.0, + 560.0, + 304.0, + 560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 556.0, + 341.0, + 556.0, + 341.0, + 589.0, + 306.0, + 589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 380.0, + 552.0, + 432.0, + 552.0, + 432.0, + 590.0, + 380.0, + 590.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 600.0, + 342.0, + 600.0, + 342.0, + 638.0, + 295.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 414.0, + 578.0, + 422.0, + 578.0, + 422.0, + 657.0, + 414.0, + 657.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 662.0, + 340.0, + 662.0, + 340.0, + 697.0, + 296.0, + 697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 700.0, + 340.0, + 700.0, + 340.0, + 735.0, + 295.0, + 735.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 379.0, + 699.0, + 481.0, + 699.0, + 481.0, + 734.0, + 379.0, + 734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 732.0, + 340.0, + 732.0, + 340.0, + 765.0, + 295.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 382.0, + 730.0, + 566.0, + 730.0, + 566.0, + 765.0, + 382.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 761.0, + 340.0, + 761.0, + 340.0, + 795.0, + 296.0, + 795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 422.0, + 759.0, + 547.0, + 759.0, + 547.0, + 802.0, + 422.0, + 802.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 792.0, + 340.0, + 792.0, + 340.0, + 825.0, + 296.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 381.0, + 790.0, + 482.0, + 790.0, + 482.0, + 826.0, + 381.0, + 826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 381.0, + 820.0, + 432.0, + 820.0, + 432.0, + 862.0, + 381.0, + 862.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 456.0, + 820.0, + 464.0, + 820.0, + 464.0, + 862.0, + 456.0, + 862.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 825.0, + 339.0, + 825.0, + 339.0, + 855.0, + 297.0, + 855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 853.0, + 340.0, + 853.0, + 340.0, + 887.0, + 296.0, + 887.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 414.0, + 843.0, + 450.0, + 843.0, + 450.0, + 903.0, + 414.0, + 903.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 636.0, + 843.0, + 823.0, + 843.0, + 823.0, + 903.0, + 636.0, + 903.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 887.0, + 340.0, + 887.0, + 340.0, + 921.0, + 296.0, + 921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 464.0, + 884.0, + 570.0, + 884.0, + 570.0, + 925.0, + 464.0, + 925.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 593.0, + 884.0, + 1099.0, + 884.0, + 1099.0, + 925.0, + 593.0, + 925.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 915.0, + 341.0, + 915.0, + 341.0, + 952.0, + 296.0, + 952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 423.0, + 917.0, + 505.0, + 917.0, + 505.0, + 951.0, + 423.0, + 951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 948.0, + 339.0, + 948.0, + 339.0, + 981.0, + 294.0, + 981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 381.0, + 946.0, + 482.0, + 946.0, + 482.0, + 982.0, + 381.0, + 982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 977.0, + 469.0, + 977.0, + 469.0, + 1010.0, + 296.0, + 1010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1007.0, + 526.0, + 1007.0, + 526.0, + 1042.0, + 297.0, + 1042.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1039.0, + 342.0, + 1039.0, + 342.0, + 1071.0, + 294.0, + 1071.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 378.0, + 1032.0, + 385.0, + 1032.0, + 385.0, + 1077.0, + 378.0, + 1077.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1067.0, + 438.0, + 1067.0, + 438.0, + 1103.0, + 294.0, + 1103.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1224.0, + 1408.0, + 1224.0, + 1408.0, + 1265.0, + 293.0, + 1265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1259.0, + 1245.0, + 1259.0, + 1245.0, + 1293.0, + 295.0, + 1293.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 5, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 3, + "poly": [ + 348, + 874, + 1373, + 874, + 1373, + 1277, + 348, + 1277 + ], + "score": 0.971 + }, + { + "category_id": 3, + "poly": [ + 318, + 244, + 1355, + 244, + 1355, + 747, + 318, + 747 + ], + "score": 0.97 + }, + { + "category_id": 3, + "poly": [ + 393, + 1464, + 1307, + 1464, + 1307, + 1877, + 393, + 1877 + ], + "score": 0.968 + }, + { + "category_id": 4, + "poly": [ + 295, + 1295, + 1407, + 1295, + 1407, + 1420, + 295, + 1420 + ], + "score": 0.96 + }, + { + "category_id": 4, + "poly": [ + 297, + 1894, + 1404, + 1894, + 1404, + 1990, + 297, + 1990 + ], + "score": 0.947 + }, + { + "category_id": 4, + "poly": [ + 292, + 764, + 1405, + 764, + 1405, + 829, + 292, + 829 + ], + "score": 0.938 + }, + { + "category_id": 2, + "poly": [ + 841, + 2060, + 858, + 2060, + 858, + 2084, + 841, + 2084 + ], + "score": 0.731 + }, + { + "category_id": 13, + "poly": [ + 1257, + 1385, + 1276, + 1385, + 1276, + 1424, + 1257, + 1424 + ], + "score": 0.85, + "latex": "\\frac { 1 } { 5 }" + }, + { + "category_id": 13, + "poly": [ + 1191, + 1296, + 1266, + 1296, + 1266, + 1325, + 1191, + 1325 + ], + "score": 0.63, + "latex": "5 0 \\mathrm { x } 5 0" + }, + { + "category_id": 13, + "poly": [ + 663, + 875, + 723, + 875, + 723, + 897, + 663, + 897 + ], + "score": 0.4, + "latex": "5 0 \\times 5 0" + }, + { + "category_id": 13, + "poly": [ + 949, + 1327, + 1023, + 1327, + 1023, + 1356, + 949, + 1356 + ], + "score": 0.31, + "latex": "5 0 \\mathrm { x } 5 0" + }, + { + "category_id": 15, + "poly": [ + 541.0, + 872.0, + 662.0, + 872.0, + 662.0, + 899.0, + 541.0, + 899.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 724.0, + 872.0, + 790.0, + 872.0, + 790.0, + 899.0, + 724.0, + 899.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 905.0, + 395.0, + 905.0, + 395.0, + 927.0, + 365.0, + 927.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 471.0, + 887.0, + 571.0, + 887.0, + 571.0, + 920.0, + 471.0, + 920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 672.0, + 905.0, + 702.0, + 905.0, + 702.0, + 928.0, + 672.0, + 928.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 770.0, + 888.0, + 885.0, + 888.0, + 885.0, + 920.0, + 770.0, + 920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 907.0, + 1083.0, + 907.0, + 1083.0, + 926.0, + 1057.0, + 926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1168.0, + 892.0, + 1275.0, + 892.0, + 1275.0, + 920.0, + 1168.0, + 920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 576.0, + 918.0, + 628.0, + 918.0, + 628.0, + 942.0, + 576.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 883.0, + 918.0, + 935.0, + 918.0, + 935.0, + 942.0, + 883.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 576.0, + 934.0, + 644.0, + 934.0, + 644.0, + 961.0, + 576.0, + 961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 883.0, + 934.0, + 952.0, + 934.0, + 952.0, + 961.0, + 883.0, + 961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 958.0, + 396.0, + 958.0, + 396.0, + 985.0, + 364.0, + 985.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 576.0, + 949.0, + 642.0, + 949.0, + 642.0, + 977.0, + 576.0, + 977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 671.0, + 961.0, + 702.0, + 961.0, + 702.0, + 982.0, + 671.0, + 982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 884.0, + 949.0, + 948.0, + 949.0, + 948.0, + 977.0, + 884.0, + 977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 962.0, + 1082.0, + 962.0, + 1082.0, + 981.0, + 1057.0, + 981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 577.0, + 970.0, + 636.0, + 970.0, + 636.0, + 991.0, + 577.0, + 991.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 884.0, + 970.0, + 943.0, + 970.0, + 943.0, + 991.0, + 884.0, + 991.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 641.0, + 1003.0, + 715.0, + 1003.0, + 715.0, + 1108.0, + 641.0, + 1108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 1014.0, + 375.0, + 1014.0, + 375.0, + 1097.0, + 345.0, + 1097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 1016.0, + 395.0, + 1016.0, + 395.0, + 1039.0, + 364.0, + 1039.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 669.0, + 1018.0, + 699.0, + 1018.0, + 699.0, + 1036.0, + 669.0, + 1036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 1072.0, + 395.0, + 1072.0, + 395.0, + 1095.0, + 364.0, + 1095.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 1073.0, + 1082.0, + 1073.0, + 1082.0, + 1092.0, + 1057.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 362.0, + 1126.0, + 396.0, + 1126.0, + 396.0, + 1153.0, + 362.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 669.0, + 1126.0, + 703.0, + 1126.0, + 703.0, + 1153.0, + 669.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 1130.0, + 1082.0, + 1130.0, + 1082.0, + 1149.0, + 1057.0, + 1149.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1108.0, + 1131.0, + 1150.0, + 1131.0, + 1150.0, + 1152.0, + 1108.0, + 1152.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1108.0, + 1146.0, + 1204.0, + 1146.0, + 1204.0, + 1165.0, + 1108.0, + 1165.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1108.0, + 1158.0, + 1219.0, + 1158.0, + 1219.0, + 1181.0, + 1108.0, + 1181.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1108.0, + 1171.0, + 1158.0, + 1171.0, + 1158.0, + 1192.0, + 1108.0, + 1192.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 429.0, + 1193.0, + 460.0, + 1193.0, + 460.0, + 1216.0, + 429.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 480.0, + 1193.0, + 511.0, + 1193.0, + 511.0, + 1216.0, + 480.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 531.0, + 1193.0, + 562.0, + 1193.0, + 562.0, + 1216.0, + 531.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 583.0, + 1193.0, + 613.0, + 1193.0, + 613.0, + 1216.0, + 583.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 632.0, + 1194.0, + 662.0, + 1194.0, + 662.0, + 1214.0, + 632.0, + 1214.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 735.0, + 1193.0, + 767.0, + 1193.0, + 767.0, + 1216.0, + 735.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 1193.0, + 818.0, + 1193.0, + 818.0, + 1215.0, + 787.0, + 1215.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 837.0, + 1193.0, + 869.0, + 1193.0, + 869.0, + 1216.0, + 837.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 890.0, + 1193.0, + 920.0, + 1193.0, + 920.0, + 1216.0, + 890.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 940.0, + 1194.0, + 971.0, + 1194.0, + 971.0, + 1216.0, + 940.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1123.0, + 1192.0, + 1151.0, + 1192.0, + 1151.0, + 1215.0, + 1123.0, + 1215.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1179.0, + 1192.0, + 1209.0, + 1192.0, + 1209.0, + 1213.0, + 1179.0, + 1213.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1234.0, + 1192.0, + 1264.0, + 1192.0, + 1264.0, + 1213.0, + 1234.0, + 1213.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1293.0, + 1193.0, + 1318.0, + 1193.0, + 1318.0, + 1212.0, + 1293.0, + 1212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1350.0, + 1194.0, + 1374.0, + 1194.0, + 1374.0, + 1211.0, + 1350.0, + 1211.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 491.0, + 1207.0, + 550.0, + 1207.0, + 550.0, + 1231.0, + 491.0, + 1231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 798.0, + 1207.0, + 858.0, + 1207.0, + 858.0, + 1231.0, + 798.0, + 1231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1196.0, + 1202.0, + 1245.0, + 1202.0, + 1245.0, + 1227.0, + 1196.0, + 1227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 643.0, + 1247.0, + 687.0, + 1247.0, + 687.0, + 1285.0, + 643.0, + 1285.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 1246.0, + 1239.0, + 1246.0, + 1239.0, + 1285.0, + 1193.0, + 1285.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1047.0, + 1004.5, + 1065.0, + 1004.5, + 1065.0, + 1094.5, + 1047.0, + 1094.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 926.0, + 1092.0, + 934.0, + 1092.0, + 934.0, + 1104.5, + 926.0, + 1104.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1117.5, + 944.0, + 1117.5, + 944.0, + 1125.5, + 941.0, + 1125.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 678.25, + 1181.5, + 707.25, + 1181.5, + 707.25, + 1216.5, + 678.25, + 1216.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 1182.5, + 401.0, + 1182.5, + 401.0, + 1214.0, + 372.0, + 1214.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1061.0, + 1184.5, + 1088.0, + 1184.5, + 1088.0, + 1211.0, + 1061.0, + 1211.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 239.0, + 744.0, + 239.0, + 744.0, + 282.0, + 333.0, + 282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 938.0, + 239.0, + 1347.0, + 239.0, + 1347.0, + 282.0, + 938.0, + 282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 647.0, + 273.0, + 718.0, + 273.0, + 718.0, + 305.0, + 647.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1252.0, + 273.0, + 1323.0, + 273.0, + 1323.0, + 305.0, + 1252.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 291.0, + 742.0, + 291.0, + 742.0, + 330.0, + 645.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1252.0, + 294.0, + 1345.0, + 294.0, + 1345.0, + 329.0, + 1252.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 332.0, + 319.0, + 367.0, + 319.0, + 367.0, + 347.0, + 332.0, + 347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 647.0, + 317.0, + 735.0, + 317.0, + 735.0, + 351.0, + 647.0, + 351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 938.0, + 319.0, + 973.0, + 319.0, + 973.0, + 347.0, + 938.0, + 347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 382.0, + 364.0, + 382.0, + 364.0, + 530.0, + 313.0, + 530.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 922.0, + 384.0, + 970.0, + 384.0, + 970.0, + 527.0, + 922.0, + 527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 334.0, + 508.0, + 365.0, + 508.0, + 365.0, + 536.0, + 334.0, + 536.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 940.0, + 508.0, + 970.0, + 508.0, + 970.0, + 536.0, + 940.0, + 536.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 573.0, + 366.0, + 573.0, + 366.0, + 600.0, + 333.0, + 600.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 938.0, + 573.0, + 972.0, + 573.0, + 972.0, + 601.0, + 938.0, + 601.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 343.0, + 639.0, + 407.0, + 639.0, + 407.0, + 672.0, + 343.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 424.0, + 649.0, + 463.0, + 649.0, + 463.0, + 673.0, + 424.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 480.0, + 647.0, + 518.0, + 647.0, + 518.0, + 673.0, + 480.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 533.0, + 647.0, + 575.0, + 647.0, + 575.0, + 673.0, + 533.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 584.0, + 646.0, + 634.0, + 646.0, + 634.0, + 675.0, + 584.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 639.0, + 646.0, + 690.0, + 646.0, + 690.0, + 675.0, + 639.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 694.0, + 649.0, + 741.0, + 649.0, + 741.0, + 673.0, + 694.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 948.0, + 640.0, + 967.0, + 640.0, + 967.0, + 660.0, + 948.0, + 660.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 646.0, + 1009.0, + 646.0, + 1009.0, + 675.0, + 979.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1035.0, + 650.0, + 1064.0, + 650.0, + 1064.0, + 673.0, + 1035.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1089.0, + 646.0, + 1120.0, + 646.0, + 1120.0, + 675.0, + 1089.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1142.0, + 646.0, + 1176.0, + 646.0, + 1176.0, + 673.0, + 1142.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 646.0, + 1233.0, + 646.0, + 1233.0, + 673.0, + 1195.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1250.0, + 647.0, + 1288.0, + 647.0, + 1288.0, + 675.0, + 1250.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1304.0, + 649.0, + 1341.0, + 649.0, + 1341.0, + 675.0, + 1304.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 530.0, + 660.0, + 579.0, + 660.0, + 579.0, + 694.0, + 530.0, + 694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1133.0, + 660.0, + 1184.0, + 660.0, + 1184.0, + 696.0, + 1133.0, + 696.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 523.0, + 713.0, + 572.0, + 713.0, + 572.0, + 756.0, + 523.0, + 756.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1128.0, + 713.0, + 1176.0, + 713.0, + 1176.0, + 756.0, + 1128.0, + 756.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 388.0, + 1475.0, + 725.0, + 1475.0, + 725.0, + 1509.0, + 388.0, + 1509.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 992.0, + 1461.0, + 1313.0, + 1461.0, + 1313.0, + 1501.0, + 992.0, + 1501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 568.0, + 1523.0, + 607.0, + 1523.0, + 607.0, + 1534.0, + 568.0, + 1534.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 624.0, + 1524.0, + 637.0, + 1524.0, + 637.0, + 1533.0, + 624.0, + 1533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 677.0, + 1578.0, + 686.0, + 1578.0, + 686.0, + 1588.0, + 677.0, + 1588.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 691.0, + 1580.0, + 696.0, + 1580.0, + 696.0, + 1585.0, + 691.0, + 1585.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 688.0, + 1590.0, + 697.0, + 1590.0, + 697.0, + 1600.0, + 688.0, + 1600.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 611.0, + 1673.0, + 619.0, + 1673.0, + 619.0, + 1694.0, + 611.0, + 1694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 428.0, + 1707.0, + 436.0, + 1707.0, + 436.0, + 1716.0, + 428.0, + 1716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 624.0, + 1709.0, + 632.0, + 1709.0, + 632.0, + 1718.0, + 624.0, + 1718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 635.0, + 1709.0, + 648.0, + 1709.0, + 648.0, + 1718.0, + 635.0, + 1718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 672.0, + 1713.0, + 693.0, + 1713.0, + 693.0, + 1788.0, + 672.0, + 1788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 686.0, + 1716.0, + 700.0, + 1716.0, + 700.0, + 1759.0, + 686.0, + 1759.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 632.0, + 1718.0, + 646.0, + 1718.0, + 646.0, + 1732.0, + 632.0, + 1732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 425.0, + 1719.0, + 437.0, + 1719.0, + 437.0, + 1732.0, + 425.0, + 1732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 622.0, + 1719.0, + 636.0, + 1719.0, + 636.0, + 1732.0, + 622.0, + 1732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 660.0, + 1719.0, + 674.0, + 1719.0, + 674.0, + 1734.0, + 660.0, + 1734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 593.0, + 1720.0, + 607.0, + 1720.0, + 607.0, + 1731.0, + 593.0, + 1731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 648.0, + 1720.0, + 661.0, + 1720.0, + 661.0, + 1732.0, + 648.0, + 1732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 441.0, + 1721.0, + 451.0, + 1721.0, + 451.0, + 1731.0, + 441.0, + 1731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 610.0, + 1721.0, + 619.0, + 1721.0, + 619.0, + 1731.0, + 610.0, + 1731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 421.0, + 1725.0, + 441.0, + 1725.0, + 441.0, + 1798.0, + 421.0, + 1798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 439.0, + 1728.0, + 455.0, + 1728.0, + 455.0, + 1746.0, + 439.0, + 1746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 619.0, + 1728.0, + 637.0, + 1728.0, + 637.0, + 1746.0, + 619.0, + 1746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 631.0, + 1728.0, + 648.0, + 1728.0, + 648.0, + 1745.0, + 631.0, + 1745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 648.0, + 1728.0, + 661.0, + 1728.0, + 661.0, + 1745.0, + 648.0, + 1745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 660.0, + 1728.0, + 676.0, + 1728.0, + 676.0, + 1749.0, + 660.0, + 1749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 566.0, + 1729.0, + 610.0, + 1729.0, + 610.0, + 1746.0, + 566.0, + 1746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 611.0, + 1731.0, + 619.0, + 1731.0, + 619.0, + 1745.0, + 611.0, + 1745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 1741.0, + 452.0, + 1741.0, + 452.0, + 1756.0, + 440.0, + 1756.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 632.0, + 1742.0, + 677.0, + 1742.0, + 677.0, + 1760.0, + 632.0, + 1760.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 685.0, + 1754.0, + 699.0, + 1754.0, + 699.0, + 1786.0, + 685.0, + 1786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 649.0, + 1755.0, + 662.0, + 1755.0, + 662.0, + 1772.0, + 649.0, + 1772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 621.0, + 1756.0, + 635.0, + 1756.0, + 635.0, + 1770.0, + 621.0, + 1770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 661.0, + 1756.0, + 674.0, + 1756.0, + 674.0, + 1772.0, + 661.0, + 1772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 611.0, + 1758.0, + 620.0, + 1758.0, + 620.0, + 1768.0, + 611.0, + 1768.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 637.0, + 1761.0, + 645.0, + 1761.0, + 645.0, + 1770.0, + 637.0, + 1770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 416.0, + 1771.0, + 426.0, + 1771.0, + 426.0, + 1791.0, + 416.0, + 1791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 1772.0, + 462.0, + 1772.0, + 462.0, + 1788.0, + 440.0, + 1788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 464.0, + 1773.0, + 480.0, + 1773.0, + 480.0, + 1785.0, + 464.0, + 1785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 608.0, + 1769.0, + 674.0, + 1769.0, + 674.0, + 1787.0, + 608.0, + 1787.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 622.0, + 1783.0, + 674.0, + 1783.0, + 674.0, + 1797.0, + 622.0, + 1797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 675.0, + 1786.0, + 686.0, + 1786.0, + 686.0, + 1794.0, + 675.0, + 1794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 527.0, + 1847.0, + 567.0, + 1847.0, + 567.0, + 1881.0, + 527.0, + 1881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1132.0, + 1847.0, + 1172.0, + 1847.0, + 1172.0, + 1881.0, + 1132.0, + 1881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 599.0, + 1772.5, + 603.0, + 1772.5, + 603.0, + 1781.5, + 599.0, + 1781.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1297.0, + 1190.0, + 1297.0, + 1190.0, + 1329.0, + 296.0, + 1329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1267.0, + 1297.0, + 1409.0, + 1297.0, + 1409.0, + 1329.0, + 1267.0, + 1329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1326.0, + 948.0, + 1326.0, + 948.0, + 1362.0, + 293.0, + 1362.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1024.0, + 1326.0, + 1406.0, + 1326.0, + 1406.0, + 1362.0, + 1024.0, + 1362.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1356.0, + 1405.0, + 1356.0, + 1405.0, + 1393.0, + 293.0, + 1393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1382.0, + 1256.0, + 1382.0, + 1256.0, + 1428.0, + 293.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1277.0, + 1382.0, + 1289.0, + 1382.0, + 1289.0, + 1428.0, + 1277.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1894.0, + 1407.0, + 1894.0, + 1407.0, + 1929.0, + 295.0, + 1929.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1925.0, + 1404.0, + 1925.0, + 1404.0, + 1959.0, + 294.0, + 1959.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1953.0, + 1385.0, + 1953.0, + 1385.0, + 1995.0, + 292.0, + 1995.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 763.0, + 1407.0, + 763.0, + 1407.0, + 802.0, + 294.0, + 802.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 795.0, + 958.0, + 795.0, + 958.0, + 833.0, + 292.0, + 833.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2059.0, + 860.0, + 2059.0, + 860.0, + 2090.0, + 839.0, + 2090.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 6, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 4, + "poly": [ + 295, + 755, + 1406, + 755, + 1406, + 906, + 295, + 906 + ], + "score": 0.97 + }, + { + "category_id": 3, + "poly": [ + 298, + 196, + 1399, + 196, + 1399, + 736, + 298, + 736 + ], + "score": 0.967 + }, + { + "category_id": 1, + "poly": [ + 242, + 975, + 1404, + 975, + 1404, + 1189, + 242, + 1189 + ], + "score": 0.967 + }, + { + "category_id": 1, + "poly": [ + 241, + 1369, + 1404, + 1369, + 1404, + 1582, + 241, + 1582 + ], + "score": 0.964 + }, + { + "category_id": 1, + "poly": [ + 242, + 1823, + 1403, + 1823, + 1403, + 2008, + 242, + 2008 + ], + "score": 0.964 + }, + { + "category_id": 1, + "poly": [ + 240, + 1595, + 1404, + 1595, + 1404, + 1810, + 240, + 1810 + ], + "score": 0.964 + }, + { + "category_id": 1, + "poly": [ + 242, + 1202, + 1403, + 1202, + 1403, + 1354, + 242, + 1354 + ], + "score": 0.962 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 858, + 2062, + 858, + 2085, + 840, + 2085 + ], + "score": 0.792 + }, + { + "category_id": 2, + "poly": [ + 339, + 202, + 564, + 202, + 564, + 218, + 339, + 218 + ], + "score": 0.091 + }, + { + "category_id": 13, + "poly": [ + 296, + 1551, + 351, + 1551, + 351, + 1581, + 296, + 1581 + ], + "score": 0.82, + "latex": "9 9 \\%" + }, + { + "category_id": 13, + "poly": [ + 772, + 1430, + 819, + 1430, + 819, + 1459, + 772, + 1459 + ], + "score": 0.66, + "latex": "9 \\mathrm { x } 9" + }, + { + "category_id": 13, + "poly": [ + 1071, + 1828, + 1089, + 1828, + 1089, + 1850, + 1071, + 1850 + ], + "score": 0.33, + "latex": "+" + }, + { + "category_id": 14, + "poly": [ + 938, + 224, + 1392, + 224, + 1392, + 247, + 938, + 247 + ], + "score": 0.26, + "latex": "{ \\mathrm { l i n e t y p e } } \\ - \\ { \\mathrm { G m G M } } \\ = - \\ { \\mathrm { c a n } } \\ & \\ - \\ { \\mathrm { P h y t u m } } \\ - \\ { \\mathrm { c l a s s } } \\ - \\ { \\mathrm { O r d e r } } \\ \\longleftarrow \\ { \\mathrm { F a m i t y } } \\ - \\ { \\mathrm { G e n u s } }" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 755.0, + 1403.0, + 755.0, + 1403.0, + 787.0, + 295.0, + 787.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 786.0, + 1406.0, + 786.0, + 1406.0, + 819.0, + 295.0, + 819.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 816.0, + 1404.0, + 816.0, + 1404.0, + 849.0, + 294.0, + 849.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 844.0, + 1406.0, + 844.0, + 1406.0, + 883.0, + 293.0, + 883.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 875.0, + 381.0, + 875.0, + 381.0, + 910.0, + 293.0, + 910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 337.0, + 200.0, + 567.0, + 200.0, + 567.0, + 220.0, + 337.0, + 220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 942.0, + 200.0, + 1151.0, + 200.0, + 1151.0, + 220.0, + 942.0, + 220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 341.0, + 227.0, + 398.0, + 227.0, + 398.0, + 247.0, + 341.0, + 247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 409.0, + 227.0, + 493.0, + 227.0, + 493.0, + 245.0, + 409.0, + 245.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 533.0, + 226.0, + 576.0, + 226.0, + 576.0, + 246.0, + 533.0, + 246.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 593.0, + 228.0, + 626.0, + 228.0, + 626.0, + 245.0, + 593.0, + 245.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 643.0, + 228.0, + 677.0, + 228.0, + 677.0, + 245.0, + 643.0, + 245.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 697.0, + 228.0, + 733.0, + 228.0, + 733.0, + 246.0, + 697.0, + 246.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 752.0, + 228.0, + 788.0, + 228.0, + 788.0, + 245.0, + 752.0, + 245.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 922.0, + 294.0, + 946.0, + 294.0, + 946.0, + 307.0, + 922.0, + 307.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 327.0, + 339.0, + 327.0, + 339.0, + 341.0, + 318.0, + 341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 922.0, + 372.0, + 947.0, + 372.0, + 947.0, + 390.0, + 922.0, + 390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 300.0, + 418.0, + 342.0, + 418.0, + 342.0, + 506.0, + 300.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 902.0, + 418.0, + 926.0, + 418.0, + 926.0, + 507.0, + 902.0, + 507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 922.0, + 452.0, + 945.0, + 452.0, + 945.0, + 471.0, + 922.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 317.0, + 526.0, + 342.0, + 526.0, + 342.0, + 544.0, + 317.0, + 544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 924.0, + 535.0, + 945.0, + 535.0, + 945.0, + 550.0, + 924.0, + 550.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 920.0, + 614.0, + 946.0, + 614.0, + 946.0, + 631.0, + 920.0, + 631.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 628.0, + 342.0, + 628.0, + 342.0, + 642.0, + 316.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 413.0, + 661.0, + 438.0, + 661.0, + 438.0, + 680.0, + 413.0, + 680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 566.0, + 662.0, + 597.0, + 662.0, + 597.0, + 680.0, + 566.0, + 680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 718.0, + 662.0, + 753.0, + 662.0, + 753.0, + 680.0, + 718.0, + 680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1019.0, + 661.0, + 1044.0, + 661.0, + 1044.0, + 680.0, + 1019.0, + 680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1171.0, + 662.0, + 1201.0, + 662.0, + 1201.0, + 680.0, + 1171.0, + 680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1324.0, + 662.0, + 1359.0, + 662.0, + 1359.0, + 680.0, + 1324.0, + 680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 512.0, + 670.0, + 616.0, + 670.0, + 616.0, + 696.0, + 512.0, + 696.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1117.0, + 671.0, + 1222.0, + 671.0, + 1222.0, + 694.0, + 1117.0, + 694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 523.0, + 705.0, + 570.0, + 705.0, + 570.0, + 744.0, + 523.0, + 744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1129.0, + 705.0, + 1175.0, + 705.0, + 1175.0, + 744.0, + 1129.0, + 744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2090.0, + 838.0, + 2090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 338.0, + 202.0, + 566.0, + 202.0, + 566.0, + 219.0, + 338.0, + 219.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 981.0, + 277.0, + 981.0, + 277.0, + 1008.0, + 241.0, + 1008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 970.0, + 1407.0, + 970.0, + 1407.0, + 1013.0, + 292.0, + 1013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1011.0, + 279.0, + 1011.0, + 279.0, + 1037.0, + 241.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1005.0, + 1404.0, + 1005.0, + 1404.0, + 1040.0, + 295.0, + 1040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1041.0, + 277.0, + 1041.0, + 277.0, + 1068.0, + 241.0, + 1068.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1034.0, + 1404.0, + 1034.0, + 1404.0, + 1069.0, + 295.0, + 1069.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 238.0, + 1071.0, + 280.0, + 1071.0, + 280.0, + 1100.0, + 238.0, + 1100.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1067.0, + 1405.0, + 1067.0, + 1405.0, + 1102.0, + 295.0, + 1102.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 238.0, + 1100.0, + 279.0, + 1100.0, + 279.0, + 1131.0, + 238.0, + 1131.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1095.0, + 1407.0, + 1095.0, + 1407.0, + 1134.0, + 293.0, + 1134.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1130.0, + 280.0, + 1130.0, + 280.0, + 1161.0, + 239.0, + 1161.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1123.0, + 1407.0, + 1123.0, + 1407.0, + 1163.0, + 291.0, + 1163.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1163.0, + 277.0, + 1163.0, + 277.0, + 1189.0, + 239.0, + 1189.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1158.0, + 398.0, + 1158.0, + 398.0, + 1190.0, + 293.0, + 1190.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1375.0, + 278.0, + 1375.0, + 278.0, + 1401.0, + 241.0, + 1401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1369.0, + 1405.0, + 1369.0, + 1405.0, + 1404.0, + 295.0, + 1404.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1404.0, + 278.0, + 1404.0, + 278.0, + 1431.0, + 240.0, + 1431.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1396.0, + 1407.0, + 1396.0, + 1407.0, + 1436.0, + 292.0, + 1436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1435.0, + 278.0, + 1435.0, + 278.0, + 1462.0, + 240.0, + 1462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1427.0, + 771.0, + 1427.0, + 771.0, + 1464.0, + 292.0, + 1464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 820.0, + 1427.0, + 1407.0, + 1427.0, + 1407.0, + 1464.0, + 820.0, + 1464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1466.0, + 278.0, + 1466.0, + 278.0, + 1492.0, + 240.0, + 1492.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1460.0, + 1404.0, + 1460.0, + 1404.0, + 1494.0, + 294.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 238.0, + 1494.0, + 279.0, + 1494.0, + 279.0, + 1524.0, + 238.0, + 1524.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1489.0, + 1408.0, + 1489.0, + 1408.0, + 1529.0, + 292.0, + 1529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1526.0, + 278.0, + 1526.0, + 278.0, + 1552.0, + 240.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1520.0, + 1405.0, + 1520.0, + 1405.0, + 1555.0, + 294.0, + 1555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1548.0, + 365.0, + 1548.0, + 365.0, + 1586.0, + 352.0, + 1586.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1557.0, + 277.0, + 1557.0, + 277.0, + 1583.0, + 240.0, + 1583.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1828.0, + 279.0, + 1828.0, + 279.0, + 1856.0, + 239.0, + 1856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1821.0, + 1070.0, + 1821.0, + 1070.0, + 1858.0, + 292.0, + 1858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1090.0, + 1821.0, + 1407.0, + 1821.0, + 1407.0, + 1858.0, + 1090.0, + 1858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1858.0, + 279.0, + 1858.0, + 279.0, + 1888.0, + 239.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1852.0, + 1407.0, + 1852.0, + 1407.0, + 1890.0, + 291.0, + 1890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1889.0, + 277.0, + 1889.0, + 277.0, + 1917.0, + 239.0, + 1917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1884.0, + 1404.0, + 1884.0, + 1404.0, + 1919.0, + 293.0, + 1919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1919.0, + 279.0, + 1919.0, + 279.0, + 1947.0, + 239.0, + 1947.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1914.0, + 1406.0, + 1914.0, + 1406.0, + 1950.0, + 293.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1950.0, + 279.0, + 1950.0, + 279.0, + 1978.0, + 239.0, + 1978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1944.0, + 1408.0, + 1944.0, + 1408.0, + 1980.0, + 290.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1980.0, + 280.0, + 1980.0, + 280.0, + 2008.0, + 239.0, + 2008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1973.0, + 1406.0, + 1973.0, + 1406.0, + 2012.0, + 292.0, + 2012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1601.0, + 278.0, + 1601.0, + 278.0, + 1628.0, + 241.0, + 1628.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1593.0, + 1408.0, + 1593.0, + 1408.0, + 1636.0, + 291.0, + 1636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1632.0, + 278.0, + 1632.0, + 278.0, + 1659.0, + 241.0, + 1659.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1625.0, + 1409.0, + 1625.0, + 1409.0, + 1665.0, + 294.0, + 1665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1663.0, + 278.0, + 1663.0, + 278.0, + 1690.0, + 241.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1658.0, + 1405.0, + 1658.0, + 1405.0, + 1693.0, + 294.0, + 1693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1693.0, + 278.0, + 1693.0, + 278.0, + 1719.0, + 240.0, + 1719.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1686.0, + 1407.0, + 1686.0, + 1407.0, + 1724.0, + 291.0, + 1724.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1724.0, + 278.0, + 1724.0, + 278.0, + 1750.0, + 241.0, + 1750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1715.0, + 1407.0, + 1715.0, + 1407.0, + 1753.0, + 291.0, + 1753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1755.0, + 277.0, + 1755.0, + 277.0, + 1781.0, + 241.0, + 1781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1746.0, + 1405.0, + 1746.0, + 1405.0, + 1785.0, + 293.0, + 1785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1784.0, + 278.0, + 1784.0, + 278.0, + 1811.0, + 241.0, + 1811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1777.0, + 640.0, + 1777.0, + 640.0, + 1816.0, + 293.0, + 1816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1206.0, + 279.0, + 1206.0, + 279.0, + 1235.0, + 239.0, + 1235.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1199.0, + 1406.0, + 1199.0, + 1406.0, + 1237.0, + 292.0, + 1237.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1239.0, + 275.0, + 1239.0, + 275.0, + 1264.0, + 241.0, + 1264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1233.0, + 1404.0, + 1233.0, + 1404.0, + 1266.0, + 294.0, + 1266.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1270.0, + 277.0, + 1270.0, + 277.0, + 1295.0, + 241.0, + 1295.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1261.0, + 1406.0, + 1261.0, + 1406.0, + 1300.0, + 292.0, + 1300.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1298.0, + 279.0, + 1298.0, + 279.0, + 1328.0, + 239.0, + 1328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1292.0, + 1404.0, + 1292.0, + 1404.0, + 1329.0, + 293.0, + 1329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1328.0, + 280.0, + 1328.0, + 280.0, + 1357.0, + 241.0, + 1357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1324.0, + 868.0, + 1324.0, + 868.0, + 1357.0, + 294.0, + 1357.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 7, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 240, + 1470, + 1404, + 1470, + 1404, + 1714, + 240, + 1714 + ], + "score": 0.972 + }, + { + "category_id": 1, + "poly": [ + 241, + 1136, + 1405, + 1136, + 1405, + 1290, + 241, + 1290 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 241, + 1303, + 1404, + 1303, + 1404, + 1456, + 241, + 1456 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 287, + 1822, + 1402, + 1822, + 1402, + 2008, + 287, + 2008 + ], + "score": 0.965 + }, + { + "category_id": 1, + "poly": [ + 245, + 904, + 1405, + 904, + 1405, + 1027, + 245, + 1027 + ], + "score": 0.96 + }, + { + "category_id": 3, + "poly": [ + 310, + 203, + 1389, + 203, + 1389, + 737, + 310, + 737 + ], + "score": 0.959 + }, + { + "category_id": 4, + "poly": [ + 295, + 755, + 1405, + 755, + 1405, + 848, + 295, + 848 + ], + "score": 0.957 + }, + { + "category_id": 0, + "poly": [ + 252, + 1755, + 509, + 1755, + 509, + 1792, + 252, + 1792 + ], + "score": 0.888 + }, + { + "category_id": 0, + "poly": [ + 257, + 1068, + 515, + 1068, + 515, + 1106, + 257, + 1106 + ], + "score": 0.882 + }, + { + "category_id": 2, + "poly": [ + 840, + 2061, + 859, + 2061, + 859, + 2084, + 840, + 2084 + ], + "score": 0.79 + }, + { + "category_id": 13, + "poly": [ + 817, + 756, + 862, + 756, + 862, + 785, + 817, + 785 + ], + "score": 0.26, + "latex": "1 0 \\mathrm { x }" + }, + { + "category_id": 15, + "poly": [ + 1204.0, + 214.0, + 1234.0, + 214.0, + 1234.0, + 232.0, + 1204.0, + 232.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 644.0, + 304.0, + 689.0, + 304.0, + 689.0, + 389.0, + 644.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1359.0, + 448.0, + 1386.0, + 448.0, + 1386.0, + 472.0, + 1359.0, + 472.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1161.0, + 667.0, + 1177.0, + 667.0, + 1177.0, + 681.0, + 1161.0, + 681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 527.0, + 708.0, + 567.0, + 708.0, + 567.0, + 741.0, + 527.0, + 741.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1130.0, + 705.0, + 1174.0, + 705.0, + 1174.0, + 744.0, + 1130.0, + 744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1239.25, + 226.5, + 1268.25, + 226.5, + 1268.25, + 244.5, + 1239.25, + 244.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 688.25, + 296.0, + 766.25, + 296.0, + 766.25, + 327.0, + 688.25, + 327.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 935.0, + 418.5, + 1050.0, + 418.5, + 1050.0, + 452.0, + 935.0, + 452.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 518.5, + 437.0, + 518.5, + 437.0, + 595.5, + 289.0, + 595.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 753.0, + 816.0, + 753.0, + 816.0, + 791.0, + 294.0, + 791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 863.0, + 753.0, + 1407.0, + 753.0, + 1407.0, + 791.0, + 863.0, + 791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 783.0, + 1405.0, + 783.0, + 1405.0, + 822.0, + 293.0, + 822.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 814.0, + 995.0, + 814.0, + 995.0, + 852.0, + 293.0, + 852.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 238.0, + 1750.0, + 515.0, + 1750.0, + 515.0, + 1799.0, + 238.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 245.0, + 1067.0, + 519.0, + 1067.0, + 519.0, + 1110.0, + 245.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2060.0, + 861.0, + 2060.0, + 861.0, + 2090.0, + 839.0, + 2090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1476.0, + 277.0, + 1476.0, + 277.0, + 1502.0, + 239.0, + 1502.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1467.0, + 1405.0, + 1467.0, + 1405.0, + 1507.0, + 293.0, + 1507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1505.0, + 278.0, + 1505.0, + 278.0, + 1534.0, + 237.0, + 1534.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1498.0, + 1407.0, + 1498.0, + 1407.0, + 1536.0, + 293.0, + 1536.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1534.0, + 280.0, + 1534.0, + 280.0, + 1564.0, + 237.0, + 1564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1532.0, + 1405.0, + 1532.0, + 1405.0, + 1565.0, + 295.0, + 1565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1565.0, + 280.0, + 1565.0, + 280.0, + 1595.0, + 237.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1558.0, + 1405.0, + 1558.0, + 1405.0, + 1599.0, + 293.0, + 1599.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1596.0, + 280.0, + 1596.0, + 280.0, + 1624.0, + 237.0, + 1624.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1591.0, + 1407.0, + 1591.0, + 1407.0, + 1627.0, + 294.0, + 1627.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 1628.0, + 278.0, + 1628.0, + 278.0, + 1652.0, + 239.0, + 1652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1620.0, + 1407.0, + 1620.0, + 1407.0, + 1656.0, + 291.0, + 1656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 238.0, + 1654.0, + 280.0, + 1654.0, + 280.0, + 1687.0, + 238.0, + 1687.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1650.0, + 1405.0, + 1650.0, + 1405.0, + 1689.0, + 291.0, + 1689.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1687.0, + 280.0, + 1687.0, + 280.0, + 1716.0, + 237.0, + 1716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1682.0, + 1017.0, + 1682.0, + 1017.0, + 1717.0, + 291.0, + 1717.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 238.0, + 1140.0, + 278.0, + 1140.0, + 278.0, + 1169.0, + 238.0, + 1169.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1134.0, + 1406.0, + 1134.0, + 1406.0, + 1172.0, + 292.0, + 1172.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1173.0, + 275.0, + 1173.0, + 275.0, + 1198.0, + 240.0, + 1198.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1164.0, + 1408.0, + 1164.0, + 1408.0, + 1204.0, + 292.0, + 1204.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1204.0, + 278.0, + 1204.0, + 278.0, + 1229.0, + 240.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1197.0, + 1406.0, + 1197.0, + 1406.0, + 1233.0, + 292.0, + 1233.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 238.0, + 1232.0, + 279.0, + 1232.0, + 279.0, + 1261.0, + 238.0, + 1261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1225.0, + 1406.0, + 1225.0, + 1406.0, + 1265.0, + 292.0, + 1265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 238.0, + 1263.0, + 281.0, + 1263.0, + 281.0, + 1291.0, + 238.0, + 1291.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1256.0, + 417.0, + 1256.0, + 417.0, + 1295.0, + 293.0, + 1295.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 238.0, + 1307.0, + 279.0, + 1307.0, + 279.0, + 1336.0, + 238.0, + 1336.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1302.0, + 1404.0, + 1302.0, + 1404.0, + 1337.0, + 294.0, + 1337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1340.0, + 278.0, + 1340.0, + 278.0, + 1365.0, + 240.0, + 1365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1333.0, + 1407.0, + 1333.0, + 1407.0, + 1368.0, + 292.0, + 1368.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1370.0, + 277.0, + 1370.0, + 277.0, + 1395.0, + 240.0, + 1395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1365.0, + 1404.0, + 1365.0, + 1404.0, + 1398.0, + 295.0, + 1398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1401.0, + 278.0, + 1401.0, + 278.0, + 1426.0, + 240.0, + 1426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1395.0, + 1405.0, + 1395.0, + 1405.0, + 1428.0, + 295.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 1431.0, + 278.0, + 1431.0, + 278.0, + 1455.0, + 240.0, + 1455.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1422.0, + 911.0, + 1422.0, + 911.0, + 1460.0, + 291.0, + 1460.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1821.0, + 1405.0, + 1821.0, + 1405.0, + 1858.0, + 294.0, + 1858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1858.0, + 1404.0, + 1858.0, + 1404.0, + 1886.0, + 296.0, + 1886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1884.0, + 1405.0, + 1884.0, + 1405.0, + 1919.0, + 293.0, + 1919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1916.0, + 1404.0, + 1916.0, + 1404.0, + 1948.0, + 295.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1944.0, + 1405.0, + 1944.0, + 1405.0, + 1981.0, + 293.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1974.0, + 1339.0, + 1974.0, + 1339.0, + 2012.0, + 293.0, + 2012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 910.0, + 278.0, + 910.0, + 278.0, + 936.0, + 241.0, + 936.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 904.0, + 1406.0, + 904.0, + 1406.0, + 939.0, + 295.0, + 939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 240.0, + 938.0, + 279.0, + 938.0, + 279.0, + 968.0, + 240.0, + 968.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 936.0, + 1404.0, + 936.0, + 1404.0, + 967.0, + 296.0, + 967.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 971.0, + 278.0, + 971.0, + 278.0, + 998.0, + 241.0, + 998.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 966.0, + 1406.0, + 966.0, + 1406.0, + 1000.0, + 293.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 241.0, + 1000.0, + 278.0, + 1000.0, + 278.0, + 1027.0, + 241.0, + 1027.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 995.0, + 562.0, + 995.0, + 562.0, + 1030.0, + 296.0, + 1030.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 8, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 235, + 177, + 1411, + 177, + 1411, + 2013, + 235, + 2013 + ], + "score": 0.88 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 866, + 2061, + 866, + 2086, + 836, + 2086 + ], + "score": 0.838 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 239.0, + 206.0, + 281.0, + 206.0, + 281.0, + 236.0, + 239.0, + 236.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 195.0, + 461.0, + 195.0, + 461.0, + 240.0, + 291.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 256.0, + 281.0, + 256.0, + 281.0, + 286.0, + 237.0, + 286.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 309.0, + 252.0, + 354.0, + 252.0, + 354.0, + 290.0, + 309.0, + 290.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 250.0, + 1405.0, + 250.0, + 1405.0, + 290.0, + 364.0, + 290.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 284.0, + 279.0, + 284.0, + 279.0, + 317.0, + 237.0, + 317.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 276.0, + 1407.0, + 276.0, + 1407.0, + 323.0, + 364.0, + 323.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 313.0, + 283.0, + 313.0, + 283.0, + 349.0, + 235.0, + 349.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 370.0, + 315.0, + 1403.0, + 315.0, + 1403.0, + 347.0, + 370.0, + 347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 345.0, + 283.0, + 345.0, + 283.0, + 381.0, + 235.0, + 381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 343.0, + 645.0, + 343.0, + 645.0, + 379.0, + 364.0, + 379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 389.0, + 279.0, + 389.0, + 279.0, + 421.0, + 237.0, + 421.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 311.0, + 387.0, + 352.0, + 387.0, + 352.0, + 419.0, + 311.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 383.0, + 1405.0, + 383.0, + 1405.0, + 421.0, + 364.0, + 421.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 417.0, + 279.0, + 417.0, + 279.0, + 450.0, + 237.0, + 450.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 413.0, + 1405.0, + 413.0, + 1405.0, + 452.0, + 368.0, + 452.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 450.0, + 281.0, + 450.0, + 281.0, + 480.0, + 237.0, + 480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 444.0, + 1409.0, + 444.0, + 1409.0, + 482.0, + 366.0, + 482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 478.0, + 279.0, + 478.0, + 279.0, + 510.0, + 237.0, + 510.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 362.0, + 472.0, + 1411.0, + 472.0, + 1411.0, + 516.0, + 362.0, + 516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 506.0, + 283.0, + 506.0, + 283.0, + 542.0, + 235.0, + 542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 504.0, + 1407.0, + 504.0, + 1407.0, + 542.0, + 364.0, + 542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 536.0, + 283.0, + 536.0, + 283.0, + 573.0, + 235.0, + 573.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 534.0, + 524.0, + 534.0, + 524.0, + 573.0, + 366.0, + 573.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 579.0, + 283.0, + 579.0, + 283.0, + 615.0, + 235.0, + 615.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 577.0, + 354.0, + 577.0, + 354.0, + 615.0, + 307.0, + 615.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 579.0, + 1405.0, + 579.0, + 1405.0, + 611.0, + 368.0, + 611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 611.0, + 279.0, + 611.0, + 279.0, + 641.0, + 237.0, + 641.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 609.0, + 1407.0, + 609.0, + 1407.0, + 641.0, + 368.0, + 641.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 637.0, + 283.0, + 637.0, + 283.0, + 674.0, + 235.0, + 674.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 637.0, + 1407.0, + 637.0, + 1407.0, + 673.0, + 364.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 671.0, + 279.0, + 671.0, + 279.0, + 704.0, + 237.0, + 704.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 667.0, + 1252.0, + 667.0, + 1252.0, + 706.0, + 366.0, + 706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 710.0, + 283.0, + 710.0, + 283.0, + 746.0, + 235.0, + 746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 710.0, + 354.0, + 710.0, + 354.0, + 746.0, + 307.0, + 746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 708.0, + 1407.0, + 708.0, + 1407.0, + 746.0, + 366.0, + 746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 744.0, + 279.0, + 744.0, + 279.0, + 774.0, + 237.0, + 774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 742.0, + 1403.0, + 742.0, + 1403.0, + 774.0, + 368.0, + 774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 774.0, + 281.0, + 774.0, + 281.0, + 805.0, + 237.0, + 805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 772.0, + 1405.0, + 772.0, + 1405.0, + 805.0, + 368.0, + 805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 805.0, + 281.0, + 805.0, + 281.0, + 835.0, + 237.0, + 835.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 799.0, + 524.0, + 799.0, + 524.0, + 837.0, + 366.0, + 837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 845.0, + 281.0, + 845.0, + 281.0, + 875.0, + 237.0, + 875.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 311.0, + 843.0, + 352.0, + 843.0, + 352.0, + 875.0, + 311.0, + 875.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 843.0, + 1403.0, + 843.0, + 1403.0, + 875.0, + 366.0, + 875.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 877.0, + 281.0, + 877.0, + 281.0, + 907.0, + 237.0, + 907.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 867.0, + 1407.0, + 867.0, + 1407.0, + 914.0, + 364.0, + 914.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 907.0, + 281.0, + 907.0, + 281.0, + 938.0, + 237.0, + 938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 899.0, + 1409.0, + 899.0, + 1409.0, + 944.0, + 366.0, + 944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 936.0, + 279.0, + 936.0, + 279.0, + 968.0, + 237.0, + 968.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 932.0, + 1405.0, + 932.0, + 1405.0, + 970.0, + 366.0, + 970.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 964.0, + 283.0, + 964.0, + 283.0, + 1000.0, + 235.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 962.0, + 646.0, + 962.0, + 646.0, + 1000.0, + 366.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1004.0, + 283.0, + 1004.0, + 283.0, + 1041.0, + 235.0, + 1041.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 1004.0, + 354.0, + 1004.0, + 354.0, + 1041.0, + 307.0, + 1041.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1006.0, + 1403.0, + 1006.0, + 1403.0, + 1039.0, + 368.0, + 1039.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1039.0, + 279.0, + 1039.0, + 279.0, + 1071.0, + 237.0, + 1071.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1035.0, + 1407.0, + 1035.0, + 1407.0, + 1073.0, + 368.0, + 1073.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1079.0, + 279.0, + 1079.0, + 279.0, + 1111.0, + 237.0, + 1111.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 1075.0, + 354.0, + 1075.0, + 354.0, + 1113.0, + 307.0, + 1113.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1075.0, + 1407.0, + 1075.0, + 1407.0, + 1113.0, + 366.0, + 1113.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1109.0, + 283.0, + 1109.0, + 283.0, + 1145.0, + 235.0, + 1145.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1105.0, + 1136.0, + 1105.0, + 1136.0, + 1143.0, + 368.0, + 1143.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1151.0, + 281.0, + 1151.0, + 281.0, + 1182.0, + 237.0, + 1182.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 1147.0, + 354.0, + 1147.0, + 354.0, + 1186.0, + 307.0, + 1186.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 1143.0, + 1407.0, + 1143.0, + 1407.0, + 1184.0, + 364.0, + 1184.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1184.0, + 281.0, + 1184.0, + 281.0, + 1214.0, + 237.0, + 1214.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 1174.0, + 1409.0, + 1174.0, + 1409.0, + 1218.0, + 365.0, + 1218.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1210.0, + 283.0, + 1210.0, + 283.0, + 1246.0, + 235.0, + 1246.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1208.0, + 1014.0, + 1208.0, + 1014.0, + 1244.0, + 366.0, + 1244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1250.0, + 283.0, + 1250.0, + 283.0, + 1287.0, + 235.0, + 1287.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 1250.0, + 354.0, + 1250.0, + 354.0, + 1287.0, + 307.0, + 1287.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1248.0, + 1405.0, + 1248.0, + 1405.0, + 1287.0, + 366.0, + 1287.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1285.0, + 279.0, + 1285.0, + 279.0, + 1315.0, + 237.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1279.0, + 1405.0, + 1279.0, + 1405.0, + 1317.0, + 366.0, + 1317.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1311.0, + 283.0, + 1311.0, + 283.0, + 1347.0, + 235.0, + 1347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1309.0, + 1407.0, + 1309.0, + 1407.0, + 1347.0, + 368.0, + 1347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1345.0, + 279.0, + 1345.0, + 279.0, + 1377.0, + 237.0, + 1377.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1337.0, + 1365.0, + 1337.0, + 1365.0, + 1375.0, + 368.0, + 1375.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1385.0, + 283.0, + 1385.0, + 283.0, + 1416.0, + 237.0, + 1416.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1383.0, + 354.0, + 1383.0, + 354.0, + 1418.0, + 295.0, + 1418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1377.0, + 1409.0, + 1377.0, + 1409.0, + 1420.0, + 363.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1416.0, + 279.0, + 1416.0, + 279.0, + 1448.0, + 237.0, + 1448.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 1410.0, + 1409.0, + 1410.0, + 1409.0, + 1454.0, + 364.0, + 1454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1444.0, + 283.0, + 1444.0, + 283.0, + 1480.0, + 235.0, + 1480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1442.0, + 1405.0, + 1442.0, + 1405.0, + 1480.0, + 366.0, + 1480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1476.0, + 279.0, + 1476.0, + 279.0, + 1508.0, + 237.0, + 1508.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 1470.0, + 943.0, + 1470.0, + 943.0, + 1510.0, + 365.0, + 1510.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1519.0, + 281.0, + 1519.0, + 281.0, + 1549.0, + 237.0, + 1549.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1514.0, + 354.0, + 1514.0, + 354.0, + 1551.0, + 293.0, + 1551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1512.0, + 1407.0, + 1512.0, + 1407.0, + 1551.0, + 368.0, + 1551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1545.0, + 283.0, + 1545.0, + 283.0, + 1581.0, + 235.0, + 1581.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1545.0, + 1409.0, + 1545.0, + 1409.0, + 1583.0, + 368.0, + 1583.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1579.0, + 281.0, + 1579.0, + 281.0, + 1609.0, + 237.0, + 1609.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1573.0, + 1409.0, + 1573.0, + 1409.0, + 1611.0, + 366.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1609.0, + 281.0, + 1609.0, + 281.0, + 1639.0, + 237.0, + 1639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1605.0, + 1407.0, + 1605.0, + 1407.0, + 1644.0, + 366.0, + 1644.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1635.0, + 283.0, + 1635.0, + 283.0, + 1672.0, + 235.0, + 1672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 370.0, + 1637.0, + 1084.0, + 1637.0, + 1084.0, + 1670.0, + 370.0, + 1670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1680.0, + 283.0, + 1680.0, + 283.0, + 1710.0, + 237.0, + 1710.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1678.0, + 354.0, + 1678.0, + 354.0, + 1712.0, + 293.0, + 1712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1676.0, + 1407.0, + 1676.0, + 1407.0, + 1714.0, + 366.0, + 1714.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1712.0, + 279.0, + 1712.0, + 279.0, + 1744.0, + 237.0, + 1744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1706.0, + 1407.0, + 1706.0, + 1407.0, + 1744.0, + 366.0, + 1744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1740.0, + 279.0, + 1740.0, + 279.0, + 1773.0, + 237.0, + 1773.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1736.0, + 560.0, + 1736.0, + 560.0, + 1775.0, + 366.0, + 1775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1779.0, + 285.0, + 1779.0, + 285.0, + 1815.0, + 235.0, + 1815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1779.0, + 354.0, + 1779.0, + 354.0, + 1813.0, + 291.0, + 1813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1771.0, + 1409.0, + 1771.0, + 1409.0, + 1819.0, + 363.0, + 1819.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1811.0, + 283.0, + 1811.0, + 283.0, + 1847.0, + 235.0, + 1847.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1809.0, + 1409.0, + 1809.0, + 1409.0, + 1847.0, + 368.0, + 1847.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1843.0, + 281.0, + 1843.0, + 281.0, + 1873.0, + 237.0, + 1873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 370.0, + 1841.0, + 1401.0, + 1841.0, + 1401.0, + 1873.0, + 370.0, + 1873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 236.0, + 1866.0, + 284.0, + 1866.0, + 284.0, + 1910.0, + 236.0, + 1910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 1865.0, + 799.0, + 1865.0, + 799.0, + 1906.0, + 365.0, + 1906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1912.0, + 285.0, + 1912.0, + 285.0, + 1948.0, + 235.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1912.0, + 354.0, + 1912.0, + 354.0, + 1946.0, + 293.0, + 1946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 1910.0, + 1407.0, + 1910.0, + 1407.0, + 1948.0, + 366.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1942.0, + 281.0, + 1942.0, + 281.0, + 1980.0, + 235.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1940.0, + 1409.0, + 1940.0, + 1409.0, + 1978.0, + 368.0, + 1978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1972.0, + 283.0, + 1972.0, + 283.0, + 2009.0, + 235.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 362.0, + 1968.0, + 1381.0, + 1968.0, + 1381.0, + 2011.0, + 362.0, + 2011.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 9, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 232, + 162, + 1412, + 162, + 1412, + 1862, + 232, + 1862 + ], + "score": 0.805 + }, + { + "category_id": 2, + "poly": [ + 835, + 2061, + 863, + 2061, + 863, + 2086, + 835, + 2086 + ], + "score": 0.751 + }, + { + "category_id": 2, + "poly": [ + 835, + 2061, + 863, + 2061, + 863, + 2086, + 835, + 2086 + ], + "score": 0.268 + }, + { + "category_id": 13, + "poly": [ + 518, + 999, + 564, + 999, + 564, + 1025, + 518, + 1025 + ], + "score": 0.29, + "latex": "{ \\tt = } 1 0" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2098.0, + 832.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2098.0, + 832.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 206.0, + 283.0, + 206.0, + 283.0, + 239.0, + 235.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 204.0, + 353.0, + 204.0, + 353.0, + 238.0, + 292.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 198.0, + 1409.0, + 198.0, + 1409.0, + 243.0, + 363.0, + 243.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 236.0, + 281.0, + 236.0, + 281.0, + 269.0, + 235.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 228.0, + 1411.0, + 228.0, + 1411.0, + 271.0, + 365.0, + 271.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 266.0, + 281.0, + 266.0, + 281.0, + 299.0, + 235.0, + 299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 262.0, + 1321.0, + 262.0, + 1321.0, + 298.0, + 369.0, + 298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 307.0, + 283.0, + 307.0, + 283.0, + 341.0, + 235.0, + 341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 307.0, + 352.0, + 307.0, + 352.0, + 341.0, + 292.0, + 341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 305.0, + 1409.0, + 305.0, + 1409.0, + 341.0, + 369.0, + 341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 334.0, + 283.0, + 334.0, + 283.0, + 374.0, + 237.0, + 374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 333.0, + 1407.0, + 333.0, + 1407.0, + 373.0, + 365.0, + 373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 371.0, + 279.0, + 371.0, + 279.0, + 399.0, + 237.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 367.0, + 1407.0, + 367.0, + 1407.0, + 403.0, + 369.0, + 403.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 399.0, + 281.0, + 399.0, + 281.0, + 433.0, + 235.0, + 433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 395.0, + 523.0, + 395.0, + 523.0, + 431.0, + 369.0, + 431.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 442.0, + 281.0, + 442.0, + 281.0, + 470.0, + 237.0, + 470.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 438.0, + 352.0, + 438.0, + 352.0, + 472.0, + 292.0, + 472.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 433.0, + 1409.0, + 433.0, + 1409.0, + 476.0, + 363.0, + 476.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 470.0, + 281.0, + 470.0, + 281.0, + 504.0, + 235.0, + 504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 463.0, + 1407.0, + 463.0, + 1407.0, + 506.0, + 363.0, + 506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 502.0, + 281.0, + 502.0, + 281.0, + 536.0, + 235.0, + 536.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 498.0, + 645.0, + 498.0, + 645.0, + 534.0, + 367.0, + 534.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 541.0, + 283.0, + 541.0, + 283.0, + 575.0, + 235.0, + 575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 541.0, + 352.0, + 541.0, + 352.0, + 575.0, + 294.0, + 575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 539.0, + 1407.0, + 539.0, + 1407.0, + 575.0, + 367.0, + 575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 571.0, + 281.0, + 571.0, + 281.0, + 605.0, + 235.0, + 605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 569.0, + 1409.0, + 569.0, + 1409.0, + 605.0, + 367.0, + 605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 603.0, + 281.0, + 603.0, + 281.0, + 637.0, + 235.0, + 637.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 601.0, + 1407.0, + 601.0, + 1407.0, + 637.0, + 369.0, + 637.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 633.0, + 281.0, + 633.0, + 281.0, + 667.0, + 235.0, + 667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 366.0, + 623.0, + 524.0, + 623.0, + 524.0, + 668.0, + 366.0, + 668.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 676.0, + 281.0, + 676.0, + 281.0, + 705.0, + 237.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 673.0, + 353.0, + 673.0, + 353.0, + 706.0, + 291.0, + 706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 361.0, + 665.0, + 1405.0, + 665.0, + 1405.0, + 712.0, + 361.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 703.0, + 283.0, + 703.0, + 283.0, + 738.0, + 235.0, + 738.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 697.0, + 630.0, + 697.0, + 630.0, + 738.0, + 367.0, + 738.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 746.0, + 283.0, + 746.0, + 283.0, + 780.0, + 235.0, + 780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 744.0, + 352.0, + 744.0, + 352.0, + 778.0, + 294.0, + 778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 742.0, + 1409.0, + 742.0, + 1409.0, + 778.0, + 367.0, + 778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 776.0, + 281.0, + 776.0, + 281.0, + 809.0, + 235.0, + 809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 770.0, + 1407.0, + 770.0, + 1407.0, + 810.0, + 363.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 806.0, + 281.0, + 806.0, + 281.0, + 840.0, + 235.0, + 840.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 800.0, + 1405.0, + 800.0, + 1405.0, + 840.0, + 365.0, + 840.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 838.0, + 281.0, + 838.0, + 281.0, + 871.0, + 235.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 834.0, + 1033.0, + 834.0, + 1033.0, + 869.0, + 367.0, + 869.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 877.0, + 283.0, + 877.0, + 283.0, + 911.0, + 235.0, + 911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 873.0, + 353.0, + 873.0, + 353.0, + 913.0, + 292.0, + 913.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 875.0, + 1407.0, + 875.0, + 1407.0, + 911.0, + 369.0, + 911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 909.0, + 283.0, + 909.0, + 283.0, + 943.0, + 235.0, + 943.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 903.0, + 1405.0, + 903.0, + 1405.0, + 943.0, + 367.0, + 943.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 939.0, + 281.0, + 939.0, + 281.0, + 973.0, + 235.0, + 973.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 937.0, + 1411.0, + 937.0, + 1411.0, + 973.0, + 369.0, + 973.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 969.0, + 281.0, + 969.0, + 281.0, + 1003.0, + 235.0, + 1003.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 965.0, + 1409.0, + 965.0, + 1409.0, + 1006.0, + 369.0, + 1006.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 999.0, + 281.0, + 999.0, + 281.0, + 1033.0, + 235.0, + 1033.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 997.0, + 517.0, + 997.0, + 517.0, + 1033.0, + 367.0, + 1033.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 565.0, + 997.0, + 1218.0, + 997.0, + 1218.0, + 1033.0, + 565.0, + 1033.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1044.0, + 281.0, + 1044.0, + 281.0, + 1072.0, + 237.0, + 1072.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1036.0, + 353.0, + 1036.0, + 353.0, + 1076.0, + 292.0, + 1076.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 1036.0, + 1409.0, + 1036.0, + 1409.0, + 1078.0, + 365.0, + 1078.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1070.0, + 281.0, + 1070.0, + 281.0, + 1106.0, + 235.0, + 1106.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1068.0, + 1409.0, + 1068.0, + 1409.0, + 1104.0, + 369.0, + 1104.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1100.0, + 281.0, + 1100.0, + 281.0, + 1134.0, + 235.0, + 1134.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1098.0, + 1409.0, + 1098.0, + 1409.0, + 1134.0, + 369.0, + 1134.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1132.0, + 281.0, + 1132.0, + 281.0, + 1166.0, + 235.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1124.0, + 1411.0, + 1124.0, + 1411.0, + 1168.0, + 363.0, + 1168.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1162.0, + 281.0, + 1162.0, + 281.0, + 1198.0, + 235.0, + 1198.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 1158.0, + 1338.0, + 1158.0, + 1338.0, + 1198.0, + 367.0, + 1198.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1205.0, + 281.0, + 1205.0, + 281.0, + 1233.0, + 237.0, + 1233.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1201.0, + 352.0, + 1201.0, + 352.0, + 1235.0, + 292.0, + 1235.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1196.0, + 1407.0, + 1196.0, + 1407.0, + 1239.0, + 363.0, + 1239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1237.0, + 281.0, + 1237.0, + 281.0, + 1265.0, + 237.0, + 1265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1226.0, + 1407.0, + 1226.0, + 1407.0, + 1271.0, + 363.0, + 1271.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1263.0, + 281.0, + 1263.0, + 281.0, + 1297.0, + 235.0, + 1297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 1259.0, + 1411.0, + 1259.0, + 1411.0, + 1301.0, + 367.0, + 1301.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1293.0, + 283.0, + 1293.0, + 283.0, + 1327.0, + 235.0, + 1327.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 1289.0, + 1409.0, + 1289.0, + 1409.0, + 1331.0, + 367.0, + 1331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1325.0, + 283.0, + 1325.0, + 283.0, + 1359.0, + 235.0, + 1359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1321.0, + 523.0, + 1321.0, + 523.0, + 1357.0, + 369.0, + 1357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1368.0, + 281.0, + 1368.0, + 281.0, + 1396.0, + 237.0, + 1396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1364.0, + 353.0, + 1364.0, + 353.0, + 1398.0, + 294.0, + 1398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1363.0, + 1405.0, + 1363.0, + 1405.0, + 1398.0, + 369.0, + 1398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1394.0, + 281.0, + 1394.0, + 281.0, + 1430.0, + 235.0, + 1430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 1389.0, + 1409.0, + 1389.0, + 1409.0, + 1434.0, + 365.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1426.0, + 281.0, + 1426.0, + 281.0, + 1460.0, + 235.0, + 1460.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 1423.0, + 1411.0, + 1423.0, + 1411.0, + 1464.0, + 367.0, + 1464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1456.0, + 281.0, + 1456.0, + 281.0, + 1490.0, + 235.0, + 1490.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1453.0, + 988.0, + 1453.0, + 988.0, + 1492.0, + 363.0, + 1492.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1498.0, + 283.0, + 1498.0, + 283.0, + 1531.0, + 235.0, + 1531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1496.0, + 353.0, + 1496.0, + 353.0, + 1530.0, + 292.0, + 1530.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 1496.0, + 1407.0, + 1496.0, + 1407.0, + 1531.0, + 367.0, + 1531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1528.0, + 281.0, + 1528.0, + 281.0, + 1561.0, + 235.0, + 1561.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1522.0, + 1411.0, + 1522.0, + 1411.0, + 1565.0, + 363.0, + 1565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1560.0, + 281.0, + 1560.0, + 281.0, + 1593.0, + 235.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1554.0, + 1030.0, + 1554.0, + 1030.0, + 1593.0, + 363.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1603.0, + 281.0, + 1603.0, + 281.0, + 1631.0, + 237.0, + 1631.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1599.0, + 352.0, + 1599.0, + 352.0, + 1633.0, + 292.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 1597.0, + 1405.0, + 1597.0, + 1405.0, + 1633.0, + 367.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1631.0, + 281.0, + 1631.0, + 281.0, + 1665.0, + 235.0, + 1665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 1625.0, + 1409.0, + 1625.0, + 1409.0, + 1664.0, + 365.0, + 1664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1661.0, + 281.0, + 1661.0, + 281.0, + 1694.0, + 235.0, + 1694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1659.0, + 1403.0, + 1659.0, + 1403.0, + 1694.0, + 369.0, + 1694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1691.0, + 281.0, + 1691.0, + 281.0, + 1726.0, + 235.0, + 1726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1689.0, + 742.0, + 1689.0, + 742.0, + 1724.0, + 369.0, + 1724.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1732.0, + 283.0, + 1732.0, + 283.0, + 1766.0, + 235.0, + 1766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1730.0, + 352.0, + 1730.0, + 352.0, + 1764.0, + 294.0, + 1764.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1724.0, + 1407.0, + 1724.0, + 1407.0, + 1768.0, + 363.0, + 1768.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1762.0, + 281.0, + 1762.0, + 281.0, + 1798.0, + 235.0, + 1798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 1756.0, + 1409.0, + 1756.0, + 1409.0, + 1800.0, + 365.0, + 1800.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 235.0, + 1794.0, + 281.0, + 1794.0, + 281.0, + 1828.0, + 235.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 1788.0, + 1412.0, + 1788.0, + 1412.0, + 1829.0, + 365.0, + 1829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 237.0, + 1826.0, + 279.0, + 1826.0, + 279.0, + 1854.0, + 237.0, + 1854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 363.0, + 1818.0, + 1380.0, + 1818.0, + 1380.0, + 1858.0, + 363.0, + 1858.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 10, + "width": 1700, + "height": 2200 + } + } +] \ No newline at end of file diff --git a/parse/dev/TJUNtiZiTKE/TJUNtiZiTKE_model.json b/parse/dev/TJUNtiZiTKE/TJUNtiZiTKE_model.json new file mode 100644 index 0000000000000000000000000000000000000000..1cbf4e290242eb0066599b206817b930564abff0 --- /dev/null +++ b/parse/dev/TJUNtiZiTKE/TJUNtiZiTKE_model.json @@ -0,0 +1,17551 @@ +[ + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 398, + 879, + 1305, + 879, + 1305, + 1334, + 398, + 1334 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 1460, + 1404, + 1460, + 1404, + 1673, + 298, + 1673 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 1688, + 1404, + 1688, + 1404, + 1931, + 298, + 1931 + ], + "score": 0.975 + }, + { + "category_id": 0, + "poly": [ + 299, + 271, + 1404, + 271, + 1404, + 382, + 299, + 382 + ], + "score": 0.959 + }, + { + "category_id": 1, + "poly": [ + 684, + 638, + 1014, + 638, + 1014, + 731, + 684, + 731 + ], + "score": 0.943 + }, + { + "category_id": 1, + "poly": [ + 926, + 499, + 1258, + 499, + 1258, + 593, + 926, + 593 + ], + "score": 0.938 + }, + { + "category_id": 1, + "poly": [ + 1048, + 638, + 1379, + 638, + 1379, + 732, + 1048, + 732 + ], + "score": 0.937 + }, + { + "category_id": 1, + "poly": [ + 321, + 639, + 649, + 639, + 649, + 730, + 321, + 730 + ], + "score": 0.932 + }, + { + "category_id": 0, + "poly": [ + 299, + 1390, + 531, + 1390, + 531, + 1429, + 299, + 1429 + ], + "score": 0.909 + }, + { + "category_id": 0, + "poly": [ + 787, + 809, + 913, + 809, + 913, + 845, + 787, + 845 + ], + "score": 0.896 + }, + { + "category_id": 2, + "poly": [ + 297, + 2033, + 1070, + 2033, + 1070, + 2062, + 297, + 2062 + ], + "score": 0.882 + }, + { + "category_id": 2, + "poly": [ + 334, + 1954, + 530, + 1954, + 530, + 1981, + 334, + 1981 + ], + "score": 0.868 + }, + { + "category_id": 1, + "poly": [ + 441, + 499, + 772, + 499, + 772, + 591, + 441, + 591 + ], + "score": 0.753 + }, + { + "category_id": 0, + "poly": [ + 531, + 500, + 690, + 500, + 690, + 530, + 531, + 530 + ], + "score": 0.159 + }, + { + "category_id": 15, + "poly": [ + 295.0, + 272.0, + 1404.0, + 272.0, + 1404.0, + 327.0, + 295.0, + 327.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 703.0, + 322.0, + 996.0, + 322.0, + 996.0, + 389.0, + 703.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1387.0, + 535.0, + 1387.0, + 535.0, + 1435.0, + 292.0, + 1435.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 782.0, + 806.0, + 920.0, + 806.0, + 920.0, + 850.0, + 782.0, + 850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2028.0, + 1073.0, + 2028.0, + 1073.0, + 2066.0, + 294.0, + 2066.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 1949.0, + 536.0, + 1949.0, + 536.0, + 1986.0, + 331.0, + 1986.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 527.0, + 497.0, + 696.0, + 497.0, + 696.0, + 534.0, + 527.0, + 534.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 878.0, + 1307.0, + 878.0, + 1307.0, + 915.0, + 394.0, + 915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 909.0, + 1309.0, + 909.0, + 1309.0, + 945.0, + 393.0, + 945.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 939.0, + 1310.0, + 939.0, + 1310.0, + 977.0, + 393.0, + 977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 971.0, + 1307.0, + 971.0, + 1307.0, + 1005.0, + 392.0, + 1005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 999.0, + 1308.0, + 999.0, + 1308.0, + 1037.0, + 392.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1031.0, + 1307.0, + 1031.0, + 1307.0, + 1065.0, + 392.0, + 1065.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1062.0, + 1306.0, + 1062.0, + 1306.0, + 1096.0, + 393.0, + 1096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1090.0, + 1306.0, + 1090.0, + 1306.0, + 1127.0, + 393.0, + 1127.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1122.0, + 1304.0, + 1122.0, + 1304.0, + 1153.0, + 394.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1153.0, + 1304.0, + 1153.0, + 1304.0, + 1185.0, + 394.0, + 1185.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1184.0, + 1305.0, + 1184.0, + 1305.0, + 1216.0, + 393.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 1215.0, + 1305.0, + 1215.0, + 1305.0, + 1246.0, + 395.0, + 1246.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1243.0, + 1304.0, + 1243.0, + 1304.0, + 1278.0, + 394.0, + 1278.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1275.0, + 1305.0, + 1275.0, + 1305.0, + 1306.0, + 394.0, + 1306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1304.0, + 734.0, + 1304.0, + 734.0, + 1338.0, + 393.0, + 1338.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1460.0, + 1408.0, + 1460.0, + 1408.0, + 1494.0, + 294.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1492.0, + 1406.0, + 1492.0, + 1406.0, + 1526.0, + 294.0, + 1526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1520.0, + 1404.0, + 1520.0, + 1404.0, + 1557.0, + 293.0, + 1557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1551.0, + 1406.0, + 1551.0, + 1406.0, + 1589.0, + 293.0, + 1589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1581.0, + 1407.0, + 1581.0, + 1407.0, + 1618.0, + 292.0, + 1618.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1612.0, + 1404.0, + 1612.0, + 1404.0, + 1646.0, + 294.0, + 1646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1641.0, + 1278.0, + 1641.0, + 1278.0, + 1678.0, + 293.0, + 1678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1688.0, + 1405.0, + 1688.0, + 1405.0, + 1725.0, + 293.0, + 1725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1718.0, + 1409.0, + 1718.0, + 1409.0, + 1755.0, + 293.0, + 1755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1749.0, + 1405.0, + 1749.0, + 1405.0, + 1782.0, + 294.0, + 1782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1780.0, + 1407.0, + 1780.0, + 1407.0, + 1813.0, + 292.0, + 1813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1810.0, + 1407.0, + 1810.0, + 1407.0, + 1844.0, + 292.0, + 1844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1839.0, + 1406.0, + 1839.0, + 1406.0, + 1874.0, + 292.0, + 1874.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1871.0, + 1405.0, + 1871.0, + 1405.0, + 1905.0, + 294.0, + 1905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1902.0, + 1076.0, + 1902.0, + 1076.0, + 1933.0, + 293.0, + 1933.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 798.0, + 634.0, + 902.0, + 634.0, + 902.0, + 671.0, + 798.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 685.0, + 666.0, + 1016.0, + 666.0, + 1016.0, + 702.0, + 685.0, + 702.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 711.0, + 700.0, + 985.0, + 700.0, + 985.0, + 733.0, + 711.0, + 733.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 987.0, + 496.0, + 1194.0, + 496.0, + 1194.0, + 535.0, + 987.0, + 535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 927.0, + 528.0, + 1258.0, + 528.0, + 1258.0, + 565.0, + 927.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 943.0, + 562.0, + 1241.0, + 562.0, + 1241.0, + 595.0, + 943.0, + 595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1147.0, + 634.0, + 1280.0, + 634.0, + 1280.0, + 674.0, + 1147.0, + 674.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1050.0, + 669.0, + 1379.0, + 669.0, + 1379.0, + 701.0, + 1050.0, + 701.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1064.0, + 700.0, + 1364.0, + 700.0, + 1364.0, + 736.0, + 1064.0, + 736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 400.0, + 636.0, + 571.0, + 636.0, + 571.0, + 674.0, + 400.0, + 674.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 669.0, + 652.0, + 669.0, + 652.0, + 700.0, + 321.0, + 700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 342.0, + 698.0, + 627.0, + 698.0, + 627.0, + 732.0, + 342.0, + 732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 529.0, + 499.0, + 695.0, + 499.0, + 695.0, + 532.0, + 529.0, + 532.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 442.0, + 530.0, + 772.0, + 530.0, + 772.0, + 564.0, + 442.0, + 564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 472.0, + 561.0, + 741.0, + 561.0, + 741.0, + 592.0, + 472.0, + 592.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 0, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1384, + 1405, + 1384, + 1405, + 1747, + 297, + 1747 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 1065, + 1404, + 1065, + 1404, + 1370, + 298, + 1370 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 298, + 1763, + 1405, + 1763, + 1405, + 2007, + 298, + 2007 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 298, + 656, + 1405, + 656, + 1405, + 901, + 298, + 901 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 299, + 460, + 1403, + 460, + 1403, + 643, + 299, + 643 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 202, + 1404, + 202, + 1404, + 447, + 298, + 447 + ], + "score": 0.979 + }, + { + "category_id": 0, + "poly": [ + 299, + 977, + 555, + 977, + 555, + 1014, + 299, + 1014 + ], + "score": 0.916 + }, + { + "category_id": 2, + "poly": [ + 841, + 2062, + 858, + 2062, + 858, + 2084, + 841, + 2084 + ], + "score": 0.715 + }, + { + "category_id": 2, + "poly": [ + 841, + 2062, + 859, + 2062, + 859, + 2084, + 841, + 2084 + ], + "score": 0.14 + }, + { + "category_id": 15, + "poly": [ + 290.0, + 973.0, + 559.0, + 973.0, + 559.0, + 1020.0, + 290.0, + 1020.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2059.0, + 862.0, + 2059.0, + 862.0, + 2093.0, + 839.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2059.0, + 862.0, + 2059.0, + 862.0, + 2093.0, + 838.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1383.0, + 1407.0, + 1383.0, + 1407.0, + 1418.0, + 295.0, + 1418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1413.0, + 1406.0, + 1413.0, + 1406.0, + 1450.0, + 294.0, + 1450.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1442.0, + 1406.0, + 1442.0, + 1406.0, + 1481.0, + 294.0, + 1481.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1468.0, + 1407.0, + 1468.0, + 1407.0, + 1517.0, + 291.0, + 1517.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1505.0, + 1406.0, + 1505.0, + 1406.0, + 1542.0, + 295.0, + 1542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1535.0, + 1406.0, + 1535.0, + 1406.0, + 1572.0, + 295.0, + 1572.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1566.0, + 1406.0, + 1566.0, + 1406.0, + 1602.0, + 292.0, + 1602.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1596.0, + 1408.0, + 1596.0, + 1408.0, + 1633.0, + 295.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1623.0, + 1405.0, + 1623.0, + 1405.0, + 1663.0, + 294.0, + 1663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1658.0, + 1405.0, + 1658.0, + 1405.0, + 1693.0, + 294.0, + 1693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1688.0, + 1405.0, + 1688.0, + 1405.0, + 1722.0, + 294.0, + 1722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1719.0, + 847.0, + 1719.0, + 847.0, + 1751.0, + 294.0, + 1751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1066.0, + 1406.0, + 1066.0, + 1406.0, + 1101.0, + 294.0, + 1101.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1096.0, + 1405.0, + 1096.0, + 1405.0, + 1131.0, + 294.0, + 1131.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1128.0, + 1404.0, + 1128.0, + 1404.0, + 1159.0, + 296.0, + 1159.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1157.0, + 1405.0, + 1157.0, + 1405.0, + 1192.0, + 294.0, + 1192.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1188.0, + 1405.0, + 1188.0, + 1405.0, + 1220.0, + 296.0, + 1220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1219.0, + 1405.0, + 1219.0, + 1405.0, + 1250.0, + 296.0, + 1250.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1249.0, + 1405.0, + 1249.0, + 1405.0, + 1281.0, + 296.0, + 1281.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1280.0, + 1405.0, + 1280.0, + 1405.0, + 1311.0, + 296.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1305.0, + 1406.0, + 1305.0, + 1406.0, + 1347.0, + 293.0, + 1347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1339.0, + 743.0, + 1339.0, + 743.0, + 1374.0, + 293.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1763.0, + 1405.0, + 1763.0, + 1405.0, + 1797.0, + 295.0, + 1797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1791.0, + 1405.0, + 1791.0, + 1405.0, + 1828.0, + 293.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1824.0, + 1407.0, + 1824.0, + 1407.0, + 1858.0, + 296.0, + 1858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1853.0, + 1405.0, + 1853.0, + 1405.0, + 1889.0, + 293.0, + 1889.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1883.0, + 1406.0, + 1883.0, + 1406.0, + 1920.0, + 295.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1911.0, + 1405.0, + 1911.0, + 1405.0, + 1950.0, + 292.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1945.0, + 1406.0, + 1945.0, + 1406.0, + 1982.0, + 293.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1975.0, + 1388.0, + 1975.0, + 1388.0, + 2010.0, + 293.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 658.0, + 1407.0, + 658.0, + 1407.0, + 691.0, + 296.0, + 691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 687.0, + 1406.0, + 687.0, + 1406.0, + 725.0, + 295.0, + 725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 720.0, + 1407.0, + 720.0, + 1407.0, + 753.0, + 295.0, + 753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 747.0, + 1405.0, + 747.0, + 1405.0, + 782.0, + 293.0, + 782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 780.0, + 1405.0, + 780.0, + 1405.0, + 813.0, + 296.0, + 813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 810.0, + 1403.0, + 810.0, + 1403.0, + 843.0, + 296.0, + 843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 840.0, + 1405.0, + 840.0, + 1405.0, + 877.0, + 293.0, + 877.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 871.0, + 501.0, + 871.0, + 501.0, + 902.0, + 293.0, + 902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 458.0, + 1405.0, + 458.0, + 1405.0, + 495.0, + 294.0, + 495.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 491.0, + 1405.0, + 491.0, + 1405.0, + 525.0, + 292.0, + 525.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 521.0, + 1405.0, + 521.0, + 1405.0, + 557.0, + 294.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 552.0, + 1404.0, + 552.0, + 1404.0, + 587.0, + 294.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 581.0, + 1405.0, + 581.0, + 1405.0, + 618.0, + 293.0, + 618.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 612.0, + 704.0, + 612.0, + 704.0, + 646.0, + 295.0, + 646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 204.0, + 1405.0, + 204.0, + 1405.0, + 237.0, + 294.0, + 237.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 235.0, + 1408.0, + 235.0, + 1408.0, + 268.0, + 296.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 263.0, + 1402.0, + 263.0, + 1402.0, + 297.0, + 296.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 295.0, + 1405.0, + 295.0, + 1405.0, + 329.0, + 292.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 326.0, + 1405.0, + 326.0, + 1405.0, + 357.0, + 294.0, + 357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 356.0, + 1405.0, + 356.0, + 1405.0, + 389.0, + 294.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 386.0, + 1404.0, + 386.0, + 1404.0, + 419.0, + 294.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 413.0, + 1237.0, + 413.0, + 1237.0, + 453.0, + 293.0, + 453.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 1, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 296, + 757, + 1405, + 757, + 1405, + 974, + 296, + 974 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 297, + 578, + 1405, + 578, + 1405, + 744, + 297, + 744 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 296, + 1430, + 1405, + 1430, + 1405, + 1555, + 296, + 1555 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 296, + 1129, + 1406, + 1129, + 1406, + 1259, + 296, + 1259 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 296, + 985, + 1405, + 985, + 1405, + 1114, + 296, + 1114 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 297, + 265, + 1405, + 265, + 1405, + 361, + 297, + 361 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 293, + 447, + 1403, + 447, + 1403, + 516, + 293, + 516 + ], + "score": 0.953 + }, + { + "category_id": 1, + "poly": [ + 295, + 1787, + 1404, + 1787, + 1404, + 1852, + 295, + 1852 + ], + "score": 0.946 + }, + { + "category_id": 1, + "poly": [ + 293, + 1655, + 1404, + 1655, + 1404, + 1720, + 293, + 1720 + ], + "score": 0.945 + }, + { + "category_id": 8, + "poly": [ + 434, + 1566, + 1263, + 1566, + 1263, + 1644, + 434, + 1644 + ], + "score": 0.944 + }, + { + "category_id": 8, + "poly": [ + 662, + 1270, + 1035, + 1270, + 1035, + 1329, + 662, + 1329 + ], + "score": 0.944 + }, + { + "category_id": 1, + "poly": [ + 294, + 1858, + 1401, + 1858, + 1401, + 1922, + 294, + 1922 + ], + "score": 0.943 + }, + { + "category_id": 8, + "poly": [ + 634, + 1732, + 1065, + 1732, + 1065, + 1777, + 634, + 1777 + ], + "score": 0.94 + }, + { + "category_id": 8, + "poly": [ + 543, + 1383, + 1154, + 1383, + 1154, + 1422, + 543, + 1422 + ], + "score": 0.94 + }, + { + "category_id": 8, + "poly": [ + 475, + 528, + 1219, + 528, + 1219, + 569, + 475, + 569 + ], + "score": 0.939 + }, + { + "category_id": 0, + "poly": [ + 295, + 392, + 1047, + 392, + 1047, + 427, + 295, + 427 + ], + "score": 0.936 + }, + { + "category_id": 1, + "poly": [ + 297, + 1340, + 1063, + 1340, + 1063, + 1374, + 297, + 1374 + ], + "score": 0.927 + }, + { + "category_id": 0, + "poly": [ + 297, + 198, + 463, + 198, + 463, + 235, + 297, + 235 + ], + "score": 0.916 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1387, + 1400, + 1387, + 1400, + 1418, + 1366, + 1418 + ], + "score": 0.897 + }, + { + "category_id": 9, + "poly": [ + 1365, + 1591, + 1400, + 1591, + 1400, + 1620, + 1365, + 1620 + ], + "score": 0.889 + }, + { + "category_id": 9, + "poly": [ + 1365, + 1952, + 1401, + 1952, + 1401, + 1983, + 1365, + 1983 + ], + "score": 0.866 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.667 + }, + { + "category_id": 8, + "poly": [ + 682, + 1932, + 1251, + 1932, + 1251, + 2004, + 682, + 2004 + ], + "score": 0.634 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.549 + }, + { + "category_id": 8, + "poly": [ + 680, + 1932, + 1251, + 1932, + 1251, + 2004, + 680, + 2004 + ], + "score": 0.349 + }, + { + "category_id": 13, + "poly": [ + 1145, + 446, + 1263, + 446, + 1263, + 484, + 1145, + 484 + ], + "score": 0.93, + "latex": "\\{ x ^ { ( k ) } \\} _ { k = 1 } ^ { n }" + }, + { + "category_id": 13, + "poly": [ + 298, + 1821, + 384, + 1821, + 384, + 1852, + 298, + 1852 + ], + "score": 0.93, + "latex": "x = Z _ { 1 }" + }, + { + "category_id": 14, + "poly": [ + 660, + 1269, + 1037, + 1269, + 1037, + 1330, + 660, + 1330 + ], + "score": 0.93, + "latex": "\\operatorname* { m i n } _ { \\theta } \\left\\{ { \\mathcal { L } } ( \\theta ) : = K { \\mathcal { L } } ( \\mathbb { Q } ^ { \\Pi ^ { * } } \\mid | \\mathbb { P } ^ { \\theta } ) \\right\\} ." + }, + { + "category_id": 14, + "poly": [ + 633, + 1730, + 1065, + 1730, + 1065, + 1776, + 633, + 1776 + ], + "score": 0.93, + "latex": "s _ { t } ^ { \\theta ^ { * } } ( z ) = \\mathbb { E } _ { Z \\sim \\mathbb { Q } ^ { \\mathrm { { n } ^ { * } } } } [ b _ { t } ( z | Z _ { 1 } ) \\mid Z _ { t } = z ] ." + }, + { + "category_id": 14, + "poly": [ + 435, + 1563, + 1262, + 1563, + 1262, + 1644, + 435, + 1644 + ], + "score": 0.93, + "latex": "\\mathcal { L } ( \\theta ) = \\mathbb { E } _ { Z \\sim \\mathbb { Q } ^ { \\mathrm { { n } ^ { * } } } } \\left[ \\frac { 1 } { 2 } \\int _ { 0 } ^ { 1 } \\left\\| \\sigma ( Z _ { t } ) ^ { - 1 } ( s _ { t } ^ { \\theta } ( Z _ { t } ) - b _ { t } ( Z _ { t } \\mid Z _ { 1 } ) ) \\right\\| _ { 2 } ^ { 2 } \\mathrm { d } t \\right] + \\mathrm { c o n s t } ," + }, + { + "category_id": 13, + "poly": [ + 373, + 1431, + 484, + 1431, + 484, + 1465, + 373, + 1465 + ], + "score": 0.93, + "latex": "b _ { t } ( Z _ { t } \\mid x )" + }, + { + "category_id": 13, + "poly": [ + 1262, + 1790, + 1344, + 1790, + 1344, + 1823, + 1262, + 1823 + ], + "score": 0.92, + "latex": "b _ { t } ( z | x )" + }, + { + "category_id": 13, + "poly": [ + 435, + 610, + 599, + 610, + 599, + 646, + 435, + 646 + ], + "score": 0.92, + "latex": "s _ { t } ^ { \\theta } \\colon { \\mathbb { R } ^ { d } } \\to { \\mathbb { R } ^ { d } }" + }, + { + "category_id": 13, + "poly": [ + 965, + 882, + 1065, + 882, + 1065, + 913, + 965, + 913 + ], + "score": 0.92, + "latex": "Z \\sim \\mathbb { Q } ^ { x }" + }, + { + "category_id": 13, + "poly": [ + 1162, + 645, + 1398, + 645, + 1398, + 679, + 1162, + 679 + ], + "score": 0.92, + "latex": "Z = \\{ Z _ { t } \\colon t \\in [ 0 , 1 ] \\}" + }, + { + "category_id": 13, + "poly": [ + 504, + 1129, + 555, + 1129, + 555, + 1164, + 504, + 1164 + ], + "score": 0.92, + "latex": "\\mathbb { Q } ^ { \\Pi ^ { * } }" + }, + { + "category_id": 13, + "poly": [ + 1274, + 1129, + 1397, + 1129, + 1397, + 1164, + 1274, + 1164 + ], + "score": 0.92, + "latex": "\\mathbb { Q } _ { 1 } ^ { \\Pi ^ { * } } = \\Pi ^ { * }" + }, + { + "category_id": 13, + "poly": [ + 438, + 987, + 631, + 987, + 631, + 1021, + 438, + 1021 + ], + "score": 0.92, + "latex": "\\mathbb { Q } ^ { x } ( Z _ { 1 } = x ) = 1" + }, + { + "category_id": 13, + "poly": [ + 576, + 913, + 663, + 913, + 663, + 942, + 576, + 942 + ], + "score": 0.92, + "latex": "Z _ { 1 } = x" + }, + { + "category_id": 13, + "poly": [ + 666, + 880, + 756, + 880, + 756, + 910, + 666, + 910 + ], + "score": 0.91, + "latex": "\\bar { \\boldsymbol { x } } \\in \\mathbb { R } ^ { d }" + }, + { + "category_id": 13, + "poly": [ + 801, + 578, + 1000, + 578, + 1000, + 613, + 801, + 613 + ], + "score": 0.91, + "latex": "\\sigma _ { t } \\colon \\mathbb { R } ^ { d } \\to \\mathbb { R } ^ { d \\times d }" + }, + { + "category_id": 13, + "poly": [ + 561, + 1822, + 642, + 1822, + 642, + 1852, + 561, + 1852 + ], + "score": 0.91, + "latex": "Z _ { t } = z" + }, + { + "category_id": 13, + "poly": [ + 1277, + 1691, + 1332, + 1691, + 1332, + 1721, + 1277, + 1721 + ], + "score": 0.91, + "latex": "\\mathcal { L } ( \\boldsymbol { \\theta } )" + }, + { + "category_id": 13, + "poly": [ + 1287, + 1165, + 1340, + 1165, + 1340, + 1199, + 1287, + 1199 + ], + "score": 0.91, + "latex": "\\mathbb { Q } ^ { \\Pi ^ { * } }" + }, + { + "category_id": 14, + "poly": [ + 445, + 1931, + 1250, + 1931, + 1250, + 2004, + 445, + 2004 + ], + "score": 0.91, + "latex": "\\mathbb { Q } ^ { x , \\mathrm { b b } } : \\mathrm { d } Z _ { t } = \\sigma _ { t } ^ { 2 } \\frac { x - Z _ { t } } { \\beta _ { 1 } - \\beta _ { t } } \\mathrm { d } t + \\sigma _ { t } \\mathrm { d } W _ { t } , Z _ { 0 } \\sim \\mathcal { N } ( x , \\beta _ { 1 } ) ," + }, + { + "category_id": 13, + "poly": [ + 1294, + 1018, + 1400, + 1018, + 1400, + 1049, + 1294, + 1049 + ], + "score": 0.91, + "latex": "Z \\sim \\mathbb { Q } ^ { x }" + }, + { + "category_id": 13, + "poly": [ + 741, + 1493, + 797, + 1493, + 797, + 1526, + 741, + 1526 + ], + "score": 0.91, + "latex": "\\mathcal { L } ( \\boldsymbol { \\theta } )" + }, + { + "category_id": 13, + "poly": [ + 903, + 1019, + 1001, + 1019, + 1001, + 1046, + 903, + 1046 + ], + "score": 0.9, + "latex": "x \\sim \\Pi ^ { * }" + }, + { + "category_id": 13, + "poly": [ + 1281, + 1861, + 1319, + 1861, + 1319, + 1891, + 1281, + 1891 + ], + "score": 0.9, + "latex": "\\mathbb { Q } ^ { x }" + }, + { + "category_id": 13, + "poly": [ + 432, + 709, + 464, + 709, + 464, + 745, + 432, + 745 + ], + "score": 0.9, + "latex": "\\mathbb { P } _ { 1 } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 521, + 882, + 559, + 882, + 559, + 913, + 521, + 913 + ], + "score": 0.9, + "latex": "\\mathbb { Q } ^ { x }" + }, + { + "category_id": 13, + "poly": [ + 905, + 1049, + 943, + 1049, + 943, + 1079, + 905, + 1079 + ], + "score": 0.9, + "latex": "\\mathbb { Q } ^ { x }" + }, + { + "category_id": 13, + "poly": [ + 346, + 676, + 379, + 676, + 379, + 712, + 346, + 712 + ], + "score": 0.9, + "latex": "\\mathbb { P } _ { t } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 629, + 1787, + 656, + 1787, + 656, + 1823, + 629, + 1823 + ], + "score": 0.89, + "latex": "s _ { t } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 705, + 1342, + 742, + 1342, + 742, + 1373, + 705, + 1373 + ], + "score": 0.89, + "latex": "\\mathbb { Q } ^ { x }" + }, + { + "category_id": 13, + "poly": [ + 779, + 988, + 817, + 988, + 817, + 1019, + 779, + 1019 + ], + "score": 0.89, + "latex": "\\mathbb { Q } ^ { x }" + }, + { + "category_id": 13, + "poly": [ + 297, + 1079, + 571, + 1079, + 571, + 1117, + 297, + 1117 + ], + "score": 0.89, + "latex": "\\begin{array} { r } { \\mathbb { Q } ^ { \\Pi ^ { * } } : = \\int \\mathbb { Q } ^ { x } ( \\cdot ) \\Pi ^ { * } ( \\mathrm { d } x ) } \\end{array}" + }, + { + "category_id": 14, + "poly": [ + 547, + 1383, + 1151, + 1383, + 1151, + 1422, + 547, + 1422 + ], + "score": 0.89, + "latex": "\\begin{array} { r } { \\mathbb { Q } ^ { x } \\colon \\quad \\mathrm { d } Z _ { t } = b _ { t } ( Z _ { t } \\mid x ) \\mathrm { d } t + \\sigma _ { t } ( Z _ { t } ) \\mathrm { d } W _ { t } , \\quad Z _ { 0 } \\sim \\mu _ { 0 } , } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 375, + 583, + 415, + 583, + 415, + 613, + 375, + 613 + ], + "score": 0.89, + "latex": "W _ { t }" + }, + { + "category_id": 13, + "poly": [ + 695, + 712, + 729, + 712, + 729, + 742, + 695, + 742 + ], + "score": 0.89, + "latex": "Z _ { 1 }" + }, + { + "category_id": 14, + "poly": [ + 481, + 527, + 1216, + 527, + 1216, + 568, + 481, + 568 + ], + "score": 0.88, + "latex": "\\begin{array} { r } { \\mathbb { P } ^ { \\theta } \\colon \\quad \\mathrm { d } Z _ { t } = s _ { t } ^ { \\theta } ( Z _ { t } ) \\mathrm { d } t + \\sigma _ { t } ( Z _ { t } ) \\mathrm { d } W _ { t } , \\quad \\forall t \\in [ 0 , 1 ] , \\quad Z _ { 0 } \\sim \\mu _ { 0 } , } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 706, + 679, + 738, + 679, + 738, + 709, + 706, + 709 + ], + "score": 0.88, + "latex": "Z _ { t }" + }, + { + "category_id": 13, + "poly": [ + 745, + 1656, + 772, + 1656, + 772, + 1685, + 745, + 1685 + ], + "score": 0.88, + "latex": "s ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 1250, + 757, + 1283, + 757, + 1283, + 787, + 1250, + 787 + ], + "score": 0.88, + "latex": "\\mathbb { P } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 591, + 643, + 624, + 643, + 624, + 673, + 591, + 673 + ], + "score": 0.87, + "latex": "\\mathbb { P } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 649, + 481, + 684, + 481, + 684, + 511, + 649, + 511 + ], + "score": 0.87, + "latex": "\\mathbb { R } ^ { d }" + }, + { + "category_id": 13, + "poly": [ + 813, + 1165, + 846, + 1165, + 846, + 1195, + 813, + 1195 + ], + "score": 0.87, + "latex": "\\mathbb { P } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 574, + 483, + 610, + 483, + 610, + 511, + 574, + 511 + ], + "score": 0.87, + "latex": "\\Pi ^ { * }" + }, + { + "category_id": 13, + "poly": [ + 1039, + 712, + 1074, + 712, + 1074, + 739, + 1039, + 739 + ], + "score": 0.87, + "latex": "\\Pi ^ { * }" + }, + { + "category_id": 13, + "poly": [ + 971, + 1689, + 1000, + 1689, + 1000, + 1715, + 971, + 1715 + ], + "score": 0.86, + "latex": "\\theta ^ { * }" + }, + { + "category_id": 13, + "poly": [ + 1297, + 616, + 1329, + 616, + 1329, + 644, + 1297, + 644 + ], + "score": 0.85, + "latex": "\\mu _ { 0 }" + }, + { + "category_id": 13, + "poly": [ + 297, + 849, + 330, + 849, + 330, + 878, + 297, + 878 + ], + "score": 0.84, + "latex": "\\mathbb { P } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 705, + 1049, + 729, + 1049, + 729, + 1075, + 705, + 1075 + ], + "score": 0.83, + "latex": "Z" + }, + { + "category_id": 13, + "poly": [ + 1230, + 681, + 1247, + 681, + 1247, + 706, + 1230, + 706 + ], + "score": 0.8, + "latex": "\\theta" + }, + { + "category_id": 13, + "poly": [ + 409, + 1053, + 428, + 1053, + 428, + 1075, + 409, + 1075 + ], + "score": 0.8, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 333, + 1023, + 352, + 1023, + 352, + 1045, + 333, + 1045 + ], + "score": 0.79, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 1005, + 1052, + 1024, + 1052, + 1024, + 1075, + 1005, + 1075 + ], + "score": 0.78, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 1217, + 614, + 1234, + 614, + 1234, + 640, + 1217, + 640 + ], + "score": 0.77, + "latex": "\\theta" + }, + { + "category_id": 13, + "poly": [ + 825, + 1660, + 839, + 1660, + 839, + 1685, + 825, + 1685 + ], + "score": 0.76, + "latex": "b" + }, + { + "category_id": 13, + "poly": [ + 1231, + 791, + 1255, + 791, + 1255, + 818, + 1231, + 818 + ], + "score": 0.76, + "latex": "Z" + }, + { + "category_id": 13, + "poly": [ + 549, + 1438, + 568, + 1438, + 568, + 1459, + 549, + 1459 + ], + "score": 0.76, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 1261, + 994, + 1280, + 994, + 1280, + 1016, + 1261, + 1016 + ], + "score": 0.75, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 504, + 1696, + 534, + 1696, + 534, + 1720, + 504, + 1720 + ], + "score": 0.75, + "latex": "\\mu _ { 0 }" + }, + { + "category_id": 13, + "poly": [ + 474, + 917, + 494, + 917, + 494, + 939, + 474, + 939 + ], + "score": 0.73, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 830, + 681, + 844, + 681, + 844, + 706, + 830, + 706 + ], + "score": 0.68, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 1385, + 1866, + 1401, + 1866, + 1401, + 1887, + 1385, + 1887 + ], + "score": 0.59, + "latex": "a" + }, + { + "category_id": 13, + "poly": [ + 1089, + 484, + 1145, + 484, + 1145, + 517, + 1089, + 517 + ], + "score": 0.35, + "latex": "[ 0 , 1 ]" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 388.0, + 1050.0, + 388.0, + 1050.0, + 436.0, + 291.0, + 436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 193.0, + 467.0, + 193.0, + 467.0, + 241.0, + 290.0, + 241.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 755.0, + 1249.0, + 755.0, + 1249.0, + 794.0, + 294.0, + 794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1284.0, + 755.0, + 1406.0, + 755.0, + 1406.0, + 794.0, + 1284.0, + 794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 790.0, + 1230.0, + 790.0, + 1230.0, + 825.0, + 295.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1256.0, + 790.0, + 1406.0, + 790.0, + 1406.0, + 825.0, + 1256.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 820.0, + 1405.0, + 820.0, + 1405.0, + 854.0, + 293.0, + 854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 846.0, + 296.0, + 846.0, + 296.0, + 888.0, + 293.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 846.0, + 1405.0, + 846.0, + 1405.0, + 888.0, + 331.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 877.0, + 520.0, + 877.0, + 520.0, + 919.0, + 294.0, + 919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 560.0, + 877.0, + 665.0, + 877.0, + 665.0, + 919.0, + 560.0, + 919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 877.0, + 964.0, + 877.0, + 964.0, + 919.0, + 757.0, + 919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1066.0, + 877.0, + 1406.0, + 877.0, + 1406.0, + 919.0, + 1066.0, + 919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 911.0, + 473.0, + 911.0, + 473.0, + 946.0, + 294.0, + 946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 495.0, + 911.0, + 575.0, + 911.0, + 575.0, + 946.0, + 495.0, + 946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 664.0, + 911.0, + 1405.0, + 911.0, + 1405.0, + 946.0, + 664.0, + 946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 942.0, + 765.0, + 942.0, + 765.0, + 974.0, + 292.0, + 974.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 576.0, + 374.0, + 576.0, + 374.0, + 616.0, + 292.0, + 616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 416.0, + 576.0, + 800.0, + 576.0, + 800.0, + 616.0, + 416.0, + 616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1001.0, + 576.0, + 1408.0, + 576.0, + 1408.0, + 616.0, + 1001.0, + 616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 605.0, + 434.0, + 605.0, + 434.0, + 649.0, + 291.0, + 649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 605.0, + 1216.0, + 605.0, + 1216.0, + 649.0, + 600.0, + 649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1235.0, + 605.0, + 1296.0, + 605.0, + 1296.0, + 649.0, + 1235.0, + 649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1330.0, + 605.0, + 1407.0, + 605.0, + 1407.0, + 649.0, + 1330.0, + 649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 641.0, + 590.0, + 641.0, + 590.0, + 681.0, + 292.0, + 681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 625.0, + 641.0, + 1161.0, + 641.0, + 1161.0, + 681.0, + 625.0, + 681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1399.0, + 641.0, + 1406.0, + 641.0, + 1406.0, + 681.0, + 1399.0, + 681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 675.0, + 345.0, + 675.0, + 345.0, + 712.0, + 294.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 380.0, + 675.0, + 705.0, + 675.0, + 705.0, + 712.0, + 380.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 739.0, + 675.0, + 829.0, + 675.0, + 829.0, + 712.0, + 739.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 845.0, + 675.0, + 1229.0, + 675.0, + 1229.0, + 712.0, + 845.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 675.0, + 1406.0, + 675.0, + 1406.0, + 712.0, + 1248.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 708.0, + 431.0, + 708.0, + 431.0, + 745.0, + 295.0, + 745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 465.0, + 708.0, + 694.0, + 708.0, + 694.0, + 745.0, + 465.0, + 745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 730.0, + 708.0, + 1038.0, + 708.0, + 1038.0, + 745.0, + 730.0, + 745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1075.0, + 708.0, + 1086.0, + 708.0, + 1086.0, + 745.0, + 1075.0, + 745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1428.0, + 372.0, + 1428.0, + 372.0, + 1468.0, + 294.0, + 1468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 485.0, + 1428.0, + 548.0, + 1428.0, + 548.0, + 1468.0, + 485.0, + 1468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 569.0, + 1428.0, + 1406.0, + 1428.0, + 1406.0, + 1468.0, + 569.0, + 1468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1456.0, + 1407.0, + 1456.0, + 1407.0, + 1501.0, + 291.0, + 1501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1489.0, + 740.0, + 1489.0, + 740.0, + 1529.0, + 293.0, + 1529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 798.0, + 1489.0, + 1405.0, + 1489.0, + 1405.0, + 1529.0, + 798.0, + 1529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1518.0, + 576.0, + 1518.0, + 576.0, + 1558.0, + 293.0, + 1558.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1128.0, + 503.0, + 1128.0, + 503.0, + 1166.0, + 291.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 556.0, + 1128.0, + 1273.0, + 1128.0, + 1273.0, + 1166.0, + 556.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1398.0, + 1128.0, + 1411.0, + 1128.0, + 1411.0, + 1166.0, + 1398.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1163.0, + 812.0, + 1163.0, + 812.0, + 1202.0, + 290.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 847.0, + 1163.0, + 1286.0, + 1163.0, + 1286.0, + 1202.0, + 847.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1341.0, + 1163.0, + 1407.0, + 1163.0, + 1407.0, + 1202.0, + 1341.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1194.0, + 1406.0, + 1194.0, + 1406.0, + 1234.0, + 293.0, + 1234.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1229.0, + 651.0, + 1229.0, + 651.0, + 1262.0, + 295.0, + 1262.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 985.0, + 437.0, + 985.0, + 437.0, + 1022.0, + 295.0, + 1022.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 632.0, + 985.0, + 778.0, + 985.0, + 778.0, + 1022.0, + 632.0, + 1022.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 818.0, + 985.0, + 1260.0, + 985.0, + 1260.0, + 1022.0, + 818.0, + 1022.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1281.0, + 985.0, + 1405.0, + 985.0, + 1405.0, + 1022.0, + 1281.0, + 1022.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1015.0, + 332.0, + 1015.0, + 332.0, + 1052.0, + 293.0, + 1052.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1015.0, + 902.0, + 1015.0, + 902.0, + 1052.0, + 353.0, + 1052.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1002.0, + 1015.0, + 1293.0, + 1015.0, + 1293.0, + 1052.0, + 1002.0, + 1052.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1046.0, + 408.0, + 1046.0, + 408.0, + 1080.0, + 293.0, + 1080.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 429.0, + 1046.0, + 704.0, + 1046.0, + 704.0, + 1080.0, + 429.0, + 1080.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 730.0, + 1046.0, + 904.0, + 1046.0, + 904.0, + 1080.0, + 730.0, + 1080.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 944.0, + 1046.0, + 1004.0, + 1046.0, + 1004.0, + 1080.0, + 944.0, + 1080.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1025.0, + 1046.0, + 1410.0, + 1046.0, + 1410.0, + 1080.0, + 1025.0, + 1080.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1074.0, + 296.0, + 1074.0, + 296.0, + 1119.0, + 293.0, + 1119.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 572.0, + 1074.0, + 582.0, + 1074.0, + 582.0, + 1119.0, + 572.0, + 1119.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 264.0, + 1406.0, + 264.0, + 1406.0, + 302.0, + 294.0, + 302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 297.0, + 1405.0, + 297.0, + 1405.0, + 331.0, + 295.0, + 331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 326.0, + 1216.0, + 326.0, + 1216.0, + 364.0, + 294.0, + 364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 286.0, + 434.0, + 1144.0, + 434.0, + 1144.0, + 500.0, + 286.0, + 500.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1264.0, + 434.0, + 1404.0, + 434.0, + 1404.0, + 500.0, + 1264.0, + 500.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 479.0, + 573.0, + 479.0, + 573.0, + 518.0, + 293.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 611.0, + 479.0, + 648.0, + 479.0, + 648.0, + 518.0, + 611.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 685.0, + 479.0, + 1088.0, + 479.0, + 1088.0, + 518.0, + 685.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1146.0, + 479.0, + 1176.0, + 479.0, + 1176.0, + 518.0, + 1146.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1785.0, + 628.0, + 1785.0, + 628.0, + 1825.0, + 294.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 657.0, + 1785.0, + 1261.0, + 1785.0, + 1261.0, + 1825.0, + 657.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1345.0, + 1785.0, + 1405.0, + 1785.0, + 1405.0, + 1825.0, + 1345.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1819.0, + 297.0, + 1819.0, + 297.0, + 1856.0, + 294.0, + 1856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 385.0, + 1819.0, + 560.0, + 1819.0, + 560.0, + 1856.0, + 385.0, + 1856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 643.0, + 1819.0, + 656.0, + 1819.0, + 656.0, + 1856.0, + 643.0, + 1856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1655.0, + 744.0, + 1655.0, + 744.0, + 1691.0, + 295.0, + 1691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 773.0, + 1655.0, + 824.0, + 1655.0, + 824.0, + 1691.0, + 773.0, + 1691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 1655.0, + 1404.0, + 1655.0, + 1404.0, + 1691.0, + 840.0, + 1691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1686.0, + 503.0, + 1686.0, + 503.0, + 1722.0, + 295.0, + 1722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 535.0, + 1686.0, + 970.0, + 1686.0, + 970.0, + 1722.0, + 535.0, + 1722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1001.0, + 1686.0, + 1276.0, + 1686.0, + 1276.0, + 1722.0, + 1001.0, + 1722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1333.0, + 1686.0, + 1363.0, + 1686.0, + 1363.0, + 1722.0, + 1333.0, + 1722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1858.0, + 1280.0, + 1858.0, + 1280.0, + 1894.0, + 296.0, + 1894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1320.0, + 1858.0, + 1384.0, + 1858.0, + 1384.0, + 1894.0, + 1320.0, + 1894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 1858.0, + 1406.0, + 1858.0, + 1406.0, + 1894.0, + 1402.0, + 1894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1889.0, + 723.0, + 1889.0, + 723.0, + 1923.0, + 296.0, + 1923.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1339.0, + 704.0, + 1339.0, + 704.0, + 1376.0, + 295.0, + 1376.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 743.0, + 1339.0, + 1064.0, + 1339.0, + 1064.0, + 1376.0, + 743.0, + 1376.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 2, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 392, + 1406, + 392, + 1406, + 759, + 297, + 759 + ], + "score": 0.984 + }, + { + "category_id": 1, + "poly": [ + 298, + 849, + 1405, + 849, + 1405, + 1034, + 298, + 1034 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 1131, + 1406, + 1131, + 1406, + 1256, + 298, + 1256 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 296, + 1907, + 1404, + 1907, + 1404, + 2015, + 296, + 2015 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 297, + 197, + 1405, + 197, + 1405, + 339, + 297, + 339 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 296, + 1702, + 1406, + 1702, + 1406, + 1801, + 296, + 1801 + ], + "score": 0.969 + }, + { + "category_id": 1, + "poly": [ + 296, + 1428, + 1406, + 1428, + 1406, + 1526, + 296, + 1526 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 300, + 1326, + 1398, + 1326, + 1398, + 1420, + 300, + 1420 + ], + "score": 0.956 + }, + { + "category_id": 1, + "poly": [ + 297, + 1587, + 1404, + 1587, + 1404, + 1679, + 297, + 1679 + ], + "score": 0.952 + }, + { + "category_id": 1, + "poly": [ + 297, + 1042, + 1400, + 1042, + 1400, + 1111, + 297, + 1111 + ], + "score": 0.94 + }, + { + "category_id": 8, + "poly": [ + 403, + 1815, + 1289, + 1815, + 1289, + 1891, + 403, + 1891 + ], + "score": 0.936 + }, + { + "category_id": 8, + "poly": [ + 334, + 1272, + 1322, + 1272, + 1322, + 1313, + 334, + 1313 + ], + "score": 0.931 + }, + { + "category_id": 0, + "poly": [ + 301, + 793, + 795, + 793, + 795, + 827, + 301, + 827 + ], + "score": 0.922 + }, + { + "category_id": 1, + "poly": [ + 292, + 1538, + 1379, + 1538, + 1379, + 1578, + 292, + 1578 + ], + "score": 0.879 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1279, + 1400, + 1279, + 1400, + 1308, + 1366, + 1308 + ], + "score": 0.865 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1838, + 1400, + 1838, + 1400, + 1867, + 1366, + 1867 + ], + "score": 0.863 + }, + { + "category_id": 2, + "poly": [ + 840, + 2061, + 859, + 2061, + 859, + 2084, + 840, + 2084 + ], + "score": 0.78 + }, + { + "category_id": 13, + "poly": [ + 967, + 236, + 1155, + 236, + 1155, + 274, + 967, + 274 + ], + "score": 0.95, + "latex": "\\mathrm { d } \\tilde { Z } _ { t } = \\sigma _ { 1 - t } \\mathrm { d } \\tilde { W } _ { t }" + }, + { + "category_id": 13, + "poly": [ + 1199, + 1460, + 1398, + 1460, + 1398, + 1496, + 1199, + 1496 + ], + "score": 0.94, + "latex": "\\mathbb { Q } ^ { A } ( Z _ { 1 } \\in A ) = 1" + }, + { + "category_id": 13, + "poly": [ + 603, + 1973, + 703, + 1973, + 703, + 2016, + 603, + 2016 + ], + "score": 0.94, + "latex": "\\sigma _ { t } ^ { 2 } \\frac { x - Z _ { t } } { \\beta _ { 1 } - \\beta _ { t } }" + }, + { + "category_id": 13, + "poly": [ + 581, + 196, + 736, + 196, + 736, + 240, + 581, + 240 + ], + "score": 0.93, + "latex": "\\begin{array} { r } { \\beta _ { t } = \\int _ { 0 } ^ { t } \\sigma _ { s } ^ { 2 } \\mathrm { d } s } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 838, + 1702, + 1078, + 1702, + 1078, + 1741, + 838, + 1741 + ], + "score": 0.93, + "latex": "U _ { t } ( z ) = \\left\\| x - z \\right\\| ^ { 2 } / 2" + }, + { + "category_id": 13, + "poly": [ + 298, + 1628, + 512, + 1628, + 512, + 1669, + 298, + 1669 + ], + "score": 0.93, + "latex": "\\begin{array} { r } { \\zeta _ { t } = \\exp ( \\int _ { 0 } ^ { t } \\alpha _ { s } \\mathrm { d } s ) } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 298, + 1461, + 496, + 1461, + 496, + 1495, + 298, + 1495 + ], + "score": 0.93, + "latex": "U ( \\cdot , t ) \\in C ^ { 2 } ( \\mathbb R ^ { d } )" + }, + { + "category_id": 13, + "poly": [ + 884, + 1077, + 1005, + 1077, + 1005, + 1110, + 884, + 1110 + ], + "score": 0.93, + "latex": "U _ { 1 } ( z ) = 0" + }, + { + "category_id": 13, + "poly": [ + 886, + 1133, + 1070, + 1133, + 1070, + 1167, + 886, + 1167 + ], + "score": 0.93, + "latex": "\\mathbb { Q } ( Z _ { 1 } \\in A ) = 1" + }, + { + "category_id": 13, + "poly": [ + 1018, + 1195, + 1104, + 1195, + 1104, + 1225, + 1018, + 1225 + ], + "score": 0.93, + "latex": "Z _ { t } \\in A" + }, + { + "category_id": 13, + "poly": [ + 1134, + 880, + 1330, + 880, + 1330, + 914, + 1134, + 914 + ], + "score": 0.93, + "latex": "\\mathbb { Q } ^ { x } ( Z _ { 1 } = x ) = 1" + }, + { + "category_id": 14, + "poly": [ + 405, + 1813, + 1293, + 1813, + 1293, + 1891, + 405, + 1891 + ], + "score": 0.92, + "latex": "\\mathbb { Q } ^ { x , \\mathrm { b b } , f } : \\mathrm { ~ } \\mathrm { ~ } \\mathrm { ~ } \\mathrm { ~ d ~ } Z _ { t } = \\left( \\sigma _ { t } f _ { t } ( Z _ { t } ) + \\sigma _ { t } ^ { 2 } \\frac { x - Z _ { t } } { \\beta _ { 1 } - \\beta _ { t } } \\right) \\mathrm { d } t + \\sigma _ { t } \\mathrm { d } W _ { t } , \\quad Z _ { 0 } \\sim \\mu _ { 0 } ." + }, + { + "category_id": 13, + "poly": [ + 1058, + 1906, + 1352, + 1906, + 1352, + 1946, + 1058, + 1946 + ], + "score": 0.92, + "latex": "\\mathbb { E } _ { \\mathbb { Q } ^ { x , \\mathrm { b b } } } [ \\left. f _ { t } ( Z _ { t } ) \\right. ^ { 2 } ] < + \\infty" + }, + { + "category_id": 13, + "poly": [ + 695, + 696, + 901, + 696, + 901, + 730, + 695, + 730 + ], + "score": 0.92, + "latex": "\\bar { \\mathbb { Q } ^ { x } } ( Z _ { 1 } = x ) = 1" + }, + { + "category_id": 13, + "poly": [ + 1129, + 1705, + 1348, + 1705, + 1348, + 1741, + 1129, + 1741 + ], + "score": 0.92, + "latex": "\\alpha _ { t } = \\sigma _ { t } ^ { 2 } / ( \\beta _ { 1 } - \\beta _ { t } )" + }, + { + "category_id": 13, + "poly": [ + 579, + 1428, + 762, + 1428, + 762, + 1462, + 579, + 1462 + ], + "score": 0.92, + "latex": "U _ { t } ( z ) = U ( z , t )" + }, + { + "category_id": 13, + "poly": [ + 378, + 1590, + 729, + 1590, + 729, + 1628, + 378, + 1628 + ], + "score": 0.92, + "latex": "\\beta _ { t } = \\mathbb { E } _ { \\mathbb { Q } ^ { A } } [ \\nabla _ { z } U _ { t } ( Z _ { t } ) ^ { \\top } \\nu _ { t } ( Z _ { t } ) ]" + }, + { + "category_id": 13, + "poly": [ + 371, + 1329, + 452, + 1329, + 452, + 1358, + 371, + 1358 + ], + "score": 0.92, + "latex": "\\alpha _ { t } > 0" + }, + { + "category_id": 13, + "poly": [ + 372, + 203, + 527, + 203, + 527, + 237, + 372, + 237 + ], + "score": 0.92, + "latex": "\\sigma _ { t } \\in [ 0 , + \\infty )" + }, + { + "category_id": 13, + "poly": [ + 298, + 1944, + 414, + 1944, + 414, + 1974, + 298, + 1974 + ], + "score": 0.91, + "latex": "\\sigma _ { t } > 0 , \\forall t" + }, + { + "category_id": 13, + "poly": [ + 355, + 1388, + 435, + 1388, + 435, + 1419, + 355, + 1419 + ], + "score": 0.91, + "latex": "\\sigma _ { t } \\mathrm { d } W _ { t }" + }, + { + "category_id": 13, + "poly": [ + 940, + 1536, + 1378, + 1536, + 1378, + 1578, + 940, + 1578 + ], + "score": 0.91, + "latex": "\\mathbb { E } _ { \\mathbb { Q } ^ { A } } [ U _ { t } ( Z _ { t } ) ] - \\| \\nabla _ { z } U _ { t } ( Z _ { t } ) \\| ^ { 2 } ] \\leq 0 , \\forall t ." + }, + { + "category_id": 13, + "poly": [ + 876, + 1043, + 942, + 1043, + 942, + 1076, + 876, + 1076 + ], + "score": 0.91, + "latex": "U _ { t } ( z )" + }, + { + "category_id": 13, + "poly": [ + 298, + 237, + 428, + 237, + 428, + 272, + 298, + 272 + ], + "score": 0.91, + "latex": "\\tilde { Z } _ { t } : = Z _ { 1 - t }" + }, + { + "category_id": 13, + "poly": [ + 365, + 273, + 459, + 273, + 459, + 307, + 365, + 307 + ], + "score": 0.91, + "latex": "\\tilde { Z } _ { 0 } = x" + }, + { + "category_id": 13, + "poly": [ + 298, + 308, + 384, + 308, + 384, + 337, + 298, + 337 + ], + "score": 0.91, + "latex": "Z _ { 1 } = x" + }, + { + "category_id": 13, + "poly": [ + 547, + 1462, + 770, + 1462, + 770, + 1496, + 547, + 1496 + ], + "score": 0.91, + "latex": "U ( z , \\cdot ) \\in C ^ { 1 } ( [ 0 , 1 ] )" + }, + { + "category_id": 13, + "poly": [ + 569, + 1075, + 689, + 1075, + 689, + 1110, + 569, + 1110 + ], + "score": 0.91, + "latex": "U _ { 1 } ( z ) \\geq 0" + }, + { + "category_id": 13, + "poly": [ + 550, + 271, + 589, + 271, + 589, + 307, + 550, + 307 + ], + "score": 0.91, + "latex": "\\tilde { W } _ { t }" + }, + { + "category_id": 13, + "poly": [ + 298, + 1074, + 391, + 1074, + 391, + 1105, + 298, + 1105 + ], + "score": 0.9, + "latex": "A \\subset { \\mathbb { R } } ^ { d }" + }, + { + "category_id": 13, + "poly": [ + 669, + 1389, + 720, + 1389, + 720, + 1419, + 669, + 1419 + ], + "score": 0.9, + "latex": "\\nu _ { t } \\mathrm { d } t" + }, + { + "category_id": 13, + "poly": [ + 713, + 1359, + 785, + 1359, + 785, + 1386, + 713, + 1386 + ], + "score": 0.9, + "latex": "t \\to 1" + }, + { + "category_id": 13, + "poly": [ + 760, + 1910, + 846, + 1910, + 846, + 1943, + 760, + 1943 + ], + "score": 0.9, + "latex": "\\mathbb { Q } ^ { x , \\mathrm { b b } , f }" + }, + { + "category_id": 13, + "poly": [ + 730, + 1075, + 831, + 1075, + 831, + 1108, + 730, + 1108 + ], + "score": 0.89, + "latex": "\\forall z \\in \\mathbb { R } ^ { d }" + }, + { + "category_id": 13, + "poly": [ + 1201, + 1389, + 1264, + 1389, + 1264, + 1416, + 1201, + 1416 + ], + "score": 0.89, + "latex": "t = 1" + }, + { + "category_id": 13, + "poly": [ + 795, + 1590, + 1341, + 1590, + 1341, + 1629, + 795, + 1629 + ], + "score": 0.89, + "latex": "\\begin{array} { r } { \\gamma _ { t } = \\mathbb { E } _ { \\mathbb { Q } ^ { A } } [ \\partial _ { t } U _ { t } ( Z _ { t } ) + \\frac { 1 } { 2 } \\mathrm { t r } ( \\nabla _ { z } ^ { 2 } U _ { t } ( Z _ { t } ) \\sigma _ { t } ^ { 2 } ( Z _ { t } ) ) ] , } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 835, + 1629, + 1156, + 1629, + 1156, + 1679, + 835, + 1679 + ], + "score": 0.89, + "latex": "\\begin{array} { r } { \\operatorname* { l i m } _ { t \\uparrow 1 } \\frac { \\zeta _ { t } } { \\int _ { 0 } ^ { t } \\zeta _ { s } ( \\beta _ { s } + \\gamma _ { s } ) \\mathrm { d } s } = + \\infty } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 677, + 882, + 715, + 882, + 715, + 912, + 677, + 912 + ], + "score": 0.89, + "latex": "\\mathbb { Q } ^ { x }" + }, + { + "category_id": 13, + "poly": [ + 848, + 1460, + 890, + 1460, + 890, + 1494, + 848, + 1494 + ], + "score": 0.89, + "latex": "\\mathbb { Q } ^ { A }" + }, + { + "category_id": 13, + "poly": [ + 1334, + 1195, + 1398, + 1195, + 1398, + 1221, + 1334, + 1221 + ], + "score": 0.89, + "latex": "t = 1" + }, + { + "category_id": 13, + "poly": [ + 478, + 1077, + 543, + 1077, + 543, + 1107, + 478, + 1107 + ], + "score": 0.88, + "latex": "t = 1" + }, + { + "category_id": 13, + "poly": [ + 298, + 1739, + 367, + 1739, + 367, + 1766, + 298, + 1766 + ], + "score": 0.88, + "latex": "\\nu = 0" + }, + { + "category_id": 14, + "poly": [ + 344, + 1273, + 1320, + 1273, + 1320, + 1312, + 344, + 1312 + ], + "score": 0.87, + "latex": "\\mathbb { Q } ^ { A } : \\quad \\mathrm { d } Z _ { t } = ( - \\alpha _ { t } \\nabla _ { z } U _ { t } ( Z _ { t } ) + \\nu _ { t } ( Z _ { t } ) ) \\mathrm { d } t + \\sigma _ { t } ( Z _ { t } ) \\mathrm { d } W _ { t } , \\qquad t \\in [ 0 , 1 ] , ~ Z _ { 0 } \\sim \\mu _ { 0 } ," + }, + { + "category_id": 13, + "poly": [ + 647, + 1135, + 673, + 1135, + 673, + 1165, + 647, + 1165 + ], + "score": 0.86, + "latex": "\\mathbb { Q }" + }, + { + "category_id": 13, + "poly": [ + 349, + 1362, + 379, + 1362, + 379, + 1388, + 349, + 1388 + ], + "score": 0.86, + "latex": "\\alpha _ { t }" + }, + { + "category_id": 13, + "poly": [ + 589, + 1632, + 775, + 1632, + 775, + 1667, + 589, + 1667 + ], + "score": 0.84, + "latex": "\\operatorname* { l i m } _ { t \\uparrow 1 } \\zeta _ { t } = + \\infty" + }, + { + "category_id": 13, + "poly": [ + 877, + 1329, + 901, + 1329, + 901, + 1355, + 877, + 1355 + ], + "score": 0.84, + "latex": "U" + }, + { + "category_id": 13, + "poly": [ + 938, + 1389, + 963, + 1389, + 963, + 1415, + 938, + 1415 + ], + "score": 0.83, + "latex": "U" + }, + { + "category_id": 13, + "poly": [ + 1156, + 1077, + 1226, + 1077, + 1226, + 1107, + 1156, + 1107 + ], + "score": 0.82, + "latex": "z \\in A" + }, + { + "category_id": 13, + "poly": [ + 867, + 1226, + 891, + 1226, + 891, + 1252, + 867, + 1252 + ], + "score": 0.82, + "latex": "A" + }, + { + "category_id": 13, + "poly": [ + 579, + 577, + 598, + 577, + 598, + 603, + 579, + 603 + ], + "score": 0.81, + "latex": "h" + }, + { + "category_id": 13, + "poly": [ + 1041, + 577, + 1060, + 577, + 1060, + 603, + 1041, + 603 + ], + "score": 0.81, + "latex": "h" + }, + { + "category_id": 13, + "poly": [ + 952, + 1333, + 971, + 1333, + 971, + 1355, + 952, + 1355 + ], + "score": 0.79, + "latex": "\\nu" + }, + { + "category_id": 13, + "poly": [ + 801, + 1135, + 825, + 1135, + 825, + 1162, + 801, + 1162 + ], + "score": 0.76, + "latex": "A" + }, + { + "category_id": 13, + "poly": [ + 1119, + 1464, + 1142, + 1464, + 1142, + 1490, + 1119, + 1490 + ], + "score": 0.74, + "latex": "A" + }, + { + "category_id": 13, + "poly": [ + 1124, + 1978, + 1140, + 1978, + 1140, + 2003, + 1124, + 2003 + ], + "score": 0.74, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 415, + 1775, + 433, + 1775, + 433, + 1796, + 415, + 1796 + ], + "score": 0.74, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 1231, + 1431, + 1255, + 1431, + 1255, + 1457, + 1231, + 1457 + ], + "score": 0.71, + "latex": "A" + }, + { + "category_id": 13, + "poly": [ + 1009, + 1917, + 1027, + 1917, + 1027, + 1939, + 1009, + 1939 + ], + "score": 0.7, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 1161, + 310, + 1175, + 310, + 1175, + 334, + 1161, + 334 + ], + "score": 0.44, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 325, + 1544, + 350, + 1544, + 350, + 1571, + 325, + 1571 + ], + "score": 0.4, + "latex": "U" + }, + { + "category_id": 13, + "poly": [ + 295, + 1542, + 350, + 1542, + 350, + 1575, + 295, + 1575 + ], + "score": 0.39, + "latex": "I ) U" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 789.0, + 799.0, + 789.0, + 799.0, + 834.0, + 293.0, + 834.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2059.0, + 863.0, + 2059.0, + 863.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 393.0, + 1406.0, + 393.0, + 1406.0, + 427.0, + 295.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 422.0, + 1404.0, + 422.0, + 1404.0, + 456.0, + 295.0, + 456.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 454.0, + 1404.0, + 454.0, + 1404.0, + 488.0, + 295.0, + 488.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 482.0, + 1409.0, + 482.0, + 1409.0, + 521.0, + 292.0, + 521.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 511.0, + 1406.0, + 511.0, + 1406.0, + 551.0, + 292.0, + 551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 546.0, + 1403.0, + 546.0, + 1403.0, + 577.0, + 296.0, + 577.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 576.0, + 578.0, + 576.0, + 578.0, + 610.0, + 292.0, + 610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 599.0, + 576.0, + 1040.0, + 576.0, + 1040.0, + 610.0, + 599.0, + 610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1061.0, + 576.0, + 1407.0, + 576.0, + 1407.0, + 610.0, + 1061.0, + 610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 606.0, + 1406.0, + 606.0, + 1406.0, + 639.0, + 294.0, + 639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 636.0, + 1406.0, + 636.0, + 1406.0, + 669.0, + 294.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 663.0, + 1406.0, + 663.0, + 1406.0, + 703.0, + 292.0, + 703.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 697.0, + 694.0, + 697.0, + 694.0, + 731.0, + 295.0, + 731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 902.0, + 697.0, + 1406.0, + 697.0, + 1406.0, + 731.0, + 902.0, + 731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 728.0, + 1044.0, + 728.0, + 1044.0, + 759.0, + 294.0, + 759.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 851.0, + 1405.0, + 851.0, + 1405.0, + 883.0, + 296.0, + 883.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 879.0, + 676.0, + 879.0, + 676.0, + 915.0, + 293.0, + 915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 716.0, + 879.0, + 1133.0, + 879.0, + 1133.0, + 915.0, + 716.0, + 915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1331.0, + 879.0, + 1406.0, + 879.0, + 1406.0, + 915.0, + 1331.0, + 915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 908.0, + 1408.0, + 908.0, + 1408.0, + 949.0, + 292.0, + 949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 939.0, + 1407.0, + 939.0, + 1407.0, + 981.0, + 291.0, + 981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 970.0, + 1406.0, + 970.0, + 1406.0, + 1010.0, + 291.0, + 1010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1001.0, + 975.0, + 1001.0, + 975.0, + 1037.0, + 293.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1130.0, + 646.0, + 1130.0, + 646.0, + 1169.0, + 293.0, + 1169.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 674.0, + 1130.0, + 800.0, + 1130.0, + 800.0, + 1169.0, + 674.0, + 1169.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 826.0, + 1130.0, + 885.0, + 1130.0, + 885.0, + 1169.0, + 826.0, + 1169.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1071.0, + 1130.0, + 1406.0, + 1130.0, + 1406.0, + 1169.0, + 1071.0, + 1169.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1163.0, + 1404.0, + 1163.0, + 1404.0, + 1196.0, + 295.0, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1192.0, + 1017.0, + 1192.0, + 1017.0, + 1229.0, + 293.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1105.0, + 1192.0, + 1333.0, + 1192.0, + 1333.0, + 1229.0, + 1105.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1399.0, + 1192.0, + 1411.0, + 1192.0, + 1411.0, + 1229.0, + 1399.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1223.0, + 866.0, + 1223.0, + 866.0, + 1258.0, + 295.0, + 1258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 892.0, + 1223.0, + 1022.0, + 1223.0, + 1022.0, + 1258.0, + 892.0, + 1258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1902.0, + 759.0, + 1902.0, + 759.0, + 1949.0, + 291.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 847.0, + 1902.0, + 1008.0, + 1902.0, + 1008.0, + 1949.0, + 847.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 1902.0, + 1057.0, + 1902.0, + 1057.0, + 1949.0, + 1028.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1353.0, + 1902.0, + 1406.0, + 1902.0, + 1406.0, + 1949.0, + 1353.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1939.0, + 297.0, + 1939.0, + 297.0, + 1978.0, + 294.0, + 1978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 415.0, + 1939.0, + 1402.0, + 1939.0, + 1402.0, + 1978.0, + 415.0, + 1978.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 704.0, + 1969.0, + 1123.0, + 1969.0, + 1123.0, + 2014.0, + 704.0, + 2014.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1141.0, + 1969.0, + 1410.0, + 1969.0, + 1410.0, + 2014.0, + 1141.0, + 2014.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 285.75, + 1964.0, + 711.75, + 1964.0, + 711.75, + 2021.0, + 285.75, + 2021.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 200.0, + 371.0, + 200.0, + 371.0, + 239.0, + 295.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 528.0, + 200.0, + 580.0, + 200.0, + 580.0, + 239.0, + 528.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 737.0, + 200.0, + 1403.0, + 200.0, + 1403.0, + 239.0, + 737.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 429.0, + 234.0, + 966.0, + 234.0, + 966.0, + 281.0, + 429.0, + 281.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1156.0, + 234.0, + 1406.0, + 234.0, + 1406.0, + 281.0, + 1156.0, + 281.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 274.0, + 364.0, + 274.0, + 364.0, + 310.0, + 294.0, + 310.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 460.0, + 274.0, + 549.0, + 274.0, + 549.0, + 310.0, + 460.0, + 310.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 590.0, + 274.0, + 1405.0, + 274.0, + 1405.0, + 310.0, + 590.0, + 310.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 385.0, + 306.0, + 1160.0, + 306.0, + 1160.0, + 341.0, + 385.0, + 341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1176.0, + 306.0, + 1379.0, + 306.0, + 1379.0, + 341.0, + 1176.0, + 341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1700.0, + 837.0, + 1700.0, + 837.0, + 1744.0, + 293.0, + 1744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1079.0, + 1700.0, + 1128.0, + 1700.0, + 1128.0, + 1744.0, + 1079.0, + 1744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1349.0, + 1700.0, + 1407.0, + 1700.0, + 1407.0, + 1744.0, + 1349.0, + 1744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1735.0, + 297.0, + 1735.0, + 297.0, + 1774.0, + 293.0, + 1774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1735.0, + 1407.0, + 1735.0, + 1407.0, + 1774.0, + 368.0, + 1774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1765.0, + 414.0, + 1765.0, + 414.0, + 1807.0, + 293.0, + 1807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 434.0, + 1765.0, + 448.0, + 1765.0, + 448.0, + 1807.0, + 434.0, + 1807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1425.0, + 578.0, + 1425.0, + 578.0, + 1467.0, + 293.0, + 1467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 763.0, + 1425.0, + 1230.0, + 1425.0, + 1230.0, + 1467.0, + 763.0, + 1467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1256.0, + 1425.0, + 1409.0, + 1425.0, + 1409.0, + 1467.0, + 1256.0, + 1467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1457.0, + 297.0, + 1457.0, + 297.0, + 1499.0, + 294.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 497.0, + 1457.0, + 546.0, + 1457.0, + 546.0, + 1499.0, + 497.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 771.0, + 1457.0, + 847.0, + 1457.0, + 847.0, + 1499.0, + 771.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 891.0, + 1457.0, + 1118.0, + 1457.0, + 1118.0, + 1499.0, + 891.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1143.0, + 1457.0, + 1198.0, + 1457.0, + 1198.0, + 1499.0, + 1143.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1399.0, + 1457.0, + 1411.0, + 1457.0, + 1411.0, + 1499.0, + 1399.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1493.0, + 547.0, + 1493.0, + 547.0, + 1526.0, + 294.0, + 1526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1323.0, + 370.0, + 1323.0, + 370.0, + 1365.0, + 292.0, + 1365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 453.0, + 1323.0, + 876.0, + 1323.0, + 876.0, + 1365.0, + 453.0, + 1365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 902.0, + 1323.0, + 951.0, + 1323.0, + 951.0, + 1365.0, + 902.0, + 1365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 972.0, + 1323.0, + 1404.0, + 1323.0, + 1404.0, + 1365.0, + 972.0, + 1365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1357.0, + 348.0, + 1357.0, + 348.0, + 1390.0, + 295.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 380.0, + 1357.0, + 712.0, + 1357.0, + 712.0, + 1390.0, + 380.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 786.0, + 1357.0, + 1404.0, + 1357.0, + 1404.0, + 1390.0, + 786.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1388.0, + 354.0, + 1388.0, + 354.0, + 1420.0, + 295.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 436.0, + 1388.0, + 668.0, + 1388.0, + 668.0, + 1420.0, + 436.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 1388.0, + 937.0, + 1388.0, + 937.0, + 1420.0, + 721.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 1388.0, + 1200.0, + 1388.0, + 1200.0, + 1420.0, + 964.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1265.0, + 1388.0, + 1277.0, + 1388.0, + 1277.0, + 1420.0, + 1265.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1587.0, + 377.0, + 1587.0, + 377.0, + 1633.0, + 292.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 730.0, + 1587.0, + 794.0, + 1587.0, + 794.0, + 1633.0, + 730.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1342.0, + 1587.0, + 1407.0, + 1587.0, + 1407.0, + 1633.0, + 1342.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 286.0, + 1620.0, + 297.0, + 1620.0, + 297.0, + 1690.0, + 286.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 513.0, + 1620.0, + 588.0, + 1620.0, + 588.0, + 1690.0, + 513.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 776.0, + 1620.0, + 834.0, + 1620.0, + 834.0, + 1690.0, + 776.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1157.0, + 1620.0, + 1179.0, + 1620.0, + 1179.0, + 1690.0, + 1157.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 971.0, + 1624.0, + 1009.0, + 1624.0, + 1009.0, + 1653.0, + 971.0, + 1653.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1040.0, + 875.0, + 1040.0, + 875.0, + 1079.0, + 293.0, + 1079.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 943.0, + 1040.0, + 1405.0, + 1040.0, + 1405.0, + 1079.0, + 943.0, + 1079.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1073.0, + 477.0, + 1073.0, + 477.0, + 1112.0, + 392.0, + 1112.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 544.0, + 1073.0, + 568.0, + 1073.0, + 568.0, + 1112.0, + 544.0, + 1112.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 690.0, + 1073.0, + 729.0, + 1073.0, + 729.0, + 1112.0, + 690.0, + 1112.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 1073.0, + 883.0, + 1073.0, + 883.0, + 1112.0, + 832.0, + 1112.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 1073.0, + 1155.0, + 1073.0, + 1155.0, + 1112.0, + 1006.0, + 1112.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1227.0, + 1073.0, + 1237.0, + 1073.0, + 1237.0, + 1112.0, + 1227.0, + 1112.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1537.0, + 294.0, + 1537.0, + 294.0, + 1581.0, + 289.0, + 1581.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1537.0, + 939.0, + 1537.0, + 939.0, + 1581.0, + 351.0, + 1581.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1379.0, + 1537.0, + 1386.0, + 1537.0, + 1386.0, + 1581.0, + 1379.0, + 1581.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 3, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1214, + 1405, + 1214, + 1405, + 1560, + 297, + 1560 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 1801, + 1404, + 1801, + 1404, + 1957, + 298, + 1957 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 1650, + 1404, + 1650, + 1404, + 1772, + 298, + 1772 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 298, + 621, + 1404, + 621, + 1404, + 745, + 298, + 745 + ], + "score": 0.972 + }, + { + "category_id": 1, + "poly": [ + 299, + 1108, + 1400, + 1108, + 1400, + 1203, + 299, + 1203 + ], + "score": 0.964 + }, + { + "category_id": 3, + "poly": [ + 319, + 220, + 1373, + 220, + 1373, + 438, + 319, + 438 + ], + "score": 0.955 + }, + { + "category_id": 4, + "poly": [ + 299, + 467, + 1398, + 467, + 1398, + 526, + 299, + 526 + ], + "score": 0.948 + }, + { + "category_id": 1, + "poly": [ + 298, + 547, + 1399, + 547, + 1399, + 609, + 298, + 609 + ], + "score": 0.943 + }, + { + "category_id": 0, + "poly": [ + 298, + 1041, + 1209, + 1041, + 1209, + 1081, + 298, + 1081 + ], + "score": 0.932 + }, + { + "category_id": 0, + "poly": [ + 299, + 1593, + 821, + 1593, + 821, + 1626, + 299, + 1626 + ], + "score": 0.925 + }, + { + "category_id": 8, + "poly": [ + 515, + 1964, + 1181, + 1964, + 1181, + 2002, + 515, + 2002 + ], + "score": 0.923 + }, + { + "category_id": 1, + "poly": [ + 318, + 816, + 1405, + 816, + 1405, + 979, + 318, + 979 + ], + "score": 0.889 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1969, + 1400, + 1969, + 1400, + 1997, + 1366, + 1997 + ], + "score": 0.854 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 858, + 2061, + 858, + 2085, + 841, + 2085 + ], + "score": 0.739 + }, + { + "category_id": 0, + "poly": [ + 297, + 776, + 871, + 776, + 871, + 809, + 297, + 809 + ], + "score": 0.334 + }, + { + "category_id": 13, + "poly": [ + 665, + 1277, + 875, + 1277, + 875, + 1313, + 665, + 1313 + ], + "score": 0.93, + "latex": "x _ { i } ^ { h } \\in \\{ e _ { 1 } , \\ldots , e _ { k } \\}" + }, + { + "category_id": 13, + "poly": [ + 935, + 1245, + 1124, + 1245, + 1124, + 1280, + 935, + 1280 + ], + "score": 0.93, + "latex": "\\boldsymbol { x } = [ x _ { i } ^ { r } , x _ { i } ^ { h } ] _ { i = 1 } ^ { \\bar { m } }" + }, + { + "category_id": 13, + "poly": [ + 1119, + 1402, + 1228, + 1402, + 1228, + 1438, + 1119, + 1438 + ], + "score": 0.93, + "latex": "x _ { i } ^ { h } \\in \\mathbb { R } ^ { k }" + }, + { + "category_id": 13, + "poly": [ + 1218, + 1245, + 1327, + 1245, + 1327, + 1280, + 1218, + 1280 + ], + "score": 0.93, + "latex": "x _ { i } ^ { r } \\in \\dot { \\mathbb { R } } ^ { 3 }" + }, + { + "category_id": 13, + "poly": [ + 965, + 1468, + 1027, + 1468, + 1027, + 1500, + 965, + 1500 + ], + "score": 0.92, + "latex": "E ( x )" + }, + { + "category_id": 13, + "poly": [ + 415, + 651, + 695, + 651, + 695, + 687, + 415, + 687 + ], + "score": 0.92, + "latex": "s _ { t } ^ { \\theta } ( z ) = \\alpha f _ { t } ( z ) \\stackrel { . } { + } \\tilde { s } _ { t } ^ { \\theta } ( z )" + }, + { + "category_id": 13, + "poly": [ + 1016, + 1281, + 1219, + 1281, + 1219, + 1313, + 1016, + 1313 + ], + "score": 0.92, + "latex": "e _ { i } = [ 0 \\cdots 1 \\cdot \\cdot \\cdot 0 ]" + }, + { + "category_id": 13, + "poly": [ + 297, + 1435, + 578, + 1435, + 578, + 1471, + 297, + 1471 + ], + "score": 0.91, + "latex": "\\hat { x } _ { i } ^ { h } = \\mathbb { I } ( x _ { i } ^ { h } = \\operatorname* { m a x } ( x _ { i } ^ { h } ) )" + }, + { + "category_id": 13, + "poly": [ + 1215, + 1467, + 1403, + 1467, + 1403, + 1501, + 1215, + 1501 + ], + "score": 0.91, + "latex": "f _ { t } ( \\cdot ) = - \\nabla E ( \\cdot )" + }, + { + "category_id": 13, + "poly": [ + 750, + 882, + 807, + 882, + 807, + 916, + 750, + 916 + ], + "score": 0.91, + "latex": "\\mathcal { L } ( \\boldsymbol { \\theta } )" + }, + { + "category_id": 13, + "poly": [ + 1018, + 579, + 1088, + 579, + 1088, + 606, + 1018, + 606 + ], + "score": 0.9, + "latex": "t \\to 1" + }, + { + "category_id": 13, + "poly": [ + 603, + 1342, + 636, + 1342, + 636, + 1376, + 603, + 1376 + ], + "score": 0.89, + "latex": "\\boldsymbol { x } _ { i } ^ { h }" + }, + { + "category_id": 13, + "poly": [ + 583, + 815, + 658, + 815, + 658, + 853, + 583, + 853 + ], + "score": 0.89, + "latex": "\\{ x ^ { ( k ) } \\}" + }, + { + "category_id": 14, + "poly": [ + 514, + 1963, + 1183, + 1963, + 1183, + 2002, + 514, + 2002 + ], + "score": 0.89, + "latex": "E ( x ) = E _ { b o n d } ( x ) + E _ { a n g l e } ( x ) + E _ { L J } ( x ) + E _ { C o u l o m b } ( x ) ." + }, + { + "category_id": 13, + "poly": [ + 409, + 1311, + 446, + 1311, + 446, + 1342, + 409, + 1342 + ], + "score": 0.88, + "latex": "\\mathbb { R } ^ { k }" + }, + { + "category_id": 13, + "poly": [ + 619, + 941, + 652, + 941, + 652, + 971, + 619, + 971 + ], + "score": 0.88, + "latex": "\\mathbb { P } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 526, + 850, + 558, + 850, + 558, + 880, + 526, + 880 + ], + "score": 0.87, + "latex": "\\mathbb { P } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 927, + 1345, + 962, + 1345, + 962, + 1372, + 927, + 1372 + ], + "score": 0.87, + "latex": "\\mathbb { R } ^ { r }" + }, + { + "category_id": 13, + "poly": [ + 529, + 620, + 561, + 620, + 561, + 650, + 529, + 650 + ], + "score": 0.86, + "latex": "\\mathbb { P } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 1213, + 549, + 1233, + 549, + 1233, + 579, + 1213, + 579 + ], + "score": 0.85, + "latex": "f" + }, + { + "category_id": 13, + "poly": [ + 461, + 549, + 481, + 549, + 481, + 579, + 461, + 579 + ], + "score": 0.85, + "latex": "f" + }, + { + "category_id": 13, + "poly": [ + 1306, + 1502, + 1331, + 1502, + 1331, + 1526, + 1306, + 1526 + ], + "score": 0.84, + "latex": "E" + }, + { + "category_id": 13, + "poly": [ + 977, + 624, + 996, + 624, + 996, + 655, + 977, + 655 + ], + "score": 0.84, + "latex": "f" + }, + { + "category_id": 13, + "poly": [ + 978, + 1314, + 998, + 1314, + 998, + 1341, + 978, + 1341 + ], + "score": 0.8, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 814, + 1712, + 833, + 1712, + 833, + 1738, + 814, + 1738 + ], + "score": 0.8, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 670, + 819, + 708, + 819, + 708, + 852, + 670, + 852 + ], + "score": 0.78, + "latex": "\\mathbb { Q } ^ { x }" + }, + { + "category_id": 13, + "poly": [ + 1296, + 658, + 1319, + 658, + 1319, + 682, + 1296, + 682 + ], + "score": 0.77, + "latex": "\\alpha" + }, + { + "category_id": 13, + "poly": [ + 829, + 1317, + 841, + 1317, + 841, + 1341, + 829, + 1341 + ], + "score": 0.77, + "latex": "i" + }, + { + "category_id": 13, + "poly": [ + 652, + 1222, + 671, + 1222, + 671, + 1243, + 652, + 1243 + ], + "score": 0.76, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 554, + 884, + 572, + 884, + 572, + 910, + 554, + 910 + ], + "score": 0.75, + "latex": "\\theta" + }, + { + "category_id": 13, + "poly": [ + 779, + 655, + 797, + 655, + 797, + 681, + 779, + 681 + ], + "score": 0.74, + "latex": "\\tilde { s }" + }, + { + "category_id": 13, + "poly": [ + 496, + 1281, + 510, + 1281, + 510, + 1307, + 496, + 1307 + ], + "score": 0.71, + "latex": "i" + }, + { + "category_id": 13, + "poly": [ + 1293, + 1283, + 1306, + 1283, + 1306, + 1307, + 1293, + 1307 + ], + "score": 0.71, + "latex": "i \\cdot" + }, + { + "category_id": 13, + "poly": [ + 1314, + 820, + 1355, + 820, + 1355, + 852, + 1314, + 852 + ], + "score": 0.55, + "latex": "f , { \\mathrm { v } }" + }, + { + "category_id": 13, + "poly": [ + 1314, + 821, + 1332, + 821, + 1332, + 852, + 1314, + 852 + ], + "score": 0.32, + "latex": "f" + }, + { + "category_id": 15, + "poly": [ + 826.0, + 241.0, + 904.0, + 241.0, + 904.0, + 283.0, + 826.0, + 283.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 911.0, + 245.0, + 935.0, + 245.0, + 935.0, + 274.0, + 911.0, + 274.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1012.0, + 240.0, + 1079.0, + 240.0, + 1079.0, + 279.0, + 1012.0, + 279.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1032.0, + 277.0, + 1061.0, + 277.0, + 1061.0, + 312.0, + 1032.0, + 312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 684.0, + 334.0, + 704.0, + 334.0, + 704.0, + 348.0, + 684.0, + 348.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 762.0, + 323.0, + 827.0, + 323.0, + 827.0, + 353.0, + 762.0, + 353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1007.0, + 305.0, + 1085.0, + 305.0, + 1085.0, + 355.0, + 1007.0, + 355.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1031.0, + 351.0, + 1061.0, + 351.0, + 1061.0, + 385.0, + 1031.0, + 385.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 338.0, + 380.0, + 428.0, + 380.0, + 428.0, + 417.0, + 338.0, + 417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 510.0, + 368.0, + 718.0, + 368.0, + 718.0, + 413.0, + 510.0, + 413.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 992.0, + 388.0, + 1103.0, + 388.0, + 1103.0, + 442.0, + 992.0, + 442.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 464.0, + 1406.0, + 464.0, + 1406.0, + 500.0, + 294.0, + 500.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 495.0, + 1398.0, + 495.0, + 1398.0, + 529.0, + 294.0, + 529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1038.0, + 1213.0, + 1038.0, + 1213.0, + 1088.0, + 289.0, + 1088.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1589.0, + 826.0, + 1589.0, + 826.0, + 1631.0, + 292.0, + 1631.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2058.0, + 861.0, + 2058.0, + 861.0, + 2093.0, + 839.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 774.0, + 872.0, + 774.0, + 872.0, + 813.0, + 296.0, + 813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1215.0, + 651.0, + 1215.0, + 651.0, + 1250.0, + 294.0, + 1250.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 672.0, + 1215.0, + 1406.0, + 1215.0, + 1406.0, + 1250.0, + 672.0, + 1250.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 286.0, + 1232.0, + 934.0, + 1232.0, + 934.0, + 1296.0, + 286.0, + 1296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1125.0, + 1232.0, + 1217.0, + 1232.0, + 1217.0, + 1296.0, + 1125.0, + 1296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1328.0, + 1232.0, + 1413.0, + 1232.0, + 1413.0, + 1296.0, + 1328.0, + 1296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1277.0, + 495.0, + 1277.0, + 495.0, + 1316.0, + 294.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 511.0, + 1277.0, + 664.0, + 1277.0, + 664.0, + 1316.0, + 511.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 876.0, + 1277.0, + 1015.0, + 1277.0, + 1015.0, + 1316.0, + 876.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1220.0, + 1277.0, + 1292.0, + 1277.0, + 1292.0, + 1316.0, + 1220.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1307.0, + 1277.0, + 1405.0, + 1277.0, + 1405.0, + 1316.0, + 1307.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1310.0, + 408.0, + 1310.0, + 408.0, + 1349.0, + 292.0, + 1349.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 447.0, + 1310.0, + 828.0, + 1310.0, + 828.0, + 1349.0, + 447.0, + 1349.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 842.0, + 1310.0, + 977.0, + 1310.0, + 977.0, + 1349.0, + 842.0, + 1349.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 999.0, + 1310.0, + 1405.0, + 1310.0, + 1405.0, + 1349.0, + 999.0, + 1349.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1339.0, + 602.0, + 1339.0, + 602.0, + 1381.0, + 291.0, + 1381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 637.0, + 1339.0, + 926.0, + 1339.0, + 926.0, + 1381.0, + 637.0, + 1381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 963.0, + 1339.0, + 1408.0, + 1339.0, + 1408.0, + 1381.0, + 963.0, + 1381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1374.0, + 1406.0, + 1374.0, + 1406.0, + 1410.0, + 294.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1397.0, + 1118.0, + 1397.0, + 1118.0, + 1444.0, + 290.0, + 1444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1229.0, + 1397.0, + 1408.0, + 1397.0, + 1408.0, + 1444.0, + 1229.0, + 1444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1426.0, + 296.0, + 1426.0, + 296.0, + 1482.0, + 289.0, + 1482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 579.0, + 1426.0, + 1412.0, + 1426.0, + 1412.0, + 1482.0, + 579.0, + 1482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1466.0, + 964.0, + 1466.0, + 964.0, + 1505.0, + 292.0, + 1505.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 1466.0, + 1214.0, + 1466.0, + 1214.0, + 1505.0, + 1028.0, + 1505.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1404.0, + 1466.0, + 1408.0, + 1466.0, + 1408.0, + 1505.0, + 1404.0, + 1505.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1495.0, + 1305.0, + 1495.0, + 1305.0, + 1535.0, + 291.0, + 1535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1332.0, + 1495.0, + 1406.0, + 1495.0, + 1406.0, + 1535.0, + 1332.0, + 1535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1525.0, + 419.0, + 1525.0, + 419.0, + 1566.0, + 293.0, + 1566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1800.0, + 1408.0, + 1800.0, + 1408.0, + 1839.0, + 293.0, + 1839.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1835.0, + 1405.0, + 1835.0, + 1405.0, + 1868.0, + 296.0, + 1868.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1864.0, + 1404.0, + 1864.0, + 1404.0, + 1898.0, + 296.0, + 1898.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1890.0, + 1409.0, + 1890.0, + 1409.0, + 1935.0, + 292.0, + 1935.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1926.0, + 1397.0, + 1926.0, + 1397.0, + 1963.0, + 296.0, + 1963.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1650.0, + 1406.0, + 1650.0, + 1406.0, + 1686.0, + 293.0, + 1686.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1679.0, + 1406.0, + 1679.0, + 1406.0, + 1718.0, + 292.0, + 1718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1711.0, + 813.0, + 1711.0, + 813.0, + 1745.0, + 292.0, + 1745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 834.0, + 1711.0, + 1405.0, + 1711.0, + 1405.0, + 1745.0, + 834.0, + 1745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1738.0, + 399.0, + 1738.0, + 399.0, + 1776.0, + 293.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 618.0, + 528.0, + 618.0, + 528.0, + 657.0, + 292.0, + 657.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 562.0, + 618.0, + 976.0, + 618.0, + 976.0, + 657.0, + 562.0, + 657.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 997.0, + 618.0, + 1405.0, + 618.0, + 1405.0, + 657.0, + 997.0, + 657.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 650.0, + 414.0, + 650.0, + 414.0, + 690.0, + 292.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 696.0, + 650.0, + 778.0, + 650.0, + 778.0, + 690.0, + 696.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 798.0, + 650.0, + 1295.0, + 650.0, + 1295.0, + 690.0, + 798.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1320.0, + 650.0, + 1406.0, + 650.0, + 1406.0, + 690.0, + 1320.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 683.0, + 1405.0, + 683.0, + 1405.0, + 719.0, + 294.0, + 719.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 715.0, + 773.0, + 715.0, + 773.0, + 747.0, + 296.0, + 747.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1109.0, + 1404.0, + 1109.0, + 1404.0, + 1143.0, + 295.0, + 1143.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1140.0, + 1404.0, + 1140.0, + 1404.0, + 1176.0, + 293.0, + 1176.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1170.0, + 993.0, + 1170.0, + 993.0, + 1207.0, + 294.0, + 1207.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 546.0, + 460.0, + 546.0, + 460.0, + 582.0, + 296.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 482.0, + 546.0, + 1212.0, + 546.0, + 1212.0, + 582.0, + 482.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1234.0, + 546.0, + 1403.0, + 546.0, + 1403.0, + 582.0, + 1234.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 577.0, + 1017.0, + 577.0, + 1017.0, + 612.0, + 296.0, + 612.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1089.0, + 577.0, + 1099.0, + 577.0, + 1099.0, + 612.0, + 1089.0, + 612.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 813.0, + 582.0, + 813.0, + 582.0, + 856.0, + 320.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 659.0, + 813.0, + 669.0, + 813.0, + 669.0, + 856.0, + 659.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 709.0, + 813.0, + 1313.0, + 813.0, + 1313.0, + 856.0, + 709.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1356.0, + 813.0, + 1407.0, + 813.0, + 1407.0, + 856.0, + 1356.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 851.0, + 525.0, + 851.0, + 525.0, + 882.0, + 320.0, + 882.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 559.0, + 851.0, + 568.0, + 851.0, + 568.0, + 882.0, + 559.0, + 882.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 881.0, + 553.0, + 881.0, + 553.0, + 917.0, + 322.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 573.0, + 881.0, + 749.0, + 881.0, + 749.0, + 917.0, + 573.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 808.0, + 881.0, + 1409.0, + 881.0, + 1409.0, + 917.0, + 808.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 909.0, + 456.0, + 909.0, + 456.0, + 946.0, + 322.0, + 946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 942.0, + 618.0, + 942.0, + 618.0, + 975.0, + 321.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 653.0, + 942.0, + 662.0, + 942.0, + 662.0, + 975.0, + 653.0, + 975.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 4, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1023, + 1404, + 1023, + 1404, + 1147, + 298, + 1147 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 296, + 1374, + 1405, + 1374, + 1405, + 1499, + 296, + 1499 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 1844, + 1404, + 1844, + 1404, + 2007, + 298, + 2007 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 297, + 1160, + 1404, + 1160, + 1404, + 1284, + 297, + 1284 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 296, + 710, + 1406, + 710, + 1406, + 835, + 296, + 835 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 296, + 199, + 1405, + 199, + 1405, + 342, + 296, + 342 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 296, + 351, + 1405, + 351, + 1405, + 501, + 296, + 501 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 297, + 597, + 1404, + 597, + 1404, + 699, + 297, + 699 + ], + "score": 0.969 + }, + { + "category_id": 1, + "poly": [ + 298, + 847, + 1404, + 847, + 1404, + 911, + 298, + 911 + ], + "score": 0.952 + }, + { + "category_id": 1, + "poly": [ + 297, + 1682, + 1399, + 1682, + 1399, + 1746, + 297, + 1746 + ], + "score": 0.952 + }, + { + "category_id": 8, + "poly": [ + 314, + 922, + 1343, + 922, + 1343, + 1010, + 314, + 1010 + ], + "score": 0.95 + }, + { + "category_id": 1, + "poly": [ + 298, + 1510, + 1397, + 1510, + 1397, + 1574, + 298, + 1574 + ], + "score": 0.947 + }, + { + "category_id": 8, + "poly": [ + 614, + 1759, + 1086, + 1759, + 1086, + 1829, + 614, + 1829 + ], + "score": 0.945 + }, + { + "category_id": 1, + "poly": [ + 294, + 513, + 1399, + 513, + 1399, + 587, + 294, + 587 + ], + "score": 0.94 + }, + { + "category_id": 0, + "poly": [ + 297, + 1317, + 857, + 1317, + 857, + 1352, + 297, + 1352 + ], + "score": 0.935 + }, + { + "category_id": 8, + "poly": [ + 662, + 1587, + 1037, + 1587, + 1037, + 1669, + 662, + 1669 + ], + "score": 0.934 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1770, + 1400, + 1770, + 1400, + 1800, + 1366, + 1800 + ], + "score": 0.891 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1606, + 1400, + 1606, + 1400, + 1636, + 1366, + 1636 + ], + "score": 0.887 + }, + { + "category_id": 2, + "poly": [ + 840, + 2061, + 859, + 2061, + 859, + 2085, + 840, + 2085 + ], + "score": 0.797 + }, + { + "category_id": 9, + "poly": [ + 1368, + 943, + 1400, + 943, + 1400, + 974, + 1368, + 974 + ], + "score": 0.754 + }, + { + "category_id": 14, + "poly": [ + 659, + 1584, + 1040, + 1584, + 1040, + 1672, + 659, + 1672 + ], + "score": 0.94, + "latex": "E _ { \\mathrm { R i e s z } } ( x ) = \\frac { 1 } { 2 } \\sum _ { j \\neq i } | | x _ { i } ^ { r } - x _ { j } ^ { r } | | ^ { - 2 } ." + }, + { + "category_id": 14, + "poly": [ + 613, + 1757, + 1083, + 1757, + 1083, + 1831, + 613, + 1831 + ], + "score": 0.94, + "latex": "E _ { \\mathrm { k n n } } ( x ) = \\sum _ { i } \\left( \\mathrm { k n n - d i s t } _ { i } ( x ^ { r } ) - \\mu _ { k n n } \\right) ^ { 2 } ," + }, + { + "category_id": 13, + "poly": [ + 708, + 427, + 811, + 427, + 811, + 468, + 708, + 468 + ], + "score": 0.94, + "latex": "\\omega ^ { 0 } ( \\hat { x } _ { i j k } ^ { h } )" + }, + { + "category_id": 13, + "poly": [ + 1011, + 276, + 1095, + 276, + 1095, + 311, + 1011, + 311 + ], + "score": 0.93, + "latex": "\\ell ^ { 0 } ( r , c )" + }, + { + "category_id": 14, + "poly": [ + 311, + 921, + 1353, + 921, + 1353, + 1010, + 311, + 1010 + ], + "score": 0.93, + "latex": "E _ { s t a t } ( \\boldsymbol { x } ) = \\sum _ { i j \\in k n n ( \\boldsymbol { x } ) } \\frac { 1 } { \\hat { \\sigma } _ { \\hat { x } _ { i j } ^ { h } } ^ { 2 } } \\left\\| \\mathrm { L e n } ( x _ { i j } ^ { r } ) - \\hat { \\mu } _ { \\hat { x } _ { i j } ^ { h } } \\right\\| ^ { 2 } + \\sum _ { i j , j k \\in k n n ( \\boldsymbol { x } ) } \\frac { 1 } { \\sigma _ { \\hat { x } _ { i j k } ^ { h } } ^ { 2 } } \\left\\| \\mathrm { A n g } ( x _ { i j k } ^ { r } ) - \\mu _ { \\hat { x } _ { i j k } ^ { h } } \\right\\| ^ { 2 } ," + }, + { + "category_id": 13, + "poly": [ + 1041, + 1084, + 1099, + 1084, + 1099, + 1120, + 1041, + 1120 + ], + "score": 0.93, + "latex": "\\hat { \\sigma } _ { r c r ^ { \\prime } } ^ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 410, + 429, + 503, + 429, + 503, + 467, + 410, + 467 + ], + "score": 0.93, + "latex": "x _ { i } ^ { r } - x _ { j } ^ { r }" + }, + { + "category_id": 13, + "poly": [ + 555, + 429, + 649, + 429, + 649, + 467, + 555, + 467 + ], + "score": 0.93, + "latex": "\\boldsymbol { x } _ { k } ^ { r } - \\boldsymbol { x } _ { j } ^ { r }" + }, + { + "category_id": 13, + "poly": [ + 803, + 512, + 1154, + 512, + 1154, + 555, + 803, + 555 + ], + "score": 0.93, + "latex": "\\begin{array} { r } { E _ { L J } ( x ) = \\sum _ { i \\neq j } e ( \\left| \\left| x _ { i } ^ { r } - x _ { j } ^ { r } \\right| \\right| ) } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 374, + 1844, + 888, + 1844, + 888, + 1888, + 374, + 1888 + ], + "score": 0.93, + "latex": "\\begin{array} { r } { \\mathrm { k n n - d i s t } _ { i } ( x ) \\ = \\ \\frac { 1 } { K } \\sum _ { j \\in { \\mathcal N } _ { K } ( x _ { i } ; x ) } | | x _ { i } ^ { r } - x _ { j } ^ { r } | | ^ { 2 } } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 299, + 238, + 421, + 238, + 421, + 280, + 299, + 280 + ], + "score": 0.92, + "latex": "\\| x _ { i } ^ { r } - x _ { j } ^ { r } \\|" + }, + { + "category_id": 13, + "poly": [ + 297, + 630, + 643, + 630, + 643, + 670, + 297, + 670 + ], + "score": 0.92, + "latex": "\\begin{array} { r l } { \\kappa \\sum _ { i j } q ( \\hat { x } _ { i } ^ { h } ) q ( \\hat { x } _ { j } ^ { h } ) / \\left\\| x _ { i } ^ { r } - x _ { j } ^ { r } \\right\\| } & { { } } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 986, + 1405, + 1044, + 1405, + 1044, + 1439, + 986, + 1439 + ], + "score": 0.91, + "latex": "\\{ x _ { i } ^ { r } \\}" + }, + { + "category_id": 13, + "poly": [ + 1058, + 632, + 1109, + 632, + 1109, + 666, + 1058, + 666 + ], + "score": 0.91, + "latex": "q ( r )" + }, + { + "category_id": 13, + "poly": [ + 1001, + 392, + 1123, + 392, + 1123, + 430, + 1001, + 430 + ], + "score": 0.91, + "latex": "\\mathrm { A n g } ( x _ { i j k } ^ { r } )" + }, + { + "category_id": 13, + "poly": [ + 369, + 1024, + 458, + 1024, + 458, + 1057, + 369, + 1057 + ], + "score": 0.91, + "latex": "k n n ( x )" + }, + { + "category_id": 13, + "poly": [ + 569, + 201, + 1155, + 201, + 1155, + 243, + 569, + 243 + ], + "score": 0.91, + "latex": "\\begin{array} { r } { E _ { b o n d } ( x ) \\ = \\ \\sum _ { i j \\in b o n d ( x ) } ( \\operatorname { L e n } ( x _ { i j } ^ { r } ) \\ - \\ \\ell ( \\hat { x } _ { i } ^ { h } , \\hat { x } _ { j } ^ { h } ) ) ^ { 2 } } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 1256, + 202, + 1402, + 202, + 1402, + 241, + 1256, + 241 + ], + "score": 0.91, + "latex": "\\mathrm { L e n } ( x _ { i j } ^ { r } ) ~ =" + }, + { + "category_id": 13, + "poly": [ + 1205, + 513, + 1404, + 513, + 1404, + 551, + 1205, + 551 + ], + "score": 0.9, + "latex": "e ( \\ell ) = ( \\sigma / \\ell ) ^ { 1 2 } -" + }, + { + "category_id": 13, + "poly": [ + 532, + 351, + 1122, + 351, + 1122, + 394, + 532, + 394 + ], + "score": 0.9, + "latex": "\\begin{array} { r } { E _ { a n g l e } ( x ) = \\sum _ { i j k \\in a n g l e ( x ) } ( \\operatorname { A n g } ( x _ { i j k } ^ { r } ) - \\omega ^ { 0 } ( \\hat { x } _ { i j k } ^ { h } ) ) ^ { 2 } } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 452, + 1976, + 521, + 1976, + 521, + 2003, + 452, + 2003 + ], + "score": 0.9, + "latex": "k = 4" + }, + { + "category_id": 13, + "poly": [ + 1212, + 598, + 1402, + 598, + 1402, + 633, + 1212, + 633 + ], + "score": 0.9, + "latex": "E _ { C o u l o m b } ( x ) \\ =" + }, + { + "category_id": 13, + "poly": [ + 296, + 550, + 389, + 550, + 389, + 587, + 296, + 587 + ], + "score": 0.9, + "latex": "2 ( \\sigma / \\ell ) ^ { 6 }" + }, + { + "category_id": 13, + "poly": [ + 838, + 1054, + 879, + 1054, + 879, + 1087, + 838, + 1087 + ], + "score": 0.89, + "latex": "\\hat { \\sigma } _ { r c } ^ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 298, + 466, + 330, + 466, + 330, + 501, + 298, + 501 + ], + "score": 0.89, + "latex": "\\hat { x } _ { k } ^ { h }" + }, + { + "category_id": 13, + "poly": [ + 660, + 1116, + 695, + 1116, + 695, + 1143, + 660, + 1143 + ], + "score": 0.88, + "latex": "c r ^ { \\prime }" + }, + { + "category_id": 13, + "poly": [ + 596, + 1890, + 657, + 1890, + 657, + 1917, + 596, + 1917 + ], + "score": 0.88, + "latex": "\\mu _ { k n n }" + }, + { + "category_id": 13, + "poly": [ + 1303, + 1853, + 1331, + 1853, + 1331, + 1879, + 1303, + 1879 + ], + "score": 0.84, + "latex": "x _ { i }" + }, + { + "category_id": 13, + "poly": [ + 482, + 241, + 578, + 241, + 578, + 276, + 482, + 276 + ], + "score": 0.84, + "latex": "b o n d ( x )" + }, + { + "category_id": 13, + "poly": [ + 951, + 1883, + 1090, + 1883, + 1090, + 1918, + 951, + 1918 + ], + "score": 0.82, + "latex": "\\scriptstyle \\mathrm { k n n - d i s t } _ { i } ( x )" + }, + { + "category_id": 13, + "poly": [ + 631, + 1055, + 734, + 1055, + 734, + 1088, + 631, + 1088 + ], + "score": 0.81, + "latex": "r , c \\in [ k ]" + }, + { + "category_id": 13, + "poly": [ + 834, + 393, + 945, + 393, + 945, + 425, + 834, + 425 + ], + "score": 0.78, + "latex": "b o u n d ( x )" + }, + { + "category_id": 13, + "poly": [ + 569, + 559, + 590, + 559, + 590, + 581, + 569, + 581 + ], + "score": 0.76, + "latex": "\\sigma" + }, + { + "category_id": 13, + "poly": [ + 735, + 637, + 756, + 637, + 756, + 660, + 735, + 660 + ], + "score": 0.76, + "latex": "\\kappa" + }, + { + "category_id": 13, + "poly": [ + 474, + 675, + 490, + 675, + 490, + 694, + 474, + 694 + ], + "score": 0.75, + "latex": "r" + }, + { + "category_id": 13, + "poly": [ + 298, + 1885, + 327, + 1885, + 327, + 1913, + 298, + 1913 + ], + "score": 0.75, + "latex": "K" + }, + { + "category_id": 13, + "poly": [ + 657, + 315, + 672, + 315, + 672, + 337, + 657, + 337 + ], + "score": 0.75, + "latex": "c" + }, + { + "category_id": 13, + "poly": [ + 590, + 316, + 606, + 316, + 606, + 337, + 590, + 337 + ], + "score": 0.74, + "latex": "r" + }, + { + "category_id": 13, + "poly": [ + 928, + 250, + 946, + 250, + 946, + 270, + 928, + 270 + ], + "score": 0.74, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 329, + 1090, + 357, + 1090, + 357, + 1113, + 329, + 1113 + ], + "score": 0.74, + "latex": "r c" + }, + { + "category_id": 13, + "poly": [ + 927, + 1086, + 987, + 1086, + 987, + 1118, + 927, + 1118 + ], + "score": 0.73, + "latex": "\\hat { \\mu } _ { r c r ^ { \\prime } }" + }, + { + "category_id": 13, + "poly": [ + 777, + 1086, + 916, + 1086, + 916, + 1119, + 777, + 1119 + ], + "score": 0.72, + "latex": "r , c , r ^ { \\prime } \\in [ k ]" + }, + { + "category_id": 13, + "poly": [ + 745, + 1056, + 788, + 1056, + 788, + 1088, + 745, + 1088 + ], + "score": 0.7, + "latex": "\\hat { \\mu } _ { r c }" + }, + { + "category_id": 13, + "poly": [ + 298, + 1066, + 316, + 1066, + 316, + 1083, + 298, + 1083 + ], + "score": 0.69, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 580, + 1120, + 610, + 1120, + 610, + 1143, + 580, + 1143 + ], + "score": 0.65, + "latex": "r c" + }, + { + "category_id": 13, + "poly": [ + 1319, + 426, + 1396, + 426, + 1396, + 467, + 1319, + 467 + ], + "score": 0.61, + "latex": "\\hat { x } _ { i } ^ { h } , \\hat { x } _ { j } ^ { h }" + }, + { + "category_id": 13, + "poly": [ + 597, + 1716, + 616, + 1716, + 616, + 1741, + 597, + 1741 + ], + "score": 0.59, + "latex": "\\mathbf { k }" + }, + { + "category_id": 13, + "poly": [ + 1274, + 354, + 1312, + 354, + 1312, + 388, + 1274, + 388 + ], + "score": 0.53, + "latex": "( x )" + }, + { + "category_id": 13, + "poly": [ + 1319, + 426, + 1352, + 426, + 1352, + 465, + 1319, + 465 + ], + "score": 0.52, + "latex": "\\hat { x } _ { i } ^ { h }" + }, + { + "category_id": 13, + "poly": [ + 1362, + 426, + 1396, + 426, + 1396, + 466, + 1362, + 466 + ], + "score": 0.52, + "latex": "\\hat { x } _ { j } ^ { h }" + }, + { + "category_id": 13, + "poly": [ + 591, + 1025, + 614, + 1025, + 614, + 1052, + 591, + 1052 + ], + "score": 0.26, + "latex": "\\mathrm { K }" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1315.0, + 860.0, + 1315.0, + 860.0, + 1358.0, + 292.0, + 1358.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2059.0, + 862.0, + 2059.0, + 862.0, + 2089.0, + 840.0, + 2089.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1024.0, + 368.0, + 1024.0, + 368.0, + 1057.0, + 296.0, + 1057.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 459.0, + 1024.0, + 590.0, + 1024.0, + 590.0, + 1057.0, + 459.0, + 1057.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 615.0, + 1024.0, + 1406.0, + 1024.0, + 1406.0, + 1057.0, + 615.0, + 1057.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1052.0, + 297.0, + 1052.0, + 297.0, + 1092.0, + 292.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 317.0, + 1052.0, + 630.0, + 1052.0, + 630.0, + 1092.0, + 317.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 735.0, + 1052.0, + 744.0, + 1052.0, + 744.0, + 1092.0, + 735.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 1052.0, + 837.0, + 1052.0, + 837.0, + 1092.0, + 789.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 880.0, + 1052.0, + 1406.0, + 1052.0, + 1406.0, + 1092.0, + 880.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1083.0, + 328.0, + 1083.0, + 328.0, + 1122.0, + 291.0, + 1122.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 358.0, + 1083.0, + 776.0, + 1083.0, + 776.0, + 1122.0, + 358.0, + 1122.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 917.0, + 1083.0, + 926.0, + 1083.0, + 926.0, + 1122.0, + 917.0, + 1122.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 988.0, + 1083.0, + 1040.0, + 1083.0, + 1040.0, + 1122.0, + 988.0, + 1122.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 1083.0, + 1406.0, + 1083.0, + 1406.0, + 1122.0, + 1100.0, + 1122.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1117.0, + 579.0, + 1117.0, + 579.0, + 1147.0, + 294.0, + 1147.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 611.0, + 1117.0, + 659.0, + 1117.0, + 659.0, + 1147.0, + 611.0, + 1147.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 696.0, + 1117.0, + 778.0, + 1117.0, + 778.0, + 1147.0, + 696.0, + 1147.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1373.0, + 1406.0, + 1373.0, + 1406.0, + 1412.0, + 293.0, + 1412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1404.0, + 985.0, + 1404.0, + 985.0, + 1441.0, + 293.0, + 1441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1045.0, + 1404.0, + 1407.0, + 1404.0, + 1407.0, + 1441.0, + 1045.0, + 1441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1434.0, + 1405.0, + 1434.0, + 1405.0, + 1473.0, + 293.0, + 1473.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1467.0, + 1209.0, + 1467.0, + 1209.0, + 1503.0, + 294.0, + 1503.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 287.0, + 1839.0, + 373.0, + 1839.0, + 373.0, + 1895.0, + 287.0, + 1895.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 889.0, + 1839.0, + 1302.0, + 1839.0, + 1302.0, + 1895.0, + 889.0, + 1895.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1332.0, + 1839.0, + 1411.0, + 1839.0, + 1411.0, + 1895.0, + 1332.0, + 1895.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1881.0, + 297.0, + 1881.0, + 297.0, + 1921.0, + 293.0, + 1921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 328.0, + 1881.0, + 595.0, + 1881.0, + 595.0, + 1921.0, + 328.0, + 1921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 658.0, + 1881.0, + 950.0, + 1881.0, + 950.0, + 1921.0, + 658.0, + 1921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1091.0, + 1881.0, + 1406.0, + 1881.0, + 1406.0, + 1921.0, + 1091.0, + 1921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1915.0, + 1405.0, + 1915.0, + 1405.0, + 1948.0, + 294.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1943.0, + 1404.0, + 1943.0, + 1404.0, + 1980.0, + 294.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1977.0, + 451.0, + 1977.0, + 451.0, + 2005.0, + 294.0, + 2005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 522.0, + 1977.0, + 533.0, + 1977.0, + 533.0, + 2005.0, + 522.0, + 2005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1160.0, + 1406.0, + 1160.0, + 1406.0, + 1196.0, + 294.0, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1192.0, + 1404.0, + 1192.0, + 1404.0, + 1224.0, + 295.0, + 1224.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1221.0, + 1405.0, + 1221.0, + 1405.0, + 1256.0, + 292.0, + 1256.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1254.0, + 738.0, + 1254.0, + 738.0, + 1285.0, + 293.0, + 1285.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 712.0, + 1404.0, + 712.0, + 1404.0, + 744.0, + 295.0, + 744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 742.0, + 1404.0, + 742.0, + 1404.0, + 775.0, + 294.0, + 775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 774.0, + 1404.0, + 774.0, + 1404.0, + 806.0, + 295.0, + 806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 803.0, + 998.0, + 803.0, + 998.0, + 836.0, + 293.0, + 836.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 196.0, + 568.0, + 196.0, + 568.0, + 248.0, + 291.0, + 248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1156.0, + 196.0, + 1255.0, + 196.0, + 1255.0, + 248.0, + 1156.0, + 248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1403.0, + 196.0, + 1406.0, + 196.0, + 1406.0, + 248.0, + 1403.0, + 248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 422.0, + 239.0, + 481.0, + 239.0, + 481.0, + 282.0, + 422.0, + 282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 579.0, + 239.0, + 927.0, + 239.0, + 927.0, + 282.0, + 579.0, + 282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 947.0, + 239.0, + 1406.0, + 239.0, + 1406.0, + 282.0, + 947.0, + 282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 277.0, + 1010.0, + 277.0, + 1010.0, + 312.0, + 294.0, + 312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1096.0, + 277.0, + 1405.0, + 277.0, + 1405.0, + 312.0, + 1096.0, + 312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 307.0, + 589.0, + 307.0, + 589.0, + 343.0, + 295.0, + 343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 607.0, + 307.0, + 656.0, + 307.0, + 656.0, + 343.0, + 607.0, + 343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 673.0, + 307.0, + 1372.0, + 307.0, + 1372.0, + 343.0, + 673.0, + 343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 344.0, + 531.0, + 344.0, + 531.0, + 401.0, + 288.0, + 401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1123.0, + 344.0, + 1273.0, + 344.0, + 1273.0, + 401.0, + 1123.0, + 401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1313.0, + 344.0, + 1410.0, + 344.0, + 1410.0, + 401.0, + 1313.0, + 401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 387.0, + 833.0, + 387.0, + 833.0, + 431.0, + 293.0, + 431.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 946.0, + 387.0, + 1000.0, + 387.0, + 1000.0, + 431.0, + 946.0, + 431.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1124.0, + 387.0, + 1407.0, + 387.0, + 1407.0, + 431.0, + 1124.0, + 431.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 419.0, + 409.0, + 419.0, + 409.0, + 477.0, + 289.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 504.0, + 419.0, + 554.0, + 419.0, + 554.0, + 477.0, + 504.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 650.0, + 419.0, + 707.0, + 419.0, + 707.0, + 477.0, + 650.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 812.0, + 419.0, + 1318.0, + 419.0, + 1318.0, + 477.0, + 812.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1397.0, + 419.0, + 1411.0, + 419.0, + 1411.0, + 477.0, + 1397.0, + 477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 460.0, + 297.0, + 460.0, + 297.0, + 504.0, + 294.0, + 504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 460.0, + 1034.0, + 460.0, + 1034.0, + 504.0, + 331.0, + 504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 594.0, + 1211.0, + 594.0, + 1211.0, + 637.0, + 292.0, + 637.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 644.0, + 628.0, + 734.0, + 628.0, + 734.0, + 671.0, + 644.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 628.0, + 1057.0, + 628.0, + 1057.0, + 671.0, + 757.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1110.0, + 628.0, + 1404.0, + 628.0, + 1404.0, + 671.0, + 1110.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 664.0, + 473.0, + 664.0, + 473.0, + 702.0, + 295.0, + 702.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 491.0, + 664.0, + 966.0, + 664.0, + 966.0, + 702.0, + 491.0, + 702.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 847.0, + 1408.0, + 847.0, + 1408.0, + 883.0, + 296.0, + 883.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 876.0, + 1371.0, + 876.0, + 1371.0, + 914.0, + 293.0, + 914.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1681.0, + 1404.0, + 1681.0, + 1404.0, + 1719.0, + 295.0, + 1719.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1710.0, + 596.0, + 1710.0, + 596.0, + 1752.0, + 292.0, + 1752.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 617.0, + 1710.0, + 1314.0, + 1710.0, + 1314.0, + 1752.0, + 617.0, + 1752.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1509.0, + 1402.0, + 1509.0, + 1402.0, + 1546.0, + 294.0, + 1546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1539.0, + 1401.0, + 1539.0, + 1401.0, + 1578.0, + 294.0, + 1578.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 507.0, + 802.0, + 507.0, + 802.0, + 557.0, + 289.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1155.0, + 507.0, + 1204.0, + 507.0, + 1204.0, + 557.0, + 1155.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 390.0, + 549.0, + 568.0, + 549.0, + 568.0, + 591.0, + 390.0, + 591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 591.0, + 549.0, + 1148.0, + 549.0, + 1148.0, + 591.0, + 591.0, + 591.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 5, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1071, + 1405, + 1071, + 1405, + 1347, + 297, + 1347 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 297, + 814, + 1406, + 814, + 1406, + 1059, + 297, + 1059 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 267, + 1406, + 267, + 1406, + 420, + 298, + 420 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 511, + 1404, + 511, + 1404, + 635, + 298, + 635 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 298, + 1884, + 1403, + 1884, + 1403, + 2007, + 298, + 2007 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 647, + 1404, + 647, + 1404, + 801, + 298, + 801 + ], + "score": 0.975 + }, + { + "category_id": 3, + "poly": [ + 303, + 1650, + 1398, + 1650, + 1398, + 1830, + 303, + 1830 + ], + "score": 0.964 + }, + { + "category_id": 6, + "poly": [ + 297, + 1374, + 1399, + 1374, + 1399, + 1433, + 297, + 1433 + ], + "score": 0.938 + }, + { + "category_id": 4, + "poly": [ + 396, + 1836, + 1298, + 1836, + 1298, + 1867, + 396, + 1867 + ], + "score": 0.929 + }, + { + "category_id": 0, + "poly": [ + 301, + 455, + 781, + 455, + 781, + 488, + 301, + 488 + ], + "score": 0.924 + }, + { + "category_id": 5, + "poly": [ + 401, + 1449, + 1296, + 1449, + 1296, + 1608, + 401, + 1608 + ], + "score": 0.919, + "html": "
QM9GEOM-DRUG
Atom Sta (%)↑Mol Sta (%)↑Valid + Unique↑Atom Sta (%)↑Mol Sta (%)↑
EN-Flow [37]85.04.90.34975.00.0
GDM[19]97.063.2=75.00.0
E-GDM[19]98.7±0.182.0±0.40.90281.30.0
Bridge98.7±0.181.8±0.20.90281.0±0.70.0
Bridge +Force (7)98.8±0.184.6±0.30.90782.4±0.80.0
" + }, + { + "category_id": 0, + "poly": [ + 297, + 198, + 519, + 198, + 519, + 237, + 297, + 237 + ], + "score": 0.916 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 858, + 2061, + 858, + 2084, + 841, + 2084 + ], + "score": 0.722 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2084, + 841, + 2084 + ], + "score": 0.101 + }, + { + "category_id": 13, + "poly": [ + 670, + 1162, + 765, + 1162, + 765, + 1197, + 670, + 1197 + ], + "score": 0.92, + "latex": "1 0 ^ { - 3 } / T" + }, + { + "category_id": 13, + "poly": [ + 1306, + 1132, + 1401, + 1132, + 1401, + 1167, + 1306, + 1167 + ], + "score": 0.91, + "latex": "1 0 ^ { - 2 } / \\check { T }" + }, + { + "category_id": 13, + "poly": [ + 550, + 1223, + 609, + 1223, + 609, + 1254, + 550, + 1254 + ], + "score": 0.9, + "latex": "1 0 ^ { - 4 }" + }, + { + "category_id": 13, + "poly": [ + 1004, + 1102, + 1064, + 1102, + 1064, + 1132, + 1004, + 1132 + ], + "score": 0.9, + "latex": "1 0 ^ { - 4 }" + }, + { + "category_id": 13, + "poly": [ + 1103, + 1164, + 1180, + 1164, + 1180, + 1193, + 1103, + 1193 + ], + "score": 0.89, + "latex": "k = 5" + }, + { + "category_id": 13, + "poly": [ + 579, + 1168, + 601, + 1168, + 601, + 1192, + 579, + 1192 + ], + "score": 0.72, + "latex": "\\alpha" + }, + { + "category_id": 13, + "poly": [ + 800, + 1165, + 875, + 1165, + 875, + 1193, + 800, + 1193 + ], + "score": 0.68, + "latex": "0 \\ w . r . t ." + }, + { + "category_id": 13, + "poly": [ + 909, + 1477, + 925, + 1477, + 925, + 1493, + 909, + 1493 + ], + "score": 0.52, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 304, + 1976, + 419, + 1976, + 419, + 2006, + 304, + 2006 + ], + "score": 0.28, + "latex": "\\mathrm { f r o m 9 8 . 7 }" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 1658.0, + 370.0, + 1658.0, + 370.0, + 1689.0, + 314.0, + 1689.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 869.0, + 1659.0, + 1009.0, + 1659.0, + 1009.0, + 1686.0, + 869.0, + 1686.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 972.0, + 1704.0, + 982.0, + 1704.0, + 982.0, + 1714.0, + 972.0, + 1714.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 746.0, + 1722.0, + 761.0, + 1722.0, + 761.0, + 1743.0, + 746.0, + 1743.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1751.0, + 768.0, + 1751.0, + 768.0, + 1764.0, + 757.0, + 1764.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 587.0, + 1762.0, + 603.0, + 1762.0, + 603.0, + 1779.0, + 587.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.25, + 1717.0, + 778.25, + 1717.0, + 778.25, + 1726.0, + 756.25, + 1726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1367.0, + 1405.0, + 1367.0, + 1405.0, + 1412.0, + 292.0, + 1412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1404.0, + 953.0, + 1404.0, + 953.0, + 1435.0, + 296.0, + 1435.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 1833.0, + 1300.0, + 1833.0, + 1300.0, + 1871.0, + 398.0, + 1871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 453.0, + 783.0, + 453.0, + 783.0, + 492.0, + 294.0, + 492.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 192.0, + 525.0, + 192.0, + 525.0, + 247.0, + 289.0, + 247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2059.0, + 860.0, + 2059.0, + 860.0, + 2091.0, + 839.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2059.0, + 860.0, + 2059.0, + 860.0, + 2092.0, + 840.0, + 2092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1069.0, + 1406.0, + 1069.0, + 1406.0, + 1107.0, + 294.0, + 1107.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1093.0, + 1003.0, + 1093.0, + 1003.0, + 1145.0, + 289.0, + 1145.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1065.0, + 1093.0, + 1410.0, + 1093.0, + 1410.0, + 1145.0, + 1065.0, + 1145.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1132.0, + 1305.0, + 1132.0, + 1305.0, + 1166.0, + 292.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1161.0, + 578.0, + 1161.0, + 578.0, + 1197.0, + 292.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 602.0, + 1161.0, + 669.0, + 1161.0, + 669.0, + 1197.0, + 602.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 766.0, + 1161.0, + 799.0, + 1161.0, + 799.0, + 1197.0, + 766.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 876.0, + 1161.0, + 1102.0, + 1161.0, + 1102.0, + 1197.0, + 876.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1181.0, + 1161.0, + 1406.0, + 1161.0, + 1406.0, + 1197.0, + 1181.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1192.0, + 1407.0, + 1192.0, + 1407.0, + 1230.0, + 294.0, + 1230.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1219.0, + 549.0, + 1219.0, + 549.0, + 1263.0, + 291.0, + 1263.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 610.0, + 1219.0, + 1406.0, + 1219.0, + 1406.0, + 1263.0, + 610.0, + 1263.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1255.0, + 1403.0, + 1255.0, + 1403.0, + 1287.0, + 295.0, + 1287.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1286.0, + 1403.0, + 1286.0, + 1403.0, + 1319.0, + 295.0, + 1319.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1317.0, + 885.0, + 1317.0, + 885.0, + 1350.0, + 295.0, + 1350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 816.0, + 1404.0, + 816.0, + 1404.0, + 849.0, + 295.0, + 849.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 848.0, + 1406.0, + 848.0, + 1406.0, + 878.0, + 296.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 875.0, + 1406.0, + 875.0, + 1406.0, + 909.0, + 295.0, + 909.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 905.0, + 1409.0, + 905.0, + 1409.0, + 941.0, + 292.0, + 941.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 936.0, + 1406.0, + 936.0, + 1406.0, + 970.0, + 295.0, + 970.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 965.0, + 1404.0, + 965.0, + 1404.0, + 1000.0, + 294.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 998.0, + 1406.0, + 998.0, + 1406.0, + 1031.0, + 294.0, + 1031.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1030.0, + 834.0, + 1030.0, + 834.0, + 1060.0, + 296.0, + 1060.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 268.0, + 1408.0, + 268.0, + 1408.0, + 301.0, + 297.0, + 301.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 295.0, + 1407.0, + 295.0, + 1407.0, + 334.0, + 293.0, + 334.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 328.0, + 1406.0, + 328.0, + 1406.0, + 365.0, + 295.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 359.0, + 1404.0, + 359.0, + 1404.0, + 392.0, + 295.0, + 392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 388.0, + 1367.0, + 388.0, + 1367.0, + 425.0, + 295.0, + 425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 510.0, + 1406.0, + 510.0, + 1406.0, + 549.0, + 292.0, + 549.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 544.0, + 1404.0, + 544.0, + 1404.0, + 577.0, + 296.0, + 577.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 573.0, + 1405.0, + 573.0, + 1405.0, + 609.0, + 294.0, + 609.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 606.0, + 1401.0, + 606.0, + 1401.0, + 638.0, + 296.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1885.0, + 1405.0, + 1885.0, + 1405.0, + 1918.0, + 296.0, + 1918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1914.0, + 1405.0, + 1914.0, + 1405.0, + 1950.0, + 293.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1947.0, + 1405.0, + 1947.0, + 1405.0, + 1979.0, + 294.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1975.0, + 303.0, + 1975.0, + 303.0, + 2009.0, + 294.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 420.0, + 1975.0, + 1403.0, + 1975.0, + 1403.0, + 2009.0, + 420.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 649.0, + 1404.0, + 649.0, + 1404.0, + 682.0, + 296.0, + 682.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 679.0, + 1404.0, + 679.0, + 1404.0, + 712.0, + 296.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 708.0, + 1407.0, + 708.0, + 1407.0, + 744.0, + 292.0, + 744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 739.0, + 1406.0, + 739.0, + 1406.0, + 776.0, + 294.0, + 776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 770.0, + 1363.0, + 770.0, + 1363.0, + 805.0, + 293.0, + 805.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 6, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1580, + 1405, + 1580, + 1405, + 1704, + 297, + 1704 + ], + "score": 0.977 + }, + { + "category_id": 5, + "poly": [ + 303, + 1124, + 1403, + 1124, + 1403, + 1252, + 303, + 1252 + ], + "score": 0.977, + "html": "
MethodAtom Stable (%)Mol Stable (%)MethodAtom Stable (%)Mol Stable (%)
Force (7),k=798.8±0.184.5±0.2Force (6)98.7±0.183.1±0.2
Force (7),k =598.8±0.184.6±0.3Force (6) w/o.bond98.7±0.182.5±0.1
Force (7), k = 398.8±0.183.9±0.3Force (6) w/o.angle98.7±0.182.4±0.2
Force (7), k =198.8±0.182.7±0.3Force (6) w/o.Long-range98.7±0.182.7±0.2
" + }, + { + "category_id": 1, + "poly": [ + 299, + 1884, + 1402, + 1884, + 1402, + 2007, + 299, + 2007 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 1716, + 1404, + 1716, + 1404, + 1871, + 298, + 1871 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 297, + 692, + 1404, + 692, + 1404, + 877, + 297, + 877 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 297, + 1293, + 1404, + 1293, + 1404, + 1477, + 297, + 1477 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 297, + 889, + 1404, + 889, + 1404, + 1044, + 297, + 1044 + ], + "score": 0.972 + }, + { + "category_id": 5, + "poly": [ + 302, + 535, + 1402, + 535, + 1402, + 655, + 302, + 655 + ], + "score": 0.972, + "html": "
Time Step
50100500
Atom Stable (%) Mol Stable (%)Atom Stable (%)Mol Stable (%)Atom Stable (%) Mol Stable (%)
EGM97.0±0.1 66.4±0.297.3±0.169.8±0.298.5±0.1 81.2±0.1
Bridge +Force (7)97.3±0.169.2±0.2 97.9±0.172.3±0.298.7±0.1 83.7±0.1
" + }, + { + "category_id": 0, + "poly": [ + 299, + 1520, + 814, + 1520, + 814, + 1555, + 299, + 1555 + ], + "score": 0.918 + }, + { + "category_id": 6, + "poly": [ + 401, + 1079, + 1286, + 1079, + 1286, + 1108, + 401, + 1108 + ], + "score": 0.906 + }, + { + "category_id": 6, + "poly": [ + 410, + 488, + 1281, + 488, + 1281, + 518, + 410, + 518 + ], + "score": 0.889 + }, + { + "category_id": 3, + "poly": [ + 303, + 259, + 1394, + 259, + 1394, + 390, + 303, + 390 + ], + "score": 0.859 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 858, + 2062, + 858, + 2085, + 840, + 2085 + ], + "score": 0.805 + }, + { + "category_id": 4, + "poly": [ + 322, + 434, + 1367, + 434, + 1367, + 467, + 322, + 467 + ], + "score": 0.627 + }, + { + "category_id": 6, + "poly": [ + 322, + 434, + 1367, + 434, + 1367, + 467, + 322, + 467 + ], + "score": 0.341 + }, + { + "category_id": 4, + "poly": [ + 810, + 210, + 900, + 210, + 900, + 234, + 810, + 234 + ], + "score": 0.132 + }, + { + "category_id": 13, + "poly": [ + 1057, + 1356, + 1130, + 1356, + 1130, + 1383, + 1057, + 1383 + ], + "score": 0.89, + "latex": "k = 5" + }, + { + "category_id": 13, + "poly": [ + 551, + 1355, + 628, + 1355, + 628, + 1384, + 551, + 1384 + ], + "score": 0.89, + "latex": "k = 7" + }, + { + "category_id": 13, + "poly": [ + 463, + 1355, + 537, + 1355, + 537, + 1385, + 463, + 1385 + ], + "score": 0.89, + "latex": "k \\leq 5" + }, + { + "category_id": 13, + "poly": [ + 890, + 1808, + 931, + 1808, + 931, + 1838, + 890, + 1838 + ], + "score": 0.88, + "latex": "5 \\%" + }, + { + "category_id": 13, + "poly": [ + 910, + 434, + 941, + 434, + 941, + 462, + 910, + 462 + ], + "score": 0.87, + "latex": "\\mathbb { P } ^ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 730, + 846, + 770, + 846, + 770, + 875, + 730, + 875 + ], + "score": 0.87, + "latex": "8 \\%" + }, + { + "category_id": 13, + "poly": [ + 720, + 1808, + 775, + 1808, + 775, + 1839, + 720, + 1839 + ], + "score": 0.87, + "latex": "8 0 \\%" + }, + { + "category_id": 13, + "poly": [ + 786, + 1808, + 841, + 1808, + 841, + 1838, + 786, + 1838 + ], + "score": 0.87, + "latex": "1 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 1111, + 846, + 1152, + 846, + 1152, + 874, + 1111, + 874 + ], + "score": 0.85, + "latex": "3 \\%" + }, + { + "category_id": 13, + "poly": [ + 1229, + 1356, + 1249, + 1356, + 1249, + 1383, + 1229, + 1383 + ], + "score": 0.79, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 370, + 1356, + 390, + 1356, + 390, + 1383, + 370, + 1383 + ], + "score": 0.76, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 596, + 582, + 627, + 582, + 627, + 605, + 596, + 605 + ], + "score": 0.42, + "latex": "\\overline { { ( \\% ) } }" + }, + { + "category_id": 13, + "poly": [ + 902, + 583, + 933, + 583, + 933, + 605, + 902, + 605 + ], + "score": 0.4, + "latex": "\\overline { { ( \\% ) } }" + }, + { + "category_id": 13, + "poly": [ + 374, + 629, + 391, + 629, + 391, + 647, + 374, + 647 + ], + "score": 0.34, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 1208, + 582, + 1238, + 582, + 1238, + 604, + 1208, + 604 + ], + "score": 0.31, + "latex": "\\overline { { ( \\% ) } }" + }, + { + "category_id": 13, + "poly": [ + 742, + 582, + 773, + 582, + 773, + 605, + 742, + 605 + ], + "score": 0.26, + "latex": "\\overline { { ( \\% ) } }" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1521.0, + 814.0, + 1521.0, + 814.0, + 1558.0, + 295.0, + 1558.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 402.0, + 1077.0, + 1291.0, + 1077.0, + 1291.0, + 1112.0, + 402.0, + 1112.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 410.0, + 484.0, + 1281.0, + 484.0, + 1281.0, + 523.0, + 410.0, + 523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 292.0, + 407.0, + 292.0, + 407.0, + 333.0, + 367.0, + 333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 434.0, + 266.0, + 908.0, + 266.0, + 908.0, + 369.0, + 434.0, + 369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2090.0, + 838.0, + 2090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 433.0, + 909.0, + 433.0, + 909.0, + 468.0, + 329.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 942.0, + 433.0, + 1368.0, + 433.0, + 1368.0, + 468.0, + 942.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 433.0, + 909.0, + 433.0, + 909.0, + 468.0, + 329.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 942.0, + 433.0, + 1368.0, + 433.0, + 1368.0, + 468.0, + 942.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 807.0, + 203.0, + 903.0, + 203.0, + 903.0, + 241.0, + 807.0, + 241.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1579.0, + 1405.0, + 1579.0, + 1405.0, + 1616.0, + 292.0, + 1616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1610.0, + 1407.0, + 1610.0, + 1407.0, + 1647.0, + 294.0, + 1647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1644.0, + 1408.0, + 1644.0, + 1408.0, + 1677.0, + 295.0, + 1677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1671.0, + 1408.0, + 1671.0, + 1408.0, + 1707.0, + 294.0, + 1707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1883.0, + 1406.0, + 1883.0, + 1406.0, + 1920.0, + 293.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1913.0, + 1406.0, + 1913.0, + 1406.0, + 1949.0, + 294.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1946.0, + 1406.0, + 1946.0, + 1406.0, + 1980.0, + 293.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1977.0, + 1194.0, + 1977.0, + 1194.0, + 2009.0, + 295.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1718.0, + 1408.0, + 1718.0, + 1408.0, + 1751.0, + 294.0, + 1751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1748.0, + 1405.0, + 1748.0, + 1405.0, + 1782.0, + 296.0, + 1782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1775.0, + 1408.0, + 1775.0, + 1408.0, + 1815.0, + 293.0, + 1815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1805.0, + 719.0, + 1805.0, + 719.0, + 1845.0, + 292.0, + 1845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 776.0, + 1805.0, + 785.0, + 1805.0, + 785.0, + 1845.0, + 776.0, + 1845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 842.0, + 1805.0, + 889.0, + 1805.0, + 889.0, + 1845.0, + 842.0, + 1845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 932.0, + 1805.0, + 1407.0, + 1805.0, + 1407.0, + 1845.0, + 932.0, + 1845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1836.0, + 455.0, + 1836.0, + 455.0, + 1873.0, + 294.0, + 1873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 693.0, + 1406.0, + 693.0, + 1406.0, + 726.0, + 293.0, + 726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 722.0, + 1406.0, + 722.0, + 1406.0, + 759.0, + 294.0, + 759.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 754.0, + 1406.0, + 754.0, + 1406.0, + 790.0, + 295.0, + 790.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 784.0, + 1406.0, + 784.0, + 1406.0, + 820.0, + 295.0, + 820.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 815.0, + 1409.0, + 815.0, + 1409.0, + 848.0, + 292.0, + 848.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 844.0, + 729.0, + 844.0, + 729.0, + 881.0, + 294.0, + 881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 771.0, + 844.0, + 1110.0, + 844.0, + 1110.0, + 881.0, + 771.0, + 881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1153.0, + 844.0, + 1312.0, + 844.0, + 1312.0, + 881.0, + 1153.0, + 881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1294.0, + 1405.0, + 1294.0, + 1405.0, + 1329.0, + 294.0, + 1329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1324.0, + 1404.0, + 1324.0, + 1404.0, + 1360.0, + 295.0, + 1360.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1352.0, + 369.0, + 1352.0, + 369.0, + 1390.0, + 294.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 391.0, + 1352.0, + 462.0, + 1352.0, + 462.0, + 1390.0, + 391.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1352.0, + 550.0, + 1352.0, + 550.0, + 1390.0, + 538.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 629.0, + 1352.0, + 1056.0, + 1352.0, + 1056.0, + 1390.0, + 629.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1131.0, + 1352.0, + 1228.0, + 1352.0, + 1228.0, + 1390.0, + 1131.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1250.0, + 1352.0, + 1406.0, + 1352.0, + 1406.0, + 1390.0, + 1250.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1384.0, + 1407.0, + 1384.0, + 1407.0, + 1419.0, + 295.0, + 1419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1415.0, + 1405.0, + 1415.0, + 1405.0, + 1451.0, + 295.0, + 1451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1446.0, + 896.0, + 1446.0, + 896.0, + 1480.0, + 295.0, + 1480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 891.0, + 1404.0, + 891.0, + 1404.0, + 924.0, + 296.0, + 924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 920.0, + 1404.0, + 920.0, + 1404.0, + 956.0, + 294.0, + 956.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 951.0, + 1404.0, + 951.0, + 1404.0, + 986.0, + 294.0, + 986.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 977.0, + 1406.0, + 977.0, + 1406.0, + 1021.0, + 292.0, + 1021.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1009.0, + 1087.0, + 1009.0, + 1087.0, + 1049.0, + 294.0, + 1049.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 7, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 374, + 1224, + 1327, + 1224, + 1327, + 1531, + 374, + 1531 + ], + "score": 0.984, + "html": "
10 Steps100 Steps
MMD↓COV↑MMD↓COV↑
CDEMDCDEMDCDEMDCDEMD
ChairDiffusion [28]14.013.2332.7229.3612.321.7947.4147.59
Bridge13.042.1446.0142.5912.471.8547.8347.13
+Riesz12.841.9547.2144.3112.311.8248.1447.42
+ Statistic12.651.8447.5845.2312.251.7848.3947.56
AirplaneDiffusion [28] Bridge3.711.3143.1239.943.281.0448.7446.38
3.441.2446.9043.463.371.0847.1146.17
+Riesz3.391.2047.1143.123.241.0948.6246.23
+ Statistic3.301.1247.0244.673.241.0648.5346.73
" + }, + { + "category_id": 1, + "poly": [ + 297, + 1611, + 1404, + 1611, + 1404, + 1826, + 297, + 1826 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 873, + 1405, + 873, + 1405, + 1149, + 297, + 1149 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 202, + 1405, + 202, + 1405, + 446, + 297, + 446 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 299, + 1839, + 1404, + 1839, + 1404, + 1931, + 299, + 1931 + ], + "score": 0.97 + }, + { + "category_id": 3, + "poly": [ + 315, + 488, + 1392, + 488, + 1392, + 706, + 315, + 706 + ], + "score": 0.964 + }, + { + "category_id": 4, + "poly": [ + 298, + 739, + 1404, + 739, + 1404, + 852, + 298, + 852 + ], + "score": 0.962 + }, + { + "category_id": 1, + "poly": [ + 297, + 1945, + 1404, + 1945, + 1404, + 2008, + 297, + 2008 + ], + "score": 0.935 + }, + { + "category_id": 0, + "poly": [ + 298, + 1558, + 745, + 1558, + 745, + 1596, + 298, + 1596 + ], + "score": 0.92 + }, + { + "category_id": 6, + "poly": [ + 378, + 1176, + 1310, + 1176, + 1310, + 1210, + 378, + 1210 + ], + "score": 0.908 + }, + { + "category_id": 2, + "poly": [ + 840, + 2061, + 858, + 2061, + 858, + 2084, + 840, + 2084 + ], + "score": 0.802 + }, + { + "category_id": 13, + "poly": [ + 985, + 262, + 1094, + 262, + 1094, + 293, + 985, + 293 + ], + "score": 0.91, + "latex": "2 \\times 1 0 ^ { - 3 }" + }, + { + "category_id": 13, + "poly": [ + 1175, + 769, + 1345, + 769, + 1345, + 797, + 1175, + 797 + ], + "score": 0.91, + "latex": "( f _ { t } = - \\nabla E _ { k n n } )" + }, + { + "category_id": 13, + "poly": [ + 746, + 769, + 926, + 769, + 926, + 797, + 746, + 797 + ], + "score": 0.9, + "latex": "( f _ { t } = - \\nabla E _ { \\mathrm { { R i e s z } } } )" + }, + { + "category_id": 13, + "poly": [ + 607, + 293, + 667, + 293, + 667, + 323, + 607, + 323 + ], + "score": 0.9, + "latex": "1 0 ^ { - 4 }" + }, + { + "category_id": 13, + "poly": [ + 377, + 769, + 458, + 769, + 458, + 797, + 377, + 797 + ], + "score": 0.87, + "latex": "f _ { t } = 0 ," + }, + { + "category_id": 13, + "poly": [ + 1001, + 1176, + 1040, + 1176, + 1040, + 1205, + 1001, + 1205 + ], + "score": 0.87, + "latex": "\\mathrm { 1 0 ^ { 3 } }" + }, + { + "category_id": 13, + "poly": [ + 793, + 298, + 815, + 298, + 815, + 323, + 793, + 323 + ], + "score": 0.77, + "latex": "\\alpha" + }, + { + "category_id": 13, + "poly": [ + 502, + 1477, + 521, + 1477, + 521, + 1497, + 502, + 1497 + ], + "score": 0.42, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 502, + 1365, + 521, + 1365, + 521, + 1384, + 502, + 1384 + ], + "score": 0.33, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 502, + 1502, + 521, + 1502, + 521, + 1523, + 502, + 1523 + ], + "score": 0.27, + "latex": "^ +" + }, + { + "category_id": 15, + "poly": [ + 638.0, + 509.0, + 766.0, + 509.0, + 766.0, + 608.0, + 638.0, + 608.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 520.0, + 523.0, + 573.0, + 523.0, + 573.0, + 560.0, + 520.0, + 560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 703.0, + 617.0, + 711.0, + 617.0, + 711.0, + 632.0, + 703.0, + 632.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 675.0, + 497.0, + 675.0, + 497.0, + 708.0, + 404.0, + 708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 690.0, + 672.0, + 765.0, + 672.0, + 765.0, + 711.0, + 690.0, + 711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 938.0, + 675.0, + 1071.0, + 675.0, + 1071.0, + 708.0, + 938.0, + 708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1193.0, + 675.0, + 1349.0, + 675.0, + 1349.0, + 708.0, + 1193.0, + 708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 514.0, + 415.0, + 514.0, + 415.0, + 612.0, + 393.0, + 612.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 738.0, + 1404.0, + 738.0, + 1404.0, + 772.0, + 296.0, + 772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 768.0, + 376.0, + 768.0, + 376.0, + 801.0, + 296.0, + 801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 459.0, + 768.0, + 745.0, + 768.0, + 745.0, + 801.0, + 459.0, + 801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 927.0, + 768.0, + 1174.0, + 768.0, + 1174.0, + 801.0, + 927.0, + 801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1346.0, + 768.0, + 1405.0, + 768.0, + 1405.0, + 801.0, + 1346.0, + 801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 793.0, + 1405.0, + 793.0, + 1405.0, + 829.0, + 295.0, + 829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 822.0, + 811.0, + 822.0, + 811.0, + 857.0, + 293.0, + 857.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1553.0, + 747.0, + 1553.0, + 747.0, + 1601.0, + 291.0, + 1601.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 376.0, + 1172.0, + 1000.0, + 1172.0, + 1000.0, + 1214.0, + 376.0, + 1214.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1041.0, + 1172.0, + 1313.0, + 1172.0, + 1313.0, + 1214.0, + 1041.0, + 1214.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2060.0, + 860.0, + 2060.0, + 860.0, + 2090.0, + 839.0, + 2090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1611.0, + 1404.0, + 1611.0, + 1404.0, + 1646.0, + 296.0, + 1646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1643.0, + 1404.0, + 1643.0, + 1404.0, + 1678.0, + 295.0, + 1678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1673.0, + 1404.0, + 1673.0, + 1404.0, + 1707.0, + 295.0, + 1707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1705.0, + 1404.0, + 1705.0, + 1404.0, + 1736.0, + 296.0, + 1736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1731.0, + 1404.0, + 1731.0, + 1404.0, + 1768.0, + 294.0, + 1768.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1763.0, + 1406.0, + 1763.0, + 1406.0, + 1801.0, + 294.0, + 1801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1792.0, + 798.0, + 1792.0, + 798.0, + 1831.0, + 294.0, + 1831.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 874.0, + 1403.0, + 874.0, + 1403.0, + 907.0, + 296.0, + 907.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 905.0, + 1405.0, + 905.0, + 1405.0, + 939.0, + 294.0, + 939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 933.0, + 1408.0, + 933.0, + 1408.0, + 973.0, + 294.0, + 973.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 965.0, + 1406.0, + 965.0, + 1406.0, + 1002.0, + 294.0, + 1002.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 994.0, + 1406.0, + 994.0, + 1406.0, + 1033.0, + 294.0, + 1033.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1021.0, + 1405.0, + 1021.0, + 1405.0, + 1066.0, + 294.0, + 1066.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1057.0, + 1406.0, + 1057.0, + 1406.0, + 1093.0, + 295.0, + 1093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1084.0, + 1407.0, + 1084.0, + 1407.0, + 1123.0, + 291.0, + 1123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1117.0, + 743.0, + 1117.0, + 743.0, + 1153.0, + 295.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 203.0, + 1403.0, + 203.0, + 1403.0, + 237.0, + 296.0, + 237.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 236.0, + 1403.0, + 236.0, + 1403.0, + 265.0, + 296.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 263.0, + 984.0, + 263.0, + 984.0, + 297.0, + 295.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1095.0, + 263.0, + 1405.0, + 263.0, + 1405.0, + 297.0, + 1095.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 292.0, + 606.0, + 292.0, + 606.0, + 329.0, + 294.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 668.0, + 292.0, + 792.0, + 292.0, + 792.0, + 329.0, + 668.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 816.0, + 292.0, + 1406.0, + 292.0, + 1406.0, + 329.0, + 816.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 325.0, + 1403.0, + 325.0, + 1403.0, + 359.0, + 294.0, + 359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 353.0, + 1403.0, + 353.0, + 1403.0, + 389.0, + 292.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 381.0, + 1406.0, + 381.0, + 1406.0, + 420.0, + 292.0, + 420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 414.0, + 414.0, + 414.0, + 414.0, + 448.0, + 294.0, + 448.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1836.0, + 1406.0, + 1836.0, + 1406.0, + 1877.0, + 292.0, + 1877.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1869.0, + 1403.0, + 1869.0, + 1403.0, + 1903.0, + 295.0, + 1903.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1902.0, + 1103.0, + 1902.0, + 1103.0, + 1932.0, + 297.0, + 1932.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1943.0, + 1404.0, + 1943.0, + 1404.0, + 1982.0, + 294.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1970.0, + 1404.0, + 1970.0, + 1404.0, + 2013.0, + 294.0, + 2013.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 8, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 299, + 297, + 1406, + 297, + 1406, + 420, + 299, + 420 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 294, + 202, + 1402, + 202, + 1402, + 265, + 294, + 265 + ], + "score": 0.945 + }, + { + "category_id": 2, + "poly": [ + 836, + 2062, + 865, + 2062, + 865, + 2085, + 836, + 2085 + ], + "score": 0.84 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 871.0, + 2058.0, + 871.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 296.0, + 1409.0, + 296.0, + 1409.0, + 331.0, + 293.0, + 331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 325.0, + 1411.0, + 325.0, + 1411.0, + 362.0, + 292.0, + 362.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 357.0, + 1407.0, + 357.0, + 1407.0, + 392.0, + 293.0, + 392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 389.0, + 1060.0, + 389.0, + 1060.0, + 424.0, + 294.0, + 424.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 203.0, + 1404.0, + 203.0, + 1404.0, + 239.0, + 296.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 231.0, + 963.0, + 231.0, + 963.0, + 270.0, + 293.0, + 270.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 9, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 0, + "poly": [ + 299, + 200, + 455, + 200, + 455, + 235, + 299, + 235 + ], + "score": 0.899 + }, + { + "category_id": 1, + "poly": [ + 296, + 139, + 1409, + 139, + 1409, + 2020, + 296, + 2020 + ], + "score": 0.864 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 863, + 2061, + 863, + 2086, + 836, + 2086 + ], + "score": 0.78 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 863, + 2061, + 863, + 2086, + 836, + 2086 + ], + "score": 0.18 + }, + { + "category_id": 15, + "poly": [ + 295.0, + 196.0, + 460.0, + 196.0, + 460.0, + 241.0, + 295.0, + 241.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2098.0, + 832.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2098.0, + 832.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 196.0, + 463.0, + 196.0, + 463.0, + 240.0, + 293.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 252.0, + 1411.0, + 252.0, + 1411.0, + 291.0, + 307.0, + 291.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 285.0, + 1404.0, + 285.0, + 1404.0, + 318.0, + 353.0, + 318.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 314.0, + 779.0, + 314.0, + 779.0, + 353.0, + 353.0, + 353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 303.0, + 363.0, + 1411.0, + 363.0, + 1411.0, + 407.0, + 303.0, + 407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 394.0, + 1408.0, + 394.0, + 1408.0, + 438.0, + 349.0, + 438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 423.0, + 606.0, + 423.0, + 606.0, + 467.0, + 354.0, + 467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 481.0, + 1406.0, + 481.0, + 1406.0, + 520.0, + 307.0, + 520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 512.0, + 1408.0, + 512.0, + 1408.0, + 551.0, + 351.0, + 551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 543.0, + 659.0, + 543.0, + 659.0, + 580.0, + 351.0, + 580.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 597.0, + 1408.0, + 597.0, + 1408.0, + 636.0, + 307.0, + 636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 628.0, + 1331.0, + 628.0, + 1331.0, + 667.0, + 353.0, + 667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 305.0, + 679.0, + 1408.0, + 679.0, + 1408.0, + 720.0, + 305.0, + 720.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 710.0, + 1404.0, + 710.0, + 1404.0, + 749.0, + 351.0, + 749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 743.0, + 951.0, + 743.0, + 951.0, + 782.0, + 353.0, + 782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 303.0, + 793.0, + 1411.0, + 793.0, + 1411.0, + 836.0, + 303.0, + 836.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 824.0, + 1406.0, + 824.0, + 1406.0, + 865.0, + 351.0, + 865.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 857.0, + 1013.0, + 857.0, + 1013.0, + 896.0, + 353.0, + 896.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 910.0, + 1402.0, + 910.0, + 1402.0, + 949.0, + 307.0, + 949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 941.0, + 1396.0, + 941.0, + 1396.0, + 980.0, + 353.0, + 980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 995.0, + 1406.0, + 995.0, + 1406.0, + 1034.0, + 307.0, + 1034.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1026.0, + 1406.0, + 1026.0, + 1406.0, + 1065.0, + 351.0, + 1065.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1059.0, + 1408.0, + 1059.0, + 1408.0, + 1092.0, + 356.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1090.0, + 1141.0, + 1090.0, + 1141.0, + 1123.0, + 353.0, + 1123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 1139.0, + 1406.0, + 1139.0, + 1406.0, + 1179.0, + 307.0, + 1179.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1170.0, + 1406.0, + 1170.0, + 1406.0, + 1210.0, + 353.0, + 1210.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1201.0, + 699.0, + 1201.0, + 699.0, + 1240.0, + 351.0, + 1240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1253.0, + 1408.0, + 1253.0, + 1408.0, + 1296.0, + 290.0, + 1296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1286.0, + 596.0, + 1286.0, + 596.0, + 1325.0, + 353.0, + 1325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 1335.0, + 1408.0, + 1335.0, + 1408.0, + 1383.0, + 288.0, + 1383.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1370.0, + 1356.0, + 1370.0, + 1356.0, + 1410.0, + 353.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1420.0, + 1408.0, + 1420.0, + 1408.0, + 1467.0, + 290.0, + 1467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1455.0, + 1406.0, + 1455.0, + 1406.0, + 1494.0, + 351.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1486.0, + 1404.0, + 1486.0, + 1404.0, + 1519.0, + 356.0, + 1519.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1515.0, + 1055.0, + 1515.0, + 1055.0, + 1554.0, + 351.0, + 1554.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1569.0, + 1406.0, + 1569.0, + 1406.0, + 1608.0, + 294.0, + 1608.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1600.0, + 1411.0, + 1600.0, + 1411.0, + 1639.0, + 353.0, + 1639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1624.0, + 1409.0, + 1624.0, + 1409.0, + 1674.0, + 347.0, + 1674.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1663.0, + 1040.0, + 1663.0, + 1040.0, + 1696.0, + 353.0, + 1696.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1717.0, + 1406.0, + 1717.0, + 1406.0, + 1750.0, + 294.0, + 1750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1748.0, + 1404.0, + 1748.0, + 1404.0, + 1781.0, + 356.0, + 1781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1777.0, + 741.0, + 1777.0, + 741.0, + 1812.0, + 349.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1829.0, + 1406.0, + 1829.0, + 1406.0, + 1868.0, + 294.0, + 1868.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1857.0, + 1406.0, + 1857.0, + 1406.0, + 1899.0, + 351.0, + 1899.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1893.0, + 996.0, + 1893.0, + 996.0, + 1926.0, + 356.0, + 1926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1944.0, + 1406.0, + 1944.0, + 1406.0, + 1983.0, + 294.0, + 1983.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1975.0, + 1170.0, + 1975.0, + 1170.0, + 2014.0, + 353.0, + 2014.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 10, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 866, + 2061, + 866, + 2086, + 836, + 2086 + ], + "score": 0.835 + }, + { + "category_id": 1, + "poly": [ + 290, + 153, + 1410, + 153, + 1410, + 2020, + 290, + 2020 + ], + "score": 0.542 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 869.0, + 2058.0, + 869.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 203.0, + 1408.0, + 203.0, + 1408.0, + 242.0, + 294.0, + 242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 232.0, + 431.0, + 232.0, + 431.0, + 273.0, + 350.0, + 273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 289.0, + 1406.0, + 289.0, + 1406.0, + 328.0, + 294.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 320.0, + 1081.0, + 320.0, + 1081.0, + 359.0, + 352.0, + 359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 373.0, + 1410.0, + 373.0, + 1410.0, + 419.0, + 290.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 406.0, + 1257.0, + 406.0, + 1257.0, + 445.0, + 354.0, + 445.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 466.0, + 1404.0, + 466.0, + 1404.0, + 498.0, + 296.0, + 498.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 494.0, + 1396.0, + 494.0, + 1396.0, + 533.0, + 354.0, + 533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 546.0, + 1408.0, + 546.0, + 1408.0, + 591.0, + 294.0, + 591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 580.0, + 1408.0, + 580.0, + 1408.0, + 619.0, + 354.0, + 619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 613.0, + 1169.0, + 613.0, + 1169.0, + 646.0, + 354.0, + 646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 666.0, + 1410.0, + 666.0, + 1410.0, + 705.0, + 294.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 695.0, + 1408.0, + 695.0, + 1408.0, + 734.0, + 354.0, + 734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 728.0, + 1410.0, + 728.0, + 1410.0, + 767.0, + 354.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 777.0, + 1410.0, + 777.0, + 1410.0, + 826.0, + 290.0, + 826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 814.0, + 1408.0, + 814.0, + 1408.0, + 853.0, + 354.0, + 853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 843.0, + 431.0, + 843.0, + 431.0, + 884.0, + 352.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 900.0, + 1404.0, + 900.0, + 1404.0, + 939.0, + 296.0, + 939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 931.0, + 1404.0, + 931.0, + 1404.0, + 970.0, + 352.0, + 970.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 955.0, + 511.0, + 955.0, + 511.0, + 1001.0, + 353.0, + 1001.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1017.0, + 1408.0, + 1017.0, + 1408.0, + 1056.0, + 296.0, + 1056.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1048.0, + 1245.0, + 1048.0, + 1245.0, + 1086.0, + 352.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1101.0, + 1410.0, + 1101.0, + 1410.0, + 1144.0, + 290.0, + 1144.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1127.0, + 1408.0, + 1127.0, + 1408.0, + 1177.0, + 350.0, + 1177.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1164.0, + 1400.0, + 1164.0, + 1400.0, + 1203.0, + 354.0, + 1203.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1222.0, + 1406.0, + 1222.0, + 1406.0, + 1261.0, + 296.0, + 1261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1250.0, + 1131.0, + 1250.0, + 1131.0, + 1289.0, + 350.0, + 1289.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1306.0, + 1404.0, + 1306.0, + 1404.0, + 1345.0, + 294.0, + 1345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1339.0, + 1408.0, + 1339.0, + 1408.0, + 1377.0, + 354.0, + 1377.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1365.0, + 489.0, + 1365.0, + 489.0, + 1404.0, + 350.0, + 1404.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1423.0, + 1410.0, + 1423.0, + 1410.0, + 1468.0, + 292.0, + 1468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1451.0, + 1408.0, + 1451.0, + 1408.0, + 1498.0, + 350.0, + 1498.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1484.0, + 643.0, + 1484.0, + 643.0, + 1523.0, + 354.0, + 1523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1541.0, + 1406.0, + 1541.0, + 1406.0, + 1580.0, + 294.0, + 1580.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1572.0, + 1410.0, + 1572.0, + 1410.0, + 1611.0, + 354.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1627.0, + 1410.0, + 1627.0, + 1410.0, + 1666.0, + 294.0, + 1666.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1658.0, + 1406.0, + 1658.0, + 1406.0, + 1699.0, + 352.0, + 1699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1689.0, + 914.0, + 1689.0, + 914.0, + 1728.0, + 354.0, + 1728.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1740.0, + 1406.0, + 1740.0, + 1406.0, + 1789.0, + 290.0, + 1789.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1775.0, + 1394.0, + 1775.0, + 1394.0, + 1816.0, + 350.0, + 1816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1832.0, + 1265.0, + 1832.0, + 1265.0, + 1871.0, + 294.0, + 1871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1881.0, + 1408.0, + 1881.0, + 1408.0, + 1931.0, + 290.0, + 1931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1918.0, + 854.0, + 1918.0, + 854.0, + 1955.0, + 352.0, + 1955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1974.0, + 1093.0, + 1974.0, + 1093.0, + 2013.0, + 294.0, + 2013.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 11, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.837 + }, + { + "category_id": 1, + "poly": [ + 293, + 113, + 1411, + 113, + 1411, + 1871, + 293, + 1871 + ], + "score": 0.62 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2057.0, + 870.0, + 2057.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 200.0, + 1410.0, + 200.0, + 1410.0, + 243.0, + 292.0, + 243.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 357.0, + 237.0, + 1410.0, + 237.0, + 1410.0, + 268.0, + 357.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 258.0, + 430.0, + 258.0, + 430.0, + 302.0, + 353.0, + 302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 311.0, + 1412.0, + 311.0, + 1412.0, + 361.0, + 289.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 348.0, + 1219.0, + 348.0, + 1219.0, + 384.0, + 353.0, + 384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 398.0, + 1406.0, + 398.0, + 1406.0, + 435.0, + 296.0, + 435.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 429.0, + 1284.0, + 429.0, + 1284.0, + 466.0, + 353.0, + 466.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 479.0, + 1408.0, + 479.0, + 1408.0, + 518.0, + 291.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 514.0, + 1025.0, + 514.0, + 1025.0, + 551.0, + 353.0, + 551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 560.0, + 1410.0, + 560.0, + 1410.0, + 605.0, + 291.0, + 605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 593.0, + 1404.0, + 593.0, + 1404.0, + 630.0, + 353.0, + 630.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 622.0, + 840.0, + 622.0, + 840.0, + 663.0, + 350.0, + 663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 677.0, + 1406.0, + 677.0, + 1406.0, + 713.0, + 294.0, + 713.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 706.0, + 740.0, + 706.0, + 740.0, + 746.0, + 348.0, + 746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 758.0, + 1410.0, + 758.0, + 1410.0, + 800.0, + 292.0, + 800.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 791.0, + 1404.0, + 791.0, + 1404.0, + 827.0, + 353.0, + 827.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 843.0, + 1406.0, + 843.0, + 1406.0, + 880.0, + 294.0, + 880.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 872.0, + 1408.0, + 872.0, + 1408.0, + 915.0, + 353.0, + 915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 903.0, + 741.0, + 903.0, + 741.0, + 942.0, + 349.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 957.0, + 1404.0, + 957.0, + 1404.0, + 994.0, + 296.0, + 994.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 986.0, + 1040.0, + 986.0, + 1040.0, + 1023.0, + 353.0, + 1023.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1037.0, + 1408.0, + 1037.0, + 1408.0, + 1079.0, + 294.0, + 1079.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1067.0, + 1410.0, + 1067.0, + 1410.0, + 1110.0, + 351.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1098.0, + 519.0, + 1098.0, + 519.0, + 1135.0, + 353.0, + 1135.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1149.0, + 1410.0, + 1149.0, + 1410.0, + 1189.0, + 291.0, + 1189.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1184.0, + 1408.0, + 1184.0, + 1408.0, + 1220.0, + 353.0, + 1220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1206.0, + 432.0, + 1206.0, + 432.0, + 1252.0, + 353.0, + 1252.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1263.0, + 1404.0, + 1263.0, + 1404.0, + 1300.0, + 294.0, + 1300.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1294.0, + 1294.0, + 1294.0, + 1294.0, + 1331.0, + 353.0, + 1331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1348.0, + 1406.0, + 1348.0, + 1406.0, + 1385.0, + 296.0, + 1385.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1379.0, + 1366.0, + 1379.0, + 1366.0, + 1416.0, + 355.0, + 1416.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1427.0, + 1408.0, + 1427.0, + 1408.0, + 1468.0, + 291.0, + 1468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1456.0, + 1408.0, + 1456.0, + 1408.0, + 1501.0, + 350.0, + 1501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1489.0, + 1166.0, + 1489.0, + 1166.0, + 1528.0, + 351.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1544.0, + 1406.0, + 1544.0, + 1406.0, + 1580.0, + 294.0, + 1580.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1569.0, + 1410.0, + 1569.0, + 1410.0, + 1615.0, + 348.0, + 1615.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1606.0, + 759.0, + 1606.0, + 759.0, + 1642.0, + 353.0, + 1642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1654.0, + 1410.0, + 1654.0, + 1410.0, + 1695.0, + 291.0, + 1695.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1681.0, + 1408.0, + 1681.0, + 1408.0, + 1728.0, + 350.0, + 1728.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1718.0, + 1282.0, + 1718.0, + 1282.0, + 1755.0, + 353.0, + 1755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1766.0, + 1408.0, + 1766.0, + 1408.0, + 1809.0, + 292.0, + 1809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1795.0, + 1412.0, + 1795.0, + 1412.0, + 1840.0, + 350.0, + 1840.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1828.0, + 638.0, + 1828.0, + 638.0, + 1867.0, + 351.0, + 1867.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 12, + "width": 1700, + "height": 2200 + } + } +] \ No newline at end of file diff --git a/parse/dev/T__V3uLix7V/T__V3uLix7V_middle.json b/parse/dev/T__V3uLix7V/T__V3uLix7V_middle.json new file mode 100644 index 0000000000000000000000000000000000000000..79cee1e3ba745a74c492cf36740d1102b109501f --- /dev/null +++ b/parse/dev/T__V3uLix7V/T__V3uLix7V_middle.json @@ -0,0 +1,48738 @@ +{ + "pdf_info": [ + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 78, + 503, + 116 + ], + "lines": [ + { + "bbox": [ + 106, + 78, + 505, + 98 + ], + "spans": [ + { + "bbox": [ + 106, + 78, + 505, + 98 + ], + "score": 1.0, + "content": "REGIONVIT: REGIONAL-TO-LOCAL ATTENTION FOR", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 97, + 289, + 118 + ], + "spans": [ + { + "bbox": [ + 106, + 97, + 289, + 118 + ], + "score": 1.0, + "content": "VISION TRANSFORMERS", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 112, + 135, + 423, + 168 + ], + "lines": [ + { + "bbox": [ + 112, + 135, + 360, + 147 + ], + "spans": [ + { + "bbox": [ + 112, + 135, + 360, + 147 + ], + "score": 1.0, + "content": "Chun-Fu (Richard) Chen, Rameswar Panda, Quanfu Fan", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 112, + 146, + 218, + 158 + ], + "spans": [ + { + "bbox": [ + 112, + 146, + 218, + 158 + ], + "score": 1.0, + "content": "MIT-IBM Watson AI Lab", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 113, + 158, + 425, + 169 + ], + "spans": [ + { + "bbox": [ + 113, + 158, + 425, + 169 + ], + "score": 1.0, + "content": "chenrich@us.ibm.com, rpanda@ibm.com, qfan@us.ibm.com", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3 + }, + { + "type": "title", + "bbox": [ + 278, + 198, + 333, + 209 + ], + "lines": [ + { + "bbox": [ + 276, + 196, + 336, + 212 + ], + "spans": [ + { + "bbox": [ + 276, + 196, + 336, + 212 + ], + "score": 1.0, + "content": "ABSTRACT", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 143, + 221, + 469, + 429 + ], + "lines": [ + { + "bbox": [ + 141, + 219, + 470, + 235 + ], + "spans": [ + { + "bbox": [ + 141, + 219, + 470, + 235 + ], + "score": 1.0, + "content": "Vision transformer (ViT) has recently shown its strong capability in achieving", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 142, + 233, + 470, + 244 + ], + "spans": [ + { + "bbox": [ + 142, + 233, + 470, + 244 + ], + "score": 1.0, + "content": "comparable results to convolutional neural networks (CNNs) on image classifica-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 141, + 243, + 469, + 255 + ], + "spans": [ + { + "bbox": [ + 141, + 243, + 469, + 255 + ], + "score": 1.0, + "content": "tion. However, vanilla ViT simply inherits the same architecture from the natural", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 141, + 255, + 470, + 267 + ], + "spans": [ + { + "bbox": [ + 141, + 255, + 470, + 267 + ], + "score": 1.0, + "content": "language processing directly, which is often not optimized for vision applications.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 141, + 265, + 470, + 278 + ], + "spans": [ + { + "bbox": [ + 141, + 265, + 470, + 278 + ], + "score": 1.0, + "content": "Motivated by this, in this paper, we propose a new architecture that adopts the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 276, + 469, + 288 + ], + "spans": [ + { + "bbox": [ + 141, + 276, + 469, + 288 + ], + "score": 1.0, + "content": "pyramid structure and employ novel regional-to-local attention rather than global", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 287, + 469, + 299 + ], + "spans": [ + { + "bbox": [ + 141, + 287, + 469, + 299 + ], + "score": 1.0, + "content": "self-attention in vision transformers. More specifically, our model first generates", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 298, + 469, + 310 + ], + "spans": [ + { + "bbox": [ + 141, + 298, + 469, + 310 + ], + "score": 1.0, + "content": "regional tokens and local tokens from an image with different patch sizes, where", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 309, + 469, + 321 + ], + "spans": [ + { + "bbox": [ + 141, + 309, + 469, + 321 + ], + "score": 1.0, + "content": "each regional token is associated with a set of local tokens based on the spatial", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 320, + 469, + 332 + ], + "spans": [ + { + "bbox": [ + 141, + 320, + 469, + 332 + ], + "score": 1.0, + "content": "location. The regional-to-local attention includes two steps: first, the regional", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 331, + 469, + 343 + ], + "spans": [ + { + "bbox": [ + 141, + 331, + 469, + 343 + ], + "score": 1.0, + "content": "self-attention extracts global information among all regional tokens and then the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 341, + 470, + 354 + ], + "spans": [ + { + "bbox": [ + 141, + 341, + 470, + 354 + ], + "score": 1.0, + "content": "local self-attention exchanges the information among one regional token and the as-", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 142, + 353, + 469, + 365 + ], + "spans": [ + { + "bbox": [ + 142, + 353, + 469, + 365 + ], + "score": 1.0, + "content": "sociated local tokens via self-attention. Therefore, even though local self-attention", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 364, + 470, + 376 + ], + "spans": [ + { + "bbox": [ + 141, + 364, + 470, + 376 + ], + "score": 1.0, + "content": "confines the scope in a local region but it can still receive global information.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 141, + 375, + 470, + 387 + ], + "spans": [ + { + "bbox": [ + 141, + 375, + 470, + 387 + ], + "score": 1.0, + "content": "Extensive experiments on four vision tasks, including image classification, ob-", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 141, + 386, + 470, + 398 + ], + "spans": [ + { + "bbox": [ + 141, + 386, + 470, + 398 + ], + "score": 1.0, + "content": "ject and keypoint detection, semantics segmentation and action recognition, show", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 141, + 396, + 471, + 409 + ], + "spans": [ + { + "bbox": [ + 141, + 396, + 471, + 409 + ], + "score": 1.0, + "content": "that our approach outperforms or is on par with state-of-the-art ViT variants in-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 141, + 408, + 470, + 420 + ], + "spans": [ + { + "bbox": [ + 141, + 408, + 470, + 420 + ], + "score": 1.0, + "content": "cluding many concurrent works. Our source codes and models are available at", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 141, + 419, + 338, + 430 + ], + "spans": [ + { + "bbox": [ + 141, + 419, + 338, + 430 + ], + "score": 1.0, + "content": "https://github.com/IBM/RegionViT.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 15 + }, + { + "type": "title", + "bbox": [ + 108, + 449, + 206, + 461 + ], + "lines": [ + { + "bbox": [ + 105, + 448, + 208, + 465 + ], + "spans": [ + { + "bbox": [ + 105, + 448, + 208, + 465 + ], + "score": 1.0, + "content": "1 INTRODUCTION", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 107, + 473, + 505, + 605 + ], + "lines": [ + { + "bbox": [ + 106, + 473, + 505, + 486 + ], + "spans": [ + { + "bbox": [ + 106, + 473, + 505, + 486 + ], + "score": 1.0, + "content": "Transformers (Vaswani et al., 2017) based on self-attention come naturally with the ability to learn", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 484, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 106, + 484, + 505, + 496 + ], + "score": 1.0, + "content": "long-range dependencies in sequential data. As important as it is to language modeling (Devlin et al.,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 495, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 506, + 507 + ], + "score": 1.0, + "content": "2019), such ability is also highly desired for many vision tasks where contextual modeling plays a", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 505, + 506, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 506, + 519 + ], + "score": 1.0, + "content": "significant role. For this reason, self-attention and transformers have been receiving an increasing", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 518, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 505, + 528 + ], + "score": 1.0, + "content": "attention in the vision community (Bello, 2021; Srinivas et al., 2021; Zhao et al., 2020; Ramachandran", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 528, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 106, + 528, + 506, + 540 + ], + "score": 1.0, + "content": "et al., 2019a; Bello et al., 2019; Hu et al., 2019; Ramachandran et al., 2019b; Wang et al., 2018).", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 540, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 106, + 540, + 505, + 550 + ], + "score": 1.0, + "content": "Especially, the recent Vision Transformers (ViT) (Dosovitskiy et al., 2021) demonstrates comparable", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 550, + 505, + 562 + ], + "spans": [ + { + "bbox": [ + 106, + 550, + 505, + 562 + ], + "score": 1.0, + "content": "image classification results against the firmly established and prevalent CNNs in computer vision (He", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 560, + 506, + 573 + ], + "spans": [ + { + "bbox": [ + 105, + 560, + 506, + 573 + ], + "score": 1.0, + "content": "et al., 2016; Tan & Le, 2019; Brock et al., 2021), albeit relying on a huge amount of training data. It", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 573, + 505, + 583 + ], + "spans": [ + { + "bbox": [ + 106, + 573, + 505, + 583 + ], + "score": 1.0, + "content": "has since then led to an explosion of interest in further investigating its potential for a wide variety of", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 583, + 505, + 594 + ], + "spans": [ + { + "bbox": [ + 106, + 583, + 505, + 594 + ], + "score": 1.0, + "content": "vision applications (Wu et al., 2021; Wang et al., 2021; Heo et al., 2021; Zhang et al., 2021a; Li et al.,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 594, + 504, + 606 + ], + "spans": [ + { + "bbox": [ + 106, + 594, + 504, + 606 + ], + "score": 1.0, + "content": "2021; Graham et al., 2021; Liu et al., 2021; Chu et al., 2021a; Yan et al., 2021; Chen et al., 2021b).", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 107, + 611, + 505, + 731 + ], + "lines": [ + { + "bbox": [ + 105, + 610, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 505, + 623 + ], + "score": 1.0, + "content": "The ViT inherits the entire architecture from the vanilla transformer (Vaswani et al., 2017), which is", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 621, + 504, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 621, + 504, + 634 + ], + "score": 1.0, + "content": "designed for natural language processing tasks, and some of those designs thus may not meet the", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "needs of vision tasks. For example, the transformer has an isotropic network structure with a fixed", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "score": 1.0, + "content": "number of tokens and unchanged embedding size, which loses the capability to model the context", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 654, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 506, + 667 + ], + "score": 1.0, + "content": "with different scales and allocates computations at different scales. As opposed to this, a majority of", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 664, + 506, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 506, + 679 + ], + "score": 1.0, + "content": "CNNs adopt a popular pyramid architecture to compute multi-scale features efficiently. Recent vision", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "transformers such as PVT (Wang et al., 2021) and PiT (Heo et al., 2021) also follow a similar pyramid", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "structure as CNNs, showing improvement on both computation and memory efficiency as well as", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "on model accuracy. Another critical bottleneck of the transformer is that the self-attention module", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 710, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 506, + 722 + ], + "score": 1.0, + "content": "has a quadratic cost in memory and computation with regard to the sequence length (i.e., the number", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "of tokens). This issue is even worse in ViT as images are 2-D, suggesting a quadratic relationship", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 43 + } + ], + "page_idx": 0, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "score": 1.0, + "content": "1", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 78, + 503, + 116 + ], + "lines": [ + { + "bbox": [ + 106, + 78, + 505, + 98 + ], + "spans": [ + { + "bbox": [ + 106, + 78, + 505, + 98 + ], + "score": 1.0, + "content": "REGIONVIT: REGIONAL-TO-LOCAL ATTENTION FOR", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 97, + 289, + 118 + ], + "spans": [ + { + "bbox": [ + 106, + 97, + 289, + 118 + ], + "score": 1.0, + "content": "VISION TRANSFORMERS", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 112, + 135, + 423, + 168 + ], + "lines": [ + { + "bbox": [ + 112, + 135, + 360, + 147 + ], + "spans": [ + { + "bbox": [ + 112, + 135, + 360, + 147 + ], + "score": 1.0, + "content": "Chun-Fu (Richard) Chen, Rameswar Panda, Quanfu Fan", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 112, + 146, + 218, + 158 + ], + "spans": [ + { + "bbox": [ + 112, + 146, + 218, + 158 + ], + "score": 1.0, + "content": "MIT-IBM Watson AI Lab", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 113, + 158, + 425, + 169 + ], + "spans": [ + { + "bbox": [ + 113, + 158, + 425, + 169 + ], + "score": 1.0, + "content": "chenrich@us.ibm.com, rpanda@ibm.com, qfan@us.ibm.com", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3, + "bbox_fs": [ + 112, + 135, + 425, + 169 + ] + }, + { + "type": "title", + "bbox": [ + 278, + 198, + 333, + 209 + ], + "lines": [ + { + "bbox": [ + 276, + 196, + 336, + 212 + ], + "spans": [ + { + "bbox": [ + 276, + 196, + 336, + 212 + ], + "score": 1.0, + "content": "ABSTRACT", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 143, + 221, + 469, + 429 + ], + "lines": [ + { + "bbox": [ + 141, + 219, + 470, + 235 + ], + "spans": [ + { + "bbox": [ + 141, + 219, + 470, + 235 + ], + "score": 1.0, + "content": "Vision transformer (ViT) has recently shown its strong capability in achieving", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 142, + 233, + 470, + 244 + ], + "spans": [ + { + "bbox": [ + 142, + 233, + 470, + 244 + ], + "score": 1.0, + "content": "comparable results to convolutional neural networks (CNNs) on image classifica-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 141, + 243, + 469, + 255 + ], + "spans": [ + { + "bbox": [ + 141, + 243, + 469, + 255 + ], + "score": 1.0, + "content": "tion. However, vanilla ViT simply inherits the same architecture from the natural", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 141, + 255, + 470, + 267 + ], + "spans": [ + { + "bbox": [ + 141, + 255, + 470, + 267 + ], + "score": 1.0, + "content": "language processing directly, which is often not optimized for vision applications.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 141, + 265, + 470, + 278 + ], + "spans": [ + { + "bbox": [ + 141, + 265, + 470, + 278 + ], + "score": 1.0, + "content": "Motivated by this, in this paper, we propose a new architecture that adopts the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 276, + 469, + 288 + ], + "spans": [ + { + "bbox": [ + 141, + 276, + 469, + 288 + ], + "score": 1.0, + "content": "pyramid structure and employ novel regional-to-local attention rather than global", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 287, + 469, + 299 + ], + "spans": [ + { + "bbox": [ + 141, + 287, + 469, + 299 + ], + "score": 1.0, + "content": "self-attention in vision transformers. More specifically, our model first generates", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 298, + 469, + 310 + ], + "spans": [ + { + "bbox": [ + 141, + 298, + 469, + 310 + ], + "score": 1.0, + "content": "regional tokens and local tokens from an image with different patch sizes, where", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 309, + 469, + 321 + ], + "spans": [ + { + "bbox": [ + 141, + 309, + 469, + 321 + ], + "score": 1.0, + "content": "each regional token is associated with a set of local tokens based on the spatial", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 320, + 469, + 332 + ], + "spans": [ + { + "bbox": [ + 141, + 320, + 469, + 332 + ], + "score": 1.0, + "content": "location. The regional-to-local attention includes two steps: first, the regional", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 331, + 469, + 343 + ], + "spans": [ + { + "bbox": [ + 141, + 331, + 469, + 343 + ], + "score": 1.0, + "content": "self-attention extracts global information among all regional tokens and then the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 341, + 470, + 354 + ], + "spans": [ + { + "bbox": [ + 141, + 341, + 470, + 354 + ], + "score": 1.0, + "content": "local self-attention exchanges the information among one regional token and the as-", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 142, + 353, + 469, + 365 + ], + "spans": [ + { + "bbox": [ + 142, + 353, + 469, + 365 + ], + "score": 1.0, + "content": "sociated local tokens via self-attention. Therefore, even though local self-attention", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 364, + 470, + 376 + ], + "spans": [ + { + "bbox": [ + 141, + 364, + 470, + 376 + ], + "score": 1.0, + "content": "confines the scope in a local region but it can still receive global information.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 141, + 375, + 470, + 387 + ], + "spans": [ + { + "bbox": [ + 141, + 375, + 470, + 387 + ], + "score": 1.0, + "content": "Extensive experiments on four vision tasks, including image classification, ob-", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 141, + 386, + 470, + 398 + ], + "spans": [ + { + "bbox": [ + 141, + 386, + 470, + 398 + ], + "score": 1.0, + "content": "ject and keypoint detection, semantics segmentation and action recognition, show", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 141, + 396, + 471, + 409 + ], + "spans": [ + { + "bbox": [ + 141, + 396, + 471, + 409 + ], + "score": 1.0, + "content": "that our approach outperforms or is on par with state-of-the-art ViT variants in-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 141, + 408, + 470, + 420 + ], + "spans": [ + { + "bbox": [ + 141, + 408, + 470, + 420 + ], + "score": 1.0, + "content": "cluding many concurrent works. Our source codes and models are available at", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 141, + 419, + 338, + 430 + ], + "spans": [ + { + "bbox": [ + 141, + 419, + 338, + 430 + ], + "score": 1.0, + "content": "https://github.com/IBM/RegionViT.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 15, + "bbox_fs": [ + 141, + 219, + 471, + 430 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 449, + 206, + 461 + ], + "lines": [ + { + "bbox": [ + 105, + 448, + 208, + 465 + ], + "spans": [ + { + "bbox": [ + 105, + 448, + 208, + 465 + ], + "score": 1.0, + "content": "1 INTRODUCTION", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 107, + 473, + 505, + 605 + ], + "lines": [ + { + "bbox": [ + 106, + 473, + 505, + 486 + ], + "spans": [ + { + "bbox": [ + 106, + 473, + 505, + 486 + ], + "score": 1.0, + "content": "Transformers (Vaswani et al., 2017) based on self-attention come naturally with the ability to learn", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 484, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 106, + 484, + 505, + 496 + ], + "score": 1.0, + "content": "long-range dependencies in sequential data. As important as it is to language modeling (Devlin et al.,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 495, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 506, + 507 + ], + "score": 1.0, + "content": "2019), such ability is also highly desired for many vision tasks where contextual modeling plays a", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 505, + 506, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 506, + 519 + ], + "score": 1.0, + "content": "significant role. For this reason, self-attention and transformers have been receiving an increasing", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 518, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 505, + 528 + ], + "score": 1.0, + "content": "attention in the vision community (Bello, 2021; Srinivas et al., 2021; Zhao et al., 2020; Ramachandran", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 528, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 106, + 528, + 506, + 540 + ], + "score": 1.0, + "content": "et al., 2019a; Bello et al., 2019; Hu et al., 2019; Ramachandran et al., 2019b; Wang et al., 2018).", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 540, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 106, + 540, + 505, + 550 + ], + "score": 1.0, + "content": "Especially, the recent Vision Transformers (ViT) (Dosovitskiy et al., 2021) demonstrates comparable", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 550, + 505, + 562 + ], + "spans": [ + { + "bbox": [ + 106, + 550, + 505, + 562 + ], + "score": 1.0, + "content": "image classification results against the firmly established and prevalent CNNs in computer vision (He", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 560, + 506, + 573 + ], + "spans": [ + { + "bbox": [ + 105, + 560, + 506, + 573 + ], + "score": 1.0, + "content": "et al., 2016; Tan & Le, 2019; Brock et al., 2021), albeit relying on a huge amount of training data. It", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 573, + 505, + 583 + ], + "spans": [ + { + "bbox": [ + 106, + 573, + 505, + 583 + ], + "score": 1.0, + "content": "has since then led to an explosion of interest in further investigating its potential for a wide variety of", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 583, + 505, + 594 + ], + "spans": [ + { + "bbox": [ + 106, + 583, + 505, + 594 + ], + "score": 1.0, + "content": "vision applications (Wu et al., 2021; Wang et al., 2021; Heo et al., 2021; Zhang et al., 2021a; Li et al.,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 594, + 504, + 606 + ], + "spans": [ + { + "bbox": [ + 106, + 594, + 504, + 606 + ], + "score": 1.0, + "content": "2021; Graham et al., 2021; Liu et al., 2021; Chu et al., 2021a; Yan et al., 2021; Chen et al., 2021b).", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 31.5, + "bbox_fs": [ + 105, + 473, + 506, + 606 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 611, + 505, + 731 + ], + "lines": [ + { + "bbox": [ + 105, + 610, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 505, + 623 + ], + "score": 1.0, + "content": "The ViT inherits the entire architecture from the vanilla transformer (Vaswani et al., 2017), which is", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 621, + 504, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 621, + 504, + 634 + ], + "score": 1.0, + "content": "designed for natural language processing tasks, and some of those designs thus may not meet the", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "needs of vision tasks. For example, the transformer has an isotropic network structure with a fixed", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "score": 1.0, + "content": "number of tokens and unchanged embedding size, which loses the capability to model the context", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 654, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 506, + 667 + ], + "score": 1.0, + "content": "with different scales and allocates computations at different scales. As opposed to this, a majority of", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 664, + 506, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 506, + 679 + ], + "score": 1.0, + "content": "CNNs adopt a popular pyramid architecture to compute multi-scale features efficiently. Recent vision", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "transformers such as PVT (Wang et al., 2021) and PiT (Heo et al., 2021) also follow a similar pyramid", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "structure as CNNs, showing improvement on both computation and memory efficiency as well as", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "on model accuracy. Another critical bottleneck of the transformer is that the self-attention module", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 710, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 506, + 722 + ], + "score": 1.0, + "content": "has a quadratic cost in memory and computation with regard to the sequence length (i.e., the number", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "of tokens). This issue is even worse in ViT as images are 2-D, suggesting a quadratic relationship", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "between the number of tokens and image resolution. As a result, ViT indicates a quadruple complexity", + "type": "text", + "cross_page": true + } + ], + "index": 0 + }, + { + "bbox": [ + 104, + 92, + 506, + 108 + ], + "spans": [ + { + "bbox": [ + 104, + 92, + 506, + 108 + ], + "score": 1.0, + "content": "w.r.t image resolution. The highly compute- and memory-intensive self-attention makes it challenging", + "type": "text", + "cross_page": true + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 505, + 118 + ], + "score": 1.0, + "content": "to train vision transformer models at fine-grained patch sizes. It also significantly undermines the", + "type": "text", + "cross_page": true + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 114, + 506, + 128 + ], + "spans": [ + { + "bbox": [ + 105, + 114, + 506, + 128 + ], + "score": 1.0, + "content": "applications of these models to tasks such as object detection and semantic segmentation, which", + "type": "text", + "cross_page": true + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 126, + 457, + 140 + ], + "spans": [ + { + "bbox": [ + 105, + 126, + 457, + 140 + ], + "score": 1.0, + "content": "benefit from or require fine feature information computed from high-resolution images.", + "type": "text", + "cross_page": true + } + ], + "index": 4 + } + ], + "index": 43, + "bbox_fs": [ + 105, + 610, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 505, + 137 + ], + "lines": [ + { + "bbox": [ + 105, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "between the number of tokens and image resolution. As a result, ViT indicates a quadruple complexity", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 104, + 92, + 506, + 108 + ], + "spans": [ + { + "bbox": [ + 104, + 92, + 506, + 108 + ], + "score": 1.0, + "content": "w.r.t image resolution. The highly compute- and memory-intensive self-attention makes it challenging", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 505, + 118 + ], + "score": 1.0, + "content": "to train vision transformer models at fine-grained patch sizes. It also significantly undermines the", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 114, + 506, + 128 + ], + "spans": [ + { + "bbox": [ + 105, + 114, + 506, + 128 + ], + "score": 1.0, + "content": "applications of these models to tasks such as object detection and semantic segmentation, which", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 126, + 457, + 140 + ], + "spans": [ + { + "bbox": [ + 105, + 126, + 457, + 140 + ], + "score": 1.0, + "content": "benefit from or require fine feature information computed from high-resolution images.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 2 + }, + { + "type": "text", + "bbox": [ + 107, + 144, + 336, + 428 + ], + "lines": [ + { + "bbox": [ + 106, + 144, + 336, + 155 + ], + "spans": [ + { + "bbox": [ + 106, + 144, + 336, + 155 + ], + "score": 1.0, + "content": "To address the aforementioned computational limitations", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 153, + 338, + 167 + ], + "spans": [ + { + "bbox": [ + 106, + 153, + 338, + 167 + ], + "score": 1.0, + "content": "of vision transformers, in this work, we develop a memory-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 165, + 337, + 177 + ], + "spans": [ + { + "bbox": [ + 106, + 165, + 337, + 177 + ], + "score": 1.0, + "content": "friendly and efficient self-attention method for transformer", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 176, + 338, + 189 + ], + "spans": [ + { + "bbox": [ + 106, + 176, + 338, + 189 + ], + "score": 1.0, + "content": "models to reach their promising potential for vision appli-", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 187, + 338, + 199 + ], + "spans": [ + { + "bbox": [ + 106, + 187, + 338, + 199 + ], + "score": 1.0, + "content": "cations. We propose a novel coarse-to-fine mechanism to", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 198, + 338, + 211 + ], + "spans": [ + { + "bbox": [ + 105, + 198, + 338, + 211 + ], + "score": 1.0, + "content": "compute self-attention in a hierarchical way. Specifically,", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 209, + 338, + 222 + ], + "spans": [ + { + "bbox": [ + 106, + 209, + 338, + 222 + ], + "score": 1.0, + "content": "our approach first divides the input image into a group of", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 219, + 337, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 219, + 286, + 233 + ], + "score": 1.0, + "content": "non-overlapping patches of large size (e.g.,", + "type": "text" + }, + { + "bbox": [ + 286, + 221, + 318, + 231 + ], + "score": 0.87, + "content": "2 8 \\times 2 8 )", + "type": "inline_equation" + }, + { + "bbox": [ + 318, + 219, + 337, + 233 + ], + "score": 1.0, + "content": "), on", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 231, + 338, + 243 + ], + "spans": [ + { + "bbox": [ + 105, + 231, + 338, + 243 + ], + "score": 1.0, + "content": "which regional tokens are computed via linear projection.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 241, + 337, + 255 + ], + "spans": [ + { + "bbox": [ + 105, + 241, + 337, + 255 + ], + "score": 1.0, + "content": "Similarly, local tokens are created for each region using", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 253, + 338, + 265 + ], + "spans": [ + { + "bbox": [ + 105, + 253, + 212, + 265 + ], + "score": 1.0, + "content": "a smaller patch size (e.g.,", + "type": "text" + }, + { + "bbox": [ + 212, + 253, + 234, + 264 + ], + "score": 0.85, + "content": "4 \\times 4 )", + "type": "inline_equation" + }, + { + "bbox": [ + 234, + 253, + 338, + 265 + ], + "score": 1.0, + "content": ". We then use a standard", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 264, + 338, + 276 + ], + "spans": [ + { + "bbox": [ + 105, + 264, + 338, + 276 + ], + "score": 1.0, + "content": "transformer to process regional and local tokens separately.", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 274, + 339, + 288 + ], + "spans": [ + { + "bbox": [ + 105, + 274, + 339, + 288 + ], + "score": 1.0, + "content": "To enable communication between the two types of tokens,", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 286, + 338, + 297 + ], + "spans": [ + { + "bbox": [ + 106, + 286, + 338, + 297 + ], + "score": 1.0, + "content": "we first perform self-attention on regional tokens (regional", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 297, + 338, + 308 + ], + "spans": [ + { + "bbox": [ + 105, + 297, + 338, + 308 + ], + "score": 1.0, + "content": "attention) and then jointly attend to the local tokens of", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 308, + 338, + 320 + ], + "spans": [ + { + "bbox": [ + 106, + 308, + 338, + 320 + ], + "score": 1.0, + "content": "each region including their associated regional token (lo-", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 319, + 338, + 331 + ], + "spans": [ + { + "bbox": [ + 105, + 319, + 338, + 331 + ], + "score": 1.0, + "content": "cal attention). By doing so, regional tokens pass global", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 330, + 338, + 341 + ], + "spans": [ + { + "bbox": [ + 106, + 330, + 338, + 341 + ], + "score": 1.0, + "content": "contextual information to local tokens efficiently while be-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 341, + 339, + 353 + ], + "spans": [ + { + "bbox": [ + 106, + 341, + 339, + 353 + ], + "score": 1.0, + "content": "ing able to effectively learn from local tokens themselves.", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 352, + 338, + 363 + ], + "spans": [ + { + "bbox": [ + 106, + 352, + 338, + 363 + ], + "score": 1.0, + "content": "For clarity, we represent this two-stage attention mecha-", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 363, + 337, + 374 + ], + "spans": [ + { + "bbox": [ + 106, + 363, + 265, + 374 + ], + "score": 1.0, + "content": "nism as Regional-to-Local Attention, or", + "type": "text" + }, + { + "bbox": [ + 266, + 363, + 284, + 373 + ], + "score": 0.47, + "content": "R 2 L", + "type": "inline_equation" + }, + { + "bbox": [ + 285, + 363, + 337, + 374 + ], + "score": 1.0, + "content": "attention for", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 374, + 337, + 385 + ], + "spans": [ + { + "bbox": [ + 106, + 374, + 337, + 385 + ], + "score": 1.0, + "content": "short (see Figure 1 for an illustration). Since both regional", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 385, + 337, + 396 + ], + "spans": [ + { + "bbox": [ + 106, + 385, + 337, + 396 + ], + "score": 1.0, + "content": "and local attention involve much fewer tokens, our R2L", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 396, + 337, + 408 + ], + "spans": [ + { + "bbox": [ + 106, + 396, + 337, + 408 + ], + "score": 1.0, + "content": "attention requires substantially less memory than regular", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 406, + 338, + 418 + ], + "spans": [ + { + "bbox": [ + 106, + 406, + 338, + 418 + ], + "score": 1.0, + "content": "global self-attention used in vision transformers. For ex-", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 416, + 337, + 431 + ], + "spans": [ + { + "bbox": [ + 105, + 416, + 337, + 431 + ], + "score": 1.0, + "content": "ample, in our default setting, the memory saving using", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 17.5 + }, + { + "type": "image", + "bbox": [ + 345, + 144, + 502, + 220 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 345, + 144, + 502, + 220 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 345, + 144, + 502, + 220 + ], + "spans": [ + { + "bbox": [ + 345, + 144, + 502, + 220 + ], + "score": 0.965, + "type": "image", + "image_path": "be438c75eaab31cb96439f283128309a198fe8950d097322f0c0e7274132054b.jpg" + } + ] + } + ], + "index": 33.5, + "virtual_lines": [ + { + "bbox": [ + 345, + 144, + 502, + 156.66666666666666 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 345, + 156.66666666666666, + 502, + 169.33333333333331 + ], + "spans": [], + "index": 32 + }, + { + "bbox": [ + 345, + 169.33333333333331, + 502, + 181.99999999999997 + ], + "spans": [], + "index": 33 + }, + { + "bbox": [ + 345, + 181.99999999999997, + 502, + 194.66666666666663 + ], + "spans": [], + "index": 34 + }, + { + "bbox": [ + 345, + 194.66666666666663, + 502, + 207.3333333333333 + ], + "spans": [], + "index": 35 + }, + { + "bbox": [ + 345, + 207.3333333333333, + 502, + 219.99999999999994 + ], + "spans": [], + "index": 36 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 344, + 226, + 505, + 426 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 343, + 225, + 505, + 237 + ], + "spans": [ + { + "bbox": [ + 343, + 225, + 505, + 237 + ], + "score": 1.0, + "content": "Figure 1: Regional-to-Local Attention", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 343, + 235, + 505, + 246 + ], + "spans": [ + { + "bbox": [ + 343, + 235, + 505, + 246 + ], + "score": 1.0, + "content": "for Vision Transformers. (a) ViT uses", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 343, + 246, + 506, + 256 + ], + "spans": [ + { + "bbox": [ + 343, + 246, + 506, + 256 + ], + "score": 1.0, + "content": "a fixed patch size through the whole net-", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 343, + 256, + 505, + 267 + ], + "spans": [ + { + "bbox": [ + 343, + 256, + 505, + 267 + ], + "score": 1.0, + "content": "work, (b) PVT adopts a pyramid structure", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 343, + 266, + 506, + 277 + ], + "spans": [ + { + "bbox": [ + 343, + 266, + 506, + 277 + ], + "score": 1.0, + "content": "to gradually enlarge the patch size in the net-", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 343, + 276, + 505, + 286 + ], + "spans": [ + { + "bbox": [ + 343, + 276, + 505, + 286 + ], + "score": 1.0, + "content": "work. Both ViT and PVT uses all tokens", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 343, + 285, + 506, + 297 + ], + "spans": [ + { + "bbox": [ + 343, + 285, + 506, + 297 + ], + "score": 1.0, + "content": "in self-attention, which are computational-", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 343, + 295, + 506, + 307 + ], + "spans": [ + { + "bbox": [ + 343, + 295, + 506, + 307 + ], + "score": 1.0, + "content": "and memory-intensive. (c) Our proposed ap-", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 343, + 306, + 505, + 316 + ], + "spans": [ + { + "bbox": [ + 343, + 306, + 505, + 316 + ], + "score": 1.0, + "content": "proach combines a pyramid structure with", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 343, + 316, + 505, + 326 + ], + "spans": [ + { + "bbox": [ + 343, + 316, + 505, + 326 + ], + "score": 1.0, + "content": "an efficient regional-to-local (R2L) attention", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 344, + 326, + 506, + 336 + ], + "spans": [ + { + "bbox": [ + 344, + 326, + 506, + 336 + ], + "score": 1.0, + "content": "mechanism to reduce computation and mem-", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 343, + 336, + 505, + 346 + ], + "spans": [ + { + "bbox": [ + 343, + 336, + 505, + 346 + ], + "score": 1.0, + "content": "ory usage. Our approach divides the input", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 343, + 345, + 506, + 356 + ], + "spans": [ + { + "bbox": [ + 343, + 345, + 506, + 356 + ], + "score": 1.0, + "content": "image into two groups of tokens, regional to-", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 344, + 356, + 505, + 366 + ], + "spans": [ + { + "bbox": [ + 344, + 356, + 505, + 366 + ], + "score": 1.0, + "content": "kens of large patch size (red) and local ones", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 343, + 365, + 506, + 376 + ], + "spans": [ + { + "bbox": [ + 343, + 365, + 506, + 376 + ], + "score": 1.0, + "content": "of small patch size (black). The two types of", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 343, + 375, + 506, + 385 + ], + "spans": [ + { + "bbox": [ + 343, + 375, + 506, + 385 + ], + "score": 1.0, + "content": "tokens communicate efficiently through R2L", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 343, + 385, + 505, + 396 + ], + "spans": [ + { + "bbox": [ + 343, + 385, + 505, + 396 + ], + "score": 1.0, + "content": "attention, which jointly attends to the local", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 343, + 395, + 505, + 406 + ], + "spans": [ + { + "bbox": [ + 343, + 395, + 505, + 406 + ], + "score": 1.0, + "content": "tokens in the same region and the associated", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 343, + 405, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 343, + 405, + 505, + 416 + ], + "score": 1.0, + "content": "regional token. Note that the numbers denote", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 344, + 416, + 489, + 426 + ], + "spans": [ + { + "bbox": [ + 344, + 416, + 489, + 426 + ], + "score": 1.0, + "content": "the patch sizes at each stage of a model.", + "type": "text" + } + ], + "index": 56 + } + ], + "index": 46.5 + } + ], + "index": 40.0 + }, + { + "type": "text", + "bbox": [ + 107, + 429, + 504, + 461 + ], + "lines": [ + { + "bbox": [ + 105, + 426, + 505, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 270, + 441 + ], + "score": 1.0, + "content": "R2L attention can be up to as much as", + "type": "text" + }, + { + "bbox": [ + 270, + 428, + 290, + 439 + ], + "score": 0.88, + "content": "73 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 426, + 505, + 441 + ], + "score": 1.0, + "content": ". We demonstrate the effectiveness of our approach", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 106, + 439, + 505, + 451 + ], + "spans": [ + { + "bbox": [ + 106, + 439, + 505, + 451 + ], + "score": 1.0, + "content": "on image classification and several downstream vision tasks including object detection and action", + "type": "text" + } + ], + "index": 58 + }, + { + "bbox": [ + 105, + 450, + 157, + 464 + ], + "spans": [ + { + "bbox": [ + 105, + 450, + 157, + 464 + ], + "score": 1.0, + "content": "recognition.", + "type": "text" + } + ], + "index": 59 + } + ], + "index": 58 + }, + { + "type": "text", + "bbox": [ + 106, + 467, + 364, + 478 + ], + "lines": [ + { + "bbox": [ + 106, + 466, + 365, + 479 + ], + "spans": [ + { + "bbox": [ + 106, + 466, + 365, + 479 + ], + "score": 1.0, + "content": "To summarize, our key contributions in this work are as follows:", + "type": "text" + } + ], + "index": 60 + } + ], + "index": 60 + }, + { + "type": "text", + "bbox": [ + 107, + 487, + 506, + 615 + ], + "lines": [ + { + "bbox": [ + 106, + 487, + 505, + 499 + ], + "spans": [ + { + "bbox": [ + 106, + 487, + 505, + 499 + ], + "score": 1.0, + "content": "1. We propose a new vision transformer (RegionViT) based on regional-to-local attention to learn", + "type": "text" + } + ], + "index": 61 + }, + { + "bbox": [ + 118, + 498, + 505, + 510 + ], + "spans": [ + { + "bbox": [ + 118, + 498, + 505, + 510 + ], + "score": 1.0, + "content": "both local and global features. Our proposed regional-to-local attention alleviates the overhead of", + "type": "text" + } + ], + "index": 62 + }, + { + "bbox": [ + 119, + 509, + 505, + 521 + ], + "spans": [ + { + "bbox": [ + 119, + 509, + 505, + 521 + ], + "score": 1.0, + "content": "standard global attention (too many tokens) and the weakness of pure local attention (no interaction", + "type": "text" + } + ], + "index": 63 + }, + { + "bbox": [ + 118, + 520, + 337, + 533 + ], + "spans": [ + { + "bbox": [ + 118, + 520, + 337, + 533 + ], + "score": 1.0, + "content": "between regions) used in existing vision transformers.", + "type": "text" + } + ], + "index": 64 + }, + { + "bbox": [ + 105, + 533, + 505, + 547 + ], + "spans": [ + { + "bbox": [ + 105, + 533, + 505, + 547 + ], + "score": 1.0, + "content": "2. Our regional-to-local attention reduces the memory complexity significantly as compared to", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 117, + 544, + 506, + 558 + ], + "spans": [ + { + "bbox": [ + 117, + 544, + 429, + 558 + ], + "score": 1.0, + "content": "standard self-attention, leading to a savings in memory complexity by about", + "type": "text" + }, + { + "bbox": [ + 429, + 545, + 473, + 557 + ], + "score": 0.93, + "content": "\\dot { \\mathcal { O } } ( N / M ^ { 2 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 474, + 544, + 506, + 558 + ], + "score": 1.0, + "content": ", where", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 119, + 554, + 506, + 570 + ], + "spans": [ + { + "bbox": [ + 119, + 556, + 131, + 566 + ], + "score": 0.75, + "content": "M", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 554, + 270, + 570 + ], + "score": 1.0, + "content": "is the window size of a region and", + "type": "text" + }, + { + "bbox": [ + 271, + 556, + 281, + 566 + ], + "score": 0.79, + "content": "N", + "type": "inline_equation" + }, + { + "bbox": [ + 281, + 554, + 506, + 570 + ], + "score": 1.0, + "content": "is the total number of tokens. This effectively allows us", + "type": "text" + } + ], + "index": 67 + }, + { + "bbox": [ + 118, + 566, + 471, + 581 + ], + "spans": [ + { + "bbox": [ + 118, + 566, + 471, + 581 + ], + "score": 1.0, + "content": "to train a more deep network for better performance while with comparable complexity.", + "type": "text" + } + ], + "index": 68 + }, + { + "bbox": [ + 106, + 582, + 506, + 595 + ], + "spans": [ + { + "bbox": [ + 106, + 582, + 506, + 595 + ], + "score": 1.0, + "content": "3. Our models outperform or on par with several concurrent works on vision transformer that exploit", + "type": "text" + } + ], + "index": 69 + }, + { + "bbox": [ + 117, + 592, + 506, + 605 + ], + "spans": [ + { + "bbox": [ + 117, + 592, + 506, + 605 + ], + "score": 1.0, + "content": "pyramid structure for image classification. Experiments also demonstrate that our models work", + "type": "text" + } + ], + "index": 70 + }, + { + "bbox": [ + 119, + 604, + 506, + 615 + ], + "spans": [ + { + "bbox": [ + 119, + 604, + 506, + 615 + ], + "score": 1.0, + "content": "well on several downstream classification tasks, including object detection and action recognition.", + "type": "text" + } + ], + "index": 71 + } + ], + "index": 66 + }, + { + "type": "title", + "bbox": [ + 108, + 630, + 211, + 643 + ], + "lines": [ + { + "bbox": [ + 104, + 628, + 213, + 645 + ], + "spans": [ + { + "bbox": [ + 104, + 628, + 213, + 645 + ], + "score": 1.0, + "content": "2 RELATED WORK", + "type": "text" + } + ], + "index": 72 + } + ], + "index": 72 + }, + { + "type": "text", + "bbox": [ + 106, + 655, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 652, + 507, + 669 + ], + "spans": [ + { + "bbox": [ + 105, + 652, + 507, + 669 + ], + "score": 1.0, + "content": "CNNs with Attention. Attention has been widely used for enhancing the features in CNNs, e.g.,", + "type": "text" + } + ], + "index": 73 + }, + { + "bbox": [ + 105, + 664, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 506, + 678 + ], + "score": 1.0, + "content": "SENet (Hu et al., 2018) and CBAM (Woo et al., 2018). Various methods have been proposed that", + "type": "text" + } + ], + "index": 74 + }, + { + "bbox": [ + 105, + 676, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 689 + ], + "score": 1.0, + "content": "combine self-attention with CNNs (Bello, 2021; Srinivas et al., 2021; Zhao et al., 2020; Ramachandran", + "type": "text" + } + ], + "index": 75 + }, + { + "bbox": [ + 104, + 685, + 507, + 702 + ], + "spans": [ + { + "bbox": [ + 104, + 685, + 507, + 702 + ], + "score": 1.0, + "content": "et al., 2019a; Bello et al., 2019; Hu et al., 2019; Ramachandran et al., 2019b; Wang et al., 2018). E.g.,", + "type": "text" + } + ], + "index": 76 + }, + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "SAN (Zhao et al., 2020) and SASA (Ramachandran et al., 2019a) attempts to replace all convolutional", + "type": "text" + } + ], + "index": 77 + }, + { + "bbox": [ + 106, + 710, + 504, + 721 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 504, + 721 + ], + "score": 1.0, + "content": "layers with local self-attention network. While these works show promising results, their complexity", + "type": "text" + } + ], + "index": 78 + }, + { + "bbox": [ + 105, + 720, + 504, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 504, + 732 + ], + "score": 1.0, + "content": "is relatively high due to the self-attention. On the other hand, BoTNet (Srinivas et al., 2021) replaces", + "type": "text" + } + ], + "index": 79 + } + ], + "index": 76 + } + ], + "page_idx": 1, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 763 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 763 + ], + "score": 1.0, + "content": "2", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 505, + 137 + ], + "lines": [], + "index": 2, + "bbox_fs": [ + 104, + 82, + 506, + 140 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 144, + 336, + 428 + ], + "lines": [ + { + "bbox": [ + 106, + 144, + 336, + 155 + ], + "spans": [ + { + "bbox": [ + 106, + 144, + 336, + 155 + ], + "score": 1.0, + "content": "To address the aforementioned computational limitations", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 153, + 338, + 167 + ], + "spans": [ + { + "bbox": [ + 106, + 153, + 338, + 167 + ], + "score": 1.0, + "content": "of vision transformers, in this work, we develop a memory-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 165, + 337, + 177 + ], + "spans": [ + { + "bbox": [ + 106, + 165, + 337, + 177 + ], + "score": 1.0, + "content": "friendly and efficient self-attention method for transformer", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 176, + 338, + 189 + ], + "spans": [ + { + "bbox": [ + 106, + 176, + 338, + 189 + ], + "score": 1.0, + "content": "models to reach their promising potential for vision appli-", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 187, + 338, + 199 + ], + "spans": [ + { + "bbox": [ + 106, + 187, + 338, + 199 + ], + "score": 1.0, + "content": "cations. We propose a novel coarse-to-fine mechanism to", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 198, + 338, + 211 + ], + "spans": [ + { + "bbox": [ + 105, + 198, + 338, + 211 + ], + "score": 1.0, + "content": "compute self-attention in a hierarchical way. Specifically,", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 209, + 338, + 222 + ], + "spans": [ + { + "bbox": [ + 106, + 209, + 338, + 222 + ], + "score": 1.0, + "content": "our approach first divides the input image into a group of", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 219, + 337, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 219, + 286, + 233 + ], + "score": 1.0, + "content": "non-overlapping patches of large size (e.g.,", + "type": "text" + }, + { + "bbox": [ + 286, + 221, + 318, + 231 + ], + "score": 0.87, + "content": "2 8 \\times 2 8 )", + "type": "inline_equation" + }, + { + "bbox": [ + 318, + 219, + 337, + 233 + ], + "score": 1.0, + "content": "), on", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 231, + 338, + 243 + ], + "spans": [ + { + "bbox": [ + 105, + 231, + 338, + 243 + ], + "score": 1.0, + "content": "which regional tokens are computed via linear projection.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 241, + 337, + 255 + ], + "spans": [ + { + "bbox": [ + 105, + 241, + 337, + 255 + ], + "score": 1.0, + "content": "Similarly, local tokens are created for each region using", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 253, + 338, + 265 + ], + "spans": [ + { + "bbox": [ + 105, + 253, + 212, + 265 + ], + "score": 1.0, + "content": "a smaller patch size (e.g.,", + "type": "text" + }, + { + "bbox": [ + 212, + 253, + 234, + 264 + ], + "score": 0.85, + "content": "4 \\times 4 )", + "type": "inline_equation" + }, + { + "bbox": [ + 234, + 253, + 338, + 265 + ], + "score": 1.0, + "content": ". We then use a standard", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 264, + 338, + 276 + ], + "spans": [ + { + "bbox": [ + 105, + 264, + 338, + 276 + ], + "score": 1.0, + "content": "transformer to process regional and local tokens separately.", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 274, + 339, + 288 + ], + "spans": [ + { + "bbox": [ + 105, + 274, + 339, + 288 + ], + "score": 1.0, + "content": "To enable communication between the two types of tokens,", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 286, + 338, + 297 + ], + "spans": [ + { + "bbox": [ + 106, + 286, + 338, + 297 + ], + "score": 1.0, + "content": "we first perform self-attention on regional tokens (regional", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 297, + 338, + 308 + ], + "spans": [ + { + "bbox": [ + 105, + 297, + 338, + 308 + ], + "score": 1.0, + "content": "attention) and then jointly attend to the local tokens of", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 308, + 338, + 320 + ], + "spans": [ + { + "bbox": [ + 106, + 308, + 338, + 320 + ], + "score": 1.0, + "content": "each region including their associated regional token (lo-", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 319, + 338, + 331 + ], + "spans": [ + { + "bbox": [ + 105, + 319, + 338, + 331 + ], + "score": 1.0, + "content": "cal attention). By doing so, regional tokens pass global", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 330, + 338, + 341 + ], + "spans": [ + { + "bbox": [ + 106, + 330, + 338, + 341 + ], + "score": 1.0, + "content": "contextual information to local tokens efficiently while be-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 341, + 339, + 353 + ], + "spans": [ + { + "bbox": [ + 106, + 341, + 339, + 353 + ], + "score": 1.0, + "content": "ing able to effectively learn from local tokens themselves.", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 352, + 338, + 363 + ], + "spans": [ + { + "bbox": [ + 106, + 352, + 338, + 363 + ], + "score": 1.0, + "content": "For clarity, we represent this two-stage attention mecha-", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 363, + 337, + 374 + ], + "spans": [ + { + "bbox": [ + 106, + 363, + 265, + 374 + ], + "score": 1.0, + "content": "nism as Regional-to-Local Attention, or", + "type": "text" + }, + { + "bbox": [ + 266, + 363, + 284, + 373 + ], + "score": 0.47, + "content": "R 2 L", + "type": "inline_equation" + }, + { + "bbox": [ + 285, + 363, + 337, + 374 + ], + "score": 1.0, + "content": "attention for", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 374, + 337, + 385 + ], + "spans": [ + { + "bbox": [ + 106, + 374, + 337, + 385 + ], + "score": 1.0, + "content": "short (see Figure 1 for an illustration). Since both regional", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 385, + 337, + 396 + ], + "spans": [ + { + "bbox": [ + 106, + 385, + 337, + 396 + ], + "score": 1.0, + "content": "and local attention involve much fewer tokens, our R2L", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 396, + 337, + 408 + ], + "spans": [ + { + "bbox": [ + 106, + 396, + 337, + 408 + ], + "score": 1.0, + "content": "attention requires substantially less memory than regular", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 406, + 338, + 418 + ], + "spans": [ + { + "bbox": [ + 106, + 406, + 338, + 418 + ], + "score": 1.0, + "content": "global self-attention used in vision transformers. For ex-", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 416, + 337, + 431 + ], + "spans": [ + { + "bbox": [ + 105, + 416, + 337, + 431 + ], + "score": 1.0, + "content": "ample, in our default setting, the memory saving using", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 17.5, + "bbox_fs": [ + 105, + 144, + 339, + 431 + ] + }, + { + "type": "image", + "bbox": [ + 345, + 144, + 502, + 220 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 345, + 144, + 502, + 220 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 345, + 144, + 502, + 220 + ], + "spans": [ + { + "bbox": [ + 345, + 144, + 502, + 220 + ], + "score": 0.965, + "type": "image", + "image_path": "be438c75eaab31cb96439f283128309a198fe8950d097322f0c0e7274132054b.jpg" + } + ] + } + ], + "index": 33.5, + "virtual_lines": [ + { + "bbox": [ + 345, + 144, + 502, + 156.66666666666666 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 345, + 156.66666666666666, + 502, + 169.33333333333331 + ], + "spans": [], + "index": 32 + }, + { + "bbox": [ + 345, + 169.33333333333331, + 502, + 181.99999999999997 + ], + "spans": [], + "index": 33 + }, + { + "bbox": [ + 345, + 181.99999999999997, + 502, + 194.66666666666663 + ], + "spans": [], + "index": 34 + }, + { + "bbox": [ + 345, + 194.66666666666663, + 502, + 207.3333333333333 + ], + "spans": [], + "index": 35 + }, + { + "bbox": [ + 345, + 207.3333333333333, + 502, + 219.99999999999994 + ], + "spans": [], + "index": 36 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 344, + 226, + 505, + 426 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 343, + 225, + 505, + 237 + ], + "spans": [ + { + "bbox": [ + 343, + 225, + 505, + 237 + ], + "score": 1.0, + "content": "Figure 1: Regional-to-Local Attention", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 343, + 235, + 505, + 246 + ], + "spans": [ + { + "bbox": [ + 343, + 235, + 505, + 246 + ], + "score": 1.0, + "content": "for Vision Transformers. (a) ViT uses", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 343, + 246, + 506, + 256 + ], + "spans": [ + { + "bbox": [ + 343, + 246, + 506, + 256 + ], + "score": 1.0, + "content": "a fixed patch size through the whole net-", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 343, + 256, + 505, + 267 + ], + "spans": [ + { + "bbox": [ + 343, + 256, + 505, + 267 + ], + "score": 1.0, + "content": "work, (b) PVT adopts a pyramid structure", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 343, + 266, + 506, + 277 + ], + "spans": [ + { + "bbox": [ + 343, + 266, + 506, + 277 + ], + "score": 1.0, + "content": "to gradually enlarge the patch size in the net-", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 343, + 276, + 505, + 286 + ], + "spans": [ + { + "bbox": [ + 343, + 276, + 505, + 286 + ], + "score": 1.0, + "content": "work. Both ViT and PVT uses all tokens", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 343, + 285, + 506, + 297 + ], + "spans": [ + { + "bbox": [ + 343, + 285, + 506, + 297 + ], + "score": 1.0, + "content": "in self-attention, which are computational-", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 343, + 295, + 506, + 307 + ], + "spans": [ + { + "bbox": [ + 343, + 295, + 506, + 307 + ], + "score": 1.0, + "content": "and memory-intensive. (c) Our proposed ap-", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 343, + 306, + 505, + 316 + ], + "spans": [ + { + "bbox": [ + 343, + 306, + 505, + 316 + ], + "score": 1.0, + "content": "proach combines a pyramid structure with", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 343, + 316, + 505, + 326 + ], + "spans": [ + { + "bbox": [ + 343, + 316, + 505, + 326 + ], + "score": 1.0, + "content": "an efficient regional-to-local (R2L) attention", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 344, + 326, + 506, + 336 + ], + "spans": [ + { + "bbox": [ + 344, + 326, + 506, + 336 + ], + "score": 1.0, + "content": "mechanism to reduce computation and mem-", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 343, + 336, + 505, + 346 + ], + "spans": [ + { + "bbox": [ + 343, + 336, + 505, + 346 + ], + "score": 1.0, + "content": "ory usage. Our approach divides the input", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 343, + 345, + 506, + 356 + ], + "spans": [ + { + "bbox": [ + 343, + 345, + 506, + 356 + ], + "score": 1.0, + "content": "image into two groups of tokens, regional to-", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 344, + 356, + 505, + 366 + ], + "spans": [ + { + "bbox": [ + 344, + 356, + 505, + 366 + ], + "score": 1.0, + "content": "kens of large patch size (red) and local ones", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 343, + 365, + 506, + 376 + ], + "spans": [ + { + "bbox": [ + 343, + 365, + 506, + 376 + ], + "score": 1.0, + "content": "of small patch size (black). The two types of", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 343, + 375, + 506, + 385 + ], + "spans": [ + { + "bbox": [ + 343, + 375, + 506, + 385 + ], + "score": 1.0, + "content": "tokens communicate efficiently through R2L", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 343, + 385, + 505, + 396 + ], + "spans": [ + { + "bbox": [ + 343, + 385, + 505, + 396 + ], + "score": 1.0, + "content": "attention, which jointly attends to the local", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 343, + 395, + 505, + 406 + ], + "spans": [ + { + "bbox": [ + 343, + 395, + 505, + 406 + ], + "score": 1.0, + "content": "tokens in the same region and the associated", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 343, + 405, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 343, + 405, + 505, + 416 + ], + "score": 1.0, + "content": "regional token. Note that the numbers denote", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 344, + 416, + 489, + 426 + ], + "spans": [ + { + "bbox": [ + 344, + 416, + 489, + 426 + ], + "score": 1.0, + "content": "the patch sizes at each stage of a model.", + "type": "text" + } + ], + "index": 56 + } + ], + "index": 46.5 + } + ], + "index": 40.0 + }, + { + "type": "text", + "bbox": [ + 107, + 429, + 504, + 461 + ], + "lines": [ + { + "bbox": [ + 105, + 426, + 505, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 270, + 441 + ], + "score": 1.0, + "content": "R2L attention can be up to as much as", + "type": "text" + }, + { + "bbox": [ + 270, + 428, + 290, + 439 + ], + "score": 0.88, + "content": "73 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 426, + 505, + 441 + ], + "score": 1.0, + "content": ". We demonstrate the effectiveness of our approach", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 106, + 439, + 505, + 451 + ], + "spans": [ + { + "bbox": [ + 106, + 439, + 505, + 451 + ], + "score": 1.0, + "content": "on image classification and several downstream vision tasks including object detection and action", + "type": "text" + } + ], + "index": 58 + }, + { + "bbox": [ + 105, + 450, + 157, + 464 + ], + "spans": [ + { + "bbox": [ + 105, + 450, + 157, + 464 + ], + "score": 1.0, + "content": "recognition.", + "type": "text" + } + ], + "index": 59 + } + ], + "index": 58, + "bbox_fs": [ + 105, + 426, + 505, + 464 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 467, + 364, + 478 + ], + "lines": [ + { + "bbox": [ + 106, + 466, + 365, + 479 + ], + "spans": [ + { + "bbox": [ + 106, + 466, + 365, + 479 + ], + "score": 1.0, + "content": "To summarize, our key contributions in this work are as follows:", + "type": "text" + } + ], + "index": 60 + } + ], + "index": 60, + "bbox_fs": [ + 106, + 466, + 365, + 479 + ] + }, + { + "type": "list", + "bbox": [ + 107, + 487, + 506, + 615 + ], + "lines": [ + { + "bbox": [ + 106, + 487, + 505, + 499 + ], + "spans": [ + { + "bbox": [ + 106, + 487, + 505, + 499 + ], + "score": 1.0, + "content": "1. We propose a new vision transformer (RegionViT) based on regional-to-local attention to learn", + "type": "text" + } + ], + "index": 61, + "is_list_start_line": true + }, + { + "bbox": [ + 118, + 498, + 505, + 510 + ], + "spans": [ + { + "bbox": [ + 118, + 498, + 505, + 510 + ], + "score": 1.0, + "content": "both local and global features. Our proposed regional-to-local attention alleviates the overhead of", + "type": "text" + } + ], + "index": 62 + }, + { + "bbox": [ + 119, + 509, + 505, + 521 + ], + "spans": [ + { + "bbox": [ + 119, + 509, + 505, + 521 + ], + "score": 1.0, + "content": "standard global attention (too many tokens) and the weakness of pure local attention (no interaction", + "type": "text" + } + ], + "index": 63 + }, + { + "bbox": [ + 118, + 520, + 337, + 533 + ], + "spans": [ + { + "bbox": [ + 118, + 520, + 337, + 533 + ], + "score": 1.0, + "content": "between regions) used in existing vision transformers.", + "type": "text" + } + ], + "index": 64, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 533, + 505, + 547 + ], + "spans": [ + { + "bbox": [ + 105, + 533, + 505, + 547 + ], + "score": 1.0, + "content": "2. Our regional-to-local attention reduces the memory complexity significantly as compared to", + "type": "text" + } + ], + "index": 65, + "is_list_start_line": true + }, + { + "bbox": [ + 117, + 544, + 506, + 558 + ], + "spans": [ + { + "bbox": [ + 117, + 544, + 429, + 558 + ], + "score": 1.0, + "content": "standard self-attention, leading to a savings in memory complexity by about", + "type": "text" + }, + { + "bbox": [ + 429, + 545, + 473, + 557 + ], + "score": 0.93, + "content": "\\dot { \\mathcal { O } } ( N / M ^ { 2 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 474, + 544, + 506, + 558 + ], + "score": 1.0, + "content": ", where", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 119, + 554, + 506, + 570 + ], + "spans": [ + { + "bbox": [ + 119, + 556, + 131, + 566 + ], + "score": 0.75, + "content": "M", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 554, + 270, + 570 + ], + "score": 1.0, + "content": "is the window size of a region and", + "type": "text" + }, + { + "bbox": [ + 271, + 556, + 281, + 566 + ], + "score": 0.79, + "content": "N", + "type": "inline_equation" + }, + { + "bbox": [ + 281, + 554, + 506, + 570 + ], + "score": 1.0, + "content": "is the total number of tokens. This effectively allows us", + "type": "text" + } + ], + "index": 67 + }, + { + "bbox": [ + 118, + 566, + 471, + 581 + ], + "spans": [ + { + "bbox": [ + 118, + 566, + 471, + 581 + ], + "score": 1.0, + "content": "to train a more deep network for better performance while with comparable complexity.", + "type": "text" + } + ], + "index": 68, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 582, + 506, + 595 + ], + "spans": [ + { + "bbox": [ + 106, + 582, + 506, + 595 + ], + "score": 1.0, + "content": "3. Our models outperform or on par with several concurrent works on vision transformer that exploit", + "type": "text" + } + ], + "index": 69, + "is_list_start_line": true + }, + { + "bbox": [ + 117, + 592, + 506, + 605 + ], + "spans": [ + { + "bbox": [ + 117, + 592, + 506, + 605 + ], + "score": 1.0, + "content": "pyramid structure for image classification. Experiments also demonstrate that our models work", + "type": "text" + } + ], + "index": 70 + }, + { + "bbox": [ + 119, + 604, + 506, + 615 + ], + "spans": [ + { + "bbox": [ + 119, + 604, + 506, + 615 + ], + "score": 1.0, + "content": "well on several downstream classification tasks, including object detection and action recognition.", + "type": "text" + } + ], + "index": 71, + "is_list_end_line": true + } + ], + "index": 66, + "bbox_fs": [ + 105, + 487, + 506, + 615 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 630, + 211, + 643 + ], + "lines": [ + { + "bbox": [ + 104, + 628, + 213, + 645 + ], + "spans": [ + { + "bbox": [ + 104, + 628, + 213, + 645 + ], + "score": 1.0, + "content": "2 RELATED WORK", + "type": "text" + } + ], + "index": 72 + } + ], + "index": 72 + }, + { + "type": "text", + "bbox": [ + 106, + 655, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 652, + 507, + 669 + ], + "spans": [ + { + "bbox": [ + 105, + 652, + 507, + 669 + ], + "score": 1.0, + "content": "CNNs with Attention. Attention has been widely used for enhancing the features in CNNs, e.g.,", + "type": "text" + } + ], + "index": 73 + }, + { + "bbox": [ + 105, + 664, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 506, + 678 + ], + "score": 1.0, + "content": "SENet (Hu et al., 2018) and CBAM (Woo et al., 2018). Various methods have been proposed that", + "type": "text" + } + ], + "index": 74 + }, + { + "bbox": [ + 105, + 676, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 689 + ], + "score": 1.0, + "content": "combine self-attention with CNNs (Bello, 2021; Srinivas et al., 2021; Zhao et al., 2020; Ramachandran", + "type": "text" + } + ], + "index": 75 + }, + { + "bbox": [ + 104, + 685, + 507, + 702 + ], + "spans": [ + { + "bbox": [ + 104, + 685, + 507, + 702 + ], + "score": 1.0, + "content": "et al., 2019a; Bello et al., 2019; Hu et al., 2019; Ramachandran et al., 2019b; Wang et al., 2018). E.g.,", + "type": "text" + } + ], + "index": 76 + }, + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "SAN (Zhao et al., 2020) and SASA (Ramachandran et al., 2019a) attempts to replace all convolutional", + "type": "text" + } + ], + "index": 77 + }, + { + "bbox": [ + 106, + 710, + 504, + 721 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 504, + 721 + ], + "score": 1.0, + "content": "layers with local self-attention network. While these works show promising results, their complexity", + "type": "text" + } + ], + "index": 78 + }, + { + "bbox": [ + 105, + 720, + 504, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 504, + 732 + ], + "score": 1.0, + "content": "is relatively high due to the self-attention. On the other hand, BoTNet (Srinivas et al., 2021) replaces", + "type": "text" + } + ], + "index": 79 + }, + { + "bbox": [ + 105, + 179, + 507, + 194 + ], + "spans": [ + { + "bbox": [ + 105, + 179, + 507, + 194 + ], + "score": 1.0, + "content": "few convolutional layers with a slightly-modified self-attention to balance computation and accuracy.", + "type": "text", + "cross_page": true + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 191, + 506, + 204 + ], + "spans": [ + { + "bbox": [ + 105, + 191, + 506, + 204 + ], + "score": 1.0, + "content": "LambdaNet work (Bello, 2021) uses the approximated self-attention to reduce the overhead of self-", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 202, + 506, + 215 + ], + "spans": [ + { + "bbox": [ + 105, + 202, + 506, + 215 + ], + "score": 1.0, + "content": "attention and make the network efficient. Few works use attention approach to define the regional of", + "type": "text", + "cross_page": true + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 213, + 506, + 226 + ], + "spans": [ + { + "bbox": [ + 105, + 213, + 506, + 226 + ], + "score": 1.0, + "content": "interest for the fine-grained visual recognition (Zheng et al., 2020; Ding et al., 2021; Wharton et al.,", + "type": "text", + "cross_page": true + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 224, + 505, + 237 + ], + "spans": [ + { + "bbox": [ + 105, + 224, + 505, + 237 + ], + "score": 1.0, + "content": "2021). In contrast, our model utilizes regional-to-local attention to alleviate the workload of global", + "type": "text", + "cross_page": true + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 236, + 490, + 248 + ], + "spans": [ + { + "bbox": [ + 106, + 236, + 490, + 248 + ], + "score": 1.0, + "content": "self-attention by local attention while still keeping the global information via regional attention.", + "type": "text", + "cross_page": true + } + ], + "index": 11 + } + ], + "index": 76, + "bbox_fs": [ + 104, + 652, + 507, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 107, + 102, + 504, + 167 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 81, + 504, + 101 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "score": 1.0, + "content": "Table 1: Comparison to related works. Most works use non-overlapped windows to group tokens, and then", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 91, + 422, + 103 + ], + "spans": [ + { + "bbox": [ + 105, + 91, + 422, + 103 + ], + "score": 1.0, + "content": "propose the corresponding methods to assure the information exchange among regions.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 107, + 102, + 504, + 167 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 102, + 504, + 167 + ], + "spans": [ + { + "bbox": [ + 107, + 102, + 504, + 167 + ], + "score": 0.98, + "html": "
MethodsStructureAttention TypesWindows of Local AttentionHandling Non-overlapped RegionsGlobal Tokens
ViT (Dosovitskiy et al.,2021)IsotropicGlobalN/AN/AN/A
PVT (Wang et al., 2021)PyramidGlobalN/AN/AN/A
Swin (Liu et al.,2021)PyramidLocalStrictly Non-overlappedShifting the windowNone
ViL (Zhang et al., 2021a)PyramidLocal + GlobalOverlappedN/A1One learnable token
Twins (Chu et al., 2021a)PyramidLocal + GlobalNon-overlappedGASubsampled from local tokens
RegionViT (Ours)PyramidLocal + RegionalNon-overlappedRegional-to-local attentionRegional tokens
", + "type": "table", + "image_path": "8901d968be699760067393e6573948b9e2a62c788948dae420e0d747f1c8c2c0.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 107, + 102, + 504, + 123.66666666666667 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 107, + 123.66666666666667, + 504, + 145.33333333333334 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 107, + 145.33333333333334, + 504, + 167.0 + ], + "spans": [], + "index": 4 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 109, + 167, + 279, + 173 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 109, + 165, + 280, + 174 + ], + "spans": [ + { + "bbox": [ + 109, + 165, + 280, + 174 + ], + "score": 1.0, + "content": "GA: global attention. 1 : not applicable as ViL used overlapped windows.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 107, + 181, + 505, + 248 + ], + "lines": [ + { + "bbox": [ + 105, + 179, + 507, + 194 + ], + "spans": [ + { + "bbox": [ + 105, + 179, + 507, + 194 + ], + "score": 1.0, + "content": "few convolutional layers with a slightly-modified self-attention to balance computation and accuracy.", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 191, + 506, + 204 + ], + "spans": [ + { + "bbox": [ + 105, + 191, + 506, + 204 + ], + "score": 1.0, + "content": "LambdaNet work (Bello, 2021) uses the approximated self-attention to reduce the overhead of self-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 202, + 506, + 215 + ], + "spans": [ + { + "bbox": [ + 105, + 202, + 506, + 215 + ], + "score": 1.0, + "content": "attention and make the network efficient. Few works use attention approach to define the regional of", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 213, + 506, + 226 + ], + "spans": [ + { + "bbox": [ + 105, + 213, + 506, + 226 + ], + "score": 1.0, + "content": "interest for the fine-grained visual recognition (Zheng et al., 2020; Ding et al., 2021; Wharton et al.,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 224, + 505, + 237 + ], + "spans": [ + { + "bbox": [ + 105, + 224, + 505, + 237 + ], + "score": 1.0, + "content": "2021). In contrast, our model utilizes regional-to-local attention to alleviate the workload of global", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 236, + 490, + 248 + ], + "spans": [ + { + "bbox": [ + 106, + 236, + 490, + 248 + ], + "score": 1.0, + "content": "self-attention by local attention while still keeping the global information via regional attention.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 107, + 253, + 505, + 516 + ], + "lines": [ + { + "bbox": [ + 106, + 252, + 505, + 264 + ], + "spans": [ + { + "bbox": [ + 106, + 252, + 505, + 264 + ], + "score": 1.0, + "content": "Vision Transformer. ViT (Dosovitskiy et al., 2021) has recently achieved comparable results", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 263, + 506, + 276 + ], + "spans": [ + { + "bbox": [ + 106, + 263, + 506, + 276 + ], + "score": 1.0, + "content": "to CNNs when using a huge amount of data, e.g., JFT300M (Sun et al., 2017). DeiT (Touvron", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 274, + 504, + 286 + ], + "spans": [ + { + "bbox": [ + 106, + 274, + 504, + 286 + ], + "score": 1.0, + "content": "et al., 2020; Wightman, 2019) subsequently proposes an efficient training scheme that allows vision", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 285, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 285, + 505, + 298 + ], + "score": 1.0, + "content": "transformer to work on par with CNNs while training only on ImageNet1K (Deng et al., 2009).", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 295, + 506, + 310 + ], + "spans": [ + { + "bbox": [ + 105, + 295, + 506, + 310 + ], + "score": 1.0, + "content": "Despite promising performance of ViT, the architecture is directly borrowed from natural language", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 307, + 506, + 319 + ], + "spans": [ + { + "bbox": [ + 105, + 307, + 506, + 319 + ], + "score": 1.0, + "content": "processing, which might not be suitable for vision applications. Motivated by this, two main line of", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 318, + 504, + 330 + ], + "spans": [ + { + "bbox": [ + 106, + 318, + 504, + 330 + ], + "score": 1.0, + "content": "research works have been developed to improve ViT. One is to enhance different components of the", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 329, + 505, + 342 + ], + "spans": [ + { + "bbox": [ + 106, + 329, + 505, + 342 + ], + "score": 1.0, + "content": "vision transformer (Yuan et al., 2021; Han et al., 2021; Touvron et al., 2021; Jiang et al., 2021) while", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 340, + 504, + 352 + ], + "spans": [ + { + "bbox": [ + 106, + 340, + 504, + 352 + ], + "score": 1.0, + "content": "still using isotropic structure (i.e., fixed token numbers and channel dimension) like ViT, e.g., while", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 352, + 505, + 363 + ], + "spans": [ + { + "bbox": [ + 106, + 352, + 505, + 363 + ], + "score": 1.0, + "content": "T2T-ViT (Yuan et al., 2021) introduces a Tokens-to-Token (T2T) transformation to encode local", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 361, + 505, + 374 + ], + "spans": [ + { + "bbox": [ + 106, + 361, + 505, + 374 + ], + "score": 1.0, + "content": "structure for each token, CrossViT propose a dual-path architecture, each with different scales, to", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 372, + 506, + 385 + ], + "spans": [ + { + "bbox": [ + 105, + 372, + 506, + 385 + ], + "score": 1.0, + "content": "learn multi-scale features (Chen et al., 2021a). CaiT (Touvron et al., 2021) proposes a layer-scalar to", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 384, + 505, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 384, + 505, + 396 + ], + "score": 1.0, + "content": "train a deeper network for better performance and LV-ViT (Jiang et al., 2021) modified how the model", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 395, + 506, + 407 + ], + "spans": [ + { + "bbox": [ + 105, + 395, + 506, + 407 + ], + "score": 1.0, + "content": "is trained when CutMix (Yun et al., 2019) augmentation is applied on ViT. Another parallel thread", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 406, + 505, + 418 + ], + "spans": [ + { + "bbox": [ + 106, + 406, + 505, + 418 + ], + "score": 1.0, + "content": "for improving vision transformer is in incorporating CNN-like pyramid structure into ViT (Wu et al.,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 416, + 505, + 429 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 505, + 429 + ], + "score": 1.0, + "content": "2021; Wang et al., 2021; Heo et al., 2021; Zhang et al., 2021a; Li et al., 2021; Graham et al., 2021;", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 426, + 505, + 440 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 505, + 440 + ], + "score": 1.0, + "content": "Liu et al., 2021; Chu et al., 2021a; Yan et al., 2021; Chen et al., 2021b). PVT (Wang et al., 2021) and", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 438, + 506, + 451 + ], + "spans": [ + { + "bbox": [ + 105, + 438, + 506, + 451 + ], + "score": 1.0, + "content": "PiT (Heo et al., 2021) introduce the pyramid structure into ViT, which makes them more suitable for", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 450, + 505, + 461 + ], + "spans": [ + { + "bbox": [ + 106, + 450, + 505, + 461 + ], + "score": 1.0, + "content": "objection detection as it can provide multi-scale features. LocalViT (Li et al., 2021) and ConT (Yan", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 461, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 106, + 461, + 505, + 472 + ], + "score": 1.0, + "content": "et al., 2021) mix convolutions with self-attention to encode locality information. Swin (Liu et al.,", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 471, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 505, + 484 + ], + "score": 1.0, + "content": "2021), ViL (Zhang et al., 2021a) and Twins (Chu et al., 2021a) limited the self-attention into a local", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 483, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 505, + 495 + ], + "score": 1.0, + "content": "region and then propose different methods to allow the interaction among each local region. Our", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 493, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 505, + 506 + ], + "score": 1.0, + "content": "model also utilizes pyramid structure and limits the self-attention in a local region; while we propose", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 505, + 464, + 517 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 464, + 517 + ], + "score": 1.0, + "content": "the regional-to-local attention to exchange the information among each region efficiently.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 23.5 + }, + { + "type": "text", + "bbox": [ + 107, + 522, + 505, + 664 + ], + "lines": [ + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "score": 1.0, + "content": "Difference from PVT, Swin, ViL and Twins. Table 1 summarizes the difference between our", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 532, + 505, + 545 + ], + "spans": [ + { + "bbox": [ + 105, + 532, + 505, + 545 + ], + "score": 1.0, + "content": "proposed approach, RegionViT with the closely related works. PVT (Wang et al., 2021) uses pyramid", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 543, + 505, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 543, + 505, + 556 + ], + "score": 1.0, + "content": "structure but the scope of self-attention is still global. While ViL (Zhang et al., 2021a) limits", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 554, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 554, + 505, + 567 + ], + "score": 1.0, + "content": "self-attention into a local region, they adopt overlapping windows as convolutions to process all", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 565, + 505, + 578 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 505, + 578 + ], + "score": 1.0, + "content": "the tokens. On the other hand, Swin (Liu et al., 2021), Twins (Chu et al., 2021a) and ours adopt", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 576, + 505, + 589 + ], + "spans": [ + { + "bbox": [ + 105, + 576, + 505, + 589 + ], + "score": 1.0, + "content": "non-overlapped windows. While Swin shifts the position of windows alternatively in the consecutive", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 588, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 505, + 600 + ], + "score": 1.0, + "content": "transformer encoder to allow the interaction between regions, Twins subsamples local tokens as the", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 598, + 505, + 611 + ], + "spans": [ + { + "bbox": [ + 105, + 598, + 505, + 611 + ], + "score": 1.0, + "content": "global tokens, and then use the global tokens as the keys for self-attention to achieve the interaction", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 610, + 505, + 621 + ], + "spans": [ + { + "bbox": [ + 106, + 610, + 505, + 621 + ], + "score": 1.0, + "content": "between regions. By contrast, our proposed method utilizes a extra set of regional tokens with", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 619, + 506, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 619, + 506, + 634 + ], + "score": 1.0, + "content": "regional-to-local attention to perform self-attention on regional tokens only for the global information", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 631, + 506, + 644 + ], + "spans": [ + { + "bbox": [ + 105, + 631, + 506, + 644 + ], + "score": 1.0, + "content": "and then each regional token is sent to the associated local tokens to pass the global information.", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 106, + 643, + 505, + 654 + ], + "spans": [ + { + "bbox": [ + 106, + 643, + 505, + 654 + ], + "score": 1.0, + "content": "One major difference from others is that the local self-attention in our approach includes one extra", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 653, + 381, + 666 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 381, + 666 + ], + "score": 1.0, + "content": "regional token to exchange the global information with local tokens.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 42 + }, + { + "type": "title", + "bbox": [ + 108, + 683, + 172, + 695 + ], + "lines": [ + { + "bbox": [ + 104, + 681, + 174, + 699 + ], + "spans": [ + { + "bbox": [ + 104, + 681, + 174, + 699 + ], + "score": 1.0, + "content": "3 METHOD", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 49 + }, + { + "type": "text", + "bbox": [ + 108, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "Our method is built on top of a vision transformer, so we first present a brief background of ViT and", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 720, + 478, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 478, + 733 + ], + "score": 1.0, + "content": "then describe our method (RegionViT) on regional-to-local attention for vision transformers.", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 50.5 + } + ], + "page_idx": 2, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 303, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 12, + "width": 9 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 107, + 102, + 504, + 167 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 81, + 504, + 101 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 505, + 93 + ], + "score": 1.0, + "content": "Table 1: Comparison to related works. Most works use non-overlapped windows to group tokens, and then", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 91, + 422, + 103 + ], + "spans": [ + { + "bbox": [ + 105, + 91, + 422, + 103 + ], + "score": 1.0, + "content": "propose the corresponding methods to assure the information exchange among regions.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 107, + 102, + 504, + 167 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 102, + 504, + 167 + ], + "spans": [ + { + "bbox": [ + 107, + 102, + 504, + 167 + ], + "score": 0.98, + "html": "
MethodsStructureAttention TypesWindows of Local AttentionHandling Non-overlapped RegionsGlobal Tokens
ViT (Dosovitskiy et al.,2021)IsotropicGlobalN/AN/AN/A
PVT (Wang et al., 2021)PyramidGlobalN/AN/AN/A
Swin (Liu et al.,2021)PyramidLocalStrictly Non-overlappedShifting the windowNone
ViL (Zhang et al., 2021a)PyramidLocal + GlobalOverlappedN/A1One learnable token
Twins (Chu et al., 2021a)PyramidLocal + GlobalNon-overlappedGASubsampled from local tokens
RegionViT (Ours)PyramidLocal + RegionalNon-overlappedRegional-to-local attentionRegional tokens
", + "type": "table", + "image_path": "8901d968be699760067393e6573948b9e2a62c788948dae420e0d747f1c8c2c0.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 107, + 102, + 504, + 123.66666666666667 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 107, + 123.66666666666667, + 504, + 145.33333333333334 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 107, + 145.33333333333334, + 504, + 167.0 + ], + "spans": [], + "index": 4 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 109, + 167, + 279, + 173 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 109, + 165, + 280, + 174 + ], + "spans": [ + { + "bbox": [ + 109, + 165, + 280, + 174 + ], + "score": 1.0, + "content": "GA: global attention. 1 : not applicable as ViL used overlapped windows.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 107, + 181, + 505, + 248 + ], + "lines": [], + "index": 8.5, + "bbox_fs": [ + 105, + 179, + 507, + 248 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 253, + 505, + 516 + ], + "lines": [ + { + "bbox": [ + 106, + 252, + 505, + 264 + ], + "spans": [ + { + "bbox": [ + 106, + 252, + 505, + 264 + ], + "score": 1.0, + "content": "Vision Transformer. ViT (Dosovitskiy et al., 2021) has recently achieved comparable results", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 263, + 506, + 276 + ], + "spans": [ + { + "bbox": [ + 106, + 263, + 506, + 276 + ], + "score": 1.0, + "content": "to CNNs when using a huge amount of data, e.g., JFT300M (Sun et al., 2017). DeiT (Touvron", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 274, + 504, + 286 + ], + "spans": [ + { + "bbox": [ + 106, + 274, + 504, + 286 + ], + "score": 1.0, + "content": "et al., 2020; Wightman, 2019) subsequently proposes an efficient training scheme that allows vision", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 285, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 285, + 505, + 298 + ], + "score": 1.0, + "content": "transformer to work on par with CNNs while training only on ImageNet1K (Deng et al., 2009).", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 295, + 506, + 310 + ], + "spans": [ + { + "bbox": [ + 105, + 295, + 506, + 310 + ], + "score": 1.0, + "content": "Despite promising performance of ViT, the architecture is directly borrowed from natural language", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 307, + 506, + 319 + ], + "spans": [ + { + "bbox": [ + 105, + 307, + 506, + 319 + ], + "score": 1.0, + "content": "processing, which might not be suitable for vision applications. Motivated by this, two main line of", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 318, + 504, + 330 + ], + "spans": [ + { + "bbox": [ + 106, + 318, + 504, + 330 + ], + "score": 1.0, + "content": "research works have been developed to improve ViT. One is to enhance different components of the", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 329, + 505, + 342 + ], + "spans": [ + { + "bbox": [ + 106, + 329, + 505, + 342 + ], + "score": 1.0, + "content": "vision transformer (Yuan et al., 2021; Han et al., 2021; Touvron et al., 2021; Jiang et al., 2021) while", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 340, + 504, + 352 + ], + "spans": [ + { + "bbox": [ + 106, + 340, + 504, + 352 + ], + "score": 1.0, + "content": "still using isotropic structure (i.e., fixed token numbers and channel dimension) like ViT, e.g., while", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 352, + 505, + 363 + ], + "spans": [ + { + "bbox": [ + 106, + 352, + 505, + 363 + ], + "score": 1.0, + "content": "T2T-ViT (Yuan et al., 2021) introduces a Tokens-to-Token (T2T) transformation to encode local", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 361, + 505, + 374 + ], + "spans": [ + { + "bbox": [ + 106, + 361, + 505, + 374 + ], + "score": 1.0, + "content": "structure for each token, CrossViT propose a dual-path architecture, each with different scales, to", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 372, + 506, + 385 + ], + "spans": [ + { + "bbox": [ + 105, + 372, + 506, + 385 + ], + "score": 1.0, + "content": "learn multi-scale features (Chen et al., 2021a). CaiT (Touvron et al., 2021) proposes a layer-scalar to", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 384, + 505, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 384, + 505, + 396 + ], + "score": 1.0, + "content": "train a deeper network for better performance and LV-ViT (Jiang et al., 2021) modified how the model", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 395, + 506, + 407 + ], + "spans": [ + { + "bbox": [ + 105, + 395, + 506, + 407 + ], + "score": 1.0, + "content": "is trained when CutMix (Yun et al., 2019) augmentation is applied on ViT. Another parallel thread", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 406, + 505, + 418 + ], + "spans": [ + { + "bbox": [ + 106, + 406, + 505, + 418 + ], + "score": 1.0, + "content": "for improving vision transformer is in incorporating CNN-like pyramid structure into ViT (Wu et al.,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 416, + 505, + 429 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 505, + 429 + ], + "score": 1.0, + "content": "2021; Wang et al., 2021; Heo et al., 2021; Zhang et al., 2021a; Li et al., 2021; Graham et al., 2021;", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 426, + 505, + 440 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 505, + 440 + ], + "score": 1.0, + "content": "Liu et al., 2021; Chu et al., 2021a; Yan et al., 2021; Chen et al., 2021b). PVT (Wang et al., 2021) and", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 438, + 506, + 451 + ], + "spans": [ + { + "bbox": [ + 105, + 438, + 506, + 451 + ], + "score": 1.0, + "content": "PiT (Heo et al., 2021) introduce the pyramid structure into ViT, which makes them more suitable for", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 450, + 505, + 461 + ], + "spans": [ + { + "bbox": [ + 106, + 450, + 505, + 461 + ], + "score": 1.0, + "content": "objection detection as it can provide multi-scale features. LocalViT (Li et al., 2021) and ConT (Yan", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 461, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 106, + 461, + 505, + 472 + ], + "score": 1.0, + "content": "et al., 2021) mix convolutions with self-attention to encode locality information. Swin (Liu et al.,", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 471, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 505, + 484 + ], + "score": 1.0, + "content": "2021), ViL (Zhang et al., 2021a) and Twins (Chu et al., 2021a) limited the self-attention into a local", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 483, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 505, + 495 + ], + "score": 1.0, + "content": "region and then propose different methods to allow the interaction among each local region. Our", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 493, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 505, + 506 + ], + "score": 1.0, + "content": "model also utilizes pyramid structure and limits the self-attention in a local region; while we propose", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 505, + 464, + 517 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 464, + 517 + ], + "score": 1.0, + "content": "the regional-to-local attention to exchange the information among each region efficiently.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 23.5, + "bbox_fs": [ + 105, + 252, + 506, + 517 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 522, + 505, + 664 + ], + "lines": [ + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "score": 1.0, + "content": "Difference from PVT, Swin, ViL and Twins. Table 1 summarizes the difference between our", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 532, + 505, + 545 + ], + "spans": [ + { + "bbox": [ + 105, + 532, + 505, + 545 + ], + "score": 1.0, + "content": "proposed approach, RegionViT with the closely related works. PVT (Wang et al., 2021) uses pyramid", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 543, + 505, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 543, + 505, + 556 + ], + "score": 1.0, + "content": "structure but the scope of self-attention is still global. While ViL (Zhang et al., 2021a) limits", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 554, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 554, + 505, + 567 + ], + "score": 1.0, + "content": "self-attention into a local region, they adopt overlapping windows as convolutions to process all", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 565, + 505, + 578 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 505, + 578 + ], + "score": 1.0, + "content": "the tokens. On the other hand, Swin (Liu et al., 2021), Twins (Chu et al., 2021a) and ours adopt", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 576, + 505, + 589 + ], + "spans": [ + { + "bbox": [ + 105, + 576, + 505, + 589 + ], + "score": 1.0, + "content": "non-overlapped windows. While Swin shifts the position of windows alternatively in the consecutive", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 588, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 505, + 600 + ], + "score": 1.0, + "content": "transformer encoder to allow the interaction between regions, Twins subsamples local tokens as the", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 598, + 505, + 611 + ], + "spans": [ + { + "bbox": [ + 105, + 598, + 505, + 611 + ], + "score": 1.0, + "content": "global tokens, and then use the global tokens as the keys for self-attention to achieve the interaction", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 610, + 505, + 621 + ], + "spans": [ + { + "bbox": [ + 106, + 610, + 505, + 621 + ], + "score": 1.0, + "content": "between regions. By contrast, our proposed method utilizes a extra set of regional tokens with", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 619, + 506, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 619, + 506, + 634 + ], + "score": 1.0, + "content": "regional-to-local attention to perform self-attention on regional tokens only for the global information", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 631, + 506, + 644 + ], + "spans": [ + { + "bbox": [ + 105, + 631, + 506, + 644 + ], + "score": 1.0, + "content": "and then each regional token is sent to the associated local tokens to pass the global information.", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 106, + 643, + 505, + 654 + ], + "spans": [ + { + "bbox": [ + 106, + 643, + 505, + 654 + ], + "score": 1.0, + "content": "One major difference from others is that the local self-attention in our approach includes one extra", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 653, + 381, + 666 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 381, + 666 + ], + "score": 1.0, + "content": "regional token to exchange the global information with local tokens.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 42, + "bbox_fs": [ + 105, + 521, + 506, + 666 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 683, + 172, + 695 + ], + "lines": [ + { + "bbox": [ + 104, + 681, + 174, + 699 + ], + "spans": [ + { + "bbox": [ + 104, + 681, + 174, + 699 + ], + "score": 1.0, + "content": "3 METHOD", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 49 + }, + { + "type": "text", + "bbox": [ + 108, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "Our method is built on top of a vision transformer, so we first present a brief background of ViT and", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 720, + 478, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 478, + 733 + ], + "score": 1.0, + "content": "then describe our method (RegionViT) on regional-to-local attention for vision transformers.", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 50.5, + "bbox_fs": [ + 106, + 709, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 108, + 80, + 504, + 138 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 80, + 504, + 138 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 80, + 504, + 138 + ], + "spans": [ + { + "bbox": [ + 108, + 80, + 504, + 138 + ], + "score": 0.954, + "type": "image", + "image_path": "4f51e129978441e6204aba5ef5a836f32d26b9a369736a313e286198ad637a3c.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 108, + 80, + 504, + 99.33333333333333 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 108, + 99.33333333333333, + 504, + 118.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 118.66666666666666, + 504, + 138.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 143, + 505, + 184 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 142, + 505, + 157 + ], + "spans": [ + { + "bbox": [ + 105, + 142, + 505, + 157 + ], + "score": 1.0, + "content": "Figure 2: Architecture of the proposed RegionViT. Two paths are in our proposed network, including", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 154, + 505, + 165 + ], + "spans": [ + { + "bbox": [ + 105, + 154, + 505, + 165 + ], + "score": 1.0, + "content": "regional tokens and local tokens, and their information is exchanged in the regional-to-local (R2L) transformer", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 163, + 505, + 175 + ], + "spans": [ + { + "bbox": [ + 106, + 163, + 505, + 175 + ], + "score": 1.0, + "content": "encoders. In the end, we average all regional tokens and use it for the classification. The tensor shape here is", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 172, + 450, + 185 + ], + "spans": [ + { + "bbox": [ + 105, + 172, + 305, + 185 + ], + "score": 1.0, + "content": "computed based on that regional tokens take a patch of", + "type": "text" + }, + { + "bbox": [ + 306, + 173, + 320, + 183 + ], + "score": 0.91, + "content": "2 8 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 320, + 172, + 436, + 185 + ], + "score": 1.0, + "content": "and local tokens take a patch of", + "type": "text" + }, + { + "bbox": [ + 436, + 173, + 446, + 183 + ], + "score": 0.84, + "content": "4 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 446, + 172, + 450, + 185 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "text", + "bbox": [ + 106, + 197, + 505, + 297 + ], + "lines": [ + { + "bbox": [ + 105, + 196, + 505, + 210 + ], + "spans": [ + { + "bbox": [ + 105, + 196, + 505, + 210 + ], + "score": 1.0, + "content": "Vision Transformer. As opposed to CNN-based approaches for image classification, ViT is a purely", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 209, + 505, + 221 + ], + "spans": [ + { + "bbox": [ + 106, + 209, + 505, + 221 + ], + "score": 1.0, + "content": "attention-based counterpart, borrowed from NLP. It consists of stacked transformer encoders, each", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 219, + 506, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 219, + 506, + 233 + ], + "score": 1.0, + "content": "of which contains a multihead self-attention (MSA) and a feed-forward network (FFN) with layer", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "spans": [ + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "score": 1.0, + "content": "normalization and residual shortcut. To classify an image, ViT first splits it into patches of fixed size,", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 241, + 505, + 254 + ], + "spans": [ + { + "bbox": [ + 105, + 241, + 124, + 254 + ], + "score": 1.0, + "content": "e.g.", + "type": "text" + }, + { + "bbox": [ + 124, + 242, + 153, + 253 + ], + "score": 0.87, + "content": "1 6 \\times 1 6", + "type": "inline_equation" + }, + { + "bbox": [ + 153, + 241, + 505, + 254 + ], + "score": 1.0, + "content": ", and then transforms them into tokens by linear projection. A class token is additionally", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 252, + 506, + 266 + ], + "spans": [ + { + "bbox": [ + 104, + 252, + 506, + 266 + ], + "score": 1.0, + "content": "prepended to the patch tokens to form the input sequence. Before the token are fed into transformer", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 264, + 505, + 276 + ], + "spans": [ + { + "bbox": [ + 106, + 264, + 505, + 276 + ], + "score": 1.0, + "content": "encoders, a learnable absolute positional embedding is added to each token to learn the position", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 274, + 505, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 274, + 505, + 287 + ], + "score": 1.0, + "content": "information. At the end of the network, the class token is used as the final feature representation for", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 285, + 329, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 285, + 329, + 298 + ], + "score": 1.0, + "content": "classification. Mathematically, ViT can be expressed as", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 11 + }, + { + "type": "interline_equation", + "bbox": [ + 189, + 302, + 420, + 332 + ], + "lines": [ + { + "bbox": [ + 189, + 302, + 420, + 332 + ], + "spans": [ + { + "bbox": [ + 189, + 302, + 420, + 332 + ], + "score": 0.92, + "content": "\\begin{array} { r l } & { \\mathbf { x } _ { 0 } = [ \\mathbf { x } _ { c l s } | | \\mathbf { x } _ { p a t c h } ] + \\mathbf { x } _ { p o s } } \\\\ & { \\mathbf { y } _ { k } = \\mathbf { x } _ { k - 1 } + \\mathtt { M S A } \\big ( \\mathtt { L N } ( \\mathbf { x } _ { k - 1 } ) \\big ) , \\mathbf { x } _ { k } = \\mathbf { y } _ { k } + \\mathtt { F F N } \\big ( \\mathtt { L N } \\big ( \\mathbf { y } _ { k } \\big ) \\big ) , } \\end{array}", + "type": "interline_equation", + "image_path": "7dfeeb8d19040ab77e4e3b785ed737ff3cd63c3753977f2581532a9ad733431d.jpg" + } + ] + } + ], + "index": 16.5, + "virtual_lines": [ + { + "bbox": [ + 189, + 302, + 420, + 317.0 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 189, + 317.0, + 420, + 332.0 + ], + "spans": [], + "index": 17 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 336, + 505, + 373 + ], + "lines": [ + { + "bbox": [ + 104, + 334, + 507, + 353 + ], + "spans": [ + { + "bbox": [ + 104, + 334, + 134, + 353 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 134, + 337, + 191, + 349 + ], + "score": 0.92, + "content": "\\mathbf { x } _ { c l s } ~ \\in ~ \\mathbb { R } ^ { 1 \\times C }", + "type": "inline_equation" + }, + { + "bbox": [ + 191, + 334, + 211, + 353 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 211, + 337, + 280, + 351 + ], + "score": 0.92, + "content": "\\mathbf { x } _ { p a t c h } \\in \\mathbb { R } ^ { N \\times C }", + "type": "inline_equation" + }, + { + "bbox": [ + 281, + 334, + 507, + 353 + ], + "score": 1.0, + "content": "are the class token and patch tokens respectively and", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 347, + 504, + 365 + ], + "spans": [ + { + "bbox": [ + 106, + 349, + 181, + 363 + ], + "score": 0.91, + "content": "\\mathbf { x } _ { p o s } \\in \\mathbb { R } ^ { ( 1 + N ) \\times C }", + "type": "inline_equation" + }, + { + "bbox": [ + 181, + 347, + 305, + 365 + ], + "score": 1.0, + "content": "is the position embedding, and", + "type": "text" + }, + { + "bbox": [ + 306, + 350, + 322, + 363 + ], + "score": 0.67, + "content": "\\left[ \\left| \\left| \\right. \\right] \\right.", + "type": "inline_equation" + }, + { + "bbox": [ + 323, + 347, + 455, + 365 + ], + "score": 1.0, + "content": "denotes the tensor concatenation.", + "type": "text" + }, + { + "bbox": [ + 456, + 351, + 477, + 361 + ], + "score": 0.32, + "content": "k , N", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 347, + 495, + 365 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 495, + 351, + 504, + 361 + ], + "score": 0.8, + "content": "C", + "type": "inline_equation" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 361, + 484, + 375 + ], + "spans": [ + { + "bbox": [ + 105, + 361, + 484, + 375 + ], + "score": 1.0, + "content": "are the layer index, the number of patch tokens and dimension of the embedding, respectively.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 107, + 378, + 505, + 445 + ], + "lines": [ + { + "bbox": [ + 105, + 378, + 505, + 391 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 505, + 391 + ], + "score": 1.0, + "content": "While the vanilla ViT is ideally capable of learning global interaction among all the patch tokens, the", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "score": 1.0, + "content": "memory complexity of self-attention becomes high when there are many tokens as the complexity", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 401, + 505, + 414 + ], + "spans": [ + { + "bbox": [ + 105, + 401, + 505, + 414 + ], + "score": 1.0, + "content": "is quadratically linear to the length of the input sequence. Moreover, the use of isotropic structure", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 411, + 505, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 505, + 424 + ], + "score": 1.0, + "content": "limits the capability of extending the vanilla ViT model to many vision applications that require", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 422, + 505, + 435 + ], + "spans": [ + { + "bbox": [ + 105, + 422, + 505, + 435 + ], + "score": 1.0, + "content": "high-resolution details, e.g., object detection. To address these issues, we propose regional-to-local", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 434, + 321, + 446 + ], + "spans": [ + { + "bbox": [ + 106, + 434, + 321, + 446 + ], + "score": 1.0, + "content": "attention for vision transformer, RegionViT for short.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 23.5 + }, + { + "type": "text", + "bbox": [ + 106, + 450, + 505, + 615 + ], + "lines": [ + { + "bbox": [ + 105, + 450, + 505, + 463 + ], + "spans": [ + { + "bbox": [ + 105, + 450, + 505, + 463 + ], + "score": 1.0, + "content": "An Overview of Our Proposed Approach (RegionViT). Figure 2 illustrates the architecture of the", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 461, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 473 + ], + "score": 1.0, + "content": "proposed vision transformer, which consists of two tokenization processes that convert an image into", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "regional (upper path) and local tokens (lower path). Each tokenization is a convolution with different", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 483, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 359, + 496 + ], + "score": 1.0, + "content": "patch sizes, e.g., in Figure 2, the patch size of regional tokens is", + "type": "text" + }, + { + "bbox": [ + 359, + 483, + 375, + 493 + ], + "score": 0.88, + "content": "2 8 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 375, + 483, + 400, + 496 + ], + "score": 1.0, + "content": "while", + "type": "text" + }, + { + "bbox": [ + 400, + 483, + 411, + 493 + ], + "score": 0.85, + "content": "4 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 411, + 483, + 505, + 496 + ], + "score": 1.0, + "content": "is used for local tokens", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 104, + 493, + 506, + 506 + ], + "spans": [ + { + "bbox": [ + 104, + 493, + 221, + 506 + ], + "score": 1.0, + "content": "with dimensions projected to", + "type": "text" + }, + { + "bbox": [ + 222, + 495, + 230, + 504 + ], + "score": 0.81, + "content": "C", + "type": "inline_equation" + }, + { + "bbox": [ + 231, + 493, + 407, + 506 + ], + "score": 1.0, + "content": ", which means that one regional token covers", + "type": "text" + }, + { + "bbox": [ + 407, + 493, + 418, + 504 + ], + "score": 0.83, + "content": "7 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 418, + 493, + 506, + 506 + ], + "score": 1.0, + "content": "local tokens based on", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 504, + 506, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 504, + 370, + 518 + ], + "score": 1.0, + "content": "the spatial locality, leading to the window size of a local region to", + "type": "text" + }, + { + "bbox": [ + 371, + 505, + 381, + 515 + ], + "score": 0.77, + "content": "7 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 381, + 504, + 506, + 518 + ], + "score": 1.0, + "content": ". At stage 1, two sets of tokens", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 515, + 507, + 530 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 507, + 530 + ], + "score": 1.0, + "content": "are passed through the proposed regional-to-local transformer encoders. However, for the later stages,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 527, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 106, + 527, + 506, + 540 + ], + "score": 1.0, + "content": "to balance the computational load and to have feature maps at different resolutions, we deploy a", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 537, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 550 + ], + "score": 1.0, + "content": "downsampling process to halve the spatial resolution while doubling the channel dimension like CNN", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 549, + 505, + 561 + ], + "spans": [ + { + "bbox": [ + 106, + 549, + 505, + 561 + ], + "score": 1.0, + "content": "on both regional and local tokens before going to the next stage. Finally, at the end of the network, we", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 560, + 505, + 572 + ], + "spans": [ + { + "bbox": [ + 105, + 560, + 505, + 572 + ], + "score": 1.0, + "content": "average the remaining regional tokens as the final embedding for the classification while the detection", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 570, + 506, + 585 + ], + "spans": [ + { + "bbox": [ + 105, + 570, + 506, + 585 + ], + "score": 1.0, + "content": "uses all local tokens at each stage since it provides more fine-grained location information. By having", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 582, + 505, + 594 + ], + "spans": [ + { + "bbox": [ + 106, + 582, + 505, + 594 + ], + "score": 1.0, + "content": "the pyramid structure, the ViT can generate multi-scale features and hence it could be easily extended", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 593, + 505, + 605 + ], + "spans": [ + { + "bbox": [ + 105, + 593, + 505, + 605 + ], + "score": 1.0, + "content": "to more vision applications, e.g., object detection, rather than image classification only. We explain", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 605, + 448, + 616 + ], + "spans": [ + { + "bbox": [ + 106, + 605, + 448, + 616 + ], + "score": 1.0, + "content": "the main components of the regional-to-local transformer encoder in the next section.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 34 + }, + { + "type": "text", + "bbox": [ + 106, + 620, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 620, + 506, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 620, + 506, + 633 + ], + "score": 1.0, + "content": "Regional-to-Local Transformer Encoder. Figure 3 shows the proposed regional-to-local (R2L)", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 630, + 506, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 506, + 645 + ], + "score": 1.0, + "content": "transformer encoder which includes R2L attention and feed-forward network (FFN). Specifically,", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 642, + 506, + 655 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 506, + 655 + ], + "score": 1.0, + "content": "within R2L attention, the regional self-attention (RSA) first involves all regional tokens to learn the", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "score": 1.0, + "content": "global information efficiently as the number of regional tokens is few and then local self-attention", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 665, + 505, + 676 + ], + "spans": [ + { + "bbox": [ + 106, + 665, + 505, + 676 + ], + "score": 1.0, + "content": "(LSA) takes the local tokens with the associated regional token to learn local feature while the", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 675, + 505, + 687 + ], + "spans": [ + { + "bbox": [ + 105, + 675, + 505, + 687 + ], + "score": 1.0, + "content": "involved regional token could provide global information at the same time. Both RSA and LSA are", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 686, + 505, + 699 + ], + "spans": [ + { + "bbox": [ + 105, + 686, + 505, + 699 + ], + "score": 1.0, + "content": "multihead self-attention (MSA) but with different input tokens. Finally, the FFN is applied to enhance", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 106, + 697, + 506, + 710 + ], + "spans": [ + { + "bbox": [ + 106, + 697, + 506, + 710 + ], + "score": 1.0, + "content": "the features. We also add layer normalization (LN) and residual shortcuts as in standard transformer", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 104, + 705, + 505, + 725 + ], + "spans": [ + { + "bbox": [ + 104, + 705, + 357, + 725 + ], + "score": 1.0, + "content": "encoders. Mathematically, given the regional and local tokens,", + "type": "text" + }, + { + "bbox": [ + 357, + 709, + 379, + 722 + ], + "score": 0.92, + "content": "\\mathbf { x } _ { r } ^ { d - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 380, + 705, + 397, + 725 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 398, + 708, + 420, + 722 + ], + "score": 0.93, + "content": "\\mathbf { x } _ { l } ^ { d - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 420, + 705, + 505, + 725 + ], + "score": 1.0, + "content": "as the inputs at layer", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 107, + 720, + 317, + 733 + ], + "spans": [ + { + "bbox": [ + 107, + 721, + 113, + 731 + ], + "score": 0.68, + "content": "d", + "type": "inline_equation" + }, + { + "bbox": [ + 114, + 720, + 317, + 733 + ], + "score": 1.0, + "content": ", the R2L transformer encoder can be expressed as:", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 46.5 + } + ], + "page_idx": 3, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 294, + 39 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 294, + 39 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 759 + ], + "lines": [] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 108, + 80, + 504, + 138 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 80, + 504, + 138 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 80, + 504, + 138 + ], + "spans": [ + { + "bbox": [ + 108, + 80, + 504, + 138 + ], + "score": 0.954, + "type": "image", + "image_path": "4f51e129978441e6204aba5ef5a836f32d26b9a369736a313e286198ad637a3c.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 108, + 80, + 504, + 99.33333333333333 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 108, + 99.33333333333333, + 504, + 118.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 118.66666666666666, + 504, + 138.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 143, + 505, + 184 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 142, + 505, + 157 + ], + "spans": [ + { + "bbox": [ + 105, + 142, + 505, + 157 + ], + "score": 1.0, + "content": "Figure 2: Architecture of the proposed RegionViT. Two paths are in our proposed network, including", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 154, + 505, + 165 + ], + "spans": [ + { + "bbox": [ + 105, + 154, + 505, + 165 + ], + "score": 1.0, + "content": "regional tokens and local tokens, and their information is exchanged in the regional-to-local (R2L) transformer", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 163, + 505, + 175 + ], + "spans": [ + { + "bbox": [ + 106, + 163, + 505, + 175 + ], + "score": 1.0, + "content": "encoders. In the end, we average all regional tokens and use it for the classification. The tensor shape here is", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 172, + 450, + 185 + ], + "spans": [ + { + "bbox": [ + 105, + 172, + 305, + 185 + ], + "score": 1.0, + "content": "computed based on that regional tokens take a patch of", + "type": "text" + }, + { + "bbox": [ + 306, + 173, + 320, + 183 + ], + "score": 0.91, + "content": "2 8 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 320, + 172, + 436, + 185 + ], + "score": 1.0, + "content": "and local tokens take a patch of", + "type": "text" + }, + { + "bbox": [ + 436, + 173, + 446, + 183 + ], + "score": 0.84, + "content": "4 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 446, + 172, + 450, + 185 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "text", + "bbox": [ + 106, + 197, + 505, + 297 + ], + "lines": [ + { + "bbox": [ + 105, + 196, + 505, + 210 + ], + "spans": [ + { + "bbox": [ + 105, + 196, + 505, + 210 + ], + "score": 1.0, + "content": "Vision Transformer. As opposed to CNN-based approaches for image classification, ViT is a purely", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 209, + 505, + 221 + ], + "spans": [ + { + "bbox": [ + 106, + 209, + 505, + 221 + ], + "score": 1.0, + "content": "attention-based counterpart, borrowed from NLP. It consists of stacked transformer encoders, each", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 219, + 506, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 219, + 506, + 233 + ], + "score": 1.0, + "content": "of which contains a multihead self-attention (MSA) and a feed-forward network (FFN) with layer", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "spans": [ + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "score": 1.0, + "content": "normalization and residual shortcut. To classify an image, ViT first splits it into patches of fixed size,", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 241, + 505, + 254 + ], + "spans": [ + { + "bbox": [ + 105, + 241, + 124, + 254 + ], + "score": 1.0, + "content": "e.g.", + "type": "text" + }, + { + "bbox": [ + 124, + 242, + 153, + 253 + ], + "score": 0.87, + "content": "1 6 \\times 1 6", + "type": "inline_equation" + }, + { + "bbox": [ + 153, + 241, + 505, + 254 + ], + "score": 1.0, + "content": ", and then transforms them into tokens by linear projection. A class token is additionally", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 252, + 506, + 266 + ], + "spans": [ + { + "bbox": [ + 104, + 252, + 506, + 266 + ], + "score": 1.0, + "content": "prepended to the patch tokens to form the input sequence. Before the token are fed into transformer", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 264, + 505, + 276 + ], + "spans": [ + { + "bbox": [ + 106, + 264, + 505, + 276 + ], + "score": 1.0, + "content": "encoders, a learnable absolute positional embedding is added to each token to learn the position", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 274, + 505, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 274, + 505, + 287 + ], + "score": 1.0, + "content": "information. At the end of the network, the class token is used as the final feature representation for", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 285, + 329, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 285, + 329, + 298 + ], + "score": 1.0, + "content": "classification. Mathematically, ViT can be expressed as", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 11, + "bbox_fs": [ + 104, + 196, + 506, + 298 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 189, + 302, + 420, + 332 + ], + "lines": [ + { + "bbox": [ + 189, + 302, + 420, + 332 + ], + "spans": [ + { + "bbox": [ + 189, + 302, + 420, + 332 + ], + "score": 0.92, + "content": "\\begin{array} { r l } & { \\mathbf { x } _ { 0 } = [ \\mathbf { x } _ { c l s } | | \\mathbf { x } _ { p a t c h } ] + \\mathbf { x } _ { p o s } } \\\\ & { \\mathbf { y } _ { k } = \\mathbf { x } _ { k - 1 } + \\mathtt { M S A } \\big ( \\mathtt { L N } ( \\mathbf { x } _ { k - 1 } ) \\big ) , \\mathbf { x } _ { k } = \\mathbf { y } _ { k } + \\mathtt { F F N } \\big ( \\mathtt { L N } \\big ( \\mathbf { y } _ { k } \\big ) \\big ) , } \\end{array}", + "type": "interline_equation", + "image_path": "7dfeeb8d19040ab77e4e3b785ed737ff3cd63c3753977f2581532a9ad733431d.jpg" + } + ] + } + ], + "index": 16.5, + "virtual_lines": [ + { + "bbox": [ + 189, + 302, + 420, + 317.0 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 189, + 317.0, + 420, + 332.0 + ], + "spans": [], + "index": 17 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 336, + 505, + 373 + ], + "lines": [ + { + "bbox": [ + 104, + 334, + 507, + 353 + ], + "spans": [ + { + "bbox": [ + 104, + 334, + 134, + 353 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 134, + 337, + 191, + 349 + ], + "score": 0.92, + "content": "\\mathbf { x } _ { c l s } ~ \\in ~ \\mathbb { R } ^ { 1 \\times C }", + "type": "inline_equation" + }, + { + "bbox": [ + 191, + 334, + 211, + 353 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 211, + 337, + 280, + 351 + ], + "score": 0.92, + "content": "\\mathbf { x } _ { p a t c h } \\in \\mathbb { R } ^ { N \\times C }", + "type": "inline_equation" + }, + { + "bbox": [ + 281, + 334, + 507, + 353 + ], + "score": 1.0, + "content": "are the class token and patch tokens respectively and", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 347, + 504, + 365 + ], + "spans": [ + { + "bbox": [ + 106, + 349, + 181, + 363 + ], + "score": 0.91, + "content": "\\mathbf { x } _ { p o s } \\in \\mathbb { R } ^ { ( 1 + N ) \\times C }", + "type": "inline_equation" + }, + { + "bbox": [ + 181, + 347, + 305, + 365 + ], + "score": 1.0, + "content": "is the position embedding, and", + "type": "text" + }, + { + "bbox": [ + 306, + 350, + 322, + 363 + ], + "score": 0.67, + "content": "\\left[ \\left| \\left| \\right. \\right] \\right.", + "type": "inline_equation" + }, + { + "bbox": [ + 323, + 347, + 455, + 365 + ], + "score": 1.0, + "content": "denotes the tensor concatenation.", + "type": "text" + }, + { + "bbox": [ + 456, + 351, + 477, + 361 + ], + "score": 0.32, + "content": "k , N", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 347, + 495, + 365 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 495, + 351, + 504, + 361 + ], + "score": 0.8, + "content": "C", + "type": "inline_equation" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 361, + 484, + 375 + ], + "spans": [ + { + "bbox": [ + 105, + 361, + 484, + 375 + ], + "score": 1.0, + "content": "are the layer index, the number of patch tokens and dimension of the embedding, respectively.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19, + "bbox_fs": [ + 104, + 334, + 507, + 375 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 378, + 505, + 445 + ], + "lines": [ + { + "bbox": [ + 105, + 378, + 505, + 391 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 505, + 391 + ], + "score": 1.0, + "content": "While the vanilla ViT is ideally capable of learning global interaction among all the patch tokens, the", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "score": 1.0, + "content": "memory complexity of self-attention becomes high when there are many tokens as the complexity", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 401, + 505, + 414 + ], + "spans": [ + { + "bbox": [ + 105, + 401, + 505, + 414 + ], + "score": 1.0, + "content": "is quadratically linear to the length of the input sequence. Moreover, the use of isotropic structure", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 411, + 505, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 505, + 424 + ], + "score": 1.0, + "content": "limits the capability of extending the vanilla ViT model to many vision applications that require", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 422, + 505, + 435 + ], + "spans": [ + { + "bbox": [ + 105, + 422, + 505, + 435 + ], + "score": 1.0, + "content": "high-resolution details, e.g., object detection. To address these issues, we propose regional-to-local", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 434, + 321, + 446 + ], + "spans": [ + { + "bbox": [ + 106, + 434, + 321, + 446 + ], + "score": 1.0, + "content": "attention for vision transformer, RegionViT for short.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 23.5, + "bbox_fs": [ + 105, + 378, + 505, + 446 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 450, + 505, + 615 + ], + "lines": [ + { + "bbox": [ + 105, + 450, + 505, + 463 + ], + "spans": [ + { + "bbox": [ + 105, + 450, + 505, + 463 + ], + "score": 1.0, + "content": "An Overview of Our Proposed Approach (RegionViT). Figure 2 illustrates the architecture of the", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 461, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 473 + ], + "score": 1.0, + "content": "proposed vision transformer, which consists of two tokenization processes that convert an image into", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "regional (upper path) and local tokens (lower path). Each tokenization is a convolution with different", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 483, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 359, + 496 + ], + "score": 1.0, + "content": "patch sizes, e.g., in Figure 2, the patch size of regional tokens is", + "type": "text" + }, + { + "bbox": [ + 359, + 483, + 375, + 493 + ], + "score": 0.88, + "content": "2 8 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 375, + 483, + 400, + 496 + ], + "score": 1.0, + "content": "while", + "type": "text" + }, + { + "bbox": [ + 400, + 483, + 411, + 493 + ], + "score": 0.85, + "content": "4 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 411, + 483, + 505, + 496 + ], + "score": 1.0, + "content": "is used for local tokens", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 104, + 493, + 506, + 506 + ], + "spans": [ + { + "bbox": [ + 104, + 493, + 221, + 506 + ], + "score": 1.0, + "content": "with dimensions projected to", + "type": "text" + }, + { + "bbox": [ + 222, + 495, + 230, + 504 + ], + "score": 0.81, + "content": "C", + "type": "inline_equation" + }, + { + "bbox": [ + 231, + 493, + 407, + 506 + ], + "score": 1.0, + "content": ", which means that one regional token covers", + "type": "text" + }, + { + "bbox": [ + 407, + 493, + 418, + 504 + ], + "score": 0.83, + "content": "7 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 418, + 493, + 506, + 506 + ], + "score": 1.0, + "content": "local tokens based on", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 504, + 506, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 504, + 370, + 518 + ], + "score": 1.0, + "content": "the spatial locality, leading to the window size of a local region to", + "type": "text" + }, + { + "bbox": [ + 371, + 505, + 381, + 515 + ], + "score": 0.77, + "content": "7 ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 381, + 504, + 506, + 518 + ], + "score": 1.0, + "content": ". At stage 1, two sets of tokens", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 515, + 507, + 530 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 507, + 530 + ], + "score": 1.0, + "content": "are passed through the proposed regional-to-local transformer encoders. However, for the later stages,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 527, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 106, + 527, + 506, + 540 + ], + "score": 1.0, + "content": "to balance the computational load and to have feature maps at different resolutions, we deploy a", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 537, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 550 + ], + "score": 1.0, + "content": "downsampling process to halve the spatial resolution while doubling the channel dimension like CNN", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 549, + 505, + 561 + ], + "spans": [ + { + "bbox": [ + 106, + 549, + 505, + 561 + ], + "score": 1.0, + "content": "on both regional and local tokens before going to the next stage. Finally, at the end of the network, we", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 560, + 505, + 572 + ], + "spans": [ + { + "bbox": [ + 105, + 560, + 505, + 572 + ], + "score": 1.0, + "content": "average the remaining regional tokens as the final embedding for the classification while the detection", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 570, + 506, + 585 + ], + "spans": [ + { + "bbox": [ + 105, + 570, + 506, + 585 + ], + "score": 1.0, + "content": "uses all local tokens at each stage since it provides more fine-grained location information. By having", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 582, + 505, + 594 + ], + "spans": [ + { + "bbox": [ + 106, + 582, + 505, + 594 + ], + "score": 1.0, + "content": "the pyramid structure, the ViT can generate multi-scale features and hence it could be easily extended", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 593, + 505, + 605 + ], + "spans": [ + { + "bbox": [ + 105, + 593, + 505, + 605 + ], + "score": 1.0, + "content": "to more vision applications, e.g., object detection, rather than image classification only. We explain", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 605, + 448, + 616 + ], + "spans": [ + { + "bbox": [ + 106, + 605, + 448, + 616 + ], + "score": 1.0, + "content": "the main components of the regional-to-local transformer encoder in the next section.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 34, + "bbox_fs": [ + 104, + 450, + 507, + 616 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 620, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 620, + 506, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 620, + 506, + 633 + ], + "score": 1.0, + "content": "Regional-to-Local Transformer Encoder. Figure 3 shows the proposed regional-to-local (R2L)", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 630, + 506, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 506, + 645 + ], + "score": 1.0, + "content": "transformer encoder which includes R2L attention and feed-forward network (FFN). Specifically,", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 642, + 506, + 655 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 506, + 655 + ], + "score": 1.0, + "content": "within R2L attention, the regional self-attention (RSA) first involves all regional tokens to learn the", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "score": 1.0, + "content": "global information efficiently as the number of regional tokens is few and then local self-attention", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 665, + 505, + 676 + ], + "spans": [ + { + "bbox": [ + 106, + 665, + 505, + 676 + ], + "score": 1.0, + "content": "(LSA) takes the local tokens with the associated regional token to learn local feature while the", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 675, + 505, + 687 + ], + "spans": [ + { + "bbox": [ + 105, + 675, + 505, + 687 + ], + "score": 1.0, + "content": "involved regional token could provide global information at the same time. Both RSA and LSA are", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 686, + 505, + 699 + ], + "spans": [ + { + "bbox": [ + 105, + 686, + 505, + 699 + ], + "score": 1.0, + "content": "multihead self-attention (MSA) but with different input tokens. Finally, the FFN is applied to enhance", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 106, + 697, + 506, + 710 + ], + "spans": [ + { + "bbox": [ + 106, + 697, + 506, + 710 + ], + "score": 1.0, + "content": "the features. We also add layer normalization (LN) and residual shortcuts as in standard transformer", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 104, + 705, + 505, + 725 + ], + "spans": [ + { + "bbox": [ + 104, + 705, + 357, + 725 + ], + "score": 1.0, + "content": "encoders. Mathematically, given the regional and local tokens,", + "type": "text" + }, + { + "bbox": [ + 357, + 709, + 379, + 722 + ], + "score": 0.92, + "content": "\\mathbf { x } _ { r } ^ { d - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 380, + 705, + 397, + 725 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 398, + 708, + 420, + 722 + ], + "score": 0.93, + "content": "\\mathbf { x } _ { l } ^ { d - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 420, + 705, + 505, + 725 + ], + "score": 1.0, + "content": "as the inputs at layer", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 107, + 720, + 317, + 733 + ], + "spans": [ + { + "bbox": [ + 107, + 721, + 113, + 731 + ], + "score": 0.68, + "content": "d", + "type": "inline_equation" + }, + { + "bbox": [ + 114, + 720, + 317, + 733 + ], + "score": 1.0, + "content": ", the R2L transformer encoder can be expressed as:", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 46.5, + "bbox_fs": [ + 104, + 620, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "interline_equation", + "bbox": [ + 172, + 93, + 441, + 130 + ], + "lines": [ + { + "bbox": [ + 173, + 93, + 441, + 130 + ], + "spans": [ + { + "bbox": [ + 173, + 93, + 441, + 130 + ], + "score": 0.4, + "content": "\\begin{array} { r l } & { \\mathbf { y } _ { r } ^ { d } = \\mathbf { x } _ { r } ^ { d - 1 } + \\mathtt { R S A } \\big ( \\mathrm { L N } ( \\mathbf { x } _ { r } ^ { d - 1 } ) \\big ) , \\mathbf { y } _ { i , j } ^ { d } = [ \\mathbf { y } _ { r _ { i , j } } ^ { d } | | \\{ \\mathbf { x } _ { l _ { i , j , m , n } } ^ { d - 1 } \\} _ { m , n \\in M } ] } \\\\ & { \\mathbf { z } _ { i , j } ^ { d } = \\mathbf { y } _ { i , j } ^ { d } + \\mathtt { L S A } \\big ( \\mathrm { L N } ( \\mathbf { y } _ { i , j } ^ { d } ) \\big ) , \\mathbf { x } _ { i , j } ^ { d } = \\mathbf { z } _ { i , j } ^ { d } + \\mathtt { F F N } \\big ( \\mathrm { L N } ( \\mathbf { z } _ { i , j } ^ { d } ) \\big ) } \\end{array}", + "type": "interline_equation", + "image_path": "c34b464a9baa0f95707a330ecf07d5e433d3e0b3d013acb853b7ec6524b2c591.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 172, + 93, + 441, + 105.33333333333333 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 172, + 105.33333333333333, + 441, + 117.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 172, + 117.66666666666666, + 441, + 130.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 131, + 506, + 177 + ], + "lines": [ + { + "bbox": [ + 106, + 132, + 505, + 144 + ], + "spans": [ + { + "bbox": [ + 106, + 132, + 132, + 144 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 132, + 132, + 146, + 143 + ], + "score": 0.72, + "content": "i , j", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 132, + 368, + 144 + ], + "score": 1.0, + "content": "are the spatial index with respect to regional tokens while", + "type": "text" + }, + { + "bbox": [ + 368, + 133, + 389, + 142 + ], + "score": 0.86, + "content": "m , n", + "type": "inline_equation" + }, + { + "bbox": [ + 389, + 132, + 505, + 144 + ], + "score": 1.0, + "content": "are the index of local token in", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 104, + 141, + 506, + 157 + ], + "spans": [ + { + "bbox": [ + 104, + 141, + 172, + 157 + ], + "score": 1.0, + "content": "the window size", + "type": "text" + }, + { + "bbox": [ + 172, + 142, + 188, + 153 + ], + "score": 0.89, + "content": "\\bar { M } ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 188, + 141, + 300, + 157 + ], + "score": 1.0, + "content": ". Note that the input of LSA,", + "type": "text" + }, + { + "bbox": [ + 300, + 142, + 317, + 156 + ], + "score": 0.91, + "content": "\\mathbf { y } _ { i , j } ^ { d }", + "type": "inline_equation" + }, + { + "bbox": [ + 317, + 141, + 506, + 157 + ], + "score": 1.0, + "content": ", includes one regional token and corresponding", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 154, + 506, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 154, + 506, + 165 + ], + "score": 1.0, + "content": "local tokens, and thus, the information between local and regional tokens are exchanged; on the other", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 164, + 442, + 178 + ], + "spans": [ + { + "bbox": [ + 105, + 164, + 179, + 178 + ], + "score": 1.0, + "content": "hand, the outputs,", + "type": "text" + }, + { + "bbox": [ + 180, + 164, + 192, + 177 + ], + "score": 0.9, + "content": "\\mathbf { x } _ { r } ^ { d }", + "type": "inline_equation" + }, + { + "bbox": [ + 192, + 164, + 209, + 178 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 210, + 164, + 222, + 177 + ], + "score": 0.9, + "content": "\\mathbf { x } _ { l } ^ { d }", + "type": "inline_equation" + }, + { + "bbox": [ + 222, + 164, + 314, + 178 + ], + "score": 1.0, + "content": ", can be extracted from", + "type": "text" + }, + { + "bbox": [ + 315, + 164, + 331, + 178 + ], + "score": 0.92, + "content": "\\mathbf { x } _ { i , j } ^ { d }", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 164, + 442, + 178 + ], + "score": 1.0, + "content": "like the top-right equation.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + }, + { + "type": "text", + "bbox": [ + 106, + 182, + 258, + 367 + ], + "lines": [ + { + "bbox": [ + 106, + 181, + 257, + 193 + ], + "spans": [ + { + "bbox": [ + 106, + 181, + 257, + 193 + ], + "score": 1.0, + "content": "The RSA exchanges information", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 192, + 257, + 204 + ], + "spans": [ + { + "bbox": [ + 105, + 192, + 257, + 204 + ], + "score": 1.0, + "content": "among all tokens, which covers the", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 203, + 257, + 215 + ], + "spans": [ + { + "bbox": [ + 106, + 203, + 257, + 215 + ], + "score": 1.0, + "content": "context of the whole image; while the", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 213, + 257, + 227 + ], + "spans": [ + { + "bbox": [ + 105, + 213, + 257, + 227 + ], + "score": 1.0, + "content": "LSA combines the features among the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 225, + 258, + 238 + ], + "spans": [ + { + "bbox": [ + 106, + 225, + 258, + 238 + ], + "score": 1.0, + "content": "tokens belonging to the spatial region,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 236, + 258, + 248 + ], + "spans": [ + { + "bbox": [ + 106, + 236, + 258, + 248 + ], + "score": 1.0, + "content": "including both regional and local to-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 247, + 258, + 258 + ], + "spans": [ + { + "bbox": [ + 106, + 247, + 258, + 258 + ], + "score": 1.0, + "content": "kens. Since the regions are divided", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 258, + 258, + 270 + ], + "spans": [ + { + "bbox": [ + 106, + 258, + 258, + 270 + ], + "score": 1.0, + "content": "by non-overlapped windows, the RSA", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 269, + 259, + 280 + ], + "spans": [ + { + "bbox": [ + 105, + 269, + 259, + 280 + ], + "score": 1.0, + "content": "is also designed to exchange the in-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 279, + 258, + 293 + ], + "spans": [ + { + "bbox": [ + 105, + 279, + 258, + 293 + ], + "score": 1.0, + "content": "formation among regions where the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 291, + 258, + 303 + ], + "spans": [ + { + "bbox": [ + 106, + 291, + 258, + 303 + ], + "score": 1.0, + "content": "LSA takes one regional token and then", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 303, + 257, + 313 + ], + "spans": [ + { + "bbox": [ + 106, + 303, + 257, + 313 + ], + "score": 1.0, + "content": "combines with it the local tokens in", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 313, + 258, + 324 + ], + "spans": [ + { + "bbox": [ + 106, + 313, + 258, + 324 + ], + "score": 1.0, + "content": "the same region. In such a case, all lo-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 324, + 258, + 336 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 258, + 336 + ], + "score": 1.0, + "content": "cal tokens are still capable of getting", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 335, + 258, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 335, + 258, + 347 + ], + "score": 1.0, + "content": "global information while being more", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 346, + 257, + 357 + ], + "spans": [ + { + "bbox": [ + 106, + 346, + 257, + 357 + ], + "score": 1.0, + "content": "focused on local neighbors. It is worth", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 357, + 258, + 369 + ], + "spans": [ + { + "bbox": [ + 106, + 357, + 258, + 369 + ], + "score": 1.0, + "content": "noting that the weights are shared be-", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 15 + }, + { + "type": "image", + "bbox": [ + 266, + 182, + 503, + 275 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 266, + 182, + 503, + 275 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 266, + 182, + 503, + 275 + ], + "spans": [ + { + "bbox": [ + 266, + 182, + 503, + 275 + ], + "score": 0.971, + "type": "image", + "image_path": "b24588ed075ddb0fab57df5144e35582830e656d60d057cf0a2c111269b41fbc.jpg" + } + ] + } + ], + "index": 27, + "virtual_lines": [ + { + "bbox": [ + 266, + 182, + 503, + 195.28571428571428 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 266, + 195.28571428571428, + 503, + 208.57142857142856 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 266, + 208.57142857142856, + 503, + 221.85714285714283 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 266, + 221.85714285714283, + 503, + 235.1428571428571 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 266, + 235.1428571428571, + 503, + 248.4285714285714 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 266, + 248.4285714285714, + 503, + 261.71428571428567 + ], + "spans": [], + "index": 29 + }, + { + "bbox": [ + 266, + 261.71428571428567, + 503, + 274.99999999999994 + ], + "spans": [], + "index": 30 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 265, + 279, + 505, + 359 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 264, + 278, + 505, + 290 + ], + "spans": [ + { + "bbox": [ + 264, + 278, + 505, + 290 + ], + "score": 1.0, + "content": "Figure 3: Illustration of Regional-to-Local (R2L) Transformer", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 264, + 289, + 505, + 300 + ], + "spans": [ + { + "bbox": [ + 264, + 289, + 505, + 300 + ], + "score": 1.0, + "content": "Encoder. All regional tokens are first passed through regional", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 264, + 298, + 505, + 310 + ], + "spans": [ + { + "bbox": [ + 264, + 298, + 505, + 310 + ], + "score": 1.0, + "content": "self-attention (RSA) to exchange the information among regions", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 264, + 309, + 505, + 320 + ], + "spans": [ + { + "bbox": [ + 264, + 309, + 505, + 320 + ], + "score": 1.0, + "content": "and then local self-attention (LSA) performs parallel self-attention", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 264, + 318, + 506, + 330 + ], + "spans": [ + { + "bbox": [ + 264, + 318, + 506, + 330 + ], + "score": 1.0, + "content": "where each takes one regional token and corresponding local to-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 264, + 329, + 505, + 339 + ], + "spans": [ + { + "bbox": [ + 264, + 329, + 505, + 339 + ], + "score": 1.0, + "content": "kens. After that, all the tokens are passed through the feed-forward", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 264, + 339, + 505, + 350 + ], + "spans": [ + { + "bbox": [ + 264, + 339, + 505, + 350 + ], + "score": 1.0, + "content": "network and split back to the regional and local tokens. RSA and", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 264, + 348, + 371, + 360 + ], + "spans": [ + { + "bbox": [ + 264, + 348, + 371, + 360 + ], + "score": 1.0, + "content": "LSA share the same weights.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 34.5 + } + ], + "index": 30.75 + }, + { + "type": "text", + "bbox": [ + 107, + 368, + 506, + 434 + ], + "lines": [ + { + "bbox": [ + 106, + 368, + 506, + 379 + ], + "spans": [ + { + "bbox": [ + 106, + 368, + 506, + 379 + ], + "score": 1.0, + "content": "tween RSA and LSA except for the layer normalization; therefore, the number of parameters won’t", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 379, + 507, + 391 + ], + "spans": [ + { + "bbox": [ + 105, + 379, + 507, + 391 + ], + "score": 1.0, + "content": "increase significantly when compared to the standard transformer encoder. With these two attentions,", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 389, + 506, + 402 + ], + "spans": [ + { + "bbox": [ + 106, + 389, + 506, + 402 + ], + "score": 1.0, + "content": "the R2L can effectively and efficiently exchange information among all regional and local tokens. In", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 401, + 506, + 414 + ], + "spans": [ + { + "bbox": [ + 105, + 401, + 506, + 414 + ], + "score": 1.0, + "content": "particular, the self-attention on regional tokens aims to extract high-level information and act as a", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 411, + 506, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 506, + 424 + ], + "score": 1.0, + "content": "bridge to pass information of local tokens from one region to other regions. On the other hand, the", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 422, + 414, + 434 + ], + "spans": [ + { + "bbox": [ + 105, + 422, + 414, + 434 + ], + "score": 1.0, + "content": "R2L attention focus on local contextual information with one regional token.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 41.5 + }, + { + "type": "text", + "bbox": [ + 106, + 439, + 506, + 517 + ], + "lines": [ + { + "bbox": [ + 105, + 438, + 507, + 453 + ], + "spans": [ + { + "bbox": [ + 105, + 438, + 507, + 453 + ], + "score": 1.0, + "content": "Moreover, the R2L transformer encoder can further reduce the memory complexity significantly.", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 104, + 450, + 507, + 463 + ], + "spans": [ + { + "bbox": [ + 104, + 450, + 301, + 463 + ], + "score": 1.0, + "content": "The memory complexity of a self-attention is", + "type": "text" + }, + { + "bbox": [ + 301, + 450, + 331, + 462 + ], + "score": 0.93, + "content": "\\mathcal { O } ( N ^ { 2 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 450, + 365, + 463 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 365, + 451, + 376, + 460 + ], + "score": 0.79, + "content": "N", + "type": "inline_equation" + }, + { + "bbox": [ + 376, + 450, + 507, + 463 + ], + "score": 1.0, + "content": "is the number of local tokens.", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 461, + 506, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 207, + 474 + ], + "score": 1.0, + "content": "When a region contains", + "type": "text" + }, + { + "bbox": [ + 207, + 462, + 219, + 471 + ], + "score": 0.74, + "content": "M", + "type": "inline_equation" + }, + { + "bbox": [ + 219, + 461, + 332, + 474 + ], + "score": 1.0, + "content": "local tokens, and there are", + "type": "text" + }, + { + "bbox": [ + 333, + 461, + 357, + 473 + ], + "score": 0.89, + "content": "N / M", + "type": "inline_equation" + }, + { + "bbox": [ + 357, + 461, + 506, + 474 + ], + "score": 1.0, + "content": "regions, the memory complexity is", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 107, + 472, + 506, + 486 + ], + "spans": [ + { + "bbox": [ + 107, + 472, + 206, + 484 + ], + "score": 0.92, + "content": "\\mathcal { O } ( ( M + 1 \\bar { ) } ^ { 2 } \\times ( N / M ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 206, + 472, + 450, + 486 + ], + "score": 1.0, + "content": "; while the complexity from the attention on regional tokens is", + "type": "text" + }, + { + "bbox": [ + 451, + 472, + 502, + 484 + ], + "score": 0.93, + "content": "\\mathcal { O } ( ( N / M ) ^ { 2 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 472, + 506, + 486 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 482, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 261, + 496 + ], + "score": 1.0, + "content": "Hence, the overall complexity becomes", + "type": "text" + }, + { + "bbox": [ + 261, + 483, + 398, + 495 + ], + "score": 0.9, + "content": "\\mathcal { \\dot { O } } ( ( M { \\dot { } } 1 ) ^ { 2 } \\times ( N / M ) + ( N / M ) ^ { 2 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 399, + 482, + 506, + 496 + ], + "score": 1.0, + "content": ". E.g., the main component", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 493, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 234, + 508 + ], + "score": 1.0, + "content": "of RegionViT is stage 3, where", + "type": "text" + }, + { + "bbox": [ + 234, + 495, + 246, + 504 + ], + "score": 0.77, + "content": "M", + "type": "inline_equation" + }, + { + "bbox": [ + 246, + 493, + 286, + 508 + ], + "score": 1.0, + "content": "is 49 and", + "type": "text" + }, + { + "bbox": [ + 286, + 495, + 296, + 504 + ], + "score": 0.81, + "content": "N", + "type": "inline_equation" + }, + { + "bbox": [ + 297, + 493, + 505, + 508 + ], + "score": 1.0, + "content": "is 196; therefore, the memory complexity saving is", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 504, + 480, + 518 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 136, + 516 + ], + "score": 0.87, + "content": "\\sim 7 3 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 504, + 480, + 518 + ], + "score": 1.0, + "content": ". The saving of each model and each stage is varied based on the model configuration.", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 48 + }, + { + "type": "text", + "bbox": [ + 106, + 522, + 505, + 600 + ], + "lines": [ + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "score": 1.0, + "content": "Relative Position Bias. The locality is an important clue for understanding visual content; therefore,", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "score": 1.0, + "content": "instead of adding absolute position embedding used by the vanilla ViT, we introduce the relative", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 105, + 543, + 506, + 557 + ], + "spans": [ + { + "bbox": [ + 105, + 543, + 506, + 557 + ], + "score": 1.0, + "content": "position bias into the attention map of R2L attention since relative position between patches (or", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 105, + 555, + 506, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 506, + 567 + ], + "score": 1.0, + "content": "pixels) are more important than the absolution position as objects in the images can be placed in an", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 105, + 565, + 506, + 578 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 506, + 578 + ], + "score": 1.0, + "content": "arbitrary way (Ramachandran et al., 2019a; Liu et al., 2021). We only add this bias to the attention", + "type": "text" + } + ], + "index": 56 + }, + { + "bbox": [ + 105, + 576, + 506, + 590 + ], + "spans": [ + { + "bbox": [ + 105, + 576, + 506, + 590 + ], + "score": 1.0, + "content": "between local tokens and not the attention between regional tokens and local tokens. Specifically, for", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 104, + 587, + 506, + 601 + ], + "spans": [ + { + "bbox": [ + 104, + 587, + 262, + 601 + ], + "score": 1.0, + "content": "a given pair of local tokens at location,", + "type": "text" + }, + { + "bbox": [ + 262, + 588, + 339, + 600 + ], + "score": 0.91, + "content": "( x _ { m } , y _ { m } ) , ( x _ { n } , y _ { n } )", + "type": "inline_equation" + }, + { + "bbox": [ + 339, + 587, + 417, + 601 + ], + "score": 1.0, + "content": ", the attention value", + "type": "text" + }, + { + "bbox": [ + 417, + 590, + 424, + 598 + ], + "score": 0.76, + "content": "a", + "type": "inline_equation" + }, + { + "bbox": [ + 425, + 587, + 506, + 601 + ], + "score": 1.0, + "content": "can be expressed as", + "type": "text" + } + ], + "index": 58 + } + ], + "index": 55 + }, + { + "type": "interline_equation", + "bbox": [ + 164, + 603, + 445, + 624 + ], + "lines": [ + { + "bbox": [ + 164, + 603, + 445, + 624 + ], + "spans": [ + { + "bbox": [ + 164, + 603, + 445, + 624 + ], + "score": 0.91, + "content": "a _ { ( x _ { m } , y _ { m } ) , ( x _ { n } , y _ { n } ) } = \\mathtt { s o f t m a x } \\left( q _ { ( x _ { m } , y _ { m } ) } k _ { ( x _ { n } , y _ { n } ) } ^ { T } + b _ { ( x _ { m } - x _ { n } , y _ { m } - y _ { n } ) } \\right) ,", + "type": "interline_equation", + "image_path": "6a7aacb1a6a88ae3d8efb902dfd1effb581d8a35b915c4b4bcec7d7ae7b3c1ab.jpg" + } + ] + } + ], + "index": 59, + "virtual_lines": [ + { + "bbox": [ + 164, + 603, + 445, + 624 + ], + "spans": [], + "index": 59 + } + ] + }, + { + "type": "text", + "bbox": [ + 109, + 628, + 504, + 655 + ], + "lines": [ + { + "bbox": [ + 105, + 624, + 507, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 134, + 643 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 201, + 624, + 352, + 643 + ], + "score": 1.0, + "content": "is taken from a learnable parameter", + "type": "text" + }, + { + "bbox": [ + 352, + 627, + 434, + 639 + ], + "score": 0.91, + "content": "B \\in \\mathbb { R } ^ { 2 M - 1 \\times 2 M - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 624, + 466, + 643 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 467, + 628, + 479, + 638 + ], + "score": 0.84, + "content": "M", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 624, + 507, + 643 + ], + "score": 1.0, + "content": "is the", + "type": "text" + } + ], + "index": 60 + }, + { + "bbox": [ + 103, + 633, + 510, + 661 + ], + "spans": [ + { + "bbox": [ + 103, + 633, + 261, + 661 + ], + "score": 1.0, + "content": "m n m n window size of a region. The first term", + "type": "text" + }, + { + "bbox": [ + 262, + 640, + 330, + 655 + ], + "score": 0.94, + "content": "q _ { ( x _ { m } , y _ { m } ) } k _ { ( x _ { n } , y _ { n } ) } ^ { T }", + "type": "inline_equation" + }, + { + "bbox": [ + 330, + 633, + 510, + 661 + ], + "score": 1.0, + "content": "is the attention value based on their content.", + "type": "text" + } + ], + "index": 61 + } + ], + "index": 60.5 + }, + { + "type": "text", + "bbox": [ + 107, + 659, + 503, + 682 + ], + "lines": [ + { + "bbox": [ + 106, + 659, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 659, + 505, + 672 + ], + "score": 1.0, + "content": "With relative position bias, our model does not require the absolute positional embedding as vanilla", + "type": "text" + } + ], + "index": 62 + }, + { + "bbox": [ + 106, + 671, + 414, + 683 + ], + "spans": [ + { + "bbox": [ + 106, + 671, + 414, + 683 + ], + "score": 1.0, + "content": "ViT since this relative position bias helps to encode the position information.", + "type": "text" + } + ], + "index": 63 + } + ], + "index": 62.5 + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "score": 1.0, + "content": "Downsampling. The spatial and channel dimension are changed along different stages of our network", + "type": "text" + } + ], + "index": 64 + }, + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 183, + 712 + ], + "score": 1.0, + "content": "by simply applying", + "type": "text" + }, + { + "bbox": [ + 184, + 699, + 206, + 709 + ], + "score": 0.88, + "content": "3 \\times 3", + "type": "inline_equation" + }, + { + "bbox": [ + 206, + 699, + 506, + 712 + ], + "score": 1.0, + "content": "depth-wise convolution with stride 2 for halving the resolution and doubling", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "the channel dimension for both regional and local tokens (weights are shared) (Heo et al., 2021). We", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 105, + 720, + 502, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 502, + 734 + ], + "score": 1.0, + "content": "also test with regular convolutions but it does not improve the accuracy with increased complexity.", + "type": "text" + } + ], + "index": 67 + } + ], + "index": 65.5 + } + ], + "page_idx": 4, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 26, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "score": 1.0, + "content": "5", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "interline_equation", + "bbox": [ + 172, + 93, + 441, + 130 + ], + "lines": [ + { + "bbox": [ + 173, + 93, + 441, + 130 + ], + "spans": [ + { + "bbox": [ + 173, + 93, + 441, + 130 + ], + "score": 0.4, + "content": "\\begin{array} { r l } & { \\mathbf { y } _ { r } ^ { d } = \\mathbf { x } _ { r } ^ { d - 1 } + \\mathtt { R S A } \\big ( \\mathrm { L N } ( \\mathbf { x } _ { r } ^ { d - 1 } ) \\big ) , \\mathbf { y } _ { i , j } ^ { d } = [ \\mathbf { y } _ { r _ { i , j } } ^ { d } | | \\{ \\mathbf { x } _ { l _ { i , j , m , n } } ^ { d - 1 } \\} _ { m , n \\in M } ] } \\\\ & { \\mathbf { z } _ { i , j } ^ { d } = \\mathbf { y } _ { i , j } ^ { d } + \\mathtt { L S A } \\big ( \\mathrm { L N } ( \\mathbf { y } _ { i , j } ^ { d } ) \\big ) , \\mathbf { x } _ { i , j } ^ { d } = \\mathbf { z } _ { i , j } ^ { d } + \\mathtt { F F N } \\big ( \\mathrm { L N } ( \\mathbf { z } _ { i , j } ^ { d } ) \\big ) } \\end{array}", + "type": "interline_equation", + "image_path": "c34b464a9baa0f95707a330ecf07d5e433d3e0b3d013acb853b7ec6524b2c591.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 172, + 93, + 441, + 105.33333333333333 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 172, + 105.33333333333333, + 441, + 117.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 172, + 117.66666666666666, + 441, + 130.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 131, + 506, + 177 + ], + "lines": [ + { + "bbox": [ + 106, + 132, + 505, + 144 + ], + "spans": [ + { + "bbox": [ + 106, + 132, + 132, + 144 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 132, + 132, + 146, + 143 + ], + "score": 0.72, + "content": "i , j", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 132, + 368, + 144 + ], + "score": 1.0, + "content": "are the spatial index with respect to regional tokens while", + "type": "text" + }, + { + "bbox": [ + 368, + 133, + 389, + 142 + ], + "score": 0.86, + "content": "m , n", + "type": "inline_equation" + }, + { + "bbox": [ + 389, + 132, + 505, + 144 + ], + "score": 1.0, + "content": "are the index of local token in", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 104, + 141, + 506, + 157 + ], + "spans": [ + { + "bbox": [ + 104, + 141, + 172, + 157 + ], + "score": 1.0, + "content": "the window size", + "type": "text" + }, + { + "bbox": [ + 172, + 142, + 188, + 153 + ], + "score": 0.89, + "content": "\\bar { M } ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 188, + 141, + 300, + 157 + ], + "score": 1.0, + "content": ". Note that the input of LSA,", + "type": "text" + }, + { + "bbox": [ + 300, + 142, + 317, + 156 + ], + "score": 0.91, + "content": "\\mathbf { y } _ { i , j } ^ { d }", + "type": "inline_equation" + }, + { + "bbox": [ + 317, + 141, + 506, + 157 + ], + "score": 1.0, + "content": ", includes one regional token and corresponding", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 154, + 506, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 154, + 506, + 165 + ], + "score": 1.0, + "content": "local tokens, and thus, the information between local and regional tokens are exchanged; on the other", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 164, + 442, + 178 + ], + "spans": [ + { + "bbox": [ + 105, + 164, + 179, + 178 + ], + "score": 1.0, + "content": "hand, the outputs,", + "type": "text" + }, + { + "bbox": [ + 180, + 164, + 192, + 177 + ], + "score": 0.9, + "content": "\\mathbf { x } _ { r } ^ { d }", + "type": "inline_equation" + }, + { + "bbox": [ + 192, + 164, + 209, + 178 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 210, + 164, + 222, + 177 + ], + "score": 0.9, + "content": "\\mathbf { x } _ { l } ^ { d }", + "type": "inline_equation" + }, + { + "bbox": [ + 222, + 164, + 314, + 178 + ], + "score": 1.0, + "content": ", can be extracted from", + "type": "text" + }, + { + "bbox": [ + 315, + 164, + 331, + 178 + ], + "score": 0.92, + "content": "\\mathbf { x } _ { i , j } ^ { d }", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 164, + 442, + 178 + ], + "score": 1.0, + "content": "like the top-right equation.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5, + "bbox_fs": [ + 104, + 132, + 506, + 178 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 182, + 258, + 367 + ], + "lines": [ + { + "bbox": [ + 106, + 181, + 257, + 193 + ], + "spans": [ + { + "bbox": [ + 106, + 181, + 257, + 193 + ], + "score": 1.0, + "content": "The RSA exchanges information", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 192, + 257, + 204 + ], + "spans": [ + { + "bbox": [ + 105, + 192, + 257, + 204 + ], + "score": 1.0, + "content": "among all tokens, which covers the", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 203, + 257, + 215 + ], + "spans": [ + { + "bbox": [ + 106, + 203, + 257, + 215 + ], + "score": 1.0, + "content": "context of the whole image; while the", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 213, + 257, + 227 + ], + "spans": [ + { + "bbox": [ + 105, + 213, + 257, + 227 + ], + "score": 1.0, + "content": "LSA combines the features among the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 225, + 258, + 238 + ], + "spans": [ + { + "bbox": [ + 106, + 225, + 258, + 238 + ], + "score": 1.0, + "content": "tokens belonging to the spatial region,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 236, + 258, + 248 + ], + "spans": [ + { + "bbox": [ + 106, + 236, + 258, + 248 + ], + "score": 1.0, + "content": "including both regional and local to-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 247, + 258, + 258 + ], + "spans": [ + { + "bbox": [ + 106, + 247, + 258, + 258 + ], + "score": 1.0, + "content": "kens. Since the regions are divided", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 258, + 258, + 270 + ], + "spans": [ + { + "bbox": [ + 106, + 258, + 258, + 270 + ], + "score": 1.0, + "content": "by non-overlapped windows, the RSA", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 269, + 259, + 280 + ], + "spans": [ + { + "bbox": [ + 105, + 269, + 259, + 280 + ], + "score": 1.0, + "content": "is also designed to exchange the in-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 279, + 258, + 293 + ], + "spans": [ + { + "bbox": [ + 105, + 279, + 258, + 293 + ], + "score": 1.0, + "content": "formation among regions where the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 291, + 258, + 303 + ], + "spans": [ + { + "bbox": [ + 106, + 291, + 258, + 303 + ], + "score": 1.0, + "content": "LSA takes one regional token and then", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 303, + 257, + 313 + ], + "spans": [ + { + "bbox": [ + 106, + 303, + 257, + 313 + ], + "score": 1.0, + "content": "combines with it the local tokens in", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 313, + 258, + 324 + ], + "spans": [ + { + "bbox": [ + 106, + 313, + 258, + 324 + ], + "score": 1.0, + "content": "the same region. In such a case, all lo-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 324, + 258, + 336 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 258, + 336 + ], + "score": 1.0, + "content": "cal tokens are still capable of getting", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 335, + 258, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 335, + 258, + 347 + ], + "score": 1.0, + "content": "global information while being more", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 346, + 257, + 357 + ], + "spans": [ + { + "bbox": [ + 106, + 346, + 257, + 357 + ], + "score": 1.0, + "content": "focused on local neighbors. It is worth", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 357, + 258, + 369 + ], + "spans": [ + { + "bbox": [ + 106, + 357, + 258, + 369 + ], + "score": 1.0, + "content": "noting that the weights are shared be-", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 15, + "bbox_fs": [ + 105, + 181, + 259, + 369 + ] + }, + { + "type": "image", + "bbox": [ + 266, + 182, + 503, + 275 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 266, + 182, + 503, + 275 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 266, + 182, + 503, + 275 + ], + "spans": [ + { + "bbox": [ + 266, + 182, + 503, + 275 + ], + "score": 0.971, + "type": "image", + "image_path": "b24588ed075ddb0fab57df5144e35582830e656d60d057cf0a2c111269b41fbc.jpg" + } + ] + } + ], + "index": 27, + "virtual_lines": [ + { + "bbox": [ + 266, + 182, + 503, + 195.28571428571428 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 266, + 195.28571428571428, + 503, + 208.57142857142856 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 266, + 208.57142857142856, + 503, + 221.85714285714283 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 266, + 221.85714285714283, + 503, + 235.1428571428571 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 266, + 235.1428571428571, + 503, + 248.4285714285714 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 266, + 248.4285714285714, + 503, + 261.71428571428567 + ], + "spans": [], + "index": 29 + }, + { + "bbox": [ + 266, + 261.71428571428567, + 503, + 274.99999999999994 + ], + "spans": [], + "index": 30 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 265, + 279, + 505, + 359 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 264, + 278, + 505, + 290 + ], + "spans": [ + { + "bbox": [ + 264, + 278, + 505, + 290 + ], + "score": 1.0, + "content": "Figure 3: Illustration of Regional-to-Local (R2L) Transformer", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 264, + 289, + 505, + 300 + ], + "spans": [ + { + "bbox": [ + 264, + 289, + 505, + 300 + ], + "score": 1.0, + "content": "Encoder. All regional tokens are first passed through regional", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 264, + 298, + 505, + 310 + ], + "spans": [ + { + "bbox": [ + 264, + 298, + 505, + 310 + ], + "score": 1.0, + "content": "self-attention (RSA) to exchange the information among regions", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 264, + 309, + 505, + 320 + ], + "spans": [ + { + "bbox": [ + 264, + 309, + 505, + 320 + ], + "score": 1.0, + "content": "and then local self-attention (LSA) performs parallel self-attention", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 264, + 318, + 506, + 330 + ], + "spans": [ + { + "bbox": [ + 264, + 318, + 506, + 330 + ], + "score": 1.0, + "content": "where each takes one regional token and corresponding local to-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 264, + 329, + 505, + 339 + ], + "spans": [ + { + "bbox": [ + 264, + 329, + 505, + 339 + ], + "score": 1.0, + "content": "kens. After that, all the tokens are passed through the feed-forward", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 264, + 339, + 505, + 350 + ], + "spans": [ + { + "bbox": [ + 264, + 339, + 505, + 350 + ], + "score": 1.0, + "content": "network and split back to the regional and local tokens. RSA and", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 264, + 348, + 371, + 360 + ], + "spans": [ + { + "bbox": [ + 264, + 348, + 371, + 360 + ], + "score": 1.0, + "content": "LSA share the same weights.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 34.5 + } + ], + "index": 30.75 + }, + { + "type": "text", + "bbox": [ + 107, + 368, + 506, + 434 + ], + "lines": [ + { + "bbox": [ + 106, + 368, + 506, + 379 + ], + "spans": [ + { + "bbox": [ + 106, + 368, + 506, + 379 + ], + "score": 1.0, + "content": "tween RSA and LSA except for the layer normalization; therefore, the number of parameters won’t", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 379, + 507, + 391 + ], + "spans": [ + { + "bbox": [ + 105, + 379, + 507, + 391 + ], + "score": 1.0, + "content": "increase significantly when compared to the standard transformer encoder. With these two attentions,", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 389, + 506, + 402 + ], + "spans": [ + { + "bbox": [ + 106, + 389, + 506, + 402 + ], + "score": 1.0, + "content": "the R2L can effectively and efficiently exchange information among all regional and local tokens. In", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 401, + 506, + 414 + ], + "spans": [ + { + "bbox": [ + 105, + 401, + 506, + 414 + ], + "score": 1.0, + "content": "particular, the self-attention on regional tokens aims to extract high-level information and act as a", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 411, + 506, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 506, + 424 + ], + "score": 1.0, + "content": "bridge to pass information of local tokens from one region to other regions. On the other hand, the", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 422, + 414, + 434 + ], + "spans": [ + { + "bbox": [ + 105, + 422, + 414, + 434 + ], + "score": 1.0, + "content": "R2L attention focus on local contextual information with one regional token.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 41.5, + "bbox_fs": [ + 105, + 368, + 507, + 434 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 439, + 506, + 517 + ], + "lines": [ + { + "bbox": [ + 105, + 438, + 507, + 453 + ], + "spans": [ + { + "bbox": [ + 105, + 438, + 507, + 453 + ], + "score": 1.0, + "content": "Moreover, the R2L transformer encoder can further reduce the memory complexity significantly.", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 104, + 450, + 507, + 463 + ], + "spans": [ + { + "bbox": [ + 104, + 450, + 301, + 463 + ], + "score": 1.0, + "content": "The memory complexity of a self-attention is", + "type": "text" + }, + { + "bbox": [ + 301, + 450, + 331, + 462 + ], + "score": 0.93, + "content": "\\mathcal { O } ( N ^ { 2 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 450, + 365, + 463 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 365, + 451, + 376, + 460 + ], + "score": 0.79, + "content": "N", + "type": "inline_equation" + }, + { + "bbox": [ + 376, + 450, + 507, + 463 + ], + "score": 1.0, + "content": "is the number of local tokens.", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 461, + 506, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 207, + 474 + ], + "score": 1.0, + "content": "When a region contains", + "type": "text" + }, + { + "bbox": [ + 207, + 462, + 219, + 471 + ], + "score": 0.74, + "content": "M", + "type": "inline_equation" + }, + { + "bbox": [ + 219, + 461, + 332, + 474 + ], + "score": 1.0, + "content": "local tokens, and there are", + "type": "text" + }, + { + "bbox": [ + 333, + 461, + 357, + 473 + ], + "score": 0.89, + "content": "N / M", + "type": "inline_equation" + }, + { + "bbox": [ + 357, + 461, + 506, + 474 + ], + "score": 1.0, + "content": "regions, the memory complexity is", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 107, + 472, + 506, + 486 + ], + "spans": [ + { + "bbox": [ + 107, + 472, + 206, + 484 + ], + "score": 0.92, + "content": "\\mathcal { O } ( ( M + 1 \\bar { ) } ^ { 2 } \\times ( N / M ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 206, + 472, + 450, + 486 + ], + "score": 1.0, + "content": "; while the complexity from the attention on regional tokens is", + "type": "text" + }, + { + "bbox": [ + 451, + 472, + 502, + 484 + ], + "score": 0.93, + "content": "\\mathcal { O } ( ( N / M ) ^ { 2 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 472, + 506, + 486 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 482, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 261, + 496 + ], + "score": 1.0, + "content": "Hence, the overall complexity becomes", + "type": "text" + }, + { + "bbox": [ + 261, + 483, + 398, + 495 + ], + "score": 0.9, + "content": "\\mathcal { \\dot { O } } ( ( M { \\dot { } } 1 ) ^ { 2 } \\times ( N / M ) + ( N / M ) ^ { 2 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 399, + 482, + 506, + 496 + ], + "score": 1.0, + "content": ". E.g., the main component", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 493, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 234, + 508 + ], + "score": 1.0, + "content": "of RegionViT is stage 3, where", + "type": "text" + }, + { + "bbox": [ + 234, + 495, + 246, + 504 + ], + "score": 0.77, + "content": "M", + "type": "inline_equation" + }, + { + "bbox": [ + 246, + 493, + 286, + 508 + ], + "score": 1.0, + "content": "is 49 and", + "type": "text" + }, + { + "bbox": [ + 286, + 495, + 296, + 504 + ], + "score": 0.81, + "content": "N", + "type": "inline_equation" + }, + { + "bbox": [ + 297, + 493, + 505, + 508 + ], + "score": 1.0, + "content": "is 196; therefore, the memory complexity saving is", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 504, + 480, + 518 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 136, + 516 + ], + "score": 0.87, + "content": "\\sim 7 3 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 504, + 480, + 518 + ], + "score": 1.0, + "content": ". The saving of each model and each stage is varied based on the model configuration.", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 48, + "bbox_fs": [ + 104, + 438, + 507, + 518 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 522, + 505, + 600 + ], + "lines": [ + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "score": 1.0, + "content": "Relative Position Bias. The locality is an important clue for understanding visual content; therefore,", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "score": 1.0, + "content": "instead of adding absolute position embedding used by the vanilla ViT, we introduce the relative", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 105, + 543, + 506, + 557 + ], + "spans": [ + { + "bbox": [ + 105, + 543, + 506, + 557 + ], + "score": 1.0, + "content": "position bias into the attention map of R2L attention since relative position between patches (or", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 105, + 555, + 506, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 506, + 567 + ], + "score": 1.0, + "content": "pixels) are more important than the absolution position as objects in the images can be placed in an", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 105, + 565, + 506, + 578 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 506, + 578 + ], + "score": 1.0, + "content": "arbitrary way (Ramachandran et al., 2019a; Liu et al., 2021). We only add this bias to the attention", + "type": "text" + } + ], + "index": 56 + }, + { + "bbox": [ + 105, + 576, + 506, + 590 + ], + "spans": [ + { + "bbox": [ + 105, + 576, + 506, + 590 + ], + "score": 1.0, + "content": "between local tokens and not the attention between regional tokens and local tokens. Specifically, for", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 104, + 587, + 506, + 601 + ], + "spans": [ + { + "bbox": [ + 104, + 587, + 262, + 601 + ], + "score": 1.0, + "content": "a given pair of local tokens at location,", + "type": "text" + }, + { + "bbox": [ + 262, + 588, + 339, + 600 + ], + "score": 0.91, + "content": "( x _ { m } , y _ { m } ) , ( x _ { n } , y _ { n } )", + "type": "inline_equation" + }, + { + "bbox": [ + 339, + 587, + 417, + 601 + ], + "score": 1.0, + "content": ", the attention value", + "type": "text" + }, + { + "bbox": [ + 417, + 590, + 424, + 598 + ], + "score": 0.76, + "content": "a", + "type": "inline_equation" + }, + { + "bbox": [ + 425, + 587, + 506, + 601 + ], + "score": 1.0, + "content": "can be expressed as", + "type": "text" + } + ], + "index": 58 + } + ], + "index": 55, + "bbox_fs": [ + 104, + 521, + 506, + 601 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 164, + 603, + 445, + 624 + ], + "lines": [ + { + "bbox": [ + 164, + 603, + 445, + 624 + ], + "spans": [ + { + "bbox": [ + 164, + 603, + 445, + 624 + ], + "score": 0.91, + "content": "a _ { ( x _ { m } , y _ { m } ) , ( x _ { n } , y _ { n } ) } = \\mathtt { s o f t m a x } \\left( q _ { ( x _ { m } , y _ { m } ) } k _ { ( x _ { n } , y _ { n } ) } ^ { T } + b _ { ( x _ { m } - x _ { n } , y _ { m } - y _ { n } ) } \\right) ,", + "type": "interline_equation", + "image_path": "6a7aacb1a6a88ae3d8efb902dfd1effb581d8a35b915c4b4bcec7d7ae7b3c1ab.jpg" + } + ] + } + ], + "index": 59, + "virtual_lines": [ + { + "bbox": [ + 164, + 603, + 445, + 624 + ], + "spans": [], + "index": 59 + } + ] + }, + { + "type": "text", + "bbox": [ + 109, + 628, + 504, + 655 + ], + "lines": [ + { + "bbox": [ + 105, + 624, + 507, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 134, + 643 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 201, + 624, + 352, + 643 + ], + "score": 1.0, + "content": "is taken from a learnable parameter", + "type": "text" + }, + { + "bbox": [ + 352, + 627, + 434, + 639 + ], + "score": 0.91, + "content": "B \\in \\mathbb { R } ^ { 2 M - 1 \\times 2 M - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 624, + 466, + 643 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 467, + 628, + 479, + 638 + ], + "score": 0.84, + "content": "M", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 624, + 507, + 643 + ], + "score": 1.0, + "content": "is the", + "type": "text" + } + ], + "index": 60 + }, + { + "bbox": [ + 103, + 633, + 510, + 661 + ], + "spans": [ + { + "bbox": [ + 103, + 633, + 261, + 661 + ], + "score": 1.0, + "content": "m n m n window size of a region. The first term", + "type": "text" + }, + { + "bbox": [ + 262, + 640, + 330, + 655 + ], + "score": 0.94, + "content": "q _ { ( x _ { m } , y _ { m } ) } k _ { ( x _ { n } , y _ { n } ) } ^ { T }", + "type": "inline_equation" + }, + { + "bbox": [ + 330, + 633, + 510, + 661 + ], + "score": 1.0, + "content": "is the attention value based on their content.", + "type": "text" + } + ], + "index": 61 + } + ], + "index": 60.5, + "bbox_fs": [ + 103, + 624, + 510, + 661 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 659, + 503, + 682 + ], + "lines": [ + { + "bbox": [ + 106, + 659, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 659, + 505, + 672 + ], + "score": 1.0, + "content": "With relative position bias, our model does not require the absolute positional embedding as vanilla", + "type": "text" + } + ], + "index": 62 + }, + { + "bbox": [ + 106, + 671, + 414, + 683 + ], + "spans": [ + { + "bbox": [ + 106, + 671, + 414, + 683 + ], + "score": 1.0, + "content": "ViT since this relative position bias helps to encode the position information.", + "type": "text" + } + ], + "index": 63 + } + ], + "index": 62.5, + "bbox_fs": [ + 106, + 659, + 505, + 683 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "score": 1.0, + "content": "Downsampling. The spatial and channel dimension are changed along different stages of our network", + "type": "text" + } + ], + "index": 64 + }, + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 183, + 712 + ], + "score": 1.0, + "content": "by simply applying", + "type": "text" + }, + { + "bbox": [ + 184, + 699, + 206, + 709 + ], + "score": 0.88, + "content": "3 \\times 3", + "type": "inline_equation" + }, + { + "bbox": [ + 206, + 699, + 506, + 712 + ], + "score": 1.0, + "content": "depth-wise convolution with stride 2 for halving the resolution and doubling", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "the channel dimension for both regional and local tokens (weights are shared) (Heo et al., 2021). We", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 105, + 720, + 502, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 502, + 734 + ], + "score": 1.0, + "content": "also test with regular convolutions but it does not improve the accuracy with increased complexity.", + "type": "text" + } + ], + "index": 67 + } + ], + "index": 65.5, + "bbox_fs": [ + 105, + 688, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 125, + 86, + 484, + 144 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 63, + 504, + 85 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 62, + 505, + 76 + ], + "spans": [ + { + "bbox": [ + 105, + 62, + 505, + 76 + ], + "score": 1.0, + "content": "Table 2: Model architectures of RegionViT. For all networks, the dimension per head is 32 and the expanding", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 74, + 504, + 85 + ], + "spans": [ + { + "bbox": [ + 106, + 74, + 125, + 85 + ], + "score": 1.0, + "content": "ratio", + "type": "text" + }, + { + "bbox": [ + 125, + 76, + 131, + 83 + ], + "score": 0.66, + "content": "r", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 74, + 474, + 85 + ], + "score": 1.0, + "content": "in FFN is 4. The patch size of local tokens is always 4 while the patch size of regional tokens is", + "type": "text" + }, + { + "bbox": [ + 474, + 74, + 501, + 83 + ], + "score": 0.89, + "content": "4 \\times M", + "type": "inline_equation" + }, + { + "bbox": [ + 501, + 74, + 504, + 85 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 125, + 86, + 484, + 144 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 125, + 86, + 484, + 144 + ], + "spans": [ + { + "bbox": [ + 125, + 86, + 484, + 144 + ], + "score": 0.978, + "html": "
ModelTokenization Local RegionalWindow size (M)Dimension (C)# of Encoders at each stage
RegionViT-Ti3-convlinear7{64,128,256,512}{2,2,8,2}
RegionViT-S3-convlinear7{96,192,384,768}{2,2,8,2}
RegionViT-M1-convlinear7{96,192,384,768}{2,2,14,2}
RegionViT-B1-convlinear7{128,256,512,1024}{2,2,14,2}
", + "type": "table", + "image_path": "dead8e6de6e69bd14273dd851771f492174a837162883f43fa29665e5f26b5d4.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 125, + 86, + 484, + 105.33333333333333 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 125, + 105.33333333333333, + 484, + 124.66666666666666 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 125, + 124.66666666666666, + 484, + 144.0 + ], + "spans": [], + "index": 4 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 126, + 144, + 480, + 158 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 128, + 142, + 435, + 152 + ], + "spans": [ + { + "bbox": [ + 128, + 142, + 188, + 152 + ], + "score": 1.0, + "content": "1-conv: Conv(out=C,", + "type": "text" + }, + { + "bbox": [ + 188, + 143, + 199, + 150 + ], + "score": 0.61, + "content": "k { = } 8", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 142, + 201, + 152 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 202, + 144, + 213, + 150 + ], + "score": 0.45, + "content": "s { = } 4", + "type": "inline_equation" + }, + { + "bbox": [ + 213, + 142, + 214, + 152 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 215, + 143, + 227, + 150 + ], + "score": 0.33, + "content": "\\overline { { { \\bf p } { = } 3 } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 142, + 295, + 152 + ], + "score": 1.0, + "content": ". 3-conv: Conv(out=C/4,", + "type": "text" + }, + { + "bbox": [ + 295, + 143, + 306, + 150 + ], + "score": 0.67, + "content": "\\overline { { \\mathbf { k } = 3 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 307, + 142, + 308, + 152 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 309, + 143, + 319, + 150 + ], + "score": 0.54, + "content": "\\overline { { \\mathsf { s } { = } 2 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 320, + 142, + 321, + 152 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 322, + 144, + 333, + 150 + ], + "score": 0.38, + "content": "\\mathrm { p } { = } 1", + "type": "inline_equation" + }, + { + "bbox": [ + 333, + 142, + 335, + 152 + ], + "score": 1.0, + "content": ")", + "type": "text" + }, + { + "bbox": [ + 336, + 144, + 345, + 150 + ], + "score": 0.42, + "content": "", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 142, + 386, + 152 + ], + "score": 1.0, + "content": "Conv(out=C/2,", + "type": "text" + }, + { + "bbox": [ + 386, + 143, + 398, + 150 + ], + "score": 0.49, + "content": "\\overline { { \\mathbf { k } = 3 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 398, + 142, + 427, + 152 + ], + "score": 1.0, + "content": ", s=2, p=1)", + "type": "text" + }, + { + "bbox": [ + 428, + 144, + 435, + 150 + ], + "score": 0.32, + "content": "", + "type": "inline_equation" + } + ], + "index": 5 + }, + { + "bbox": [ + 129, + 150, + 479, + 158 + ], + "spans": [ + { + "bbox": [ + 129, + 150, + 166, + 158 + ], + "score": 1.0, + "content": "Conv(out=C,", + "type": "text" + }, + { + "bbox": [ + 166, + 150, + 178, + 158 + ], + "score": 0.51, + "content": "\\mathrm { k } = 3", + "type": "inline_equation" + }, + { + "bbox": [ + 178, + 150, + 179, + 158 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 180, + 150, + 190, + 158 + ], + "score": 0.61, + "content": "\\mathrm { s } { = } 1", + "type": "inline_equation" + }, + { + "bbox": [ + 191, + 150, + 192, + 158 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 192, + 151, + 205, + 158 + ], + "score": 0.27, + "content": "\\mathbf { p } { = } 1 ,", + "type": "inline_equation" + }, + { + "bbox": [ + 205, + 150, + 479, + 158 + ], + "score": 1.0, + "content": "), where C is the channel dimension of the first block. LayerNorm and GeLU are added between Conv.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 106, + 162, + 505, + 228 + ], + "lines": [ + { + "bbox": [ + 106, + 162, + 506, + 173 + ], + "spans": [ + { + "bbox": [ + 106, + 162, + 449, + 173 + ], + "score": 1.0, + "content": "Input Tokenization. The tokenization for local tokens can be implemented by using a", + "type": "text" + }, + { + "bbox": [ + 450, + 162, + 469, + 172 + ], + "score": 0.88, + "content": "4 \\times 4", + "type": "inline_equation" + }, + { + "bbox": [ + 469, + 162, + 506, + 173 + ], + "score": 1.0, + "content": "convolu-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 173, + 505, + 185 + ], + "spans": [ + { + "bbox": [ + 106, + 173, + 197, + 185 + ], + "score": 1.0, + "content": "tion with channel size", + "type": "text" + }, + { + "bbox": [ + 197, + 173, + 207, + 183 + ], + "score": 0.77, + "content": "C", + "type": "inline_equation" + }, + { + "bbox": [ + 207, + 173, + 505, + 185 + ], + "score": 1.0, + "content": "and stride 4 (i.e. non-overlapped). However, as pointed out in T2T (Yuan", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 183, + 505, + 196 + ], + "spans": [ + { + "bbox": [ + 105, + 183, + 505, + 196 + ], + "score": 1.0, + "content": "et al., 2021), CrossViT (Chen et al., 2021a), and PiT (Heo et al., 2021), using a stronger but still", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 195, + 505, + 207 + ], + "spans": [ + { + "bbox": [ + 106, + 195, + 505, + 207 + ], + "score": 1.0, + "content": "simple subnetwork for the tokenization could further improve the performance, especially for the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 207, + 504, + 218 + ], + "spans": [ + { + "bbox": [ + 106, + 207, + 504, + 218 + ], + "score": 1.0, + "content": "smaller models. Thus, we adopt two input tokenizations in our model, one still contains only one", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 217, + 491, + 228 + ], + "spans": [ + { + "bbox": [ + 106, + 217, + 491, + 228 + ], + "score": 1.0, + "content": "convolutional layer and another one contains three convolutional layers (see Table 2 for details).", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 9.5 + }, + { + "type": "text", + "bbox": [ + 107, + 234, + 505, + 278 + ], + "lines": [ + { + "bbox": [ + 105, + 233, + 506, + 246 + ], + "spans": [ + { + "bbox": [ + 105, + 233, + 506, + 246 + ], + "score": 1.0, + "content": "Regional Tokens. We adopt a simple approach to generate regional tokens, that is, using linear", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 245, + 505, + 257 + ], + "spans": [ + { + "bbox": [ + 106, + 245, + 505, + 257 + ], + "score": 1.0, + "content": "projection with larger patch sizes in comparison to the patch size for the local tokens. E.g., as shown", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 255, + 505, + 269 + ], + "spans": [ + { + "bbox": [ + 105, + 255, + 282, + 269 + ], + "score": 1.0, + "content": "in Figure 2, the patch size for local tokens is", + "type": "text" + }, + { + "bbox": [ + 282, + 256, + 302, + 266 + ], + "score": 0.88, + "content": "4 \\times 4", + "type": "inline_equation" + }, + { + "bbox": [ + 302, + 255, + 405, + 269 + ], + "score": 1.0, + "content": ", we simply use patch size", + "type": "text" + }, + { + "bbox": [ + 405, + 256, + 434, + 266 + ], + "score": 0.89, + "content": "2 8 \\times 2 8", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 255, + 505, + 269 + ], + "score": 1.0, + "content": "to split the image", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 267, + 363, + 279 + ], + "spans": [ + { + "bbox": [ + 106, + 267, + 363, + 279 + ], + "score": 1.0, + "content": "and then project each patch linearly to generate regional tokens.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 14.5 + }, + { + "type": "text", + "bbox": [ + 107, + 283, + 504, + 306 + ], + "lines": [ + { + "bbox": [ + 106, + 284, + 505, + 295 + ], + "spans": [ + { + "bbox": [ + 106, + 284, + 505, + 295 + ], + "score": 1.0, + "content": "Table 2 shows our RegionViT models with different configurations. By default, the window size is 7", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 294, + 479, + 307 + ], + "spans": [ + { + "bbox": [ + 105, + 294, + 468, + 307 + ], + "score": 1.0, + "content": "while we also experiment with a larger window size (14), those models are annotated with", + "type": "text" + }, + { + "bbox": [ + 468, + 296, + 475, + 304 + ], + "score": 0.74, + "content": "^ +", + "type": "inline_equation" + }, + { + "bbox": [ + 475, + 294, + 479, + 307 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5 + }, + { + "type": "title", + "bbox": [ + 108, + 322, + 200, + 335 + ], + "lines": [ + { + "bbox": [ + 104, + 320, + 201, + 337 + ], + "spans": [ + { + "bbox": [ + 104, + 320, + 201, + 337 + ], + "score": 1.0, + "content": "4 EXPERIMENTS", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "title", + "bbox": [ + 108, + 347, + 235, + 358 + ], + "lines": [ + { + "bbox": [ + 105, + 347, + 236, + 360 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 236, + 360 + ], + "score": 1.0, + "content": "4.1 IMAGE CLASSIFICATION", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 367, + 506, + 444 + ], + "lines": [ + { + "bbox": [ + 106, + 367, + 505, + 380 + ], + "spans": [ + { + "bbox": [ + 106, + 367, + 505, + 380 + ], + "score": 1.0, + "content": "Datasets. We use ImageNet1K (Deng et al., 2009) (IN1K) and ImageNet21K (Deng et al., 2009)", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 378, + 506, + 390 + ], + "spans": [ + { + "bbox": [ + 106, + 378, + 506, + 390 + ], + "score": 1.0, + "content": "(IN21K) to validate our method. ImageNet1K contains 1.28 million training images and 50k val-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 389, + 505, + 401 + ], + "spans": [ + { + "bbox": [ + 106, + 389, + 505, + 401 + ], + "score": 1.0, + "content": "idation images over 1k classes, and ImageNet21K is a large-scale dataset that consists of around", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 400, + 506, + 412 + ], + "spans": [ + { + "bbox": [ + 105, + 400, + 506, + 412 + ], + "score": 1.0, + "content": "14 million images over 21,841 classes. We use all images for training and then finetune the model", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 411, + 506, + 422 + ], + "spans": [ + { + "bbox": [ + 106, + 411, + 506, + 422 + ], + "score": 1.0, + "content": "on ImageNet1K. Moreover, we also perform the transfer learning from ImageNet1K to five down-", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 422, + 506, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 422, + 506, + 434 + ], + "score": 1.0, + "content": "stream datasets, including CIFAR10 (Krizhevsky et al., 2009), CIFAR100 (Krizhevsky et al., 2009),", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 433, + 506, + 445 + ], + "spans": [ + { + "bbox": [ + 106, + 433, + 506, + 445 + ], + "score": 1.0, + "content": "IIIPets (Parkhi et al., 2012), StandfordCars (Krause et al., 2013) and ChestXRay8 (Wang et al., 2017).", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 106, + 450, + 505, + 538 + ], + "lines": [ + { + "bbox": [ + 105, + 448, + 506, + 463 + ], + "spans": [ + { + "bbox": [ + 105, + 448, + 506, + 463 + ], + "score": 1.0, + "content": "Training and Evaluation. We follow DeiT (Touvron et al., 2020) to train our models on IN1K except", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 460, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 106, + 460, + 505, + 473 + ], + "score": 1.0, + "content": "that we use batch size 4,096 with a base learning rate 0.004 and the warm-up epochs is 50. We adopt", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 471, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 505, + 484 + ], + "score": 1.0, + "content": "the AdamW (Loshchilov & Hutter, 2019) optimizer with cosine learning rate scheduler (Loshchilov", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 482, + 507, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 507, + 496 + ], + "score": 1.0, + "content": "& Hutter, 2017). We apply Mixup (Zhang et al., 2018), CutMix (Yun et al., 2019), RandomEras-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 104, + 493, + 506, + 506 + ], + "spans": [ + { + "bbox": [ + 104, + 493, + 506, + 506 + ], + "score": 1.0, + "content": "ing (Zhong et al., 2020), label smoothing (Szegedy et al., 2016), RandAugment (Cubuk et al., 2020)", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 505, + 505, + 517 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 437, + 517 + ], + "score": 1.0, + "content": "and instance repetition (Hoffer et al., 2020). During training, we random cropped a", + "type": "text" + }, + { + "bbox": [ + 437, + 505, + 477, + 515 + ], + "score": 0.9, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 477, + 505, + 505, + 517 + ], + "score": 1.0, + "content": "region", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 516, + 505, + 527 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 147, + 527 + ], + "score": 1.0, + "content": "and take a", + "type": "text" + }, + { + "bbox": [ + 148, + 516, + 186, + 527 + ], + "score": 0.9, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 187, + 516, + 505, + 527 + ], + "score": 1.0, + "content": "center crop after resizing the shorter side to 256 for evaluation. We used a similar", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 527, + 452, + 539 + ], + "spans": [ + { + "bbox": [ + 106, + 527, + 452, + 539 + ], + "score": 1.0, + "content": "setting for IN21K and transfer learning, and more details can be found in Section A.1.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 31.5 + }, + { + "type": "table", + "bbox": [ + 111, + 579, + 300, + 729 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 557, + 504, + 577 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 556, + 505, + 568 + ], + "spans": [ + { + "bbox": [ + 105, + 556, + 505, + 568 + ], + "score": 1.0, + "content": "Table 3: Comparisons with recent pyramid-like structure-based ViT models on ImageNet1K. The bold numbers", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 567, + 270, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 567, + 270, + 578 + ], + "score": 1.0, + "content": "indicate the best number within each section.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5 + }, + { + "type": "table_body", + "bbox": [ + 111, + 579, + 300, + 729 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 111, + 579, + 300, + 729 + ], + "spans": [ + { + "bbox": [ + 111, + 579, + 300, + 729 + ], + "score": 0.966, + "html": "
ModelParams (M)FLOPs (G)Acc. (%)
PiT-XS (Heo et al., 2021)10.61.478.1
ConT-S(Yan et al.,2021)10.11.576.5
PVT-T(Wang et al.,2021)13.21.975.1
ConViT-Ti+ (d'Ascoli et al., 2021)10.02.076.7
Twins-SVT-S (Chu et al.,2021a)24.02.881.7
PiT-S (Heo et al., 2021)23.52.980.9
ConT-M(Yan et al., 2021)19.23.180.2
Twins-PCPVT-S (Chu et al., 2021a)24.13.781.2
PVT-S (Wang et al., 2021)24.53.879.8
RegionViT-Ti13.82.480.4
RegionViT-Ti+14.32.781.5
DeiT-S (Touvron et al., 2020)22.14.679.9
CvT-13 (Wu et al.,2021)20.04.581.6
Swin-T (Liu et al., 2021)29.04.581.3
LocalViT-S (Li et al.,2021)22.44.680.8
ViL-S (Zhang et al., 2021a)24.64.982.0
Visformer-S (Chen et al.,2021b)40.24.982.3
ConViT-S (d'Ascoli et al.,2021)27.05.481.3
NesT-S (Zhang et al., 2021b)17.05.881.5
RegionViT-S30.65.382.6
RegionViT-S+31.35.783.3
", + "type": "table", + "image_path": "0b6d088036456b2a4c78444f0295f127489f30e37f3c7eae0c15593287ace348.jpg" + } + ] + } + ], + "index": 43.5, + "virtual_lines": [ + { + "bbox": [ + 111, + 579, + 300, + 591.5 + ], + "spans": [], + "index": 38 + }, + { + "bbox": [ + 111, + 591.5, + 300, + 604.0 + ], + "spans": [], + "index": 39 + }, + { + "bbox": [ + 111, + 604.0, + 300, + 616.5 + ], + "spans": [], + "index": 40 + }, + { + "bbox": [ + 111, + 616.5, + 300, + 629.0 + ], + "spans": [], + "index": 41 + }, + { + "bbox": [ + 111, + 629.0, + 300, + 641.5 + ], + "spans": [], + "index": 42 + }, + { + "bbox": [ + 111, + 641.5, + 300, + 654.0 + ], + "spans": [], + "index": 43 + }, + { + "bbox": [ + 111, + 654.0, + 300, + 666.5 + ], + "spans": [], + "index": 44 + }, + { + "bbox": [ + 111, + 666.5, + 300, + 679.0 + ], + "spans": [], + "index": 45 + }, + { + "bbox": [ + 111, + 679.0, + 300, + 691.5 + ], + "spans": [], + "index": 46 + }, + { + "bbox": [ + 111, + 691.5, + 300, + 704.0 + ], + "spans": [], + "index": 47 + }, + { + "bbox": [ + 111, + 704.0, + 300, + 716.5 + ], + "spans": [], + "index": 48 + }, + { + "bbox": [ + 111, + 716.5, + 300, + 729.0 + ], + "spans": [], + "index": 49 + } + ] + } + ], + "index": 40.0 + }, + { + "type": "table", + "bbox": [ + 311, + 579, + 499, + 729 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 311, + 579, + 499, + 729 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 311, + 579, + 499, + 729 + ], + "spans": [ + { + "bbox": [ + 311, + 579, + 499, + 729 + ], + "score": 0.97, + "html": "
ModelParams (M)FLOPs (G)Acc. (%)
ConT-M(Yan et al., 2021)39.66.481.8
Twins-PCPVT-B (Chu et al.,2021a)43.86.482.7
PVT-M(Wang et al., 2021)44.26.781.2
CvT-21(Wu et al.,2021)32.07.182.5
Twins-SVT-B (Chu et al., 2021a)568.383.2
ViL-M (Zhang et al.,2021a)39.78.783.3
Swin-S (Liu et al., 2021)50.08.783.0
PVT-L (Wang et al.,2021)61.49.881.7
ConViT-S+ (d'Ascoli et al., 2021)48.010.082.2
NesT-S (Zhang et al., 2021b)38.010.483.3
RegionViT-M41.27.483.1
RegionViT-M+42.07.983.4
DeiT-B (Touvron et al.,2020)86.617.681.8
PiT-B (Heo et al., 2021)73.812.582.0
ViL-B (Zhang et al., 2021a)55.713.483.2
Twins-SVT-L(Chu et al.,2021a)99.214.883.7
Swin-B (Liu et al.,2021)88.015.483.5
ConViT-B (d'Ascoli et al.,2021)86.017.082.4
NesT-B (Zhang et al., 2021b)68.017.983.8
RegionViT-B72.713.083.2
RegionViT-B+73.813.683.8
", + "type": "table", + "image_path": "30316ee497884d718c99f74f5a3484680fc2081650e7990719ee5b51f13908ad.jpg" + } + ] + } + ], + "index": 55.5, + "virtual_lines": [ + { + "bbox": [ + 311, + 579, + 499, + 591.5 + ], + "spans": [], + "index": 50 + }, + { + "bbox": [ + 311, + 591.5, + 499, + 604.0 + ], + "spans": [], + "index": 51 + }, + { + "bbox": [ + 311, + 604.0, + 499, + 616.5 + ], + "spans": [], + "index": 52 + }, + { + "bbox": [ + 311, + 616.5, + 499, + 629.0 + ], + "spans": [], + "index": 53 + }, + { + "bbox": [ + 311, + 629.0, + 499, + 641.5 + ], + "spans": [], + "index": 54 + }, + { + "bbox": [ + 311, + 641.5, + 499, + 654.0 + ], + "spans": [], + "index": 55 + }, + { + "bbox": [ + 311, + 654.0, + 499, + 666.5 + ], + "spans": [], + "index": 56 + }, + { + "bbox": [ + 311, + 666.5, + 499, + 679.0 + ], + "spans": [], + "index": 57 + }, + { + "bbox": [ + 311, + 679.0, + 499, + 691.5 + ], + "spans": [], + "index": 58 + }, + { + "bbox": [ + 311, + 691.5, + 499, + 704.0 + ], + "spans": [], + "index": 59 + }, + { + "bbox": [ + 311, + 704.0, + 499, + 716.5 + ], + "spans": [], + "index": 60 + }, + { + "bbox": [ + 311, + 716.5, + 499, + 729.0 + ], + "spans": [], + "index": 61 + } + ] + } + ], + "index": 55.5 + } + ], + "page_idx": 5, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 310, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 310, + 762 + ], + "score": 1.0, + "content": "6", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 125, + 86, + 484, + 144 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 63, + 504, + 85 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 62, + 505, + 76 + ], + "spans": [ + { + "bbox": [ + 105, + 62, + 505, + 76 + ], + "score": 1.0, + "content": "Table 2: Model architectures of RegionViT. For all networks, the dimension per head is 32 and the expanding", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 74, + 504, + 85 + ], + "spans": [ + { + "bbox": [ + 106, + 74, + 125, + 85 + ], + "score": 1.0, + "content": "ratio", + "type": "text" + }, + { + "bbox": [ + 125, + 76, + 131, + 83 + ], + "score": 0.66, + "content": "r", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 74, + 474, + 85 + ], + "score": 1.0, + "content": "in FFN is 4. The patch size of local tokens is always 4 while the patch size of regional tokens is", + "type": "text" + }, + { + "bbox": [ + 474, + 74, + 501, + 83 + ], + "score": 0.89, + "content": "4 \\times M", + "type": "inline_equation" + }, + { + "bbox": [ + 501, + 74, + 504, + 85 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 125, + 86, + 484, + 144 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 125, + 86, + 484, + 144 + ], + "spans": [ + { + "bbox": [ + 125, + 86, + 484, + 144 + ], + "score": 0.978, + "html": "
ModelTokenization Local RegionalWindow size (M)Dimension (C)# of Encoders at each stage
RegionViT-Ti3-convlinear7{64,128,256,512}{2,2,8,2}
RegionViT-S3-convlinear7{96,192,384,768}{2,2,8,2}
RegionViT-M1-convlinear7{96,192,384,768}{2,2,14,2}
RegionViT-B1-convlinear7{128,256,512,1024}{2,2,14,2}
", + "type": "table", + "image_path": "dead8e6de6e69bd14273dd851771f492174a837162883f43fa29665e5f26b5d4.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 125, + 86, + 484, + 105.33333333333333 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 125, + 105.33333333333333, + 484, + 124.66666666666666 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 125, + 124.66666666666666, + 484, + 144.0 + ], + "spans": [], + "index": 4 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 126, + 144, + 480, + 158 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 128, + 142, + 435, + 152 + ], + "spans": [ + { + "bbox": [ + 128, + 142, + 188, + 152 + ], + "score": 1.0, + "content": "1-conv: Conv(out=C,", + "type": "text" + }, + { + "bbox": [ + 188, + 143, + 199, + 150 + ], + "score": 0.61, + "content": "k { = } 8", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 142, + 201, + 152 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 202, + 144, + 213, + 150 + ], + "score": 0.45, + "content": "s { = } 4", + "type": "inline_equation" + }, + { + "bbox": [ + 213, + 142, + 214, + 152 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 215, + 143, + 227, + 150 + ], + "score": 0.33, + "content": "\\overline { { { \\bf p } { = } 3 } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 142, + 295, + 152 + ], + "score": 1.0, + "content": ". 3-conv: Conv(out=C/4,", + "type": "text" + }, + { + "bbox": [ + 295, + 143, + 306, + 150 + ], + "score": 0.67, + "content": "\\overline { { \\mathbf { k } = 3 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 307, + 142, + 308, + 152 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 309, + 143, + 319, + 150 + ], + "score": 0.54, + "content": "\\overline { { \\mathsf { s } { = } 2 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 320, + 142, + 321, + 152 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 322, + 144, + 333, + 150 + ], + "score": 0.38, + "content": "\\mathrm { p } { = } 1", + "type": "inline_equation" + }, + { + "bbox": [ + 333, + 142, + 335, + 152 + ], + "score": 1.0, + "content": ")", + "type": "text" + }, + { + "bbox": [ + 336, + 144, + 345, + 150 + ], + "score": 0.42, + "content": "", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 142, + 386, + 152 + ], + "score": 1.0, + "content": "Conv(out=C/2,", + "type": "text" + }, + { + "bbox": [ + 386, + 143, + 398, + 150 + ], + "score": 0.49, + "content": "\\overline { { \\mathbf { k } = 3 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 398, + 142, + 427, + 152 + ], + "score": 1.0, + "content": ", s=2, p=1)", + "type": "text" + }, + { + "bbox": [ + 428, + 144, + 435, + 150 + ], + "score": 0.32, + "content": "", + "type": "inline_equation" + } + ], + "index": 5 + }, + { + "bbox": [ + 129, + 150, + 479, + 158 + ], + "spans": [ + { + "bbox": [ + 129, + 150, + 166, + 158 + ], + "score": 1.0, + "content": "Conv(out=C,", + "type": "text" + }, + { + "bbox": [ + 166, + 150, + 178, + 158 + ], + "score": 0.51, + "content": "\\mathrm { k } = 3", + "type": "inline_equation" + }, + { + "bbox": [ + 178, + 150, + 179, + 158 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 180, + 150, + 190, + 158 + ], + "score": 0.61, + "content": "\\mathrm { s } { = } 1", + "type": "inline_equation" + }, + { + "bbox": [ + 191, + 150, + 192, + 158 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 192, + 151, + 205, + 158 + ], + "score": 0.27, + "content": "\\mathbf { p } { = } 1 ,", + "type": "inline_equation" + }, + { + "bbox": [ + 205, + 150, + 479, + 158 + ], + "score": 1.0, + "content": "), where C is the channel dimension of the first block. LayerNorm and GeLU are added between Conv.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 106, + 162, + 505, + 228 + ], + "lines": [ + { + "bbox": [ + 106, + 162, + 506, + 173 + ], + "spans": [ + { + "bbox": [ + 106, + 162, + 449, + 173 + ], + "score": 1.0, + "content": "Input Tokenization. The tokenization for local tokens can be implemented by using a", + "type": "text" + }, + { + "bbox": [ + 450, + 162, + 469, + 172 + ], + "score": 0.88, + "content": "4 \\times 4", + "type": "inline_equation" + }, + { + "bbox": [ + 469, + 162, + 506, + 173 + ], + "score": 1.0, + "content": "convolu-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 173, + 505, + 185 + ], + "spans": [ + { + "bbox": [ + 106, + 173, + 197, + 185 + ], + "score": 1.0, + "content": "tion with channel size", + "type": "text" + }, + { + "bbox": [ + 197, + 173, + 207, + 183 + ], + "score": 0.77, + "content": "C", + "type": "inline_equation" + }, + { + "bbox": [ + 207, + 173, + 505, + 185 + ], + "score": 1.0, + "content": "and stride 4 (i.e. non-overlapped). However, as pointed out in T2T (Yuan", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 183, + 505, + 196 + ], + "spans": [ + { + "bbox": [ + 105, + 183, + 505, + 196 + ], + "score": 1.0, + "content": "et al., 2021), CrossViT (Chen et al., 2021a), and PiT (Heo et al., 2021), using a stronger but still", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 195, + 505, + 207 + ], + "spans": [ + { + "bbox": [ + 106, + 195, + 505, + 207 + ], + "score": 1.0, + "content": "simple subnetwork for the tokenization could further improve the performance, especially for the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 207, + 504, + 218 + ], + "spans": [ + { + "bbox": [ + 106, + 207, + 504, + 218 + ], + "score": 1.0, + "content": "smaller models. Thus, we adopt two input tokenizations in our model, one still contains only one", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 217, + 491, + 228 + ], + "spans": [ + { + "bbox": [ + 106, + 217, + 491, + 228 + ], + "score": 1.0, + "content": "convolutional layer and another one contains three convolutional layers (see Table 2 for details).", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 9.5, + "bbox_fs": [ + 105, + 162, + 506, + 228 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 234, + 505, + 278 + ], + "lines": [ + { + "bbox": [ + 105, + 233, + 506, + 246 + ], + "spans": [ + { + "bbox": [ + 105, + 233, + 506, + 246 + ], + "score": 1.0, + "content": "Regional Tokens. We adopt a simple approach to generate regional tokens, that is, using linear", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 245, + 505, + 257 + ], + "spans": [ + { + "bbox": [ + 106, + 245, + 505, + 257 + ], + "score": 1.0, + "content": "projection with larger patch sizes in comparison to the patch size for the local tokens. E.g., as shown", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 255, + 505, + 269 + ], + "spans": [ + { + "bbox": [ + 105, + 255, + 282, + 269 + ], + "score": 1.0, + "content": "in Figure 2, the patch size for local tokens is", + "type": "text" + }, + { + "bbox": [ + 282, + 256, + 302, + 266 + ], + "score": 0.88, + "content": "4 \\times 4", + "type": "inline_equation" + }, + { + "bbox": [ + 302, + 255, + 405, + 269 + ], + "score": 1.0, + "content": ", we simply use patch size", + "type": "text" + }, + { + "bbox": [ + 405, + 256, + 434, + 266 + ], + "score": 0.89, + "content": "2 8 \\times 2 8", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 255, + 505, + 269 + ], + "score": 1.0, + "content": "to split the image", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 267, + 363, + 279 + ], + "spans": [ + { + "bbox": [ + 106, + 267, + 363, + 279 + ], + "score": 1.0, + "content": "and then project each patch linearly to generate regional tokens.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 14.5, + "bbox_fs": [ + 105, + 233, + 506, + 279 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 283, + 504, + 306 + ], + "lines": [ + { + "bbox": [ + 106, + 284, + 505, + 295 + ], + "spans": [ + { + "bbox": [ + 106, + 284, + 505, + 295 + ], + "score": 1.0, + "content": "Table 2 shows our RegionViT models with different configurations. By default, the window size is 7", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 294, + 479, + 307 + ], + "spans": [ + { + "bbox": [ + 105, + 294, + 468, + 307 + ], + "score": 1.0, + "content": "while we also experiment with a larger window size (14), those models are annotated with", + "type": "text" + }, + { + "bbox": [ + 468, + 296, + 475, + 304 + ], + "score": 0.74, + "content": "^ +", + "type": "inline_equation" + }, + { + "bbox": [ + 475, + 294, + 479, + 307 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5, + "bbox_fs": [ + 105, + 284, + 505, + 307 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 322, + 200, + 335 + ], + "lines": [ + { + "bbox": [ + 104, + 320, + 201, + 337 + ], + "spans": [ + { + "bbox": [ + 104, + 320, + 201, + 337 + ], + "score": 1.0, + "content": "4 EXPERIMENTS", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "title", + "bbox": [ + 108, + 347, + 235, + 358 + ], + "lines": [ + { + "bbox": [ + 105, + 347, + 236, + 360 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 236, + 360 + ], + "score": 1.0, + "content": "4.1 IMAGE CLASSIFICATION", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 367, + 506, + 444 + ], + "lines": [ + { + "bbox": [ + 106, + 367, + 505, + 380 + ], + "spans": [ + { + "bbox": [ + 106, + 367, + 505, + 380 + ], + "score": 1.0, + "content": "Datasets. We use ImageNet1K (Deng et al., 2009) (IN1K) and ImageNet21K (Deng et al., 2009)", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 378, + 506, + 390 + ], + "spans": [ + { + "bbox": [ + 106, + 378, + 506, + 390 + ], + "score": 1.0, + "content": "(IN21K) to validate our method. ImageNet1K contains 1.28 million training images and 50k val-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 389, + 505, + 401 + ], + "spans": [ + { + "bbox": [ + 106, + 389, + 505, + 401 + ], + "score": 1.0, + "content": "idation images over 1k classes, and ImageNet21K is a large-scale dataset that consists of around", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 400, + 506, + 412 + ], + "spans": [ + { + "bbox": [ + 105, + 400, + 506, + 412 + ], + "score": 1.0, + "content": "14 million images over 21,841 classes. We use all images for training and then finetune the model", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 411, + 506, + 422 + ], + "spans": [ + { + "bbox": [ + 106, + 411, + 506, + 422 + ], + "score": 1.0, + "content": "on ImageNet1K. Moreover, we also perform the transfer learning from ImageNet1K to five down-", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 422, + 506, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 422, + 506, + 434 + ], + "score": 1.0, + "content": "stream datasets, including CIFAR10 (Krizhevsky et al., 2009), CIFAR100 (Krizhevsky et al., 2009),", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 433, + 506, + 445 + ], + "spans": [ + { + "bbox": [ + 106, + 433, + 506, + 445 + ], + "score": 1.0, + "content": "IIIPets (Parkhi et al., 2012), StandfordCars (Krause et al., 2013) and ChestXRay8 (Wang et al., 2017).", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 24, + "bbox_fs": [ + 105, + 367, + 506, + 445 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 450, + 505, + 538 + ], + "lines": [ + { + "bbox": [ + 105, + 448, + 506, + 463 + ], + "spans": [ + { + "bbox": [ + 105, + 448, + 506, + 463 + ], + "score": 1.0, + "content": "Training and Evaluation. We follow DeiT (Touvron et al., 2020) to train our models on IN1K except", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 460, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 106, + 460, + 505, + 473 + ], + "score": 1.0, + "content": "that we use batch size 4,096 with a base learning rate 0.004 and the warm-up epochs is 50. We adopt", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 471, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 505, + 484 + ], + "score": 1.0, + "content": "the AdamW (Loshchilov & Hutter, 2019) optimizer with cosine learning rate scheduler (Loshchilov", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 482, + 507, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 507, + 496 + ], + "score": 1.0, + "content": "& Hutter, 2017). We apply Mixup (Zhang et al., 2018), CutMix (Yun et al., 2019), RandomEras-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 104, + 493, + 506, + 506 + ], + "spans": [ + { + "bbox": [ + 104, + 493, + 506, + 506 + ], + "score": 1.0, + "content": "ing (Zhong et al., 2020), label smoothing (Szegedy et al., 2016), RandAugment (Cubuk et al., 2020)", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 505, + 505, + 517 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 437, + 517 + ], + "score": 1.0, + "content": "and instance repetition (Hoffer et al., 2020). During training, we random cropped a", + "type": "text" + }, + { + "bbox": [ + 437, + 505, + 477, + 515 + ], + "score": 0.9, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 477, + 505, + 505, + 517 + ], + "score": 1.0, + "content": "region", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 516, + 505, + 527 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 147, + 527 + ], + "score": 1.0, + "content": "and take a", + "type": "text" + }, + { + "bbox": [ + 148, + 516, + 186, + 527 + ], + "score": 0.9, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 187, + 516, + 505, + 527 + ], + "score": 1.0, + "content": "center crop after resizing the shorter side to 256 for evaluation. We used a similar", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 527, + 452, + 539 + ], + "spans": [ + { + "bbox": [ + 106, + 527, + 452, + 539 + ], + "score": 1.0, + "content": "setting for IN21K and transfer learning, and more details can be found in Section A.1.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 31.5, + "bbox_fs": [ + 104, + 448, + 507, + 539 + ] + }, + { + "type": "table", + "bbox": [ + 111, + 579, + 300, + 729 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 557, + 504, + 577 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 556, + 505, + 568 + ], + "spans": [ + { + "bbox": [ + 105, + 556, + 505, + 568 + ], + "score": 1.0, + "content": "Table 3: Comparisons with recent pyramid-like structure-based ViT models on ImageNet1K. The bold numbers", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 567, + 270, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 567, + 270, + 578 + ], + "score": 1.0, + "content": "indicate the best number within each section.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5 + }, + { + "type": "table_body", + "bbox": [ + 111, + 579, + 300, + 729 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 111, + 579, + 300, + 729 + ], + "spans": [ + { + "bbox": [ + 111, + 579, + 300, + 729 + ], + "score": 0.966, + "html": "
ModelParams (M)FLOPs (G)Acc. (%)
PiT-XS (Heo et al., 2021)10.61.478.1
ConT-S(Yan et al.,2021)10.11.576.5
PVT-T(Wang et al.,2021)13.21.975.1
ConViT-Ti+ (d'Ascoli et al., 2021)10.02.076.7
Twins-SVT-S (Chu et al.,2021a)24.02.881.7
PiT-S (Heo et al., 2021)23.52.980.9
ConT-M(Yan et al., 2021)19.23.180.2
Twins-PCPVT-S (Chu et al., 2021a)24.13.781.2
PVT-S (Wang et al., 2021)24.53.879.8
RegionViT-Ti13.82.480.4
RegionViT-Ti+14.32.781.5
DeiT-S (Touvron et al., 2020)22.14.679.9
CvT-13 (Wu et al.,2021)20.04.581.6
Swin-T (Liu et al., 2021)29.04.581.3
LocalViT-S (Li et al.,2021)22.44.680.8
ViL-S (Zhang et al., 2021a)24.64.982.0
Visformer-S (Chen et al.,2021b)40.24.982.3
ConViT-S (d'Ascoli et al.,2021)27.05.481.3
NesT-S (Zhang et al., 2021b)17.05.881.5
RegionViT-S30.65.382.6
RegionViT-S+31.35.783.3
", + "type": "table", + "image_path": "0b6d088036456b2a4c78444f0295f127489f30e37f3c7eae0c15593287ace348.jpg" + } + ] + } + ], + "index": 43.5, + "virtual_lines": [ + { + "bbox": [ + 111, + 579, + 300, + 591.5 + ], + "spans": [], + "index": 38 + }, + { + "bbox": [ + 111, + 591.5, + 300, + 604.0 + ], + "spans": [], + "index": 39 + }, + { + "bbox": [ + 111, + 604.0, + 300, + 616.5 + ], + "spans": [], + "index": 40 + }, + { + "bbox": [ + 111, + 616.5, + 300, + 629.0 + ], + "spans": [], + "index": 41 + }, + { + "bbox": [ + 111, + 629.0, + 300, + 641.5 + ], + "spans": [], + "index": 42 + }, + { + "bbox": [ + 111, + 641.5, + 300, + 654.0 + ], + "spans": [], + "index": 43 + }, + { + "bbox": [ + 111, + 654.0, + 300, + 666.5 + ], + "spans": [], + "index": 44 + }, + { + "bbox": [ + 111, + 666.5, + 300, + 679.0 + ], + "spans": [], + "index": 45 + }, + { + "bbox": [ + 111, + 679.0, + 300, + 691.5 + ], + "spans": [], + "index": 46 + }, + { + "bbox": [ + 111, + 691.5, + 300, + 704.0 + ], + "spans": [], + "index": 47 + }, + { + "bbox": [ + 111, + 704.0, + 300, + 716.5 + ], + "spans": [], + "index": 48 + }, + { + "bbox": [ + 111, + 716.5, + 300, + 729.0 + ], + "spans": [], + "index": 49 + } + ] + } + ], + "index": 40.0 + }, + { + "type": "table", + "bbox": [ + 311, + 579, + 499, + 729 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 311, + 579, + 499, + 729 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 311, + 579, + 499, + 729 + ], + "spans": [ + { + "bbox": [ + 311, + 579, + 499, + 729 + ], + "score": 0.97, + "html": "
ModelParams (M)FLOPs (G)Acc. (%)
ConT-M(Yan et al., 2021)39.66.481.8
Twins-PCPVT-B (Chu et al.,2021a)43.86.482.7
PVT-M(Wang et al., 2021)44.26.781.2
CvT-21(Wu et al.,2021)32.07.182.5
Twins-SVT-B (Chu et al., 2021a)568.383.2
ViL-M (Zhang et al.,2021a)39.78.783.3
Swin-S (Liu et al., 2021)50.08.783.0
PVT-L (Wang et al.,2021)61.49.881.7
ConViT-S+ (d'Ascoli et al., 2021)48.010.082.2
NesT-S (Zhang et al., 2021b)38.010.483.3
RegionViT-M41.27.483.1
RegionViT-M+42.07.983.4
DeiT-B (Touvron et al.,2020)86.617.681.8
PiT-B (Heo et al., 2021)73.812.582.0
ViL-B (Zhang et al., 2021a)55.713.483.2
Twins-SVT-L(Chu et al.,2021a)99.214.883.7
Swin-B (Liu et al.,2021)88.015.483.5
ConViT-B (d'Ascoli et al.,2021)86.017.082.4
NesT-B (Zhang et al., 2021b)68.017.983.8
RegionViT-B72.713.083.2
RegionViT-B+73.813.683.8
", + "type": "table", + "image_path": "30316ee497884d718c99f74f5a3484680fc2081650e7990719ee5b51f13908ad.jpg" + } + ] + } + ], + "index": 55.5, + "virtual_lines": [ + { + "bbox": [ + 311, + 579, + 499, + 591.5 + ], + "spans": [], + "index": 50 + }, + { + "bbox": [ + 311, + 591.5, + 499, + 604.0 + ], + "spans": [], + "index": 51 + }, + { + "bbox": [ + 311, + 604.0, + 499, + 616.5 + ], + "spans": [], + "index": 52 + }, + { + "bbox": [ + 311, + 616.5, + 499, + 629.0 + ], + "spans": [], + "index": 53 + }, + { + "bbox": [ + 311, + 629.0, + 499, + 641.5 + ], + "spans": [], + "index": 54 + }, + { + "bbox": [ + 311, + 641.5, + 499, + 654.0 + ], + "spans": [], + "index": 55 + }, + { + "bbox": [ + 311, + 654.0, + 499, + 666.5 + ], + "spans": [], + "index": 56 + }, + { + "bbox": [ + 311, + 666.5, + 499, + 679.0 + ], + "spans": [], + "index": 57 + }, + { + "bbox": [ + 311, + 679.0, + 499, + 691.5 + ], + "spans": [], + "index": 58 + }, + { + "bbox": [ + 311, + 691.5, + 499, + 704.0 + ], + "spans": [], + "index": 59 + }, + { + "bbox": [ + 311, + 704.0, + 499, + 716.5 + ], + "spans": [], + "index": 60 + }, + { + "bbox": [ + 311, + 716.5, + 499, + 729.0 + ], + "spans": [], + "index": 61 + } + ] + } + ], + "index": 55.5 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 110, + 103, + 271, + 152 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 195, + 81, + 415, + 91 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 194, + 79, + 417, + 93 + ], + "spans": [ + { + "bbox": [ + 194, + 79, + 417, + 93 + ], + "score": 1.0, + "content": "Table 4: Results on IN1K with IN21K and transfer learning.", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table_body", + "bbox": [ + 110, + 103, + 271, + 152 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 110, + 103, + 271, + 152 + ], + "spans": [ + { + "bbox": [ + 110, + 103, + 271, + 152 + ], + "score": 0.969, + "html": "
ModelParams (M)FLOPs (G)Acc. (%)
ViL-B55.743.786.0
Swin-L197.0103.987.3
CvT-W24277193.287.7
RegionViT-B+76.542.686.5
", + "type": "table", + "image_path": "b9a68b9626f4e665e8c0b30bb2575152e0cc4b84f98388a4325bb1e0f0c3ffdd.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 110, + 103, + 271, + 119.33333333333333 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 110, + 119.33333333333333, + 271, + 135.66666666666666 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 110, + 135.66666666666666, + 271, + 152.0 + ], + "spans": [], + "index": 6 + } + ] + } + ], + "index": 2.0 + }, + { + "type": "table", + "bbox": [ + 280, + 105, + 501, + 150 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 313, + 94, + 468, + 104 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 313, + 93, + 469, + 105 + ], + "spans": [ + { + "bbox": [ + 313, + 93, + 469, + 105 + ], + "score": 1.0, + "content": "(b) Transfer learning on downstream tasks.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1 + }, + { + "type": "table_body", + "bbox": [ + 280, + 105, + 501, + 150 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 280, + 105, + 501, + 150 + ], + "spans": [ + { + "bbox": [ + 280, + 105, + 501, + 150 + ], + "score": 0.939, + "html": "
CIFAR10CIFAR100PetStanfordCarsChestXRay8
DeiT-S99.190.994.991.555.4
DeiT-B99.190.894.491.755.8
RegionViT-S98.990.095.392.857.8
RegionViT-M99.090.895.591.958.3
", + "type": "table", + "image_path": "70abcc8e23c30803312873285cb17d3453dd900952470700ec0f0f7b3495387d.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 280, + 105, + 501, + 120.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 280, + 120.0, + 501, + 135.0 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 280, + 135.0, + 501, + 150.0 + ], + "spans": [], + "index": 7 + } + ] + } + ], + "index": 3.0 + }, + { + "type": "table", + "bbox": [ + 125, + 187, + 485, + 392 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 105, + 165, + 503, + 186 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 164, + 505, + 177 + ], + "spans": [ + { + "bbox": [ + 105, + 164, + 358, + 177 + ], + "score": 1.0, + "content": "Table 5: Object detection performance on MS COCO val2017 with", + "type": "text" + }, + { + "bbox": [ + 359, + 166, + 371, + 175 + ], + "score": 0.85, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 372, + 164, + 388, + 177 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 388, + 166, + 401, + 175 + ], + "score": 0.87, + "content": "3 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 402, + 164, + 505, + 177 + ], + "score": 1.0, + "content": "schedule. The bold number", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 174, + 469, + 186 + ], + "spans": [ + { + "bbox": [ + 105, + 174, + 363, + 186 + ], + "score": 1.0, + "content": "indicates the best number within the section, and for MaskRCNN, both", + "type": "text" + }, + { + "bbox": [ + 364, + 175, + 380, + 185 + ], + "score": 0.85, + "content": "\\mathsf { A P } ^ { b }", + "type": "inline_equation" + }, + { + "bbox": [ + 380, + 174, + 396, + 186 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 397, + 176, + 416, + 185 + ], + "score": 0.82, + "content": "\\mathsf { A P } ^ { m }", + "type": "inline_equation" + }, + { + "bbox": [ + 416, + 174, + 469, + 186 + ], + "score": 1.0, + "content": "are annotated.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8.5 + }, + { + "type": "table_body", + "bbox": [ + 125, + 187, + 485, + 392 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 125, + 187, + 485, + 392 + ], + "spans": [ + { + "bbox": [ + 125, + 187, + 485, + 392 + ], + "score": 0.984, + "html": "
BackboneParamsFLOPsRetineNet3× APParams (M)FLOPsMaskRCNN
(M)1× AP(G)1× APb |APmAPb3× APm
ResNet5037.723436.339.044.2260.038.034.441.0
ConT-M (Yan et al.,2021)27.0217.239.311140.538.1137.1 1
PVT-S (Wang et al., 2021)34.240.442.244.140.437.843.039.6
ViL-S (Zhang et al., 2021a)35.7252.241.642.945.0174.341.838.543.439.6
Swin-T (Liu et al., 2021)38.524541.543.947.826442.239.146.041.6
Twins-SVT-S (w/PEG) (Chu et al.,2021a)34.320943.045.644.022843.540.346.842.6
RegionViT-S40.8192.642.245.850.1171.342.539.546.342.3
RegionViT-S+41.5204.243.146.950.9182.943.540.447.343.4
RegionViT-S+ w/PEG41.6204.343.946.750.9183.044.240.847.643.4
ResNet10156.731538.540.963.233640.436.442.838.5
ResNeXt101-32x4d56.4319.139.941.462.834041.937.544.039.2
PVT-M(Wang et al., 2021)53.941.943.263.942.039.044.240.5
ViL-M (Zhang et al.,2021a)50.8338.942.943.760.1261.143.439.744.640.7
Swin-S (Liu et al., 2021)59.833544.546.369.135444.840.947.642.8
Twins-SVT-B (w/PEG) (Chu et al., 2021a)67.032245.346.976.334045.241.548.043.0
RegionViT-B83.4308.943.346.192.2287.943.540.147.243.0
RegionViT-B+84.4328.144.246.993.2307.144.541.048.143.5
RegionViT-B+ w/PEG84.5328.244.646.993.2307.245.441.648.343.5
ResNeXt101-64x4d95.547341.01101.949342.838.4
PVT-L (Wang et al., 2021)71.134542.681.036442.939.511
ViL-B (Zhang et al., 2021a)66.7443.044.344.776.1365.145.141.045.741.3
Swin-B (Liu et al.,2021)98.447744.71107.249645.541.3二
Twins-SVT-L (w/ PEG) (Chu et al.,2021a)110.945545.7119.747445.941.61
RegionViT-B+ w/PEGt84.5506.446.148.293.2464.446.342.449.244.5
", + "type": "table", + "image_path": "66777bf52b14fa40310ff41b4f4bfebc777136622d1b5eefa4be49593243ef6c.jpg" + } + ] + } + ], + "index": 11, + "virtual_lines": [ + { + "bbox": [ + 125, + 187, + 485, + 255.33333333333331 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 125, + 255.33333333333331, + 485, + 323.66666666666663 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 125, + 323.66666666666663, + 485, + 391.99999999999994 + ], + "spans": [], + "index": 12 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 127, + 393, + 479, + 400 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 130, + 392, + 478, + 401 + ], + "spans": [ + { + "bbox": [ + 130, + 392, + 452, + 401 + ], + "score": 1.0, + "content": "The reported results of Swin are from Twins as the original paper does not include resutls with ImageNet1K weights. †: input resolution is", + "type": "text" + }, + { + "bbox": [ + 453, + 393, + 478, + 400 + ], + "score": 0.48, + "content": "\\overline { { 8 9 6 \\times 1 3 4 4 } }", + "type": "inline_equation" + } + ], + "index": 13 + } + ], + "index": 13 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 106, + 423, + 505, + 501 + ], + "lines": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "score": 1.0, + "content": "Results on ImageNet1K. Table 3 shows the results on ImageNet1K where the methods listed all", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 434, + 506, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 447 + ], + "score": 1.0, + "content": "adopt CNN-like pyramid structure into the ViT and are all very recent and concurrent works. For", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 445, + 505, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 445, + 505, + 459 + ], + "score": 1.0, + "content": "the smaller models (Ti, S), RegionViT achieves a better trade-off between accuracy and complexity", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 456, + 506, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 506, + 470 + ], + "score": 1.0, + "content": "(parameters or FLOPs); on the other hand, for the larger models (M and B), it obtains better accuracy", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 467, + 506, + 481 + ], + "spans": [ + { + "bbox": [ + 105, + 467, + 506, + 481 + ], + "score": 1.0, + "content": "while having fewer FLOPs and parameters. The efficiency of RegionViT comes from the proposed", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 478, + 505, + 491 + ], + "spans": [ + { + "bbox": [ + 105, + 478, + 505, + 491 + ], + "score": 1.0, + "content": "R2L transformer encoder, and hence with such efficiency, it enables the network to be wide and deep", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 488, + 302, + 503 + ], + "spans": [ + { + "bbox": [ + 105, + 488, + 302, + 503 + ], + "score": 1.0, + "content": "for better accuracy with comparable complexity.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 17 + }, + { + "type": "text", + "bbox": [ + 107, + 506, + 504, + 539 + ], + "lines": [ + { + "bbox": [ + 105, + 505, + 505, + 520 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 505, + 520 + ], + "score": 1.0, + "content": "Results on ImageNet21K. Table 4a shows that our model achieves a good accuracy-efficiency", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 517, + 505, + 529 + ], + "spans": [ + { + "bbox": [ + 105, + 517, + 330, + 529 + ], + "score": 1.0, + "content": "tradeoff with ImageNet21K for pretraining. RegionViT-", + "type": "text" + }, + { + "bbox": [ + 330, + 518, + 344, + 528 + ], + "score": 0.81, + "content": "\\mathbf { \\cdot B + }", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 517, + 390, + 529 + ], + "score": 1.0, + "content": "only needs", + "type": "text" + }, + { + "bbox": [ + 390, + 518, + 410, + 528 + ], + "score": 0.86, + "content": "2 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 410, + 517, + 505, + 529 + ], + "score": 1.0, + "content": "parameters and FLOPs", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 528, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 506, + 540 + ], + "score": 1.0, + "content": "compared to CvT-W24 (Wu et al., 2021) and half parameters and FLOPs to Swin-L (Liu et al., 2021).", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 107, + 545, + 505, + 600 + ], + "lines": [ + { + "bbox": [ + 106, + 545, + 505, + 557 + ], + "spans": [ + { + "bbox": [ + 106, + 545, + 505, + 557 + ], + "score": 1.0, + "content": "Results on Transfer Learning. Table 4b shows the transfer learning results with ImageNet1K", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 555, + 506, + 569 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 408, + 569 + ], + "score": 1.0, + "content": "pretrained weights. Our model outperforms DeiT (Touvron et al., 2020) by", + "type": "text" + }, + { + "bbox": [ + 409, + 556, + 437, + 567 + ], + "score": 0.88, + "content": "2 \\sim 3 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 437, + 555, + 506, + 569 + ], + "score": 1.0, + "content": "on ChestXRay8,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 567, + 506, + 580 + ], + "spans": [ + { + "bbox": [ + 105, + 567, + 506, + 580 + ], + "score": 1.0, + "content": "which has a larger domain gap from ImageNet1K than other four datasets. We think this is because", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 577, + 505, + 591 + ], + "spans": [ + { + "bbox": [ + 106, + 577, + 505, + 591 + ], + "score": 1.0, + "content": "the hierarchical feature models could provide better generalization for the domain gap compared to", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 589, + 409, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 589, + 409, + 601 + ], + "score": 1.0, + "content": "DeiT which uses isotropic spatial resolution throughout the whole network.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 26 + }, + { + "type": "title", + "bbox": [ + 107, + 616, + 358, + 627 + ], + "lines": [ + { + "bbox": [ + 106, + 616, + 360, + 629 + ], + "spans": [ + { + "bbox": [ + 106, + 616, + 360, + 629 + ], + "score": 1.0, + "content": "4.2 OBJECT DETECTION AND SEMANTIC SEGMENTATION", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 637, + 505, + 693 + ], + "lines": [ + { + "bbox": [ + 105, + 638, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 105, + 638, + 505, + 650 + ], + "score": 1.0, + "content": "Dataset. We use MS COCO 2017 to validate our models (Lin et al., 2014) on object detection. COCO", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 648, + 506, + 661 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 506, + 661 + ], + "score": 1.0, + "content": "2017 dataset contains 118K images for training and 5K images for validation across 80 categories.", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 659, + 506, + 671 + ], + "spans": [ + { + "bbox": [ + 106, + 659, + 506, + 671 + ], + "score": 1.0, + "content": "We also test our model on keypoint detection and results are shown in Sec. B. On the other hand,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 671, + 506, + 682 + ], + "spans": [ + { + "bbox": [ + 105, + 671, + 506, + 682 + ], + "score": 1.0, + "content": "we validate our model with semantic segmentation on ADE20K (Zhou et al., 2017). The ADE20K", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 682, + 396, + 694 + ], + "spans": [ + { + "bbox": [ + 106, + 682, + 172, + 694 + ], + "score": 1.0, + "content": "dataset contains", + "type": "text" + }, + { + "bbox": [ + 172, + 682, + 189, + 692 + ], + "score": 0.5, + "content": "2 0 \\mathrm { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 682, + 396, + 694 + ], + "score": 1.0, + "content": "images in the training set, 2k images for validation.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 32 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 698, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 698, + 506, + 711 + ], + "score": 1.0, + "content": "Training and Evaluation. For object detection, we adopt RetinaNet (Lin et al., 2017) and MaskR-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "CNN (He et al., 2017) as our detection framework. We simply replace the backbone with RegionViT,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 720, + 505, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 505, + 732 + ], + "score": 1.0, + "content": "and then output the local tokens at each stages as multi-scale features for the detector. The regional", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36 + } + ], + "page_idx": 6, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 105, + 25, + 294, + 39 + ], + "spans": [ + { + "bbox": [ + 105, + 25, + 294, + 39 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 303, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 13, + "width": 7 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 110, + 103, + 271, + 152 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 195, + 81, + 415, + 91 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 194, + 79, + 417, + 93 + ], + "spans": [ + { + "bbox": [ + 194, + 79, + 417, + 93 + ], + "score": 1.0, + "content": "Table 4: Results on IN1K with IN21K and transfer learning.", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table_body", + "bbox": [ + 110, + 103, + 271, + 152 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 110, + 103, + 271, + 152 + ], + "spans": [ + { + "bbox": [ + 110, + 103, + 271, + 152 + ], + "score": 0.969, + "html": "
ModelParams (M)FLOPs (G)Acc. (%)
ViL-B55.743.786.0
Swin-L197.0103.987.3
CvT-W24277193.287.7
RegionViT-B+76.542.686.5
", + "type": "table", + "image_path": "b9a68b9626f4e665e8c0b30bb2575152e0cc4b84f98388a4325bb1e0f0c3ffdd.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 110, + 103, + 271, + 119.33333333333333 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 110, + 119.33333333333333, + 271, + 135.66666666666666 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 110, + 135.66666666666666, + 271, + 152.0 + ], + "spans": [], + "index": 6 + } + ] + } + ], + "index": 2.0 + }, + { + "type": "table", + "bbox": [ + 280, + 105, + 501, + 150 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 313, + 94, + 468, + 104 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 313, + 93, + 469, + 105 + ], + "spans": [ + { + "bbox": [ + 313, + 93, + 469, + 105 + ], + "score": 1.0, + "content": "(b) Transfer learning on downstream tasks.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1 + }, + { + "type": "table_body", + "bbox": [ + 280, + 105, + 501, + 150 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 280, + 105, + 501, + 150 + ], + "spans": [ + { + "bbox": [ + 280, + 105, + 501, + 150 + ], + "score": 0.939, + "html": "
CIFAR10CIFAR100PetStanfordCarsChestXRay8
DeiT-S99.190.994.991.555.4
DeiT-B99.190.894.491.755.8
RegionViT-S98.990.095.392.857.8
RegionViT-M99.090.895.591.958.3
", + "type": "table", + "image_path": "70abcc8e23c30803312873285cb17d3453dd900952470700ec0f0f7b3495387d.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 280, + 105, + 501, + 120.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 280, + 120.0, + 501, + 135.0 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 280, + 135.0, + 501, + 150.0 + ], + "spans": [], + "index": 7 + } + ] + } + ], + "index": 3.0 + }, + { + "type": "table", + "bbox": [ + 125, + 187, + 485, + 392 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 105, + 165, + 503, + 186 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 164, + 505, + 177 + ], + "spans": [ + { + "bbox": [ + 105, + 164, + 358, + 177 + ], + "score": 1.0, + "content": "Table 5: Object detection performance on MS COCO val2017 with", + "type": "text" + }, + { + "bbox": [ + 359, + 166, + 371, + 175 + ], + "score": 0.85, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 372, + 164, + 388, + 177 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 388, + 166, + 401, + 175 + ], + "score": 0.87, + "content": "3 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 402, + 164, + 505, + 177 + ], + "score": 1.0, + "content": "schedule. The bold number", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 174, + 469, + 186 + ], + "spans": [ + { + "bbox": [ + 105, + 174, + 363, + 186 + ], + "score": 1.0, + "content": "indicates the best number within the section, and for MaskRCNN, both", + "type": "text" + }, + { + "bbox": [ + 364, + 175, + 380, + 185 + ], + "score": 0.85, + "content": "\\mathsf { A P } ^ { b }", + "type": "inline_equation" + }, + { + "bbox": [ + 380, + 174, + 396, + 186 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 397, + 176, + 416, + 185 + ], + "score": 0.82, + "content": "\\mathsf { A P } ^ { m }", + "type": "inline_equation" + }, + { + "bbox": [ + 416, + 174, + 469, + 186 + ], + "score": 1.0, + "content": "are annotated.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8.5 + }, + { + "type": "table_body", + "bbox": [ + 125, + 187, + 485, + 392 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 125, + 187, + 485, + 392 + ], + "spans": [ + { + "bbox": [ + 125, + 187, + 485, + 392 + ], + "score": 0.984, + "html": "
BackboneParamsFLOPsRetineNet3× APParams (M)FLOPsMaskRCNN
(M)1× AP(G)1× APb |APmAPb3× APm
ResNet5037.723436.339.044.2260.038.034.441.0
ConT-M (Yan et al.,2021)27.0217.239.311140.538.1137.1 1
PVT-S (Wang et al., 2021)34.240.442.244.140.437.843.039.6
ViL-S (Zhang et al., 2021a)35.7252.241.642.945.0174.341.838.543.439.6
Swin-T (Liu et al., 2021)38.524541.543.947.826442.239.146.041.6
Twins-SVT-S (w/PEG) (Chu et al.,2021a)34.320943.045.644.022843.540.346.842.6
RegionViT-S40.8192.642.245.850.1171.342.539.546.342.3
RegionViT-S+41.5204.243.146.950.9182.943.540.447.343.4
RegionViT-S+ w/PEG41.6204.343.946.750.9183.044.240.847.643.4
ResNet10156.731538.540.963.233640.436.442.838.5
ResNeXt101-32x4d56.4319.139.941.462.834041.937.544.039.2
PVT-M(Wang et al., 2021)53.941.943.263.942.039.044.240.5
ViL-M (Zhang et al.,2021a)50.8338.942.943.760.1261.143.439.744.640.7
Swin-S (Liu et al., 2021)59.833544.546.369.135444.840.947.642.8
Twins-SVT-B (w/PEG) (Chu et al., 2021a)67.032245.346.976.334045.241.548.043.0
RegionViT-B83.4308.943.346.192.2287.943.540.147.243.0
RegionViT-B+84.4328.144.246.993.2307.144.541.048.143.5
RegionViT-B+ w/PEG84.5328.244.646.993.2307.245.441.648.343.5
ResNeXt101-64x4d95.547341.01101.949342.838.4
PVT-L (Wang et al., 2021)71.134542.681.036442.939.511
ViL-B (Zhang et al., 2021a)66.7443.044.344.776.1365.145.141.045.741.3
Swin-B (Liu et al.,2021)98.447744.71107.249645.541.3二
Twins-SVT-L (w/ PEG) (Chu et al.,2021a)110.945545.7119.747445.941.61
RegionViT-B+ w/PEGt84.5506.446.148.293.2464.446.342.449.244.5
", + "type": "table", + "image_path": "66777bf52b14fa40310ff41b4f4bfebc777136622d1b5eefa4be49593243ef6c.jpg" + } + ] + } + ], + "index": 11, + "virtual_lines": [ + { + "bbox": [ + 125, + 187, + 485, + 255.33333333333331 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 125, + 255.33333333333331, + 485, + 323.66666666666663 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 125, + 323.66666666666663, + 485, + 391.99999999999994 + ], + "spans": [], + "index": 12 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 127, + 393, + 479, + 400 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 130, + 392, + 478, + 401 + ], + "spans": [ + { + "bbox": [ + 130, + 392, + 452, + 401 + ], + "score": 1.0, + "content": "The reported results of Swin are from Twins as the original paper does not include resutls with ImageNet1K weights. †: input resolution is", + "type": "text" + }, + { + "bbox": [ + 453, + 393, + 478, + 400 + ], + "score": 0.48, + "content": "\\overline { { 8 9 6 \\times 1 3 4 4 } }", + "type": "inline_equation" + } + ], + "index": 13 + } + ], + "index": 13 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 106, + 423, + 505, + 501 + ], + "lines": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "score": 1.0, + "content": "Results on ImageNet1K. Table 3 shows the results on ImageNet1K where the methods listed all", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 434, + 506, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 447 + ], + "score": 1.0, + "content": "adopt CNN-like pyramid structure into the ViT and are all very recent and concurrent works. For", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 445, + 505, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 445, + 505, + 459 + ], + "score": 1.0, + "content": "the smaller models (Ti, S), RegionViT achieves a better trade-off between accuracy and complexity", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 456, + 506, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 506, + 470 + ], + "score": 1.0, + "content": "(parameters or FLOPs); on the other hand, for the larger models (M and B), it obtains better accuracy", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 467, + 506, + 481 + ], + "spans": [ + { + "bbox": [ + 105, + 467, + 506, + 481 + ], + "score": 1.0, + "content": "while having fewer FLOPs and parameters. The efficiency of RegionViT comes from the proposed", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 478, + 505, + 491 + ], + "spans": [ + { + "bbox": [ + 105, + 478, + 505, + 491 + ], + "score": 1.0, + "content": "R2L transformer encoder, and hence with such efficiency, it enables the network to be wide and deep", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 488, + 302, + 503 + ], + "spans": [ + { + "bbox": [ + 105, + 488, + 302, + 503 + ], + "score": 1.0, + "content": "for better accuracy with comparable complexity.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 17, + "bbox_fs": [ + 105, + 424, + 506, + 503 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 506, + 504, + 539 + ], + "lines": [ + { + "bbox": [ + 105, + 505, + 505, + 520 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 505, + 520 + ], + "score": 1.0, + "content": "Results on ImageNet21K. Table 4a shows that our model achieves a good accuracy-efficiency", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 517, + 505, + 529 + ], + "spans": [ + { + "bbox": [ + 105, + 517, + 330, + 529 + ], + "score": 1.0, + "content": "tradeoff with ImageNet21K for pretraining. RegionViT-", + "type": "text" + }, + { + "bbox": [ + 330, + 518, + 344, + 528 + ], + "score": 0.81, + "content": "\\mathbf { \\cdot B + }", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 517, + 390, + 529 + ], + "score": 1.0, + "content": "only needs", + "type": "text" + }, + { + "bbox": [ + 390, + 518, + 410, + 528 + ], + "score": 0.86, + "content": "2 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 410, + 517, + 505, + 529 + ], + "score": 1.0, + "content": "parameters and FLOPs", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 528, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 506, + 540 + ], + "score": 1.0, + "content": "compared to CvT-W24 (Wu et al., 2021) and half parameters and FLOPs to Swin-L (Liu et al., 2021).", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22, + "bbox_fs": [ + 105, + 505, + 506, + 540 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 545, + 505, + 600 + ], + "lines": [ + { + "bbox": [ + 106, + 545, + 505, + 557 + ], + "spans": [ + { + "bbox": [ + 106, + 545, + 505, + 557 + ], + "score": 1.0, + "content": "Results on Transfer Learning. Table 4b shows the transfer learning results with ImageNet1K", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 555, + 506, + 569 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 408, + 569 + ], + "score": 1.0, + "content": "pretrained weights. Our model outperforms DeiT (Touvron et al., 2020) by", + "type": "text" + }, + { + "bbox": [ + 409, + 556, + 437, + 567 + ], + "score": 0.88, + "content": "2 \\sim 3 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 437, + 555, + 506, + 569 + ], + "score": 1.0, + "content": "on ChestXRay8,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 567, + 506, + 580 + ], + "spans": [ + { + "bbox": [ + 105, + 567, + 506, + 580 + ], + "score": 1.0, + "content": "which has a larger domain gap from ImageNet1K than other four datasets. We think this is because", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 577, + 505, + 591 + ], + "spans": [ + { + "bbox": [ + 106, + 577, + 505, + 591 + ], + "score": 1.0, + "content": "the hierarchical feature models could provide better generalization for the domain gap compared to", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 589, + 409, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 589, + 409, + 601 + ], + "score": 1.0, + "content": "DeiT which uses isotropic spatial resolution throughout the whole network.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 26, + "bbox_fs": [ + 105, + 545, + 506, + 601 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 616, + 358, + 627 + ], + "lines": [ + { + "bbox": [ + 106, + 616, + 360, + 629 + ], + "spans": [ + { + "bbox": [ + 106, + 616, + 360, + 629 + ], + "score": 1.0, + "content": "4.2 OBJECT DETECTION AND SEMANTIC SEGMENTATION", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 637, + 505, + 693 + ], + "lines": [ + { + "bbox": [ + 105, + 638, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 105, + 638, + 505, + 650 + ], + "score": 1.0, + "content": "Dataset. We use MS COCO 2017 to validate our models (Lin et al., 2014) on object detection. COCO", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 648, + 506, + 661 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 506, + 661 + ], + "score": 1.0, + "content": "2017 dataset contains 118K images for training and 5K images for validation across 80 categories.", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 659, + 506, + 671 + ], + "spans": [ + { + "bbox": [ + 106, + 659, + 506, + 671 + ], + "score": 1.0, + "content": "We also test our model on keypoint detection and results are shown in Sec. B. On the other hand,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 671, + 506, + 682 + ], + "spans": [ + { + "bbox": [ + 105, + 671, + 506, + 682 + ], + "score": 1.0, + "content": "we validate our model with semantic segmentation on ADE20K (Zhou et al., 2017). The ADE20K", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 682, + 396, + 694 + ], + "spans": [ + { + "bbox": [ + 106, + 682, + 172, + 694 + ], + "score": 1.0, + "content": "dataset contains", + "type": "text" + }, + { + "bbox": [ + 172, + 682, + 189, + 692 + ], + "score": 0.5, + "content": "2 0 \\mathrm { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 682, + 396, + 694 + ], + "score": 1.0, + "content": "images in the training set, 2k images for validation.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 32, + "bbox_fs": [ + 105, + 638, + 506, + 694 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 698, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 698, + 506, + 711 + ], + "score": 1.0, + "content": "Training and Evaluation. For object detection, we adopt RetinaNet (Lin et al., 2017) and MaskR-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "CNN (He et al., 2017) as our detection framework. We simply replace the backbone with RegionViT,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 720, + 505, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 505, + 732 + ], + "score": 1.0, + "content": "and then output the local tokens at each stages as multi-scale features for the detector. The regional", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 82, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 505, + 93 + ], + "score": 1.0, + "content": "tokens are not used in the detection. Before sending the features to the detector framework, we add", + "type": "text", + "cross_page": true + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 93, + 505, + 105 + ], + "spans": [ + { + "bbox": [ + 105, + 93, + 505, + 105 + ], + "score": 1.0, + "content": "new layer normalization layers to normalize the features. We use RegionViT-S and RegionViT-B", + "type": "text", + "cross_page": true + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 505, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 505, + 117 + ], + "score": 1.0, + "content": "as the backbone and the weights are initialized from ImageNet1K pretraining. The shorter side and", + "type": "text", + "cross_page": true + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 115, + 505, + 127 + ], + "spans": [ + { + "bbox": [ + 106, + 115, + 505, + 127 + ], + "score": 1.0, + "content": "longer side of the input image is resized to 672 and 1,120, respectively. We train our models based", + "type": "text", + "cross_page": true + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 126, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 106, + 126, + 178, + 138 + ], + "score": 1.0, + "content": "on the settings in", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 179, + 127, + 193, + 137 + ], + "score": 0.86, + "content": "1 \\times", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 193, + 126, + 211, + 138 + ], + "score": 1.0, + "content": "and", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 211, + 127, + 226, + 137 + ], + "score": 0.87, + "content": "3 \\times", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 226, + 126, + 505, + 138 + ], + "score": 1.0, + "content": "schedule in Detectron2 (Wu et al., 2019) for object detection. More", + "type": "text", + "cross_page": true + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "spans": [ + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "score": 1.0, + "content": "training details could be found in A.2. For semantic segmentation, we adopt Semantic FPN as the", + "type": "text", + "cross_page": true + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 148, + 505, + 160 + ], + "spans": [ + { + "bbox": [ + 106, + 148, + 505, + 160 + ], + "score": 1.0, + "content": "framework (Kirillov et al., 2019) and use RegionViT as the backbone. We initialize models with", + "type": "text", + "cross_page": true + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 159, + 505, + 171 + ], + "spans": [ + { + "bbox": [ + 106, + 159, + 505, + 171 + ], + "score": 1.0, + "content": "ImageNet1K weights, and mostly follow the training receipt in Twins’ paper (Chu et al., 2021a).", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 170, + 295, + 182 + ], + "spans": [ + { + "bbox": [ + 106, + 170, + 295, + 182 + ], + "score": 1.0, + "content": "More training details can be found in Sec. A.3.", + "type": "text", + "cross_page": true + } + ], + "index": 8 + } + ], + "index": 36, + "bbox_fs": [ + 106, + 698, + 506, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 181 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 505, + 93 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 505, + 93 + ], + "score": 1.0, + "content": "tokens are not used in the detection. Before sending the features to the detector framework, we add", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 93, + 505, + 105 + ], + "spans": [ + { + "bbox": [ + 105, + 93, + 505, + 105 + ], + "score": 1.0, + "content": "new layer normalization layers to normalize the features. We use RegionViT-S and RegionViT-B", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 505, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 505, + 117 + ], + "score": 1.0, + "content": "as the backbone and the weights are initialized from ImageNet1K pretraining. The shorter side and", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 115, + 505, + 127 + ], + "spans": [ + { + "bbox": [ + 106, + 115, + 505, + 127 + ], + "score": 1.0, + "content": "longer side of the input image is resized to 672 and 1,120, respectively. We train our models based", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 126, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 106, + 126, + 178, + 138 + ], + "score": 1.0, + "content": "on the settings in", + "type": "text" + }, + { + "bbox": [ + 179, + 127, + 193, + 137 + ], + "score": 0.86, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 193, + 126, + 211, + 138 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 211, + 127, + 226, + 137 + ], + "score": 0.87, + "content": "3 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 226, + 126, + 505, + 138 + ], + "score": 1.0, + "content": "schedule in Detectron2 (Wu et al., 2019) for object detection. More", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "spans": [ + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "score": 1.0, + "content": "training details could be found in A.2. For semantic segmentation, we adopt Semantic FPN as the", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 148, + 505, + 160 + ], + "spans": [ + { + "bbox": [ + 106, + 148, + 505, + 160 + ], + "score": 1.0, + "content": "framework (Kirillov et al., 2019) and use RegionViT as the backbone. We initialize models with", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 159, + 505, + 171 + ], + "spans": [ + { + "bbox": [ + 106, + 159, + 505, + 171 + ], + "score": 1.0, + "content": "ImageNet1K weights, and mostly follow the training receipt in Twins’ paper (Chu et al., 2021a).", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 170, + 295, + 182 + ], + "spans": [ + { + "bbox": [ + 106, + 170, + 295, + 182 + ], + "score": 1.0, + "content": "More training details can be found in Sec. A.3.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 107, + 187, + 505, + 286 + ], + "lines": [ + { + "bbox": [ + 106, + 187, + 506, + 199 + ], + "spans": [ + { + "bbox": [ + 106, + 187, + 506, + 199 + ], + "score": 1.0, + "content": "Results on Object Detection. Table 5 compares RegionViT with other methods. For the smaller", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 198, + 505, + 210 + ], + "spans": [ + { + "bbox": [ + 106, + 198, + 256, + 210 + ], + "score": 1.0, + "content": "models, RegionViT-S and RegionViT-", + "type": "text" + }, + { + "bbox": [ + 257, + 199, + 269, + 209 + ], + "score": 0.84, + "content": "^ { S + }", + "type": "inline_equation" + }, + { + "bbox": [ + 269, + 198, + 505, + 210 + ], + "score": 1.0, + "content": "achieve similar accuracy while being moderately efficient in", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 209, + 506, + 222 + ], + "spans": [ + { + "bbox": [ + 105, + 209, + 506, + 222 + ], + "score": 1.0, + "content": "terms of FLOPs. We find that the position encoding generator (PEG) proposed by (Chu et al., 2021b)", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 219, + 505, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 219, + 505, + 233 + ], + "score": 1.0, + "content": "could help the detection accuracy substantially, so we also provide the results with PEG, which are", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "spans": [ + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "score": 1.0, + "content": "also used in Twins (Chu et al., 2021a). For the middle size models with RetinaNet, RegionViT-B are", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 242, + 505, + 255 + ], + "spans": [ + { + "bbox": [ + 105, + 242, + 233, + 255 + ], + "score": 1.0, + "content": "slightly worse than Twins with", + "type": "text" + }, + { + "bbox": [ + 233, + 243, + 247, + 253 + ], + "score": 0.86, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 248, + 242, + 505, + 255 + ], + "score": 1.0, + "content": "schedule; however, our model catches up the performance with", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 252, + 505, + 266 + ], + "spans": [ + { + "bbox": [ + 106, + 253, + 120, + 263 + ], + "score": 0.86, + "content": "3 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 121, + 252, + 433, + 266 + ], + "score": 1.0, + "content": "schedule. When comparing the large models under similar FLOPs (RegionViT-", + "type": "text" + }, + { + "bbox": [ + 433, + 253, + 447, + 263 + ], + "score": 0.71, + "content": "\\mathbf { \\cdot B + }", + "type": "inline_equation" + }, + { + "bbox": [ + 447, + 252, + 460, + 266 + ], + "score": 1.0, + "content": "w/", + "type": "text" + }, + { + "bbox": [ + 460, + 253, + 484, + 264 + ], + "score": 0.57, + "content": "\\mathrm { P E G \\dag }", + "type": "inline_equation" + }, + { + "bbox": [ + 485, + 252, + 505, + 266 + ], + "score": 1.0, + "content": "), we", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 263, + 505, + 276 + ], + "spans": [ + { + "bbox": [ + 105, + 263, + 505, + 276 + ], + "score": 1.0, + "content": "further train our model with the similar resolution used by other works, and observe that our models", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 275, + 484, + 288 + ], + "spans": [ + { + "bbox": [ + 106, + 275, + 429, + 288 + ], + "score": 1.0, + "content": "outperform all others while being more parameter efficient (models marked with", + "type": "text" + }, + { + "bbox": [ + 429, + 275, + 435, + 286 + ], + "score": 0.82, + "content": "\\dagger", + "type": "inline_equation" + }, + { + "bbox": [ + 436, + 275, + 484, + 288 + ], + "score": 1.0, + "content": "in Table 5).", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 107, + 292, + 317, + 369 + ], + "lines": [ + { + "bbox": [ + 106, + 292, + 318, + 303 + ], + "spans": [ + { + "bbox": [ + 106, + 292, + 318, + 303 + ], + "score": 1.0, + "content": "Results on Semantic Segmentation. Table 6 com-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 303, + 317, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 303, + 317, + 314 + ], + "score": 1.0, + "content": "pares ours to Swin (Liu et al., 2021) and Twins. As", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 313, + 318, + 325 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 318, + 325 + ], + "score": 1.0, + "content": "PEG is useful for object detection, we show the re-", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 324, + 318, + 336 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 318, + 336 + ], + "score": 1.0, + "content": "sults with PEG. Our models achieved significant im-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 336, + 317, + 347 + ], + "spans": [ + { + "bbox": [ + 105, + 336, + 317, + 347 + ], + "score": 1.0, + "content": "provement with similar FLOPs for both models. This", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 347, + 318, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 318, + 358 + ], + "score": 1.0, + "content": "suggests that the proposed R2L attention can effec-", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 358, + 220, + 369 + ], + "spans": [ + { + "bbox": [ + 106, + 358, + 220, + 369 + ], + "score": 1.0, + "content": "tively model global context.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21 + }, + { + "type": "table", + "bbox": [ + 324, + 305, + 504, + 365 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 327, + 294, + 502, + 304 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 325, + 292, + 504, + 306 + ], + "spans": [ + { + "bbox": [ + 325, + 292, + 504, + 306 + ], + "score": 1.0, + "content": "Table 6: Performance on semantic segmentation.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "table_body", + "bbox": [ + 324, + 305, + 504, + 365 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 324, + 305, + 504, + 365 + ], + "spans": [ + { + "bbox": [ + 324, + 305, + 504, + 365 + ], + "score": 0.977, + "html": "
ModelFLOPs (G)Params (M)mIoU (%)
Swin-T*4631.941.5
Twins-SVT-S3728.343.2
RegionViT-S+ (w/PEG)3735.745.3
Swin-S*7053.245.2
Twins-SVT-B6760.445.3
RegionViT-B+ (w/PEG)6778.347.5
", + "type": "table", + "image_path": "9a755f71d04bcbe40fe9f1e9a07d283be8a26c838520f23c77a0f33273d07ffa.jpg" + } + ] + } + ], + "index": 28, + "virtual_lines": [ + { + "bbox": [ + 324, + 305, + 504, + 317.0 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 324, + 317.0, + 504, + 329.0 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 324, + 329.0, + 504, + 341.0 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 324, + 341.0, + 504, + 353.0 + ], + "spans": [], + "index": 29 + }, + { + "bbox": [ + 324, + 353.0, + 504, + 365.0 + ], + "spans": [], + "index": 30 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 328, + 366, + 480, + 372 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 327, + 363, + 481, + 374 + ], + "spans": [ + { + "bbox": [ + 327, + 363, + 481, + 374 + ], + "score": 1.0, + "content": "∗ : Numbers are cited from the reproduced results of Twins’ paper.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31 + } + ], + "index": 28 + }, + { + "type": "title", + "bbox": [ + 108, + 389, + 229, + 399 + ], + "lines": [ + { + "bbox": [ + 106, + 387, + 231, + 401 + ], + "spans": [ + { + "bbox": [ + 106, + 387, + 231, + 401 + ], + "score": 1.0, + "content": "4.3 ACTION RECOGNITION", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 32 + }, + { + "type": "text", + "bbox": [ + 106, + 411, + 504, + 455 + ], + "lines": [ + { + "bbox": [ + 106, + 411, + 505, + 423 + ], + "spans": [ + { + "bbox": [ + 106, + 411, + 505, + 423 + ], + "score": 1.0, + "content": "Datasets and Setup. We validate our approach on Kinetics400 (K400) (Kay et al., 2017) and", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 422, + 506, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 422, + 506, + 434 + ], + "score": 1.0, + "content": "Something-Something V2 (SSV2) (Goyal et al., 2017). We adopt the divided-space-time attention", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 433, + 506, + 446 + ], + "spans": [ + { + "bbox": [ + 105, + 433, + 506, + 446 + ], + "score": 1.0, + "content": "in TimeSformer (Bertasius et al., 2021) as the temporal modeling to perform action recognition", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 444, + 325, + 456 + ], + "spans": [ + { + "bbox": [ + 105, + 444, + 325, + 456 + ], + "score": 1.0, + "content": "experiments. More details could be found in Sec. A.4.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34.5 + }, + { + "type": "text", + "bbox": [ + 107, + 461, + 297, + 527 + ], + "lines": [ + { + "bbox": [ + 106, + 460, + 297, + 473 + ], + "spans": [ + { + "bbox": [ + 106, + 460, + 297, + 473 + ], + "score": 1.0, + "content": "Results. Table 7 shows that with RegionViT", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 471, + 297, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 297, + 484 + ], + "score": 1.0, + "content": "as the backbone, the model can be much more", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 483, + 299, + 495 + ], + "spans": [ + { + "bbox": [ + 106, + 483, + 299, + 495 + ], + "score": 1.0, + "content": "efficient with competitive accuracy to TimeS-", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 494, + 298, + 505 + ], + "spans": [ + { + "bbox": [ + 106, + 494, + 298, + 505 + ], + "score": 1.0, + "content": "former, which uses vanilla ViT as the back-", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 505, + 298, + 517 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 298, + 517 + ], + "score": 1.0, + "content": "bone. RegionViT-M could reduce more than", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 516, + 299, + 528 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 127, + 527 + ], + "score": 0.84, + "content": "50 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 127, + 516, + 299, + 528 + ], + "score": 1.0, + "content": "FLOPs and parameters than the TimeS-", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 39.5 + }, + { + "type": "table", + "bbox": [ + 304, + 474, + 504, + 514 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 322, + 464, + 484, + 474 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 321, + 462, + 486, + 475 + ], + "spans": [ + { + "bbox": [ + 321, + 462, + 486, + 475 + ], + "score": 1.0, + "content": "Table 7: Performance on action recognition.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 43 + }, + { + "type": "table_body", + "bbox": [ + 304, + 474, + 504, + 514 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 304, + 474, + 504, + 514 + ], + "spans": [ + { + "bbox": [ + 304, + 474, + 504, + 514 + ], + "score": 0.967, + "html": "
ModelFLOPs* (G)Params (M)K400 Acc. (%)SSV2 Acc.(%)
TimeSformer197121.475.859.5
x TimeSformer†197121.477.159.2
RegionViT-S59.442.976.659.7
RegionViT-M83.157.577.659.8
", + "type": "table", + "image_path": "518b4ba3b1852554ba2c0b622d00dc063763da99dc3b392c48e4e9d21d3a3cd2.jpg" + } + ] + } + ], + "index": 45, + "virtual_lines": [ + { + "bbox": [ + 304, + 474, + 504, + 487.3333333333333 + ], + "spans": [], + "index": 44 + }, + { + "bbox": [ + 304, + 487.3333333333333, + 504, + 500.66666666666663 + ], + "spans": [], + "index": 45 + }, + { + "bbox": [ + 304, + 500.66666666666663, + 504, + 514.0 + ], + "spans": [], + "index": 46 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 307, + 514, + 432, + 520 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 308, + 513, + 432, + 522 + ], + "spans": [ + { + "bbox": [ + 308, + 513, + 432, + 522 + ], + "score": 1.0, + "content": "∗ : FLOPs of single crop, †: retrained with the same setting.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 47 + } + ], + "index": 45 + }, + { + "type": "text", + "bbox": [ + 107, + 528, + 504, + 549 + ], + "lines": [ + { + "bbox": [ + 106, + 527, + 505, + 539 + ], + "spans": [ + { + "bbox": [ + 106, + 527, + 505, + 539 + ], + "score": 1.0, + "content": "former. This not only shows the importance of spatial modeling in action recognition but also", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 537, + 454, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 454, + 550 + ], + "score": 1.0, + "content": "validates that our proposed model can also be extended for efficient action recognition.", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 48.5 + }, + { + "type": "title", + "bbox": [ + 108, + 569, + 208, + 580 + ], + "lines": [ + { + "bbox": [ + 106, + 568, + 210, + 581 + ], + "spans": [ + { + "bbox": [ + 106, + 568, + 210, + 581 + ], + "score": 1.0, + "content": "4.4 ABLATION STUDY", + "type": "text" + } + ], + "index": 50 + } + ], + "index": 50 + }, + { + "type": "text", + "bbox": [ + 106, + 592, + 505, + 614 + ], + "lines": [ + { + "bbox": [ + 106, + 591, + 506, + 605 + ], + "spans": [ + { + "bbox": [ + 106, + 591, + 506, + 605 + ], + "score": 1.0, + "content": "We perform the following experiments to verify the effectiveness of different components in Region-", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 106, + 602, + 346, + 615 + ], + "spans": [ + { + "bbox": [ + 106, + 602, + 346, + 615 + ], + "score": 1.0, + "content": "ViT. All experiments are conducted based on RegionViT-S.", + "type": "text" + } + ], + "index": 52 + } + ], + "index": 51.5 + }, + { + "type": "text", + "bbox": [ + 107, + 619, + 298, + 686 + ], + "lines": [ + { + "bbox": [ + 106, + 619, + 297, + 631 + ], + "spans": [ + { + "bbox": [ + 106, + 619, + 297, + 631 + ], + "score": 1.0, + "content": "Regional Tokens. Table 8 shows the results", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 106, + 631, + 297, + 642 + ], + "spans": [ + { + "bbox": [ + 106, + 631, + 297, + 642 + ], + "score": 1.0, + "content": "with and without regional tokens on three tasks,", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 105, + 641, + 297, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 297, + 654 + ], + "score": 1.0, + "content": "and Table A5 includes FLOPs and parameters", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 106, + 653, + 298, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 653, + 298, + 664 + ], + "score": 1.0, + "content": "comparison. For image classification, the re-", + "type": "text" + } + ], + "index": 56 + }, + { + "bbox": [ + 106, + 664, + 298, + 676 + ], + "spans": [ + { + "bbox": [ + 106, + 664, + 234, + 676 + ], + "score": 1.0, + "content": "gional tokens provide around", + "type": "text" + }, + { + "bbox": [ + 234, + 664, + 257, + 674 + ], + "score": 0.86, + "content": "0 . 4 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 257, + 664, + 298, + 676 + ], + "score": 1.0, + "content": "improve-", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 105, + 674, + 299, + 687 + ], + "spans": [ + { + "bbox": [ + 105, + 674, + 299, + 687 + ], + "score": 1.0, + "content": "ment with negligible overhead in both compu-", + "type": "text" + } + ], + "index": 58 + } + ], + "index": 55.5 + }, + { + "type": "table", + "bbox": [ + 304, + 633, + 505, + 685 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 335, + 622, + 473, + 632 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 334, + 620, + 474, + 634 + ], + "spans": [ + { + "bbox": [ + 334, + 620, + 474, + 634 + ], + "score": 1.0, + "content": "Table 8: Ablation on regional tokens.", + "type": "text" + } + ], + "index": 59 + } + ], + "index": 59 + }, + { + "type": "table_body", + "bbox": [ + 304, + 633, + 505, + 685 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 304, + 633, + 505, + 685 + ], + "spans": [ + { + "bbox": [ + 304, + 633, + 505, + 685 + ], + "score": 0.971, + "html": "
Dataset Regional TokensIN1K Acc.MS COCO MaskRCNN RetinaNetADE20K SemanticFPN
(APb/APm) 40.5/37.8(APb) 40.7(mIoU) 42.3
N Y82.2 82.642.5/39.542.243.7
", + "type": "table", + "image_path": "2b3d805d53ef1929b5f5aaea19faff8082637679fed6d1d68796360c501ea01d.jpg" + } + ] + } + ], + "index": 61.5, + "virtual_lines": [ + { + "bbox": [ + 304, + 633, + 505, + 646.0 + ], + "spans": [], + "index": 60 + }, + { + "bbox": [ + 304, + 646.0, + 505, + 659.0 + ], + "spans": [], + "index": 61 + }, + { + "bbox": [ + 304, + 659.0, + 505, + 672.0 + ], + "spans": [], + "index": 62 + }, + { + "bbox": [ + 304, + 672.0, + 505, + 685.0 + ], + "spans": [], + "index": 63 + } + ] + } + ], + "index": 60.25 + }, + { + "type": "text", + "bbox": [ + 108, + 686, + 505, + 730 + ], + "lines": [ + { + "bbox": [ + 106, + 686, + 505, + 698 + ], + "spans": [ + { + "bbox": [ + 106, + 686, + 137, + 698 + ], + "score": 1.0, + "content": "tations", + "type": "text" + }, + { + "bbox": [ + 137, + 686, + 158, + 697 + ], + "score": 0.84, + "content": "( 6 \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 686, + 226, + 698 + ], + "score": 1.0, + "content": "and parameters", + "type": "text" + }, + { + "bbox": [ + 226, + 686, + 254, + 697 + ], + "score": 0.88, + "content": "( 0 . 7 \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 255, + 686, + 505, + 698 + ], + "score": 1.0, + "content": ". On the other hand, the regional tokens clearly improve the", + "type": "text" + } + ], + "index": 64 + }, + { + "bbox": [ + 105, + 696, + 505, + 709 + ], + "spans": [ + { + "bbox": [ + 105, + 696, + 449, + 709 + ], + "score": 1.0, + "content": "performance of object detection and semantic segmentation with negligible overhead (", + "type": "text" + }, + { + "bbox": [ + 449, + 697, + 471, + 707 + ], + "score": 0.88, + "content": "< 2 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 472, + 696, + 505, + 709 + ], + "score": 1.0, + "content": "on both", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 104, + 707, + 506, + 720 + ], + "spans": [ + { + "bbox": [ + 104, + 707, + 506, + 720 + ], + "score": 1.0, + "content": "FLOPs and parameters). This is because dense prediction tasks require more multi-scale features", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 105, + 718, + 485, + 731 + ], + "spans": [ + { + "bbox": [ + 105, + 718, + 485, + 731 + ], + "score": 1.0, + "content": "with global contextual information (provided by the regional tokens) than image classification.", + "type": "text" + } + ], + "index": 67 + } + ], + "index": 65.5 + } + ], + "page_idx": 7, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 300, + 750, + 309, + 761 + ], + "spans": [ + { + "bbox": [ + 300, + 750, + 309, + 761 + ], + "score": 1.0, + "content": "8", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 181 + ], + "lines": [], + "index": 4, + "bbox_fs": [ + 105, + 82, + 505, + 182 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 187, + 505, + 286 + ], + "lines": [ + { + "bbox": [ + 106, + 187, + 506, + 199 + ], + "spans": [ + { + "bbox": [ + 106, + 187, + 506, + 199 + ], + "score": 1.0, + "content": "Results on Object Detection. Table 5 compares RegionViT with other methods. For the smaller", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 198, + 505, + 210 + ], + "spans": [ + { + "bbox": [ + 106, + 198, + 256, + 210 + ], + "score": 1.0, + "content": "models, RegionViT-S and RegionViT-", + "type": "text" + }, + { + "bbox": [ + 257, + 199, + 269, + 209 + ], + "score": 0.84, + "content": "^ { S + }", + "type": "inline_equation" + }, + { + "bbox": [ + 269, + 198, + 505, + 210 + ], + "score": 1.0, + "content": "achieve similar accuracy while being moderately efficient in", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 209, + 506, + 222 + ], + "spans": [ + { + "bbox": [ + 105, + 209, + 506, + 222 + ], + "score": 1.0, + "content": "terms of FLOPs. We find that the position encoding generator (PEG) proposed by (Chu et al., 2021b)", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 219, + 505, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 219, + 505, + 233 + ], + "score": 1.0, + "content": "could help the detection accuracy substantially, so we also provide the results with PEG, which are", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "spans": [ + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "score": 1.0, + "content": "also used in Twins (Chu et al., 2021a). For the middle size models with RetinaNet, RegionViT-B are", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 242, + 505, + 255 + ], + "spans": [ + { + "bbox": [ + 105, + 242, + 233, + 255 + ], + "score": 1.0, + "content": "slightly worse than Twins with", + "type": "text" + }, + { + "bbox": [ + 233, + 243, + 247, + 253 + ], + "score": 0.86, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 248, + 242, + 505, + 255 + ], + "score": 1.0, + "content": "schedule; however, our model catches up the performance with", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 252, + 505, + 266 + ], + "spans": [ + { + "bbox": [ + 106, + 253, + 120, + 263 + ], + "score": 0.86, + "content": "3 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 121, + 252, + 433, + 266 + ], + "score": 1.0, + "content": "schedule. When comparing the large models under similar FLOPs (RegionViT-", + "type": "text" + }, + { + "bbox": [ + 433, + 253, + 447, + 263 + ], + "score": 0.71, + "content": "\\mathbf { \\cdot B + }", + "type": "inline_equation" + }, + { + "bbox": [ + 447, + 252, + 460, + 266 + ], + "score": 1.0, + "content": "w/", + "type": "text" + }, + { + "bbox": [ + 460, + 253, + 484, + 264 + ], + "score": 0.57, + "content": "\\mathrm { P E G \\dag }", + "type": "inline_equation" + }, + { + "bbox": [ + 485, + 252, + 505, + 266 + ], + "score": 1.0, + "content": "), we", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 263, + 505, + 276 + ], + "spans": [ + { + "bbox": [ + 105, + 263, + 505, + 276 + ], + "score": 1.0, + "content": "further train our model with the similar resolution used by other works, and observe that our models", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 275, + 484, + 288 + ], + "spans": [ + { + "bbox": [ + 106, + 275, + 429, + 288 + ], + "score": 1.0, + "content": "outperform all others while being more parameter efficient (models marked with", + "type": "text" + }, + { + "bbox": [ + 429, + 275, + 435, + 286 + ], + "score": 0.82, + "content": "\\dagger", + "type": "inline_equation" + }, + { + "bbox": [ + 436, + 275, + 484, + 288 + ], + "score": 1.0, + "content": "in Table 5).", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 13, + "bbox_fs": [ + 105, + 187, + 506, + 288 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 292, + 317, + 369 + ], + "lines": [ + { + "bbox": [ + 106, + 292, + 318, + 303 + ], + "spans": [ + { + "bbox": [ + 106, + 292, + 318, + 303 + ], + "score": 1.0, + "content": "Results on Semantic Segmentation. Table 6 com-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 303, + 317, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 303, + 317, + 314 + ], + "score": 1.0, + "content": "pares ours to Swin (Liu et al., 2021) and Twins. As", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 313, + 318, + 325 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 318, + 325 + ], + "score": 1.0, + "content": "PEG is useful for object detection, we show the re-", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 324, + 318, + 336 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 318, + 336 + ], + "score": 1.0, + "content": "sults with PEG. Our models achieved significant im-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 336, + 317, + 347 + ], + "spans": [ + { + "bbox": [ + 105, + 336, + 317, + 347 + ], + "score": 1.0, + "content": "provement with similar FLOPs for both models. This", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 347, + 318, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 318, + 358 + ], + "score": 1.0, + "content": "suggests that the proposed R2L attention can effec-", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 358, + 220, + 369 + ], + "spans": [ + { + "bbox": [ + 106, + 358, + 220, + 369 + ], + "score": 1.0, + "content": "tively model global context.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21, + "bbox_fs": [ + 105, + 292, + 318, + 369 + ] + }, + { + "type": "table", + "bbox": [ + 324, + 305, + 504, + 365 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 327, + 294, + 502, + 304 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 325, + 292, + 504, + 306 + ], + "spans": [ + { + "bbox": [ + 325, + 292, + 504, + 306 + ], + "score": 1.0, + "content": "Table 6: Performance on semantic segmentation.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "table_body", + "bbox": [ + 324, + 305, + 504, + 365 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 324, + 305, + 504, + 365 + ], + "spans": [ + { + "bbox": [ + 324, + 305, + 504, + 365 + ], + "score": 0.977, + "html": "
ModelFLOPs (G)Params (M)mIoU (%)
Swin-T*4631.941.5
Twins-SVT-S3728.343.2
RegionViT-S+ (w/PEG)3735.745.3
Swin-S*7053.245.2
Twins-SVT-B6760.445.3
RegionViT-B+ (w/PEG)6778.347.5
", + "type": "table", + "image_path": "9a755f71d04bcbe40fe9f1e9a07d283be8a26c838520f23c77a0f33273d07ffa.jpg" + } + ] + } + ], + "index": 28, + "virtual_lines": [ + { + "bbox": [ + 324, + 305, + 504, + 317.0 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 324, + 317.0, + 504, + 329.0 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 324, + 329.0, + 504, + 341.0 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 324, + 341.0, + 504, + 353.0 + ], + "spans": [], + "index": 29 + }, + { + "bbox": [ + 324, + 353.0, + 504, + 365.0 + ], + "spans": [], + "index": 30 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 328, + 366, + 480, + 372 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 327, + 363, + 481, + 374 + ], + "spans": [ + { + "bbox": [ + 327, + 363, + 481, + 374 + ], + "score": 1.0, + "content": "∗ : Numbers are cited from the reproduced results of Twins’ paper.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31 + } + ], + "index": 28 + }, + { + "type": "title", + "bbox": [ + 108, + 389, + 229, + 399 + ], + "lines": [ + { + "bbox": [ + 106, + 387, + 231, + 401 + ], + "spans": [ + { + "bbox": [ + 106, + 387, + 231, + 401 + ], + "score": 1.0, + "content": "4.3 ACTION RECOGNITION", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 32 + }, + { + "type": "text", + "bbox": [ + 106, + 411, + 504, + 455 + ], + "lines": [ + { + "bbox": [ + 106, + 411, + 505, + 423 + ], + "spans": [ + { + "bbox": [ + 106, + 411, + 505, + 423 + ], + "score": 1.0, + "content": "Datasets and Setup. We validate our approach on Kinetics400 (K400) (Kay et al., 2017) and", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 422, + 506, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 422, + 506, + 434 + ], + "score": 1.0, + "content": "Something-Something V2 (SSV2) (Goyal et al., 2017). We adopt the divided-space-time attention", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 433, + 506, + 446 + ], + "spans": [ + { + "bbox": [ + 105, + 433, + 506, + 446 + ], + "score": 1.0, + "content": "in TimeSformer (Bertasius et al., 2021) as the temporal modeling to perform action recognition", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 444, + 325, + 456 + ], + "spans": [ + { + "bbox": [ + 105, + 444, + 325, + 456 + ], + "score": 1.0, + "content": "experiments. More details could be found in Sec. A.4.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34.5, + "bbox_fs": [ + 105, + 411, + 506, + 456 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 461, + 297, + 527 + ], + "lines": [ + { + "bbox": [ + 106, + 460, + 297, + 473 + ], + "spans": [ + { + "bbox": [ + 106, + 460, + 297, + 473 + ], + "score": 1.0, + "content": "Results. Table 7 shows that with RegionViT", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 471, + 297, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 297, + 484 + ], + "score": 1.0, + "content": "as the backbone, the model can be much more", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 483, + 299, + 495 + ], + "spans": [ + { + "bbox": [ + 106, + 483, + 299, + 495 + ], + "score": 1.0, + "content": "efficient with competitive accuracy to TimeS-", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 494, + 298, + 505 + ], + "spans": [ + { + "bbox": [ + 106, + 494, + 298, + 505 + ], + "score": 1.0, + "content": "former, which uses vanilla ViT as the back-", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 505, + 298, + 517 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 298, + 517 + ], + "score": 1.0, + "content": "bone. RegionViT-M could reduce more than", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 516, + 299, + 528 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 127, + 527 + ], + "score": 0.84, + "content": "50 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 127, + 516, + 299, + 528 + ], + "score": 1.0, + "content": "FLOPs and parameters than the TimeS-", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 39.5, + "bbox_fs": [ + 105, + 460, + 299, + 528 + ] + }, + { + "type": "table", + "bbox": [ + 304, + 474, + 504, + 514 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 322, + 464, + 484, + 474 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 321, + 462, + 486, + 475 + ], + "spans": [ + { + "bbox": [ + 321, + 462, + 486, + 475 + ], + "score": 1.0, + "content": "Table 7: Performance on action recognition.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 43 + }, + { + "type": "table_body", + "bbox": [ + 304, + 474, + 504, + 514 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 304, + 474, + 504, + 514 + ], + "spans": [ + { + "bbox": [ + 304, + 474, + 504, + 514 + ], + "score": 0.967, + "html": "
ModelFLOPs* (G)Params (M)K400 Acc. (%)SSV2 Acc.(%)
TimeSformer197121.475.859.5
x TimeSformer†197121.477.159.2
RegionViT-S59.442.976.659.7
RegionViT-M83.157.577.659.8
", + "type": "table", + "image_path": "518b4ba3b1852554ba2c0b622d00dc063763da99dc3b392c48e4e9d21d3a3cd2.jpg" + } + ] + } + ], + "index": 45, + "virtual_lines": [ + { + "bbox": [ + 304, + 474, + 504, + 487.3333333333333 + ], + "spans": [], + "index": 44 + }, + { + "bbox": [ + 304, + 487.3333333333333, + 504, + 500.66666666666663 + ], + "spans": [], + "index": 45 + }, + { + "bbox": [ + 304, + 500.66666666666663, + 504, + 514.0 + ], + "spans": [], + "index": 46 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 307, + 514, + 432, + 520 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 308, + 513, + 432, + 522 + ], + "spans": [ + { + "bbox": [ + 308, + 513, + 432, + 522 + ], + "score": 1.0, + "content": "∗ : FLOPs of single crop, †: retrained with the same setting.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 47 + } + ], + "index": 45 + }, + { + "type": "text", + "bbox": [ + 107, + 528, + 504, + 549 + ], + "lines": [ + { + "bbox": [ + 106, + 527, + 505, + 539 + ], + "spans": [ + { + "bbox": [ + 106, + 527, + 505, + 539 + ], + "score": 1.0, + "content": "former. This not only shows the importance of spatial modeling in action recognition but also", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 537, + 454, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 454, + 550 + ], + "score": 1.0, + "content": "validates that our proposed model can also be extended for efficient action recognition.", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 48.5, + "bbox_fs": [ + 105, + 527, + 505, + 550 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 569, + 208, + 580 + ], + "lines": [ + { + "bbox": [ + 106, + 568, + 210, + 581 + ], + "spans": [ + { + "bbox": [ + 106, + 568, + 210, + 581 + ], + "score": 1.0, + "content": "4.4 ABLATION STUDY", + "type": "text" + } + ], + "index": 50 + } + ], + "index": 50 + }, + { + "type": "text", + "bbox": [ + 106, + 592, + 505, + 614 + ], + "lines": [ + { + "bbox": [ + 106, + 591, + 506, + 605 + ], + "spans": [ + { + "bbox": [ + 106, + 591, + 506, + 605 + ], + "score": 1.0, + "content": "We perform the following experiments to verify the effectiveness of different components in Region-", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 106, + 602, + 346, + 615 + ], + "spans": [ + { + "bbox": [ + 106, + 602, + 346, + 615 + ], + "score": 1.0, + "content": "ViT. All experiments are conducted based on RegionViT-S.", + "type": "text" + } + ], + "index": 52 + } + ], + "index": 51.5, + "bbox_fs": [ + 106, + 591, + 506, + 615 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 619, + 298, + 686 + ], + "lines": [ + { + "bbox": [ + 106, + 619, + 297, + 631 + ], + "spans": [ + { + "bbox": [ + 106, + 619, + 297, + 631 + ], + "score": 1.0, + "content": "Regional Tokens. Table 8 shows the results", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 106, + 631, + 297, + 642 + ], + "spans": [ + { + "bbox": [ + 106, + 631, + 297, + 642 + ], + "score": 1.0, + "content": "with and without regional tokens on three tasks,", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 105, + 641, + 297, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 297, + 654 + ], + "score": 1.0, + "content": "and Table A5 includes FLOPs and parameters", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 106, + 653, + 298, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 653, + 298, + 664 + ], + "score": 1.0, + "content": "comparison. For image classification, the re-", + "type": "text" + } + ], + "index": 56 + }, + { + "bbox": [ + 106, + 664, + 298, + 676 + ], + "spans": [ + { + "bbox": [ + 106, + 664, + 234, + 676 + ], + "score": 1.0, + "content": "gional tokens provide around", + "type": "text" + }, + { + "bbox": [ + 234, + 664, + 257, + 674 + ], + "score": 0.86, + "content": "0 . 4 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 257, + 664, + 298, + 676 + ], + "score": 1.0, + "content": "improve-", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 105, + 674, + 299, + 687 + ], + "spans": [ + { + "bbox": [ + 105, + 674, + 299, + 687 + ], + "score": 1.0, + "content": "ment with negligible overhead in both compu-", + "type": "text" + } + ], + "index": 58 + } + ], + "index": 55.5, + "bbox_fs": [ + 105, + 619, + 299, + 687 + ] + }, + { + "type": "table", + "bbox": [ + 304, + 633, + 505, + 685 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 335, + 622, + 473, + 632 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 334, + 620, + 474, + 634 + ], + "spans": [ + { + "bbox": [ + 334, + 620, + 474, + 634 + ], + "score": 1.0, + "content": "Table 8: Ablation on regional tokens.", + "type": "text" + } + ], + "index": 59 + } + ], + "index": 59 + }, + { + "type": "table_body", + "bbox": [ + 304, + 633, + 505, + 685 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 304, + 633, + 505, + 685 + ], + "spans": [ + { + "bbox": [ + 304, + 633, + 505, + 685 + ], + "score": 0.971, + "html": "
Dataset Regional TokensIN1K Acc.MS COCO MaskRCNN RetinaNetADE20K SemanticFPN
(APb/APm) 40.5/37.8(APb) 40.7(mIoU) 42.3
N Y82.2 82.642.5/39.542.243.7
", + "type": "table", + "image_path": "2b3d805d53ef1929b5f5aaea19faff8082637679fed6d1d68796360c501ea01d.jpg" + } + ] + } + ], + "index": 61.5, + "virtual_lines": [ + { + "bbox": [ + 304, + 633, + 505, + 646.0 + ], + "spans": [], + "index": 60 + }, + { + "bbox": [ + 304, + 646.0, + 505, + 659.0 + ], + "spans": [], + "index": 61 + }, + { + "bbox": [ + 304, + 659.0, + 505, + 672.0 + ], + "spans": [], + "index": 62 + }, + { + "bbox": [ + 304, + 672.0, + 505, + 685.0 + ], + "spans": [], + "index": 63 + } + ] + } + ], + "index": 60.25 + }, + { + "type": "text", + "bbox": [ + 108, + 686, + 505, + 730 + ], + "lines": [ + { + "bbox": [ + 106, + 686, + 505, + 698 + ], + "spans": [ + { + "bbox": [ + 106, + 686, + 137, + 698 + ], + "score": 1.0, + "content": "tations", + "type": "text" + }, + { + "bbox": [ + 137, + 686, + 158, + 697 + ], + "score": 0.84, + "content": "( 6 \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 686, + 226, + 698 + ], + "score": 1.0, + "content": "and parameters", + "type": "text" + }, + { + "bbox": [ + 226, + 686, + 254, + 697 + ], + "score": 0.88, + "content": "( 0 . 7 \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 255, + 686, + 505, + 698 + ], + "score": 1.0, + "content": ". On the other hand, the regional tokens clearly improve the", + "type": "text" + } + ], + "index": 64 + }, + { + "bbox": [ + 105, + 696, + 505, + 709 + ], + "spans": [ + { + "bbox": [ + 105, + 696, + 449, + 709 + ], + "score": 1.0, + "content": "performance of object detection and semantic segmentation with negligible overhead (", + "type": "text" + }, + { + "bbox": [ + 449, + 697, + 471, + 707 + ], + "score": 0.88, + "content": "< 2 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 472, + 696, + 505, + 709 + ], + "score": 1.0, + "content": "on both", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 104, + 707, + 506, + 720 + ], + "spans": [ + { + "bbox": [ + 104, + 707, + 506, + 720 + ], + "score": 1.0, + "content": "FLOPs and parameters). This is because dense prediction tasks require more multi-scale features", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 105, + 718, + 485, + 731 + ], + "spans": [ + { + "bbox": [ + 105, + 718, + 485, + 731 + ], + "score": 1.0, + "content": "with global contextual information (provided by the regional tokens) than image classification.", + "type": "text" + } + ], + "index": 67 + } + ], + "index": 65.5, + "bbox_fs": [ + 104, + 686, + 506, + 731 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 83, + 297, + 137 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 298, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 298, + 95 + ], + "score": 1.0, + "content": "Downsampling. Table 9 shows different ap-", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 94, + 296, + 105 + ], + "spans": [ + { + "bbox": [ + 105, + 94, + 276, + 105 + ], + "score": 1.0, + "content": "proaches for downsampling the patches.", + "type": "text" + }, + { + "bbox": [ + 276, + 94, + 296, + 104 + ], + "score": 0.83, + "content": "3 \\times 3", + "type": "inline_equation" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 104, + 298, + 115 + ], + "spans": [ + { + "bbox": [ + 106, + 104, + 258, + 115 + ], + "score": 1.0, + "content": "kernel size achieves better results than", + "type": "text" + }, + { + "bbox": [ + 259, + 105, + 278, + 115 + ], + "score": 0.88, + "content": "2 \\times 2", + "type": "inline_equation" + }, + { + "bbox": [ + 279, + 104, + 298, + 115 + ], + "score": 1.0, + "content": "con-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 114, + 298, + 127 + ], + "spans": [ + { + "bbox": [ + 105, + 114, + 298, + 127 + ], + "score": 1.0, + "content": "volution, because the kernel is non-overlapped", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 126, + 299, + 138 + ], + "spans": [ + { + "bbox": [ + 106, + 126, + 126, + 138 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 126, + 127, + 146, + 137 + ], + "score": 0.87, + "content": "2 \\times 2", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 126, + 299, + 138 + ], + "score": 1.0, + "content": "convolution, which limits the interac-", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 2 + }, + { + "type": "table", + "bbox": [ + 304, + 96, + 504, + 138 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 319, + 84, + 487, + 95 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 319, + 83, + 488, + 97 + ], + "spans": [ + { + "bbox": [ + 319, + 83, + 488, + 97 + ], + "score": 1.0, + "content": "Table 9: Different downsampling approaches.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "table_body", + "bbox": [ + 304, + 96, + 504, + 138 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 304, + 96, + 504, + 138 + ], + "spans": [ + { + "bbox": [ + 304, + 96, + 504, + 138 + ], + "score": 0.969, + "html": "
DownsamplingParamsFLOPs (G)IN1K Acc. (%)
2×2 convolution32.15.582.3
3×3 convolution34.15.782.5
3×3 depth-wise conv.30.65.382.6
", + "type": "table", + "image_path": "34b8c71a505dd7e761c717f1d52783fa1ed69d59742d33ed118fc72e59a2e539.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 304, + 96, + 504, + 110.0 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 304, + 110.0, + 504, + 124.0 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 304, + 124.0, + 504, + 138.0 + ], + "spans": [], + "index": 9 + } + ] + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 108, + 138, + 507, + 160 + ], + "lines": [ + { + "bbox": [ + 106, + 137, + 505, + 150 + ], + "spans": [ + { + "bbox": [ + 106, + 137, + 505, + 150 + ], + "score": 1.0, + "content": "tion among local tokens. Moreover, using regular convolution does not improve the performance but", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 147, + 507, + 162 + ], + "spans": [ + { + "bbox": [ + 105, + 147, + 319, + 162 + ], + "score": 1.0, + "content": "increases computations and parameters. Thus, we use", + "type": "text" + }, + { + "bbox": [ + 319, + 149, + 339, + 159 + ], + "score": 0.86, + "content": "3 \\times 3", + "type": "inline_equation" + }, + { + "bbox": [ + 339, + 147, + 507, + 162 + ], + "score": 1.0, + "content": "depthwise convolution for downsampling.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 106, + 165, + 297, + 221 + ], + "lines": [ + { + "bbox": [ + 106, + 165, + 298, + 177 + ], + "spans": [ + { + "bbox": [ + 106, + 165, + 298, + 177 + ], + "score": 1.0, + "content": "Weight sharing between RSA and LSA as", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 176, + 298, + 188 + ], + "spans": [ + { + "bbox": [ + 106, + 176, + 298, + 188 + ], + "score": 1.0, + "content": "well as downsampling. Table 10 shows the re-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 187, + 297, + 199 + ], + "spans": [ + { + "bbox": [ + 106, + 187, + 297, + 199 + ], + "score": 1.0, + "content": "sults with and without weight sharing. As seen", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 197, + 297, + 210 + ], + "spans": [ + { + "bbox": [ + 105, + 197, + 297, + 210 + ], + "score": 1.0, + "content": "from the table, using separated weights for RSA", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 208, + 298, + 223 + ], + "spans": [ + { + "bbox": [ + 106, + 208, + 298, + 223 + ], + "score": 1.0, + "content": "and LSA only slightly improves the accuracy", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 105, + 221, + 505, + 243 + ], + "lines": [ + { + "bbox": [ + 106, + 220, + 505, + 232 + ], + "spans": [ + { + "bbox": [ + 106, + 220, + 505, + 232 + ], + "score": 1.0, + "content": "but significantly increases the model size. This suggests sharing parameters between RSA and LSA", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 231, + 372, + 243 + ], + "spans": [ + { + "bbox": [ + 105, + 231, + 372, + 243 + ], + "score": 1.0, + "content": "suffices for learning local and global information in our approach.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5 + }, + { + "type": "table", + "bbox": [ + 304, + 179, + 504, + 217 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 356, + 167, + 452, + 178 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 354, + 165, + 454, + 181 + ], + "spans": [ + { + "bbox": [ + 354, + 165, + 454, + 181 + ], + "score": 1.0, + "content": "Table 10: Weight sharing.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "table_body", + "bbox": [ + 304, + 179, + 504, + 217 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 304, + 179, + 504, + 217 + ], + "spans": [ + { + "bbox": [ + 304, + 179, + 504, + 217 + ], + "score": 0.96, + "html": "
Weight sharingParamsFLOPs (G)IN1K Acc. (%)
N40.45.382.7
Y30.65.382.6
", + "type": "table", + "image_path": "ade46fff21f8473d5206e19825c1ca1182433ecda061114ef2b68a3f4999d873.jpg" + } + ] + } + ], + "index": 19.5, + "virtual_lines": [ + { + "bbox": [ + 304, + 179, + 504, + 198.0 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 304, + 198.0, + 504, + 217.0 + ], + "spans": [], + "index": 20 + } + ] + } + ], + "index": 18.75 + }, + { + "type": "text", + "bbox": [ + 106, + 248, + 297, + 303 + ], + "lines": [ + { + "bbox": [ + 106, + 247, + 298, + 259 + ], + "spans": [ + { + "bbox": [ + 106, + 247, + 298, + 259 + ], + "score": 1.0, + "content": "Keys in LSA. Table 11 studies the number of", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 258, + 297, + 271 + ], + "spans": [ + { + "bbox": [ + 105, + 258, + 297, + 271 + ], + "score": 1.0, + "content": "regional tokens in LSA. When using all regional", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 270, + 297, + 281 + ], + "spans": [ + { + "bbox": [ + 106, + 270, + 297, + 281 + ], + "score": 1.0, + "content": "tokens in the LSA, it provides the local tokens", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 281, + 298, + 293 + ], + "spans": [ + { + "bbox": [ + 106, + 281, + 298, + 293 + ], + "score": 1.0, + "content": "the possibility to explore all global information.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 291, + 297, + 304 + ], + "spans": [ + { + "bbox": [ + 105, + 291, + 259, + 304 + ], + "score": 1.0, + "content": "Nonetheless, this model only improves", + "type": "text" + }, + { + "bbox": [ + 259, + 292, + 281, + 302 + ], + "score": 0.87, + "content": "0 . 1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 282, + 291, + 297, + 304 + ], + "score": 1.0, + "content": "but", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 23 + }, + { + "type": "table", + "bbox": [ + 304, + 262, + 504, + 298 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 359, + 250, + 448, + 261 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 358, + 249, + 450, + 262 + ], + "spans": [ + { + "bbox": [ + 358, + 249, + 450, + 262 + ], + "score": 1.0, + "content": "Table 11: Keys in LSA.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 26 + }, + { + "type": "table_body", + "bbox": [ + 304, + 262, + 504, + 298 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 304, + 262, + 504, + 298 + ], + "spans": [ + { + "bbox": [ + 304, + 262, + 504, + 298 + ], + "score": 0.967, + "html": "
Regional tokensParamsFLOPs (G)IN1K Acc. (%)
All30.66.082.7
Corresponding30.65.382.6
", + "type": "table", + "image_path": "02feafa313da8136ea97cf1fc43c19477a597fbad9be75bc30c0bbc917625e70.jpg" + } + ] + } + ], + "index": 27.5, + "virtual_lines": [ + { + "bbox": [ + 304, + 262, + 504, + 280.0 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 304, + 280.0, + 504, + 298.0 + ], + "spans": [], + "index": 28 + } + ] + } + ], + "index": 26.75 + }, + { + "type": "text", + "bbox": [ + 106, + 303, + 505, + 325 + ], + "lines": [ + { + "bbox": [ + 105, + 303, + 504, + 315 + ], + "spans": [ + { + "bbox": [ + 105, + 303, + 127, + 315 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 127, + 303, + 147, + 313 + ], + "score": 0.87, + "content": "13 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 147, + 303, + 504, + 315 + ], + "score": 1.0, + "content": "more computations, which suggests that when considering the regional tokens, the one", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 313, + 412, + 326 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 412, + 326 + ], + "score": 1.0, + "content": "associated with the current region is sufficient to achieve good performance.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5 + }, + { + "type": "text", + "bbox": [ + 107, + 331, + 297, + 396 + ], + "lines": [ + { + "bbox": [ + 106, + 329, + 298, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 329, + 298, + 343 + ], + "score": 1.0, + "content": "Position information. Table 12 compares per-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 342, + 297, + 353 + ], + "spans": [ + { + "bbox": [ + 106, + 342, + 297, + 353 + ], + "score": 1.0, + "content": "formance of different combinations of absolute", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 353, + 297, + 365 + ], + "spans": [ + { + "bbox": [ + 105, + 353, + 297, + 365 + ], + "score": 1.0, + "content": "position embedding (APE) and relative position", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 363, + 298, + 374 + ], + "spans": [ + { + "bbox": [ + 106, + 363, + 298, + 374 + ], + "score": 1.0, + "content": "bias (RPB). The models with RPB achieve bet-", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 374, + 299, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 374, + 299, + 387 + ], + "score": 1.0, + "content": "ter accuracy with similar FLOPs and parameters.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 385, + 298, + 398 + ], + "spans": [ + { + "bbox": [ + 105, + 385, + 298, + 398 + ], + "score": 1.0, + "content": "Although the model with APE could slightly im-", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 33.5 + }, + { + "type": "table", + "bbox": [ + 304, + 344, + 504, + 391 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 328, + 333, + 480, + 343 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 327, + 331, + 481, + 344 + ], + "spans": [ + { + "bbox": [ + 327, + 331, + 481, + 344 + ], + "score": 1.0, + "content": "Table 12: Different position information.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "table_body", + "bbox": [ + 304, + 344, + 504, + 391 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 304, + 344, + 504, + 391 + ], + "spans": [ + { + "bbox": [ + 304, + 344, + 504, + 391 + ], + "score": 0.971, + "html": "
Abs. pos.Rel. pos.Params (M)FLOPs (G)IN1K Acc. (%)
NN30.65.382.4
YN30.95.382.2
YY30.95.382.7
NY30.65.382.6
", + "type": "table", + "image_path": "5fde92d3500b48401daf0f3493d4bdca1755d22ee0419743ec863dc937e4e07f.jpg" + } + ] + } + ], + "index": 39, + "virtual_lines": [ + { + "bbox": [ + 304, + 344, + 504, + 359.6666666666667 + ], + "spans": [], + "index": 38 + }, + { + "bbox": [ + 304, + 359.6666666666667, + 504, + 375.33333333333337 + ], + "spans": [], + "index": 39 + }, + { + "bbox": [ + 304, + 375.33333333333337, + 504, + 391.00000000000006 + ], + "spans": [], + "index": 40 + } + ] + } + ], + "index": 38.0 + }, + { + "type": "text", + "bbox": [ + 107, + 397, + 505, + 419 + ], + "lines": [ + { + "bbox": [ + 105, + 396, + 506, + 409 + ], + "spans": [ + { + "bbox": [ + 105, + 396, + 506, + 409 + ], + "score": 1.0, + "content": "prove the performance, it limits the model to run at a fixed resolution, which is not suitable for vision", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 407, + 506, + 420 + ], + "spans": [ + { + "bbox": [ + 105, + 407, + 506, + 420 + ], + "score": 1.0, + "content": "tasks where image size could be varied. Thus, we only adopt the relative position bias in our models.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41.5 + }, + { + "type": "text", + "bbox": [ + 106, + 424, + 298, + 479 + ], + "lines": [ + { + "bbox": [ + 106, + 423, + 297, + 437 + ], + "spans": [ + { + "bbox": [ + 106, + 423, + 297, + 437 + ], + "score": 1.0, + "content": "Overlapped windows. Table 13 compares the", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 434, + 299, + 448 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 299, + 448 + ], + "score": 1.0, + "content": "results with and without overlapped windows.", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 446, + 298, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 298, + 459 + ], + "score": 1.0, + "content": "The overlapping ratio of neighboring windows", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 457, + 298, + 469 + ], + "spans": [ + { + "bbox": [ + 105, + 457, + 116, + 469 + ], + "score": 1.0, + "content": "is", + "type": "text" + }, + { + "bbox": [ + 117, + 457, + 137, + 468 + ], + "score": 0.86, + "content": "50 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 457, + 298, + 469 + ], + "score": 1.0, + "content": "and hence this model allows more in-", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 468, + 298, + 481 + ], + "spans": [ + { + "bbox": [ + 105, + 468, + 298, + 481 + ], + "score": 1.0, + "content": "teractions among local tokens. It improves", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 45 + }, + { + "type": "table", + "bbox": [ + 304, + 438, + 504, + 475 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 340, + 426, + 468, + 437 + ], + "group_id": 4, + "lines": [ + { + "bbox": [ + 339, + 424, + 469, + 438 + ], + "spans": [ + { + "bbox": [ + 339, + 424, + 469, + 438 + ], + "score": 1.0, + "content": "Table 13: Overlapped windows.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 48 + }, + { + "type": "table_body", + "bbox": [ + 304, + 438, + 504, + 475 + ], + "group_id": 4, + "lines": [ + { + "bbox": [ + 304, + 438, + 504, + 475 + ], + "spans": [ + { + "bbox": [ + 304, + 438, + 504, + 475 + ], + "score": 0.965, + "html": "
ModelParamsFLOPs (G)IN1K Acc. (%)
overlapped30.618.882.8
non-overlapped30.65.382.6
", + "type": "table", + "image_path": "cbd48a837d0fcb2bd0b8853bf01a0e4804700eca5e82b229be0695ca1dc25d16.jpg" + } + ] + } + ], + "index": 49.5, + "virtual_lines": [ + { + "bbox": [ + 304, + 438, + 504, + 456.5 + ], + "spans": [], + "index": 49 + }, + { + "bbox": [ + 304, + 456.5, + 504, + 475.0 + ], + "spans": [], + "index": 50 + } + ] + } + ], + "index": 48.75 + }, + { + "type": "text", + "bbox": [ + 106, + 479, + 505, + 502 + ], + "lines": [ + { + "bbox": [ + 105, + 478, + 505, + 492 + ], + "spans": [ + { + "bbox": [ + 105, + 478, + 173, + 492 + ], + "score": 1.0, + "content": "RegionViT-S by", + "type": "text" + }, + { + "bbox": [ + 173, + 479, + 195, + 489 + ], + "score": 0.86, + "content": "0 . 2 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 195, + 478, + 305, + 492 + ], + "score": 1.0, + "content": "but increases the FLOPs by", + "type": "text" + }, + { + "bbox": [ + 305, + 479, + 326, + 490 + ], + "score": 0.87, + "content": "3 . 5 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 326, + 478, + 505, + 492 + ], + "score": 1.0, + "content": ". This suggests that the information exchange", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 105, + 489, + 451, + 503 + ], + "spans": [ + { + "bbox": [ + 105, + 489, + 451, + 503 + ], + "score": 1.0, + "content": "between the border of windows is sufficiently covered by the proposed R2L attention.", + "type": "text" + } + ], + "index": 52 + } + ], + "index": 51.5 + }, + { + "type": "text", + "bbox": [ + 106, + 507, + 297, + 562 + ], + "lines": [ + { + "bbox": [ + 106, + 506, + 297, + 518 + ], + "spans": [ + { + "bbox": [ + 106, + 506, + 297, + 518 + ], + "score": 1.0, + "content": "Comparison between R2L attention and", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 106, + 517, + 297, + 529 + ], + "spans": [ + { + "bbox": [ + 106, + 517, + 297, + 529 + ], + "score": 1.0, + "content": "shifted-window attention. Table 14 shows", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 105, + 529, + 298, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 529, + 298, + 540 + ], + "score": 1.0, + "content": "the comparison between shifted-window atten-", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 105, + 539, + 299, + 551 + ], + "spans": [ + { + "bbox": [ + 105, + 539, + 299, + 551 + ], + "score": 1.0, + "content": "tion (Liu et al., 2021) and proposed R2L atten-", + "type": "text" + } + ], + "index": 56 + }, + { + "bbox": [ + 106, + 551, + 297, + 562 + ], + "spans": [ + { + "bbox": [ + 106, + 551, + 297, + 562 + ], + "score": 1.0, + "content": "tion. R2L attention outperforms shifted-window", + "type": "text" + } + ], + "index": 57 + } + ], + "index": 55 + }, + { + "type": "text", + "bbox": [ + 108, + 562, + 504, + 584 + ], + "lines": [ + { + "bbox": [ + 106, + 561, + 505, + 574 + ], + "spans": [ + { + "bbox": [ + 106, + 561, + 159, + 574 + ], + "score": 1.0, + "content": "attention by", + "type": "text" + }, + { + "bbox": [ + 159, + 562, + 182, + 573 + ], + "score": 0.86, + "content": "0 . 7 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 182, + 561, + 505, + 574 + ], + "score": 1.0, + "content": ", which shows that our proposed R2L attention can effectively model global", + "type": "text" + } + ], + "index": 58 + }, + { + "bbox": [ + 105, + 572, + 277, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 277, + 586 + ], + "score": 1.0, + "content": "information to achieve good performance.", + "type": "text" + } + ], + "index": 59 + } + ], + "index": 58.5 + }, + { + "type": "table", + "bbox": [ + 304, + 521, + 504, + 558 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 312, + 509, + 493, + 519 + ], + "group_id": 5, + "lines": [ + { + "bbox": [ + 313, + 508, + 495, + 521 + ], + "spans": [ + { + "bbox": [ + 313, + 508, + 495, + 521 + ], + "score": 1.0, + "content": "Table 14: Comparison of different attentions.", + "type": "text" + } + ], + "index": 60 + } + ], + "index": 60 + }, + { + "type": "table_body", + "bbox": [ + 304, + 521, + 504, + 558 + ], + "group_id": 5, + "lines": [ + { + "bbox": [ + 304, + 521, + 504, + 558 + ], + "spans": [ + { + "bbox": [ + 304, + 521, + 504, + 558 + ], + "score": 0.963, + "html": "
ModelParamsFLOPs (G)IN1K Acc. (%)
Shifted-window29.04.581.3
R2L32.15.282.0
", + "type": "table", + "image_path": "bd9e65592297ebcfc30c495d4720dbd19d12d0cc359d374492af69bfeee00aa5.jpg" + } + ] + } + ], + "index": 61.5, + "virtual_lines": [ + { + "bbox": [ + 304, + 521, + 504, + 539.5 + ], + "spans": [], + "index": 61 + }, + { + "bbox": [ + 304, + 539.5, + 504, + 558.0 + ], + "spans": [], + "index": 62 + } + ] + } + ], + "index": 60.75 + }, + { + "type": "title", + "bbox": [ + 108, + 600, + 195, + 613 + ], + "lines": [ + { + "bbox": [ + 105, + 599, + 197, + 616 + ], + "spans": [ + { + "bbox": [ + 105, + 599, + 197, + 616 + ], + "score": 1.0, + "content": "5 CONCLUSION", + "type": "text" + } + ], + "index": 63 + } + ], + "index": 63 + }, + { + "type": "text", + "bbox": [ + 106, + 625, + 505, + 714 + ], + "lines": [ + { + "bbox": [ + 105, + 624, + 506, + 638 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 506, + 638 + ], + "score": 1.0, + "content": "In this paper, we propose a new ViT architecture that exploits the pyramid structure used in most of", + "type": "text" + } + ], + "index": 64 + }, + { + "bbox": [ + 105, + 635, + 506, + 649 + ], + "spans": [ + { + "bbox": [ + 105, + 635, + 506, + 649 + ], + "score": 1.0, + "content": "CNNs to provide multi-scale features and hence, the models can be more easily extended to different", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 106, + 647, + 505, + 659 + ], + "spans": [ + { + "bbox": [ + 106, + 647, + 505, + 659 + ], + "score": 1.0, + "content": "vision applications, like object detection. Moreover, the proposed regional-to-local attention relaxes", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 106, + 658, + 505, + 671 + ], + "spans": [ + { + "bbox": [ + 106, + 658, + 505, + 671 + ], + "score": 1.0, + "content": "the memory overhead of performing self-attention on the fine-grained image tokens by limiting the", + "type": "text" + } + ], + "index": 67 + }, + { + "bbox": [ + 105, + 669, + 506, + 682 + ], + "spans": [ + { + "bbox": [ + 105, + 669, + 506, + 682 + ], + "score": 1.0, + "content": "scope of attention but still keeping the capability to explore global information. Extensive experiments", + "type": "text" + } + ], + "index": 68 + }, + { + "bbox": [ + 106, + 681, + 505, + 693 + ], + "spans": [ + { + "bbox": [ + 106, + 681, + 505, + 693 + ], + "score": 1.0, + "content": "on several standard benchmark datasets well demonstrate that our proposed models outperform or are", + "type": "text" + } + ], + "index": 69 + }, + { + "bbox": [ + 106, + 691, + 506, + 704 + ], + "spans": [ + { + "bbox": [ + 106, + 691, + 506, + 704 + ], + "score": 1.0, + "content": "on par with many concurrent ViT variants on four vision applications, including image classification,", + "type": "text" + } + ], + "index": 70 + }, + { + "bbox": [ + 105, + 702, + 416, + 715 + ], + "spans": [ + { + "bbox": [ + 105, + 702, + 416, + 715 + ], + "score": 1.0, + "content": "object and keypoint detection, semantic segmentation and action recognition.", + "type": "text" + } + ], + "index": 71 + } + ], + "index": 67.5 + } + ], + "page_idx": 8, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 105, + 25, + 294, + 39 + ], + "spans": [ + { + "bbox": [ + 105, + 25, + 294, + 39 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "score": 1.0, + "content": "9", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 83, + 297, + 137 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 298, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 298, + 95 + ], + "score": 1.0, + "content": "Downsampling. Table 9 shows different ap-", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 94, + 296, + 105 + ], + "spans": [ + { + "bbox": [ + 105, + 94, + 276, + 105 + ], + "score": 1.0, + "content": "proaches for downsampling the patches.", + "type": "text" + }, + { + "bbox": [ + 276, + 94, + 296, + 104 + ], + "score": 0.83, + "content": "3 \\times 3", + "type": "inline_equation" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 104, + 298, + 115 + ], + "spans": [ + { + "bbox": [ + 106, + 104, + 258, + 115 + ], + "score": 1.0, + "content": "kernel size achieves better results than", + "type": "text" + }, + { + "bbox": [ + 259, + 105, + 278, + 115 + ], + "score": 0.88, + "content": "2 \\times 2", + "type": "inline_equation" + }, + { + "bbox": [ + 279, + 104, + 298, + 115 + ], + "score": 1.0, + "content": "con-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 114, + 298, + 127 + ], + "spans": [ + { + "bbox": [ + 105, + 114, + 298, + 127 + ], + "score": 1.0, + "content": "volution, because the kernel is non-overlapped", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 126, + 299, + 138 + ], + "spans": [ + { + "bbox": [ + 106, + 126, + 126, + 138 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 126, + 127, + 146, + 137 + ], + "score": 0.87, + "content": "2 \\times 2", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 126, + 299, + 138 + ], + "score": 1.0, + "content": "convolution, which limits the interac-", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 2, + "bbox_fs": [ + 105, + 82, + 299, + 138 + ] + }, + { + "type": "table", + "bbox": [ + 304, + 96, + 504, + 138 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 319, + 84, + 487, + 95 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 319, + 83, + 488, + 97 + ], + "spans": [ + { + "bbox": [ + 319, + 83, + 488, + 97 + ], + "score": 1.0, + "content": "Table 9: Different downsampling approaches.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "table_body", + "bbox": [ + 304, + 96, + 504, + 138 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 304, + 96, + 504, + 138 + ], + "spans": [ + { + "bbox": [ + 304, + 96, + 504, + 138 + ], + "score": 0.969, + "html": "
DownsamplingParamsFLOPs (G)IN1K Acc. (%)
2×2 convolution32.15.582.3
3×3 convolution34.15.782.5
3×3 depth-wise conv.30.65.382.6
", + "type": "table", + "image_path": "34b8c71a505dd7e761c717f1d52783fa1ed69d59742d33ed118fc72e59a2e539.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 304, + 96, + 504, + 110.0 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 304, + 110.0, + 504, + 124.0 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 304, + 124.0, + 504, + 138.0 + ], + "spans": [], + "index": 9 + } + ] + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 108, + 138, + 507, + 160 + ], + "lines": [ + { + "bbox": [ + 106, + 137, + 505, + 150 + ], + "spans": [ + { + "bbox": [ + 106, + 137, + 505, + 150 + ], + "score": 1.0, + "content": "tion among local tokens. Moreover, using regular convolution does not improve the performance but", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 147, + 507, + 162 + ], + "spans": [ + { + "bbox": [ + 105, + 147, + 319, + 162 + ], + "score": 1.0, + "content": "increases computations and parameters. Thus, we use", + "type": "text" + }, + { + "bbox": [ + 319, + 149, + 339, + 159 + ], + "score": 0.86, + "content": "3 \\times 3", + "type": "inline_equation" + }, + { + "bbox": [ + 339, + 147, + 507, + 162 + ], + "score": 1.0, + "content": "depthwise convolution for downsampling.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 8.5, + "bbox_fs": [ + 105, + 137, + 507, + 162 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 165, + 297, + 221 + ], + "lines": [ + { + "bbox": [ + 106, + 165, + 298, + 177 + ], + "spans": [ + { + "bbox": [ + 106, + 165, + 298, + 177 + ], + "score": 1.0, + "content": "Weight sharing between RSA and LSA as", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 176, + 298, + 188 + ], + "spans": [ + { + "bbox": [ + 106, + 176, + 298, + 188 + ], + "score": 1.0, + "content": "well as downsampling. Table 10 shows the re-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 187, + 297, + 199 + ], + "spans": [ + { + "bbox": [ + 106, + 187, + 297, + 199 + ], + "score": 1.0, + "content": "sults with and without weight sharing. As seen", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 197, + 297, + 210 + ], + "spans": [ + { + "bbox": [ + 105, + 197, + 297, + 210 + ], + "score": 1.0, + "content": "from the table, using separated weights for RSA", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 208, + 298, + 223 + ], + "spans": [ + { + "bbox": [ + 106, + 208, + 298, + 223 + ], + "score": 1.0, + "content": "and LSA only slightly improves the accuracy", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 13, + "bbox_fs": [ + 105, + 165, + 298, + 223 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 221, + 505, + 243 + ], + "lines": [ + { + "bbox": [ + 106, + 220, + 505, + 232 + ], + "spans": [ + { + "bbox": [ + 106, + 220, + 505, + 232 + ], + "score": 1.0, + "content": "but significantly increases the model size. This suggests sharing parameters between RSA and LSA", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 231, + 372, + 243 + ], + "spans": [ + { + "bbox": [ + 105, + 231, + 372, + 243 + ], + "score": 1.0, + "content": "suffices for learning local and global information in our approach.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5, + "bbox_fs": [ + 105, + 220, + 505, + 243 + ] + }, + { + "type": "table", + "bbox": [ + 304, + 179, + 504, + 217 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 356, + 167, + 452, + 178 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 354, + 165, + 454, + 181 + ], + "spans": [ + { + "bbox": [ + 354, + 165, + 454, + 181 + ], + "score": 1.0, + "content": "Table 10: Weight sharing.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "table_body", + "bbox": [ + 304, + 179, + 504, + 217 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 304, + 179, + 504, + 217 + ], + "spans": [ + { + "bbox": [ + 304, + 179, + 504, + 217 + ], + "score": 0.96, + "html": "
Weight sharingParamsFLOPs (G)IN1K Acc. (%)
N40.45.382.7
Y30.65.382.6
", + "type": "table", + "image_path": "ade46fff21f8473d5206e19825c1ca1182433ecda061114ef2b68a3f4999d873.jpg" + } + ] + } + ], + "index": 19.5, + "virtual_lines": [ + { + "bbox": [ + 304, + 179, + 504, + 198.0 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 304, + 198.0, + 504, + 217.0 + ], + "spans": [], + "index": 20 + } + ] + } + ], + "index": 18.75 + }, + { + "type": "text", + "bbox": [ + 106, + 248, + 297, + 303 + ], + "lines": [ + { + "bbox": [ + 106, + 247, + 298, + 259 + ], + "spans": [ + { + "bbox": [ + 106, + 247, + 298, + 259 + ], + "score": 1.0, + "content": "Keys in LSA. Table 11 studies the number of", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 258, + 297, + 271 + ], + "spans": [ + { + "bbox": [ + 105, + 258, + 297, + 271 + ], + "score": 1.0, + "content": "regional tokens in LSA. When using all regional", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 270, + 297, + 281 + ], + "spans": [ + { + "bbox": [ + 106, + 270, + 297, + 281 + ], + "score": 1.0, + "content": "tokens in the LSA, it provides the local tokens", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 281, + 298, + 293 + ], + "spans": [ + { + "bbox": [ + 106, + 281, + 298, + 293 + ], + "score": 1.0, + "content": "the possibility to explore all global information.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 291, + 297, + 304 + ], + "spans": [ + { + "bbox": [ + 105, + 291, + 259, + 304 + ], + "score": 1.0, + "content": "Nonetheless, this model only improves", + "type": "text" + }, + { + "bbox": [ + 259, + 292, + 281, + 302 + ], + "score": 0.87, + "content": "0 . 1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 282, + 291, + 297, + 304 + ], + "score": 1.0, + "content": "but", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 23, + "bbox_fs": [ + 105, + 247, + 298, + 304 + ] + }, + { + "type": "table", + "bbox": [ + 304, + 262, + 504, + 298 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 359, + 250, + 448, + 261 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 358, + 249, + 450, + 262 + ], + "spans": [ + { + "bbox": [ + 358, + 249, + 450, + 262 + ], + "score": 1.0, + "content": "Table 11: Keys in LSA.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 26 + }, + { + "type": "table_body", + "bbox": [ + 304, + 262, + 504, + 298 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 304, + 262, + 504, + 298 + ], + "spans": [ + { + "bbox": [ + 304, + 262, + 504, + 298 + ], + "score": 0.967, + "html": "
Regional tokensParamsFLOPs (G)IN1K Acc. (%)
All30.66.082.7
Corresponding30.65.382.6
", + "type": "table", + "image_path": "02feafa313da8136ea97cf1fc43c19477a597fbad9be75bc30c0bbc917625e70.jpg" + } + ] + } + ], + "index": 27.5, + "virtual_lines": [ + { + "bbox": [ + 304, + 262, + 504, + 280.0 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 304, + 280.0, + 504, + 298.0 + ], + "spans": [], + "index": 28 + } + ] + } + ], + "index": 26.75 + }, + { + "type": "text", + "bbox": [ + 106, + 303, + 505, + 325 + ], + "lines": [ + { + "bbox": [ + 105, + 303, + 504, + 315 + ], + "spans": [ + { + "bbox": [ + 105, + 303, + 127, + 315 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 127, + 303, + 147, + 313 + ], + "score": 0.87, + "content": "13 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 147, + 303, + 504, + 315 + ], + "score": 1.0, + "content": "more computations, which suggests that when considering the regional tokens, the one", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 313, + 412, + 326 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 412, + 326 + ], + "score": 1.0, + "content": "associated with the current region is sufficient to achieve good performance.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5, + "bbox_fs": [ + 105, + 303, + 504, + 326 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 331, + 297, + 396 + ], + "lines": [ + { + "bbox": [ + 106, + 329, + 298, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 329, + 298, + 343 + ], + "score": 1.0, + "content": "Position information. Table 12 compares per-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 342, + 297, + 353 + ], + "spans": [ + { + "bbox": [ + 106, + 342, + 297, + 353 + ], + "score": 1.0, + "content": "formance of different combinations of absolute", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 353, + 297, + 365 + ], + "spans": [ + { + "bbox": [ + 105, + 353, + 297, + 365 + ], + "score": 1.0, + "content": "position embedding (APE) and relative position", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 363, + 298, + 374 + ], + "spans": [ + { + "bbox": [ + 106, + 363, + 298, + 374 + ], + "score": 1.0, + "content": "bias (RPB). The models with RPB achieve bet-", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 374, + 299, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 374, + 299, + 387 + ], + "score": 1.0, + "content": "ter accuracy with similar FLOPs and parameters.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 385, + 298, + 398 + ], + "spans": [ + { + "bbox": [ + 105, + 385, + 298, + 398 + ], + "score": 1.0, + "content": "Although the model with APE could slightly im-", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 33.5, + "bbox_fs": [ + 105, + 329, + 299, + 398 + ] + }, + { + "type": "table", + "bbox": [ + 304, + 344, + 504, + 391 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 328, + 333, + 480, + 343 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 327, + 331, + 481, + 344 + ], + "spans": [ + { + "bbox": [ + 327, + 331, + 481, + 344 + ], + "score": 1.0, + "content": "Table 12: Different position information.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "table_body", + "bbox": [ + 304, + 344, + 504, + 391 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 304, + 344, + 504, + 391 + ], + "spans": [ + { + "bbox": [ + 304, + 344, + 504, + 391 + ], + "score": 0.971, + "html": "
Abs. pos.Rel. pos.Params (M)FLOPs (G)IN1K Acc. (%)
NN30.65.382.4
YN30.95.382.2
YY30.95.382.7
NY30.65.382.6
", + "type": "table", + "image_path": "5fde92d3500b48401daf0f3493d4bdca1755d22ee0419743ec863dc937e4e07f.jpg" + } + ] + } + ], + "index": 39, + "virtual_lines": [ + { + "bbox": [ + 304, + 344, + 504, + 359.6666666666667 + ], + "spans": [], + "index": 38 + }, + { + "bbox": [ + 304, + 359.6666666666667, + 504, + 375.33333333333337 + ], + "spans": [], + "index": 39 + }, + { + "bbox": [ + 304, + 375.33333333333337, + 504, + 391.00000000000006 + ], + "spans": [], + "index": 40 + } + ] + } + ], + "index": 38.0 + }, + { + "type": "text", + "bbox": [ + 107, + 397, + 505, + 419 + ], + "lines": [ + { + "bbox": [ + 105, + 396, + 506, + 409 + ], + "spans": [ + { + "bbox": [ + 105, + 396, + 506, + 409 + ], + "score": 1.0, + "content": "prove the performance, it limits the model to run at a fixed resolution, which is not suitable for vision", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 407, + 506, + 420 + ], + "spans": [ + { + "bbox": [ + 105, + 407, + 506, + 420 + ], + "score": 1.0, + "content": "tasks where image size could be varied. Thus, we only adopt the relative position bias in our models.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41.5, + "bbox_fs": [ + 105, + 396, + 506, + 420 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 424, + 298, + 479 + ], + "lines": [ + { + "bbox": [ + 106, + 423, + 297, + 437 + ], + "spans": [ + { + "bbox": [ + 106, + 423, + 297, + 437 + ], + "score": 1.0, + "content": "Overlapped windows. Table 13 compares the", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 434, + 299, + 448 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 299, + 448 + ], + "score": 1.0, + "content": "results with and without overlapped windows.", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 446, + 298, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 298, + 459 + ], + "score": 1.0, + "content": "The overlapping ratio of neighboring windows", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 457, + 298, + 469 + ], + "spans": [ + { + "bbox": [ + 105, + 457, + 116, + 469 + ], + "score": 1.0, + "content": "is", + "type": "text" + }, + { + "bbox": [ + 117, + 457, + 137, + 468 + ], + "score": 0.86, + "content": "50 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 457, + 298, + 469 + ], + "score": 1.0, + "content": "and hence this model allows more in-", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 468, + 298, + 481 + ], + "spans": [ + { + "bbox": [ + 105, + 468, + 298, + 481 + ], + "score": 1.0, + "content": "teractions among local tokens. It improves", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 45, + "bbox_fs": [ + 105, + 423, + 299, + 481 + ] + }, + { + "type": "table", + "bbox": [ + 304, + 438, + 504, + 475 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 340, + 426, + 468, + 437 + ], + "group_id": 4, + "lines": [ + { + "bbox": [ + 339, + 424, + 469, + 438 + ], + "spans": [ + { + "bbox": [ + 339, + 424, + 469, + 438 + ], + "score": 1.0, + "content": "Table 13: Overlapped windows.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 48 + }, + { + "type": "table_body", + "bbox": [ + 304, + 438, + 504, + 475 + ], + "group_id": 4, + "lines": [ + { + "bbox": [ + 304, + 438, + 504, + 475 + ], + "spans": [ + { + "bbox": [ + 304, + 438, + 504, + 475 + ], + "score": 0.965, + "html": "
ModelParamsFLOPs (G)IN1K Acc. (%)
overlapped30.618.882.8
non-overlapped30.65.382.6
", + "type": "table", + "image_path": "cbd48a837d0fcb2bd0b8853bf01a0e4804700eca5e82b229be0695ca1dc25d16.jpg" + } + ] + } + ], + "index": 49.5, + "virtual_lines": [ + { + "bbox": [ + 304, + 438, + 504, + 456.5 + ], + "spans": [], + "index": 49 + }, + { + "bbox": [ + 304, + 456.5, + 504, + 475.0 + ], + "spans": [], + "index": 50 + } + ] + } + ], + "index": 48.75 + }, + { + "type": "text", + "bbox": [ + 106, + 479, + 505, + 502 + ], + "lines": [ + { + "bbox": [ + 105, + 478, + 505, + 492 + ], + "spans": [ + { + "bbox": [ + 105, + 478, + 173, + 492 + ], + "score": 1.0, + "content": "RegionViT-S by", + "type": "text" + }, + { + "bbox": [ + 173, + 479, + 195, + 489 + ], + "score": 0.86, + "content": "0 . 2 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 195, + 478, + 305, + 492 + ], + "score": 1.0, + "content": "but increases the FLOPs by", + "type": "text" + }, + { + "bbox": [ + 305, + 479, + 326, + 490 + ], + "score": 0.87, + "content": "3 . 5 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 326, + 478, + 505, + 492 + ], + "score": 1.0, + "content": ". This suggests that the information exchange", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 105, + 489, + 451, + 503 + ], + "spans": [ + { + "bbox": [ + 105, + 489, + 451, + 503 + ], + "score": 1.0, + "content": "between the border of windows is sufficiently covered by the proposed R2L attention.", + "type": "text" + } + ], + "index": 52 + } + ], + "index": 51.5, + "bbox_fs": [ + 105, + 478, + 505, + 503 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 507, + 297, + 562 + ], + "lines": [ + { + "bbox": [ + 106, + 506, + 297, + 518 + ], + "spans": [ + { + "bbox": [ + 106, + 506, + 297, + 518 + ], + "score": 1.0, + "content": "Comparison between R2L attention and", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 106, + 517, + 297, + 529 + ], + "spans": [ + { + "bbox": [ + 106, + 517, + 297, + 529 + ], + "score": 1.0, + "content": "shifted-window attention. Table 14 shows", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 105, + 529, + 298, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 529, + 298, + 540 + ], + "score": 1.0, + "content": "the comparison between shifted-window atten-", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 105, + 539, + 299, + 551 + ], + "spans": [ + { + "bbox": [ + 105, + 539, + 299, + 551 + ], + "score": 1.0, + "content": "tion (Liu et al., 2021) and proposed R2L atten-", + "type": "text" + } + ], + "index": 56 + }, + { + "bbox": [ + 106, + 551, + 297, + 562 + ], + "spans": [ + { + "bbox": [ + 106, + 551, + 297, + 562 + ], + "score": 1.0, + "content": "tion. R2L attention outperforms shifted-window", + "type": "text" + } + ], + "index": 57 + } + ], + "index": 55, + "bbox_fs": [ + 105, + 506, + 299, + 562 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 562, + 504, + 584 + ], + "lines": [ + { + "bbox": [ + 106, + 561, + 505, + 574 + ], + "spans": [ + { + "bbox": [ + 106, + 561, + 159, + 574 + ], + "score": 1.0, + "content": "attention by", + "type": "text" + }, + { + "bbox": [ + 159, + 562, + 182, + 573 + ], + "score": 0.86, + "content": "0 . 7 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 182, + 561, + 505, + 574 + ], + "score": 1.0, + "content": ", which shows that our proposed R2L attention can effectively model global", + "type": "text" + } + ], + "index": 58 + }, + { + "bbox": [ + 105, + 572, + 277, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 277, + 586 + ], + "score": 1.0, + "content": "information to achieve good performance.", + "type": "text" + } + ], + "index": 59 + } + ], + "index": 58.5, + "bbox_fs": [ + 105, + 561, + 505, + 586 + ] + }, + { + "type": "table", + "bbox": [ + 304, + 521, + 504, + 558 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 312, + 509, + 493, + 519 + ], + "group_id": 5, + "lines": [ + { + "bbox": [ + 313, + 508, + 495, + 521 + ], + "spans": [ + { + "bbox": [ + 313, + 508, + 495, + 521 + ], + "score": 1.0, + "content": "Table 14: Comparison of different attentions.", + "type": "text" + } + ], + "index": 60 + } + ], + "index": 60 + }, + { + "type": "table_body", + "bbox": [ + 304, + 521, + 504, + 558 + ], + "group_id": 5, + "lines": [ + { + "bbox": [ + 304, + 521, + 504, + 558 + ], + "spans": [ + { + "bbox": [ + 304, + 521, + 504, + 558 + ], + "score": 0.963, + "html": "
ModelParamsFLOPs (G)IN1K Acc. (%)
Shifted-window29.04.581.3
R2L32.15.282.0
", + "type": "table", + "image_path": "bd9e65592297ebcfc30c495d4720dbd19d12d0cc359d374492af69bfeee00aa5.jpg" + } + ] + } + ], + "index": 61.5, + "virtual_lines": [ + { + "bbox": [ + 304, + 521, + 504, + 539.5 + ], + "spans": [], + "index": 61 + }, + { + "bbox": [ + 304, + 539.5, + 504, + 558.0 + ], + "spans": [], + "index": 62 + } + ] + } + ], + "index": 60.75 + }, + { + "type": "title", + "bbox": [ + 108, + 600, + 195, + 613 + ], + "lines": [ + { + "bbox": [ + 105, + 599, + 197, + 616 + ], + "spans": [ + { + "bbox": [ + 105, + 599, + 197, + 616 + ], + "score": 1.0, + "content": "5 CONCLUSION", + "type": "text" + } + ], + "index": 63 + } + ], + "index": 63 + }, + { + "type": "text", + "bbox": [ + 106, + 625, + 505, + 714 + ], + "lines": [ + { + "bbox": [ + 105, + 624, + 506, + 638 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 506, + 638 + ], + "score": 1.0, + "content": "In this paper, we propose a new ViT architecture that exploits the pyramid structure used in most of", + "type": "text" + } + ], + "index": 64 + }, + { + "bbox": [ + 105, + 635, + 506, + 649 + ], + "spans": [ + { + "bbox": [ + 105, + 635, + 506, + 649 + ], + "score": 1.0, + "content": "CNNs to provide multi-scale features and hence, the models can be more easily extended to different", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 106, + 647, + 505, + 659 + ], + "spans": [ + { + "bbox": [ + 106, + 647, + 505, + 659 + ], + "score": 1.0, + "content": "vision applications, like object detection. Moreover, the proposed regional-to-local attention relaxes", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 106, + 658, + 505, + 671 + ], + "spans": [ + { + "bbox": [ + 106, + 658, + 505, + 671 + ], + "score": 1.0, + "content": "the memory overhead of performing self-attention on the fine-grained image tokens by limiting the", + "type": "text" + } + ], + "index": 67 + }, + { + "bbox": [ + 105, + 669, + 506, + 682 + ], + "spans": [ + { + "bbox": [ + 105, + 669, + 506, + 682 + ], + "score": 1.0, + "content": "scope of attention but still keeping the capability to explore global information. Extensive experiments", + "type": "text" + } + ], + "index": 68 + }, + { + "bbox": [ + 106, + 681, + 505, + 693 + ], + "spans": [ + { + "bbox": [ + 106, + 681, + 505, + 693 + ], + "score": 1.0, + "content": "on several standard benchmark datasets well demonstrate that our proposed models outperform or are", + "type": "text" + } + ], + "index": 69 + }, + { + "bbox": [ + 106, + 691, + 506, + 704 + ], + "spans": [ + { + "bbox": [ + 106, + 691, + 506, + 704 + ], + "score": 1.0, + "content": "on par with many concurrent ViT variants on four vision applications, including image classification,", + "type": "text" + } + ], + "index": 70 + }, + { + "bbox": [ + 105, + 702, + 416, + 715 + ], + "spans": [ + { + "bbox": [ + 105, + 702, + 416, + 715 + ], + "score": 1.0, + "content": "object and keypoint detection, semantic segmentation and action recognition.", + "type": "text" + } + ], + "index": 71 + } + ], + "index": 67.5, + "bbox_fs": [ + 105, + 624, + 506, + 715 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 81, + 210, + 94 + ], + "lines": [ + { + "bbox": [ + 106, + 81, + 213, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 213, + 95 + ], + "score": 1.0, + "content": "ETHICS STATEMENT", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 106, + 505, + 216 + ], + "lines": [ + { + "bbox": [ + 106, + 106, + 506, + 118 + ], + "spans": [ + { + "bbox": [ + 106, + 106, + 506, + 118 + ], + "score": 1.0, + "content": "Our work introduces a memory-friendly and efficient self-attention method for transformer models,", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 117, + 505, + 130 + ], + "spans": [ + { + "bbox": [ + 105, + 117, + 505, + 130 + ], + "score": 1.0, + "content": "which have wide-ranging applications to image classification, object detection, human activity", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 128, + 505, + 141 + ], + "spans": [ + { + "bbox": [ + 105, + 128, + 505, + 141 + ], + "score": 1.0, + "content": "recognition, etc. By improving efficiency, our work can have a positive effect on these applications", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 139, + 505, + 152 + ], + "spans": [ + { + "bbox": [ + 106, + 139, + 505, + 152 + ], + "score": 1.0, + "content": "in society, allowing faster processing. E.g., this could enable faster responses to detected events", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 150, + 505, + 163 + ], + "spans": [ + { + "bbox": [ + 106, + 150, + 505, + 163 + ], + "score": 1.0, + "content": "such as medical emergencies or detecting defects in manufacturing parts. Lower computation costs", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 160, + 505, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 160, + 505, + 174 + ], + "score": 1.0, + "content": "could also save energy and therefore have a positive impact on the environment. Negative impacts", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 170, + 506, + 186 + ], + "spans": [ + { + "bbox": [ + 105, + 170, + 506, + 186 + ], + "score": 1.0, + "content": "of our research are difficult to predict, however, it shares many of the pitfalls associated with deep", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 182, + 506, + 196 + ], + "spans": [ + { + "bbox": [ + 105, + 182, + 506, + 196 + ], + "score": 1.0, + "content": "classification models. E.g., Image and video classification systems have negative implications on", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 194, + 506, + 207 + ], + "spans": [ + { + "bbox": [ + 106, + 194, + 506, + 207 + ], + "score": 1.0, + "content": "privacy and could be used by malicious actors or governments to infringe on the privacy of citizens.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 205, + 483, + 217 + ], + "spans": [ + { + "bbox": [ + 106, + 205, + 483, + 217 + ], + "score": 1.0, + "content": "Future research into private and ethical aspects of visual recognition is an important direction.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 5.5 + }, + { + "type": "title", + "bbox": [ + 108, + 232, + 267, + 245 + ], + "lines": [ + { + "bbox": [ + 106, + 232, + 270, + 246 + ], + "spans": [ + { + "bbox": [ + 106, + 232, + 270, + 246 + ], + "score": 1.0, + "content": "REPRODUCIBILITY STATEMENT", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 107, + 257, + 505, + 324 + ], + "lines": [ + { + "bbox": [ + 105, + 256, + 506, + 271 + ], + "spans": [ + { + "bbox": [ + 105, + 256, + 506, + 271 + ], + "score": 1.0, + "content": "In order to reproduce our results, we describe the details of the hyperparameters used in our training", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 268, + 506, + 281 + ], + "spans": [ + { + "bbox": [ + 105, + 268, + 506, + 281 + ], + "score": 1.0, + "content": "in Appendix A for all the tasks and we also attach our codes for image classification as part of", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 279, + 505, + 292 + ], + "spans": [ + { + "bbox": [ + 105, + 279, + 505, + 292 + ], + "score": 1.0, + "content": "the supplemental material. We will publicly release all the codes and models after acceptance. In", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 290, + 506, + 303 + ], + "spans": [ + { + "bbox": [ + 105, + 290, + 506, + 303 + ], + "score": 1.0, + "content": "the attached codes, we set all the hyperparameters as their default value, so that users can re-run", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 301, + 506, + 313 + ], + "spans": [ + { + "bbox": [ + 105, + 301, + 506, + 313 + ], + "score": 1.0, + "content": "our experiments with exactly the same hyperparameters. The README file also describes the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 313, + 404, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 404, + 324 + ], + "score": 1.0, + "content": "requirement to build necessary Python environment for running the codes.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 14.5 + } + ], + "page_idx": 9, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 752, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 81, + 210, + 94 + ], + "lines": [ + { + "bbox": [ + 106, + 81, + 213, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 213, + 95 + ], + "score": 1.0, + "content": "ETHICS STATEMENT", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 106, + 505, + 216 + ], + "lines": [ + { + "bbox": [ + 106, + 106, + 506, + 118 + ], + "spans": [ + { + "bbox": [ + 106, + 106, + 506, + 118 + ], + "score": 1.0, + "content": "Our work introduces a memory-friendly and efficient self-attention method for transformer models,", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 117, + 505, + 130 + ], + "spans": [ + { + "bbox": [ + 105, + 117, + 505, + 130 + ], + "score": 1.0, + "content": "which have wide-ranging applications to image classification, object detection, human activity", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 128, + 505, + 141 + ], + "spans": [ + { + "bbox": [ + 105, + 128, + 505, + 141 + ], + "score": 1.0, + "content": "recognition, etc. By improving efficiency, our work can have a positive effect on these applications", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 139, + 505, + 152 + ], + "spans": [ + { + "bbox": [ + 106, + 139, + 505, + 152 + ], + "score": 1.0, + "content": "in society, allowing faster processing. E.g., this could enable faster responses to detected events", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 150, + 505, + 163 + ], + "spans": [ + { + "bbox": [ + 106, + 150, + 505, + 163 + ], + "score": 1.0, + "content": "such as medical emergencies or detecting defects in manufacturing parts. Lower computation costs", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 160, + 505, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 160, + 505, + 174 + ], + "score": 1.0, + "content": "could also save energy and therefore have a positive impact on the environment. Negative impacts", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 170, + 506, + 186 + ], + "spans": [ + { + "bbox": [ + 105, + 170, + 506, + 186 + ], + "score": 1.0, + "content": "of our research are difficult to predict, however, it shares many of the pitfalls associated with deep", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 182, + 506, + 196 + ], + "spans": [ + { + "bbox": [ + 105, + 182, + 506, + 196 + ], + "score": 1.0, + "content": "classification models. E.g., Image and video classification systems have negative implications on", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 194, + 506, + 207 + ], + "spans": [ + { + "bbox": [ + 106, + 194, + 506, + 207 + ], + "score": 1.0, + "content": "privacy and could be used by malicious actors or governments to infringe on the privacy of citizens.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 205, + 483, + 217 + ], + "spans": [ + { + "bbox": [ + 106, + 205, + 483, + 217 + ], + "score": 1.0, + "content": "Future research into private and ethical aspects of visual recognition is an important direction.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 5.5, + "bbox_fs": [ + 105, + 106, + 506, + 217 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 232, + 267, + 245 + ], + "lines": [ + { + "bbox": [ + 106, + 232, + 270, + 246 + ], + "spans": [ + { + "bbox": [ + 106, + 232, + 270, + 246 + ], + "score": 1.0, + "content": "REPRODUCIBILITY STATEMENT", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 107, + 257, + 505, + 324 + ], + "lines": [ + { + "bbox": [ + 105, + 256, + 506, + 271 + ], + "spans": [ + { + "bbox": [ + 105, + 256, + 506, + 271 + ], + "score": 1.0, + "content": "In order to reproduce our results, we describe the details of the hyperparameters used in our training", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 268, + 506, + 281 + ], + "spans": [ + { + "bbox": [ + 105, + 268, + 506, + 281 + ], + "score": 1.0, + "content": "in Appendix A for all the tasks and we also attach our codes for image classification as part of", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 279, + 505, + 292 + ], + "spans": [ + { + "bbox": [ + 105, + 279, + 505, + 292 + ], + "score": 1.0, + "content": "the supplemental material. We will publicly release all the codes and models after acceptance. In", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 290, + 506, + 303 + ], + "spans": [ + { + "bbox": [ + 105, + 290, + 506, + 303 + ], + "score": 1.0, + "content": "the attached codes, we set all the hyperparameters as their default value, so that users can re-run", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 301, + 506, + 313 + ], + "spans": [ + { + "bbox": [ + 105, + 301, + 506, + 313 + ], + "score": 1.0, + "content": "our experiments with exactly the same hyperparameters. The README file also describes the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 313, + 404, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 404, + 324 + ], + "score": 1.0, + "content": "requirement to build necessary Python environment for running the codes.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 14.5, + "bbox_fs": [ + 105, + 256, + 506, + 324 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 82, + 176, + 93 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 176, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 176, + 95 + ], + "score": 1.0, + "content": "REFERENCES", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 100, + 504, + 133 + ], + "lines": [ + { + "bbox": [ + 105, + 99, + 506, + 112 + ], + "spans": [ + { + "bbox": [ + 105, + 99, + 506, + 112 + ], + "score": 1.0, + "content": "Irwan Bello. Lambdanetworks: Modeling long-range interactions without attention. In International", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 116, + 110, + 506, + 124 + ], + "spans": [ + { + "bbox": [ + 116, + 110, + 506, + 124 + ], + "score": 1.0, + "content": "Conference on Learning Representations, 2021. URL https://openreview.net/forum?", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 116, + 122, + 205, + 135 + ], + "spans": [ + { + "bbox": [ + 116, + 122, + 205, + 135 + ], + "score": 1.0, + "content": "id=xTJEN-ggl1b.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2 + }, + { + "type": "text", + "bbox": [ + 108, + 140, + 504, + 174 + ], + "lines": [ + { + "bbox": [ + 106, + 140, + 505, + 153 + ], + "spans": [ + { + "bbox": [ + 106, + 140, + 505, + 153 + ], + "score": 1.0, + "content": "Irwan Bello, Barret Zoph, Ashish Vaswani, Jonathon Shlens, and Quoc V Le. Attention augmented", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 151, + 505, + 164 + ], + "spans": [ + { + "bbox": [ + 115, + 151, + 505, + 164 + ], + "score": 1.0, + "content": "convolutional networks. In Proceedings of the IEEE/CVF International Conference on Computer", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 116, + 162, + 235, + 175 + ], + "spans": [ + { + "bbox": [ + 116, + 162, + 235, + 175 + ], + "score": 1.0, + "content": "Vision, pp. 3286–3295, 2019.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 105, + 180, + 505, + 204 + ], + "lines": [ + { + "bbox": [ + 106, + 180, + 506, + 193 + ], + "spans": [ + { + "bbox": [ + 106, + 180, + 506, + 193 + ], + "score": 1.0, + "content": "Gedas Bertasius, Heng Wang, and Lorenzo Torresani. Is Space-Time Attention All You Need for", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 116, + 191, + 315, + 204 + ], + "spans": [ + { + "bbox": [ + 116, + 191, + 315, + 204 + ], + "score": 1.0, + "content": "Video Understanding? arXiv.org, February 2021.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 105, + 209, + 505, + 233 + ], + "lines": [ + { + "bbox": [ + 106, + 210, + 505, + 223 + ], + "spans": [ + { + "bbox": [ + 106, + 210, + 505, + 223 + ], + "score": 1.0, + "content": "Andrew Brock, Soham De, Samuel L Smith, and Karen Simonyan. High-Performance Large-Scale", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 115, + 221, + 346, + 233 + ], + "spans": [ + { + "bbox": [ + 115, + 221, + 346, + 233 + ], + "score": 1.0, + "content": "Image Recognition Without Normalization. arXiv, 2021.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "text", + "bbox": [ + 106, + 239, + 505, + 262 + ], + "lines": [ + { + "bbox": [ + 106, + 240, + 505, + 252 + ], + "spans": [ + { + "bbox": [ + 106, + 240, + 505, + 252 + ], + "score": 1.0, + "content": "Chun-Fu Chen, Quanfu Fan, and Rameswar Panda. CrossViT: Cross-Attention Multi-Scale Vision", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 116, + 250, + 370, + 263 + ], + "spans": [ + { + "bbox": [ + 116, + 250, + 370, + 263 + ], + "score": 1.0, + "content": "Transformer for Image Classification. arXiv.org, March 2021a.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5 + }, + { + "type": "text", + "bbox": [ + 106, + 268, + 505, + 292 + ], + "lines": [ + { + "bbox": [ + 106, + 268, + 505, + 282 + ], + "spans": [ + { + "bbox": [ + 106, + 268, + 505, + 282 + ], + "score": 1.0, + "content": "Zhengsu Chen, Lingxi Xie, Jianwei Niu, Xuefeng Liu, Longhui Wei, and Qi Tian. Visformer: The", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 116, + 279, + 328, + 293 + ], + "spans": [ + { + "bbox": [ + 116, + 279, + 328, + 293 + ], + "score": 1.0, + "content": "Vision-friendly Transformer. arxiv.org, April 2021b.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5 + }, + { + "type": "text", + "bbox": [ + 108, + 298, + 505, + 332 + ], + "lines": [ + { + "bbox": [ + 106, + 298, + 506, + 311 + ], + "spans": [ + { + "bbox": [ + 106, + 298, + 506, + 311 + ], + "score": 1.0, + "content": "Xiangxiang Chu, Zhi Tian, Yuqing Wang, Bo Zhang, Haibing Ren, Xiaolin Wei, Huaxia Xia, and", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 308, + 507, + 323 + ], + "spans": [ + { + "bbox": [ + 115, + 308, + 507, + 323 + ], + "score": 1.0, + "content": "Chunhua Shen. Twins: Revisiting Spatial Attention Design in Vision Transformers. arXiv.org,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 114, + 320, + 169, + 334 + ], + "spans": [ + { + "bbox": [ + 114, + 320, + 169, + 334 + ], + "score": 1.0, + "content": "April 2021a.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 105, + 338, + 505, + 362 + ], + "lines": [ + { + "bbox": [ + 106, + 339, + 506, + 351 + ], + "spans": [ + { + "bbox": [ + 106, + 339, + 506, + 351 + ], + "score": 1.0, + "content": "Xiangxiang Chu, Zhi Tian, Bo Zhang, Xinlong Wang, Xiaolin Wei, Huaxia Xia, and Chunhua Shen.", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 117, + 350, + 411, + 362 + ], + "spans": [ + { + "bbox": [ + 117, + 350, + 411, + 362 + ], + "score": 1.0, + "content": "Conditional Positional Encodings for Vision Transformers. arXiv, 2021b.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18.5 + }, + { + "type": "text", + "bbox": [ + 106, + 367, + 506, + 412 + ], + "lines": [ + { + "bbox": [ + 106, + 368, + 506, + 381 + ], + "spans": [ + { + "bbox": [ + 106, + 368, + 506, + 381 + ], + "score": 1.0, + "content": "Ekin Dogus Cubuk, Barret Zoph, Jon Shlens, and Quoc Le. RandAugment: Practical Automated", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 115, + 379, + 506, + 392 + ], + "spans": [ + { + "bbox": [ + 115, + 379, + 506, + 392 + ], + "score": 1.0, + "content": "Data Augmentation with a Reduced Search Space. In H Larochelle, M Ranzato, R Hadsell, M F", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 390, + 507, + 403 + ], + "spans": [ + { + "bbox": [ + 115, + 390, + 507, + 403 + ], + "score": 1.0, + "content": "Balcan, and H Lin (eds.), Advances in Neural Information Processing Systems, pp. 18613–18624.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 117, + 402, + 239, + 413 + ], + "spans": [ + { + "bbox": [ + 117, + 402, + 239, + 413 + ], + "score": 1.0, + "content": "Curran Associates, Inc., 2020.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 21.5 + }, + { + "type": "text", + "bbox": [ + 108, + 419, + 505, + 453 + ], + "lines": [ + { + "bbox": [ + 106, + 419, + 507, + 433 + ], + "spans": [ + { + "bbox": [ + 106, + 419, + 507, + 433 + ], + "score": 1.0, + "content": "Stephane d’Ascoli, Hugo Touvron, Matthew Leavitt, Ari Morcos, Giulio Biroli, and Levent Sagun. ´", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 430, + 507, + 444 + ], + "spans": [ + { + "bbox": [ + 115, + 430, + 507, + 444 + ], + "score": 1.0, + "content": "ConViT: Improving Vision Transformers with Soft Convolutional Inductive Biases. arXiv.org,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 115, + 440, + 265, + 456 + ], + "spans": [ + { + "bbox": [ + 115, + 440, + 265, + 456 + ], + "score": 1.0, + "content": "cs.CV, 03 2021. URL arXiv.org.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 106, + 459, + 505, + 493 + ], + "lines": [ + { + "bbox": [ + 105, + 459, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 505, + 473 + ], + "score": 1.0, + "content": "Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 115, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition,", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 115, + 482, + 218, + 494 + ], + "spans": [ + { + "bbox": [ + 115, + 482, + 218, + 494 + ], + "score": 1.0, + "content": "pp. 248–255. Ieee, 2009.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 106, + 500, + 506, + 567 + ], + "lines": [ + { + "bbox": [ + 105, + 500, + 506, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 500, + 506, + 513 + ], + "score": 1.0, + "content": "Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: Pre-training of deep", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 510, + 506, + 524 + ], + "spans": [ + { + "bbox": [ + 115, + 510, + 506, + 524 + ], + "score": 1.0, + "content": "bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 521, + 506, + 537 + ], + "spans": [ + { + "bbox": [ + 115, + 521, + 506, + 537 + ], + "score": 1.0, + "content": "the North American Chapter of the Association for Computational Linguistics: Human Language", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 115, + 533, + 506, + 546 + ], + "spans": [ + { + "bbox": [ + 115, + 533, + 506, + 546 + ], + "score": 1.0, + "content": "Technologies, Volume 1 (Long and Short Papers), pp. 4171–4186, Minneapolis, Minnesota, June", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 543, + 507, + 557 + ], + "spans": [ + { + "bbox": [ + 115, + 543, + 507, + 557 + ], + "score": 1.0, + "content": "2019. Association for Computational Linguistics. doi: 10.18653/v1/N19-1423. URL https:", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 117, + 556, + 330, + 567 + ], + "spans": [ + { + "bbox": [ + 117, + 556, + 330, + 567 + ], + "score": 1.0, + "content": "//www.aclweb.org/anthology/N19-1423.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 32.5 + }, + { + "type": "text", + "bbox": [ + 107, + 573, + 506, + 618 + ], + "lines": [ + { + "bbox": [ + 106, + 574, + 507, + 587 + ], + "spans": [ + { + "bbox": [ + 106, + 574, + 507, + 587 + ], + "score": 1.0, + "content": "Yifeng Ding, Zhanyu Ma, Shaoguo Wen, Jiyang Xie, Dongliang Chang, Zhongwei Si, Ming Wu,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 584, + 506, + 598 + ], + "spans": [ + { + "bbox": [ + 115, + 584, + 506, + 598 + ], + "score": 1.0, + "content": "and Haibin Ling. Ap-cnn: Weakly supervised attention pyramid convolutional neural network for", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 596, + 507, + 609 + ], + "spans": [ + { + "bbox": [ + 115, + 596, + 507, + 609 + ], + "score": 1.0, + "content": "fine-grained visual classification. IEEE Transactions on Image Processing, 30:2826–2836, 2021.", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 116, + 608, + 249, + 618 + ], + "spans": [ + { + "bbox": [ + 116, + 608, + 249, + 618 + ], + "score": 1.0, + "content": "doi: 10.1109/TIP.2021.3055617.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 37.5 + }, + { + "type": "text", + "bbox": [ + 107, + 625, + 506, + 681 + ], + "lines": [ + { + "bbox": [ + 105, + 624, + 506, + 637 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 506, + 637 + ], + "score": 1.0, + "content": "Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 635, + 506, + 649 + ], + "spans": [ + { + "bbox": [ + 115, + 635, + 506, + 649 + ], + "score": 1.0, + "content": "Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 646, + 507, + 660 + ], + "spans": [ + { + "bbox": [ + 115, + 646, + 507, + 660 + ], + "score": 1.0, + "content": "and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale.", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 115, + 657, + 507, + 671 + ], + "spans": [ + { + "bbox": [ + 115, + 657, + 507, + 671 + ], + "score": 1.0, + "content": "In International Conference on Learning Representations, 2021. URL https://openreview.", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 115, + 667, + 259, + 683 + ], + "spans": [ + { + "bbox": [ + 115, + 667, + 259, + 683 + ], + "score": 1.0, + "content": "net/forum?id=YicbFdNTTy.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 42 + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 506, + 731 + ], + "lines": [ + { + "bbox": [ + 106, + 687, + 505, + 699 + ], + "spans": [ + { + "bbox": [ + 106, + 687, + 505, + 699 + ], + "score": 1.0, + "content": "Raghav Goyal, Samira Ebrahimi Kahou, Vincent Michalski, Joanna Materzynska, Susanne Westphal,", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 114, + 698, + 507, + 711 + ], + "spans": [ + { + "bbox": [ + 114, + 698, + 507, + 711 + ], + "score": 1.0, + "content": "Heuna Kim, Valentin Haenel, Ingo Fruend, Peter Yianilos, Moritz Mueller-Freitag, et al. The”", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 115, + 709, + 507, + 722 + ], + "spans": [ + { + "bbox": [ + 115, + 709, + 507, + 722 + ], + "score": 1.0, + "content": "something something” video database for learning and evaluating visual common sense. In ICCV,", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 115, + 720, + 142, + 732 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 142, + 732 + ], + "score": 1.0, + "content": "2017.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 46.5 + } + ], + "page_idx": 10, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 765 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 765 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 82, + 176, + 93 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 176, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 176, + 95 + ], + "score": 1.0, + "content": "REFERENCES", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 100, + 504, + 133 + ], + "lines": [ + { + "bbox": [ + 105, + 99, + 506, + 112 + ], + "spans": [ + { + "bbox": [ + 105, + 99, + 506, + 112 + ], + "score": 1.0, + "content": "Irwan Bello. Lambdanetworks: Modeling long-range interactions without attention. In International", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 116, + 110, + 506, + 124 + ], + "spans": [ + { + "bbox": [ + 116, + 110, + 506, + 124 + ], + "score": 1.0, + "content": "Conference on Learning Representations, 2021. URL https://openreview.net/forum?", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 116, + 122, + 205, + 135 + ], + "spans": [ + { + "bbox": [ + 116, + 122, + 205, + 135 + ], + "score": 1.0, + "content": "id=xTJEN-ggl1b.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2, + "bbox_fs": [ + 105, + 99, + 506, + 135 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 140, + 504, + 174 + ], + "lines": [ + { + "bbox": [ + 106, + 140, + 505, + 153 + ], + "spans": [ + { + "bbox": [ + 106, + 140, + 505, + 153 + ], + "score": 1.0, + "content": "Irwan Bello, Barret Zoph, Ashish Vaswani, Jonathon Shlens, and Quoc V Le. Attention augmented", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 151, + 505, + 164 + ], + "spans": [ + { + "bbox": [ + 115, + 151, + 505, + 164 + ], + "score": 1.0, + "content": "convolutional networks. In Proceedings of the IEEE/CVF International Conference on Computer", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 116, + 162, + 235, + 175 + ], + "spans": [ + { + "bbox": [ + 116, + 162, + 235, + 175 + ], + "score": 1.0, + "content": "Vision, pp. 3286–3295, 2019.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5, + "bbox_fs": [ + 106, + 140, + 505, + 175 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 180, + 505, + 204 + ], + "lines": [ + { + "bbox": [ + 106, + 180, + 506, + 193 + ], + "spans": [ + { + "bbox": [ + 106, + 180, + 506, + 193 + ], + "score": 1.0, + "content": "Gedas Bertasius, Heng Wang, and Lorenzo Torresani. Is Space-Time Attention All You Need for", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 116, + 191, + 315, + 204 + ], + "spans": [ + { + "bbox": [ + 116, + 191, + 315, + 204 + ], + "score": 1.0, + "content": "Video Understanding? arXiv.org, February 2021.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5, + "bbox_fs": [ + 106, + 180, + 506, + 204 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 209, + 505, + 233 + ], + "lines": [ + { + "bbox": [ + 106, + 210, + 505, + 223 + ], + "spans": [ + { + "bbox": [ + 106, + 210, + 505, + 223 + ], + "score": 1.0, + "content": "Andrew Brock, Soham De, Samuel L Smith, and Karen Simonyan. High-Performance Large-Scale", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 115, + 221, + 346, + 233 + ], + "spans": [ + { + "bbox": [ + 115, + 221, + 346, + 233 + ], + "score": 1.0, + "content": "Image Recognition Without Normalization. arXiv, 2021.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5, + "bbox_fs": [ + 106, + 210, + 505, + 233 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 239, + 505, + 262 + ], + "lines": [ + { + "bbox": [ + 106, + 240, + 505, + 252 + ], + "spans": [ + { + "bbox": [ + 106, + 240, + 505, + 252 + ], + "score": 1.0, + "content": "Chun-Fu Chen, Quanfu Fan, and Rameswar Panda. CrossViT: Cross-Attention Multi-Scale Vision", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 116, + 250, + 370, + 263 + ], + "spans": [ + { + "bbox": [ + 116, + 250, + 370, + 263 + ], + "score": 1.0, + "content": "Transformer for Image Classification. arXiv.org, March 2021a.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5, + "bbox_fs": [ + 106, + 240, + 505, + 263 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 268, + 505, + 292 + ], + "lines": [ + { + "bbox": [ + 106, + 268, + 505, + 282 + ], + "spans": [ + { + "bbox": [ + 106, + 268, + 505, + 282 + ], + "score": 1.0, + "content": "Zhengsu Chen, Lingxi Xie, Jianwei Niu, Xuefeng Liu, Longhui Wei, and Qi Tian. Visformer: The", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 116, + 279, + 328, + 293 + ], + "spans": [ + { + "bbox": [ + 116, + 279, + 328, + 293 + ], + "score": 1.0, + "content": "Vision-friendly Transformer. arxiv.org, April 2021b.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5, + "bbox_fs": [ + 106, + 268, + 505, + 293 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 298, + 505, + 332 + ], + "lines": [ + { + "bbox": [ + 106, + 298, + 506, + 311 + ], + "spans": [ + { + "bbox": [ + 106, + 298, + 506, + 311 + ], + "score": 1.0, + "content": "Xiangxiang Chu, Zhi Tian, Yuqing Wang, Bo Zhang, Haibing Ren, Xiaolin Wei, Huaxia Xia, and", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 308, + 507, + 323 + ], + "spans": [ + { + "bbox": [ + 115, + 308, + 507, + 323 + ], + "score": 1.0, + "content": "Chunhua Shen. Twins: Revisiting Spatial Attention Design in Vision Transformers. arXiv.org,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 114, + 320, + 169, + 334 + ], + "spans": [ + { + "bbox": [ + 114, + 320, + 169, + 334 + ], + "score": 1.0, + "content": "April 2021a.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16, + "bbox_fs": [ + 106, + 298, + 507, + 334 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 338, + 505, + 362 + ], + "lines": [ + { + "bbox": [ + 106, + 339, + 506, + 351 + ], + "spans": [ + { + "bbox": [ + 106, + 339, + 506, + 351 + ], + "score": 1.0, + "content": "Xiangxiang Chu, Zhi Tian, Bo Zhang, Xinlong Wang, Xiaolin Wei, Huaxia Xia, and Chunhua Shen.", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 117, + 350, + 411, + 362 + ], + "spans": [ + { + "bbox": [ + 117, + 350, + 411, + 362 + ], + "score": 1.0, + "content": "Conditional Positional Encodings for Vision Transformers. arXiv, 2021b.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18.5, + "bbox_fs": [ + 106, + 339, + 506, + 362 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 367, + 506, + 412 + ], + "lines": [ + { + "bbox": [ + 106, + 368, + 506, + 381 + ], + "spans": [ + { + "bbox": [ + 106, + 368, + 506, + 381 + ], + "score": 1.0, + "content": "Ekin Dogus Cubuk, Barret Zoph, Jon Shlens, and Quoc Le. RandAugment: Practical Automated", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 115, + 379, + 506, + 392 + ], + "spans": [ + { + "bbox": [ + 115, + 379, + 506, + 392 + ], + "score": 1.0, + "content": "Data Augmentation with a Reduced Search Space. In H Larochelle, M Ranzato, R Hadsell, M F", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 390, + 507, + 403 + ], + "spans": [ + { + "bbox": [ + 115, + 390, + 507, + 403 + ], + "score": 1.0, + "content": "Balcan, and H Lin (eds.), Advances in Neural Information Processing Systems, pp. 18613–18624.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 117, + 402, + 239, + 413 + ], + "spans": [ + { + "bbox": [ + 117, + 402, + 239, + 413 + ], + "score": 1.0, + "content": "Curran Associates, Inc., 2020.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 21.5, + "bbox_fs": [ + 106, + 368, + 507, + 413 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 419, + 505, + 453 + ], + "lines": [ + { + "bbox": [ + 106, + 419, + 507, + 433 + ], + "spans": [ + { + "bbox": [ + 106, + 419, + 507, + 433 + ], + "score": 1.0, + "content": "Stephane d’Ascoli, Hugo Touvron, Matthew Leavitt, Ari Morcos, Giulio Biroli, and Levent Sagun. ´", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 430, + 507, + 444 + ], + "spans": [ + { + "bbox": [ + 115, + 430, + 507, + 444 + ], + "score": 1.0, + "content": "ConViT: Improving Vision Transformers with Soft Convolutional Inductive Biases. arXiv.org,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 115, + 440, + 265, + 456 + ], + "spans": [ + { + "bbox": [ + 115, + 440, + 265, + 456 + ], + "score": 1.0, + "content": "cs.CV, 03 2021. URL arXiv.org.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25, + "bbox_fs": [ + 106, + 419, + 507, + 456 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 459, + 505, + 493 + ], + "lines": [ + { + "bbox": [ + 105, + 459, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 505, + 473 + ], + "score": 1.0, + "content": "Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 115, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition,", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 115, + 482, + 218, + 494 + ], + "spans": [ + { + "bbox": [ + 115, + 482, + 218, + 494 + ], + "score": 1.0, + "content": "pp. 248–255. Ieee, 2009.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 28, + "bbox_fs": [ + 105, + 459, + 506, + 494 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 500, + 506, + 567 + ], + "lines": [ + { + "bbox": [ + 105, + 500, + 506, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 500, + 506, + 513 + ], + "score": 1.0, + "content": "Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: Pre-training of deep", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 510, + 506, + 524 + ], + "spans": [ + { + "bbox": [ + 115, + 510, + 506, + 524 + ], + "score": 1.0, + "content": "bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 521, + 506, + 537 + ], + "spans": [ + { + "bbox": [ + 115, + 521, + 506, + 537 + ], + "score": 1.0, + "content": "the North American Chapter of the Association for Computational Linguistics: Human Language", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 115, + 533, + 506, + 546 + ], + "spans": [ + { + "bbox": [ + 115, + 533, + 506, + 546 + ], + "score": 1.0, + "content": "Technologies, Volume 1 (Long and Short Papers), pp. 4171–4186, Minneapolis, Minnesota, June", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 543, + 507, + 557 + ], + "spans": [ + { + "bbox": [ + 115, + 543, + 507, + 557 + ], + "score": 1.0, + "content": "2019. Association for Computational Linguistics. doi: 10.18653/v1/N19-1423. URL https:", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 117, + 556, + 330, + 567 + ], + "spans": [ + { + "bbox": [ + 117, + 556, + 330, + 567 + ], + "score": 1.0, + "content": "//www.aclweb.org/anthology/N19-1423.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 32.5, + "bbox_fs": [ + 105, + 500, + 507, + 567 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 573, + 506, + 618 + ], + "lines": [ + { + "bbox": [ + 106, + 574, + 507, + 587 + ], + "spans": [ + { + "bbox": [ + 106, + 574, + 507, + 587 + ], + "score": 1.0, + "content": "Yifeng Ding, Zhanyu Ma, Shaoguo Wen, Jiyang Xie, Dongliang Chang, Zhongwei Si, Ming Wu,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 584, + 506, + 598 + ], + "spans": [ + { + "bbox": [ + 115, + 584, + 506, + 598 + ], + "score": 1.0, + "content": "and Haibin Ling. Ap-cnn: Weakly supervised attention pyramid convolutional neural network for", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 596, + 507, + 609 + ], + "spans": [ + { + "bbox": [ + 115, + 596, + 507, + 609 + ], + "score": 1.0, + "content": "fine-grained visual classification. IEEE Transactions on Image Processing, 30:2826–2836, 2021.", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 116, + 608, + 249, + 618 + ], + "spans": [ + { + "bbox": [ + 116, + 608, + 249, + 618 + ], + "score": 1.0, + "content": "doi: 10.1109/TIP.2021.3055617.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 37.5, + "bbox_fs": [ + 106, + 574, + 507, + 618 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 625, + 506, + 681 + ], + "lines": [ + { + "bbox": [ + 105, + 624, + 506, + 637 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 506, + 637 + ], + "score": 1.0, + "content": "Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 635, + 506, + 649 + ], + "spans": [ + { + "bbox": [ + 115, + 635, + 506, + 649 + ], + "score": 1.0, + "content": "Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 646, + 507, + 660 + ], + "spans": [ + { + "bbox": [ + 115, + 646, + 507, + 660 + ], + "score": 1.0, + "content": "and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale.", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 115, + 657, + 507, + 671 + ], + "spans": [ + { + "bbox": [ + 115, + 657, + 507, + 671 + ], + "score": 1.0, + "content": "In International Conference on Learning Representations, 2021. URL https://openreview.", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 115, + 667, + 259, + 683 + ], + "spans": [ + { + "bbox": [ + 115, + 667, + 259, + 683 + ], + "score": 1.0, + "content": "net/forum?id=YicbFdNTTy.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 42, + "bbox_fs": [ + 105, + 624, + 507, + 683 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 506, + 731 + ], + "lines": [ + { + "bbox": [ + 106, + 687, + 505, + 699 + ], + "spans": [ + { + "bbox": [ + 106, + 687, + 505, + 699 + ], + "score": 1.0, + "content": "Raghav Goyal, Samira Ebrahimi Kahou, Vincent Michalski, Joanna Materzynska, Susanne Westphal,", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 114, + 698, + 507, + 711 + ], + "spans": [ + { + "bbox": [ + 114, + 698, + 507, + 711 + ], + "score": 1.0, + "content": "Heuna Kim, Valentin Haenel, Ingo Fruend, Peter Yianilos, Moritz Mueller-Freitag, et al. The”", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 115, + 709, + 507, + 722 + ], + "spans": [ + { + "bbox": [ + 115, + 709, + 507, + 722 + ], + "score": 1.0, + "content": "something something” video database for learning and evaluating visual common sense. In ICCV,", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 115, + 720, + 142, + 732 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 142, + 732 + ], + "score": 1.0, + "content": "2017.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 46.5, + "bbox_fs": [ + 106, + 687, + 507, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 506, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 506, + 96 + ], + "score": 1.0, + "content": "Ben Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Herve J ´ egou, ´", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 116, + 93, + 507, + 106 + ], + "spans": [ + { + "bbox": [ + 116, + 93, + 507, + 106 + ], + "score": 1.0, + "content": "and Matthijs Douze. LeViT: a Vision Transformer in ConvNet’s Clothing for Faster Inference.", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 105, + 208, + 116 + ], + "spans": [ + { + "bbox": [ + 115, + 105, + 208, + 116 + ], + "score": 1.0, + "content": "arXiv.org, April 2021.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 106, + 124, + 504, + 147 + ], + "lines": [ + { + "bbox": [ + 105, + 123, + 505, + 137 + ], + "spans": [ + { + "bbox": [ + 105, + 123, + 505, + 137 + ], + "score": 1.0, + "content": "Kai Han, An Xiao, Enhua Wu, Jianyuan Guo, Chunjing Xu, and Yunhe Wang. Transformer in", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 116, + 136, + 331, + 147 + ], + "spans": [ + { + "bbox": [ + 116, + 136, + 331, + 147 + ], + "score": 1.0, + "content": "transformer. arXiv preprint arXiv:2103.00112, 2021.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 107, + 155, + 505, + 188 + ], + "lines": [ + { + "bbox": [ + 105, + 154, + 505, + 169 + ], + "spans": [ + { + "bbox": [ + 105, + 154, + 505, + 169 + ], + "score": 1.0, + "content": "Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep Residual Learning for Image", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 166, + 505, + 179 + ], + "spans": [ + { + "bbox": [ + 115, + 166, + 505, + 179 + ], + "score": 1.0, + "content": "Recognition. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 115, + 176, + 143, + 189 + ], + "spans": [ + { + "bbox": [ + 115, + 176, + 143, + 189 + ], + "score": 1.0, + "content": "2016.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 106, + 196, + 505, + 220 + ], + "lines": [ + { + "bbox": [ + 106, + 197, + 505, + 210 + ], + "spans": [ + { + "bbox": [ + 106, + 197, + 505, + 210 + ], + "score": 1.0, + "content": "Kaiming He, Georgia Gkioxari, Piotr Dollar, and Ross Girshick. Mask r-cnn. In Proceedings of the", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 116, + 208, + 402, + 221 + ], + "spans": [ + { + "bbox": [ + 116, + 208, + 402, + 221 + ], + "score": 1.0, + "content": "IEEE International Conference on Computer Vision (ICCV), Oct 2017.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 106, + 227, + 504, + 251 + ], + "lines": [ + { + "bbox": [ + 106, + 228, + 506, + 241 + ], + "spans": [ + { + "bbox": [ + 106, + 228, + 506, + 241 + ], + "score": 1.0, + "content": "Byeongho Heo, Sangdoo Yun, Dongyoon Han, Sanghyuk Chun, Junsuk Choe, and Seong Joon Oh.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 116, + 239, + 438, + 252 + ], + "spans": [ + { + "bbox": [ + 116, + 239, + 438, + 252 + ], + "score": 1.0, + "content": "Rethinking Spatial Dimensions of Vision Transformers. arXiv.org, March 2021.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5 + }, + { + "type": "text", + "bbox": [ + 109, + 258, + 504, + 293 + ], + "lines": [ + { + "bbox": [ + 106, + 258, + 506, + 272 + ], + "spans": [ + { + "bbox": [ + 106, + 258, + 506, + 272 + ], + "score": 1.0, + "content": "Elad Hoffer, Tal Ben-Nun, Itay Hubara, Niv Giladi, Torsten Hoefler, and Daniel Soudry. Augment", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 114, + 270, + 505, + 283 + ], + "spans": [ + { + "bbox": [ + 114, + 270, + 505, + 283 + ], + "score": 1.0, + "content": "your batch: Improving generalization through instance repetition. In Proceedings of the IEEE/CVF", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 280, + 429, + 295 + ], + "spans": [ + { + "bbox": [ + 115, + 280, + 429, + 295 + ], + "score": 1.0, + "content": "Conference on Computer Vision and Pattern Recognition (CVPR), June 2020.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 108, + 300, + 505, + 334 + ], + "lines": [ + { + "bbox": [ + 106, + 300, + 506, + 315 + ], + "spans": [ + { + "bbox": [ + 106, + 300, + 506, + 315 + ], + "score": 1.0, + "content": "Han Hu, Zheng Zhang, Zhenda Xie, and Stephen Lin. Local relation networks for image recognition.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 312, + 506, + 324 + ], + "spans": [ + { + "bbox": [ + 115, + 312, + 506, + 324 + ], + "score": 1.0, + "content": "In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 3464–3473,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 114, + 322, + 143, + 335 + ], + "spans": [ + { + "bbox": [ + 114, + 322, + 143, + 335 + ], + "score": 1.0, + "content": "2019.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 106, + 343, + 504, + 366 + ], + "lines": [ + { + "bbox": [ + 105, + 342, + 506, + 356 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 506, + 356 + ], + "score": 1.0, + "content": "J. Hu, L. Shen, and G. Sun. Squeeze-and-excitation networks. In 2018 IEEE/CVF Conference on", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 116, + 354, + 506, + 367 + ], + "spans": [ + { + "bbox": [ + 116, + 354, + 506, + 367 + ], + "score": 1.0, + "content": "Computer Vision and Pattern Recognition, pp. 7132–7141, 2018. doi: 10.1109/CVPR.2018.00745.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18.5 + }, + { + "type": "text", + "bbox": [ + 107, + 373, + 506, + 407 + ], + "lines": [ + { + "bbox": [ + 104, + 371, + 507, + 388 + ], + "spans": [ + { + "bbox": [ + 104, + 371, + 507, + 388 + ], + "score": 1.0, + "content": "Zihang Jiang, Qibin Hou, Li Yuan, Daquan Zhou, Xiaojie Jin, Anran Wang, and Jiashi Feng.", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 115, + 384, + 505, + 398 + ], + "spans": [ + { + "bbox": [ + 115, + 384, + 228, + 398 + ], + "score": 1.0, + "content": "Token Labeling: Training a", + "type": "text" + }, + { + "bbox": [ + 228, + 385, + 255, + 396 + ], + "score": 0.82, + "content": "8 5 . 4 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 256, + 384, + 505, + 398 + ], + "score": 1.0, + "content": "Top-1 Accuracy Vision Transformer with 56M Parameters on", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 396, + 322, + 410 + ], + "spans": [ + { + "bbox": [ + 115, + 396, + 322, + 410 + ], + "score": 1.0, + "content": "ImageNet. arxiv.org, 04 2021. URL arxiv.org.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21 + }, + { + "type": "text", + "bbox": [ + 107, + 415, + 506, + 449 + ], + "lines": [ + { + "bbox": [ + 106, + 415, + 506, + 429 + ], + "spans": [ + { + "bbox": [ + 106, + 415, + 506, + 429 + ], + "score": 1.0, + "content": "Will Kay, Joao Carreira, Karen Simonyan, Brian Zhang, Chloe Hillier, Sudheendra Vijayanarasimhan,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 115, + 427, + 506, + 439 + ], + "spans": [ + { + "bbox": [ + 115, + 427, + 506, + 439 + ], + "score": 1.0, + "content": "Fabio Viola, Tim Green, Trevor Back, Paul Natsev, et al. The kinetics human action video dataset.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 438, + 220, + 449 + ], + "spans": [ + { + "bbox": [ + 115, + 438, + 220, + 449 + ], + "score": 1.0, + "content": "arXiv:1705.06950, 2017.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 107, + 457, + 505, + 491 + ], + "lines": [ + { + "bbox": [ + 105, + 457, + 506, + 471 + ], + "spans": [ + { + "bbox": [ + 105, + 457, + 506, + 471 + ], + "score": 1.0, + "content": "Alexander Kirillov, Ross Girshick, Kaiming He, and Piotr Dollar. Panoptic feature pyramid networks.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 115, + 468, + 507, + 483 + ], + "spans": [ + { + "bbox": [ + 115, + 468, + 507, + 483 + ], + "score": 1.0, + "content": "In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 479, + 163, + 492 + ], + "spans": [ + { + "bbox": [ + 115, + 479, + 163, + 492 + ], + "score": 1.0, + "content": "June 2019.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 106, + 499, + 504, + 533 + ], + "lines": [ + { + "bbox": [ + 105, + 499, + 505, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 499, + 505, + 513 + ], + "score": 1.0, + "content": "Jonathan Krause, Michael Stark, Jia Deng, and Li Fei-Fei. 3d object representations for fine-grained", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 511, + 505, + 523 + ], + "spans": [ + { + "bbox": [ + 116, + 511, + 505, + 523 + ], + "score": 1.0, + "content": "categorization. In 4th International IEEE Workshop on 3D Representation and Recognition", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 522, + 264, + 534 + ], + "spans": [ + { + "bbox": [ + 115, + 522, + 264, + 534 + ], + "score": 1.0, + "content": "(3dRR-13), Sydney, Australia, 2013.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 105, + 541, + 504, + 554 + ], + "lines": [ + { + "bbox": [ + 106, + 540, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 106, + 540, + 505, + 555 + ], + "score": 1.0, + "content": "Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layers of features from tiny images. 2009.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 32 + }, + { + "type": "text", + "bbox": [ + 106, + 561, + 504, + 585 + ], + "lines": [ + { + "bbox": [ + 106, + 560, + 505, + 576 + ], + "spans": [ + { + "bbox": [ + 106, + 560, + 505, + 576 + ], + "score": 1.0, + "content": "Yawei Li, Kai Zhang, Jiezhang Cao, Radu Timofte, and Luc Van Gool. LocalViT: Bringing Locality", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 573, + 306, + 585 + ], + "spans": [ + { + "bbox": [ + 115, + 573, + 306, + 585 + ], + "score": 1.0, + "content": "to Vision Transformers. arXiv.org, April 2021.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5 + }, + { + "type": "text", + "bbox": [ + 107, + 592, + 506, + 648 + ], + "lines": [ + { + "bbox": [ + 106, + 593, + 505, + 605 + ], + "spans": [ + { + "bbox": [ + 106, + 593, + 505, + 605 + ], + "score": 1.0, + "content": "Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 116, + 604, + 505, + 615 + ], + "spans": [ + { + "bbox": [ + 116, + 604, + 505, + 615 + ], + "score": 1.0, + "content": "Dollar, and C Lawrence Zitnick. Microsoft COCO: Common Objects in Context. In David ´", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 614, + 506, + 627 + ], + "spans": [ + { + "bbox": [ + 115, + 614, + 506, + 627 + ], + "score": 1.0, + "content": "Fleet, Tomas Pajdla, Bernt Schiele, and Tinne Tuytelaars (eds.), Computer Vision – ECCV 2014:", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 113, + 623, + 507, + 641 + ], + "spans": [ + { + "bbox": [ + 113, + 623, + 507, + 641 + ], + "score": 1.0, + "content": "13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V, pp.", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 636, + 348, + 649 + ], + "spans": [ + { + "bbox": [ + 115, + 636, + 348, + 649 + ], + "score": 1.0, + "content": "740–755. Springer International Publishing, Cham, 2014.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 106, + 656, + 504, + 690 + ], + "lines": [ + { + "bbox": [ + 105, + 655, + 506, + 669 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 506, + 669 + ], + "score": 1.0, + "content": "Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollar. Focal loss for dense object", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 667, + 506, + 680 + ], + "spans": [ + { + "bbox": [ + 115, + 667, + 506, + 680 + ], + "score": 1.0, + "content": "detection. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), Oct", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 677, + 144, + 690 + ], + "spans": [ + { + "bbox": [ + 115, + 677, + 144, + 690 + ], + "score": 1.0, + "content": "2017.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41 + }, + { + "type": "text", + "bbox": [ + 107, + 698, + 505, + 731 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 711 + ], + "score": 1.0, + "content": "Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo.", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 115, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 115, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows. arXiv.org, March", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 114, + 719, + 143, + 732 + ], + "spans": [ + { + "bbox": [ + 114, + 719, + 143, + 732 + ], + "score": 1.0, + "content": "2021.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44 + } + ], + "page_idx": 11, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "12", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 506, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 506, + 96 + ], + "score": 1.0, + "content": "Ben Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Herve J ´ egou, ´", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 116, + 93, + 507, + 106 + ], + "spans": [ + { + "bbox": [ + 116, + 93, + 507, + 106 + ], + "score": 1.0, + "content": "and Matthijs Douze. LeViT: a Vision Transformer in ConvNet’s Clothing for Faster Inference.", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 105, + 208, + 116 + ], + "spans": [ + { + "bbox": [ + 115, + 105, + 208, + 116 + ], + "score": 1.0, + "content": "arXiv.org, April 2021.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1, + "bbox_fs": [ + 105, + 81, + 507, + 116 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 124, + 504, + 147 + ], + "lines": [ + { + "bbox": [ + 105, + 123, + 505, + 137 + ], + "spans": [ + { + "bbox": [ + 105, + 123, + 505, + 137 + ], + "score": 1.0, + "content": "Kai Han, An Xiao, Enhua Wu, Jianyuan Guo, Chunjing Xu, and Yunhe Wang. Transformer in", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 116, + 136, + 331, + 147 + ], + "spans": [ + { + "bbox": [ + 116, + 136, + 331, + 147 + ], + "score": 1.0, + "content": "transformer. arXiv preprint arXiv:2103.00112, 2021.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5, + "bbox_fs": [ + 105, + 123, + 505, + 147 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 155, + 505, + 188 + ], + "lines": [ + { + "bbox": [ + 105, + 154, + 505, + 169 + ], + "spans": [ + { + "bbox": [ + 105, + 154, + 505, + 169 + ], + "score": 1.0, + "content": "Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep Residual Learning for Image", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 166, + 505, + 179 + ], + "spans": [ + { + "bbox": [ + 115, + 166, + 505, + 179 + ], + "score": 1.0, + "content": "Recognition. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 115, + 176, + 143, + 189 + ], + "spans": [ + { + "bbox": [ + 115, + 176, + 143, + 189 + ], + "score": 1.0, + "content": "2016.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6, + "bbox_fs": [ + 105, + 154, + 505, + 189 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 196, + 505, + 220 + ], + "lines": [ + { + "bbox": [ + 106, + 197, + 505, + 210 + ], + "spans": [ + { + "bbox": [ + 106, + 197, + 505, + 210 + ], + "score": 1.0, + "content": "Kaiming He, Georgia Gkioxari, Piotr Dollar, and Ross Girshick. Mask r-cnn. In Proceedings of the", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 116, + 208, + 402, + 221 + ], + "spans": [ + { + "bbox": [ + 116, + 208, + 402, + 221 + ], + "score": 1.0, + "content": "IEEE International Conference on Computer Vision (ICCV), Oct 2017.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8.5, + "bbox_fs": [ + 106, + 197, + 505, + 221 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 227, + 504, + 251 + ], + "lines": [ + { + "bbox": [ + 106, + 228, + 506, + 241 + ], + "spans": [ + { + "bbox": [ + 106, + 228, + 506, + 241 + ], + "score": 1.0, + "content": "Byeongho Heo, Sangdoo Yun, Dongyoon Han, Sanghyuk Chun, Junsuk Choe, and Seong Joon Oh.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 116, + 239, + 438, + 252 + ], + "spans": [ + { + "bbox": [ + 116, + 239, + 438, + 252 + ], + "score": 1.0, + "content": "Rethinking Spatial Dimensions of Vision Transformers. arXiv.org, March 2021.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5, + "bbox_fs": [ + 106, + 228, + 506, + 252 + ] + }, + { + "type": "text", + "bbox": [ + 109, + 258, + 504, + 293 + ], + "lines": [ + { + "bbox": [ + 106, + 258, + 506, + 272 + ], + "spans": [ + { + "bbox": [ + 106, + 258, + 506, + 272 + ], + "score": 1.0, + "content": "Elad Hoffer, Tal Ben-Nun, Itay Hubara, Niv Giladi, Torsten Hoefler, and Daniel Soudry. Augment", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 114, + 270, + 505, + 283 + ], + "spans": [ + { + "bbox": [ + 114, + 270, + 505, + 283 + ], + "score": 1.0, + "content": "your batch: Improving generalization through instance repetition. In Proceedings of the IEEE/CVF", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 280, + 429, + 295 + ], + "spans": [ + { + "bbox": [ + 115, + 280, + 429, + 295 + ], + "score": 1.0, + "content": "Conference on Computer Vision and Pattern Recognition (CVPR), June 2020.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13, + "bbox_fs": [ + 106, + 258, + 506, + 295 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 300, + 505, + 334 + ], + "lines": [ + { + "bbox": [ + 106, + 300, + 506, + 315 + ], + "spans": [ + { + "bbox": [ + 106, + 300, + 506, + 315 + ], + "score": 1.0, + "content": "Han Hu, Zheng Zhang, Zhenda Xie, and Stephen Lin. Local relation networks for image recognition.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 312, + 506, + 324 + ], + "spans": [ + { + "bbox": [ + 115, + 312, + 506, + 324 + ], + "score": 1.0, + "content": "In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 3464–3473,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 114, + 322, + 143, + 335 + ], + "spans": [ + { + "bbox": [ + 114, + 322, + 143, + 335 + ], + "score": 1.0, + "content": "2019.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16, + "bbox_fs": [ + 106, + 300, + 506, + 335 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 343, + 504, + 366 + ], + "lines": [ + { + "bbox": [ + 105, + 342, + 506, + 356 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 506, + 356 + ], + "score": 1.0, + "content": "J. Hu, L. Shen, and G. Sun. Squeeze-and-excitation networks. In 2018 IEEE/CVF Conference on", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 116, + 354, + 506, + 367 + ], + "spans": [ + { + "bbox": [ + 116, + 354, + 506, + 367 + ], + "score": 1.0, + "content": "Computer Vision and Pattern Recognition, pp. 7132–7141, 2018. doi: 10.1109/CVPR.2018.00745.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18.5, + "bbox_fs": [ + 105, + 342, + 506, + 367 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 373, + 506, + 407 + ], + "lines": [ + { + "bbox": [ + 104, + 371, + 507, + 388 + ], + "spans": [ + { + "bbox": [ + 104, + 371, + 507, + 388 + ], + "score": 1.0, + "content": "Zihang Jiang, Qibin Hou, Li Yuan, Daquan Zhou, Xiaojie Jin, Anran Wang, and Jiashi Feng.", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 115, + 384, + 505, + 398 + ], + "spans": [ + { + "bbox": [ + 115, + 384, + 228, + 398 + ], + "score": 1.0, + "content": "Token Labeling: Training a", + "type": "text" + }, + { + "bbox": [ + 228, + 385, + 255, + 396 + ], + "score": 0.82, + "content": "8 5 . 4 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 256, + 384, + 505, + 398 + ], + "score": 1.0, + "content": "Top-1 Accuracy Vision Transformer with 56M Parameters on", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 396, + 322, + 410 + ], + "spans": [ + { + "bbox": [ + 115, + 396, + 322, + 410 + ], + "score": 1.0, + "content": "ImageNet. arxiv.org, 04 2021. URL arxiv.org.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21, + "bbox_fs": [ + 104, + 371, + 507, + 410 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 415, + 506, + 449 + ], + "lines": [ + { + "bbox": [ + 106, + 415, + 506, + 429 + ], + "spans": [ + { + "bbox": [ + 106, + 415, + 506, + 429 + ], + "score": 1.0, + "content": "Will Kay, Joao Carreira, Karen Simonyan, Brian Zhang, Chloe Hillier, Sudheendra Vijayanarasimhan,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 115, + 427, + 506, + 439 + ], + "spans": [ + { + "bbox": [ + 115, + 427, + 506, + 439 + ], + "score": 1.0, + "content": "Fabio Viola, Tim Green, Trevor Back, Paul Natsev, et al. The kinetics human action video dataset.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 438, + 220, + 449 + ], + "spans": [ + { + "bbox": [ + 115, + 438, + 220, + 449 + ], + "score": 1.0, + "content": "arXiv:1705.06950, 2017.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24, + "bbox_fs": [ + 106, + 415, + 506, + 449 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 457, + 505, + 491 + ], + "lines": [ + { + "bbox": [ + 105, + 457, + 506, + 471 + ], + "spans": [ + { + "bbox": [ + 105, + 457, + 506, + 471 + ], + "score": 1.0, + "content": "Alexander Kirillov, Ross Girshick, Kaiming He, and Piotr Dollar. Panoptic feature pyramid networks.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 115, + 468, + 507, + 483 + ], + "spans": [ + { + "bbox": [ + 115, + 468, + 507, + 483 + ], + "score": 1.0, + "content": "In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 479, + 163, + 492 + ], + "spans": [ + { + "bbox": [ + 115, + 479, + 163, + 492 + ], + "score": 1.0, + "content": "June 2019.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27, + "bbox_fs": [ + 105, + 457, + 507, + 492 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 499, + 504, + 533 + ], + "lines": [ + { + "bbox": [ + 105, + 499, + 505, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 499, + 505, + 513 + ], + "score": 1.0, + "content": "Jonathan Krause, Michael Stark, Jia Deng, and Li Fei-Fei. 3d object representations for fine-grained", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 511, + 505, + 523 + ], + "spans": [ + { + "bbox": [ + 116, + 511, + 505, + 523 + ], + "score": 1.0, + "content": "categorization. In 4th International IEEE Workshop on 3D Representation and Recognition", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 522, + 264, + 534 + ], + "spans": [ + { + "bbox": [ + 115, + 522, + 264, + 534 + ], + "score": 1.0, + "content": "(3dRR-13), Sydney, Australia, 2013.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30, + "bbox_fs": [ + 105, + 499, + 505, + 534 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 541, + 504, + 554 + ], + "lines": [ + { + "bbox": [ + 106, + 540, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 106, + 540, + 505, + 555 + ], + "score": 1.0, + "content": "Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layers of features from tiny images. 2009.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 32, + "bbox_fs": [ + 106, + 540, + 505, + 555 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 561, + 504, + 585 + ], + "lines": [ + { + "bbox": [ + 106, + 560, + 505, + 576 + ], + "spans": [ + { + "bbox": [ + 106, + 560, + 505, + 576 + ], + "score": 1.0, + "content": "Yawei Li, Kai Zhang, Jiezhang Cao, Radu Timofte, and Luc Van Gool. LocalViT: Bringing Locality", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 573, + 306, + 585 + ], + "spans": [ + { + "bbox": [ + 115, + 573, + 306, + 585 + ], + "score": 1.0, + "content": "to Vision Transformers. arXiv.org, April 2021.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5, + "bbox_fs": [ + 106, + 560, + 505, + 585 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 592, + 506, + 648 + ], + "lines": [ + { + "bbox": [ + 106, + 593, + 505, + 605 + ], + "spans": [ + { + "bbox": [ + 106, + 593, + 505, + 605 + ], + "score": 1.0, + "content": "Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 116, + 604, + 505, + 615 + ], + "spans": [ + { + "bbox": [ + 116, + 604, + 505, + 615 + ], + "score": 1.0, + "content": "Dollar, and C Lawrence Zitnick. Microsoft COCO: Common Objects in Context. In David ´", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 614, + 506, + 627 + ], + "spans": [ + { + "bbox": [ + 115, + 614, + 506, + 627 + ], + "score": 1.0, + "content": "Fleet, Tomas Pajdla, Bernt Schiele, and Tinne Tuytelaars (eds.), Computer Vision – ECCV 2014:", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 113, + 623, + 507, + 641 + ], + "spans": [ + { + "bbox": [ + 113, + 623, + 507, + 641 + ], + "score": 1.0, + "content": "13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V, pp.", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 636, + 348, + 649 + ], + "spans": [ + { + "bbox": [ + 115, + 636, + 348, + 649 + ], + "score": 1.0, + "content": "740–755. Springer International Publishing, Cham, 2014.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 37, + "bbox_fs": [ + 106, + 593, + 507, + 649 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 656, + 504, + 690 + ], + "lines": [ + { + "bbox": [ + 105, + 655, + 506, + 669 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 506, + 669 + ], + "score": 1.0, + "content": "Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollar. Focal loss for dense object", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 667, + 506, + 680 + ], + "spans": [ + { + "bbox": [ + 115, + 667, + 506, + 680 + ], + "score": 1.0, + "content": "detection. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), Oct", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 677, + 144, + 690 + ], + "spans": [ + { + "bbox": [ + 115, + 677, + 144, + 690 + ], + "score": 1.0, + "content": "2017.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41, + "bbox_fs": [ + 105, + 655, + 506, + 690 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 698, + 505, + 731 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 711 + ], + "score": 1.0, + "content": "Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo.", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 115, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 115, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows. arXiv.org, March", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 114, + 719, + 143, + 732 + ], + "spans": [ + { + "bbox": [ + 114, + 719, + 143, + 732 + ], + "score": 1.0, + "content": "2021.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44, + "bbox_fs": [ + 105, + 698, + 506, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 104, + 82, + 505, + 105 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 506, + 95 + ], + "score": 1.0, + "content": "Ilya Loshchilov and Frank Hutter. Sgdr: Stochastic gradient descent with warm restarts. In Interna-", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 336, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 336, + 106 + ], + "score": 1.0, + "content": "tional Conference on Learning Representations, 2017.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 106, + 111, + 503, + 145 + ], + "lines": [ + { + "bbox": [ + 106, + 111, + 505, + 124 + ], + "spans": [ + { + "bbox": [ + 106, + 111, + 505, + 124 + ], + "score": 1.0, + "content": "Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In International Confer-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 115, + 123, + 505, + 135 + ], + "spans": [ + { + "bbox": [ + 115, + 123, + 505, + 135 + ], + "score": 1.0, + "content": "ence on Learning Representations, 2019. URL https://openreview.net/forum?id=", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 133, + 181, + 147 + ], + "spans": [ + { + "bbox": [ + 115, + 133, + 181, + 147 + ], + "score": 1.0, + "content": "Bkg6RiCqY7.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 106, + 152, + 505, + 175 + ], + "lines": [ + { + "bbox": [ + 106, + 153, + 505, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 153, + 505, + 165 + ], + "score": 1.0, + "content": "Omkar M. Parkhi, Andrea Vedaldi, Andrew Zisserman, and C. V. Jawahar. Cats and dogs. In IEEE", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 163, + 374, + 176 + ], + "spans": [ + { + "bbox": [ + 115, + 163, + 374, + 176 + ], + "score": 1.0, + "content": "Conference on Computer Vision and Pattern Recognition, 2012.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + }, + { + "type": "text", + "bbox": [ + 106, + 181, + 506, + 227 + ], + "lines": [ + { + "bbox": [ + 105, + 182, + 507, + 194 + ], + "spans": [ + { + "bbox": [ + 105, + 182, + 507, + 194 + ], + "score": 1.0, + "content": "Prajit Ramachandran, Niki Parmar, Ashish Vaswani, Irwan Bello, Anselm Levskaya, and Jon Shlens.", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 116, + 193, + 505, + 205 + ], + "spans": [ + { + "bbox": [ + 116, + 193, + 505, + 205 + ], + "score": 1.0, + "content": "Stand-Alone Self-Attention in Vision Models. In H Wallach, H Larochelle, A Beygelzimer, F d", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 115, + 204, + 507, + 217 + ], + "spans": [ + { + "bbox": [ + 115, + 204, + 507, + 217 + ], + "score": 1.0, + "content": "Alch e Buc, E Fox, and R Garnett (eds.), Advances in Neural Information Processing Systems.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 116, + 216, + 244, + 227 + ], + "spans": [ + { + "bbox": [ + 116, + 216, + 244, + 227 + ], + "score": 1.0, + "content": "Curran Associates, Inc., 2019a.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 107, + 233, + 504, + 257 + ], + "lines": [ + { + "bbox": [ + 105, + 233, + 505, + 246 + ], + "spans": [ + { + "bbox": [ + 105, + 233, + 505, + 246 + ], + "score": 1.0, + "content": "Prajit Ramachandran, Niki Parmar, Ashish Vaswani, Irwan Bello, Anselm Levskaya, and Jonathon", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 115, + 244, + 493, + 257 + ], + "spans": [ + { + "bbox": [ + 115, + 244, + 493, + 257 + ], + "score": 1.0, + "content": "Shlens. Stand-alone self-attention in vision models. arXiv preprint arXiv:1906.05909, 2019b.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5 + }, + { + "type": "text", + "bbox": [ + 106, + 263, + 504, + 286 + ], + "lines": [ + { + "bbox": [ + 106, + 263, + 507, + 276 + ], + "spans": [ + { + "bbox": [ + 106, + 263, + 507, + 276 + ], + "score": 1.0, + "content": "Aravind Srinivas, Tsung-Yi Lin, Niki Parmar, Jonathon Shlens, Pieter Abbeel, and Ashish Vaswani.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 116, + 274, + 469, + 287 + ], + "spans": [ + { + "bbox": [ + 116, + 274, + 469, + 287 + ], + "score": 1.0, + "content": "Bottleneck transformers for visual recognition. arXiv preprint arXiv:2101.11605, 2021.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5 + }, + { + "type": "text", + "bbox": [ + 108, + 292, + 505, + 326 + ], + "lines": [ + { + "bbox": [ + 105, + 291, + 505, + 307 + ], + "spans": [ + { + "bbox": [ + 105, + 291, + 505, + 307 + ], + "score": 1.0, + "content": "C. Sun, A. Shrivastava, S. Singh, and A. Gupta. Revisiting unreasonable effectiveness of data in deep", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 303, + 507, + 317 + ], + "spans": [ + { + "bbox": [ + 115, + 303, + 507, + 317 + ], + "score": 1.0, + "content": "learning era. In 2017 IEEE International Conference on Computer Vision (ICCV), pp. 843–852,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 116, + 315, + 258, + 326 + ], + "spans": [ + { + "bbox": [ + 116, + 315, + 258, + 326 + ], + "score": 1.0, + "content": "2017. doi: 10.1109/ICCV.2017.97.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 107, + 333, + 506, + 367 + ], + "lines": [ + { + "bbox": [ + 106, + 333, + 505, + 346 + ], + "spans": [ + { + "bbox": [ + 106, + 333, + 505, + 346 + ], + "score": 1.0, + "content": "Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. Rethinking the", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 344, + 505, + 357 + ], + "spans": [ + { + "bbox": [ + 115, + 344, + 505, + 357 + ], + "score": 1.0, + "content": "inception architecture for computer vision. In Proceedings of the IEEE Conference on Computer", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 354, + 325, + 368 + ], + "spans": [ + { + "bbox": [ + 115, + 354, + 325, + 368 + ], + "score": 1.0, + "content": "Vision and Pattern Recognition (CVPR), June 2016.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 106, + 373, + 505, + 418 + ], + "lines": [ + { + "bbox": [ + 105, + 374, + 506, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 374, + 506, + 387 + ], + "score": 1.0, + "content": "Mingxing Tan and Quoc Le. EfficientNet: Rethinking Model Scaling for Convolutional Neural", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 385, + 505, + 397 + ], + "spans": [ + { + "bbox": [ + 115, + 385, + 505, + 397 + ], + "score": 1.0, + "content": "Networks. In Kamalika Chaudhuri and Ruslan Salakhutdinov (eds.), Proceedings of the 36th", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 396, + 507, + 409 + ], + "spans": [ + { + "bbox": [ + 115, + 396, + 507, + 409 + ], + "score": 1.0, + "content": "International Conference on Machine Learning, pp. 6105–6114, Long Beach, California, USA,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 115, + 407, + 194, + 418 + ], + "spans": [ + { + "bbox": [ + 115, + 407, + 194, + 418 + ], + "score": 1.0, + "content": "June 2019. PMLR.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 22.5 + }, + { + "type": "text", + "bbox": [ + 106, + 425, + 506, + 459 + ], + "lines": [ + { + "bbox": [ + 106, + 425, + 505, + 438 + ], + "spans": [ + { + "bbox": [ + 106, + 425, + 505, + 438 + ], + "score": 1.0, + "content": "Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Herve´", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 115, + 436, + 505, + 450 + ], + "spans": [ + { + "bbox": [ + 115, + 436, + 505, + 450 + ], + "score": 1.0, + "content": "Jegou. Training data-efficient image transformers & distillation through attention. ´ arXiv preprint", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 115, + 448, + 219, + 458 + ], + "spans": [ + { + "bbox": [ + 115, + 448, + 219, + 458 + ], + "score": 1.0, + "content": "arXiv:2012.12877, 2020.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26 + }, + { + "type": "text", + "bbox": [ + 105, + 465, + 504, + 489 + ], + "lines": [ + { + "bbox": [ + 105, + 464, + 506, + 480 + ], + "spans": [ + { + "bbox": [ + 105, + 464, + 506, + 480 + ], + "score": 1.0, + "content": "Hugo Touvron, Matthieu Cord, Alexandre Sablayrolles, Gabriel Synnaeve, and Herve J ´ egou. Going ´", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 115, + 474, + 444, + 492 + ], + "spans": [ + { + "bbox": [ + 115, + 474, + 444, + 492 + ], + "score": 1.0, + "content": "deeper with Image Transformers. arXiv.org, cs.CV, 03 2021. URL arXiv.org.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 28.5 + }, + { + "type": "text", + "bbox": [ + 106, + 495, + 506, + 540 + ], + "lines": [ + { + "bbox": [ + 106, + 496, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 106, + 496, + 505, + 508 + ], + "score": 1.0, + "content": "Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 506, + 506, + 520 + ], + "spans": [ + { + "bbox": [ + 115, + 506, + 506, + 520 + ], + "score": 1.0, + "content": "Kaiser, and Illia Polosukhin. Attention is All you Need. In I Guyon, U V Luxburg, S Bengio,", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 116, + 518, + 505, + 530 + ], + "spans": [ + { + "bbox": [ + 116, + 518, + 505, + 530 + ], + "score": 1.0, + "content": "H Wallach, R Fergus, S Vishwanathan, and R Garnett (eds.), Advances in Neural Information", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 116, + 529, + 322, + 541 + ], + "spans": [ + { + "bbox": [ + 116, + 529, + 322, + 541 + ], + "score": 1.0, + "content": "Processing Systems. Curran Associates, Inc., 2017.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 106, + 546, + 505, + 581 + ], + "lines": [ + { + "bbox": [ + 105, + 546, + 506, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 546, + 506, + 560 + ], + "score": 1.0, + "content": "Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo,", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 116, + 559, + 505, + 571 + ], + "spans": [ + { + "bbox": [ + 116, + 559, + 505, + 571 + ], + "score": 1.0, + "content": "and Ling Shao. Pyramid vision transformer: A versatile backbone for dense prediction without", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 569, + 198, + 581 + ], + "spans": [ + { + "bbox": [ + 115, + 569, + 198, + 581 + ], + "score": 1.0, + "content": "convolutions, 2021.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35 + }, + { + "type": "text", + "bbox": [ + 107, + 587, + 506, + 621 + ], + "lines": [ + { + "bbox": [ + 106, + 588, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 505, + 600 + ], + "score": 1.0, + "content": "Xiaolong Wang, Ross Girshick, Abhinav Gupta, and Kaiming He. Non-local neural networks. In", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 598, + 506, + 612 + ], + "spans": [ + { + "bbox": [ + 115, + 598, + 506, + 612 + ], + "score": 1.0, + "content": "Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 7794–7803,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 608, + 143, + 622 + ], + "spans": [ + { + "bbox": [ + 115, + 608, + 143, + 622 + ], + "score": 1.0, + "content": "2018.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 107, + 628, + 506, + 673 + ], + "lines": [ + { + "bbox": [ + 105, + 628, + 507, + 641 + ], + "spans": [ + { + "bbox": [ + 105, + 628, + 507, + 641 + ], + "score": 1.0, + "content": "Xiaosong Wang, Yifan Peng, Le Lu, Zhiyong Lu, Mohammadhadi Bagheri, and Ronald M Summers.", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 116, + 640, + 506, + 651 + ], + "spans": [ + { + "bbox": [ + 116, + 640, + 256, + 651 + ], + "score": 1.0, + "content": "Chestx-ray8: Hospital-scale chest", + "type": "text" + }, + { + "bbox": [ + 256, + 641, + 263, + 650 + ], + "score": 0.37, + "content": "\\mathbf { X }", + "type": "inline_equation" + }, + { + "bbox": [ + 263, + 640, + 506, + 651 + ], + "score": 1.0, + "content": "-ray database and benchmarks on weakly-supervised classi-", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 649, + 505, + 663 + ], + "spans": [ + { + "bbox": [ + 115, + 649, + 505, + 663 + ], + "score": 1.0, + "content": "fication and localization of common thorax diseases. In Proceedings of the IEEE conference on", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 115, + 662, + 372, + 673 + ], + "spans": [ + { + "bbox": [ + 115, + 662, + 372, + 673 + ], + "score": 1.0, + "content": "computer vision and pattern recognition, pp. 2097–2106, 2017.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 41.5 + }, + { + "type": "text", + "bbox": [ + 107, + 680, + 504, + 703 + ], + "lines": [ + { + "bbox": [ + 106, + 680, + 504, + 691 + ], + "spans": [ + { + "bbox": [ + 106, + 680, + 504, + 691 + ], + "score": 1.0, + "content": "Zachary Wharton, Ardhendu Behera, and Asish Bera. An attention-driven hierarchical multi-scale", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 691, + 292, + 703 + ], + "spans": [ + { + "bbox": [ + 115, + 691, + 292, + 703 + ], + "score": 1.0, + "content": "representation for visual recognition, 2021.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44.5 + }, + { + "type": "text", + "bbox": [ + 108, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 708, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 183, + 722 + ], + "score": 1.0, + "content": "Ross Wightman.", + "type": "text" + }, + { + "bbox": [ + 201, + 708, + 311, + 722 + ], + "score": 1.0, + "content": "Pytorch image models.", + "type": "text" + }, + { + "bbox": [ + 329, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "https://github.com/rwightman/", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 115, + 721, + 268, + 732 + ], + "spans": [ + { + "bbox": [ + 115, + 721, + 268, + 732 + ], + "score": 1.0, + "content": "pytorch-image-models, 2019.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 46.5 + } + ], + "page_idx": 12, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 104, + 82, + 505, + 105 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 506, + 95 + ], + "score": 1.0, + "content": "Ilya Loshchilov and Frank Hutter. Sgdr: Stochastic gradient descent with warm restarts. In Interna-", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 336, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 336, + 106 + ], + "score": 1.0, + "content": "tional Conference on Learning Representations, 2017.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5, + "bbox_fs": [ + 105, + 81, + 506, + 106 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 111, + 503, + 145 + ], + "lines": [ + { + "bbox": [ + 106, + 111, + 505, + 124 + ], + "spans": [ + { + "bbox": [ + 106, + 111, + 505, + 124 + ], + "score": 1.0, + "content": "Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In International Confer-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 115, + 123, + 505, + 135 + ], + "spans": [ + { + "bbox": [ + 115, + 123, + 505, + 135 + ], + "score": 1.0, + "content": "ence on Learning Representations, 2019. URL https://openreview.net/forum?id=", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 133, + 181, + 147 + ], + "spans": [ + { + "bbox": [ + 115, + 133, + 181, + 147 + ], + "score": 1.0, + "content": "Bkg6RiCqY7.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3, + "bbox_fs": [ + 106, + 111, + 505, + 147 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 152, + 505, + 175 + ], + "lines": [ + { + "bbox": [ + 106, + 153, + 505, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 153, + 505, + 165 + ], + "score": 1.0, + "content": "Omkar M. Parkhi, Andrea Vedaldi, Andrew Zisserman, and C. V. Jawahar. Cats and dogs. In IEEE", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 163, + 374, + 176 + ], + "spans": [ + { + "bbox": [ + 115, + 163, + 374, + 176 + ], + "score": 1.0, + "content": "Conference on Computer Vision and Pattern Recognition, 2012.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5, + "bbox_fs": [ + 106, + 153, + 505, + 176 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 181, + 506, + 227 + ], + "lines": [ + { + "bbox": [ + 105, + 182, + 507, + 194 + ], + "spans": [ + { + "bbox": [ + 105, + 182, + 507, + 194 + ], + "score": 1.0, + "content": "Prajit Ramachandran, Niki Parmar, Ashish Vaswani, Irwan Bello, Anselm Levskaya, and Jon Shlens.", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 116, + 193, + 505, + 205 + ], + "spans": [ + { + "bbox": [ + 116, + 193, + 505, + 205 + ], + "score": 1.0, + "content": "Stand-Alone Self-Attention in Vision Models. In H Wallach, H Larochelle, A Beygelzimer, F d", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 115, + 204, + 507, + 217 + ], + "spans": [ + { + "bbox": [ + 115, + 204, + 507, + 217 + ], + "score": 1.0, + "content": "Alch e Buc, E Fox, and R Garnett (eds.), Advances in Neural Information Processing Systems.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 116, + 216, + 244, + 227 + ], + "spans": [ + { + "bbox": [ + 116, + 216, + 244, + 227 + ], + "score": 1.0, + "content": "Curran Associates, Inc., 2019a.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 8.5, + "bbox_fs": [ + 105, + 182, + 507, + 227 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 233, + 504, + 257 + ], + "lines": [ + { + "bbox": [ + 105, + 233, + 505, + 246 + ], + "spans": [ + { + "bbox": [ + 105, + 233, + 505, + 246 + ], + "score": 1.0, + "content": "Prajit Ramachandran, Niki Parmar, Ashish Vaswani, Irwan Bello, Anselm Levskaya, and Jonathon", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 115, + 244, + 493, + 257 + ], + "spans": [ + { + "bbox": [ + 115, + 244, + 493, + 257 + ], + "score": 1.0, + "content": "Shlens. Stand-alone self-attention in vision models. arXiv preprint arXiv:1906.05909, 2019b.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5, + "bbox_fs": [ + 105, + 233, + 505, + 257 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 263, + 504, + 286 + ], + "lines": [ + { + "bbox": [ + 106, + 263, + 507, + 276 + ], + "spans": [ + { + "bbox": [ + 106, + 263, + 507, + 276 + ], + "score": 1.0, + "content": "Aravind Srinivas, Tsung-Yi Lin, Niki Parmar, Jonathon Shlens, Pieter Abbeel, and Ashish Vaswani.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 116, + 274, + 469, + 287 + ], + "spans": [ + { + "bbox": [ + 116, + 274, + 469, + 287 + ], + "score": 1.0, + "content": "Bottleneck transformers for visual recognition. arXiv preprint arXiv:2101.11605, 2021.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5, + "bbox_fs": [ + 106, + 263, + 507, + 287 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 292, + 505, + 326 + ], + "lines": [ + { + "bbox": [ + 105, + 291, + 505, + 307 + ], + "spans": [ + { + "bbox": [ + 105, + 291, + 505, + 307 + ], + "score": 1.0, + "content": "C. Sun, A. Shrivastava, S. Singh, and A. Gupta. Revisiting unreasonable effectiveness of data in deep", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 303, + 507, + 317 + ], + "spans": [ + { + "bbox": [ + 115, + 303, + 507, + 317 + ], + "score": 1.0, + "content": "learning era. In 2017 IEEE International Conference on Computer Vision (ICCV), pp. 843–852,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 116, + 315, + 258, + 326 + ], + "spans": [ + { + "bbox": [ + 116, + 315, + 258, + 326 + ], + "score": 1.0, + "content": "2017. doi: 10.1109/ICCV.2017.97.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16, + "bbox_fs": [ + 105, + 291, + 507, + 326 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 333, + 506, + 367 + ], + "lines": [ + { + "bbox": [ + 106, + 333, + 505, + 346 + ], + "spans": [ + { + "bbox": [ + 106, + 333, + 505, + 346 + ], + "score": 1.0, + "content": "Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. Rethinking the", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 344, + 505, + 357 + ], + "spans": [ + { + "bbox": [ + 115, + 344, + 505, + 357 + ], + "score": 1.0, + "content": "inception architecture for computer vision. In Proceedings of the IEEE Conference on Computer", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 354, + 325, + 368 + ], + "spans": [ + { + "bbox": [ + 115, + 354, + 325, + 368 + ], + "score": 1.0, + "content": "Vision and Pattern Recognition (CVPR), June 2016.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19, + "bbox_fs": [ + 106, + 333, + 505, + 368 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 373, + 505, + 418 + ], + "lines": [ + { + "bbox": [ + 105, + 374, + 506, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 374, + 506, + 387 + ], + "score": 1.0, + "content": "Mingxing Tan and Quoc Le. EfficientNet: Rethinking Model Scaling for Convolutional Neural", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 385, + 505, + 397 + ], + "spans": [ + { + "bbox": [ + 115, + 385, + 505, + 397 + ], + "score": 1.0, + "content": "Networks. In Kamalika Chaudhuri and Ruslan Salakhutdinov (eds.), Proceedings of the 36th", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 396, + 507, + 409 + ], + "spans": [ + { + "bbox": [ + 115, + 396, + 507, + 409 + ], + "score": 1.0, + "content": "International Conference on Machine Learning, pp. 6105–6114, Long Beach, California, USA,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 115, + 407, + 194, + 418 + ], + "spans": [ + { + "bbox": [ + 115, + 407, + 194, + 418 + ], + "score": 1.0, + "content": "June 2019. PMLR.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 22.5, + "bbox_fs": [ + 105, + 374, + 507, + 418 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 425, + 506, + 459 + ], + "lines": [ + { + "bbox": [ + 106, + 425, + 505, + 438 + ], + "spans": [ + { + "bbox": [ + 106, + 425, + 505, + 438 + ], + "score": 1.0, + "content": "Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Herve´", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 115, + 436, + 505, + 450 + ], + "spans": [ + { + "bbox": [ + 115, + 436, + 505, + 450 + ], + "score": 1.0, + "content": "Jegou. Training data-efficient image transformers & distillation through attention. ´ arXiv preprint", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 115, + 448, + 219, + 458 + ], + "spans": [ + { + "bbox": [ + 115, + 448, + 219, + 458 + ], + "score": 1.0, + "content": "arXiv:2012.12877, 2020.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26, + "bbox_fs": [ + 106, + 425, + 505, + 458 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 465, + 504, + 489 + ], + "lines": [ + { + "bbox": [ + 105, + 464, + 506, + 480 + ], + "spans": [ + { + "bbox": [ + 105, + 464, + 506, + 480 + ], + "score": 1.0, + "content": "Hugo Touvron, Matthieu Cord, Alexandre Sablayrolles, Gabriel Synnaeve, and Herve J ´ egou. Going ´", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 115, + 474, + 444, + 492 + ], + "spans": [ + { + "bbox": [ + 115, + 474, + 444, + 492 + ], + "score": 1.0, + "content": "deeper with Image Transformers. arXiv.org, cs.CV, 03 2021. URL arXiv.org.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 28.5, + "bbox_fs": [ + 105, + 464, + 506, + 492 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 495, + 506, + 540 + ], + "lines": [ + { + "bbox": [ + 106, + 496, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 106, + 496, + 505, + 508 + ], + "score": 1.0, + "content": "Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 506, + 506, + 520 + ], + "spans": [ + { + "bbox": [ + 115, + 506, + 506, + 520 + ], + "score": 1.0, + "content": "Kaiser, and Illia Polosukhin. Attention is All you Need. In I Guyon, U V Luxburg, S Bengio,", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 116, + 518, + 505, + 530 + ], + "spans": [ + { + "bbox": [ + 116, + 518, + 505, + 530 + ], + "score": 1.0, + "content": "H Wallach, R Fergus, S Vishwanathan, and R Garnett (eds.), Advances in Neural Information", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 116, + 529, + 322, + 541 + ], + "spans": [ + { + "bbox": [ + 116, + 529, + 322, + 541 + ], + "score": 1.0, + "content": "Processing Systems. Curran Associates, Inc., 2017.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 31.5, + "bbox_fs": [ + 106, + 496, + 506, + 541 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 546, + 505, + 581 + ], + "lines": [ + { + "bbox": [ + 105, + 546, + 506, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 546, + 506, + 560 + ], + "score": 1.0, + "content": "Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo,", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 116, + 559, + 505, + 571 + ], + "spans": [ + { + "bbox": [ + 116, + 559, + 505, + 571 + ], + "score": 1.0, + "content": "and Ling Shao. Pyramid vision transformer: A versatile backbone for dense prediction without", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 569, + 198, + 581 + ], + "spans": [ + { + "bbox": [ + 115, + 569, + 198, + 581 + ], + "score": 1.0, + "content": "convolutions, 2021.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35, + "bbox_fs": [ + 105, + 546, + 506, + 581 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 587, + 506, + 621 + ], + "lines": [ + { + "bbox": [ + 106, + 588, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 505, + 600 + ], + "score": 1.0, + "content": "Xiaolong Wang, Ross Girshick, Abhinav Gupta, and Kaiming He. Non-local neural networks. In", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 598, + 506, + 612 + ], + "spans": [ + { + "bbox": [ + 115, + 598, + 506, + 612 + ], + "score": 1.0, + "content": "Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 7794–7803,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 608, + 143, + 622 + ], + "spans": [ + { + "bbox": [ + 115, + 608, + 143, + 622 + ], + "score": 1.0, + "content": "2018.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38, + "bbox_fs": [ + 106, + 588, + 506, + 622 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 628, + 506, + 673 + ], + "lines": [ + { + "bbox": [ + 105, + 628, + 507, + 641 + ], + "spans": [ + { + "bbox": [ + 105, + 628, + 507, + 641 + ], + "score": 1.0, + "content": "Xiaosong Wang, Yifan Peng, Le Lu, Zhiyong Lu, Mohammadhadi Bagheri, and Ronald M Summers.", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 116, + 640, + 506, + 651 + ], + "spans": [ + { + "bbox": [ + 116, + 640, + 256, + 651 + ], + "score": 1.0, + "content": "Chestx-ray8: Hospital-scale chest", + "type": "text" + }, + { + "bbox": [ + 256, + 641, + 263, + 650 + ], + "score": 0.37, + "content": "\\mathbf { X }", + "type": "inline_equation" + }, + { + "bbox": [ + 263, + 640, + 506, + 651 + ], + "score": 1.0, + "content": "-ray database and benchmarks on weakly-supervised classi-", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 649, + 505, + 663 + ], + "spans": [ + { + "bbox": [ + 115, + 649, + 505, + 663 + ], + "score": 1.0, + "content": "fication and localization of common thorax diseases. In Proceedings of the IEEE conference on", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 115, + 662, + 372, + 673 + ], + "spans": [ + { + "bbox": [ + 115, + 662, + 372, + 673 + ], + "score": 1.0, + "content": "computer vision and pattern recognition, pp. 2097–2106, 2017.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 41.5, + "bbox_fs": [ + 105, + 628, + 507, + 673 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 680, + 504, + 703 + ], + "lines": [ + { + "bbox": [ + 106, + 680, + 504, + 691 + ], + "spans": [ + { + "bbox": [ + 106, + 680, + 504, + 691 + ], + "score": 1.0, + "content": "Zachary Wharton, Ardhendu Behera, and Asish Bera. An attention-driven hierarchical multi-scale", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 691, + 292, + 703 + ], + "spans": [ + { + "bbox": [ + 115, + 691, + 292, + 703 + ], + "score": 1.0, + "content": "representation for visual recognition, 2021.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44.5, + "bbox_fs": [ + 106, + 680, + 504, + 703 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 708, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 183, + 722 + ], + "score": 1.0, + "content": "Ross Wightman.", + "type": "text" + }, + { + "bbox": [ + 201, + 708, + 311, + 722 + ], + "score": 1.0, + "content": "Pytorch image models.", + "type": "text" + }, + { + "bbox": [ + 329, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "https://github.com/rwightman/", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 115, + 721, + 268, + 732 + ], + "spans": [ + { + "bbox": [ + 115, + 721, + 268, + 732 + ], + "score": 1.0, + "content": "pytorch-image-models, 2019.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 46.5, + "bbox_fs": [ + 105, + 708, + 505, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "score": 1.0, + "content": "Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon. Cbam: Convolutional block", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 506, + 107 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 506, + 107 + ], + "score": 1.0, + "content": "attention module. In Proceedings of the European Conference on Computer Vision (ECCV),", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 116, + 104, + 186, + 117 + ], + "spans": [ + { + "bbox": [ + 116, + 104, + 186, + 117 + ], + "score": 1.0, + "content": "September 2018.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 106, + 123, + 505, + 146 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 506, + 136 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 506, + 136 + ], + "score": 1.0, + "content": "Haiping Wu, Bin Xiao, Noel Codella, Mengchen Liu, Xiyang Dai, Lu Yuan, and Lei Zhang. CvT:", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 116, + 134, + 416, + 147 + ], + "spans": [ + { + "bbox": [ + 116, + 134, + 416, + 147 + ], + "score": 1.0, + "content": "Introducing Convolutions to Vision Transformers. arXiv.org, March 2021.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 106, + 153, + 504, + 176 + ], + "lines": [ + { + "bbox": [ + 106, + 152, + 506, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 152, + 506, + 165 + ], + "score": 1.0, + "content": "Yuxin Wu, Alexander Kirillov, Francisco Massa, Wan-Yen Lo, and Ross Girshick. Detectron2.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 164, + 422, + 176 + ], + "spans": [ + { + "bbox": [ + 115, + 164, + 422, + 176 + ], + "score": 1.0, + "content": "https://github.com/facebookresearch/detectron2, 2019.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + }, + { + "type": "text", + "bbox": [ + 105, + 182, + 505, + 206 + ], + "lines": [ + { + "bbox": [ + 106, + 183, + 505, + 196 + ], + "spans": [ + { + "bbox": [ + 106, + 183, + 505, + 196 + ], + "score": 1.0, + "content": "Haotian Yan, Zhe Li, Weijian Li, Changhu Wang, Ming Wu, and Chuang Zhang. ConTNet: Why not", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 116, + 194, + 412, + 207 + ], + "spans": [ + { + "bbox": [ + 116, + 194, + 412, + 207 + ], + "score": 1.0, + "content": "use convolution and transformer at the same time? arXiv.org, April 2021.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 106, + 212, + 505, + 236 + ], + "lines": [ + { + "bbox": [ + 105, + 212, + 505, + 225 + ], + "spans": [ + { + "bbox": [ + 105, + 212, + 505, + 225 + ], + "score": 1.0, + "content": "Li Yuan, Yunpeng Chen, Tao Wang, Weihao Yu, Yujun Shi, Francis EH Tay, Jiashi Feng, and", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 116, + 223, + 507, + 236 + ], + "spans": [ + { + "bbox": [ + 116, + 223, + 507, + 236 + ], + "score": 1.0, + "content": "Shuicheng Yan. Tokens-to-token vit: Training vision transformers from scratch on imagenet, 2021.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "text", + "bbox": [ + 106, + 242, + 506, + 287 + ], + "lines": [ + { + "bbox": [ + 105, + 242, + 506, + 256 + ], + "spans": [ + { + "bbox": [ + 105, + 242, + 506, + 256 + ], + "score": 1.0, + "content": "Sangdoo Yun, Dongyoon Han, Seong Joon Oh, Sanghyuk Chun, Junsuk Choe, and Youngjoon", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 115, + 253, + 506, + 267 + ], + "spans": [ + { + "bbox": [ + 115, + 253, + 506, + 267 + ], + "score": 1.0, + "content": "Yoo. CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features. In", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 114, + 264, + 506, + 278 + ], + "spans": [ + { + "bbox": [ + 114, + 264, + 506, + 278 + ], + "score": 1.0, + "content": "Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), October", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 275, + 142, + 287 + ], + "spans": [ + { + "bbox": [ + 115, + 275, + 142, + 287 + ], + "score": 1.0, + "content": "2019.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 12.5 + }, + { + "type": "text", + "bbox": [ + 107, + 294, + 505, + 328 + ], + "lines": [ + { + "bbox": [ + 105, + 294, + 506, + 308 + ], + "spans": [ + { + "bbox": [ + 105, + 294, + 506, + 308 + ], + "score": 1.0, + "content": "Hongyi Zhang, Moustapha Cisse, Yann N. Dauphin, and David Lopez-Paz. mixup: Beyond empirical", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 305, + 507, + 318 + ], + "spans": [ + { + "bbox": [ + 115, + 305, + 507, + 318 + ], + "score": 1.0, + "content": "risk minimization. In International Conference on Learning Representations, 2018. URL https:", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 117, + 316, + 331, + 329 + ], + "spans": [ + { + "bbox": [ + 117, + 316, + 331, + 329 + ], + "score": 1.0, + "content": "//openreview.net/forum?id=r1Ddp1-Rb.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 107, + 335, + 505, + 369 + ], + "lines": [ + { + "bbox": [ + 105, + 335, + 506, + 348 + ], + "spans": [ + { + "bbox": [ + 105, + 335, + 506, + 348 + ], + "score": 1.0, + "content": "Pengchuan Zhang, Xiyang Dai, Jianwei Yang, Bin Xiao, Lu Yuan, Lei Zhang, and Jianfeng Gao.", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 344, + 506, + 360 + ], + "spans": [ + { + "bbox": [ + 115, + 344, + 506, + 360 + ], + "score": 1.0, + "content": "Multi-Scale Vision Longformer: A New Vision Transformer for High-Resolution Image Encoding.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 116, + 357, + 217, + 370 + ], + "spans": [ + { + "bbox": [ + 116, + 357, + 217, + 370 + ], + "score": 1.0, + "content": "arXiv.org, March 2021a.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 105, + 375, + 505, + 399 + ], + "lines": [ + { + "bbox": [ + 105, + 375, + 506, + 389 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 506, + 389 + ], + "score": 1.0, + "content": "Zizhao Zhang, Han Zhang, Long Zhao, Ting Chen, and Tomas Pfister. Aggregating Nested Trans-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 116, + 388, + 210, + 399 + ], + "spans": [ + { + "bbox": [ + 116, + 388, + 210, + 399 + ], + "score": 1.0, + "content": "formers. arXiv, 2021b.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5 + }, + { + "type": "text", + "bbox": [ + 106, + 405, + 505, + 439 + ], + "lines": [ + { + "bbox": [ + 105, + 405, + 505, + 419 + ], + "spans": [ + { + "bbox": [ + 105, + 405, + 505, + 419 + ], + "score": 1.0, + "content": "Hengshuang Zhao, Jiaya Jia, and Vladlen Koltun. Exploring self-attention for image recognition. In", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 114, + 415, + 507, + 432 + ], + "spans": [ + { + "bbox": [ + 114, + 415, + 507, + 432 + ], + "score": 1.0, + "content": "Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 426, + 163, + 440 + ], + "spans": [ + { + "bbox": [ + 115, + 426, + 163, + 440 + ], + "score": 1.0, + "content": "June 2020.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 106, + 447, + 504, + 481 + ], + "lines": [ + { + "bbox": [ + 105, + 447, + 506, + 460 + ], + "spans": [ + { + "bbox": [ + 105, + 447, + 506, + 460 + ], + "score": 1.0, + "content": "Heliang Zheng, Jianlong Fu, Zheng-Jun Zha, Jiebo Luo, and Tao Mei. Learning Rich Part Hierarchies", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 116, + 457, + 506, + 471 + ], + "spans": [ + { + "bbox": [ + 116, + 457, + 506, + 471 + ], + "score": 1.0, + "content": "With Progressive Attention Networks for Fine-Grained Image Recognition. IEEE Transactions on", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 116, + 469, + 473, + 482 + ], + "spans": [ + { + "bbox": [ + 116, + 469, + 473, + 482 + ], + "score": 1.0, + "content": "Image Processing, 29:476–488, 2020. ISSN 1057-7149. doi: 10.1109/tip.2019.2921876.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 107, + 487, + 504, + 521 + ], + "lines": [ + { + "bbox": [ + 105, + 487, + 506, + 501 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 506, + 501 + ], + "score": 1.0, + "content": "Zhun Zhong, Liang Zheng, Guoliang Kang, Shaozi Li, and Yi Yang. Random Erasing Data Augmen-", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 115, + 498, + 506, + 511 + ], + "spans": [ + { + "bbox": [ + 115, + 498, + 506, + 511 + ], + "score": 1.0, + "content": "tation. Proceedings of the AAAI Conference on Artificial Intelligence, 34(07):13001–13008, April", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 509, + 142, + 521 + ], + "spans": [ + { + "bbox": [ + 115, + 509, + 142, + 521 + ], + "score": 1.0, + "content": "2020.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 108, + 528, + 505, + 563 + ], + "lines": [ + { + "bbox": [ + 106, + 529, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 106, + 529, + 505, + 541 + ], + "score": 1.0, + "content": "Bolei Zhou, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Scene", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 115, + 539, + 506, + 552 + ], + "spans": [ + { + "bbox": [ + 115, + 539, + 506, + 552 + ], + "score": 1.0, + "content": "parsing through ade20k dataset. In Proceedings of the IEEE Conference on Computer Vision and", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 550, + 279, + 564 + ], + "spans": [ + { + "bbox": [ + 115, + 550, + 279, + 564 + ], + "score": 1.0, + "content": "Pattern Recognition (CVPR), July 2017.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33 + } + ], + "page_idx": 13, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "14", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "score": 1.0, + "content": "Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon. Cbam: Convolutional block", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 506, + 107 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 506, + 107 + ], + "score": 1.0, + "content": "attention module. In Proceedings of the European Conference on Computer Vision (ECCV),", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 116, + 104, + 186, + 117 + ], + "spans": [ + { + "bbox": [ + 116, + 104, + 186, + 117 + ], + "score": 1.0, + "content": "September 2018.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1, + "bbox_fs": [ + 105, + 82, + 506, + 117 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 123, + 505, + 146 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 506, + 136 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 506, + 136 + ], + "score": 1.0, + "content": "Haiping Wu, Bin Xiao, Noel Codella, Mengchen Liu, Xiyang Dai, Lu Yuan, and Lei Zhang. CvT:", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 116, + 134, + 416, + 147 + ], + "spans": [ + { + "bbox": [ + 116, + 134, + 416, + 147 + ], + "score": 1.0, + "content": "Introducing Convolutions to Vision Transformers. arXiv.org, March 2021.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5, + "bbox_fs": [ + 106, + 123, + 506, + 147 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 153, + 504, + 176 + ], + "lines": [ + { + "bbox": [ + 106, + 152, + 506, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 152, + 506, + 165 + ], + "score": 1.0, + "content": "Yuxin Wu, Alexander Kirillov, Francisco Massa, Wan-Yen Lo, and Ross Girshick. Detectron2.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 164, + 422, + 176 + ], + "spans": [ + { + "bbox": [ + 115, + 164, + 422, + 176 + ], + "score": 1.0, + "content": "https://github.com/facebookresearch/detectron2, 2019.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5, + "bbox_fs": [ + 106, + 152, + 506, + 176 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 182, + 505, + 206 + ], + "lines": [ + { + "bbox": [ + 106, + 183, + 505, + 196 + ], + "spans": [ + { + "bbox": [ + 106, + 183, + 505, + 196 + ], + "score": 1.0, + "content": "Haotian Yan, Zhe Li, Weijian Li, Changhu Wang, Ming Wu, and Chuang Zhang. ConTNet: Why not", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 116, + 194, + 412, + 207 + ], + "spans": [ + { + "bbox": [ + 116, + 194, + 412, + 207 + ], + "score": 1.0, + "content": "use convolution and transformer at the same time? arXiv.org, April 2021.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5, + "bbox_fs": [ + 106, + 183, + 505, + 207 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 212, + 505, + 236 + ], + "lines": [ + { + "bbox": [ + 105, + 212, + 505, + 225 + ], + "spans": [ + { + "bbox": [ + 105, + 212, + 505, + 225 + ], + "score": 1.0, + "content": "Li Yuan, Yunpeng Chen, Tao Wang, Weihao Yu, Yujun Shi, Francis EH Tay, Jiashi Feng, and", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 116, + 223, + 507, + 236 + ], + "spans": [ + { + "bbox": [ + 116, + 223, + 507, + 236 + ], + "score": 1.0, + "content": "Shuicheng Yan. Tokens-to-token vit: Training vision transformers from scratch on imagenet, 2021.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5, + "bbox_fs": [ + 105, + 212, + 507, + 236 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 242, + 506, + 287 + ], + "lines": [ + { + "bbox": [ + 105, + 242, + 506, + 256 + ], + "spans": [ + { + "bbox": [ + 105, + 242, + 506, + 256 + ], + "score": 1.0, + "content": "Sangdoo Yun, Dongyoon Han, Seong Joon Oh, Sanghyuk Chun, Junsuk Choe, and Youngjoon", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 115, + 253, + 506, + 267 + ], + "spans": [ + { + "bbox": [ + 115, + 253, + 506, + 267 + ], + "score": 1.0, + "content": "Yoo. CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features. In", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 114, + 264, + 506, + 278 + ], + "spans": [ + { + "bbox": [ + 114, + 264, + 506, + 278 + ], + "score": 1.0, + "content": "Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), October", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 275, + 142, + 287 + ], + "spans": [ + { + "bbox": [ + 115, + 275, + 142, + 287 + ], + "score": 1.0, + "content": "2019.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 12.5, + "bbox_fs": [ + 105, + 242, + 506, + 287 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 294, + 505, + 328 + ], + "lines": [ + { + "bbox": [ + 105, + 294, + 506, + 308 + ], + "spans": [ + { + "bbox": [ + 105, + 294, + 506, + 308 + ], + "score": 1.0, + "content": "Hongyi Zhang, Moustapha Cisse, Yann N. Dauphin, and David Lopez-Paz. mixup: Beyond empirical", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 305, + 507, + 318 + ], + "spans": [ + { + "bbox": [ + 115, + 305, + 507, + 318 + ], + "score": 1.0, + "content": "risk minimization. In International Conference on Learning Representations, 2018. URL https:", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 117, + 316, + 331, + 329 + ], + "spans": [ + { + "bbox": [ + 117, + 316, + 331, + 329 + ], + "score": 1.0, + "content": "//openreview.net/forum?id=r1Ddp1-Rb.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16, + "bbox_fs": [ + 105, + 294, + 507, + 329 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 335, + 505, + 369 + ], + "lines": [ + { + "bbox": [ + 105, + 335, + 506, + 348 + ], + "spans": [ + { + "bbox": [ + 105, + 335, + 506, + 348 + ], + "score": 1.0, + "content": "Pengchuan Zhang, Xiyang Dai, Jianwei Yang, Bin Xiao, Lu Yuan, Lei Zhang, and Jianfeng Gao.", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 344, + 506, + 360 + ], + "spans": [ + { + "bbox": [ + 115, + 344, + 506, + 360 + ], + "score": 1.0, + "content": "Multi-Scale Vision Longformer: A New Vision Transformer for High-Resolution Image Encoding.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 116, + 357, + 217, + 370 + ], + "spans": [ + { + "bbox": [ + 116, + 357, + 217, + 370 + ], + "score": 1.0, + "content": "arXiv.org, March 2021a.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19, + "bbox_fs": [ + 105, + 335, + 506, + 370 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 375, + 505, + 399 + ], + "lines": [ + { + "bbox": [ + 105, + 375, + 506, + 389 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 506, + 389 + ], + "score": 1.0, + "content": "Zizhao Zhang, Han Zhang, Long Zhao, Ting Chen, and Tomas Pfister. Aggregating Nested Trans-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 116, + 388, + 210, + 399 + ], + "spans": [ + { + "bbox": [ + 116, + 388, + 210, + 399 + ], + "score": 1.0, + "content": "formers. arXiv, 2021b.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5, + "bbox_fs": [ + 105, + 375, + 506, + 399 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 405, + 505, + 439 + ], + "lines": [ + { + "bbox": [ + 105, + 405, + 505, + 419 + ], + "spans": [ + { + "bbox": [ + 105, + 405, + 505, + 419 + ], + "score": 1.0, + "content": "Hengshuang Zhao, Jiaya Jia, and Vladlen Koltun. Exploring self-attention for image recognition. In", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 114, + 415, + 507, + 432 + ], + "spans": [ + { + "bbox": [ + 114, + 415, + 507, + 432 + ], + "score": 1.0, + "content": "Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 426, + 163, + 440 + ], + "spans": [ + { + "bbox": [ + 115, + 426, + 163, + 440 + ], + "score": 1.0, + "content": "June 2020.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24, + "bbox_fs": [ + 105, + 405, + 507, + 440 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 447, + 504, + 481 + ], + "lines": [ + { + "bbox": [ + 105, + 447, + 506, + 460 + ], + "spans": [ + { + "bbox": [ + 105, + 447, + 506, + 460 + ], + "score": 1.0, + "content": "Heliang Zheng, Jianlong Fu, Zheng-Jun Zha, Jiebo Luo, and Tao Mei. Learning Rich Part Hierarchies", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 116, + 457, + 506, + 471 + ], + "spans": [ + { + "bbox": [ + 116, + 457, + 506, + 471 + ], + "score": 1.0, + "content": "With Progressive Attention Networks for Fine-Grained Image Recognition. IEEE Transactions on", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 116, + 469, + 473, + 482 + ], + "spans": [ + { + "bbox": [ + 116, + 469, + 473, + 482 + ], + "score": 1.0, + "content": "Image Processing, 29:476–488, 2020. ISSN 1057-7149. doi: 10.1109/tip.2019.2921876.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27, + "bbox_fs": [ + 105, + 447, + 506, + 482 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 487, + 504, + 521 + ], + "lines": [ + { + "bbox": [ + 105, + 487, + 506, + 501 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 506, + 501 + ], + "score": 1.0, + "content": "Zhun Zhong, Liang Zheng, Guoliang Kang, Shaozi Li, and Yi Yang. Random Erasing Data Augmen-", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 115, + 498, + 506, + 511 + ], + "spans": [ + { + "bbox": [ + 115, + 498, + 506, + 511 + ], + "score": 1.0, + "content": "tation. Proceedings of the AAAI Conference on Artificial Intelligence, 34(07):13001–13008, April", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 509, + 142, + 521 + ], + "spans": [ + { + "bbox": [ + 115, + 509, + 142, + 521 + ], + "score": 1.0, + "content": "2020.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30, + "bbox_fs": [ + 105, + 487, + 506, + 521 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 528, + 505, + 563 + ], + "lines": [ + { + "bbox": [ + 106, + 529, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 106, + 529, + 505, + 541 + ], + "score": 1.0, + "content": "Bolei Zhou, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Scene", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 115, + 539, + 506, + 552 + ], + "spans": [ + { + "bbox": [ + 115, + 539, + 506, + 552 + ], + "score": 1.0, + "content": "parsing through ade20k dataset. In Proceedings of the IEEE Conference on Computer Vision and", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 550, + 279, + 564 + ], + "spans": [ + { + "bbox": [ + 115, + 550, + 279, + 564 + ], + "score": 1.0, + "content": "Pattern Recognition (CVPR), July 2017.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33, + "bbox_fs": [ + 106, + 529, + 506, + 564 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 92, + 505, + 315 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 117, + 81, + 490, + 92 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 119, + 79, + 492, + 93 + ], + "spans": [ + { + "bbox": [ + 119, + 79, + 492, + 93 + ], + "score": 1.0, + "content": "Table A1: Details of training settings for image classification on ImageNet1K and ImageNet21K.", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table_body", + "bbox": [ + 106, + 92, + 505, + 315 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 92, + 505, + 315 + ], + "spans": [ + { + "bbox": [ + 106, + 92, + 505, + 315 + ], + "score": 0.985, + "html": "
IN1KIN21KFinetune to IN1K@384²Transfer
Batch size4,0964,0961,024768
Epochs300120301,000
OptimizerAdamWAdamWAdamWSGD
Weight Decay0.050.011e-81e-4
Linear-rate Scheduler (Initial LR)Cosine (0.004)Cosine (0.001)Cosine (0.0002)Cosine (0.01)
Warmup Epochs50505
Warmup linear-rateLinear (1e-6)
Scheduler (Initial LR)
Data Aug.RandAugment (m=9,n=2)
Mixup (α)0.8
CutMix (α) Random Erasing1.0 0.25
Instance
Repetition*3
Drop-path0.1
Label Smoothing0.0 0.1
", + "type": "table", + "image_path": "4c05ce7a783a22cd075cd7fe4d1fc76bd16bb5a5daf2d620b9c064f816863f9c.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 106, + 92, + 505, + 166.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 166.33333333333331, + 505, + 240.66666666666663 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 106, + 240.66666666666663, + 505, + 314.99999999999994 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 111, + 314, + 214, + 324 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 112, + 312, + 215, + 325 + ], + "spans": [ + { + "bbox": [ + 112, + 312, + 215, + 325 + ], + "score": 1.0, + "content": "∗: disabled for RegionViT-Ti.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + } + ], + "index": 2 + }, + { + "type": "title", + "bbox": [ + 108, + 349, + 162, + 360 + ], + "lines": [ + { + "bbox": [ + 106, + 347, + 163, + 363 + ], + "spans": [ + { + "bbox": [ + 106, + 347, + 163, + 363 + ], + "score": 1.0, + "content": "APPENDIX", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 107, + 375, + 505, + 442 + ], + "lines": [ + { + "bbox": [ + 106, + 375, + 505, + 388 + ], + "spans": [ + { + "bbox": [ + 106, + 375, + 505, + 388 + ], + "score": 1.0, + "content": "Summary This appendix contains the following additional details. First, in Sec. A, we describe the", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 387, + 506, + 398 + ], + "spans": [ + { + "bbox": [ + 106, + 387, + 506, + 398 + ], + "score": 1.0, + "content": "training details on image classification, object and keypoint detection, semantic segmentation and ac-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 398, + 506, + 410 + ], + "spans": [ + { + "bbox": [ + 106, + 398, + 506, + 410 + ], + "score": 1.0, + "content": "tion recognition separately. Then, we provide detailed results on object detection and ablation studies.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 408, + 506, + 421 + ], + "spans": [ + { + "bbox": [ + 106, + 408, + 506, + 421 + ], + "score": 1.0, + "content": "The supplementary materials include our codes to reproduce our results on image classification and", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 419, + 506, + 432 + ], + "spans": [ + { + "bbox": [ + 105, + 419, + 506, + 432 + ], + "score": 1.0, + "content": "the README file in the attached codes (RegionViT Code.zip) also provides the detailed instructions", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 432, + 245, + 442 + ], + "spans": [ + { + "bbox": [ + 106, + 432, + 245, + 442 + ], + "score": 1.0, + "content": "to train and evaluate the networks.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 8.5 + }, + { + "type": "title", + "bbox": [ + 108, + 462, + 229, + 475 + ], + "lines": [ + { + "bbox": [ + 105, + 461, + 230, + 477 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 230, + 477 + ], + "score": 1.0, + "content": "A TRAINING DETAILS", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 12 + }, + { + "type": "title", + "bbox": [ + 108, + 490, + 237, + 501 + ], + "lines": [ + { + "bbox": [ + 106, + 489, + 238, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 238, + 502 + ], + "score": 1.0, + "content": "A.1 IMAGE CLASSIFICATION", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 106, + 511, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 511, + 505, + 525 + ], + "spans": [ + { + "bbox": [ + 105, + 511, + 505, + 525 + ], + "score": 1.0, + "content": "We follow DeiT (Touvron et al., 2020) to train our models on ImageNet1K (Deng et al., 2009) except", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 522, + 505, + 535 + ], + "spans": [ + { + "bbox": [ + 105, + 522, + 505, + 535 + ], + "score": 1.0, + "content": "that we use batch size 4,096 with base learning rate 0.004 and the warm-up epochs is 50. We adopt", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 534, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 534, + 505, + 546 + ], + "score": 1.0, + "content": "the AdamW (Loshchilov & Hutter, 2019) optimizer with cosine learning rate scheduler (Loshchilov", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 544, + 506, + 558 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 506, + 558 + ], + "score": 1.0, + "content": "& Hutter, 2017). We apply Mixup (Zhang et al., 2018), CutMix (Yun et al., 2019), RandomEras-", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 556, + 505, + 568 + ], + "spans": [ + { + "bbox": [ + 106, + 556, + 505, + 568 + ], + "score": 1.0, + "content": "ing (Zhong et al., 2020), label smoothing (Szegedy et al., 2016), RandAugment (Cubuk et al., 2020)", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 567, + 505, + 579 + ], + "spans": [ + { + "bbox": [ + 105, + 567, + 436, + 579 + ], + "score": 1.0, + "content": "and instance repetition (Hoffer et al., 2020). During training, we randomly crop a", + "type": "text" + }, + { + "bbox": [ + 436, + 567, + 476, + 578 + ], + "score": 0.89, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 476, + 567, + 505, + 579 + ], + "score": 1.0, + "content": "region", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 577, + 505, + 591 + ], + "spans": [ + { + "bbox": [ + 105, + 577, + 147, + 591 + ], + "score": 1.0, + "content": "and take a", + "type": "text" + }, + { + "bbox": [ + 148, + 578, + 186, + 589 + ], + "score": 0.89, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 187, + 577, + 505, + 591 + ], + "score": 1.0, + "content": "center crop after resizing the shorter side to 256 for evaluation. While pretraining", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 590, + 505, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 590, + 505, + 601 + ], + "score": 1.0, + "content": "on ImageNet21K (Deng et al., 2009), we use similar settings with slight modifications. We train the", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 600, + 506, + 613 + ], + "spans": [ + { + "bbox": [ + 106, + 600, + 506, + 613 + ], + "score": 1.0, + "content": "model using 120 epochs with 5 epochs for warmup, weight-decay of 0.01 and base learning rate of", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "score": 1.0, + "content": "0.001. For transfer learning experiments, we finetune the ImageNet1K pretrained models with 1,000", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 621, + 506, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 506, + 635 + ], + "score": 1.0, + "content": "epochs, batch size of 768, learning rate of 0.01, SGD optimizer, weight decay of 0.0001, and using", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 633, + 506, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 506, + 645 + ], + "score": 1.0, + "content": "the same data augmentation in training on ImageNet1K. During evaluation, we resize the shorter", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 642, + 507, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 272, + 657 + ], + "score": 1.0, + "content": "side of an image to 256 and then take a", + "type": "text" + }, + { + "bbox": [ + 273, + 644, + 313, + 654 + ], + "score": 0.89, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 313, + 642, + 507, + 657 + ], + "score": 1.0, + "content": "region at the center and report top-1 accuracy.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 654, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 506, + 667 + ], + "score": 1.0, + "content": "For finetuning experiments from ImageNet21K, we finetune the models with higher resolution of", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 665, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 145, + 676 + ], + "score": 0.88, + "content": "3 8 4 \\times 3 8 4", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 665, + 505, + 678 + ], + "score": 1.0, + "content": ", for 30 epochs with cosine learning rate scheduler, base linear rate of 0.002, weight-decay", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 677, + 505, + 688 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 424, + 688 + ], + "score": 1.0, + "content": "of 1e-8. Moreover, as it is trained on larger resolution, we adjust the window size", + "type": "text" + }, + { + "bbox": [ + 425, + 677, + 437, + 687 + ], + "score": 0.72, + "content": "M", + "type": "inline_equation" + }, + { + "bbox": [ + 437, + 677, + 505, + 688 + ], + "score": 1.0, + "content": "to have the same", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 687, + 506, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 701 + ], + "score": 1.0, + "content": "number of regional tokens at the first stage, e.g., we change the window size to 12 for the the models", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 146, + 711 + ], + "score": 1.0, + "content": "trained at", + "type": "text" + }, + { + "bbox": [ + 146, + 699, + 185, + 709 + ], + "score": 0.89, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "with window size 7. We adopt the bicubic interpolation to upsample the weights", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "of tokenizations and relative position bias. During evaluation, we directly resize the shorter side to", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 720, + 423, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 206, + 732 + ], + "score": 1.0, + "content": "384 and then take center", + "type": "text" + }, + { + "bbox": [ + 206, + 721, + 245, + 731 + ], + "score": 0.88, + "content": "3 8 4 \\times 3 8 4", + "type": "inline_equation" + }, + { + "bbox": [ + 246, + 720, + 423, + 732 + ], + "score": 1.0, + "content": "crop. We trained the models with 32 GPUs.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 23.5 + } + ], + "page_idx": 14, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 92, + 505, + 315 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 117, + 81, + 490, + 92 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 119, + 79, + 492, + 93 + ], + "spans": [ + { + "bbox": [ + 119, + 79, + 492, + 93 + ], + "score": 1.0, + "content": "Table A1: Details of training settings for image classification on ImageNet1K and ImageNet21K.", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table_body", + "bbox": [ + 106, + 92, + 505, + 315 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 92, + 505, + 315 + ], + "spans": [ + { + "bbox": [ + 106, + 92, + 505, + 315 + ], + "score": 0.985, + "html": "
IN1KIN21KFinetune to IN1K@384²Transfer
Batch size4,0964,0961,024768
Epochs300120301,000
OptimizerAdamWAdamWAdamWSGD
Weight Decay0.050.011e-81e-4
Linear-rate Scheduler (Initial LR)Cosine (0.004)Cosine (0.001)Cosine (0.0002)Cosine (0.01)
Warmup Epochs50505
Warmup linear-rateLinear (1e-6)
Scheduler (Initial LR)
Data Aug.RandAugment (m=9,n=2)
Mixup (α)0.8
CutMix (α) Random Erasing1.0 0.25
Instance
Repetition*3
Drop-path0.1
Label Smoothing0.0 0.1
", + "type": "table", + "image_path": "4c05ce7a783a22cd075cd7fe4d1fc76bd16bb5a5daf2d620b9c064f816863f9c.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 106, + 92, + 505, + 166.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 166.33333333333331, + 505, + 240.66666666666663 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 106, + 240.66666666666663, + 505, + 314.99999999999994 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 111, + 314, + 214, + 324 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 112, + 312, + 215, + 325 + ], + "spans": [ + { + "bbox": [ + 112, + 312, + 215, + 325 + ], + "score": 1.0, + "content": "∗: disabled for RegionViT-Ti.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + } + ], + "index": 2 + }, + { + "type": "title", + "bbox": [ + 108, + 349, + 162, + 360 + ], + "lines": [ + { + "bbox": [ + 106, + 347, + 163, + 363 + ], + "spans": [ + { + "bbox": [ + 106, + 347, + 163, + 363 + ], + "score": 1.0, + "content": "APPENDIX", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 107, + 375, + 505, + 442 + ], + "lines": [ + { + "bbox": [ + 106, + 375, + 505, + 388 + ], + "spans": [ + { + "bbox": [ + 106, + 375, + 505, + 388 + ], + "score": 1.0, + "content": "Summary This appendix contains the following additional details. First, in Sec. A, we describe the", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 387, + 506, + 398 + ], + "spans": [ + { + "bbox": [ + 106, + 387, + 506, + 398 + ], + "score": 1.0, + "content": "training details on image classification, object and keypoint detection, semantic segmentation and ac-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 398, + 506, + 410 + ], + "spans": [ + { + "bbox": [ + 106, + 398, + 506, + 410 + ], + "score": 1.0, + "content": "tion recognition separately. Then, we provide detailed results on object detection and ablation studies.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 408, + 506, + 421 + ], + "spans": [ + { + "bbox": [ + 106, + 408, + 506, + 421 + ], + "score": 1.0, + "content": "The supplementary materials include our codes to reproduce our results on image classification and", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 419, + 506, + 432 + ], + "spans": [ + { + "bbox": [ + 105, + 419, + 506, + 432 + ], + "score": 1.0, + "content": "the README file in the attached codes (RegionViT Code.zip) also provides the detailed instructions", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 432, + 245, + 442 + ], + "spans": [ + { + "bbox": [ + 106, + 432, + 245, + 442 + ], + "score": 1.0, + "content": "to train and evaluate the networks.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 8.5, + "bbox_fs": [ + 105, + 375, + 506, + 442 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 462, + 229, + 475 + ], + "lines": [ + { + "bbox": [ + 105, + 461, + 230, + 477 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 230, + 477 + ], + "score": 1.0, + "content": "A TRAINING DETAILS", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 12 + }, + { + "type": "title", + "bbox": [ + 108, + 490, + 237, + 501 + ], + "lines": [ + { + "bbox": [ + 106, + 489, + 238, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 238, + 502 + ], + "score": 1.0, + "content": "A.1 IMAGE CLASSIFICATION", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 106, + 511, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 511, + 505, + 525 + ], + "spans": [ + { + "bbox": [ + 105, + 511, + 505, + 525 + ], + "score": 1.0, + "content": "We follow DeiT (Touvron et al., 2020) to train our models on ImageNet1K (Deng et al., 2009) except", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 522, + 505, + 535 + ], + "spans": [ + { + "bbox": [ + 105, + 522, + 505, + 535 + ], + "score": 1.0, + "content": "that we use batch size 4,096 with base learning rate 0.004 and the warm-up epochs is 50. We adopt", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 534, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 534, + 505, + 546 + ], + "score": 1.0, + "content": "the AdamW (Loshchilov & Hutter, 2019) optimizer with cosine learning rate scheduler (Loshchilov", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 544, + 506, + 558 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 506, + 558 + ], + "score": 1.0, + "content": "& Hutter, 2017). We apply Mixup (Zhang et al., 2018), CutMix (Yun et al., 2019), RandomEras-", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 556, + 505, + 568 + ], + "spans": [ + { + "bbox": [ + 106, + 556, + 505, + 568 + ], + "score": 1.0, + "content": "ing (Zhong et al., 2020), label smoothing (Szegedy et al., 2016), RandAugment (Cubuk et al., 2020)", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 567, + 505, + 579 + ], + "spans": [ + { + "bbox": [ + 105, + 567, + 436, + 579 + ], + "score": 1.0, + "content": "and instance repetition (Hoffer et al., 2020). During training, we randomly crop a", + "type": "text" + }, + { + "bbox": [ + 436, + 567, + 476, + 578 + ], + "score": 0.89, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 476, + 567, + 505, + 579 + ], + "score": 1.0, + "content": "region", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 577, + 505, + 591 + ], + "spans": [ + { + "bbox": [ + 105, + 577, + 147, + 591 + ], + "score": 1.0, + "content": "and take a", + "type": "text" + }, + { + "bbox": [ + 148, + 578, + 186, + 589 + ], + "score": 0.89, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 187, + 577, + 505, + 591 + ], + "score": 1.0, + "content": "center crop after resizing the shorter side to 256 for evaluation. While pretraining", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 590, + 505, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 590, + 505, + 601 + ], + "score": 1.0, + "content": "on ImageNet21K (Deng et al., 2009), we use similar settings with slight modifications. We train the", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 600, + 506, + 613 + ], + "spans": [ + { + "bbox": [ + 106, + 600, + 506, + 613 + ], + "score": 1.0, + "content": "model using 120 epochs with 5 epochs for warmup, weight-decay of 0.01 and base learning rate of", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "score": 1.0, + "content": "0.001. For transfer learning experiments, we finetune the ImageNet1K pretrained models with 1,000", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 621, + 506, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 506, + 635 + ], + "score": 1.0, + "content": "epochs, batch size of 768, learning rate of 0.01, SGD optimizer, weight decay of 0.0001, and using", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 633, + 506, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 506, + 645 + ], + "score": 1.0, + "content": "the same data augmentation in training on ImageNet1K. During evaluation, we resize the shorter", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 642, + 507, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 272, + 657 + ], + "score": 1.0, + "content": "side of an image to 256 and then take a", + "type": "text" + }, + { + "bbox": [ + 273, + 644, + 313, + 654 + ], + "score": 0.89, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 313, + 642, + 507, + 657 + ], + "score": 1.0, + "content": "region at the center and report top-1 accuracy.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 654, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 506, + 667 + ], + "score": 1.0, + "content": "For finetuning experiments from ImageNet21K, we finetune the models with higher resolution of", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 665, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 145, + 676 + ], + "score": 0.88, + "content": "3 8 4 \\times 3 8 4", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 665, + 505, + 678 + ], + "score": 1.0, + "content": ", for 30 epochs with cosine learning rate scheduler, base linear rate of 0.002, weight-decay", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 677, + 505, + 688 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 424, + 688 + ], + "score": 1.0, + "content": "of 1e-8. Moreover, as it is trained on larger resolution, we adjust the window size", + "type": "text" + }, + { + "bbox": [ + 425, + 677, + 437, + 687 + ], + "score": 0.72, + "content": "M", + "type": "inline_equation" + }, + { + "bbox": [ + 437, + 677, + 505, + 688 + ], + "score": 1.0, + "content": "to have the same", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 687, + 506, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 701 + ], + "score": 1.0, + "content": "number of regional tokens at the first stage, e.g., we change the window size to 12 for the the models", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 146, + 711 + ], + "score": 1.0, + "content": "trained at", + "type": "text" + }, + { + "bbox": [ + 146, + 699, + 185, + 709 + ], + "score": 0.89, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "with window size 7. We adopt the bicubic interpolation to upsample the weights", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "of tokenizations and relative position bias. During evaluation, we directly resize the shorter side to", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 720, + 423, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 206, + 732 + ], + "score": 1.0, + "content": "384 and then take center", + "type": "text" + }, + { + "bbox": [ + 206, + 721, + 245, + 731 + ], + "score": 0.88, + "content": "3 8 4 \\times 3 8 4", + "type": "inline_equation" + }, + { + "bbox": [ + 246, + 720, + 423, + 732 + ], + "score": 1.0, + "content": "crop. We trained the models with 32 GPUs.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 23.5, + "bbox_fs": [ + 105, + 511, + 507, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 104, + 505, + 263 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 108, + 81, + 504, + 101 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 81, + 505, + 92 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 357, + 92 + ], + "score": 1.0, + "content": "Table A2: Object detection performance on the COCO val2017 with", + "type": "text" + }, + { + "bbox": [ + 357, + 82, + 370, + 91 + ], + "score": 0.87, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 370, + 81, + 505, + 92 + ], + "score": 1.0, + "content": "schedule. The bold number indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 90, + 436, + 102 + ], + "spans": [ + { + "bbox": [ + 106, + 90, + 329, + 102 + ], + "score": 1.0, + "content": "the best number within the section, and for MaskRCNN, both", + "type": "text" + }, + { + "bbox": [ + 330, + 90, + 346, + 100 + ], + "score": 0.84, + "content": "\\mathsf { A P } ^ { b }", + "type": "inline_equation" + }, + { + "bbox": [ + 346, + 90, + 362, + 102 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 362, + 91, + 382, + 100 + ], + "score": 0.81, + "content": "\\mathbf { A P } ^ { m }", + "type": "inline_equation" + }, + { + "bbox": [ + 382, + 90, + 436, + 102 + ], + "score": 1.0, + "content": "are annotated.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 106, + 104, + 505, + 263 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 104, + 505, + 263 + ], + "spans": [ + { + "bbox": [ + 106, + 104, + 505, + 263 + ], + "score": 0.982, + "html": "
BackboneParams (M)FLOPs (G)AP AP50RetineNet AP75 APsParamsFLOPsMaskRCNN AP5
1APMAPL(M)(G)APbAPAPmAP0APm5
ResNet5037.723436.355.338.619.340.048.844.2260.038.058.641.434.455.136.7
ConT-M(Yan et al.,2021)27.0217.239.359.341.823.143.151.940.5一38.1
PVT-S (Wang et al.,2021)34.240.461.343.025.042.955.744.140.462.943.837.860.140.3
ViL-S (Zhang et al.,2021a)35.7252.241.662.544.124.944.656.245.0174.341.864.145.138.561.141.4
Swin-T(Liu et al.,2021)38.524541.562.144.225.144.955.547.826442.264.646.239.161.642.0
Twins-SVT-S(w/PEG)(Chu etal.,21a)34.320943.064.246.328.046.457.544.022843.566.047.340.363.243.4
RegionViT-S40.8192.642.264.145.127.545.455.350.1171.342.565.846.139.562.842.2
RegionViT-S+41.5204.243.164.846.229.646.656.150.9182.943.566.947.540.463.743.4
RegionViT-S+W/PEG41.6204.343.965.547.328.547.357.950.9183.044.267.348.240.864.144.0
ResNet10131557.738.8
ResNeXt101-32x4d56.7 56.4319.138.5 39.957.8 59.641.2 42.721.4 22.342.6 44.251.1 52.563.233640.4 41.961.1 62.544.236.4 37.559.440.2
PVT-M(Wang et al.,2021)53.941.963.144.325.044.957.662.8 63.934042.064.445.9 45.639.061.642.1
ViL-M(Zhang et al.,021a)50.8338.942.964.045.427.046.157.260.1261.143.465.947.039.762.842.1
Swin-S (Liu et al.,2021)59.833544.565.747.527.448.059.969.135444.866.648.940.963.444.2
Twins-SVT-B(w/PEG)(Chu etal.,221a)67.032245.366.748.128.548.960.676.334045.267.649.341.564.544.8
RegionViT-B83.4308.943.365.246.429.246.457.092.2287.943.566.747.440.163.443.0
RegionViT-B+84.4328.144.266.247.129.247.558.693.2307.144.567.648.741.064.443.9
RegionViT-B+ W/PEG84.5328.244.666.447.629.647.659.093.2307.245.468.449.641.665.244.8
ResNeXt101-64x4d95.547341.060.944.023.945.254.0101.949342.863.847.338.460.641.3
PVT-L (Wang et al., 2021)71.134542.663.745.425.846.058.481.036442.965.046.639.561.942.5
ViL-B (Zhang et al., 2021a)66.7443.044.365.547.128.947.958.376.1365.145.167.249.341.064.344.2
Swin-B (Liu etal.,2021)98.447744.765.949.2111107.249645.51141.311
Twins-SVT-L (w/PEG) (Chu et al.,2021a)110.945545.767.149.2一一一119.747445.9一一41.6 42.466.245.6
RegionViT-B+W/PEGt84.5506.446.168.049.530.549.960.193.2464.446.369.151.2
", + "type": "table", + "image_path": "5b1324c86399f9630626e664ab46f330df1e21f77843d28ef3a276ab33b2d5be.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 106, + 104, + 505, + 157.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 106, + 157.0, + 505, + 210.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 106, + 210.0, + 505, + 263.0 + ], + "spans": [], + "index": 4 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 114, + 263, + 463, + 270 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 111, + 262, + 462, + 270 + ], + "spans": [ + { + "bbox": [ + 111, + 262, + 462, + 270 + ], + "score": 1.0, + "content": "The reported results of Swin (Liu et al., 2021) are from Twins (Chu et al., 2021a) as the original paper does not include resutls with ImageNet1K weights. †: input resolution is 896×1344.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + } + ], + "index": 3 + }, + { + "type": "title", + "bbox": [ + 107, + 279, + 343, + 289 + ], + "lines": [ + { + "bbox": [ + 105, + 277, + 344, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 344, + 291 + ], + "score": 1.0, + "content": "A.2 OBJECT DETECTION AND KEYPOINT DETECTION", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 106, + 300, + 505, + 443 + ], + "lines": [ + { + "bbox": [ + 105, + 300, + 506, + 312 + ], + "spans": [ + { + "bbox": [ + 105, + 300, + 506, + 312 + ], + "score": 1.0, + "content": "We adopt RetinaNet (Lin et al., 2017) and MaskRCNN (He et al., 2017) as our detection framework", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 311, + 506, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 311, + 506, + 324 + ], + "score": 1.0, + "content": "and use KeypoinyRCNN for keypoint detection, and we simply replace the backbone with RegionViT,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 323, + 505, + 335 + ], + "spans": [ + { + "bbox": [ + 106, + 323, + 505, + 335 + ], + "score": 1.0, + "content": "and then output the local tokens at each stages as multi-scale features for the detector. The regional", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 333, + 505, + 345 + ], + "spans": [ + { + "bbox": [ + 106, + 333, + 505, + 345 + ], + "score": 1.0, + "content": "tokens are not used in the detection. Before sending the features to the detector framework, we add", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 344, + 505, + 356 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 505, + 356 + ], + "score": 1.0, + "content": "new layer normalization layers to normalize the features. We use RegionViT-S and RegionViT-B", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 355, + 506, + 368 + ], + "spans": [ + { + "bbox": [ + 105, + 355, + 506, + 368 + ], + "score": 1.0, + "content": "as the backbone and the weights are initialized from ImageNet1K pretraining. We train our models", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 366, + 505, + 378 + ], + "spans": [ + { + "bbox": [ + 106, + 366, + 204, + 378 + ], + "score": 1.0, + "content": "based on the settings in", + "type": "text" + }, + { + "bbox": [ + 204, + 367, + 218, + 377 + ], + "score": 0.87, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 219, + 366, + 505, + 378 + ], + "score": 1.0, + "content": "schedule in Detectron2 (Wu et al., 2019), i.e., the batch size is 16 and", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 376, + 506, + 390 + ], + "spans": [ + { + "bbox": [ + 105, + 376, + 290, + 390 + ], + "score": 1.0, + "content": "the learning rate is set to 0.0001 which drops", + "type": "text" + }, + { + "bbox": [ + 290, + 378, + 309, + 388 + ], + "score": 0.88, + "content": "1 0 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 376, + 506, + 390 + ], + "score": 1.0, + "content": "at the 60,000-th and 80,000-th iteration. We use", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 388, + 506, + 401 + ], + "spans": [ + { + "bbox": [ + 105, + 388, + 506, + 401 + ], + "score": 1.0, + "content": "AdamW optimizer with weight-decay of 0.05, and resize the shorter and longer side of an image to", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 399, + 506, + 412 + ], + "spans": [ + { + "bbox": [ + 106, + 399, + 506, + 412 + ], + "score": 1.0, + "content": "672 and 1,120, respectively. We also adopt drop-path when finetuning the detector, with a rate set to", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 410, + 504, + 422 + ], + "spans": [ + { + "bbox": [ + 105, + 410, + 504, + 422 + ], + "score": 1.0, + "content": "0.2 for both RetinaNet and MaskRCNN. We trained the models with 8 GPUs. When training with", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 421, + 505, + 433 + ], + "spans": [ + { + "bbox": [ + 106, + 421, + 172, + 433 + ], + "score": 1.0, + "content": "longer schedule", + "type": "text" + }, + { + "bbox": [ + 172, + 422, + 192, + 433 + ], + "score": 0.88, + "content": "( 3 \\times )", + "type": "inline_equation" + }, + { + "bbox": [ + 192, + 421, + 505, + 433 + ], + "score": 1.0, + "content": ", we adopt stronger data augmentation (multi-scale training) used in Swin (Liu", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 432, + 476, + 444 + ], + "spans": [ + { + "bbox": [ + 106, + 432, + 421, + 444 + ], + "score": 1.0, + "content": "et al., 2021) or Twins (Chu et al., 2021a) rather than fixed-size training used in", + "type": "text" + }, + { + "bbox": [ + 422, + 433, + 436, + 443 + ], + "score": 0.86, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 436, + 432, + 476, + 444 + ], + "score": 1.0, + "content": "schedule.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 13 + }, + { + "type": "title", + "bbox": [ + 108, + 461, + 249, + 472 + ], + "lines": [ + { + "bbox": [ + 106, + 461, + 251, + 474 + ], + "spans": [ + { + "bbox": [ + 106, + 461, + 251, + 474 + ], + "score": 1.0, + "content": "A.3 SEMANTIC SEGMENTATION", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 483, + 505, + 571 + ], + "lines": [ + { + "bbox": [ + 105, + 482, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 505, + 496 + ], + "score": 1.0, + "content": "We adopt Semantic FPN as our semantic segmentation framework (Kirillov et al., 2019) by replacing", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 493, + 504, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 504, + 506 + ], + "score": 1.0, + "content": "the backbone network with RegionViT like the object detection task. We train the models with", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 504, + 505, + 517 + ], + "spans": [ + { + "bbox": [ + 105, + 504, + 505, + 517 + ], + "score": 1.0, + "content": "ImageNet1K pretrained weights. We mostly follow the training receipt in Twins (Chu et al., 2021a)", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 516, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 505, + 528 + ], + "score": 1.0, + "content": "The shorter side of the image is resized to 448 and the longer side won’t exceed 1792, and then we", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 526, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 133, + 541 + ], + "score": 1.0, + "content": "crop a", + "type": "text" + }, + { + "bbox": [ + 134, + 527, + 174, + 538 + ], + "score": 0.89, + "content": "4 4 8 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 174, + 526, + 505, + 541 + ], + "score": 1.0, + "content": "region for training. During the evaluation, we take the whole image after resizing", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 538, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 538, + 505, + 550 + ], + "score": 1.0, + "content": "as illustrated above. We train the model with AdamW optimizer with learning rate 1e-4 for 80k", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 549, + 505, + 561 + ], + "spans": [ + { + "bbox": [ + 105, + 549, + 505, + 561 + ], + "score": 1.0, + "content": "iterations, and the learning rate is decayed with poly schedule (power is 0.9). The weight decay and", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 560, + 427, + 572 + ], + "spans": [ + { + "bbox": [ + 106, + 560, + 427, + 572 + ], + "score": 1.0, + "content": "drop-path rate is 0.05 and 0.2, respectively. We trained all models with 8 GPUs.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 24.5 + }, + { + "type": "title", + "bbox": [ + 108, + 589, + 232, + 600 + ], + "lines": [ + { + "bbox": [ + 106, + 588, + 233, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 233, + 601 + ], + "score": 1.0, + "content": "A.4 ACTION RECOGNITION", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 610, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "score": 1.0, + "content": "We adopt the divided-space-time attention in TimeSformer (Bertasius et al., 2021) as the temporal", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "score": 1.0, + "content": "modeling to perform action recognition experiments. More specifically, we plugin the temporal", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "attention before every R2L transformer encoder and finetune the model pretrained with ImageNet1K", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 644, + 505, + 656 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 505, + 656 + ], + "score": 1.0, + "content": "(instead of ImageNet21K). We use the AdamW optimizer (Loshchilov & Hutter, 2019) with base", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "score": 1.0, + "content": "linear rate of 2e-4 and weight-decay of 0.0001. We apply the same Mixup (Zhang et al., 2018),", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 665, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 505, + 678 + ], + "score": 1.0, + "content": "CutMix (Yun et al., 2019) and drop path (Tan & Le, 2019) in the image classification. For the", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 676, + 504, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 373, + 689 + ], + "score": 1.0, + "content": "input, we uniformly sample 8 frames from the whole video and the", + "type": "text" + }, + { + "bbox": [ + 374, + 677, + 413, + 687 + ], + "score": 0.89, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 413, + 676, + 504, + 689 + ], + "score": 1.0, + "content": "region is cropped after", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "score": 1.0, + "content": "shorter side of images is resized to the range of 256 to 320, resulting the size of the input video as", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 158, + 710 + ], + "score": 0.88, + "content": "8 \\times 2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 698, + 429, + 712 + ], + "score": 1.0, + "content": ". During evaluation, we use the same way to sample frames but take", + "type": "text" + }, + { + "bbox": [ + 429, + 699, + 476, + 709 + ], + "score": 0.87, + "content": "3 2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 476, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "spatial", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "crops (top-left, center and bottom-right) after resizing shorter side of image to 256, and then ensemble", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 721, + 474, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 474, + 732 + ], + "score": 1.0, + "content": "the predictions as final prediction for the input video. We trained the models with 16 GPUs.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 35 + } + ], + "page_idx": 15, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 752, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 104, + 505, + 263 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 108, + 81, + 504, + 101 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 81, + 505, + 92 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 357, + 92 + ], + "score": 1.0, + "content": "Table A2: Object detection performance on the COCO val2017 with", + "type": "text" + }, + { + "bbox": [ + 357, + 82, + 370, + 91 + ], + "score": 0.87, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 370, + 81, + 505, + 92 + ], + "score": 1.0, + "content": "schedule. The bold number indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 90, + 436, + 102 + ], + "spans": [ + { + "bbox": [ + 106, + 90, + 329, + 102 + ], + "score": 1.0, + "content": "the best number within the section, and for MaskRCNN, both", + "type": "text" + }, + { + "bbox": [ + 330, + 90, + 346, + 100 + ], + "score": 0.84, + "content": "\\mathsf { A P } ^ { b }", + "type": "inline_equation" + }, + { + "bbox": [ + 346, + 90, + 362, + 102 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 362, + 91, + 382, + 100 + ], + "score": 0.81, + "content": "\\mathbf { A P } ^ { m }", + "type": "inline_equation" + }, + { + "bbox": [ + 382, + 90, + 436, + 102 + ], + "score": 1.0, + "content": "are annotated.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 106, + 104, + 505, + 263 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 104, + 505, + 263 + ], + "spans": [ + { + "bbox": [ + 106, + 104, + 505, + 263 + ], + "score": 0.982, + "html": "
BackboneParams (M)FLOPs (G)AP AP50RetineNet AP75 APsParamsFLOPsMaskRCNN AP5
1APMAPL(M)(G)APbAPAPmAP0APm5
ResNet5037.723436.355.338.619.340.048.844.2260.038.058.641.434.455.136.7
ConT-M(Yan et al.,2021)27.0217.239.359.341.823.143.151.940.5一38.1
PVT-S (Wang et al.,2021)34.240.461.343.025.042.955.744.140.462.943.837.860.140.3
ViL-S (Zhang et al.,2021a)35.7252.241.662.544.124.944.656.245.0174.341.864.145.138.561.141.4
Swin-T(Liu et al.,2021)38.524541.562.144.225.144.955.547.826442.264.646.239.161.642.0
Twins-SVT-S(w/PEG)(Chu etal.,21a)34.320943.064.246.328.046.457.544.022843.566.047.340.363.243.4
RegionViT-S40.8192.642.264.145.127.545.455.350.1171.342.565.846.139.562.842.2
RegionViT-S+41.5204.243.164.846.229.646.656.150.9182.943.566.947.540.463.743.4
RegionViT-S+W/PEG41.6204.343.965.547.328.547.357.950.9183.044.267.348.240.864.144.0
ResNet10131557.738.8
ResNeXt101-32x4d56.7 56.4319.138.5 39.957.8 59.641.2 42.721.4 22.342.6 44.251.1 52.563.233640.4 41.961.1 62.544.236.4 37.559.440.2
PVT-M(Wang et al.,2021)53.941.963.144.325.044.957.662.8 63.934042.064.445.9 45.639.061.642.1
ViL-M(Zhang et al.,021a)50.8338.942.964.045.427.046.157.260.1261.143.465.947.039.762.842.1
Swin-S (Liu et al.,2021)59.833544.565.747.527.448.059.969.135444.866.648.940.963.444.2
Twins-SVT-B(w/PEG)(Chu etal.,221a)67.032245.366.748.128.548.960.676.334045.267.649.341.564.544.8
RegionViT-B83.4308.943.365.246.429.246.457.092.2287.943.566.747.440.163.443.0
RegionViT-B+84.4328.144.266.247.129.247.558.693.2307.144.567.648.741.064.443.9
RegionViT-B+ W/PEG84.5328.244.666.447.629.647.659.093.2307.245.468.449.641.665.244.8
ResNeXt101-64x4d95.547341.060.944.023.945.254.0101.949342.863.847.338.460.641.3
PVT-L (Wang et al., 2021)71.134542.663.745.425.846.058.481.036442.965.046.639.561.942.5
ViL-B (Zhang et al., 2021a)66.7443.044.365.547.128.947.958.376.1365.145.167.249.341.064.344.2
Swin-B (Liu etal.,2021)98.447744.765.949.2111107.249645.51141.311
Twins-SVT-L (w/PEG) (Chu et al.,2021a)110.945545.767.149.2一一一119.747445.9一一41.6 42.466.245.6
RegionViT-B+W/PEGt84.5506.446.168.049.530.549.960.193.2464.446.369.151.2
", + "type": "table", + "image_path": "5b1324c86399f9630626e664ab46f330df1e21f77843d28ef3a276ab33b2d5be.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 106, + 104, + 505, + 157.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 106, + 157.0, + 505, + 210.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 106, + 210.0, + 505, + 263.0 + ], + "spans": [], + "index": 4 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 114, + 263, + 463, + 270 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 111, + 262, + 462, + 270 + ], + "spans": [ + { + "bbox": [ + 111, + 262, + 462, + 270 + ], + "score": 1.0, + "content": "The reported results of Swin (Liu et al., 2021) are from Twins (Chu et al., 2021a) as the original paper does not include resutls with ImageNet1K weights. †: input resolution is 896×1344.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + } + ], + "index": 3 + }, + { + "type": "title", + "bbox": [ + 107, + 279, + 343, + 289 + ], + "lines": [ + { + "bbox": [ + 105, + 277, + 344, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 344, + 291 + ], + "score": 1.0, + "content": "A.2 OBJECT DETECTION AND KEYPOINT DETECTION", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 106, + 300, + 505, + 443 + ], + "lines": [ + { + "bbox": [ + 105, + 300, + 506, + 312 + ], + "spans": [ + { + "bbox": [ + 105, + 300, + 506, + 312 + ], + "score": 1.0, + "content": "We adopt RetinaNet (Lin et al., 2017) and MaskRCNN (He et al., 2017) as our detection framework", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 311, + 506, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 311, + 506, + 324 + ], + "score": 1.0, + "content": "and use KeypoinyRCNN for keypoint detection, and we simply replace the backbone with RegionViT,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 323, + 505, + 335 + ], + "spans": [ + { + "bbox": [ + 106, + 323, + 505, + 335 + ], + "score": 1.0, + "content": "and then output the local tokens at each stages as multi-scale features for the detector. The regional", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 333, + 505, + 345 + ], + "spans": [ + { + "bbox": [ + 106, + 333, + 505, + 345 + ], + "score": 1.0, + "content": "tokens are not used in the detection. Before sending the features to the detector framework, we add", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 344, + 505, + 356 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 505, + 356 + ], + "score": 1.0, + "content": "new layer normalization layers to normalize the features. We use RegionViT-S and RegionViT-B", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 355, + 506, + 368 + ], + "spans": [ + { + "bbox": [ + 105, + 355, + 506, + 368 + ], + "score": 1.0, + "content": "as the backbone and the weights are initialized from ImageNet1K pretraining. We train our models", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 366, + 505, + 378 + ], + "spans": [ + { + "bbox": [ + 106, + 366, + 204, + 378 + ], + "score": 1.0, + "content": "based on the settings in", + "type": "text" + }, + { + "bbox": [ + 204, + 367, + 218, + 377 + ], + "score": 0.87, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 219, + 366, + 505, + 378 + ], + "score": 1.0, + "content": "schedule in Detectron2 (Wu et al., 2019), i.e., the batch size is 16 and", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 376, + 506, + 390 + ], + "spans": [ + { + "bbox": [ + 105, + 376, + 290, + 390 + ], + "score": 1.0, + "content": "the learning rate is set to 0.0001 which drops", + "type": "text" + }, + { + "bbox": [ + 290, + 378, + 309, + 388 + ], + "score": 0.88, + "content": "1 0 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 376, + 506, + 390 + ], + "score": 1.0, + "content": "at the 60,000-th and 80,000-th iteration. We use", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 388, + 506, + 401 + ], + "spans": [ + { + "bbox": [ + 105, + 388, + 506, + 401 + ], + "score": 1.0, + "content": "AdamW optimizer with weight-decay of 0.05, and resize the shorter and longer side of an image to", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 399, + 506, + 412 + ], + "spans": [ + { + "bbox": [ + 106, + 399, + 506, + 412 + ], + "score": 1.0, + "content": "672 and 1,120, respectively. We also adopt drop-path when finetuning the detector, with a rate set to", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 410, + 504, + 422 + ], + "spans": [ + { + "bbox": [ + 105, + 410, + 504, + 422 + ], + "score": 1.0, + "content": "0.2 for both RetinaNet and MaskRCNN. We trained the models with 8 GPUs. When training with", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 421, + 505, + 433 + ], + "spans": [ + { + "bbox": [ + 106, + 421, + 172, + 433 + ], + "score": 1.0, + "content": "longer schedule", + "type": "text" + }, + { + "bbox": [ + 172, + 422, + 192, + 433 + ], + "score": 0.88, + "content": "( 3 \\times )", + "type": "inline_equation" + }, + { + "bbox": [ + 192, + 421, + 505, + 433 + ], + "score": 1.0, + "content": ", we adopt stronger data augmentation (multi-scale training) used in Swin (Liu", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 432, + 476, + 444 + ], + "spans": [ + { + "bbox": [ + 106, + 432, + 421, + 444 + ], + "score": 1.0, + "content": "et al., 2021) or Twins (Chu et al., 2021a) rather than fixed-size training used in", + "type": "text" + }, + { + "bbox": [ + 422, + 433, + 436, + 443 + ], + "score": 0.86, + "content": "1 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 436, + 432, + 476, + 444 + ], + "score": 1.0, + "content": "schedule.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 13, + "bbox_fs": [ + 105, + 300, + 506, + 444 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 461, + 249, + 472 + ], + "lines": [ + { + "bbox": [ + 106, + 461, + 251, + 474 + ], + "spans": [ + { + "bbox": [ + 106, + 461, + 251, + 474 + ], + "score": 1.0, + "content": "A.3 SEMANTIC SEGMENTATION", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 483, + 505, + 571 + ], + "lines": [ + { + "bbox": [ + 105, + 482, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 505, + 496 + ], + "score": 1.0, + "content": "We adopt Semantic FPN as our semantic segmentation framework (Kirillov et al., 2019) by replacing", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 493, + 504, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 504, + 506 + ], + "score": 1.0, + "content": "the backbone network with RegionViT like the object detection task. We train the models with", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 504, + 505, + 517 + ], + "spans": [ + { + "bbox": [ + 105, + 504, + 505, + 517 + ], + "score": 1.0, + "content": "ImageNet1K pretrained weights. We mostly follow the training receipt in Twins (Chu et al., 2021a)", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 516, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 505, + 528 + ], + "score": 1.0, + "content": "The shorter side of the image is resized to 448 and the longer side won’t exceed 1792, and then we", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 526, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 133, + 541 + ], + "score": 1.0, + "content": "crop a", + "type": "text" + }, + { + "bbox": [ + 134, + 527, + 174, + 538 + ], + "score": 0.89, + "content": "4 4 8 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 174, + 526, + 505, + 541 + ], + "score": 1.0, + "content": "region for training. During the evaluation, we take the whole image after resizing", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 538, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 538, + 505, + 550 + ], + "score": 1.0, + "content": "as illustrated above. We train the model with AdamW optimizer with learning rate 1e-4 for 80k", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 549, + 505, + 561 + ], + "spans": [ + { + "bbox": [ + 105, + 549, + 505, + 561 + ], + "score": 1.0, + "content": "iterations, and the learning rate is decayed with poly schedule (power is 0.9). The weight decay and", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 560, + 427, + 572 + ], + "spans": [ + { + "bbox": [ + 106, + 560, + 427, + 572 + ], + "score": 1.0, + "content": "drop-path rate is 0.05 and 0.2, respectively. We trained all models with 8 GPUs.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 24.5, + "bbox_fs": [ + 105, + 482, + 505, + 572 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 589, + 232, + 600 + ], + "lines": [ + { + "bbox": [ + 106, + 588, + 233, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 233, + 601 + ], + "score": 1.0, + "content": "A.4 ACTION RECOGNITION", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 610, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "score": 1.0, + "content": "We adopt the divided-space-time attention in TimeSformer (Bertasius et al., 2021) as the temporal", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "score": 1.0, + "content": "modeling to perform action recognition experiments. More specifically, we plugin the temporal", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "attention before every R2L transformer encoder and finetune the model pretrained with ImageNet1K", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 644, + 505, + 656 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 505, + 656 + ], + "score": 1.0, + "content": "(instead of ImageNet21K). We use the AdamW optimizer (Loshchilov & Hutter, 2019) with base", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "score": 1.0, + "content": "linear rate of 2e-4 and weight-decay of 0.0001. We apply the same Mixup (Zhang et al., 2018),", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 665, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 505, + 678 + ], + "score": 1.0, + "content": "CutMix (Yun et al., 2019) and drop path (Tan & Le, 2019) in the image classification. For the", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 676, + 504, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 373, + 689 + ], + "score": 1.0, + "content": "input, we uniformly sample 8 frames from the whole video and the", + "type": "text" + }, + { + "bbox": [ + 374, + 677, + 413, + 687 + ], + "score": 0.89, + "content": "2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 413, + 676, + 504, + 689 + ], + "score": 1.0, + "content": "region is cropped after", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 699 + ], + "score": 1.0, + "content": "shorter side of images is resized to the range of 256 to 320, resulting the size of the input video as", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 158, + 710 + ], + "score": 0.88, + "content": "8 \\times 2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 698, + 429, + 712 + ], + "score": 1.0, + "content": ". During evaluation, we use the same way to sample frames but take", + "type": "text" + }, + { + "bbox": [ + 429, + 699, + 476, + 709 + ], + "score": 0.87, + "content": "3 2 2 4 \\times 2 2 4", + "type": "inline_equation" + }, + { + "bbox": [ + 476, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "spatial", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "crops (top-left, center and bottom-right) after resizing shorter side of image to 256, and then ensemble", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 721, + 474, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 474, + 732 + ], + "score": 1.0, + "content": "the predictions as final prediction for the input video. We trained the models with 16 GPUs.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 35, + "bbox_fs": [ + 105, + 611, + 506, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 103, + 505, + 235 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 108, + 81, + 504, + 101 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 81, + 505, + 92 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 357, + 92 + ], + "score": 1.0, + "content": "Table A3: Object detection performance on the COCO val2017 with", + "type": "text" + }, + { + "bbox": [ + 357, + 82, + 370, + 91 + ], + "score": 0.87, + "content": "3 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 370, + 81, + 505, + 92 + ], + "score": 1.0, + "content": "schedule. The bold number indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 90, + 436, + 102 + ], + "spans": [ + { + "bbox": [ + 106, + 90, + 329, + 102 + ], + "score": 1.0, + "content": "the best number within the section, and for MaskRCNN, both", + "type": "text" + }, + { + "bbox": [ + 330, + 90, + 346, + 100 + ], + "score": 0.84, + "content": "\\mathsf { A P } ^ { b }", + "type": "inline_equation" + }, + { + "bbox": [ + 346, + 90, + 362, + 102 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 362, + 91, + 382, + 100 + ], + "score": 0.81, + "content": "\\mathbf { A P } ^ { m }", + "type": "inline_equation" + }, + { + "bbox": [ + 382, + 90, + 436, + 102 + ], + "score": 1.0, + "content": "are annotated.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 106, + 103, + 505, + 235 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 103, + 505, + 235 + ], + "spans": [ + { + "bbox": [ + 106, + 103, + 505, + 235 + ], + "score": 0.982, + "html": "
BackboneParams (M)FLOPs (G)AP AP50RetineNet AP75|APsParams (M)FLOPs GAPbMaskRCNN
APM|APLAPAP5|APmAPAPm5
ResNet5037.723439.058.441.822.442.851.644.2260.041.061.744.937.158.440.1
PVT-S (Wang etal.,2021)34.2142.262.745.026.245.257.244.124543.065.346.939.962.542.8
ViL-S (Zhang etal.,2021a)35.7252.242.963.845.627.846.456.345.0174.343.464.947.039.662.142.4
Swin-T(Liu et al., 2021)38.524543.964.847.128.447.257.847.826446.068.250.241.665.144.8
Twins-SVT-S (w/PEG)(Chu et al., 2021a)34.320945.667.148.629.849.360.044.022846.869.251.242.666.345.8
RegionViT-S40.8192.645.867.249.230.050.060.550.1171.346.368.850.642.365.545.7
RegionViT-S+41.5204.246.968.350.731.151.062.050.9182.947.369.552.043.166.446.7
RegionViT-S+ w/PEG41.6204.346.768.250.230.750.862.450.9183.047.670.052.043.467.147.0
ResNet10156.731540.944.063.233642.847.138.560.141.3
ResNeXt101-32x4d56.4319.141.460.1 61.044.323.7 23.945.0 45.553.8 53.762.834044.063.2 64.448.039.261.441.9
PVT-M(Wang etal.,021)53.9143.263.846.127.346.358.963.930244.266.048.240.563.143.5
ViL-M(Zhang etal.,2021a)50.8338.943.764.646.427.947.156.960.1261.144.666.348.540.763.843.7
Swin-S (Liuetal.,221)59.833546.367.449.831.150.360.969.135447.669.452.542.866.546.4
Twins-SVT-B (w/ PEG)(Chu et al., 2021a)67.032246.968.050.231.750.361.876.334048.069.552.743.066.846.6
RegionViT-B83.4308.946.167.849.131.550.261.292.2287.947.269.151.743.066.446.5
RegionViT-B+84.4328.146.968.650.130.850.762.693.2307.148.170.252.543.567.147.1
RegionViT-B+ W/PEG84.5328.246.968.350.331.150.562.493.2307.248.370.152.843.567.147.0
ViL-B(Zhang etal.,2021a)66.7443.044.765.547.629.948.058.176.1365.145.767.249.941.364.444.5
RegionViT-B+w/PEGt84.5506.448.269.951.534.351.561.793.2464.449.271.053.744.568.448.3
", + "type": "table", + "image_path": "bf53fc4a89ff6cc60c01468d30dcae85e82af638183271a3fdf8e0b3472f115d.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 106, + 103, + 505, + 147.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 106, + 147.0, + 505, + 191.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 106, + 191.0, + 505, + 235.0 + ], + "spans": [], + "index": 4 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 116, + 235, + 462, + 241 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 114, + 234, + 464, + 242 + ], + "spans": [ + { + "bbox": [ + 114, + 234, + 464, + 242 + ], + "score": 1.0, + "content": "he reported results of Swin (Liu et al., 2021) are from Twins (Chu et al., 2021a) as the original paper does not include resutls with ImageNet1K weights. †: input resolution is 896×1344.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + } + ], + "index": 3 + }, + { + "type": "table", + "bbox": [ + 118, + 264, + 492, + 310 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 207, + 253, + 403, + 263 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 205, + 250, + 405, + 266 + ], + "spans": [ + { + "bbox": [ + 205, + 250, + 405, + 266 + ], + "score": 1.0, + "content": "Table A4: Performance on person keypoint detection.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + }, + { + "type": "table_body", + "bbox": [ + 118, + 264, + 492, + 310 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 118, + 264, + 492, + 310 + ], + "spans": [ + { + "bbox": [ + 118, + 264, + 492, + 310 + ], + "score": 0.976, + "html": "
ModelFLOPs (G)Params (M)bbox AP (%)keypoint AP (%)
ResNet-50137.759.153.664.0
RegionViT-S+ (w/ PEG)172.265.756.066.1
", + "type": "table", + "image_path": "48007dead85e9ec542551c051ee803ce4bae2908e0aa7159d746e4fdd02d2b09.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 118, + 264, + 492, + 279.3333333333333 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 118, + 279.3333333333333, + 492, + 294.66666666666663 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 118, + 294.66666666666663, + 492, + 309.99999999999994 + ], + "spans": [], + "index": 9 + } + ] + } + ], + "index": 7.0 + }, + { + "type": "title", + "bbox": [ + 108, + 329, + 265, + 342 + ], + "lines": [ + { + "bbox": [ + 105, + 327, + 267, + 343 + ], + "spans": [ + { + "bbox": [ + 105, + 327, + 267, + 343 + ], + "score": 1.0, + "content": "B MORE DETAILED RESULTS", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 10 + }, + { + "type": "text", + "bbox": [ + 107, + 353, + 505, + 409 + ], + "lines": [ + { + "bbox": [ + 106, + 353, + 505, + 366 + ], + "spans": [ + { + "bbox": [ + 106, + 353, + 505, + 366 + ], + "score": 1.0, + "content": "We provided more details of some Tables shown in the main paper here. Table A2 and Table A3", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 365, + 506, + 377 + ], + "spans": [ + { + "bbox": [ + 106, + 365, + 506, + 377 + ], + "score": 1.0, + "content": "include more metrics as compared to Table 5. Table A4 shows the results on keypoint detection.", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 375, + 506, + 389 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 506, + 389 + ], + "score": 1.0, + "content": "RegionViT outperforms ResNet-50 with moderate increases in FLOPs and parameters. This suggests", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 386, + 505, + 399 + ], + "spans": [ + { + "bbox": [ + 106, + 386, + 505, + 399 + ], + "score": 1.0, + "content": "that RegionViT can model the global context in keypoint detection as well. Table A5 includes", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 398, + 450, + 410 + ], + "spans": [ + { + "bbox": [ + 105, + 398, + 450, + 410 + ], + "score": 1.0, + "content": "complexity comparison for the ablation study on regional tokens as shown in Table 8.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 13 + } + ], + "page_idx": 16, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 103, + 505, + 235 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 108, + 81, + 504, + 101 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 81, + 505, + 92 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 357, + 92 + ], + "score": 1.0, + "content": "Table A3: Object detection performance on the COCO val2017 with", + "type": "text" + }, + { + "bbox": [ + 357, + 82, + 370, + 91 + ], + "score": 0.87, + "content": "3 \\times", + "type": "inline_equation" + }, + { + "bbox": [ + 370, + 81, + 505, + 92 + ], + "score": 1.0, + "content": "schedule. The bold number indicates", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 90, + 436, + 102 + ], + "spans": [ + { + "bbox": [ + 106, + 90, + 329, + 102 + ], + "score": 1.0, + "content": "the best number within the section, and for MaskRCNN, both", + "type": "text" + }, + { + "bbox": [ + 330, + 90, + 346, + 100 + ], + "score": 0.84, + "content": "\\mathsf { A P } ^ { b }", + "type": "inline_equation" + }, + { + "bbox": [ + 346, + 90, + 362, + 102 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 362, + 91, + 382, + 100 + ], + "score": 0.81, + "content": "\\mathbf { A P } ^ { m }", + "type": "inline_equation" + }, + { + "bbox": [ + 382, + 90, + 436, + 102 + ], + "score": 1.0, + "content": "are annotated.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 106, + 103, + 505, + 235 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 103, + 505, + 235 + ], + "spans": [ + { + "bbox": [ + 106, + 103, + 505, + 235 + ], + "score": 0.982, + "html": "
BackboneParams (M)FLOPs (G)AP AP50RetineNet AP75|APsParams (M)FLOPs GAPbMaskRCNN
APM|APLAPAP5|APmAPAPm5
ResNet5037.723439.058.441.822.442.851.644.2260.041.061.744.937.158.440.1
PVT-S (Wang etal.,2021)34.2142.262.745.026.245.257.244.124543.065.346.939.962.542.8
ViL-S (Zhang etal.,2021a)35.7252.242.963.845.627.846.456.345.0174.343.464.947.039.662.142.4
Swin-T(Liu et al., 2021)38.524543.964.847.128.447.257.847.826446.068.250.241.665.144.8
Twins-SVT-S (w/PEG)(Chu et al., 2021a)34.320945.667.148.629.849.360.044.022846.869.251.242.666.345.8
RegionViT-S40.8192.645.867.249.230.050.060.550.1171.346.368.850.642.365.545.7
RegionViT-S+41.5204.246.968.350.731.151.062.050.9182.947.369.552.043.166.446.7
RegionViT-S+ w/PEG41.6204.346.768.250.230.750.862.450.9183.047.670.052.043.467.147.0
ResNet10156.731540.944.063.233642.847.138.560.141.3
ResNeXt101-32x4d56.4319.141.460.1 61.044.323.7 23.945.0 45.553.8 53.762.834044.063.2 64.448.039.261.441.9
PVT-M(Wang etal.,021)53.9143.263.846.127.346.358.963.930244.266.048.240.563.143.5
ViL-M(Zhang etal.,2021a)50.8338.943.764.646.427.947.156.960.1261.144.666.348.540.763.843.7
Swin-S (Liuetal.,221)59.833546.367.449.831.150.360.969.135447.669.452.542.866.546.4
Twins-SVT-B (w/ PEG)(Chu et al., 2021a)67.032246.968.050.231.750.361.876.334048.069.552.743.066.846.6
RegionViT-B83.4308.946.167.849.131.550.261.292.2287.947.269.151.743.066.446.5
RegionViT-B+84.4328.146.968.650.130.850.762.693.2307.148.170.252.543.567.147.1
RegionViT-B+ W/PEG84.5328.246.968.350.331.150.562.493.2307.248.370.152.843.567.147.0
ViL-B(Zhang etal.,2021a)66.7443.044.765.547.629.948.058.176.1365.145.767.249.941.364.444.5
RegionViT-B+w/PEGt84.5506.448.269.951.534.351.561.793.2464.449.271.053.744.568.448.3
", + "type": "table", + "image_path": "bf53fc4a89ff6cc60c01468d30dcae85e82af638183271a3fdf8e0b3472f115d.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 106, + 103, + 505, + 147.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 106, + 147.0, + 505, + 191.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 106, + 191.0, + 505, + 235.0 + ], + "spans": [], + "index": 4 + } + ] + }, + { + "type": "table_footnote", + "bbox": [ + 116, + 235, + 462, + 241 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 114, + 234, + 464, + 242 + ], + "spans": [ + { + "bbox": [ + 114, + 234, + 464, + 242 + ], + "score": 1.0, + "content": "he reported results of Swin (Liu et al., 2021) are from Twins (Chu et al., 2021a) as the original paper does not include resutls with ImageNet1K weights. †: input resolution is 896×1344.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + } + ], + "index": 3 + }, + { + "type": "table", + "bbox": [ + 118, + 264, + 492, + 310 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 207, + 253, + 403, + 263 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 205, + 250, + 405, + 266 + ], + "spans": [ + { + "bbox": [ + 205, + 250, + 405, + 266 + ], + "score": 1.0, + "content": "Table A4: Performance on person keypoint detection.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + }, + { + "type": "table_body", + "bbox": [ + 118, + 264, + 492, + 310 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 118, + 264, + 492, + 310 + ], + "spans": [ + { + "bbox": [ + 118, + 264, + 492, + 310 + ], + "score": 0.976, + "html": "
ModelFLOPs (G)Params (M)bbox AP (%)keypoint AP (%)
ResNet-50137.759.153.664.0
RegionViT-S+ (w/ PEG)172.265.756.066.1
", + "type": "table", + "image_path": "48007dead85e9ec542551c051ee803ce4bae2908e0aa7159d746e4fdd02d2b09.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 118, + 264, + 492, + 279.3333333333333 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 118, + 279.3333333333333, + 492, + 294.66666666666663 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 118, + 294.66666666666663, + 492, + 309.99999999999994 + ], + "spans": [], + "index": 9 + } + ] + } + ], + "index": 7.0 + }, + { + "type": "title", + "bbox": [ + 108, + 329, + 265, + 342 + ], + "lines": [ + { + "bbox": [ + 105, + 327, + 267, + 343 + ], + "spans": [ + { + "bbox": [ + 105, + 327, + 267, + 343 + ], + "score": 1.0, + "content": "B MORE DETAILED RESULTS", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 10 + }, + { + "type": "text", + "bbox": [ + 107, + 353, + 505, + 409 + ], + "lines": [ + { + "bbox": [ + 106, + 353, + 505, + 366 + ], + "spans": [ + { + "bbox": [ + 106, + 353, + 505, + 366 + ], + "score": 1.0, + "content": "We provided more details of some Tables shown in the main paper here. Table A2 and Table A3", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 365, + 506, + 377 + ], + "spans": [ + { + "bbox": [ + 106, + 365, + 506, + 377 + ], + "score": 1.0, + "content": "include more metrics as compared to Table 5. Table A4 shows the results on keypoint detection.", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 375, + 506, + 389 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 506, + 389 + ], + "score": 1.0, + "content": "RegionViT outperforms ResNet-50 with moderate increases in FLOPs and parameters. This suggests", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 386, + 505, + 399 + ], + "spans": [ + { + "bbox": [ + 106, + 386, + 505, + 399 + ], + "score": 1.0, + "content": "that RegionViT can model the global context in keypoint detection as well. Table A5 includes", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 398, + 450, + 410 + ], + "spans": [ + { + "bbox": [ + 105, + 398, + 450, + 410 + ], + "score": 1.0, + "content": "complexity comparison for the ablation study on regional tokens as shown in Table 8.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 13, + "bbox_fs": [ + 105, + 353, + 506, + 410 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 154, + 185, + 456, + 368 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 165, + 172, + 442, + 183 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 166, + 170, + 443, + 185 + ], + "spans": [ + { + "bbox": [ + 166, + 170, + 443, + 185 + ], + "score": 1.0, + "content": "Table A5: Performance w/ and w/o regional tokens on RegionViT-S.", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table_body", + "bbox": [ + 154, + 185, + 456, + 368 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 154, + 185, + 456, + 368 + ], + "spans": [ + { + "bbox": [ + 154, + 185, + 456, + 368 + ], + "score": 0.985, + "html": "
Regional TokensFLOPs (G) Params (M)ImageNet1K Acc.(%)
N5.030.482.2
Y5.330.682.6
MaskRCNNMS COCO (APb/APm)
N168.549.940.5/37.8
Y171.350.142.2/39.0
RetinaNetMS COCO (APb)
N189.840.540.7
Y192.640.841.7
SemanticFPNADE20K (mIoU)
N36.834.742.3
Y37.335.043.7
", + "type": "table", + "image_path": "30f0531d409549be7bb2f2b55c11824413373404161f84a0882854b5ae222a45.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 154, + 185, + 456, + 246.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 154, + 246.0, + 456, + 307.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 154, + 307.0, + 456, + 368.0 + ], + "spans": [], + "index": 3 + } + ] + } + ], + "index": 1.0 + }, + { + "type": "table", + "bbox": [ + 132, + 570, + 478, + 638 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 239, + 559, + 372, + 569 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 238, + 558, + 372, + 571 + ], + "spans": [ + { + "bbox": [ + 238, + 558, + 372, + 571 + ], + "score": 1.0, + "content": "Table A6: Throughput comparison.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "table_body", + "bbox": [ + 132, + 570, + 478, + 638 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 132, + 570, + 478, + 638 + ], + "spans": [ + { + "bbox": [ + 132, + 570, + 478, + 638 + ], + "score": 0.978, + "html": "
ModelAcc. (%)Params (M)FLOPs (G)Throughput (images/sec)
Swin-T81.329.04.51129
Swin-S83.050.08.7710
RegionViT-S82.630.65.3823
RegionViT-M83.141.27.4706
", + "type": "table", + "image_path": "8d4f08a814398e8f9f9f118c9808a21595d2685ec643977254ba584ab97c60e3.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 132, + 570, + 478, + 592.6666666666666 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 132, + 592.6666666666666, + 478, + 615.3333333333333 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 132, + 615.3333333333333, + 478, + 637.9999999999999 + ], + "spans": [], + "index": 7 + } + ] + } + ], + "index": 5.0 + } + ], + "page_idx": 17, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 763 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 763 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 154, + 185, + 456, + 368 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 165, + 172, + 442, + 183 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 166, + 170, + 443, + 185 + ], + "spans": [ + { + "bbox": [ + 166, + 170, + 443, + 185 + ], + "score": 1.0, + "content": "Table A5: Performance w/ and w/o regional tokens on RegionViT-S.", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table_body", + "bbox": [ + 154, + 185, + 456, + 368 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 154, + 185, + 456, + 368 + ], + "spans": [ + { + "bbox": [ + 154, + 185, + 456, + 368 + ], + "score": 0.985, + "html": "
Regional TokensFLOPs (G) Params (M)ImageNet1K Acc.(%)
N5.030.482.2
Y5.330.682.6
MaskRCNNMS COCO (APb/APm)
N168.549.940.5/37.8
Y171.350.142.2/39.0
RetinaNetMS COCO (APb)
N189.840.540.7
Y192.640.841.7
SemanticFPNADE20K (mIoU)
N36.834.742.3
Y37.335.043.7
", + "type": "table", + "image_path": "30f0531d409549be7bb2f2b55c11824413373404161f84a0882854b5ae222a45.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 154, + 185, + 456, + 246.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 154, + 246.0, + 456, + 307.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 154, + 307.0, + 456, + 368.0 + ], + "spans": [], + "index": 3 + } + ] + } + ], + "index": 1.0 + }, + { + "type": "table", + "bbox": [ + 132, + 570, + 478, + 638 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 239, + 559, + 372, + 569 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 238, + 558, + 372, + 571 + ], + "spans": [ + { + "bbox": [ + 238, + 558, + 372, + 571 + ], + "score": 1.0, + "content": "Table A6: Throughput comparison.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "table_body", + "bbox": [ + 132, + 570, + 478, + 638 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 132, + 570, + 478, + 638 + ], + "spans": [ + { + "bbox": [ + 132, + 570, + 478, + 638 + ], + "score": 0.978, + "html": "
ModelAcc. (%)Params (M)FLOPs (G)Throughput (images/sec)
Swin-T81.329.04.51129
Swin-S83.050.08.7710
RegionViT-S82.630.65.3823
RegionViT-M83.141.27.4706
", + "type": "table", + "image_path": "8d4f08a814398e8f9f9f118c9808a21595d2685ec643977254ba584ab97c60e3.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 132, + 570, + 478, + 592.6666666666666 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 132, + 592.6666666666666, + 478, + 615.3333333333333 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 132, + 615.3333333333333, + 478, + 637.9999999999999 + ], + "spans": [], + "index": 7 + } + ] + } + ], + "index": 5.0 + } + ] + } + ], + "_backend": "pipeline", + "_version_name": "2.2.2" +} \ No newline at end of file diff --git a/parse/dev/UaXD4Al3mdb/images/0c453ea1504fb873b5a8e1a2e4b4e9d29ff1ebdd9aa292d7b432f265ea0ab4c9.jpg b/parse/dev/UaXD4Al3mdb/images/0c453ea1504fb873b5a8e1a2e4b4e9d29ff1ebdd9aa292d7b432f265ea0ab4c9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..eb3fd78ce66436e27e2e09310cf7382d04a0241e --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/0c453ea1504fb873b5a8e1a2e4b4e9d29ff1ebdd9aa292d7b432f265ea0ab4c9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e009d2dbfee7c8207343129cd2bec213ad961374fb19058d19fdac1531f1540f +size 10040 diff --git a/parse/dev/UaXD4Al3mdb/images/1ba2d0d47c56dca778a4b2b810280ad0169af971d0b78dc432d51e761c535330.jpg b/parse/dev/UaXD4Al3mdb/images/1ba2d0d47c56dca778a4b2b810280ad0169af971d0b78dc432d51e761c535330.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7ec2c582c5fdeb36a69d310d9ea3026abf36976c --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/1ba2d0d47c56dca778a4b2b810280ad0169af971d0b78dc432d51e761c535330.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:937a0682974809d7072972ad3c5e9cd56d730d449a1f7676b29fa3997b6d0a91 +size 5884 diff --git a/parse/dev/UaXD4Al3mdb/images/3f211eba51de277917e4096ec9d5650bcb7965a8bc4d1e665e7a52d200b56e5e.jpg b/parse/dev/UaXD4Al3mdb/images/3f211eba51de277917e4096ec9d5650bcb7965a8bc4d1e665e7a52d200b56e5e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7d1d8a45c1fa5e7b94b1a78ca648cfffc543c432 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/3f211eba51de277917e4096ec9d5650bcb7965a8bc4d1e665e7a52d200b56e5e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4296f516a4b0563fef77f3c4410ca11301f1d1fba1d4e22c12c2c3eededc9a43 +size 25750 diff --git a/parse/dev/UaXD4Al3mdb/images/436eeeec636e0eb2f72bf1eeea3c99dadffcf3aa6014679468dde1b96b65de70.jpg b/parse/dev/UaXD4Al3mdb/images/436eeeec636e0eb2f72bf1eeea3c99dadffcf3aa6014679468dde1b96b65de70.jpg new file mode 100644 index 0000000000000000000000000000000000000000..06bb295f15c386179d73a2c310ff0441f0200106 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/436eeeec636e0eb2f72bf1eeea3c99dadffcf3aa6014679468dde1b96b65de70.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b523f929e696da25c0fdda30561e0114393629b1c1e219cc3f8bcdc1d6e3b6ed +size 46485 diff --git a/parse/dev/UaXD4Al3mdb/images/53d500b82406a84178aec5587400e5bba0386b0d5252dfbb6196ac52fb737902.jpg b/parse/dev/UaXD4Al3mdb/images/53d500b82406a84178aec5587400e5bba0386b0d5252dfbb6196ac52fb737902.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8e84365ec1b0fdbf3af2782a1884522adf955ba4 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/53d500b82406a84178aec5587400e5bba0386b0d5252dfbb6196ac52fb737902.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:510b4cd893142ea3397bd93388e2dcf5ab7426c69756f8a96cee8ca6b72e8891 +size 86881 diff --git a/parse/dev/UaXD4Al3mdb/images/602f673325fd0fb0b3149bd8daa4422913459b3ae157d61d9f9edfd431ab3809.jpg b/parse/dev/UaXD4Al3mdb/images/602f673325fd0fb0b3149bd8daa4422913459b3ae157d61d9f9edfd431ab3809.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c0a86c4dcab60499b9f1b4a34a0fb378b5af73f5 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/602f673325fd0fb0b3149bd8daa4422913459b3ae157d61d9f9edfd431ab3809.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e767b8e33ef319b23c8cd586c0367694b3dc812b0e1c48b1298ef5fe2a35f601 +size 10866 diff --git a/parse/dev/UaXD4Al3mdb/images/6064b7508ff4fe6718183f84e4b61a60b90eac71897e4287ac14ff2564a8b0d8.jpg b/parse/dev/UaXD4Al3mdb/images/6064b7508ff4fe6718183f84e4b61a60b90eac71897e4287ac14ff2564a8b0d8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1fd15519646a1655d28c1a017432acef53020108 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/6064b7508ff4fe6718183f84e4b61a60b90eac71897e4287ac14ff2564a8b0d8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eab768cbab345228614632b8eaee21a29ddd06f9c5b5a2805b39abcd31a4fcfa +size 25938 diff --git a/parse/dev/UaXD4Al3mdb/images/6b64eaf8acda880898da64a9ab2d48522bbe16c876ec064a9fac674880ba0757.jpg b/parse/dev/UaXD4Al3mdb/images/6b64eaf8acda880898da64a9ab2d48522bbe16c876ec064a9fac674880ba0757.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d30d141c5bdf6ecbd18ff8366fe458505adebc22 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/6b64eaf8acda880898da64a9ab2d48522bbe16c876ec064a9fac674880ba0757.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bbdf1b7739d4472440773a0c17ed98919a9ac0b00a3feb44f9451e3d11824689 +size 6614 diff --git a/parse/dev/UaXD4Al3mdb/images/7a05ddb2bdc3e7e4c2077af9f6e057ec8744a2d6be69cd3bca1f35b0e7dccbe7.jpg b/parse/dev/UaXD4Al3mdb/images/7a05ddb2bdc3e7e4c2077af9f6e057ec8744a2d6be69cd3bca1f35b0e7dccbe7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ea57dd82df6bc043cd528ca4848d1c550553f070 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/7a05ddb2bdc3e7e4c2077af9f6e057ec8744a2d6be69cd3bca1f35b0e7dccbe7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d9c1729e68bf3447aeba8e7313e88d5a8352c3688abc89be96b22fc5d7c5afe +size 5201 diff --git a/parse/dev/UaXD4Al3mdb/images/809bcd02e3155cdbbebd93e99d1a81624624de621b70e10a109d811956dab577.jpg b/parse/dev/UaXD4Al3mdb/images/809bcd02e3155cdbbebd93e99d1a81624624de621b70e10a109d811956dab577.jpg new file mode 100644 index 0000000000000000000000000000000000000000..46f90e9eebd019ad9126d84061de54aa9c990071 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/809bcd02e3155cdbbebd93e99d1a81624624de621b70e10a109d811956dab577.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:737cea90686c1317cea5de61df0ade9a4a483f0f599dd7a142c277e377f0ebdb +size 56791 diff --git a/parse/dev/UaXD4Al3mdb/images/8ff4bbbba00f6f5c35be210e10f2f71f199ba77e6ae592260198482b44b93b9b.jpg b/parse/dev/UaXD4Al3mdb/images/8ff4bbbba00f6f5c35be210e10f2f71f199ba77e6ae592260198482b44b93b9b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d7a6d193e53199ed9c4a8d6e841387d35af3924a --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/8ff4bbbba00f6f5c35be210e10f2f71f199ba77e6ae592260198482b44b93b9b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86e1a3af825b8e65882ee558a67e87f9f9d3ded4dd6d27f24130510cdbd0d00d +size 9537 diff --git a/parse/dev/UaXD4Al3mdb/images/9ef63fbdc5ce26164fe93437fc299de837a6cd9d0d1d3113106fc14376f4a371.jpg b/parse/dev/UaXD4Al3mdb/images/9ef63fbdc5ce26164fe93437fc299de837a6cd9d0d1d3113106fc14376f4a371.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0fb6671f2f8c4f220b5652f2f7562177d12da2c2 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/9ef63fbdc5ce26164fe93437fc299de837a6cd9d0d1d3113106fc14376f4a371.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2af5d2e5479b1761fb89d3b4d99ea19d13508ecbb55d9e20173708dfcd77efb8 +size 44685 diff --git a/parse/dev/UaXD4Al3mdb/images/d47b8dc04b34455230758031fc824c7f2501bb5ea89300154a2eef35c065d67c.jpg b/parse/dev/UaXD4Al3mdb/images/d47b8dc04b34455230758031fc824c7f2501bb5ea89300154a2eef35c065d67c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7d89a71c86af813d022f60c1ee9494674d43da33 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/d47b8dc04b34455230758031fc824c7f2501bb5ea89300154a2eef35c065d67c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24b21b2b1fbf69198c09555ec4e862e4bba5b64ea7ea17dcf82b9bab215456be +size 20525 diff --git a/parse/dev/UaXD4Al3mdb/images/f2c6b910d837c6e4c5c2bef6f560eccce0d56509c4a6fe5191e553249e994115.jpg b/parse/dev/UaXD4Al3mdb/images/f2c6b910d837c6e4c5c2bef6f560eccce0d56509c4a6fe5191e553249e994115.jpg new file mode 100644 index 0000000000000000000000000000000000000000..82c33e8967abc48316425eadabb9159e655b1902 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/f2c6b910d837c6e4c5c2bef6f560eccce0d56509c4a6fe5191e553249e994115.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43f60b905c677d2fae0ae45ac6d7e585ea45f9e7f472a7f154b4f011e17aad66 +size 43406 diff --git a/parse/dev/UaXD4Al3mdb/images/f940629f68a811776ee853f66c31bb63c073d896bc37822cd6bb54041c507012.jpg b/parse/dev/UaXD4Al3mdb/images/f940629f68a811776ee853f66c31bb63c073d896bc37822cd6bb54041c507012.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5b01f3f0777fdaf44e2a739c60f5057294f80727 --- /dev/null +++ b/parse/dev/UaXD4Al3mdb/images/f940629f68a811776ee853f66c31bb63c073d896bc37822cd6bb54041c507012.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44206043e430367e0174c83d3a9c981757bf26bde199c6718e0315436874519b +size 166847 diff --git a/parse/dev/XsZ5YebcCz/images/005c04a33b8525b98239edc9bebd5600306db9962db04612faafca5ba84da861.jpg b/parse/dev/XsZ5YebcCz/images/005c04a33b8525b98239edc9bebd5600306db9962db04612faafca5ba84da861.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7e43463a5d6adfba37698a26a4ea496d3b0d7032 --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/005c04a33b8525b98239edc9bebd5600306db9962db04612faafca5ba84da861.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1bbdbd8b92a5924955282c40ef805fe670aa10fe2c6f88235f95501f1bdd2afa +size 8415 diff --git a/parse/dev/XsZ5YebcCz/images/10588f0f74515b5e02251a11e3718944e8266309e367afc186e1604d77c6300c.jpg b/parse/dev/XsZ5YebcCz/images/10588f0f74515b5e02251a11e3718944e8266309e367afc186e1604d77c6300c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..98bd98177c6e6175e565bfb01cba223405411d9e --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/10588f0f74515b5e02251a11e3718944e8266309e367afc186e1604d77c6300c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f96912a3781204f6b149d00cd59872273404b07ec03c7cfd8aa373cd2262761f +size 6034 diff --git a/parse/dev/XsZ5YebcCz/images/1eb0d24ed1a8a8b7753c215c9eddae7dab29f99692b3c2246159b67834026397.jpg b/parse/dev/XsZ5YebcCz/images/1eb0d24ed1a8a8b7753c215c9eddae7dab29f99692b3c2246159b67834026397.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9ae1414dc38c7768c7b9e35d9ea821301ad6eeca --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/1eb0d24ed1a8a8b7753c215c9eddae7dab29f99692b3c2246159b67834026397.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50a41ec447b1ae04e4c62c443b4ba9250e4195071d45df1f170244cedfff365e +size 6542 diff --git a/parse/dev/XsZ5YebcCz/images/43748bd7958feef4df8a46c052d8dafd916a5b69ead4361b887fe1672947d296.jpg b/parse/dev/XsZ5YebcCz/images/43748bd7958feef4df8a46c052d8dafd916a5b69ead4361b887fe1672947d296.jpg new file mode 100644 index 0000000000000000000000000000000000000000..55ffd3c0d9b42ec9d7927bd51a1ed5f647e852dd --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/43748bd7958feef4df8a46c052d8dafd916a5b69ead4361b887fe1672947d296.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5da14590535116f2ce9ef3623e55032ca192c31f9e13371f867bdd5bb8cdd7b0 +size 30495 diff --git a/parse/dev/XsZ5YebcCz/images/6151ae8c9361d47cae040a9749f3d699b3a669590a8e80b1cb8a82891805430d.jpg b/parse/dev/XsZ5YebcCz/images/6151ae8c9361d47cae040a9749f3d699b3a669590a8e80b1cb8a82891805430d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..24d7da0b10502c2782c0f31d0f45d068cc3adb04 --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/6151ae8c9361d47cae040a9749f3d699b3a669590a8e80b1cb8a82891805430d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86b5cb382b0bbac8054bf65ae70c0215b07d6e9d00172cc7f0da90308e712d2d +size 48387 diff --git a/parse/dev/XsZ5YebcCz/images/62e4c4d5e512d3d89a00f40c94b7160a7fee037ff757e461b9600a8e5c1d6909.jpg b/parse/dev/XsZ5YebcCz/images/62e4c4d5e512d3d89a00f40c94b7160a7fee037ff757e461b9600a8e5c1d6909.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bdf92c510628197c1b8601b620d26d8495cc5be7 --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/62e4c4d5e512d3d89a00f40c94b7160a7fee037ff757e461b9600a8e5c1d6909.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5f35f1ec6d82b476e84622bc4a718c5b39ac339f96944b8dbb5c7daf26197fca +size 37656 diff --git a/parse/dev/XsZ5YebcCz/images/97ef96a1ed1d8d969d4790f87ed6e9e5faf24d42e3344ed8283b0aca00a0ec6c.jpg b/parse/dev/XsZ5YebcCz/images/97ef96a1ed1d8d969d4790f87ed6e9e5faf24d42e3344ed8283b0aca00a0ec6c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ca08597e50e95ae3f298a7db6eb499b733b87ac3 --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/97ef96a1ed1d8d969d4790f87ed6e9e5faf24d42e3344ed8283b0aca00a0ec6c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:637e924595722a1f9fba9569863a3ca863ffa0cb9f9fdef0b94c6e8545b94bc6 +size 7291 diff --git a/parse/dev/XsZ5YebcCz/images/983dd9cff224f46f0c342bc8c52f968f599a6d19da6a312a88b352f5154afe06.jpg b/parse/dev/XsZ5YebcCz/images/983dd9cff224f46f0c342bc8c52f968f599a6d19da6a312a88b352f5154afe06.jpg new file mode 100644 index 0000000000000000000000000000000000000000..82cd074688b2fa31978474aae1f73128db4304e9 --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/983dd9cff224f46f0c342bc8c52f968f599a6d19da6a312a88b352f5154afe06.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e4e17401aa75fef9991dcb6f88ab0e64dc031e6711c38eee9cdda47bb4ebed6 +size 6539 diff --git a/parse/dev/XsZ5YebcCz/images/9f3e7472e83e0a44082159e938d8abdaecd179da37e54b2858750ce83bd07cfd.jpg b/parse/dev/XsZ5YebcCz/images/9f3e7472e83e0a44082159e938d8abdaecd179da37e54b2858750ce83bd07cfd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..828c6a5315e58535e06fe3b6eeffe3509f0a3d88 --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/9f3e7472e83e0a44082159e938d8abdaecd179da37e54b2858750ce83bd07cfd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f225a6614a66ba8cd109d82c6c0af1c40de8cf2d51469741b35843af84e90244 +size 25838 diff --git a/parse/dev/XsZ5YebcCz/images/a8ff82a9818ff6ded39331de3c7a759a1efa8a903e976761851245bfecec0acb.jpg b/parse/dev/XsZ5YebcCz/images/a8ff82a9818ff6ded39331de3c7a759a1efa8a903e976761851245bfecec0acb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..019c900f2d42acbafc9a6acf7e7e65e172ec63a7 --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/a8ff82a9818ff6ded39331de3c7a759a1efa8a903e976761851245bfecec0acb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:746baf504cf238fd72b1091860083f1c201e7d7570497c70100de3c77477918b +size 31430 diff --git a/parse/dev/XsZ5YebcCz/images/c499ac854571be9fb82d0528ce414c8a064f5f709552c3d21462b76a77b5f877.jpg b/parse/dev/XsZ5YebcCz/images/c499ac854571be9fb82d0528ce414c8a064f5f709552c3d21462b76a77b5f877.jpg new file mode 100644 index 0000000000000000000000000000000000000000..81d8bdf492ba4f3fe98ce0dff7e490ce53baceca --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/c499ac854571be9fb82d0528ce414c8a064f5f709552c3d21462b76a77b5f877.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4f0e19d14f2c9d86f5c2f58dc8fa5f0e758ef337374110d40b465cf14f82d121 +size 8446 diff --git a/parse/dev/XsZ5YebcCz/images/c865b06259e6d45037bc8da1d40d9ae6af6b9a96d68ad622376f9537271da031.jpg b/parse/dev/XsZ5YebcCz/images/c865b06259e6d45037bc8da1d40d9ae6af6b9a96d68ad622376f9537271da031.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c08b4cf6bb6cd992ee67ff381e4598f43192331b --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/c865b06259e6d45037bc8da1d40d9ae6af6b9a96d68ad622376f9537271da031.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0cbace0e1ca52a0b74db62bed0deed5ab92b3f5c2749317f521133ef3ad8e57c +size 13176 diff --git a/parse/dev/XsZ5YebcCz/images/e1e34d8f4f75d5e81faf2e587a7429991db188f20f7bdfebc2df7de3972e7c6e.jpg b/parse/dev/XsZ5YebcCz/images/e1e34d8f4f75d5e81faf2e587a7429991db188f20f7bdfebc2df7de3972e7c6e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b01fcd719aba5ccfbaee711e72014012839e8f0c --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/e1e34d8f4f75d5e81faf2e587a7429991db188f20f7bdfebc2df7de3972e7c6e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65da3f5df4e2724f7ebbc09ebe4224270cacb396297abcda522902aef8d5e851 +size 31128 diff --git a/parse/dev/XsZ5YebcCz/images/e609e2b9a2859de7da1b1d2cb850e678c05b248dd0a82d6918b4e70d64339c9b.jpg b/parse/dev/XsZ5YebcCz/images/e609e2b9a2859de7da1b1d2cb850e678c05b248dd0a82d6918b4e70d64339c9b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5db62ac503ed66c6866a65e0d822f946b0021e0f --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/e609e2b9a2859de7da1b1d2cb850e678c05b248dd0a82d6918b4e70d64339c9b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ad0cb20c21216d523c232862ae7344a46ab399d1feae8b1d8740a941e35551e +size 23955 diff --git a/parse/dev/XsZ5YebcCz/images/e830f4aa08348dbfa6119fbb0ca9843572868c0cde0085794815b48d81e76175.jpg b/parse/dev/XsZ5YebcCz/images/e830f4aa08348dbfa6119fbb0ca9843572868c0cde0085794815b48d81e76175.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1ab7d61aacff1ba9c644dd05508cbec3faed478c --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/e830f4aa08348dbfa6119fbb0ca9843572868c0cde0085794815b48d81e76175.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc51ed6c0012f4a155f453007b5795faf212a5263457848579aec2f7649cff09 +size 136048 diff --git a/parse/dev/XsZ5YebcCz/images/fa8dbba34591351713fe459f9a5a9723c65c98498e8a47661dbe257c9ac828c6.jpg b/parse/dev/XsZ5YebcCz/images/fa8dbba34591351713fe459f9a5a9723c65c98498e8a47661dbe257c9ac828c6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0e64b477423848d4d75a4993af4bb2ace8ecc22f --- /dev/null +++ b/parse/dev/XsZ5YebcCz/images/fa8dbba34591351713fe459f9a5a9723c65c98498e8a47661dbe257c9ac828c6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41255185a74a3bc8ed1a5c43f3a27e88094132e3386d53474bbcf2f9d53ec05e +size 14997 diff --git a/parse/dev/Yo0s4qp_UMR/images/0209bf69d9800a3c8691c7ad9870c8b3e5e10893b8c9463f29ae7b06f66edf68.jpg b/parse/dev/Yo0s4qp_UMR/images/0209bf69d9800a3c8691c7ad9870c8b3e5e10893b8c9463f29ae7b06f66edf68.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1ad76ab445def5e08fef89ef6e02c0ff1081cda8 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/0209bf69d9800a3c8691c7ad9870c8b3e5e10893b8c9463f29ae7b06f66edf68.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e1118f0b6d058c46da4c9f874886eb0354a3e56f8dbbdfead50b90974e5e473b +size 4134 diff --git a/parse/dev/Yo0s4qp_UMR/images/02e5d083eacdd68ad7766920dc60b9bc14261c4d2578249d776666d8045b684e.jpg b/parse/dev/Yo0s4qp_UMR/images/02e5d083eacdd68ad7766920dc60b9bc14261c4d2578249d776666d8045b684e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..81a6928d5ba71046843d957707abdc96ba2ee333 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/02e5d083eacdd68ad7766920dc60b9bc14261c4d2578249d776666d8045b684e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:32e83782ceb5b43312bf76272d7aa3123ce4790e6240d48afc4fefc96cefbd05 +size 31010 diff --git a/parse/dev/Yo0s4qp_UMR/images/2e411dfa2b9791ff93c476e18bcdeabcd6045bf0e881a1c39c3f8743cdd56fc0.jpg b/parse/dev/Yo0s4qp_UMR/images/2e411dfa2b9791ff93c476e18bcdeabcd6045bf0e881a1c39c3f8743cdd56fc0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..64968ac2718f4d47aa59b9bf343f2518919da983 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/2e411dfa2b9791ff93c476e18bcdeabcd6045bf0e881a1c39c3f8743cdd56fc0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7bfdb74766d65e936db061af4cc5191e5f559eb1a48c657cd95b6751f0465f5c +size 5605 diff --git a/parse/dev/Yo0s4qp_UMR/images/3a06076b9fc6d610ec683780cc4cd8f6ef18a3b5e544e7b98bce40d2541d1e26.jpg b/parse/dev/Yo0s4qp_UMR/images/3a06076b9fc6d610ec683780cc4cd8f6ef18a3b5e544e7b98bce40d2541d1e26.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c5277314d7cda695d4086eed5ce9cb41538193b1 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/3a06076b9fc6d610ec683780cc4cd8f6ef18a3b5e544e7b98bce40d2541d1e26.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:81992d2cfaff7622c2c6c00d28e927e16e9f388fd2658133d29b055d85570681 +size 5319 diff --git a/parse/dev/Yo0s4qp_UMR/images/5e72a3b1c9197c1e6d8978b36c36c7345edc21ede33dab63c47567f298b8f56a.jpg b/parse/dev/Yo0s4qp_UMR/images/5e72a3b1c9197c1e6d8978b36c36c7345edc21ede33dab63c47567f298b8f56a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..23452c22fac46fd806dd71e4f5c69d5a50795345 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/5e72a3b1c9197c1e6d8978b36c36c7345edc21ede33dab63c47567f298b8f56a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8fae12d8bdb1c374f83c9662fd60fb6244eb2090bbbc52375bf2a8495f4fc335 +size 57431 diff --git a/parse/dev/Yo0s4qp_UMR/images/a17d9947254d341597afbb378d4bde6d1babd3fcdda71e05033eb14b2c82df80.jpg b/parse/dev/Yo0s4qp_UMR/images/a17d9947254d341597afbb378d4bde6d1babd3fcdda71e05033eb14b2c82df80.jpg new file mode 100644 index 0000000000000000000000000000000000000000..83e3758b21c520d706cfdaeb90e6668357454128 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/a17d9947254d341597afbb378d4bde6d1babd3fcdda71e05033eb14b2c82df80.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ed13fb8f6127761dced89bc7508b577ad2503074c604ac7ad9fb985d2195532 +size 4053 diff --git a/parse/dev/Yo0s4qp_UMR/images/a2995b422d221b5820c061c7aae2c291d569daedc137130abf966319d05b59ad.jpg b/parse/dev/Yo0s4qp_UMR/images/a2995b422d221b5820c061c7aae2c291d569daedc137130abf966319d05b59ad.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9e6d203daaa4de7adea2dd2c2d7fd4add8f1462b --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/a2995b422d221b5820c061c7aae2c291d569daedc137130abf966319d05b59ad.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b372b0fccb29faefc76b8d87a6122a68da78979638b7ec4801505c1696d9e3d +size 30037 diff --git a/parse/dev/Yo0s4qp_UMR/images/a7fdc4110b697c986604c95f811fcceac55acdd57a25051919ac257cebbffba7.jpg b/parse/dev/Yo0s4qp_UMR/images/a7fdc4110b697c986604c95f811fcceac55acdd57a25051919ac257cebbffba7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f4ad1c51f251ebbd1912fc90a9c9ff959bd02b18 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/a7fdc4110b697c986604c95f811fcceac55acdd57a25051919ac257cebbffba7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b69b5bef1273c547384d3d9086e48d7fd977e2fe62d65adea69d131331f1fcc +size 4513 diff --git a/parse/dev/Yo0s4qp_UMR/images/d8a1488f4ac94fcba9d774c8cdd8a9082181dc8120edbcd969c3bcaeb117c0ee.jpg b/parse/dev/Yo0s4qp_UMR/images/d8a1488f4ac94fcba9d774c8cdd8a9082181dc8120edbcd969c3bcaeb117c0ee.jpg new file mode 100644 index 0000000000000000000000000000000000000000..46dced7889d7d391bf923d30d2d3b5c9d3dadda6 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/d8a1488f4ac94fcba9d774c8cdd8a9082181dc8120edbcd969c3bcaeb117c0ee.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:98c93b9f715139b5344ca139116621b2a1fcb28bc5c73982c87daad3ac3ccdba +size 42934 diff --git a/parse/dev/Yo0s4qp_UMR/images/dc2005c01952b7890c907d0d90b20dd1c2b2a85fc3d6562c83ee3d52298d0ffb.jpg b/parse/dev/Yo0s4qp_UMR/images/dc2005c01952b7890c907d0d90b20dd1c2b2a85fc3d6562c83ee3d52298d0ffb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4f83bc5472c017f755e049a8bc78def0ab0264e6 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/dc2005c01952b7890c907d0d90b20dd1c2b2a85fc3d6562c83ee3d52298d0ffb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d5775338e142ae05477476431e8d063519a7cadcd40cb3f251038e3351ddfe59 +size 7224 diff --git a/parse/dev/Yo0s4qp_UMR/images/dfbe593c05537ccb230959612b42a99359ada15be5c0d17a4c070f3fdc0eedc3.jpg b/parse/dev/Yo0s4qp_UMR/images/dfbe593c05537ccb230959612b42a99359ada15be5c0d17a4c070f3fdc0eedc3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7ef53d9af3a3c240a888117f7099341156c35c2c --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/dfbe593c05537ccb230959612b42a99359ada15be5c0d17a4c070f3fdc0eedc3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:52c65138587589ee2ef0372c7d4e23192a9fdf673de9248a42588a21787647c6 +size 15030 diff --git a/parse/dev/Yo0s4qp_UMR/images/e66d34315d4ed36ec806b233fe5262a49ffc13427173c81d956b896a44aad661.jpg b/parse/dev/Yo0s4qp_UMR/images/e66d34315d4ed36ec806b233fe5262a49ffc13427173c81d956b896a44aad661.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6b6c3e47d7f739fba30ebac6d962e99d0d6d4d06 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/e66d34315d4ed36ec806b233fe5262a49ffc13427173c81d956b896a44aad661.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d6f54b1389ade6f85c7651979e8a1938dada1a88145ef6c3bfece33f57c531f +size 10082 diff --git a/parse/dev/Yo0s4qp_UMR/images/e85808dcc13778457f8dfd7670925e6dd8f71e152c941fd4da5ec50fa991e172.jpg b/parse/dev/Yo0s4qp_UMR/images/e85808dcc13778457f8dfd7670925e6dd8f71e152c941fd4da5ec50fa991e172.jpg new file mode 100644 index 0000000000000000000000000000000000000000..66db7152ffa18237d6abfa074b20bdcd0365f9f2 --- /dev/null +++ b/parse/dev/Yo0s4qp_UMR/images/e85808dcc13778457f8dfd7670925e6dd8f71e152c941fd4da5ec50fa991e172.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5f5fd384c689859933fe4b1d367950b623b4c678d4fce9362f02cba13c62a55 +size 5993 diff --git a/parse/dev/a6NvoZ5DLoe/images/0d9f37094b57c7d3d7abd24dfe5251f1f6b197322ca122f8a582da6e2ade3688.jpg b/parse/dev/a6NvoZ5DLoe/images/0d9f37094b57c7d3d7abd24dfe5251f1f6b197322ca122f8a582da6e2ade3688.jpg new file mode 100644 index 0000000000000000000000000000000000000000..33320c7f2fc569aa9222f32ae02005813e9e35a3 --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/0d9f37094b57c7d3d7abd24dfe5251f1f6b197322ca122f8a582da6e2ade3688.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75d4a865906b88963da6509f2bb417cb61130e2cbe98c9565e00976dc32d9938 +size 22454 diff --git a/parse/dev/a6NvoZ5DLoe/images/130ed901cdfcb0707b1662c907f0ca3cd7ffee1ee2641ab6d8b0b71370939f06.jpg b/parse/dev/a6NvoZ5DLoe/images/130ed901cdfcb0707b1662c907f0ca3cd7ffee1ee2641ab6d8b0b71370939f06.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cdafb551bcf32360e1f0d57d6ebfb5b939183ffb --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/130ed901cdfcb0707b1662c907f0ca3cd7ffee1ee2641ab6d8b0b71370939f06.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:83c30f0c27f13366890a019fe28c8c83692c31254b7837382a19c150636b731e +size 10404 diff --git a/parse/dev/a6NvoZ5DLoe/images/19d199fddf1d95b0b9acdb841afb0161e8a7f42f1d2576d03e1d16cf1688dbc9.jpg b/parse/dev/a6NvoZ5DLoe/images/19d199fddf1d95b0b9acdb841afb0161e8a7f42f1d2576d03e1d16cf1688dbc9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2c1ed9e6bf4805b1d33b8b85138968e206a496d1 --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/19d199fddf1d95b0b9acdb841afb0161e8a7f42f1d2576d03e1d16cf1688dbc9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad785711fdd1aa2471a9a06f13e0c5ab3ca9d4f9404242cf9588b62ff6994d20 +size 54307 diff --git a/parse/dev/a6NvoZ5DLoe/images/2a2a0b48dfbd54c6cf8aaeecddfe58c7e7589d51cb62f095f513998efc58b637.jpg b/parse/dev/a6NvoZ5DLoe/images/2a2a0b48dfbd54c6cf8aaeecddfe58c7e7589d51cb62f095f513998efc58b637.jpg new file mode 100644 index 0000000000000000000000000000000000000000..01799b3a61a0f4ef4dadd1d6c81b9655a109c850 --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/2a2a0b48dfbd54c6cf8aaeecddfe58c7e7589d51cb62f095f513998efc58b637.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f45dcdedba1184bb101cc40be149383428258a23b43579d60dadbc670d7abaa2 +size 58886 diff --git a/parse/dev/a6NvoZ5DLoe/images/3c908dcb4ee782da90d563a5876e5fd8b61ca72308f3fc14c28dca8a4987bcba.jpg b/parse/dev/a6NvoZ5DLoe/images/3c908dcb4ee782da90d563a5876e5fd8b61ca72308f3fc14c28dca8a4987bcba.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cff1a37e20e94a234b3e458e30066ad5baa5ebff --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/3c908dcb4ee782da90d563a5876e5fd8b61ca72308f3fc14c28dca8a4987bcba.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0fdf918d94114b60e024595c1ff312bfd8eb31b58cf165918cbad5d84bf0dd7e +size 21406 diff --git a/parse/dev/a6NvoZ5DLoe/images/3dbcd5efa23c6c3fb14257d85f5d11cd2bf718edfa48545ae6c8b1c4e13372ff.jpg b/parse/dev/a6NvoZ5DLoe/images/3dbcd5efa23c6c3fb14257d85f5d11cd2bf718edfa48545ae6c8b1c4e13372ff.jpg new file mode 100644 index 0000000000000000000000000000000000000000..76b70a88688fcd048d1e8ce3be68daa65b34c212 --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/3dbcd5efa23c6c3fb14257d85f5d11cd2bf718edfa48545ae6c8b1c4e13372ff.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a6782bda4b3315d3aca48d5c2725a5a29379ac760fc6d86f821c66254d46537 +size 4301 diff --git a/parse/dev/a6NvoZ5DLoe/images/4a28b510312ef9ea68dcfabfbd81cc16c232fcc685e92e3439c1c2de93b28e9a.jpg b/parse/dev/a6NvoZ5DLoe/images/4a28b510312ef9ea68dcfabfbd81cc16c232fcc685e92e3439c1c2de93b28e9a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0d1e3efee092ee6e1d06dfd7dddc8120d3859515 --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/4a28b510312ef9ea68dcfabfbd81cc16c232fcc685e92e3439c1c2de93b28e9a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02beb1ea29f163201efb7694ec2a93c009399669044493ce69dbec44d4a2e0bc +size 41759 diff --git a/parse/dev/a6NvoZ5DLoe/images/4c2430d5add1a767687c9f209ad59cb6998c4ae2e65e9f0aa2b74a9eb39efa48.jpg b/parse/dev/a6NvoZ5DLoe/images/4c2430d5add1a767687c9f209ad59cb6998c4ae2e65e9f0aa2b74a9eb39efa48.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8b391f1a8f0897b74532cc101616a74adbb9dfd2 --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/4c2430d5add1a767687c9f209ad59cb6998c4ae2e65e9f0aa2b74a9eb39efa48.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f74caa803edd12ae93d6392ddb4504c3f1015b2dd49ae51959e74b3c7cfbce8 +size 17227 diff --git a/parse/dev/a6NvoZ5DLoe/images/6c1342051899b421ca0dc688654fd6dbc42add049507cb235afdc37ea844fbeb.jpg b/parse/dev/a6NvoZ5DLoe/images/6c1342051899b421ca0dc688654fd6dbc42add049507cb235afdc37ea844fbeb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..80abd5a9bf18bda51131c91ea29298f744e11a6a --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/6c1342051899b421ca0dc688654fd6dbc42add049507cb235afdc37ea844fbeb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fd37ce579fe6f3011345a9009c66b0f1e92252cc101de548efd9261c10b19fdb +size 11535 diff --git a/parse/dev/a6NvoZ5DLoe/images/7083d65d551627d042a21decaa25bf06e657a2e867400ae638752f699a24960b.jpg b/parse/dev/a6NvoZ5DLoe/images/7083d65d551627d042a21decaa25bf06e657a2e867400ae638752f699a24960b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4ba94af2c7de7e386ef0d8e4f6deecfc8946114e --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/7083d65d551627d042a21decaa25bf06e657a2e867400ae638752f699a24960b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a95f4c104cf9d937d6bf2a477c9044d5bd150b3291b7791176edad02d7a3fec5 +size 80704 diff --git a/parse/dev/a6NvoZ5DLoe/images/709da4b0ceb9013bd7c71e63c19926cd2c798fbce18a58300e5695187cec3023.jpg b/parse/dev/a6NvoZ5DLoe/images/709da4b0ceb9013bd7c71e63c19926cd2c798fbce18a58300e5695187cec3023.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f2eea97cbcc44fc3e6030ef036e18094891a0901 --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/709da4b0ceb9013bd7c71e63c19926cd2c798fbce18a58300e5695187cec3023.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d6a5a68bb47b8c4f6870fbbcb1168d301ac7aec7f72764928adb430b936c52f4 +size 5281 diff --git a/parse/dev/a6NvoZ5DLoe/images/87528c817fe73604142f62ee6e0d8d06fd74e5c9a36abadba289a003b94edb67.jpg b/parse/dev/a6NvoZ5DLoe/images/87528c817fe73604142f62ee6e0d8d06fd74e5c9a36abadba289a003b94edb67.jpg new file mode 100644 index 0000000000000000000000000000000000000000..50a0ad63c7d46128815e4586c1644531cb39219f --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/87528c817fe73604142f62ee6e0d8d06fd74e5c9a36abadba289a003b94edb67.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45dc1e15fca8fe632debf1ee421f85f0291c23abb43390eb4fd43bc99872e6ce +size 60774 diff --git a/parse/dev/a6NvoZ5DLoe/images/9bf17d192e9c1040cec0a29756dd246a5f98b6f6bb3d062cf3981eadb6049e71.jpg b/parse/dev/a6NvoZ5DLoe/images/9bf17d192e9c1040cec0a29756dd246a5f98b6f6bb3d062cf3981eadb6049e71.jpg new file mode 100644 index 0000000000000000000000000000000000000000..eb6c87dd61c365ba6c8d441faf4131eae23389c9 --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/9bf17d192e9c1040cec0a29756dd246a5f98b6f6bb3d062cf3981eadb6049e71.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75dc7a4c8a2d0895be4bced674c46a27d4f578367f8e95babe1b37d02c3e4a4b +size 3955 diff --git a/parse/dev/a6NvoZ5DLoe/images/b44c24a1c5e8f05e8992e5bf1c61ff3ddb3c6387cecbf6a9fe4e645afa3fca7b.jpg b/parse/dev/a6NvoZ5DLoe/images/b44c24a1c5e8f05e8992e5bf1c61ff3ddb3c6387cecbf6a9fe4e645afa3fca7b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..248e70fd17733b9e3deb438383ba930843df8689 --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/b44c24a1c5e8f05e8992e5bf1c61ff3ddb3c6387cecbf6a9fe4e645afa3fca7b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:546731dda27b8d70733cd41fa23d487fdc966adf447af4f63443563ffc0fe6a7 +size 6683 diff --git a/parse/dev/a6NvoZ5DLoe/images/cbcaa941a8288759368287797a349f15ac8f740d6270ebc0daa81f890f2b8000.jpg b/parse/dev/a6NvoZ5DLoe/images/cbcaa941a8288759368287797a349f15ac8f740d6270ebc0daa81f890f2b8000.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0d3be089aa6a063a19dc7746859e7d93654a858f --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/cbcaa941a8288759368287797a349f15ac8f740d6270ebc0daa81f890f2b8000.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df22a1755da425333c24dad45da5c02a8fd4b49f88e34630e34bbe0dd01fdef6 +size 3320 diff --git a/parse/dev/a6NvoZ5DLoe/images/ccc3044c887030e4065776e0c2dbb22b1efdcba2cb8c9713385823b409169c06.jpg b/parse/dev/a6NvoZ5DLoe/images/ccc3044c887030e4065776e0c2dbb22b1efdcba2cb8c9713385823b409169c06.jpg new file mode 100644 index 0000000000000000000000000000000000000000..941d58279d0e8e85309a342a66556c68905c5b24 --- /dev/null +++ b/parse/dev/a6NvoZ5DLoe/images/ccc3044c887030e4065776e0c2dbb22b1efdcba2cb8c9713385823b409169c06.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43d5080ae865e75bbc59013ab016a24b9f9247f81ee46c12085c06e1bc015bf5 +size 31041 diff --git a/parse/dev/cGDAkQo1C0p/images/4740973eea46730c1be546110e00dd25add8f7a9b0715889a1bc9ad7b6736b05.jpg b/parse/dev/cGDAkQo1C0p/images/4740973eea46730c1be546110e00dd25add8f7a9b0715889a1bc9ad7b6736b05.jpg new file mode 100644 index 0000000000000000000000000000000000000000..09170f73e1a2dc8999315c26395813d5d0076e9c --- /dev/null +++ b/parse/dev/cGDAkQo1C0p/images/4740973eea46730c1be546110e00dd25add8f7a9b0715889a1bc9ad7b6736b05.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07bb9e00ebefd6af5c108a5ffb96837c0103a7b1f6334bd0b870eb8e41a263a3 +size 8204 diff --git a/parse/dev/cJPkX1g9PQS/images/0daf6f18dab40fa3d960ec8612f0d4da86818d0e681f48dd5dcbad89d6c578ed.jpg b/parse/dev/cJPkX1g9PQS/images/0daf6f18dab40fa3d960ec8612f0d4da86818d0e681f48dd5dcbad89d6c578ed.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fbda6d6c4036a9ea23814d0aa5c2cf816b564cf1 --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/0daf6f18dab40fa3d960ec8612f0d4da86818d0e681f48dd5dcbad89d6c578ed.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7523d3ae27449b43e08a345095a9b0f03b83d598f2972cfdbf3ffda8d8b9aa32 +size 3139 diff --git a/parse/dev/cJPkX1g9PQS/images/1b1e867d91637c4e79bab8a6246f54e3a19dd41221050de2a084556836695ca8.jpg b/parse/dev/cJPkX1g9PQS/images/1b1e867d91637c4e79bab8a6246f54e3a19dd41221050de2a084556836695ca8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a3f8f1df8bcced3e80b3fd05d8cddb49057b3572 --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/1b1e867d91637c4e79bab8a6246f54e3a19dd41221050de2a084556836695ca8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:40ba90d255c1a16b764c66a19883d27e8fba7a6190d927af92022815803e7405 +size 79251 diff --git a/parse/dev/cJPkX1g9PQS/images/237e4bc07b171a830423814265c89b10bb3557f789d26217d60d33d749ef7417.jpg b/parse/dev/cJPkX1g9PQS/images/237e4bc07b171a830423814265c89b10bb3557f789d26217d60d33d749ef7417.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8ac9cfdc89287f6874b20c36f1882d4c6bef20a3 --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/237e4bc07b171a830423814265c89b10bb3557f789d26217d60d33d749ef7417.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a082116f460ca0165e0850f0c930dd0df673ce56c63a0da917bcd4be7ef7d789 +size 12840 diff --git a/parse/dev/cJPkX1g9PQS/images/3f5ff71f26480152d8ad1bff0fe0f121b4bfeb8db790225ff661e6fc916288f1.jpg b/parse/dev/cJPkX1g9PQS/images/3f5ff71f26480152d8ad1bff0fe0f121b4bfeb8db790225ff661e6fc916288f1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..422ec2e8aec3e6e7f879af5c4ed622c22efff488 --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/3f5ff71f26480152d8ad1bff0fe0f121b4bfeb8db790225ff661e6fc916288f1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f63f76e669119cad937167dc8b44619e25b78716f47b3455730d0916a749c38 +size 52738 diff --git a/parse/dev/cJPkX1g9PQS/images/599b82d734eecaabb93eb20ea5d89e663190cd80eeee47cd484489d9033cd2b7.jpg b/parse/dev/cJPkX1g9PQS/images/599b82d734eecaabb93eb20ea5d89e663190cd80eeee47cd484489d9033cd2b7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ab9c110fd871c4e90ffbca842fd1ed1f1698f1cd --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/599b82d734eecaabb93eb20ea5d89e663190cd80eeee47cd484489d9033cd2b7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:acf0724752f782dbf88eab2d2bec0cd98c190ae05bedd8ff949bd4949da09b9a +size 43115 diff --git a/parse/dev/cJPkX1g9PQS/images/5b5509c8b1aa4d3c69b60073f52daa4294dd4ccb0448515196e8d287c492f34b.jpg b/parse/dev/cJPkX1g9PQS/images/5b5509c8b1aa4d3c69b60073f52daa4294dd4ccb0448515196e8d287c492f34b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3b2f8901259302849fc1873c646124d58ff6485b --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/5b5509c8b1aa4d3c69b60073f52daa4294dd4ccb0448515196e8d287c492f34b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13e17d1774ece302dbb6ff2d352ae98bcdadfb25f7485ccd4befa551fef1fa87 +size 47021 diff --git a/parse/dev/cJPkX1g9PQS/images/770c8b3f3235528837c4945720e937b2757d9b17b1e6cb6fd2a42db03574138f.jpg b/parse/dev/cJPkX1g9PQS/images/770c8b3f3235528837c4945720e937b2757d9b17b1e6cb6fd2a42db03574138f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..941d95c110bdda8b78664441158c47253ea3efda --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/770c8b3f3235528837c4945720e937b2757d9b17b1e6cb6fd2a42db03574138f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02a8cd42135ca3388bf6bd0aeebd063d5c2f3ddf93d371075ececfecd9e4dbf4 +size 64018 diff --git a/parse/dev/cJPkX1g9PQS/images/7bfb703ef14c8f47e4359574338ee9b995f9ac7618fdc2a4784efd18a303cc16.jpg b/parse/dev/cJPkX1g9PQS/images/7bfb703ef14c8f47e4359574338ee9b995f9ac7618fdc2a4784efd18a303cc16.jpg new file mode 100644 index 0000000000000000000000000000000000000000..309324f3ce57565bc840f76c8a4d38e2f16bd9c3 --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/7bfb703ef14c8f47e4359574338ee9b995f9ac7618fdc2a4784efd18a303cc16.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23750776af86653bbb9bc02e30b042c564cffa6dbab65587cc3bafc69833ea8d +size 2994 diff --git a/parse/dev/cJPkX1g9PQS/images/862e8b6570dc757cf9df8f76b818a562f25803bacb9427fdb3d56d7ec6ae07ee.jpg b/parse/dev/cJPkX1g9PQS/images/862e8b6570dc757cf9df8f76b818a562f25803bacb9427fdb3d56d7ec6ae07ee.jpg new file mode 100644 index 0000000000000000000000000000000000000000..aab79c2461018ecac24458a5d574ddd28c49aa01 --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/862e8b6570dc757cf9df8f76b818a562f25803bacb9427fdb3d56d7ec6ae07ee.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d6065919ac0104ab140dd4db68245f129233c41992763ac2b7f95cf1fc879447 +size 59868 diff --git a/parse/dev/cJPkX1g9PQS/images/8dbacbb6c09f9acf73ac26d0542725399b44d44c4b6eeacd1c5f7843dbe7e755.jpg b/parse/dev/cJPkX1g9PQS/images/8dbacbb6c09f9acf73ac26d0542725399b44d44c4b6eeacd1c5f7843dbe7e755.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a7810e7bd6d892561423a9d84b3270038795a760 --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/8dbacbb6c09f9acf73ac26d0542725399b44d44c4b6eeacd1c5f7843dbe7e755.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee9e13b020053044976c1cf7b5533dacdc3744dd72935ec0a8e9ccddba8c0784 +size 125066 diff --git a/parse/dev/cJPkX1g9PQS/images/8faf3d71105bf5d4f60db1df9c9af9c06a5c03233b4d5f13576165a3bbc6467d.jpg b/parse/dev/cJPkX1g9PQS/images/8faf3d71105bf5d4f60db1df9c9af9c06a5c03233b4d5f13576165a3bbc6467d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..64ca3abba0d24dd7820eddb9be459c44ddc07194 --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/8faf3d71105bf5d4f60db1df9c9af9c06a5c03233b4d5f13576165a3bbc6467d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ac87aefc3236d9c2c539cf2c49273c2321b7ea459f4c7a2f8b5a189097b051f5 +size 97866 diff --git a/parse/dev/cJPkX1g9PQS/images/a2065b45711674c1ff47aed93bbceb3759a430c9f060e5e582c75ee25a3926c8.jpg b/parse/dev/cJPkX1g9PQS/images/a2065b45711674c1ff47aed93bbceb3759a430c9f060e5e582c75ee25a3926c8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2af57e03d2d45e5937681ad8924a9c789576800b --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/a2065b45711674c1ff47aed93bbceb3759a430c9f060e5e582c75ee25a3926c8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:01b12a8c2b8a9db243cc55db859e335e06028eff08d15f73f0839097781870b8 +size 9173 diff --git a/parse/dev/cJPkX1g9PQS/images/a8eb74f3bb306891b2369d7ab34f86a2517655ce5eab462ffe58e0d8d69eed2f.jpg b/parse/dev/cJPkX1g9PQS/images/a8eb74f3bb306891b2369d7ab34f86a2517655ce5eab462ffe58e0d8d69eed2f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..311ef5ae4983364ec2b4a0bbd14c1b9bcd67626d --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/a8eb74f3bb306891b2369d7ab34f86a2517655ce5eab462ffe58e0d8d69eed2f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ac7669fed4ee430f2b673bb062236b437253ab92e1f962ce210b2a5317b4522 +size 51297 diff --git a/parse/dev/cJPkX1g9PQS/images/e81d2177e5f4efabbee3d9d423dab4c2929907f28fc8dfeae9e49ada2de4754b.jpg b/parse/dev/cJPkX1g9PQS/images/e81d2177e5f4efabbee3d9d423dab4c2929907f28fc8dfeae9e49ada2de4754b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a01dc83143865dfd597c38df5a09f354cee5ddb3 --- /dev/null +++ b/parse/dev/cJPkX1g9PQS/images/e81d2177e5f4efabbee3d9d423dab4c2929907f28fc8dfeae9e49ada2de4754b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7bf98bccba5029f6d5a16c3b4341619a7c0b63d4919f1f41e876ffcce9b81d46 +size 5781 diff --git a/parse/dev/cZAi1yWpiXQ/cZAi1yWpiXQ_model.json b/parse/dev/cZAi1yWpiXQ/cZAi1yWpiXQ_model.json new file mode 100644 index 0000000000000000000000000000000000000000..ecfedc8f20e6690e92533a623d3964b20e31266b --- /dev/null +++ b/parse/dev/cZAi1yWpiXQ/cZAi1yWpiXQ_model.json @@ -0,0 +1,26887 @@ +[ + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 398, + 746, + 1302, + 746, + 1302, + 1234, + 398, + 1234 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 1586, + 1404, + 1586, + 1404, + 1952, + 298, + 1952 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 299, + 1386, + 1403, + 1386, + 1403, + 1571, + 299, + 1571 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 314, + 448, + 1386, + 448, + 1386, + 589, + 314, + 589 + ], + "score": 0.965 + }, + { + "category_id": 0, + "poly": [ + 299, + 220, + 1402, + 220, + 1402, + 323, + 299, + 323 + ], + "score": 0.96 + }, + { + "category_id": 1, + "poly": [ + 314, + 369, + 1359, + 369, + 1359, + 446, + 314, + 446 + ], + "score": 0.926 + }, + { + "category_id": 2, + "poly": [ + 337, + 2005, + 872, + 2005, + 872, + 2034, + 337, + 2034 + ], + "score": 0.91 + }, + { + "category_id": 0, + "poly": [ + 302, + 1312, + 573, + 1312, + 573, + 1346, + 302, + 1346 + ], + "score": 0.896 + }, + { + "category_id": 0, + "poly": [ + 773, + 670, + 927, + 670, + 927, + 704, + 773, + 704 + ], + "score": 0.89 + }, + { + "category_id": 2, + "poly": [ + 299, + 75, + 816, + 75, + 816, + 105, + 299, + 105 + ], + "score": 0.88 + }, + { + "category_id": 2, + "poly": [ + 841, + 2088, + 857, + 2088, + 857, + 2112, + 841, + 2112 + ], + "score": 0.715 + }, + { + "category_id": 15, + "poly": [ + 295.0, + 216.0, + 1406.0, + 216.0, + 1406.0, + 274.0, + 295.0, + 274.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 278.0, + 827.0, + 278.0, + 827.0, + 326.0, + 297.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 340.0, + 2001.0, + 877.0, + 2001.0, + 877.0, + 2040.0, + 340.0, + 2040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1308.0, + 579.0, + 1308.0, + 579.0, + 1354.0, + 294.0, + 1354.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 770.0, + 669.0, + 932.0, + 669.0, + 932.0, + 707.0, + 770.0, + 707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 295.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2087.0, + 860.0, + 2087.0, + 860.0, + 2117.0, + 840.0, + 2117.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 746.0, + 1303.0, + 746.0, + 1303.0, + 780.0, + 395.0, + 780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 776.0, + 1306.0, + 776.0, + 1306.0, + 813.0, + 393.0, + 813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 805.0, + 1304.0, + 805.0, + 1304.0, + 843.0, + 393.0, + 843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 839.0, + 1305.0, + 839.0, + 1305.0, + 871.0, + 395.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 867.0, + 1306.0, + 867.0, + 1306.0, + 901.0, + 394.0, + 901.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 898.0, + 1306.0, + 898.0, + 1306.0, + 932.0, + 394.0, + 932.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 931.0, + 1306.0, + 931.0, + 1306.0, + 961.0, + 393.0, + 961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 961.0, + 1305.0, + 961.0, + 1305.0, + 989.0, + 395.0, + 989.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 991.0, + 1305.0, + 991.0, + 1305.0, + 1023.0, + 394.0, + 1023.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1020.0, + 1305.0, + 1020.0, + 1305.0, + 1053.0, + 394.0, + 1053.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1050.0, + 1305.0, + 1050.0, + 1305.0, + 1085.0, + 394.0, + 1085.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1082.0, + 1306.0, + 1082.0, + 1306.0, + 1116.0, + 392.0, + 1116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1111.0, + 1306.0, + 1111.0, + 1306.0, + 1145.0, + 393.0, + 1145.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 1144.0, + 1305.0, + 1144.0, + 1305.0, + 1176.0, + 395.0, + 1176.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1172.0, + 1306.0, + 1172.0, + 1306.0, + 1208.0, + 393.0, + 1208.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 1201.0, + 991.0, + 1201.0, + 991.0, + 1239.0, + 395.0, + 1239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1585.0, + 1405.0, + 1585.0, + 1405.0, + 1622.0, + 293.0, + 1622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1616.0, + 1403.0, + 1616.0, + 1403.0, + 1651.0, + 293.0, + 1651.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1644.0, + 1405.0, + 1644.0, + 1405.0, + 1683.0, + 293.0, + 1683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1678.0, + 1405.0, + 1678.0, + 1405.0, + 1711.0, + 293.0, + 1711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1710.0, + 1404.0, + 1710.0, + 1404.0, + 1742.0, + 293.0, + 1742.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1737.0, + 1405.0, + 1737.0, + 1405.0, + 1773.0, + 293.0, + 1773.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1771.0, + 1404.0, + 1771.0, + 1404.0, + 1802.0, + 296.0, + 1802.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1799.0, + 1405.0, + 1799.0, + 1405.0, + 1835.0, + 292.0, + 1835.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1831.0, + 1406.0, + 1831.0, + 1406.0, + 1865.0, + 292.0, + 1865.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1856.0, + 1405.0, + 1856.0, + 1405.0, + 1898.0, + 292.0, + 1898.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1892.0, + 1404.0, + 1892.0, + 1404.0, + 1924.0, + 293.0, + 1924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1922.0, + 1402.0, + 1922.0, + 1402.0, + 1956.0, + 294.0, + 1956.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1386.0, + 1405.0, + 1386.0, + 1405.0, + 1421.0, + 295.0, + 1421.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1419.0, + 1404.0, + 1419.0, + 1404.0, + 1451.0, + 297.0, + 1451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1448.0, + 1404.0, + 1448.0, + 1404.0, + 1484.0, + 295.0, + 1484.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1476.0, + 1403.0, + 1476.0, + 1403.0, + 1514.0, + 294.0, + 1514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1508.0, + 1403.0, + 1508.0, + 1403.0, + 1543.0, + 295.0, + 1543.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1540.0, + 1382.0, + 1540.0, + 1382.0, + 1571.0, + 297.0, + 1571.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 311.0, + 441.0, + 1214.0, + 441.0, + 1214.0, + 488.0, + 311.0, + 488.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 473.0, + 964.0, + 473.0, + 964.0, + 527.0, + 308.0, + 527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 510.0, + 1391.0, + 510.0, + 1391.0, + 564.0, + 307.0, + 564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 310.0, + 550.0, + 1291.0, + 550.0, + 1291.0, + 596.0, + 310.0, + 596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 564.0, + 365.0, + 804.0, + 365.0, + 804.0, + 418.0, + 564.0, + 418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 809.0, + 364.0, + 1361.0, + 364.0, + 1361.0, + 414.0, + 809.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 461.0, + 406.0, + 926.0, + 406.0, + 926.0, + 452.0, + 461.0, + 452.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 364.0, + 562.0, + 364.0, + 562.0, + 420.0, + 302.0, + 420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 299.75, + 400.5, + 450.75, + 400.5, + 450.75, + 451.5, + 299.75, + 451.5 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 0, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 461, + 1403, + 461, + 1403, + 826, + 298, + 826 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 298, + 843, + 1404, + 843, + 1404, + 1209, + 298, + 1209 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 298, + 1711, + 1403, + 1711, + 1403, + 1957, + 298, + 1957 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 298, + 229, + 1403, + 229, + 1403, + 445, + 298, + 445 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 372, + 1284, + 1403, + 1284, + 1403, + 1591, + 372, + 1591 + ], + "score": 0.956 + }, + { + "category_id": 1, + "poly": [ + 298, + 1972, + 1399, + 1972, + 1399, + 2034, + 298, + 2034 + ], + "score": 0.934 + }, + { + "category_id": 1, + "poly": [ + 298, + 1224, + 606, + 1224, + 606, + 1256, + 298, + 1256 + ], + "score": 0.926 + }, + { + "category_id": 0, + "poly": [ + 303, + 1643, + 1121, + 1643, + 1121, + 1675, + 303, + 1675 + ], + "score": 0.898 + }, + { + "category_id": 2, + "poly": [ + 300, + 76, + 815, + 76, + 815, + 104, + 300, + 104 + ], + "score": 0.895 + }, + { + "category_id": 2, + "poly": [ + 841, + 2088, + 858, + 2088, + 858, + 2112, + 841, + 2112 + ], + "score": 0.707 + }, + { + "category_id": 2, + "poly": [ + 841, + 2088, + 859, + 2088, + 859, + 2112, + 841, + 2112 + ], + "score": 0.14 + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1638.0, + 1126.0, + 1638.0, + 1126.0, + 1680.0, + 295.0, + 1680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 296.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2085.0, + 862.0, + 2085.0, + 862.0, + 2120.0, + 839.0, + 2120.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2085.0, + 862.0, + 2085.0, + 862.0, + 2120.0, + 838.0, + 2120.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 458.0, + 1403.0, + 458.0, + 1403.0, + 495.0, + 294.0, + 495.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 490.0, + 1404.0, + 490.0, + 1404.0, + 526.0, + 292.0, + 526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 521.0, + 1403.0, + 521.0, + 1403.0, + 556.0, + 293.0, + 556.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 549.0, + 1407.0, + 549.0, + 1407.0, + 588.0, + 293.0, + 588.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 583.0, + 1405.0, + 583.0, + 1405.0, + 617.0, + 294.0, + 617.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 613.0, + 1404.0, + 613.0, + 1404.0, + 646.0, + 294.0, + 646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 640.0, + 1405.0, + 640.0, + 1405.0, + 679.0, + 293.0, + 679.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 674.0, + 1405.0, + 674.0, + 1405.0, + 708.0, + 294.0, + 708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 704.0, + 1404.0, + 704.0, + 1404.0, + 738.0, + 294.0, + 738.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 737.0, + 1402.0, + 737.0, + 1402.0, + 767.0, + 296.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 767.0, + 1404.0, + 767.0, + 1404.0, + 797.0, + 296.0, + 797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 796.0, + 835.0, + 796.0, + 835.0, + 829.0, + 293.0, + 829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 842.0, + 1404.0, + 842.0, + 1404.0, + 876.0, + 296.0, + 876.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 873.0, + 1401.0, + 873.0, + 1401.0, + 903.0, + 296.0, + 903.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 901.0, + 1404.0, + 901.0, + 1404.0, + 937.0, + 293.0, + 937.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 935.0, + 1404.0, + 935.0, + 1404.0, + 965.0, + 296.0, + 965.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 962.0, + 1405.0, + 962.0, + 1405.0, + 999.0, + 293.0, + 999.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 993.0, + 1405.0, + 993.0, + 1405.0, + 1028.0, + 293.0, + 1028.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1022.0, + 1405.0, + 1022.0, + 1405.0, + 1058.0, + 293.0, + 1058.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1055.0, + 1404.0, + 1055.0, + 1404.0, + 1089.0, + 293.0, + 1089.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1084.0, + 1402.0, + 1084.0, + 1402.0, + 1118.0, + 294.0, + 1118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1112.0, + 1404.0, + 1112.0, + 1404.0, + 1152.0, + 292.0, + 1152.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1146.0, + 1405.0, + 1146.0, + 1405.0, + 1180.0, + 294.0, + 1180.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1178.0, + 1343.0, + 1178.0, + 1343.0, + 1212.0, + 294.0, + 1212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1710.0, + 1404.0, + 1710.0, + 1404.0, + 1746.0, + 294.0, + 1746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1743.0, + 1403.0, + 1743.0, + 1403.0, + 1776.0, + 294.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1770.0, + 1404.0, + 1770.0, + 1404.0, + 1809.0, + 292.0, + 1809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1804.0, + 1403.0, + 1804.0, + 1403.0, + 1838.0, + 294.0, + 1838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1836.0, + 1402.0, + 1836.0, + 1402.0, + 1866.0, + 296.0, + 1866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1864.0, + 1404.0, + 1864.0, + 1404.0, + 1898.0, + 293.0, + 1898.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1893.0, + 1404.0, + 1893.0, + 1404.0, + 1930.0, + 293.0, + 1930.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1925.0, + 1167.0, + 1925.0, + 1167.0, + 1959.0, + 294.0, + 1959.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 231.0, + 1404.0, + 231.0, + 1404.0, + 265.0, + 294.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 262.0, + 1405.0, + 262.0, + 1405.0, + 296.0, + 294.0, + 296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 291.0, + 1403.0, + 291.0, + 1403.0, + 326.0, + 294.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 318.0, + 1405.0, + 318.0, + 1405.0, + 358.0, + 293.0, + 358.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 351.0, + 1403.0, + 351.0, + 1403.0, + 386.0, + 293.0, + 386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 383.0, + 1405.0, + 383.0, + 1405.0, + 417.0, + 294.0, + 417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 412.0, + 1117.0, + 412.0, + 1117.0, + 452.0, + 292.0, + 452.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 1281.0, + 1404.0, + 1281.0, + 1404.0, + 1320.0, + 369.0, + 1320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1311.0, + 1323.0, + 1311.0, + 1323.0, + 1352.0, + 394.0, + 1352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 370.0, + 1358.0, + 1404.0, + 1358.0, + 1404.0, + 1397.0, + 370.0, + 1397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1392.0, + 1405.0, + 1392.0, + 1405.0, + 1425.0, + 394.0, + 1425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1424.0, + 1270.0, + 1424.0, + 1270.0, + 1454.0, + 393.0, + 1454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 1466.0, + 1404.0, + 1466.0, + 1404.0, + 1500.0, + 372.0, + 1500.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1499.0, + 1405.0, + 1499.0, + 1405.0, + 1530.0, + 393.0, + 1530.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 1530.0, + 1405.0, + 1530.0, + 1405.0, + 1562.0, + 395.0, + 1562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1559.0, + 1181.0, + 1559.0, + 1181.0, + 1594.0, + 393.0, + 1594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1971.0, + 1404.0, + 1971.0, + 1404.0, + 2007.0, + 296.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 2003.0, + 1404.0, + 2003.0, + 1404.0, + 2036.0, + 293.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1220.0, + 610.0, + 1220.0, + 610.0, + 1259.0, + 297.0, + 1259.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 1, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 230, + 1405, + 230, + 1405, + 566, + 298, + 566 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 299, + 580, + 879, + 580, + 879, + 1129, + 299, + 1129 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 1274, + 1403, + 1274, + 1403, + 1457, + 298, + 1457 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 297, + 1586, + 1404, + 1586, + 1404, + 1834, + 297, + 1834 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 298, + 1130, + 1403, + 1130, + 1403, + 1259, + 298, + 1259 + ], + "score": 0.968 + }, + { + "category_id": 2, + "poly": [ + 296, + 1865, + 1403, + 1865, + 1403, + 2033, + 296, + 2033 + ], + "score": 0.963 + }, + { + "category_id": 4, + "poly": [ + 901, + 838, + 1403, + 838, + 1403, + 1088, + 901, + 1088 + ], + "score": 0.957 + }, + { + "category_id": 1, + "poly": [ + 299, + 1510, + 1405, + 1510, + 1405, + 1573, + 299, + 1573 + ], + "score": 0.946 + }, + { + "category_id": 3, + "poly": [ + 1005, + 614, + 1293, + 614, + 1293, + 819, + 1005, + 819 + ], + "score": 0.944 + }, + { + "category_id": 8, + "poly": [ + 717, + 1460, + 980, + 1460, + 980, + 1498, + 717, + 1498 + ], + "score": 0.934 + }, + { + "category_id": 2, + "poly": [ + 300, + 76, + 815, + 76, + 815, + 104, + 300, + 104 + ], + "score": 0.883 + }, + { + "category_id": 9, + "poly": [ + 1367, + 1465, + 1399, + 1465, + 1399, + 1494, + 1367, + 1494 + ], + "score": 0.863 + }, + { + "category_id": 2, + "poly": [ + 841, + 2088, + 858, + 2088, + 858, + 2112, + 841, + 2112 + ], + "score": 0.663 + }, + { + "category_id": 2, + "poly": [ + 841, + 2088, + 859, + 2088, + 859, + 2112, + 841, + 2112 + ], + "score": 0.295 + }, + { + "category_id": 13, + "poly": [ + 981, + 961, + 1200, + 961, + 1200, + 998, + 981, + 998 + ], + "score": 0.93, + "latex": "C , S , X , Y , E , \\tilde { X } , \\theta" + }, + { + "category_id": 14, + "poly": [ + 716, + 1460, + 981, + 1460, + 981, + 1497, + 716, + 1497 + ], + "score": 0.93, + "latex": "E : = \\mathcal { M } ( X , Y , \\pmb \\theta , U _ { E } ) ," + }, + { + "category_id": 13, + "poly": [ + 853, + 1740, + 954, + 1740, + 954, + 1773, + 853, + 1773 + ], + "score": 0.92, + "latex": "h \\left( X ; \\pmb { \\theta } \\right)" + }, + { + "category_id": 13, + "poly": [ + 1094, + 1160, + 1282, + 1160, + 1282, + 1199, + 1094, + 1199 + ], + "score": 0.92, + "latex": "( X , Y , \\pmb \\theta ) E ^ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 626, + 1895, + 702, + 1895, + 702, + 1924, + 626, + 1924 + ], + "score": 0.91, + "latex": "( X , Y )" + }, + { + "category_id": 13, + "poly": [ + 531, + 887, + 685, + 887, + 685, + 915, + 531, + 915 + ], + "score": 0.91, + "latex": "C \\right. X \\left. S" + }, + { + "category_id": 13, + "poly": [ + 298, + 977, + 397, + 977, + 397, + 1006, + 298, + 1006 + ], + "score": 0.91, + "latex": "C Y" + }, + { + "category_id": 13, + "poly": [ + 385, + 2004, + 451, + 2004, + 451, + 2034, + 385, + 2034 + ], + "score": 0.91, + "latex": "( Y , \\pmb \\theta )" + }, + { + "category_id": 13, + "poly": [ + 432, + 1680, + 542, + 1680, + 542, + 1711, + 432, + 1711 + ], + "score": 0.9, + "latex": "P \\left( X , Y \\right)" + }, + { + "category_id": 13, + "poly": [ + 635, + 1512, + 676, + 1512, + 676, + 1542, + 635, + 1542 + ], + "score": 0.9, + "latex": "U _ { E }" + }, + { + "category_id": 13, + "poly": [ + 473, + 1649, + 602, + 1649, + 602, + 1681, + 473, + 1681 + ], + "score": 0.89, + "latex": "d o ( E = { \\bf 0 } )" + }, + { + "category_id": 13, + "poly": [ + 506, + 1741, + 554, + 1741, + 554, + 1773, + 506, + 1773 + ], + "score": 0.86, + "latex": "\\ell ( \\cdot )" + }, + { + "category_id": 13, + "poly": [ + 585, + 1428, + 609, + 1428, + 609, + 1453, + 585, + 1453 + ], + "score": 0.85, + "latex": "E" + }, + { + "category_id": 13, + "poly": [ + 593, + 975, + 623, + 975, + 623, + 1005, + 593, + 1005 + ], + "score": 0.84, + "latex": "\\tilde { X }" + }, + { + "category_id": 13, + "poly": [ + 1040, + 1951, + 1065, + 1951, + 1065, + 1975, + 1040, + 1975 + ], + "score": 0.83, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 441, + 1040, + 467, + 1040, + 467, + 1066, + 441, + 1066 + ], + "score": 0.82, + "latex": "E" + }, + { + "category_id": 13, + "poly": [ + 507, + 1712, + 543, + 1712, + 543, + 1737, + 507, + 1737 + ], + "score": 0.82, + "latex": "\\mathcal { M }" + }, + { + "category_id": 13, + "poly": [ + 487, + 1009, + 513, + 1009, + 513, + 1036, + 487, + 1036 + ], + "score": 0.81, + "latex": "E" + }, + { + "category_id": 13, + "poly": [ + 773, + 584, + 796, + 584, + 796, + 613, + 773, + 613 + ], + "score": 0.81, + "latex": "\\mathcal { G }" + }, + { + "category_id": 13, + "poly": [ + 713, + 1009, + 741, + 1009, + 741, + 1036, + 713, + 1036 + ], + "score": 0.81, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 616, + 1398, + 641, + 1398, + 641, + 1424, + 616, + 1424 + ], + "score": 0.81, + "latex": "E" + }, + { + "category_id": 13, + "poly": [ + 471, + 1590, + 496, + 1590, + 496, + 1616, + 471, + 1616 + ], + "score": 0.81, + "latex": "E" + }, + { + "category_id": 13, + "poly": [ + 538, + 705, + 565, + 705, + 565, + 731, + 538, + 731 + ], + "score": 0.81, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 957, + 1198, + 979, + 1198, + 979, + 1227, + 957, + 1227 + ], + "score": 0.81, + "latex": "\\mathcal { G }" + }, + { + "category_id": 13, + "poly": [ + 385, + 1650, + 410, + 1650, + 410, + 1678, + 385, + 1678 + ], + "score": 0.8, + "latex": "E" + }, + { + "category_id": 13, + "poly": [ + 298, + 1071, + 323, + 1071, + 323, + 1097, + 298, + 1097 + ], + "score": 0.8, + "latex": "E" + }, + { + "category_id": 13, + "poly": [ + 1000, + 1896, + 1026, + 1896, + 1026, + 1920, + 1000, + 1920 + ], + "score": 0.8, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 822, + 855, + 878, + 855, + 878, + 885, + 822, + 885 + ], + "score": 0.79, + "latex": " { S } \\perp \\perp" + }, + { + "category_id": 13, + "poly": [ + 394, + 857, + 416, + 857, + 416, + 883, + 394, + 883 + ], + "score": 0.78, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 692, + 1071, + 721, + 1071, + 721, + 1098, + 692, + 1098 + ], + "score": 0.78, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 343, + 1131, + 363, + 1131, + 363, + 1157, + 343, + 1157 + ], + "score": 0.77, + "latex": "\\pmb \\theta" + }, + { + "category_id": 13, + "poly": [ + 445, + 796, + 470, + 796, + 470, + 823, + 445, + 823 + ], + "score": 0.77, + "latex": "C" + }, + { + "category_id": 13, + "poly": [ + 298, + 1711, + 323, + 1711, + 323, + 1737, + 298, + 1737 + ], + "score": 0.76, + "latex": "E" + }, + { + "category_id": 13, + "poly": [ + 458, + 1100, + 484, + 1100, + 484, + 1128, + 458, + 1128 + ], + "score": 0.75, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 297, + 887, + 322, + 887, + 322, + 915, + 297, + 915 + ], + "score": 0.74, + "latex": "C" + }, + { + "category_id": 13, + "poly": [ + 452, + 1803, + 473, + 1803, + 473, + 1829, + 452, + 1829 + ], + "score": 0.7, + "latex": "\\pmb { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 888, + 1868, + 907, + 1868, + 907, + 1892, + 888, + 1892 + ], + "score": 0.66, + "latex": "\\pmb \\theta" + }, + { + "category_id": 15, + "poly": [ + 332.0, + 1859.0, + 887.0, + 1859.0, + 887.0, + 1900.0, + 332.0, + 1900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 908.0, + 1859.0, + 1408.0, + 1859.0, + 1408.0, + 1900.0, + 908.0, + 1900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1892.0, + 625.0, + 1892.0, + 625.0, + 1926.0, + 293.0, + 1926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 703.0, + 1892.0, + 999.0, + 1892.0, + 999.0, + 1926.0, + 703.0, + 1926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1027.0, + 1892.0, + 1405.0, + 1892.0, + 1405.0, + 1926.0, + 1027.0, + 1926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1921.0, + 1404.0, + 1921.0, + 1404.0, + 1951.0, + 296.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1949.0, + 1039.0, + 1949.0, + 1039.0, + 1979.0, + 294.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1066.0, + 1949.0, + 1403.0, + 1949.0, + 1403.0, + 1979.0, + 1066.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1978.0, + 1405.0, + 1978.0, + 1405.0, + 2008.0, + 296.0, + 2008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 2003.0, + 384.0, + 2003.0, + 384.0, + 2036.0, + 295.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 452.0, + 2003.0, + 650.0, + 2003.0, + 650.0, + 2036.0, + 452.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 901.0, + 838.0, + 1404.0, + 838.0, + 1404.0, + 873.0, + 901.0, + 873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 901.0, + 869.0, + 1405.0, + 869.0, + 1405.0, + 902.0, + 901.0, + 902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 898.0, + 900.0, + 1406.0, + 900.0, + 1406.0, + 932.0, + 898.0, + 932.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 897.0, + 929.0, + 1406.0, + 929.0, + 1406.0, + 964.0, + 897.0, + 964.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 901.0, + 963.0, + 980.0, + 963.0, + 980.0, + 998.0, + 901.0, + 998.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1201.0, + 963.0, + 1404.0, + 963.0, + 1404.0, + 998.0, + 1201.0, + 998.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 901.0, + 995.0, + 1405.0, + 995.0, + 1405.0, + 1028.0, + 901.0, + 1028.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 901.0, + 1026.0, + 1406.0, + 1026.0, + 1406.0, + 1059.0, + 901.0, + 1059.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 1053.0, + 1226.0, + 1053.0, + 1226.0, + 1092.0, + 899.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1209.0, + 632.0, + 1227.0, + 632.0, + 1227.0, + 656.0, + 1209.0, + 656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1136.0, + 696.0, + 1175.0, + 696.0, + 1175.0, + 737.0, + 1136.0, + 737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1268.0, + 705.0, + 1291.0, + 705.0, + 1291.0, + 731.0, + 1268.0, + 731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1209.0, + 772.0, + 1240.0, + 772.0, + 1240.0, + 814.0, + 1209.0, + 814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1079.0, + 778.0, + 1108.0, + 778.0, + 1108.0, + 809.0, + 1079.0, + 809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 296.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2085.0, + 860.0, + 2085.0, + 860.0, + 2116.0, + 839.0, + 2116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2085.0, + 862.0, + 2085.0, + 862.0, + 2118.0, + 838.0, + 2118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 230.0, + 1403.0, + 230.0, + 1403.0, + 264.0, + 295.0, + 264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 262.0, + 1406.0, + 262.0, + 1406.0, + 296.0, + 295.0, + 296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 291.0, + 1405.0, + 291.0, + 1405.0, + 326.0, + 293.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 321.0, + 1405.0, + 321.0, + 1405.0, + 356.0, + 293.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 352.0, + 1405.0, + 352.0, + 1405.0, + 387.0, + 295.0, + 387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 383.0, + 1406.0, + 383.0, + 1406.0, + 417.0, + 293.0, + 417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 413.0, + 1405.0, + 413.0, + 1405.0, + 449.0, + 292.0, + 449.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 444.0, + 1405.0, + 444.0, + 1405.0, + 479.0, + 295.0, + 479.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 474.0, + 1405.0, + 474.0, + 1405.0, + 508.0, + 293.0, + 508.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 503.0, + 1406.0, + 503.0, + 1406.0, + 542.0, + 293.0, + 542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 537.0, + 1352.0, + 537.0, + 1352.0, + 568.0, + 296.0, + 568.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 582.0, + 772.0, + 582.0, + 772.0, + 616.0, + 295.0, + 616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 797.0, + 582.0, + 880.0, + 582.0, + 880.0, + 616.0, + 797.0, + 616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 612.0, + 881.0, + 612.0, + 881.0, + 647.0, + 295.0, + 647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 642.0, + 880.0, + 642.0, + 880.0, + 676.0, + 296.0, + 676.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 674.0, + 880.0, + 674.0, + 880.0, + 705.0, + 296.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 702.0, + 537.0, + 702.0, + 537.0, + 737.0, + 294.0, + 737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 566.0, + 702.0, + 881.0, + 702.0, + 881.0, + 737.0, + 566.0, + 737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 731.0, + 883.0, + 731.0, + 883.0, + 772.0, + 293.0, + 772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 766.0, + 880.0, + 766.0, + 880.0, + 798.0, + 296.0, + 798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 795.0, + 444.0, + 795.0, + 444.0, + 829.0, + 294.0, + 829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 471.0, + 795.0, + 882.0, + 795.0, + 882.0, + 829.0, + 471.0, + 829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 827.0, + 881.0, + 827.0, + 881.0, + 858.0, + 293.0, + 858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 854.0, + 393.0, + 854.0, + 393.0, + 888.0, + 294.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 417.0, + 854.0, + 821.0, + 854.0, + 821.0, + 888.0, + 417.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 885.0, + 530.0, + 885.0, + 530.0, + 919.0, + 323.0, + 919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 686.0, + 885.0, + 880.0, + 885.0, + 880.0, + 919.0, + 686.0, + 919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 916.0, + 881.0, + 916.0, + 881.0, + 950.0, + 295.0, + 950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 947.0, + 880.0, + 947.0, + 880.0, + 979.0, + 295.0, + 979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 975.0, + 592.0, + 975.0, + 592.0, + 1010.0, + 398.0, + 1010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 624.0, + 975.0, + 881.0, + 975.0, + 881.0, + 1010.0, + 624.0, + 1010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1009.0, + 486.0, + 1009.0, + 486.0, + 1040.0, + 296.0, + 1040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 514.0, + 1009.0, + 712.0, + 1009.0, + 712.0, + 1040.0, + 514.0, + 1040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 742.0, + 1009.0, + 881.0, + 1009.0, + 881.0, + 1040.0, + 742.0, + 1040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1038.0, + 440.0, + 1038.0, + 440.0, + 1073.0, + 292.0, + 1073.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 468.0, + 1038.0, + 882.0, + 1038.0, + 882.0, + 1073.0, + 468.0, + 1073.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1067.0, + 691.0, + 1067.0, + 691.0, + 1104.0, + 324.0, + 1104.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 722.0, + 1067.0, + 882.0, + 1067.0, + 882.0, + 1104.0, + 722.0, + 1104.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1099.0, + 457.0, + 1099.0, + 457.0, + 1134.0, + 295.0, + 1134.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 485.0, + 1099.0, + 881.0, + 1099.0, + 881.0, + 1134.0, + 485.0, + 1134.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1275.0, + 1404.0, + 1275.0, + 1404.0, + 1306.0, + 296.0, + 1306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1305.0, + 1403.0, + 1305.0, + 1403.0, + 1336.0, + 294.0, + 1336.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1336.0, + 1404.0, + 1336.0, + 1404.0, + 1368.0, + 296.0, + 1368.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1364.0, + 1404.0, + 1364.0, + 1404.0, + 1400.0, + 294.0, + 1400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1396.0, + 615.0, + 1396.0, + 615.0, + 1431.0, + 294.0, + 1431.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 642.0, + 1396.0, + 1405.0, + 1396.0, + 1405.0, + 1431.0, + 642.0, + 1431.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1429.0, + 584.0, + 1429.0, + 584.0, + 1459.0, + 293.0, + 1459.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 610.0, + 1429.0, + 620.0, + 1429.0, + 620.0, + 1459.0, + 610.0, + 1459.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1588.0, + 470.0, + 1588.0, + 470.0, + 1622.0, + 295.0, + 1622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 497.0, + 1588.0, + 1404.0, + 1588.0, + 1404.0, + 1622.0, + 497.0, + 1622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1619.0, + 1405.0, + 1619.0, + 1405.0, + 1650.0, + 295.0, + 1650.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1648.0, + 384.0, + 1648.0, + 384.0, + 1682.0, + 294.0, + 1682.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 411.0, + 1648.0, + 472.0, + 1648.0, + 472.0, + 1682.0, + 411.0, + 1682.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 603.0, + 1648.0, + 1405.0, + 1648.0, + 1405.0, + 1682.0, + 603.0, + 1682.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1677.0, + 431.0, + 1677.0, + 431.0, + 1714.0, + 294.0, + 1714.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 543.0, + 1677.0, + 1405.0, + 1677.0, + 1405.0, + 1714.0, + 543.0, + 1714.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1708.0, + 506.0, + 1708.0, + 506.0, + 1741.0, + 324.0, + 1741.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 544.0, + 1708.0, + 1404.0, + 1708.0, + 1404.0, + 1741.0, + 544.0, + 1741.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1740.0, + 505.0, + 1740.0, + 505.0, + 1774.0, + 295.0, + 1774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 555.0, + 1740.0, + 852.0, + 1740.0, + 852.0, + 1774.0, + 555.0, + 1774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 955.0, + 1740.0, + 1404.0, + 1740.0, + 1404.0, + 1774.0, + 955.0, + 1774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1768.0, + 1402.0, + 1768.0, + 1402.0, + 1804.0, + 294.0, + 1804.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1801.0, + 451.0, + 1801.0, + 451.0, + 1834.0, + 295.0, + 1834.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 474.0, + 1801.0, + 1405.0, + 1801.0, + 1405.0, + 1834.0, + 474.0, + 1834.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1129.0, + 342.0, + 1129.0, + 342.0, + 1162.0, + 294.0, + 1162.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 1129.0, + 1404.0, + 1129.0, + 1404.0, + 1162.0, + 364.0, + 1162.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1160.0, + 1093.0, + 1160.0, + 1093.0, + 1203.0, + 291.0, + 1203.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1283.0, + 1160.0, + 1404.0, + 1160.0, + 1404.0, + 1203.0, + 1283.0, + 1203.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1194.0, + 956.0, + 1194.0, + 956.0, + 1232.0, + 293.0, + 1232.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 980.0, + 1194.0, + 1403.0, + 1194.0, + 1403.0, + 1232.0, + 980.0, + 1232.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1228.0, + 666.0, + 1228.0, + 666.0, + 1261.0, + 296.0, + 1261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1508.0, + 634.0, + 1508.0, + 634.0, + 1544.0, + 295.0, + 1544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 677.0, + 1508.0, + 1405.0, + 1508.0, + 1405.0, + 1544.0, + 677.0, + 1544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1540.0, + 669.0, + 1540.0, + 669.0, + 1576.0, + 294.0, + 1576.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 2, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 855, + 1404, + 855, + 1404, + 1261, + 297, + 1261 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 297, + 1499, + 1404, + 1499, + 1404, + 1919, + 297, + 1919 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 298, + 638, + 1403, + 638, + 1403, + 761, + 298, + 761 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 299, + 1275, + 1402, + 1275, + 1402, + 1405, + 299, + 1405 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 297, + 386, + 1404, + 386, + 1404, + 520, + 297, + 520 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 300, + 228, + 1401, + 228, + 1401, + 322, + 300, + 322 + ], + "score": 0.96 + }, + { + "category_id": 2, + "poly": [ + 298, + 1943, + 1403, + 1943, + 1403, + 2035, + 298, + 2035 + ], + "score": 0.956 + }, + { + "category_id": 8, + "poly": [ + 635, + 323, + 1062, + 323, + 1062, + 380, + 635, + 380 + ], + "score": 0.949 + }, + { + "category_id": 8, + "poly": [ + 417, + 1415, + 1280, + 1415, + 1280, + 1487, + 417, + 1487 + ], + "score": 0.942 + }, + { + "category_id": 0, + "poly": [ + 299, + 568, + 481, + 568, + 481, + 604, + 299, + 604 + ], + "score": 0.908 + }, + { + "category_id": 2, + "poly": [ + 299, + 75, + 815, + 75, + 815, + 105, + 299, + 105 + ], + "score": 0.906 + }, + { + "category_id": 9, + "poly": [ + 1366, + 329, + 1399, + 329, + 1399, + 358, + 1366, + 358 + ], + "score": 0.886 + }, + { + "category_id": 0, + "poly": [ + 298, + 800, + 874, + 800, + 874, + 830, + 298, + 830 + ], + "score": 0.882 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1428, + 1400, + 1428, + 1400, + 1458, + 1366, + 1458 + ], + "score": 0.877 + }, + { + "category_id": 2, + "poly": [ + 841, + 2089, + 858, + 2089, + 858, + 2111, + 841, + 2111 + ], + "score": 0.778 + }, + { + "category_id": 13, + "poly": [ + 991, + 887, + 1226, + 887, + 1226, + 921, + 991, + 921 + ], + "score": 0.94, + "latex": "S \\to \\underline { { X } } C \\to Y" + }, + { + "category_id": 13, + "poly": [ + 1158, + 448, + 1275, + 448, + 1275, + 485, + 1158, + 485 + ], + "score": 0.94, + "latex": "P _ { \\theta } ( { \\tilde { X } } , Y )" + }, + { + "category_id": 13, + "poly": [ + 809, + 1531, + 925, + 1531, + 925, + 1568, + 809, + 1568 + ], + "score": 0.94, + "latex": "P _ { \\theta } ( Y | \\tilde { X } )" + }, + { + "category_id": 13, + "poly": [ + 1178, + 1793, + 1320, + 1793, + 1320, + 1829, + 1178, + 1829 + ], + "score": 0.94, + "latex": "P _ { \\theta } ( Y | \\tilde { X } , s )" + }, + { + "category_id": 13, + "poly": [ + 1245, + 1531, + 1353, + 1531, + 1353, + 1568, + 1245, + 1568 + ], + "score": 0.94, + "latex": "\\dot { P } _ { \\theta } ( s | \\tilde { X } )" + }, + { + "category_id": 13, + "poly": [ + 588, + 1826, + 717, + 1826, + 717, + 1859, + 588, + 1859 + ], + "score": 0.93, + "latex": "P ( \\boldsymbol { Y } | \\boldsymbol { X } , \\boldsymbol { s } )" + }, + { + "category_id": 13, + "poly": [ + 650, + 1534, + 756, + 1534, + 756, + 1568, + 650, + 1568 + ], + "score": 0.93, + "latex": "P ( { Y \\vert { X } } )" + }, + { + "category_id": 14, + "poly": [ + 633, + 318, + 1065, + 318, + 1065, + 382, + 633, + 382 + ], + "score": 0.93, + "latex": "E _ { a d v } = \\underset { E ^ { \\prime } \\in \\mathbb { B } } { \\arg \\operatorname* { m a x } } \\ell ( h ( \\boldsymbol { X } + \\boldsymbol { E ^ { \\prime } } ; \\pmb { \\theta } ) , \\boldsymbol { Y } ) ." + }, + { + "category_id": 13, + "poly": [ + 758, + 1368, + 874, + 1368, + 874, + 1405, + 758, + 1405 + ], + "score": 0.93, + "latex": "P _ { \\theta } ( Y | \\tilde { X } )" + }, + { + "category_id": 13, + "poly": [ + 580, + 483, + 698, + 483, + 698, + 519, + 580, + 519 + ], + "score": 0.93, + "latex": "P _ { \\theta } ( { \\tilde { X } } , Y )" + }, + { + "category_id": 13, + "poly": [ + 297, + 1758, + 439, + 1758, + 439, + 1796, + 297, + 1796 + ], + "score": 0.93, + "latex": "P _ { \\theta } ( Y | \\tilde { X } , s )" + }, + { + "category_id": 13, + "poly": [ + 1096, + 486, + 1197, + 486, + 1197, + 519, + 1096, + 519 + ], + "score": 0.92, + "latex": "h \\left( X ; \\theta \\right)" + }, + { + "category_id": 13, + "poly": [ + 298, + 1371, + 403, + 1371, + 403, + 1405, + 298, + 1405 + ], + "score": 0.92, + "latex": "P ( Y | X )" + }, + { + "category_id": 13, + "poly": [ + 558, + 418, + 885, + 418, + 885, + 451, + 558, + 451 + ], + "score": 0.92, + "latex": "E _ { a d v } : = \\mathcal { M } _ { a d v } ( X , Y , \\pmb \\theta , U _ { E } )" + }, + { + "category_id": 13, + "poly": [ + 1122, + 389, + 1182, + 389, + 1182, + 419, + 1122, + 419 + ], + "score": 0.91, + "latex": "E _ { a d v }" + }, + { + "category_id": 13, + "poly": [ + 712, + 262, + 772, + 262, + 772, + 292, + 712, + 292 + ], + "score": 0.91, + "latex": "E _ { a d v }" + }, + { + "category_id": 14, + "poly": [ + 417, + 1414, + 1279, + 1414, + 1279, + 1488, + 417, + 1488 + ], + "score": 0.91, + "latex": "P ( \\boldsymbol { Y } | \\boldsymbol { X } ) = \\sum _ { s \\in \\mathbb { S } } P ( s | \\boldsymbol { X } ) P ( \\boldsymbol { Y } | \\boldsymbol { X } , s ) , P _ { \\theta } ( \\boldsymbol { Y } | \\tilde { \\boldsymbol { X } } ) = \\sum _ { s \\in \\mathbb { S } } P _ { \\theta } ( s | \\tilde { \\boldsymbol { X } } ) P _ { \\theta } ( \\boldsymbol { Y } | \\tilde { \\boldsymbol { X } } , s ) ," + }, + { + "category_id": 13, + "poly": [ + 950, + 232, + 1009, + 232, + 1009, + 262, + 950, + 262 + ], + "score": 0.91, + "latex": "E _ { a d v }" + }, + { + "category_id": 13, + "poly": [ + 784, + 1796, + 855, + 1796, + 855, + 1828, + 784, + 1828 + ], + "score": 0.91, + "latex": "\\mathcal { M } _ { a d v }" + }, + { + "category_id": 13, + "poly": [ + 430, + 419, + 501, + 419, + 501, + 450, + 430, + 450 + ], + "score": 0.91, + "latex": "\\mathcal { M } _ { a d v }" + }, + { + "category_id": 13, + "poly": [ + 1118, + 262, + 1177, + 262, + 1177, + 292, + 1118, + 292 + ], + "score": 0.9, + "latex": "E _ { a d v }" + }, + { + "category_id": 13, + "poly": [ + 632, + 454, + 698, + 454, + 698, + 483, + 632, + 483 + ], + "score": 0.9, + "latex": "M _ { a d v }" + }, + { + "category_id": 13, + "poly": [ + 437, + 291, + 484, + 291, + 484, + 324, + 437, + 324 + ], + "score": 0.88, + "latex": "\\ell ( \\cdot )" + }, + { + "category_id": 13, + "poly": [ + 1220, + 1979, + 1245, + 1979, + 1245, + 2009, + 1220, + 2009 + ], + "score": 0.87, + "latex": "\\ell _ { p }" + }, + { + "category_id": 13, + "poly": [ + 712, + 1946, + 740, + 1946, + 740, + 1975, + 712, + 1975 + ], + "score": 0.85, + "latex": "\\underline { { \\boldsymbol X } }" + }, + { + "category_id": 13, + "poly": [ + 298, + 1565, + 439, + 1565, + 439, + 1602, + 298, + 1602 + ], + "score": 0.85, + "latex": "P _ { \\theta } ( Y | \\tilde { X } , s )" + }, + { + "category_id": 13, + "poly": [ + 460, + 1565, + 568, + 1565, + 568, + 1602, + 460, + 1602 + ], + "score": 0.83, + "latex": "P _ { \\theta } ( s | \\tilde { X } )" + }, + { + "category_id": 13, + "poly": [ + 564, + 889, + 587, + 889, + 587, + 918, + 564, + 918 + ], + "score": 0.83, + "latex": "\\mathcal { G }" + }, + { + "category_id": 13, + "poly": [ + 1036, + 1889, + 1057, + 1889, + 1057, + 1915, + 1036, + 1915 + ], + "score": 0.82, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 1032, + 1946, + 1057, + 1946, + 1057, + 1969, + 1032, + 1969 + ], + "score": 0.82, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 644, + 925, + 672, + 925, + 672, + 953, + 644, + 953 + ], + "score": 0.81, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 926, + 1018, + 948, + 1018, + 948, + 1044, + 926, + 1044 + ], + "score": 0.81, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 403, + 1730, + 425, + 1730, + 425, + 1756, + 403, + 1756 + ], + "score": 0.81, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 374, + 389, + 397, + 389, + 397, + 415, + 374, + 415 + ], + "score": 0.81, + "latex": "\\mathbb { B }" + }, + { + "category_id": 13, + "poly": [ + 393, + 926, + 415, + 926, + 415, + 954, + 393, + 954 + ], + "score": 0.8, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 767, + 1502, + 789, + 1502, + 789, + 1529, + 767, + 1529 + ], + "score": 0.79, + "latex": "\\mathbb { S }" + }, + { + "category_id": 13, + "poly": [ + 1325, + 1730, + 1346, + 1730, + 1346, + 1757, + 1325, + 1757 + ], + "score": 0.79, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 961, + 1889, + 987, + 1889, + 987, + 1915, + 961, + 1915 + ], + "score": 0.78, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 404, + 487, + 424, + 487, + 424, + 514, + 404, + 514 + ], + "score": 0.78, + "latex": "\\pmb \\theta" + }, + { + "category_id": 13, + "poly": [ + 748, + 987, + 769, + 987, + 769, + 1015, + 748, + 1015 + ], + "score": 0.78, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 1375, + 487, + 1393, + 487, + 1393, + 514, + 1375, + 514 + ], + "score": 0.77, + "latex": "\\pmb \\theta" + }, + { + "category_id": 13, + "poly": [ + 1091, + 1169, + 1117, + 1169, + 1117, + 1196, + 1091, + 1196 + ], + "score": 0.76, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 548, + 926, + 574, + 926, + 574, + 953, + 548, + 953 + ], + "score": 0.74, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 1250, + 1729, + 1276, + 1729, + 1276, + 1757, + 1250, + 1757 + ], + "score": 0.73, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 978, + 1508, + 996, + 1508, + 996, + 1529, + 978, + 1529 + ], + "score": 0.71, + "latex": "s" + }, + { + "category_id": 13, + "poly": [ + 847, + 1017, + 875, + 1017, + 875, + 1045, + 847, + 1045 + ], + "score": 0.71, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 1167, + 1170, + 1189, + 1170, + 1189, + 1197, + 1167, + 1197 + ], + "score": 0.71, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 674, + 987, + 700, + 987, + 700, + 1015, + 674, + 1015 + ], + "score": 0.71, + "latex": "Y" + }, + { + "category_id": 15, + "poly": [ + 330.0, + 1937.0, + 711.0, + 1937.0, + 711.0, + 1979.0, + 330.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 741.0, + 1937.0, + 1031.0, + 1937.0, + 1031.0, + 1979.0, + 741.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1058.0, + 1937.0, + 1152.0, + 1937.0, + 1152.0, + 1979.0, + 1058.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 328.0, + 1970.0, + 1219.0, + 1970.0, + 1219.0, + 2013.0, + 328.0, + 2013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1246.0, + 1970.0, + 1405.0, + 1970.0, + 1405.0, + 2013.0, + 1246.0, + 2013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 2002.0, + 946.0, + 2002.0, + 946.0, + 2038.0, + 295.0, + 2038.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 563.0, + 486.0, + 563.0, + 486.0, + 613.0, + 292.0, + 613.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 295.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 799.0, + 879.0, + 799.0, + 879.0, + 833.0, + 294.0, + 833.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2086.0, + 862.0, + 2086.0, + 862.0, + 2118.0, + 838.0, + 2118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 856.0, + 1402.0, + 856.0, + 1402.0, + 890.0, + 295.0, + 890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 887.0, + 563.0, + 887.0, + 563.0, + 922.0, + 292.0, + 922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 588.0, + 887.0, + 990.0, + 887.0, + 990.0, + 922.0, + 588.0, + 922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1227.0, + 887.0, + 1405.0, + 887.0, + 1405.0, + 922.0, + 1227.0, + 922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 923.0, + 392.0, + 923.0, + 392.0, + 961.0, + 295.0, + 961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 416.0, + 923.0, + 547.0, + 923.0, + 547.0, + 961.0, + 416.0, + 961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 575.0, + 923.0, + 643.0, + 923.0, + 643.0, + 961.0, + 575.0, + 961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 673.0, + 923.0, + 1404.0, + 923.0, + 1404.0, + 961.0, + 673.0, + 961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 953.0, + 1405.0, + 953.0, + 1405.0, + 992.0, + 292.0, + 992.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 985.0, + 673.0, + 985.0, + 673.0, + 1022.0, + 294.0, + 1022.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 701.0, + 985.0, + 747.0, + 985.0, + 747.0, + 1022.0, + 701.0, + 1022.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 770.0, + 985.0, + 1404.0, + 985.0, + 1404.0, + 1022.0, + 770.0, + 1022.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1018.0, + 846.0, + 1018.0, + 846.0, + 1048.0, + 296.0, + 1048.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 876.0, + 1018.0, + 925.0, + 1018.0, + 925.0, + 1048.0, + 876.0, + 1048.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 949.0, + 1018.0, + 1405.0, + 1018.0, + 1405.0, + 1048.0, + 949.0, + 1048.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1048.0, + 1405.0, + 1048.0, + 1405.0, + 1082.0, + 295.0, + 1082.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1076.0, + 1402.0, + 1076.0, + 1402.0, + 1113.0, + 294.0, + 1113.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1104.0, + 1405.0, + 1104.0, + 1405.0, + 1147.0, + 292.0, + 1147.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1136.0, + 1406.0, + 1136.0, + 1406.0, + 1176.0, + 292.0, + 1176.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1167.0, + 1090.0, + 1167.0, + 1090.0, + 1202.0, + 295.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1118.0, + 1167.0, + 1166.0, + 1167.0, + 1166.0, + 1202.0, + 1118.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1190.0, + 1167.0, + 1402.0, + 1167.0, + 1402.0, + 1202.0, + 1190.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1199.0, + 1405.0, + 1199.0, + 1405.0, + 1233.0, + 294.0, + 1233.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1230.0, + 1338.0, + 1230.0, + 1338.0, + 1264.0, + 295.0, + 1264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1500.0, + 766.0, + 1500.0, + 766.0, + 1535.0, + 296.0, + 1535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 790.0, + 1500.0, + 977.0, + 1500.0, + 977.0, + 1535.0, + 790.0, + 1535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 997.0, + 1500.0, + 1404.0, + 1500.0, + 1404.0, + 1535.0, + 997.0, + 1535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1534.0, + 649.0, + 1534.0, + 649.0, + 1569.0, + 295.0, + 1569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 1534.0, + 808.0, + 1534.0, + 808.0, + 1569.0, + 757.0, + 1569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 926.0, + 1534.0, + 1244.0, + 1534.0, + 1244.0, + 1569.0, + 926.0, + 1569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1354.0, + 1534.0, + 1404.0, + 1534.0, + 1404.0, + 1569.0, + 1354.0, + 1569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 1566.0, + 459.0, + 1566.0, + 459.0, + 1604.0, + 440.0, + 1604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 569.0, + 1566.0, + 1404.0, + 1566.0, + 1404.0, + 1604.0, + 569.0, + 1604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1599.0, + 1404.0, + 1599.0, + 1404.0, + 1634.0, + 295.0, + 1634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1626.0, + 1405.0, + 1626.0, + 1405.0, + 1667.0, + 294.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1659.0, + 1406.0, + 1659.0, + 1406.0, + 1695.0, + 294.0, + 1695.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1694.0, + 1406.0, + 1694.0, + 1406.0, + 1733.0, + 294.0, + 1733.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1728.0, + 402.0, + 1728.0, + 402.0, + 1763.0, + 295.0, + 1763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 426.0, + 1728.0, + 1249.0, + 1728.0, + 1249.0, + 1763.0, + 426.0, + 1763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1277.0, + 1728.0, + 1324.0, + 1728.0, + 1324.0, + 1763.0, + 1277.0, + 1763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1347.0, + 1728.0, + 1406.0, + 1728.0, + 1406.0, + 1763.0, + 1347.0, + 1763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 1760.0, + 1405.0, + 1760.0, + 1405.0, + 1799.0, + 440.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1795.0, + 783.0, + 1795.0, + 783.0, + 1830.0, + 294.0, + 1830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 856.0, + 1795.0, + 1177.0, + 1795.0, + 1177.0, + 1830.0, + 856.0, + 1830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1321.0, + 1795.0, + 1405.0, + 1795.0, + 1405.0, + 1830.0, + 1321.0, + 1830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1823.0, + 587.0, + 1823.0, + 587.0, + 1863.0, + 292.0, + 1863.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 718.0, + 1823.0, + 1405.0, + 1823.0, + 1405.0, + 1863.0, + 718.0, + 1863.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1859.0, + 1402.0, + 1859.0, + 1402.0, + 1890.0, + 295.0, + 1890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1884.0, + 960.0, + 1884.0, + 960.0, + 1925.0, + 292.0, + 1925.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 988.0, + 1884.0, + 1035.0, + 1884.0, + 1035.0, + 1925.0, + 988.0, + 1925.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1058.0, + 1884.0, + 1405.0, + 1884.0, + 1405.0, + 1925.0, + 1058.0, + 1925.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 639.0, + 1403.0, + 639.0, + 1403.0, + 672.0, + 296.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 669.0, + 1405.0, + 669.0, + 1405.0, + 703.0, + 293.0, + 703.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 702.0, + 1405.0, + 702.0, + 1405.0, + 734.0, + 294.0, + 734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 730.0, + 1331.0, + 730.0, + 1331.0, + 765.0, + 293.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1273.0, + 1406.0, + 1273.0, + 1406.0, + 1312.0, + 293.0, + 1312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1309.0, + 1404.0, + 1309.0, + 1404.0, + 1339.0, + 297.0, + 1339.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1336.0, + 1404.0, + 1336.0, + 1404.0, + 1371.0, + 294.0, + 1371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1370.0, + 297.0, + 1370.0, + 297.0, + 1407.0, + 294.0, + 1407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 1370.0, + 757.0, + 1370.0, + 757.0, + 1407.0, + 404.0, + 1407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 382.0, + 373.0, + 382.0, + 373.0, + 424.0, + 294.0, + 424.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 382.0, + 1121.0, + 382.0, + 1121.0, + 424.0, + 398.0, + 424.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1183.0, + 382.0, + 1406.0, + 382.0, + 1406.0, + 424.0, + 1183.0, + 424.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 416.0, + 429.0, + 416.0, + 429.0, + 453.0, + 292.0, + 453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 502.0, + 416.0, + 557.0, + 416.0, + 557.0, + 453.0, + 502.0, + 453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 886.0, + 416.0, + 1405.0, + 416.0, + 1405.0, + 453.0, + 886.0, + 453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 449.0, + 631.0, + 449.0, + 631.0, + 491.0, + 292.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 699.0, + 449.0, + 1157.0, + 449.0, + 1157.0, + 491.0, + 699.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1276.0, + 449.0, + 1405.0, + 449.0, + 1405.0, + 491.0, + 1276.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 486.0, + 403.0, + 486.0, + 403.0, + 520.0, + 295.0, + 520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 425.0, + 486.0, + 579.0, + 486.0, + 579.0, + 520.0, + 425.0, + 520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 699.0, + 486.0, + 1095.0, + 486.0, + 1095.0, + 520.0, + 699.0, + 520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1198.0, + 486.0, + 1374.0, + 486.0, + 1374.0, + 520.0, + 1198.0, + 520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1394.0, + 486.0, + 1404.0, + 486.0, + 1404.0, + 520.0, + 1394.0, + 520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 229.0, + 949.0, + 229.0, + 949.0, + 267.0, + 295.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1010.0, + 229.0, + 1405.0, + 229.0, + 1405.0, + 267.0, + 1010.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 258.0, + 711.0, + 258.0, + 711.0, + 298.0, + 294.0, + 298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 773.0, + 258.0, + 1117.0, + 258.0, + 1117.0, + 298.0, + 773.0, + 298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1178.0, + 258.0, + 1405.0, + 258.0, + 1405.0, + 298.0, + 1178.0, + 298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 291.0, + 436.0, + 291.0, + 436.0, + 326.0, + 295.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 485.0, + 291.0, + 496.0, + 291.0, + 496.0, + 326.0, + 485.0, + 326.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 3, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 695, + 1405, + 695, + 1405, + 1101, + 297, + 1101 + ], + "score": 0.985 + }, + { + "category_id": 1, + "poly": [ + 297, + 1189, + 1405, + 1189, + 1405, + 1387, + 297, + 1387 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 527, + 1404, + 527, + 1404, + 681, + 298, + 681 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 298, + 1659, + 1404, + 1659, + 1404, + 1847, + 298, + 1847 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 299, + 307, + 1402, + 307, + 1402, + 431, + 299, + 431 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 301, + 1399, + 1402, + 1399, + 1402, + 1498, + 301, + 1498 + ], + "score": 0.972 + }, + { + "category_id": 8, + "poly": [ + 426, + 1512, + 1276, + 1512, + 1276, + 1648, + 426, + 1648 + ], + "score": 0.965 + }, + { + "category_id": 1, + "poly": [ + 294, + 229, + 1400, + 229, + 1400, + 293, + 294, + 293 + ], + "score": 0.953 + }, + { + "category_id": 1, + "poly": [ + 298, + 1970, + 1402, + 1970, + 1402, + 2035, + 298, + 2035 + ], + "score": 0.942 + }, + { + "category_id": 8, + "poly": [ + 459, + 1117, + 1239, + 1117, + 1239, + 1163, + 459, + 1163 + ], + "score": 0.924 + }, + { + "category_id": 1, + "poly": [ + 292, + 1859, + 1358, + 1859, + 1358, + 1898, + 292, + 1898 + ], + "score": 0.918 + }, + { + "category_id": 8, + "poly": [ + 427, + 1908, + 1272, + 1908, + 1272, + 1954, + 427, + 1954 + ], + "score": 0.915 + }, + { + "category_id": 2, + "poly": [ + 300, + 75, + 815, + 75, + 815, + 105, + 300, + 105 + ], + "score": 0.903 + }, + { + "category_id": 0, + "poly": [ + 299, + 471, + 1034, + 471, + 1034, + 501, + 299, + 501 + ], + "score": 0.898 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1612, + 1400, + 1612, + 1400, + 1641, + 1366, + 1641 + ], + "score": 0.882 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1131, + 1400, + 1131, + 1400, + 1160, + 1366, + 1160 + ], + "score": 0.876 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1917, + 1400, + 1917, + 1400, + 1946, + 1366, + 1946 + ], + "score": 0.868 + }, + { + "category_id": 2, + "poly": [ + 841, + 2088, + 859, + 2088, + 859, + 2112, + 841, + 2112 + ], + "score": 0.764 + }, + { + "category_id": 13, + "poly": [ + 1158, + 974, + 1300, + 974, + 1300, + 1010, + 1158, + 1010 + ], + "score": 0.94, + "latex": "P _ { \\theta } ( Y \\vert \\tilde { X } , s )" + }, + { + "category_id": 13, + "poly": [ + 554, + 1349, + 651, + 1349, + 651, + 1386, + 554, + 1386 + ], + "score": 0.94, + "latex": "h ( { \\tilde { X } } ; \\theta )" + }, + { + "category_id": 13, + "poly": [ + 1063, + 879, + 1205, + 879, + 1205, + 916, + 1063, + 916 + ], + "score": 0.94, + "latex": "P _ { \\theta } ( Y | \\tilde { X } , s )" + }, + { + "category_id": 13, + "poly": [ + 756, + 1860, + 886, + 1860, + 886, + 1897, + 756, + 1897 + ], + "score": 0.94, + "latex": "P ( \\boldsymbol { Y } | \\tilde { X } , s )" + }, + { + "category_id": 13, + "poly": [ + 938, + 1860, + 1079, + 1860, + 1079, + 1897, + 938, + 1897 + ], + "score": 0.94, + "latex": "P _ { \\theta } ( Y | \\tilde { X } , s )" + }, + { + "category_id": 13, + "poly": [ + 434, + 977, + 563, + 977, + 563, + 1010, + 434, + 1010 + ], + "score": 0.94, + "latex": "P ( \\boldsymbol { Y } | \\boldsymbol { X } , \\boldsymbol { s } )" + }, + { + "category_id": 13, + "poly": [ + 856, + 1285, + 952, + 1285, + 952, + 1322, + 856, + 1322 + ], + "score": 0.93, + "latex": "h ( { \\tilde { X } } ; \\theta )" + }, + { + "category_id": 13, + "poly": [ + 681, + 1461, + 796, + 1461, + 796, + 1498, + 681, + 1498 + ], + "score": 0.93, + "latex": "P _ { \\theta } ( Y | \\tilde { X } )" + }, + { + "category_id": 13, + "poly": [ + 523, + 1464, + 629, + 1464, + 629, + 1498, + 523, + 1498 + ], + "score": 0.93, + "latex": "P ( { Y \\vert { X } } )" + }, + { + "category_id": 14, + "poly": [ + 424, + 1510, + 1273, + 1510, + 1273, + 1652, + 424, + 1652 + ], + "score": 0.93, + "latex": "\\begin{array} { r l } & { \\quad K L \\left( P _ { \\theta } \\left( Y \\vert \\tilde { X } \\right) , Q _ { \\theta } \\left( Y \\vert \\tilde { X } \\right) \\right) + \\gamma K L \\left( P \\left( Y \\vert X \\right) , Q _ { \\theta } \\left( Y \\vert X \\right) \\right) } \\\\ & { = \\mathbb { E } _ { \\left( \\tilde { X } , Y \\right) \\sim P _ { \\theta } \\left( \\tilde { X } , Y \\right) } C E \\left( h \\left( \\tilde { X } ; \\theta \\right) , Y \\right) + \\gamma \\mathbb { E } _ { \\left( X , Y \\right) \\sim P \\left( X , Y \\right) } C E \\left( h \\left( X ; \\theta \\right) , Y \\right) } \\\\ & { \\approx \\mathbb { E } _ { \\left( X , Y \\right) \\sim P \\left( X , Y \\right) } C E \\left( h \\left( X + E _ { a d v } ; \\theta \\right) , Y \\right) + \\gamma C E \\left( h \\left( X ; \\theta \\right) , Y \\right) , } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 1156, + 1661, + 1277, + 1661, + 1277, + 1694, + 1156, + 1694 + ], + "score": 0.93, + "latex": "Q _ { \\theta } ( Y | X )" + }, + { + "category_id": 13, + "poly": [ + 1263, + 882, + 1393, + 882, + 1393, + 916, + 1263, + 916 + ], + "score": 0.93, + "latex": "P ( \\boldsymbol { Y } | \\boldsymbol { X } , \\boldsymbol { s } )" + }, + { + "category_id": 13, + "poly": [ + 375, + 1969, + 544, + 1969, + 544, + 2008, + 375, + 2008 + ], + "score": 0.92, + "latex": "g \\left( s \\left( X \\right) ; W _ { g } \\right)" + }, + { + "category_id": 13, + "poly": [ + 806, + 1692, + 906, + 1692, + 906, + 1725, + 806, + 1725 + ], + "score": 0.92, + "latex": "h \\left( X ; \\pmb { \\theta } \\right)" + }, + { + "category_id": 13, + "poly": [ + 1251, + 1753, + 1370, + 1753, + 1370, + 1784, + 1251, + 1784 + ], + "score": 0.92, + "latex": "X + E _ { a d v }" + }, + { + "category_id": 13, + "poly": [ + 670, + 2002, + 739, + 2002, + 739, + 2035, + 670, + 2035 + ], + "score": 0.92, + "latex": "s \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 360, + 1190, + 411, + 1190, + 411, + 1223, + 360, + 1223 + ], + "score": 0.9, + "latex": "d ( \\cdot )" + }, + { + "category_id": 13, + "poly": [ + 1184, + 1191, + 1259, + 1191, + 1259, + 1220, + 1184, + 1220 + ], + "score": 0.9, + "latex": "\\lambda > 0" + }, + { + "category_id": 14, + "poly": [ + 456, + 1116, + 1240, + 1116, + 1240, + 1172, + 456, + 1172 + ], + "score": 0.89, + "latex": "\\operatorname* { m i n } _ { \\theta } \\ d \\big ( P \\left( Y | X \\right) , P _ { \\theta } \\left( Y | \\tilde { X } \\right) \\big ) + \\lambda \\mathbb { E } _ { s } d \\big ( P \\left( Y | X , s \\right) , P _ { \\theta } \\left( Y | \\tilde { X } , s \\right) \\big ) ." + }, + { + "category_id": 14, + "poly": [ + 426, + 1909, + 1269, + 1909, + 1269, + 1953, + 426, + 1953 + ], + "score": 0.89, + "latex": "\\mathbb { E } _ { ( X , Y ) \\sim P ( X , Y ) } C E \\left( g \\left( s \\left( X + E _ { a d v } \\right) ; W _ { g } \\right) , Y \\right) + \\beta \\left( g \\left( s \\left( X \\right) ; W _ { g } \\right) , Y \\right) ," + }, + { + "category_id": 13, + "poly": [ + 823, + 1662, + 870, + 1662, + 870, + 1689, + 823, + 1689 + ], + "score": 0.87, + "latex": "C E" + }, + { + "category_id": 13, + "poly": [ + 1081, + 1220, + 1110, + 1220, + 1110, + 1252, + 1081, + 1252 + ], + "score": 0.86, + "latex": "\\tilde { X }" + }, + { + "category_id": 13, + "poly": [ + 373, + 1662, + 420, + 1662, + 420, + 1690, + 373, + 1690 + ], + "score": 0.85, + "latex": "K L" + }, + { + "category_id": 13, + "poly": [ + 1221, + 2004, + 1249, + 2004, + 1249, + 2030, + 1221, + 2030 + ], + "score": 0.85, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 964, + 1696, + 984, + 1696, + 984, + 1725, + 964, + 1725 + ], + "score": 0.84, + "latex": "\\gamma" + }, + { + "category_id": 13, + "poly": [ + 1251, + 1786, + 1278, + 1786, + 1278, + 1811, + 1251, + 1811 + ], + "score": 0.84, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 567, + 1040, + 595, + 1040, + 595, + 1067, + 567, + 1067 + ], + "score": 0.83, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 1381, + 2004, + 1400, + 2004, + 1400, + 2035, + 1381, + 2035 + ], + "score": 0.83, + "latex": "\\beta" + }, + { + "category_id": 13, + "poly": [ + 1002, + 1224, + 1031, + 1224, + 1031, + 1252, + 1002, + 1252 + ], + "score": 0.83, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 491, + 1041, + 517, + 1041, + 517, + 1067, + 491, + 1067 + ], + "score": 0.79, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 473, + 262, + 495, + 262, + 495, + 289, + 473, + 289 + ], + "score": 0.78, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 715, + 884, + 742, + 884, + 742, + 911, + 715, + 911 + ], + "score": 0.78, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 972, + 884, + 995, + 884, + 995, + 912, + 972, + 912 + ], + "score": 0.74, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 1303, + 2008, + 1319, + 2008, + 1319, + 2030, + 1303, + 2030 + ], + "score": 0.72, + "latex": "s" + }, + { + "category_id": 13, + "poly": [ + 396, + 262, + 422, + 262, + 422, + 289, + 396, + 289 + ], + "score": 0.7, + "latex": "Y" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 296.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 470.0, + 1037.0, + 470.0, + 1037.0, + 504.0, + 294.0, + 504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2085.0, + 862.0, + 2085.0, + 862.0, + 2119.0, + 838.0, + 2119.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 696.0, + 1403.0, + 696.0, + 1403.0, + 730.0, + 295.0, + 730.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 729.0, + 1402.0, + 729.0, + 1402.0, + 759.0, + 296.0, + 759.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 756.0, + 1405.0, + 756.0, + 1405.0, + 791.0, + 294.0, + 791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 789.0, + 1405.0, + 789.0, + 1405.0, + 823.0, + 295.0, + 823.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 820.0, + 1403.0, + 820.0, + 1403.0, + 851.0, + 296.0, + 851.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 852.0, + 1402.0, + 852.0, + 1402.0, + 882.0, + 296.0, + 882.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 881.0, + 714.0, + 881.0, + 714.0, + 917.0, + 294.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 743.0, + 881.0, + 971.0, + 881.0, + 971.0, + 917.0, + 743.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 996.0, + 881.0, + 1062.0, + 881.0, + 1062.0, + 917.0, + 996.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1206.0, + 881.0, + 1262.0, + 881.0, + 1262.0, + 917.0, + 1206.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1394.0, + 881.0, + 1402.0, + 881.0, + 1402.0, + 917.0, + 1394.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 914.0, + 1403.0, + 914.0, + 1403.0, + 948.0, + 296.0, + 948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 942.0, + 1405.0, + 942.0, + 1405.0, + 980.0, + 294.0, + 980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 978.0, + 433.0, + 978.0, + 433.0, + 1012.0, + 295.0, + 1012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 564.0, + 978.0, + 1157.0, + 978.0, + 1157.0, + 1012.0, + 564.0, + 1012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1301.0, + 978.0, + 1405.0, + 978.0, + 1405.0, + 1012.0, + 1301.0, + 1012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1007.0, + 1403.0, + 1007.0, + 1403.0, + 1043.0, + 294.0, + 1043.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1039.0, + 490.0, + 1039.0, + 490.0, + 1073.0, + 295.0, + 1073.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 518.0, + 1039.0, + 566.0, + 1039.0, + 566.0, + 1073.0, + 518.0, + 1073.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 596.0, + 1039.0, + 1405.0, + 1039.0, + 1405.0, + 1073.0, + 596.0, + 1073.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1067.0, + 1269.0, + 1067.0, + 1269.0, + 1106.0, + 296.0, + 1106.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1187.0, + 359.0, + 1187.0, + 359.0, + 1225.0, + 294.0, + 1225.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 412.0, + 1187.0, + 1183.0, + 1187.0, + 1183.0, + 1225.0, + 412.0, + 1225.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1260.0, + 1187.0, + 1405.0, + 1187.0, + 1405.0, + 1225.0, + 1260.0, + 1225.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1219.0, + 1001.0, + 1219.0, + 1001.0, + 1260.0, + 294.0, + 1260.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1032.0, + 1219.0, + 1080.0, + 1219.0, + 1080.0, + 1260.0, + 1032.0, + 1260.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1111.0, + 1219.0, + 1407.0, + 1219.0, + 1407.0, + 1260.0, + 1111.0, + 1260.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1252.0, + 1405.0, + 1252.0, + 1405.0, + 1291.0, + 294.0, + 1291.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1288.0, + 855.0, + 1288.0, + 855.0, + 1324.0, + 294.0, + 1324.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 953.0, + 1288.0, + 1406.0, + 1288.0, + 1406.0, + 1324.0, + 953.0, + 1324.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1317.0, + 1405.0, + 1317.0, + 1405.0, + 1355.0, + 295.0, + 1355.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1353.0, + 553.0, + 1353.0, + 553.0, + 1386.0, + 296.0, + 1386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 652.0, + 1353.0, + 1392.0, + 1353.0, + 1392.0, + 1386.0, + 652.0, + 1386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 529.0, + 1404.0, + 529.0, + 1404.0, + 562.0, + 296.0, + 562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 559.0, + 1401.0, + 559.0, + 1401.0, + 592.0, + 294.0, + 592.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 587.0, + 1406.0, + 587.0, + 1406.0, + 623.0, + 293.0, + 623.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 619.0, + 1405.0, + 619.0, + 1405.0, + 656.0, + 294.0, + 656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 649.0, + 691.0, + 649.0, + 691.0, + 685.0, + 293.0, + 685.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1658.0, + 372.0, + 1658.0, + 372.0, + 1698.0, + 295.0, + 1698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 421.0, + 1658.0, + 822.0, + 1658.0, + 822.0, + 1698.0, + 421.0, + 1698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 871.0, + 1658.0, + 1155.0, + 1658.0, + 1155.0, + 1698.0, + 871.0, + 1698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1278.0, + 1658.0, + 1404.0, + 1658.0, + 1404.0, + 1698.0, + 1278.0, + 1698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1689.0, + 805.0, + 1689.0, + 805.0, + 1729.0, + 293.0, + 1729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 907.0, + 1689.0, + 963.0, + 1689.0, + 963.0, + 1729.0, + 907.0, + 1729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 985.0, + 1689.0, + 1406.0, + 1689.0, + 1406.0, + 1729.0, + 985.0, + 1729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1722.0, + 1405.0, + 1722.0, + 1405.0, + 1758.0, + 294.0, + 1758.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1752.0, + 1250.0, + 1752.0, + 1250.0, + 1788.0, + 294.0, + 1788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1371.0, + 1752.0, + 1405.0, + 1752.0, + 1405.0, + 1788.0, + 1371.0, + 1788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1783.0, + 1250.0, + 1783.0, + 1250.0, + 1819.0, + 293.0, + 1819.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1279.0, + 1783.0, + 1405.0, + 1783.0, + 1405.0, + 1819.0, + 1279.0, + 1819.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1811.0, + 784.0, + 1811.0, + 784.0, + 1849.0, + 293.0, + 1849.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 308.0, + 1404.0, + 308.0, + 1404.0, + 341.0, + 295.0, + 341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 337.0, + 1406.0, + 337.0, + 1406.0, + 374.0, + 293.0, + 374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 365.0, + 1406.0, + 365.0, + 1406.0, + 404.0, + 293.0, + 404.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 398.0, + 1204.0, + 398.0, + 1204.0, + 436.0, + 294.0, + 436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1398.0, + 1402.0, + 1398.0, + 1402.0, + 1436.0, + 295.0, + 1436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1430.0, + 1406.0, + 1430.0, + 1406.0, + 1468.0, + 295.0, + 1468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1465.0, + 522.0, + 1465.0, + 522.0, + 1499.0, + 297.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 630.0, + 1465.0, + 680.0, + 1465.0, + 680.0, + 1499.0, + 630.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 797.0, + 1465.0, + 913.0, + 1465.0, + 913.0, + 1499.0, + 797.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 229.0, + 1405.0, + 229.0, + 1405.0, + 265.0, + 296.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 262.0, + 395.0, + 262.0, + 395.0, + 294.0, + 297.0, + 294.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 423.0, + 262.0, + 472.0, + 262.0, + 472.0, + 294.0, + 423.0, + 294.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 496.0, + 262.0, + 1135.0, + 262.0, + 1135.0, + 294.0, + 496.0, + 294.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 1970.0, + 374.0, + 1970.0, + 374.0, + 2006.0, + 298.0, + 2006.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 545.0, + 1970.0, + 1404.0, + 1970.0, + 1404.0, + 2006.0, + 545.0, + 2006.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2001.0, + 669.0, + 2001.0, + 669.0, + 2036.0, + 294.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 740.0, + 2001.0, + 1220.0, + 2001.0, + 1220.0, + 2036.0, + 740.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1250.0, + 2001.0, + 1302.0, + 2001.0, + 1302.0, + 2036.0, + 1250.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1320.0, + 2001.0, + 1380.0, + 2001.0, + 1380.0, + 2036.0, + 1320.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1401.0, + 2001.0, + 1406.0, + 2001.0, + 1406.0, + 2036.0, + 1401.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1853.0, + 755.0, + 1853.0, + 755.0, + 1908.0, + 291.0, + 1908.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 887.0, + 1853.0, + 937.0, + 1853.0, + 937.0, + 1908.0, + 887.0, + 1908.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1080.0, + 1853.0, + 1367.0, + 1853.0, + 1367.0, + 1908.0, + 1080.0, + 1908.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 4, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1574, + 1405, + 1574, + 1405, + 2036, + 297, + 2036 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 296, + 888, + 1404, + 888, + 1404, + 1146, + 296, + 1146 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 689, + 1405, + 689, + 1405, + 875, + 297, + 875 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 296, + 411, + 1405, + 411, + 1405, + 598, + 296, + 598 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 300, + 1467, + 1404, + 1467, + 1404, + 1563, + 300, + 1563 + ], + "score": 0.963 + }, + { + "category_id": 8, + "poly": [ + 443, + 1356, + 1256, + 1356, + 1256, + 1459, + 443, + 1459 + ], + "score": 0.955 + }, + { + "category_id": 8, + "poly": [ + 411, + 301, + 1271, + 301, + 1271, + 405, + 411, + 405 + ], + "score": 0.95 + }, + { + "category_id": 1, + "poly": [ + 294, + 228, + 1403, + 228, + 1403, + 293, + 294, + 293 + ], + "score": 0.947 + }, + { + "category_id": 8, + "poly": [ + 431, + 1247, + 1263, + 1247, + 1263, + 1291, + 431, + 1291 + ], + "score": 0.915 + }, + { + "category_id": 2, + "poly": [ + 299, + 74, + 816, + 74, + 816, + 105, + 299, + 105 + ], + "score": 0.914 + }, + { + "category_id": 0, + "poly": [ + 298, + 634, + 1185, + 634, + 1185, + 665, + 298, + 665 + ], + "score": 0.912 + }, + { + "category_id": 1, + "poly": [ + 298, + 1303, + 1306, + 1303, + 1306, + 1345, + 298, + 1345 + ], + "score": 0.91 + }, + { + "category_id": 8, + "poly": [ + 422, + 1158, + 1274, + 1158, + 1274, + 1202, + 422, + 1202 + ], + "score": 0.895 + }, + { + "category_id": 9, + "poly": [ + 1352, + 1389, + 1400, + 1389, + 1400, + 1419, + 1352, + 1419 + ], + "score": 0.893 + }, + { + "category_id": 1, + "poly": [ + 298, + 1208, + 1386, + 1208, + 1386, + 1241, + 298, + 1241 + ], + "score": 0.887 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1165, + 1400, + 1165, + 1400, + 1194, + 1366, + 1194 + ], + "score": 0.881 + }, + { + "category_id": 9, + "poly": [ + 1366, + 338, + 1400, + 338, + 1400, + 368, + 1366, + 368 + ], + "score": 0.865 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1254, + 1400, + 1254, + 1400, + 1283, + 1366, + 1283 + ], + "score": 0.864 + }, + { + "category_id": 2, + "poly": [ + 840, + 2088, + 859, + 2088, + 859, + 2112, + 840, + 2112 + ], + "score": 0.781 + }, + { + "category_id": 14, + "poly": [ + 441, + 1352, + 1256, + 1352, + 1256, + 1461, + 441, + 1461 + ], + "score": 0.94, + "latex": "P \\left( \\boldsymbol { Y } = i | \\overline { { g } } \\left( \\mu \\left( \\boldsymbol { X } \\right) ; W _ { g } \\right) \\right) = \\frac { e ^ { W _ { g , i } ^ { \\top } \\mu \\left( \\boldsymbol { X } \\right) } } { \\sum _ { j } e ^ { W _ { g , j } ^ { \\top } \\mu \\left( \\boldsymbol { X } \\right) + \\frac { \\sigma ^ { 2 } } { 2 } \\left( W _ { g , j } - W _ { g , i } \\right) ^ { \\top } \\left( W _ { g , j } - W _ { g , i } \\right) } } ." + }, + { + "category_id": 13, + "poly": [ + 298, + 1102, + 592, + 1102, + 592, + 1149, + 298, + 1149 + ], + "score": 0.93, + "latex": "\\hat { s } \\left( X \\right) \\sim \\mathcal { N } \\left( \\mu \\left( X \\right) , \\sigma ^ { 2 } I \\right)" + }, + { + "category_id": 13, + "poly": [ + 783, + 1304, + 956, + 1304, + 956, + 1346, + 783, + 1346 + ], + "score": 0.93, + "latex": "\\overline { { g } } \\left( \\mu \\left( X \\right) ; W _ { g } \\right)" + }, + { + "category_id": 13, + "poly": [ + 857, + 2002, + 1167, + 2002, + 1167, + 2036, + 857, + 2036 + ], + "score": 0.92, + "latex": "\\hat { s } \\left( X \\right) = \\mu \\left( X ; \\bar { W _ { s } } \\right) + \\sigma n" + }, + { + "category_id": 13, + "poly": [ + 595, + 1607, + 664, + 1607, + 664, + 1640, + 595, + 1640 + ], + "score": 0.92, + "latex": "{ \\hat { s } } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 959, + 1012, + 1033, + 1012, + 1033, + 1044, + 959, + 1044 + ], + "score": 0.91, + "latex": "\\mu \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 764, + 1789, + 833, + 1789, + 833, + 1822, + 764, + 1822 + ], + "score": 0.91, + "latex": "{ \\hat { s } } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 1220, + 1698, + 1289, + 1698, + 1289, + 1731, + 1220, + 1731 + ], + "score": 0.91, + "latex": "{ \\hat { c } } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 1331, + 1499, + 1400, + 1499, + 1400, + 1532, + 1331, + 1532 + ], + "score": 0.91, + "latex": "{ \\hat { s } } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 368, + 1499, + 437, + 1499, + 437, + 1532, + 368, + 1532 + ], + "score": 0.91, + "latex": "{ \\hat { s } } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 1196, + 1576, + 1265, + 1576, + 1265, + 1609, + 1196, + 1609 + ], + "score": 0.91, + "latex": "s \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 1150, + 981, + 1219, + 981, + 1219, + 1014, + 1150, + 1014 + ], + "score": 0.91, + "latex": "{ \\hat { s } } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 599, + 1011, + 871, + 1011, + 871, + 1044, + 599, + 1044 + ], + "score": 0.91, + "latex": "\\hat { s } \\left( X \\right) \\cdot \\bar { \\mathcal { N } } \\left( \\bar { \\mu } \\left( X \\right) , \\Sigma \\right)" + }, + { + "category_id": 13, + "poly": [ + 500, + 1820, + 569, + 1820, + 569, + 1853, + 500, + 1853 + ], + "score": 0.91, + "latex": "\\hat { c } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 784, + 1728, + 972, + 1728, + 972, + 1761, + 784, + 1761 + ], + "score": 0.91, + "latex": "\\hat { c } \\left( X \\right) \\perp \\perp \\hat { s } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 419, + 842, + 487, + 842, + 487, + 875, + 419, + 875 + ], + "score": 0.9, + "latex": "s \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 636, + 1789, + 705, + 1789, + 705, + 1821, + 636, + 1821 + ], + "score": 0.9, + "latex": "\\hat { c } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 512, + 506, + 581, + 506, + 581, + 536, + 512, + 536 + ], + "score": 0.9, + "latex": "\\gamma = 0" + }, + { + "category_id": 13, + "poly": [ + 901, + 981, + 970, + 981, + 970, + 1013, + 901, + 1013 + ], + "score": 0.9, + "latex": "s \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 297, + 1728, + 367, + 1728, + 367, + 1761, + 297, + 1761 + ], + "score": 0.9, + "latex": "{ \\hat { s } } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 835, + 1469, + 877, + 1469, + 877, + 1503, + 835, + 1503 + ], + "score": 0.9, + "latex": "W _ { g }" + }, + { + "category_id": 13, + "poly": [ + 345, + 505, + 415, + 505, + 415, + 533, + 345, + 533 + ], + "score": 0.89, + "latex": "\\lambda = 0" + }, + { + "category_id": 13, + "poly": [ + 472, + 1942, + 1002, + 1942, + 1002, + 1975, + 472, + 1975 + ], + "score": 0.89, + "latex": "{ \\hat { c } } \\left( X \\right) ~ = ~ { \\hat { c } } \\left( r \\left( X ; W _ { r } \\right) ; W _ { c } \\right) ~ = ~ W _ { c } r \\left( X ; W _ { r } \\right)" + }, + { + "category_id": 13, + "poly": [ + 1123, + 1943, + 1164, + 1943, + 1164, + 1973, + 1123, + 1973 + ], + "score": 0.89, + "latex": "W _ { c }" + }, + { + "category_id": 13, + "poly": [ + 621, + 1821, + 690, + 1821, + 690, + 1852, + 621, + 1852 + ], + "score": 0.89, + "latex": "{ \\hat { s } } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 1258, + 2003, + 1300, + 2003, + 1300, + 2034, + 1258, + 2034 + ], + "score": 0.88, + "latex": "W _ { s }" + }, + { + "category_id": 14, + "poly": [ + 435, + 1246, + 1260, + 1246, + 1260, + 1291, + 435, + 1291 + ], + "score": 0.88, + "latex": "\\mathbb { E } _ { \\hat { s } ( X ) \\sim \\mathcal { N } \\left( \\mu \\left( X \\right) , \\sigma ^ { 2 } I \\right) } C E \\left( g \\left( \\hat { s } \\left( X \\right) ; W _ { g } \\right) , Y \\right) \\leq C E \\left( \\overline { { g } } \\left( \\mu \\left( X \\right) ; W _ { g } \\right) , Y \\right) ," + }, + { + "category_id": 13, + "poly": [ + 612, + 1302, + 647, + 1302, + 647, + 1337, + 612, + 1337 + ], + "score": 0.88, + "latex": "i ^ { t h }" + }, + { + "category_id": 14, + "poly": [ + 408, + 300, + 1290, + 300, + 1290, + 407, + 408, + 407 + ], + "score": 0.88, + "latex": "\\begin{array} { r l } & { \\displaystyle \\operatorname* { m i n } _ { \\theta , W _ { g } } \\mathbb { E } ( x , Y ) { \\sim } P ( X , Y ) C E \\left( h \\left( X + E _ { a d v } ; \\theta \\right) , Y \\right) + \\gamma C E \\left( h \\left( X ; \\theta \\right) , Y \\right) } \\\\ & { \\quad \\quad \\quad + \\lambda \\left( \\mathbb { E } _ { s } C E \\left( g \\left( s \\left( X + E _ { a d v } \\right) ; W _ { g } \\right) , Y \\right) + \\beta C E \\left( g \\left( s \\left( X \\right) ; W _ { g } \\right) , Y \\right) \\right) . } \\end{array}" + }, + { + "category_id": 14, + "poly": [ + 420, + 1157, + 1275, + 1157, + 1275, + 1202, + 420, + 1202 + ], + "score": 0.86, + "latex": "\\mathbb { E } _ { s } C E \\left( g \\left( s \\left( X \\right) ; W _ { g } \\right) , Y \\right) \\approx \\mathbb { E } _ { \\hat { s } \\left( X \\right) \\sim \\mathcal { N } \\left( \\mu \\left( X \\right) , \\sigma ^ { 2 } I \\right) } C E \\left( g \\left( \\hat { s } \\left( X \\right) ; W _ { g } \\right) , Y \\right) ." + }, + { + "category_id": 13, + "poly": [ + 873, + 449, + 891, + 449, + 891, + 476, + 873, + 476 + ], + "score": 0.84, + "latex": "g" + }, + { + "category_id": 13, + "poly": [ + 938, + 445, + 958, + 445, + 958, + 471, + 938, + 471 + ], + "score": 0.83, + "latex": "h" + }, + { + "category_id": 13, + "poly": [ + 1088, + 1074, + 1112, + 1074, + 1112, + 1100, + 1088, + 1100 + ], + "score": 0.82, + "latex": "\\Sigma" + }, + { + "category_id": 13, + "poly": [ + 1373, + 813, + 1401, + 813, + 1401, + 840, + 1373, + 840 + ], + "score": 0.82, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 627, + 1045, + 651, + 1045, + 651, + 1070, + 627, + 1070 + ], + "score": 0.81, + "latex": "\\Sigma" + }, + { + "category_id": 13, + "poly": [ + 423, + 447, + 442, + 447, + 442, + 476, + 423, + 476 + ], + "score": 0.81, + "latex": "g" + }, + { + "category_id": 13, + "poly": [ + 530, + 1475, + 549, + 1475, + 549, + 1501, + 530, + 1501 + ], + "score": 0.8, + "latex": "g" + }, + { + "category_id": 13, + "poly": [ + 634, + 419, + 653, + 419, + 653, + 444, + 634, + 444 + ], + "score": 0.8, + "latex": "g" + }, + { + "category_id": 13, + "poly": [ + 880, + 414, + 899, + 414, + 899, + 440, + 880, + 440 + ], + "score": 0.8, + "latex": "h" + }, + { + "category_id": 13, + "poly": [ + 1121, + 1669, + 1142, + 1669, + 1142, + 1698, + 1121, + 1698 + ], + "score": 0.79, + "latex": "\\mathcal { G }" + }, + { + "category_id": 13, + "poly": [ + 564, + 445, + 583, + 445, + 583, + 471, + 564, + 471 + ], + "score": 0.77, + "latex": "h" + }, + { + "category_id": 13, + "poly": [ + 378, + 753, + 399, + 753, + 399, + 779, + 378, + 779 + ], + "score": 0.74, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 297, + 752, + 323, + 752, + 323, + 779, + 297, + 779 + ], + "score": 0.73, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 344, + 847, + 360, + 847, + 360, + 870, + 344, + 870 + ], + "score": 0.67, + "latex": "s" + }, + { + "category_id": 13, + "poly": [ + 931, + 789, + 947, + 789, + 947, + 810, + 931, + 810 + ], + "score": 0.64, + "latex": "s" + }, + { + "category_id": 13, + "poly": [ + 857, + 1700, + 879, + 1700, + 879, + 1726, + 857, + 1726 + ], + "score": 0.6, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 788, + 1699, + 813, + 1699, + 813, + 1726, + 788, + 1726 + ], + "score": 0.56, + "latex": "C" + }, + { + "category_id": 13, + "poly": [ + 788, + 1699, + 880, + 1699, + 880, + 1727, + 788, + 1727 + ], + "score": 0.27, + "latex": "C \\perp \\perp S" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 72.0, + 817.0, + 72.0, + 817.0, + 108.0, + 295.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 631.0, + 1190.0, + 631.0, + 1190.0, + 669.0, + 293.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2086.0, + 862.0, + 2086.0, + 862.0, + 2116.0, + 840.0, + 2116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1574.0, + 1195.0, + 1574.0, + 1195.0, + 1613.0, + 294.0, + 1613.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1266.0, + 1574.0, + 1406.0, + 1574.0, + 1406.0, + 1613.0, + 1266.0, + 1613.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1606.0, + 594.0, + 1606.0, + 594.0, + 1641.0, + 294.0, + 1641.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 665.0, + 1606.0, + 1403.0, + 1606.0, + 1403.0, + 1641.0, + 665.0, + 1641.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1636.0, + 1405.0, + 1636.0, + 1405.0, + 1672.0, + 292.0, + 1672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1668.0, + 1120.0, + 1668.0, + 1120.0, + 1703.0, + 294.0, + 1703.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1143.0, + 1668.0, + 1405.0, + 1668.0, + 1405.0, + 1703.0, + 1143.0, + 1703.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1696.0, + 787.0, + 1696.0, + 787.0, + 1734.0, + 294.0, + 1734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 881.0, + 1696.0, + 1219.0, + 1696.0, + 1219.0, + 1734.0, + 881.0, + 1734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1290.0, + 1696.0, + 1405.0, + 1696.0, + 1405.0, + 1734.0, + 1290.0, + 1734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1728.0, + 783.0, + 1728.0, + 783.0, + 1764.0, + 368.0, + 1764.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 973.0, + 1728.0, + 1407.0, + 1728.0, + 1407.0, + 1764.0, + 973.0, + 1764.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1759.0, + 1405.0, + 1759.0, + 1405.0, + 1792.0, + 292.0, + 1792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1787.0, + 635.0, + 1787.0, + 635.0, + 1825.0, + 292.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 706.0, + 1787.0, + 763.0, + 1787.0, + 763.0, + 1825.0, + 706.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 834.0, + 1787.0, + 1406.0, + 1787.0, + 1406.0, + 1825.0, + 834.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1820.0, + 499.0, + 1820.0, + 499.0, + 1855.0, + 294.0, + 1855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 1820.0, + 620.0, + 1820.0, + 620.0, + 1855.0, + 570.0, + 1855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 691.0, + 1820.0, + 1403.0, + 1820.0, + 1403.0, + 1855.0, + 691.0, + 1855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1848.0, + 1406.0, + 1848.0, + 1406.0, + 1888.0, + 292.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1882.0, + 1405.0, + 1882.0, + 1405.0, + 1917.0, + 295.0, + 1917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1910.0, + 1402.0, + 1910.0, + 1402.0, + 1946.0, + 294.0, + 1946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1940.0, + 471.0, + 1940.0, + 471.0, + 1979.0, + 292.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1003.0, + 1940.0, + 1122.0, + 1940.0, + 1122.0, + 1979.0, + 1003.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1165.0, + 1940.0, + 1406.0, + 1940.0, + 1406.0, + 1979.0, + 1165.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1973.0, + 1405.0, + 1973.0, + 1405.0, + 2008.0, + 295.0, + 2008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2001.0, + 856.0, + 2001.0, + 856.0, + 2040.0, + 294.0, + 2040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1168.0, + 2001.0, + 1257.0, + 2001.0, + 1257.0, + 2040.0, + 1168.0, + 2040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1301.0, + 2001.0, + 1405.0, + 2001.0, + 1405.0, + 2040.0, + 1301.0, + 2040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 889.0, + 1404.0, + 889.0, + 1404.0, + 924.0, + 295.0, + 924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 920.0, + 1405.0, + 920.0, + 1405.0, + 955.0, + 295.0, + 955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 947.0, + 1405.0, + 947.0, + 1405.0, + 986.0, + 294.0, + 986.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 981.0, + 900.0, + 981.0, + 900.0, + 1016.0, + 294.0, + 1016.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 971.0, + 981.0, + 1149.0, + 981.0, + 1149.0, + 1016.0, + 971.0, + 1016.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1220.0, + 981.0, + 1406.0, + 981.0, + 1406.0, + 1016.0, + 1220.0, + 1016.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1011.0, + 598.0, + 1011.0, + 598.0, + 1046.0, + 295.0, + 1046.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 872.0, + 1011.0, + 958.0, + 1011.0, + 958.0, + 1046.0, + 872.0, + 1046.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1034.0, + 1011.0, + 1406.0, + 1011.0, + 1406.0, + 1046.0, + 1034.0, + 1046.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1042.0, + 626.0, + 1042.0, + 626.0, + 1077.0, + 294.0, + 1077.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 652.0, + 1042.0, + 1406.0, + 1042.0, + 1406.0, + 1077.0, + 652.0, + 1077.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1068.0, + 1087.0, + 1068.0, + 1087.0, + 1110.0, + 294.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1113.0, + 1068.0, + 1406.0, + 1068.0, + 1406.0, + 1110.0, + 1113.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1105.0, + 297.0, + 1105.0, + 297.0, + 1148.0, + 294.0, + 1148.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 593.0, + 1105.0, + 1123.0, + 1105.0, + 1123.0, + 1148.0, + 593.0, + 1148.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 687.0, + 1405.0, + 687.0, + 1405.0, + 726.0, + 294.0, + 726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 722.0, + 1405.0, + 722.0, + 1405.0, + 754.0, + 296.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 749.0, + 296.0, + 749.0, + 296.0, + 786.0, + 292.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 749.0, + 377.0, + 749.0, + 377.0, + 786.0, + 324.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 400.0, + 749.0, + 1407.0, + 749.0, + 1407.0, + 786.0, + 400.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 782.0, + 930.0, + 782.0, + 930.0, + 818.0, + 294.0, + 818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 948.0, + 782.0, + 1405.0, + 782.0, + 1405.0, + 818.0, + 948.0, + 818.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 811.0, + 1372.0, + 811.0, + 1372.0, + 847.0, + 292.0, + 847.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 841.0, + 343.0, + 841.0, + 343.0, + 878.0, + 294.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 361.0, + 841.0, + 418.0, + 841.0, + 418.0, + 878.0, + 361.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 488.0, + 841.0, + 929.0, + 841.0, + 929.0, + 878.0, + 488.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 411.0, + 633.0, + 411.0, + 633.0, + 447.0, + 295.0, + 447.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 654.0, + 411.0, + 879.0, + 411.0, + 879.0, + 447.0, + 654.0, + 447.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 900.0, + 411.0, + 1405.0, + 411.0, + 1405.0, + 447.0, + 900.0, + 447.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 440.0, + 422.0, + 440.0, + 422.0, + 480.0, + 293.0, + 480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 443.0, + 440.0, + 563.0, + 440.0, + 563.0, + 480.0, + 443.0, + 480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 584.0, + 440.0, + 872.0, + 440.0, + 872.0, + 480.0, + 584.0, + 480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 892.0, + 440.0, + 937.0, + 440.0, + 937.0, + 480.0, + 892.0, + 480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 959.0, + 440.0, + 1403.0, + 440.0, + 1403.0, + 480.0, + 959.0, + 480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 472.0, + 1405.0, + 472.0, + 1405.0, + 508.0, + 294.0, + 508.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 502.0, + 344.0, + 502.0, + 344.0, + 540.0, + 293.0, + 540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 416.0, + 502.0, + 511.0, + 502.0, + 511.0, + 540.0, + 416.0, + 540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 582.0, + 502.0, + 1403.0, + 502.0, + 1403.0, + 540.0, + 582.0, + 540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 535.0, + 1405.0, + 535.0, + 1405.0, + 567.0, + 295.0, + 567.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 562.0, + 1136.0, + 562.0, + 1136.0, + 602.0, + 294.0, + 602.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1466.0, + 529.0, + 1466.0, + 529.0, + 1504.0, + 294.0, + 1504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 550.0, + 1466.0, + 834.0, + 1466.0, + 834.0, + 1504.0, + 550.0, + 1504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 878.0, + 1466.0, + 1405.0, + 1466.0, + 1405.0, + 1504.0, + 878.0, + 1504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1499.0, + 367.0, + 1499.0, + 367.0, + 1533.0, + 296.0, + 1533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 438.0, + 1499.0, + 1330.0, + 1499.0, + 1330.0, + 1533.0, + 438.0, + 1533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1530.0, + 851.0, + 1530.0, + 851.0, + 1564.0, + 295.0, + 1564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 231.0, + 1404.0, + 231.0, + 1404.0, + 263.0, + 294.0, + 263.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 263.0, + 729.0, + 263.0, + 729.0, + 295.0, + 297.0, + 295.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1296.0, + 611.0, + 1296.0, + 611.0, + 1352.0, + 291.0, + 1352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 648.0, + 1296.0, + 782.0, + 1296.0, + 782.0, + 1352.0, + 648.0, + 1352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 957.0, + 1296.0, + 1311.0, + 1296.0, + 1311.0, + 1352.0, + 957.0, + 1352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1204.0, + 1388.0, + 1204.0, + 1388.0, + 1247.0, + 294.0, + 1247.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 5, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 998, + 1404, + 998, + 1404, + 1461, + 297, + 1461 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 297, + 228, + 1403, + 228, + 1403, + 488, + 297, + 488 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 1476, + 1404, + 1476, + 1404, + 1752, + 297, + 1752 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 667, + 1403, + 667, + 1403, + 852, + 297, + 852 + ], + "score": 0.977 + }, + { + "category_id": 8, + "poly": [ + 420, + 549, + 1276, + 549, + 1276, + 655, + 420, + 655 + ], + "score": 0.957 + }, + { + "category_id": 1, + "poly": [ + 303, + 1849, + 1401, + 1849, + 1401, + 1942, + 303, + 1942 + ], + "score": 0.957 + }, + { + "category_id": 1, + "poly": [ + 293, + 502, + 1378, + 502, + 1378, + 535, + 293, + 535 + ], + "score": 0.917 + }, + { + "category_id": 2, + "poly": [ + 323, + 1972, + 1409, + 1972, + 1409, + 2034, + 323, + 2034 + ], + "score": 0.901 + }, + { + "category_id": 0, + "poly": [ + 298, + 939, + 459, + 939, + 459, + 970, + 298, + 970 + ], + "score": 0.898 + }, + { + "category_id": 9, + "poly": [ + 1352, + 588, + 1399, + 588, + 1399, + 619, + 1352, + 619 + ], + "score": 0.897 + }, + { + "category_id": 2, + "poly": [ + 300, + 75, + 815, + 75, + 815, + 104, + 300, + 104 + ], + "score": 0.894 + }, + { + "category_id": 0, + "poly": [ + 300, + 884, + 541, + 884, + 541, + 918, + 300, + 918 + ], + "score": 0.889 + }, + { + "category_id": 0, + "poly": [ + 301, + 1791, + 682, + 1791, + 682, + 1822, + 301, + 1822 + ], + "score": 0.882 + }, + { + "category_id": 2, + "poly": [ + 841, + 2087, + 858, + 2087, + 858, + 2111, + 841, + 2111 + ], + "score": 0.7 + }, + { + "category_id": 2, + "poly": [ + 841, + 2087, + 859, + 2087, + 859, + 2111, + 841, + 2111 + ], + "score": 0.16 + }, + { + "category_id": 2, + "poly": [ + 321, + 1973, + 1412, + 1973, + 1412, + 2034, + 321, + 2034 + ], + "score": 0.151 + }, + { + "category_id": 13, + "poly": [ + 297, + 1244, + 394, + 1244, + 394, + 1279, + 297, + 1279 + ], + "score": 0.93, + "latex": "h ( X ; \\pmb \\theta )" + }, + { + "category_id": 13, + "poly": [ + 298, + 394, + 515, + 394, + 515, + 429, + 298, + 429 + ], + "score": 0.92, + "latex": "\\langle a , b \\rangle _ { M } = \\langle a , M b \\rangle" + }, + { + "category_id": 13, + "poly": [ + 765, + 425, + 885, + 425, + 885, + 457, + 765, + 457 + ], + "score": 0.92, + "latex": "r \\left( X ; W _ { r } \\right)" + }, + { + "category_id": 13, + "poly": [ + 756, + 1060, + 830, + 1060, + 830, + 1090, + 756, + 1090 + ], + "score": 0.91, + "latex": "\\gamma = 0" + }, + { + "category_id": 13, + "poly": [ + 523, + 455, + 597, + 455, + 597, + 488, + 523, + 488 + ], + "score": 0.91, + "latex": "\\mu \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 1211, + 1121, + 1281, + 1121, + 1281, + 1151, + 1211, + 1151 + ], + "score": 0.91, + "latex": "\\gamma = 0" + }, + { + "category_id": 13, + "poly": [ + 528, + 265, + 651, + 265, + 651, + 298, + 528, + 298 + ], + "score": 0.91, + "latex": "\\mu \\left( X ; W _ { s } \\right)" + }, + { + "category_id": 14, + "poly": [ + 420, + 548, + 1282, + 548, + 1282, + 657, + 420, + 657 + ], + "score": 0.9, + "latex": "\\begin{array} { r l } & { \\underset { \\theta , W _ { g } } { \\operatorname* { m i n } } \\mathbb { E } _ { ( X , Y ) \\sim P ( X , Y ) } C E \\left( h \\left( X + E _ { a d v } ; \\pmb { \\theta } \\right) , Y \\right) + \\gamma C E \\left( h \\left( X ; \\pmb { \\theta } \\right) , Y \\right) } \\\\ & { } \\\\ & { \\quad \\quad \\quad + \\lambda \\left( C E \\left( \\overline { { g } } \\left( \\mu \\left( X + E _ { a d v } \\right) ; W _ { g } \\right) , Y \\right) + \\beta C E \\left( \\overline { { g } } \\left( \\mu \\left( X \\right) ; W _ { g } \\right) , Y \\right) \\right) . } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 1025, + 297, + 1145, + 297, + 1145, + 329, + 1025, + 329 + ], + "score": 0.9, + "latex": "r \\left( X ; W _ { r } \\right)" + }, + { + "category_id": 13, + "poly": [ + 1055, + 1339, + 1137, + 1339, + 1137, + 1368, + 1055, + 1368 + ], + "score": 0.9, + "latex": "\\lambda ~ = ~ 0" + }, + { + "category_id": 13, + "poly": [ + 1296, + 328, + 1336, + 328, + 1336, + 359, + 1296, + 359 + ], + "score": 0.89, + "latex": "W _ { s }" + }, + { + "category_id": 13, + "poly": [ + 1332, + 1120, + 1401, + 1120, + 1401, + 1148, + 1332, + 1148 + ], + "score": 0.89, + "latex": "\\lambda > 0" + }, + { + "category_id": 13, + "poly": [ + 345, + 456, + 385, + 456, + 385, + 487, + 345, + 487 + ], + "score": 0.89, + "latex": "W _ { c }" + }, + { + "category_id": 13, + "poly": [ + 1052, + 1507, + 1087, + 1507, + 1087, + 1537, + 1052, + 1537 + ], + "score": 0.89, + "latex": "\\ell _ { \\infty }" + }, + { + "category_id": 13, + "poly": [ + 809, + 364, + 849, + 364, + 849, + 396, + 809, + 396 + ], + "score": 0.88, + "latex": "W _ { c }" + }, + { + "category_id": 13, + "poly": [ + 297, + 297, + 866, + 297, + 866, + 330, + 297, + 330 + ], + "score": 0.84, + "latex": "\\mu \\left( X ; W _ { s } \\right) = \\mu \\left( r \\left( X ; W _ { r } \\right) ; W _ { s } \\right) = W _ { s } r \\left( X ; W _ { r } \\right)" + }, + { + "category_id": 13, + "poly": [ + 949, + 359, + 1210, + 359, + 1210, + 397, + 949, + 397 + ], + "score": 0.83, + "latex": "\\left( W _ { c } \\right) ^ { \\perp } \\perp _ { M } \\mathrm { k e r } \\left( W _ { s } \\right) ^ { \\perp }" + }, + { + "category_id": 13, + "poly": [ + 947, + 327, + 1016, + 327, + 1016, + 359, + 947, + 359 + ], + "score": 0.82, + "latex": "\\hat { c } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 372, + 670, + 399, + 670, + 399, + 697, + 372, + 697 + ], + "score": 0.8, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 941, + 1029, + 963, + 1029, + 963, + 1056, + 941, + 1056 + ], + "score": 0.79, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 790, + 1401, + 812, + 1401, + 812, + 1427, + 790, + 1427 + ], + "score": 0.79, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 629, + 330, + 662, + 330, + 662, + 355, + 629, + 355 + ], + "score": 0.79, + "latex": "M" + }, + { + "category_id": 13, + "poly": [ + 297, + 1089, + 319, + 1089, + 319, + 1117, + 297, + 1117 + ], + "score": 0.77, + "latex": "S" + }, + { + "category_id": 13, + "poly": [ + 697, + 1028, + 723, + 1028, + 723, + 1055, + 697, + 1055 + ], + "score": 0.76, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 528, + 1278, + 555, + 1278, + 555, + 1305, + 528, + 1305 + ], + "score": 0.75, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 763, + 237, + 786, + 237, + 786, + 258, + 763, + 258 + ], + "score": 0.72, + "latex": "\\textbf { \\em n }" + }, + { + "category_id": 13, + "poly": [ + 1052, + 329, + 1122, + 329, + 1122, + 359, + 1052, + 359 + ], + "score": 0.68, + "latex": "{ \\hat { s } } \\left( X \\right)" + }, + { + "category_id": 13, + "poly": [ + 1326, + 1059, + 1352, + 1059, + 1352, + 1086, + 1326, + 1086 + ], + "score": 0.68, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 708, + 1400, + 735, + 1400, + 735, + 1427, + 708, + 1427 + ], + "score": 0.66, + "latex": "Y" + }, + { + "category_id": 13, + "poly": [ + 947, + 328, + 1121, + 328, + 1121, + 359, + 947, + 359 + ], + "score": 0.3, + "latex": "\\hat { c } \\left( X \\right) \\perp \\perp \\hat { s } \\left( X \\right)" + }, + { + "category_id": 15, + "poly": [ + 327.0, + 1965.0, + 1275.0, + 1965.0, + 1275.0, + 2007.0, + 327.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 327.0, + 1997.0, + 1405.0, + 1997.0, + 1405.0, + 2041.0, + 327.0, + 2041.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 934.0, + 463.0, + 934.0, + 463.0, + 975.0, + 292.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 72.0, + 816.0, + 72.0, + 816.0, + 108.0, + 296.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 880.0, + 545.0, + 880.0, + 545.0, + 923.0, + 293.0, + 923.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1790.0, + 686.0, + 1790.0, + 686.0, + 1826.0, + 294.0, + 1826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2086.0, + 860.0, + 2086.0, + 860.0, + 2117.0, + 839.0, + 2117.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2086.0, + 860.0, + 2086.0, + 860.0, + 2118.0, + 839.0, + 2118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 328.0, + 1964.0, + 1273.0, + 1964.0, + 1273.0, + 2007.0, + 328.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 1998.0, + 1405.0, + 1998.0, + 1405.0, + 2039.0, + 331.0, + 2039.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 998.0, + 1404.0, + 998.0, + 1404.0, + 1030.0, + 296.0, + 1030.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1026.0, + 696.0, + 1026.0, + 696.0, + 1062.0, + 291.0, + 1062.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 724.0, + 1026.0, + 940.0, + 1026.0, + 940.0, + 1062.0, + 724.0, + 1062.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 1026.0, + 1407.0, + 1026.0, + 1407.0, + 1062.0, + 964.0, + 1062.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1057.0, + 755.0, + 1057.0, + 755.0, + 1093.0, + 292.0, + 1093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 831.0, + 1057.0, + 1325.0, + 1057.0, + 1325.0, + 1093.0, + 831.0, + 1093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1353.0, + 1057.0, + 1406.0, + 1057.0, + 1406.0, + 1093.0, + 1353.0, + 1093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1086.0, + 1404.0, + 1086.0, + 1404.0, + 1123.0, + 320.0, + 1123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1118.0, + 1210.0, + 1118.0, + 1210.0, + 1153.0, + 295.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1282.0, + 1118.0, + 1331.0, + 1118.0, + 1331.0, + 1153.0, + 1282.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 1118.0, + 1405.0, + 1118.0, + 1405.0, + 1153.0, + 1402.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1150.0, + 1405.0, + 1150.0, + 1405.0, + 1182.0, + 295.0, + 1182.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1179.0, + 1404.0, + 1179.0, + 1404.0, + 1214.0, + 294.0, + 1214.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1208.0, + 1404.0, + 1208.0, + 1404.0, + 1245.0, + 294.0, + 1245.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 1243.0, + 1405.0, + 1243.0, + 1405.0, + 1282.0, + 395.0, + 1282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1276.0, + 527.0, + 1276.0, + 527.0, + 1311.0, + 292.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 556.0, + 1276.0, + 1405.0, + 1276.0, + 1405.0, + 1311.0, + 556.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1306.0, + 1404.0, + 1306.0, + 1404.0, + 1341.0, + 292.0, + 1341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1340.0, + 1054.0, + 1340.0, + 1054.0, + 1372.0, + 296.0, + 1372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1138.0, + 1340.0, + 1405.0, + 1340.0, + 1405.0, + 1372.0, + 1138.0, + 1372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1368.0, + 1406.0, + 1368.0, + 1406.0, + 1403.0, + 292.0, + 1403.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1399.0, + 707.0, + 1399.0, + 707.0, + 1434.0, + 294.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 736.0, + 1399.0, + 789.0, + 1399.0, + 789.0, + 1434.0, + 736.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 813.0, + 1399.0, + 1405.0, + 1399.0, + 1405.0, + 1434.0, + 813.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1428.0, + 691.0, + 1428.0, + 691.0, + 1465.0, + 292.0, + 1465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 229.0, + 762.0, + 229.0, + 762.0, + 265.0, + 292.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 229.0, + 1404.0, + 229.0, + 1404.0, + 265.0, + 787.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 264.0, + 527.0, + 264.0, + 527.0, + 303.0, + 293.0, + 303.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 652.0, + 264.0, + 1404.0, + 264.0, + 1404.0, + 303.0, + 652.0, + 303.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 296.0, + 296.0, + 296.0, + 296.0, + 332.0, + 291.0, + 332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 867.0, + 296.0, + 1024.0, + 296.0, + 1024.0, + 332.0, + 867.0, + 332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1146.0, + 296.0, + 1404.0, + 296.0, + 1404.0, + 332.0, + 1146.0, + 332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 324.0, + 628.0, + 324.0, + 628.0, + 364.0, + 291.0, + 364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 663.0, + 324.0, + 946.0, + 324.0, + 946.0, + 364.0, + 663.0, + 364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1123.0, + 324.0, + 1295.0, + 324.0, + 1295.0, + 364.0, + 1123.0, + 364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1337.0, + 324.0, + 1407.0, + 324.0, + 1407.0, + 364.0, + 1337.0, + 364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 361.0, + 808.0, + 361.0, + 808.0, + 399.0, + 292.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 850.0, + 361.0, + 948.0, + 361.0, + 948.0, + 399.0, + 850.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1211.0, + 361.0, + 1405.0, + 361.0, + 1405.0, + 399.0, + 1211.0, + 399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 516.0, + 394.0, + 1405.0, + 394.0, + 1405.0, + 429.0, + 516.0, + 429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 423.0, + 764.0, + 423.0, + 764.0, + 463.0, + 291.0, + 463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 886.0, + 423.0, + 1407.0, + 423.0, + 1407.0, + 463.0, + 886.0, + 463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 455.0, + 344.0, + 455.0, + 344.0, + 491.0, + 293.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 386.0, + 455.0, + 522.0, + 455.0, + 522.0, + 491.0, + 386.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 598.0, + 455.0, + 918.0, + 455.0, + 918.0, + 491.0, + 598.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1476.0, + 1405.0, + 1476.0, + 1405.0, + 1508.0, + 296.0, + 1508.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1507.0, + 1051.0, + 1507.0, + 1051.0, + 1540.0, + 295.0, + 1540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1088.0, + 1507.0, + 1404.0, + 1507.0, + 1404.0, + 1540.0, + 1088.0, + 1540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1534.0, + 1405.0, + 1534.0, + 1405.0, + 1571.0, + 292.0, + 1571.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1564.0, + 1405.0, + 1564.0, + 1405.0, + 1602.0, + 295.0, + 1602.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1598.0, + 1405.0, + 1598.0, + 1405.0, + 1631.0, + 296.0, + 1631.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1624.0, + 1404.0, + 1624.0, + 1404.0, + 1662.0, + 294.0, + 1662.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1657.0, + 1406.0, + 1657.0, + 1406.0, + 1692.0, + 294.0, + 1692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1686.0, + 1405.0, + 1686.0, + 1405.0, + 1726.0, + 294.0, + 1726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1721.0, + 1125.0, + 1721.0, + 1125.0, + 1754.0, + 295.0, + 1754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 667.0, + 371.0, + 667.0, + 371.0, + 702.0, + 295.0, + 702.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 400.0, + 667.0, + 1405.0, + 667.0, + 1405.0, + 702.0, + 400.0, + 702.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 698.0, + 1403.0, + 698.0, + 1403.0, + 730.0, + 296.0, + 730.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 730.0, + 1404.0, + 730.0, + 1404.0, + 762.0, + 295.0, + 762.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 761.0, + 1404.0, + 761.0, + 1404.0, + 793.0, + 296.0, + 793.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 791.0, + 1405.0, + 791.0, + 1405.0, + 823.0, + 296.0, + 823.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 817.0, + 1015.0, + 817.0, + 1015.0, + 855.0, + 294.0, + 855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 1848.0, + 1405.0, + 1848.0, + 1405.0, + 1882.0, + 298.0, + 1882.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1876.0, + 1406.0, + 1876.0, + 1406.0, + 1916.0, + 295.0, + 1916.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 1910.0, + 1404.0, + 1910.0, + 1404.0, + 1944.0, + 298.0, + 1944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 500.0, + 1381.0, + 500.0, + 1381.0, + 539.0, + 296.0, + 539.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 6, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 298, + 573, + 1392, + 573, + 1392, + 750, + 298, + 750 + ], + "score": 0.981, + "html": "
MethodBest checkpointLast checkpoint
NaturalFGSMPGD-20CW-20NaturalFGSMPGD-20CW-20
Madry83.5656.6951.9251.0084.6554.3746.3846.73
CausalAdv-M80.42±0.3957.98±0.2154.44±0.1852.51±0.2583.72±0.4159.17±0.2451.82±0.1950.93±0.27
TRADES81.3957.2553.6451.3982.9157.9552.8051.27
CausalAdv-T81.22±0.2758.97±0.1754.55±0.1652.95±0.2681.62±0.3058.90±0.1653.64±0.1452.70±0.37
" + }, + { + "category_id": 1, + "poly": [ + 298, + 1117, + 879, + 1117, + 879, + 1696, + 298, + 1696 + ], + "score": 0.98 + }, + { + "category_id": 5, + "poly": [ + 298, + 303, + 1394, + 303, + 1394, + 481, + 298, + 481 + ], + "score": 0.978, + "html": "
MethodBest checkpointLast checkpoint
NaturalFGSMPGD-40CW-40NaturalFGSMPGD-40CW-40
Madry99.4897.8295.7595.9299.4796.5294.3394.45
CausalAdv-M99.53±0.0498.02±0.0796.37±0.1296.47±0.1799.49±0.0896.83±0.1094.67±0.1494.84±0.19
TRADES99.3997.2296.5596.6699.3696.7694.8994.91
CausalAdv-T99.49±0.0697.82±0.0796.72±0.1096.78±0.1599.49±0.0497.32±0.0896.63±0.1396.69±0.21
" + }, + { + "category_id": 4, + "poly": [ + 902, + 1452, + 1402, + 1452, + 1402, + 1760, + 902, + 1760 + ], + "score": 0.969 + }, + { + "category_id": 3, + "poly": [ + 937, + 1122, + 1338, + 1122, + 1338, + 1426, + 937, + 1426 + ], + "score": 0.968 + }, + { + "category_id": 1, + "poly": [ + 299, + 1911, + 1402, + 1911, + 1402, + 2034, + 299, + 2034 + ], + "score": 0.967 + }, + { + "category_id": 1, + "poly": [ + 298, + 1775, + 1403, + 1775, + 1403, + 1896, + 298, + 1896 + ], + "score": 0.965 + }, + { + "category_id": 1, + "poly": [ + 297, + 790, + 1403, + 790, + 1403, + 1038, + 297, + 1038 + ], + "score": 0.957 + }, + { + "category_id": 6, + "poly": [ + 298, + 224, + 1401, + 224, + 1401, + 288, + 298, + 288 + ], + "score": 0.927 + }, + { + "category_id": 1, + "poly": [ + 298, + 1713, + 879, + 1713, + 879, + 1773, + 298, + 1773 + ], + "score": 0.926 + }, + { + "category_id": 2, + "poly": [ + 299, + 75, + 815, + 75, + 815, + 105, + 299, + 105 + ], + "score": 0.9 + }, + { + "category_id": 0, + "poly": [ + 299, + 1059, + 515, + 1059, + 515, + 1090, + 299, + 1090 + ], + "score": 0.894 + }, + { + "category_id": 6, + "poly": [ + 299, + 493, + 1398, + 493, + 1398, + 557, + 299, + 557 + ], + "score": 0.814 + }, + { + "category_id": 2, + "poly": [ + 840, + 2088, + 858, + 2088, + 858, + 2112, + 840, + 2112 + ], + "score": 0.79 + }, + { + "category_id": 13, + "poly": [ + 347, + 1482, + 436, + 1482, + 436, + 1513, + 347, + 1513 + ], + "score": 0.9, + "latex": "4 8 . 0 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 353, + 524, + 471, + 524, + 471, + 557, + 353, + 557 + ], + "score": 0.89, + "latex": "\\epsilon = 8 / 2 5 5" + }, + { + "category_id": 13, + "poly": [ + 471, + 1834, + 560, + 1834, + 560, + 1865, + 471, + 1865 + ], + "score": 0.89, + "latex": "5 1 . 6 8 \\%" + }, + { + "category_id": 13, + "poly": [ + 1047, + 1834, + 1136, + 1834, + 1136, + 1865, + 1047, + 1865 + ], + "score": 0.88, + "latex": "5 1 . 8 2 \\%" + }, + { + "category_id": 13, + "poly": [ + 789, + 1452, + 879, + 1452, + 879, + 1482, + 789, + 1482 + ], + "score": 0.88, + "latex": "5 1 . 8 2 \\%" + }, + { + "category_id": 13, + "poly": [ + 665, + 2002, + 753, + 2002, + 753, + 2033, + 665, + 2033 + ], + "score": 0.88, + "latex": "4 9 . 5 8 \\%" + }, + { + "category_id": 13, + "poly": [ + 1163, + 2002, + 1253, + 2002, + 1253, + 2033, + 1163, + 2033 + ], + "score": 0.87, + "latex": "5 2 . 4 6 \\%" + }, + { + "category_id": 13, + "poly": [ + 1258, + 225, + 1342, + 225, + 1342, + 253, + 1258, + 253 + ], + "score": 0.87, + "latex": "\\epsilon = 0 . 3" + }, + { + "category_id": 13, + "poly": [ + 355, + 1269, + 459, + 1269, + 459, + 1300, + 355, + 1300 + ], + "score": 0.85, + "latex": "( 5 { \\bar { 9 } } . 1 7 \\% )" + }, + { + "category_id": 13, + "poly": [ + 299, + 1360, + 402, + 1360, + 402, + 1392, + 299, + 1392 + ], + "score": 0.85, + "latex": "( 8 3 . 6 5 \\% )" + }, + { + "category_id": 13, + "poly": [ + 623, + 1299, + 724, + 1299, + 724, + 1330, + 623, + 1330 + ], + "score": 0.84, + "latex": "( 5 0 . 9 3 \\% )" + }, + { + "category_id": 13, + "poly": [ + 653, + 1391, + 755, + 1391, + 755, + 1423, + 653, + 1423 + ], + "score": 0.83, + "latex": "( 5 1 . 8 2 \\% )" + }, + { + "category_id": 13, + "poly": [ + 580, + 1360, + 683, + 1360, + 683, + 1391, + 580, + 1391 + ], + "score": 0.82, + "latex": "( 8 3 . 5 7 \\% )" + }, + { + "category_id": 13, + "poly": [ + 938, + 2002, + 1028, + 2002, + 1028, + 2033, + 938, + 2033 + ], + "score": 0.81, + "latex": "5 1 . 5 6 \\%" + }, + { + "category_id": 13, + "poly": [ + 340, + 1300, + 441, + 1300, + 441, + 1330, + 340, + 1330 + ], + "score": 0.78, + "latex": "( 5 1 . 8 2 \\% )" + }, + { + "category_id": 13, + "poly": [ + 662, + 495, + 701, + 495, + 701, + 524, + 662, + 524 + ], + "score": 0.64, + "latex": "( \\% )" + }, + { + "category_id": 13, + "poly": [ + 660, + 226, + 696, + 226, + 696, + 254, + 660, + 254 + ], + "score": 0.62, + "latex": "( \\% )" + }, + { + "category_id": 13, + "poly": [ + 814, + 717, + 916, + 717, + 916, + 744, + 814, + 744 + ], + "score": 0.49, + "latex": "\\mathbf { 5 } 2 . 9 5 { \\scriptstyle \\pm 0 . 2 6 }" + }, + { + "category_id": 13, + "poly": [ + 931, + 717, + 1033, + 717, + 1033, + 744, + 931, + 744 + ], + "score": 0.43, + "latex": "8 1 . 6 2 { \\scriptstyle \\pm 0 . 3 0 }" + }, + { + "category_id": 13, + "poly": [ + 814, + 660, + 916, + 660, + 916, + 687, + 814, + 687 + ], + "score": 0.32, + "latex": "{ \\pm 2 . 5 1 \\pm 0 . 2 5 }" + }, + { + "category_id": 13, + "poly": [ + 698, + 661, + 801, + 661, + 801, + 687, + 698, + 687 + ], + "score": 0.31, + "latex": "{ \\pm 4 . 4 4 } { \\scriptstyle \\pm 0 . 1 8 }" + }, + { + "category_id": 13, + "poly": [ + 931, + 661, + 1032, + 661, + 1032, + 687, + 931, + 687 + ], + "score": 0.3, + "latex": "8 3 . 7 2 { \\scriptstyle \\pm 0 . 4 1 }" + }, + { + "category_id": 13, + "poly": [ + 583, + 391, + 684, + 391, + 684, + 418, + 583, + 418 + ], + "score": 0.27, + "latex": "\\mathbf { 9 8 . 0 2 { \\scriptstyle \\pm 0 . 0 7 } }" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 1451.0, + 1405.0, + 1451.0, + 1405.0, + 1486.0, + 899.0, + 1486.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 1478.0, + 1404.0, + 1478.0, + 1404.0, + 1511.0, + 899.0, + 1511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 900.0, + 1508.0, + 1403.0, + 1508.0, + 1403.0, + 1538.0, + 900.0, + 1538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 1535.0, + 1404.0, + 1535.0, + 1404.0, + 1568.0, + 899.0, + 1568.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 898.0, + 1564.0, + 1405.0, + 1564.0, + 1405.0, + 1593.0, + 898.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 898.0, + 1592.0, + 1405.0, + 1592.0, + 1405.0, + 1622.0, + 898.0, + 1622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 900.0, + 1620.0, + 1405.0, + 1620.0, + 1405.0, + 1650.0, + 900.0, + 1650.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 900.0, + 1646.0, + 1404.0, + 1646.0, + 1404.0, + 1676.0, + 900.0, + 1676.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 900.0, + 1676.0, + 1403.0, + 1676.0, + 1403.0, + 1703.0, + 900.0, + 1703.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 900.0, + 1702.0, + 1404.0, + 1702.0, + 1404.0, + 1734.0, + 900.0, + 1734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 1725.0, + 985.0, + 1725.0, + 985.0, + 1765.0, + 899.0, + 1765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 955.0, + 1147.0, + 982.0, + 1147.0, + 982.0, + 1169.0, + 955.0, + 1169.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1066.0, + 1160.0, + 1073.0, + 1160.0, + 1073.0, + 1168.0, + 1066.0, + 1168.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 956.0, + 1186.0, + 982.0, + 1186.0, + 982.0, + 1208.0, + 956.0, + 1208.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 937.0, + 1208.0, + 962.0, + 1208.0, + 962.0, + 1311.0, + 937.0, + 1311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 955.0, + 1224.0, + 980.0, + 1224.0, + 980.0, + 1247.0, + 955.0, + 1247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 1244.0, + 1014.0, + 1244.0, + 1014.0, + 1252.0, + 1006.0, + 1252.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 957.0, + 1262.0, + 980.0, + 1262.0, + 980.0, + 1282.0, + 957.0, + 1282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1094.0, + 1263.0, + 1108.0, + 1263.0, + 1108.0, + 1275.0, + 1094.0, + 1275.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 957.0, + 1300.0, + 980.0, + 1300.0, + 980.0, + 1320.0, + 957.0, + 1320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1194.0, + 1297.0, + 1277.0, + 1297.0, + 1277.0, + 1320.0, + 1194.0, + 1320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1194.0, + 1318.0, + 1278.0, + 1318.0, + 1278.0, + 1340.0, + 1194.0, + 1340.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 956.0, + 1337.0, + 982.0, + 1337.0, + 982.0, + 1359.0, + 956.0, + 1359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1194.0, + 1336.0, + 1325.0, + 1336.0, + 1325.0, + 1360.0, + 1194.0, + 1360.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1194.0, + 1357.0, + 1325.0, + 1357.0, + 1325.0, + 1381.0, + 1194.0, + 1381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 956.0, + 1375.0, + 987.0, + 1375.0, + 987.0, + 1396.0, + 956.0, + 1396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 992.0, + 1392.0, + 1007.0, + 1392.0, + 1007.0, + 1409.0, + 992.0, + 1409.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1027.0, + 1392.0, + 1053.0, + 1392.0, + 1053.0, + 1410.0, + 1027.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1066.0, + 1392.0, + 1092.0, + 1392.0, + 1092.0, + 1410.0, + 1066.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1107.0, + 1392.0, + 1133.0, + 1392.0, + 1133.0, + 1410.0, + 1107.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1140.0, + 1390.0, + 1256.0, + 1390.0, + 1256.0, + 1411.0, + 1140.0, + 1411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1260.0, + 1390.0, + 1335.0, + 1390.0, + 1335.0, + 1411.0, + 1260.0, + 1411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1130.0, + 1406.0, + 1188.0, + 1406.0, + 1188.0, + 1429.0, + 1130.0, + 1429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1036.75, + 1164.5, + 1071.75, + 1164.5, + 1071.75, + 1181.0, + 1036.75, + 1181.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1072.75, + 1158.0, + 1099.75, + 1158.0, + 1099.75, + 1170.5, + 1072.75, + 1170.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1001.0, + 1251.0, + 1016.0, + 1251.0, + 1016.0, + 1257.5, + 1001.0, + 1257.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 989.0, + 1324.5, + 1028.0, + 1324.5, + 1028.0, + 1334.5, + 989.0, + 1334.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 223.0, + 659.0, + 223.0, + 659.0, + 259.0, + 296.0, + 259.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 697.0, + 223.0, + 1257.0, + 223.0, + 1257.0, + 259.0, + 697.0, + 259.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1343.0, + 223.0, + 1405.0, + 223.0, + 1405.0, + 259.0, + 1343.0, + 259.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 253.0, + 1111.0, + 253.0, + 1111.0, + 289.0, + 296.0, + 289.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 295.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1053.0, + 520.0, + 1053.0, + 520.0, + 1097.0, + 293.0, + 1097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 492.0, + 661.0, + 492.0, + 661.0, + 528.0, + 296.0, + 528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 702.0, + 492.0, + 1404.0, + 492.0, + 1404.0, + 528.0, + 702.0, + 528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 524.0, + 352.0, + 524.0, + 352.0, + 560.0, + 296.0, + 560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 472.0, + 524.0, + 1347.0, + 524.0, + 1347.0, + 560.0, + 472.0, + 560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 836.0, + 2085.0, + 859.0, + 2085.0, + 859.0, + 2116.0, + 836.0, + 2116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1116.0, + 880.0, + 1116.0, + 880.0, + 1150.0, + 297.0, + 1150.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1146.0, + 880.0, + 1146.0, + 880.0, + 1182.0, + 293.0, + 1182.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1177.0, + 880.0, + 1177.0, + 880.0, + 1211.0, + 295.0, + 1211.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1206.0, + 881.0, + 1206.0, + 881.0, + 1243.0, + 294.0, + 1243.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1239.0, + 882.0, + 1239.0, + 882.0, + 1272.0, + 293.0, + 1272.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1268.0, + 354.0, + 1268.0, + 354.0, + 1302.0, + 294.0, + 1302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 460.0, + 1268.0, + 882.0, + 1268.0, + 882.0, + 1302.0, + 460.0, + 1302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1296.0, + 339.0, + 1296.0, + 339.0, + 1333.0, + 294.0, + 1333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 442.0, + 1296.0, + 622.0, + 1296.0, + 622.0, + 1333.0, + 442.0, + 1333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 725.0, + 1296.0, + 882.0, + 1296.0, + 882.0, + 1333.0, + 725.0, + 1333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1330.0, + 882.0, + 1330.0, + 882.0, + 1361.0, + 295.0, + 1361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1358.0, + 298.0, + 1358.0, + 298.0, + 1393.0, + 295.0, + 1393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 403.0, + 1358.0, + 579.0, + 1358.0, + 579.0, + 1393.0, + 403.0, + 1393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 684.0, + 1358.0, + 883.0, + 1358.0, + 883.0, + 1393.0, + 684.0, + 1393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1387.0, + 652.0, + 1387.0, + 652.0, + 1427.0, + 293.0, + 1427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 1387.0, + 882.0, + 1387.0, + 882.0, + 1427.0, + 756.0, + 1427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1423.0, + 881.0, + 1423.0, + 881.0, + 1454.0, + 295.0, + 1454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1451.0, + 788.0, + 1451.0, + 788.0, + 1484.0, + 294.0, + 1484.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1482.0, + 346.0, + 1482.0, + 346.0, + 1517.0, + 295.0, + 1517.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 437.0, + 1482.0, + 882.0, + 1482.0, + 882.0, + 1517.0, + 437.0, + 1517.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1513.0, + 882.0, + 1513.0, + 882.0, + 1547.0, + 296.0, + 1547.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1540.0, + 882.0, + 1540.0, + 882.0, + 1578.0, + 293.0, + 1578.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1569.0, + 882.0, + 1569.0, + 882.0, + 1610.0, + 294.0, + 1610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1604.0, + 880.0, + 1604.0, + 880.0, + 1638.0, + 294.0, + 1638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1634.0, + 882.0, + 1634.0, + 882.0, + 1668.0, + 295.0, + 1668.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1665.0, + 765.0, + 1665.0, + 765.0, + 1699.0, + 295.0, + 1699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1909.0, + 1406.0, + 1909.0, + 1406.0, + 1946.0, + 294.0, + 1946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1942.0, + 1404.0, + 1942.0, + 1404.0, + 1975.0, + 295.0, + 1975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1973.0, + 1406.0, + 1973.0, + 1406.0, + 2005.0, + 294.0, + 2005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2003.0, + 664.0, + 2003.0, + 664.0, + 2035.0, + 294.0, + 2035.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 754.0, + 2003.0, + 937.0, + 2003.0, + 937.0, + 2035.0, + 754.0, + 2035.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1029.0, + 2003.0, + 1162.0, + 2003.0, + 1162.0, + 2035.0, + 1029.0, + 2035.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1254.0, + 2003.0, + 1406.0, + 2003.0, + 1406.0, + 2035.0, + 1254.0, + 2035.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1773.0, + 1404.0, + 1773.0, + 1404.0, + 1806.0, + 294.0, + 1806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1802.0, + 1403.0, + 1802.0, + 1403.0, + 1837.0, + 293.0, + 1837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1833.0, + 470.0, + 1833.0, + 470.0, + 1869.0, + 293.0, + 1869.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 561.0, + 1833.0, + 1046.0, + 1833.0, + 1046.0, + 1869.0, + 561.0, + 1869.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1137.0, + 1833.0, + 1404.0, + 1833.0, + 1404.0, + 1869.0, + 1137.0, + 1869.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1865.0, + 719.0, + 1865.0, + 719.0, + 1900.0, + 293.0, + 1900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 790.0, + 1405.0, + 790.0, + 1405.0, + 826.0, + 294.0, + 826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 822.0, + 1404.0, + 822.0, + 1404.0, + 856.0, + 295.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 854.0, + 1401.0, + 854.0, + 1401.0, + 884.0, + 296.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 883.0, + 1406.0, + 883.0, + 1406.0, + 918.0, + 291.0, + 918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 914.0, + 1407.0, + 914.0, + 1407.0, + 948.0, + 295.0, + 948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 944.0, + 1405.0, + 944.0, + 1405.0, + 977.0, + 293.0, + 977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 973.0, + 1403.0, + 973.0, + 1403.0, + 1011.0, + 294.0, + 1011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1007.0, + 1145.0, + 1007.0, + 1145.0, + 1037.0, + 296.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1710.0, + 882.0, + 1710.0, + 882.0, + 1746.0, + 297.0, + 1746.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1742.0, + 881.0, + 1742.0, + 881.0, + 1776.0, + 296.0, + 1776.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 7, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 297, + 303, + 1393, + 303, + 1393, + 481, + 297, + 481 + ], + "score": 0.982, + "html": "
MethodBest checkpointLast checkpoint
NaturalFGSMPGD-20CW-20NaturalFGSMPGD-20CW-20
Madry55.9828.3925.1524.0455.0825.3521.6321.42
CausalAdv-M54.07±0.1729.76±0.1627.62±0.1325.44±0.1554.81±0.2326.83±0.1923.34±0.1522.93±0.13
TRADES53.8529.0427.9124.0953.5429.2926.8023.79
CausalAdv-T53.17±0.3930.66±0.2028.57±0.1825.74±0.1854.79±0.4130.81±0.4028.51±0.3525.32±0.27
" + }, + { + "category_id": 1, + "poly": [ + 298, + 1697, + 1404, + 1697, + 1404, + 2034, + 298, + 2034 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 298, + 955, + 1404, + 955, + 1404, + 1202, + 298, + 1202 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 628, + 1404, + 628, + 1404, + 843, + 298, + 843 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 298, + 1416, + 1403, + 1416, + 1403, + 1599, + 298, + 1599 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 1217, + 1403, + 1217, + 1403, + 1401, + 298, + 1401 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 519, + 1402, + 519, + 1402, + 612, + 298, + 612 + ], + "score": 0.97 + }, + { + "category_id": 6, + "poly": [ + 297, + 223, + 1402, + 223, + 1402, + 287, + 297, + 287 + ], + "score": 0.927 + }, + { + "category_id": 0, + "poly": [ + 301, + 887, + 581, + 887, + 581, + 922, + 301, + 922 + ], + "score": 0.899 + }, + { + "category_id": 2, + "poly": [ + 299, + 76, + 815, + 76, + 815, + 104, + 299, + 104 + ], + "score": 0.896 + }, + { + "category_id": 0, + "poly": [ + 299, + 1644, + 544, + 1644, + 544, + 1680, + 299, + 1680 + ], + "score": 0.887 + }, + { + "category_id": 2, + "poly": [ + 840, + 2088, + 858, + 2088, + 858, + 2111, + 840, + 2111 + ], + "score": 0.792 + }, + { + "category_id": 13, + "poly": [ + 353, + 254, + 471, + 254, + 471, + 287, + 353, + 287 + ], + "score": 0.9, + "latex": "\\epsilon = 8 / 2 5 5" + }, + { + "category_id": 13, + "poly": [ + 669, + 811, + 759, + 811, + 759, + 841, + 669, + 841 + ], + "score": 0.9, + "latex": "5 4 . 4 4 \\%" + }, + { + "category_id": 13, + "poly": [ + 544, + 780, + 633, + 780, + 633, + 810, + 544, + 810 + ], + "score": 0.89, + "latex": "5 1 . 9 2 \\%" + }, + { + "category_id": 13, + "poly": [ + 1100, + 750, + 1154, + 750, + 1154, + 780, + 1100, + 780 + ], + "score": 0.89, + "latex": "4 4 \\%" + }, + { + "category_id": 13, + "poly": [ + 1030, + 811, + 1120, + 811, + 1120, + 842, + 1030, + 842 + ], + "score": 0.88, + "latex": "5 0 . 9 3 \\%" + }, + { + "category_id": 13, + "poly": [ + 773, + 519, + 863, + 519, + 863, + 551, + 773, + 551 + ], + "score": 0.87, + "latex": "5 3 . 7 4 \\%" + }, + { + "category_id": 13, + "poly": [ + 331, + 519, + 421, + 519, + 421, + 551, + 331, + 551 + ], + "score": 0.87, + "latex": "5 4 . 0 9 \\%" + }, + { + "category_id": 13, + "poly": [ + 658, + 226, + 699, + 226, + 699, + 255, + 658, + 255 + ], + "score": 0.67, + "latex": "( \\% )" + }, + { + "category_id": 13, + "poly": [ + 698, + 391, + 800, + 391, + 800, + 418, + 698, + 418 + ], + "score": 0.48, + "latex": "2 7 . 6 2 \\substack { \\pm 0 . 1 3 }" + }, + { + "category_id": 13, + "poly": [ + 814, + 447, + 916, + 447, + 916, + 474, + 814, + 474 + ], + "score": 0.46, + "latex": "2 5 . 7 4 \\substack { \\pm 0 . 1 8 }" + }, + { + "category_id": 13, + "poly": [ + 467, + 447, + 569, + 447, + 569, + 474, + 467, + 474 + ], + "score": 0.44, + "latex": "5 3 . 1 7 { \\scriptstyle \\pm 0 . 3 9 }" + }, + { + "category_id": 13, + "poly": [ + 814, + 391, + 915, + 391, + 915, + 417, + 814, + 417 + ], + "score": 0.34, + "latex": "2 5 . 4 4 { \\scriptstyle \\pm 0 . 1 5 }" + }, + { + "category_id": 13, + "poly": [ + 583, + 391, + 684, + 391, + 684, + 418, + 583, + 418 + ], + "score": 0.32, + "latex": "\\mathbf { 2 9 . 7 6 { \\scriptstyle \\pm 0 . 1 6 } }" + }, + { + "category_id": 13, + "poly": [ + 467, + 392, + 569, + 392, + 569, + 417, + 467, + 417 + ], + "score": 0.31, + "latex": "5 4 . 0 7 { \\scriptstyle \\pm 0 . 1 7 }" + }, + { + "category_id": 13, + "poly": [ + 931, + 447, + 1032, + 447, + 1032, + 474, + 931, + 474 + ], + "score": 0.28, + "latex": "\\mathbf { 5 4 . 7 9 } { \\scriptstyle \\pm 0 . 4 1 }" + }, + { + "category_id": 13, + "poly": [ + 698, + 447, + 801, + 447, + 801, + 475, + 698, + 475 + ], + "score": 0.25, + "latex": "2 8 . 5 7 { \\scriptstyle \\pm 0 . 1 8 }" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 223.0, + 657.0, + 223.0, + 657.0, + 259.0, + 296.0, + 259.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 700.0, + 223.0, + 1404.0, + 223.0, + 1404.0, + 259.0, + 700.0, + 259.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 254.0, + 352.0, + 254.0, + 352.0, + 290.0, + 297.0, + 290.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 472.0, + 254.0, + 1347.0, + 254.0, + 1347.0, + 290.0, + 472.0, + 290.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 884.0, + 586.0, + 884.0, + 586.0, + 928.0, + 293.0, + 928.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 72.0, + 818.0, + 72.0, + 818.0, + 110.0, + 294.0, + 110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1637.0, + 550.0, + 1637.0, + 550.0, + 1689.0, + 292.0, + 1689.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2087.0, + 860.0, + 2087.0, + 860.0, + 2117.0, + 839.0, + 2117.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1699.0, + 1404.0, + 1699.0, + 1404.0, + 1734.0, + 294.0, + 1734.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1728.0, + 1405.0, + 1728.0, + 1405.0, + 1765.0, + 295.0, + 1765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1757.0, + 1405.0, + 1757.0, + 1405.0, + 1795.0, + 293.0, + 1795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1788.0, + 1403.0, + 1788.0, + 1403.0, + 1826.0, + 293.0, + 1826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1821.0, + 1405.0, + 1821.0, + 1405.0, + 1856.0, + 294.0, + 1856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1852.0, + 1405.0, + 1852.0, + 1405.0, + 1887.0, + 294.0, + 1887.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1880.0, + 1404.0, + 1880.0, + 1404.0, + 1916.0, + 293.0, + 1916.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1911.0, + 1404.0, + 1911.0, + 1404.0, + 1946.0, + 294.0, + 1946.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1942.0, + 1404.0, + 1942.0, + 1404.0, + 1977.0, + 294.0, + 1977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1972.0, + 1404.0, + 1972.0, + 1404.0, + 2007.0, + 294.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 2002.0, + 1215.0, + 2002.0, + 1215.0, + 2039.0, + 293.0, + 2039.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 954.0, + 1404.0, + 954.0, + 1404.0, + 988.0, + 296.0, + 988.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 988.0, + 1404.0, + 988.0, + 1404.0, + 1018.0, + 296.0, + 1018.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1017.0, + 1405.0, + 1017.0, + 1405.0, + 1051.0, + 294.0, + 1051.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1047.0, + 1404.0, + 1047.0, + 1404.0, + 1080.0, + 293.0, + 1080.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1077.0, + 1404.0, + 1077.0, + 1404.0, + 1111.0, + 296.0, + 1111.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1107.0, + 1405.0, + 1107.0, + 1405.0, + 1141.0, + 294.0, + 1141.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1138.0, + 1403.0, + 1138.0, + 1403.0, + 1172.0, + 293.0, + 1172.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1169.0, + 729.0, + 1169.0, + 729.0, + 1204.0, + 293.0, + 1204.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 623.0, + 1405.0, + 623.0, + 1405.0, + 665.0, + 292.0, + 665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 662.0, + 1402.0, + 662.0, + 1402.0, + 692.0, + 297.0, + 692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 688.0, + 1405.0, + 688.0, + 1405.0, + 726.0, + 293.0, + 726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 720.0, + 1402.0, + 720.0, + 1402.0, + 754.0, + 294.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 751.0, + 1099.0, + 751.0, + 1099.0, + 782.0, + 296.0, + 782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1155.0, + 751.0, + 1404.0, + 751.0, + 1404.0, + 782.0, + 1155.0, + 782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 782.0, + 543.0, + 782.0, + 543.0, + 813.0, + 296.0, + 813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 634.0, + 782.0, + 1405.0, + 782.0, + 1405.0, + 813.0, + 634.0, + 813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 812.0, + 668.0, + 812.0, + 668.0, + 846.0, + 294.0, + 846.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 760.0, + 812.0, + 1029.0, + 812.0, + 1029.0, + 846.0, + 760.0, + 846.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1121.0, + 812.0, + 1372.0, + 812.0, + 1372.0, + 846.0, + 1121.0, + 846.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1413.0, + 1405.0, + 1413.0, + 1405.0, + 1451.0, + 296.0, + 1451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1444.0, + 1405.0, + 1444.0, + 1405.0, + 1480.0, + 294.0, + 1480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1476.0, + 1407.0, + 1476.0, + 1407.0, + 1511.0, + 293.0, + 1511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1505.0, + 1405.0, + 1505.0, + 1405.0, + 1543.0, + 293.0, + 1543.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1539.0, + 1404.0, + 1539.0, + 1404.0, + 1571.0, + 296.0, + 1571.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1567.0, + 1217.0, + 1567.0, + 1217.0, + 1602.0, + 296.0, + 1602.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1218.0, + 1403.0, + 1218.0, + 1403.0, + 1249.0, + 296.0, + 1249.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1246.0, + 1404.0, + 1246.0, + 1404.0, + 1280.0, + 294.0, + 1280.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1276.0, + 1404.0, + 1276.0, + 1404.0, + 1310.0, + 293.0, + 1310.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1309.0, + 1404.0, + 1309.0, + 1404.0, + 1341.0, + 296.0, + 1341.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1336.0, + 1407.0, + 1336.0, + 1407.0, + 1372.0, + 294.0, + 1372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1370.0, + 1271.0, + 1370.0, + 1271.0, + 1402.0, + 296.0, + 1402.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 515.0, + 330.0, + 515.0, + 330.0, + 557.0, + 292.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 422.0, + 515.0, + 772.0, + 515.0, + 772.0, + 557.0, + 422.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 864.0, + 515.0, + 1406.0, + 515.0, + 1406.0, + 557.0, + 864.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 552.0, + 1404.0, + 552.0, + 1404.0, + 584.0, + 293.0, + 584.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 582.0, + 953.0, + 582.0, + 953.0, + 616.0, + 296.0, + 616.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 8, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 295, + 1405, + 295, + 1405, + 509, + 298, + 509 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 299, + 624, + 1403, + 624, + 1403, + 747, + 299, + 747 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 290, + 863, + 1400, + 863, + 1400, + 924, + 290, + 924 + ], + "score": 0.931 + }, + { + "category_id": 0, + "poly": [ + 302, + 225, + 704, + 225, + 704, + 261, + 302, + 261 + ], + "score": 0.906 + }, + { + "category_id": 0, + "poly": [ + 301, + 792, + 796, + 792, + 796, + 828, + 301, + 828 + ], + "score": 0.902 + }, + { + "category_id": 0, + "poly": [ + 301, + 555, + 639, + 555, + 639, + 591, + 301, + 591 + ], + "score": 0.878 + }, + { + "category_id": 2, + "poly": [ + 299, + 75, + 816, + 75, + 816, + 105, + 299, + 105 + ], + "score": 0.873 + }, + { + "category_id": 2, + "poly": [ + 836, + 2088, + 865, + 2088, + 865, + 2112, + 836, + 2112 + ], + "score": 0.833 + }, + { + "category_id": 15, + "poly": [ + 293.0, + 225.0, + 708.0, + 225.0, + 708.0, + 272.0, + 293.0, + 272.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 789.0, + 800.0, + 789.0, + 800.0, + 832.0, + 291.0, + 832.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 552.0, + 642.0, + 552.0, + 642.0, + 595.0, + 293.0, + 595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 295.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2084.0, + 871.0, + 2084.0, + 871.0, + 2123.0, + 832.0, + 2123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 295.0, + 1406.0, + 295.0, + 1406.0, + 333.0, + 295.0, + 333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 323.0, + 1405.0, + 323.0, + 1405.0, + 364.0, + 292.0, + 364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 358.0, + 1403.0, + 358.0, + 1403.0, + 392.0, + 295.0, + 392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 387.0, + 1403.0, + 387.0, + 1403.0, + 422.0, + 293.0, + 422.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 418.0, + 1406.0, + 418.0, + 1406.0, + 456.0, + 293.0, + 456.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 446.0, + 1405.0, + 446.0, + 1405.0, + 484.0, + 292.0, + 484.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 480.0, + 927.0, + 480.0, + 927.0, + 511.0, + 296.0, + 511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 623.0, + 1407.0, + 623.0, + 1407.0, + 660.0, + 293.0, + 660.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 656.0, + 1403.0, + 656.0, + 1403.0, + 689.0, + 295.0, + 689.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 684.0, + 1403.0, + 684.0, + 1403.0, + 723.0, + 293.0, + 723.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 718.0, + 905.0, + 718.0, + 905.0, + 751.0, + 295.0, + 751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 861.0, + 1401.0, + 861.0, + 1401.0, + 896.0, + 295.0, + 896.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 892.0, + 1010.0, + 892.0, + 1010.0, + 927.0, + 297.0, + 927.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 9, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 294, + 238, + 1407, + 238, + 1407, + 2038, + 294, + 2038 + ], + "score": 0.915 + }, + { + "category_id": 2, + "poly": [ + 300, + 75, + 815, + 75, + 815, + 105, + 300, + 105 + ], + "score": 0.883 + }, + { + "category_id": 2, + "poly": [ + 836, + 2088, + 863, + 2088, + 863, + 2113, + 836, + 2113 + ], + "score": 0.825 + }, + { + "category_id": 0, + "poly": [ + 300, + 228, + 488, + 228, + 488, + 261, + 300, + 261 + ], + "score": 0.79 + }, + { + "category_id": 13, + "poly": [ + 800, + 684, + 849, + 684, + 849, + 714, + 800, + 714 + ], + "score": 0.39, + "latex": "( S P )" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 296.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 868.0, + 2085.0, + 868.0, + 2125.0, + 832.0, + 2125.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 227.0, + 491.0, + 227.0, + 491.0, + 265.0, + 295.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 232.0, + 491.0, + 232.0, + 491.0, + 263.0, + 298.0, + 263.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 309.0, + 1405.0, + 309.0, + 1405.0, + 340.0, + 298.0, + 340.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 336.0, + 1407.0, + 336.0, + 1407.0, + 374.0, + 324.0, + 374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 366.0, + 752.0, + 366.0, + 752.0, + 404.0, + 318.0, + 404.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 400.0, + 1403.0, + 400.0, + 1403.0, + 437.0, + 296.0, + 437.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 429.0, + 1405.0, + 429.0, + 1405.0, + 471.0, + 318.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 459.0, + 400.0, + 459.0, + 400.0, + 499.0, + 320.0, + 499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 493.0, + 1403.0, + 493.0, + 1403.0, + 530.0, + 294.0, + 530.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 523.0, + 1407.0, + 523.0, + 1407.0, + 564.0, + 318.0, + 564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 556.0, + 1405.0, + 556.0, + 1405.0, + 594.0, + 322.0, + 594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 586.0, + 563.0, + 586.0, + 563.0, + 623.0, + 322.0, + 623.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 618.0, + 1373.0, + 618.0, + 1373.0, + 653.0, + 292.0, + 653.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 649.0, + 1409.0, + 649.0, + 1409.0, + 691.0, + 290.0, + 691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 683.0, + 799.0, + 683.0, + 799.0, + 720.0, + 320.0, + 720.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 850.0, + 683.0, + 1130.0, + 683.0, + 1130.0, + 720.0, + 850.0, + 720.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 716.0, + 1407.0, + 716.0, + 1407.0, + 754.0, + 296.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 750.0, + 1403.0, + 750.0, + 1403.0, + 782.0, + 322.0, + 782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 774.0, + 400.0, + 774.0, + 400.0, + 813.0, + 320.0, + 813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 809.0, + 1407.0, + 809.0, + 1407.0, + 847.0, + 296.0, + 847.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 837.0, + 1409.0, + 837.0, + 1409.0, + 883.0, + 318.0, + 883.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 869.0, + 668.0, + 869.0, + 668.0, + 906.0, + 316.0, + 906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 906.0, + 1401.0, + 906.0, + 1401.0, + 938.0, + 294.0, + 938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 930.0, + 1405.0, + 930.0, + 1405.0, + 976.0, + 320.0, + 976.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 966.0, + 733.0, + 966.0, + 733.0, + 1003.0, + 322.0, + 1003.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 997.0, + 1405.0, + 997.0, + 1405.0, + 1035.0, + 294.0, + 1035.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1029.0, + 1403.0, + 1029.0, + 1403.0, + 1067.0, + 322.0, + 1067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1061.0, + 553.0, + 1061.0, + 553.0, + 1092.0, + 324.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1092.0, + 1403.0, + 1092.0, + 1403.0, + 1130.0, + 296.0, + 1130.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1124.0, + 1407.0, + 1124.0, + 1407.0, + 1162.0, + 322.0, + 1162.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1154.0, + 934.0, + 1154.0, + 934.0, + 1191.0, + 324.0, + 1191.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1183.0, + 1403.0, + 1183.0, + 1403.0, + 1223.0, + 292.0, + 1223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1219.0, + 637.0, + 1219.0, + 637.0, + 1251.0, + 324.0, + 1251.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1253.0, + 1405.0, + 1253.0, + 1405.0, + 1284.0, + 296.0, + 1284.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1280.0, + 1405.0, + 1280.0, + 1405.0, + 1318.0, + 322.0, + 1318.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1312.0, + 1403.0, + 1312.0, + 1403.0, + 1350.0, + 296.0, + 1350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1346.0, + 1405.0, + 1346.0, + 1405.0, + 1383.0, + 322.0, + 1383.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1377.0, + 483.0, + 1377.0, + 483.0, + 1409.0, + 324.0, + 1409.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1407.0, + 1407.0, + 1407.0, + 1407.0, + 1445.0, + 296.0, + 1445.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1439.0, + 1405.0, + 1439.0, + 1405.0, + 1476.0, + 320.0, + 1476.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1467.0, + 1180.0, + 1467.0, + 1180.0, + 1504.0, + 322.0, + 1504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1498.0, + 1405.0, + 1498.0, + 1405.0, + 1540.0, + 292.0, + 1540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1532.0, + 1407.0, + 1532.0, + 1407.0, + 1569.0, + 322.0, + 1569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 1560.0, + 888.0, + 1560.0, + 888.0, + 1601.0, + 316.0, + 1601.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1599.0, + 1405.0, + 1599.0, + 1405.0, + 1631.0, + 296.0, + 1631.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1627.0, + 1405.0, + 1627.0, + 1405.0, + 1664.0, + 322.0, + 1664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 1655.0, + 1407.0, + 1655.0, + 1407.0, + 1696.0, + 316.0, + 1696.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1690.0, + 1405.0, + 1690.0, + 1405.0, + 1728.0, + 294.0, + 1728.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 1718.0, + 1170.0, + 1718.0, + 1170.0, + 1759.0, + 316.0, + 1759.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1754.0, + 1405.0, + 1754.0, + 1405.0, + 1791.0, + 296.0, + 1791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1785.0, + 1403.0, + 1785.0, + 1403.0, + 1823.0, + 320.0, + 1823.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1813.0, + 467.0, + 1813.0, + 467.0, + 1850.0, + 320.0, + 1850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1847.0, + 1403.0, + 1847.0, + 1403.0, + 1884.0, + 294.0, + 1884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1878.0, + 1405.0, + 1878.0, + 1405.0, + 1916.0, + 322.0, + 1916.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 1906.0, + 922.0, + 1906.0, + 922.0, + 1947.0, + 316.0, + 1947.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1944.0, + 1401.0, + 1944.0, + 1401.0, + 1981.0, + 296.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 1967.0, + 1405.0, + 1967.0, + 1405.0, + 2015.0, + 318.0, + 2015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 2001.0, + 507.0, + 2001.0, + 507.0, + 2039.0, + 324.0, + 2039.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 10, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 294, + 207, + 1409, + 207, + 1409, + 1996, + 294, + 1996 + ], + "score": 0.939 + }, + { + "category_id": 2, + "poly": [ + 299, + 75, + 815, + 75, + 815, + 105, + 299, + 105 + ], + "score": 0.887 + }, + { + "category_id": 2, + "poly": [ + 836, + 2088, + 865, + 2088, + 865, + 2113, + 836, + 2113 + ], + "score": 0.829 + }, + { + "category_id": 15, + "poly": [ + 295.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 295.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 870.0, + 2085.0, + 870.0, + 2124.0, + 832.0, + 2124.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 230.0, + 1407.0, + 230.0, + 1407.0, + 267.0, + 294.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 259.0, + 1409.0, + 259.0, + 1409.0, + 303.0, + 320.0, + 303.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 287.0, + 1053.0, + 287.0, + 1053.0, + 334.0, + 316.0, + 334.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 320.0, + 1407.0, + 320.0, + 1407.0, + 366.0, + 292.0, + 366.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 356.0, + 939.0, + 356.0, + 939.0, + 393.0, + 318.0, + 393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 385.0, + 1405.0, + 385.0, + 1405.0, + 427.0, + 290.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 417.0, + 1407.0, + 417.0, + 1407.0, + 460.0, + 318.0, + 460.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 448.0, + 1099.0, + 448.0, + 1099.0, + 492.0, + 318.0, + 492.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 478.0, + 1407.0, + 478.0, + 1407.0, + 523.0, + 292.0, + 523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 511.0, + 713.0, + 511.0, + 713.0, + 552.0, + 314.0, + 552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 547.0, + 1403.0, + 547.0, + 1403.0, + 584.0, + 296.0, + 584.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 572.0, + 1407.0, + 572.0, + 1407.0, + 619.0, + 320.0, + 619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 608.0, + 537.0, + 608.0, + 537.0, + 643.0, + 318.0, + 643.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 637.0, + 1403.0, + 637.0, + 1403.0, + 678.0, + 294.0, + 678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 667.0, + 1405.0, + 667.0, + 1405.0, + 714.0, + 318.0, + 714.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 700.0, + 680.0, + 700.0, + 680.0, + 743.0, + 318.0, + 743.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 730.0, + 1407.0, + 730.0, + 1407.0, + 775.0, + 290.0, + 775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 763.0, + 889.0, + 763.0, + 889.0, + 806.0, + 318.0, + 806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 793.0, + 1405.0, + 793.0, + 1405.0, + 840.0, + 292.0, + 840.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 828.0, + 396.0, + 828.0, + 396.0, + 867.0, + 320.0, + 867.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 857.0, + 1405.0, + 857.0, + 1405.0, + 903.0, + 292.0, + 903.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 887.0, + 1407.0, + 887.0, + 1407.0, + 932.0, + 316.0, + 932.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 924.0, + 394.0, + 924.0, + 394.0, + 958.0, + 322.0, + 958.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 952.0, + 1409.0, + 952.0, + 1409.0, + 995.0, + 294.0, + 995.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 983.0, + 1163.0, + 983.0, + 1163.0, + 1025.0, + 316.0, + 1025.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1017.0, + 1407.0, + 1017.0, + 1407.0, + 1060.0, + 294.0, + 1060.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1050.0, + 1243.0, + 1050.0, + 1243.0, + 1088.0, + 322.0, + 1088.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1078.0, + 1405.0, + 1078.0, + 1405.0, + 1123.0, + 292.0, + 1123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1111.0, + 1407.0, + 1111.0, + 1407.0, + 1155.0, + 320.0, + 1155.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1145.0, + 1385.0, + 1145.0, + 1385.0, + 1182.0, + 320.0, + 1182.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1178.0, + 1403.0, + 1178.0, + 1403.0, + 1216.0, + 296.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1206.0, + 1407.0, + 1206.0, + 1407.0, + 1249.0, + 320.0, + 1249.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 1237.0, + 699.0, + 1237.0, + 699.0, + 1275.0, + 318.0, + 1275.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1269.0, + 1405.0, + 1269.0, + 1405.0, + 1312.0, + 292.0, + 1312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 1298.0, + 1405.0, + 1298.0, + 1405.0, + 1344.0, + 316.0, + 1344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1330.0, + 630.0, + 1330.0, + 630.0, + 1367.0, + 320.0, + 1367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1365.0, + 1405.0, + 1365.0, + 1405.0, + 1403.0, + 294.0, + 1403.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1397.0, + 1403.0, + 1397.0, + 1403.0, + 1434.0, + 322.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 312.0, + 1422.0, + 1005.0, + 1422.0, + 1005.0, + 1469.0, + 312.0, + 1469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1460.0, + 1403.0, + 1460.0, + 1403.0, + 1497.0, + 294.0, + 1497.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1489.0, + 1405.0, + 1489.0, + 1405.0, + 1527.0, + 320.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 1519.0, + 728.0, + 1519.0, + 728.0, + 1562.0, + 318.0, + 1562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1550.0, + 1405.0, + 1550.0, + 1405.0, + 1593.0, + 294.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 1578.0, + 1407.0, + 1578.0, + 1407.0, + 1629.0, + 314.0, + 1629.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1615.0, + 540.0, + 1615.0, + 540.0, + 1652.0, + 320.0, + 1652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1645.0, + 1405.0, + 1645.0, + 1405.0, + 1688.0, + 294.0, + 1688.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1676.0, + 1407.0, + 1676.0, + 1407.0, + 1719.0, + 320.0, + 1719.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1708.0, + 534.0, + 1708.0, + 534.0, + 1745.0, + 324.0, + 1745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1741.0, + 1407.0, + 1741.0, + 1407.0, + 1778.0, + 294.0, + 1778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1771.0, + 1407.0, + 1771.0, + 1407.0, + 1814.0, + 320.0, + 1814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1804.0, + 1269.0, + 1804.0, + 1269.0, + 1841.0, + 320.0, + 1841.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1833.0, + 953.0, + 1833.0, + 953.0, + 1877.0, + 292.0, + 1877.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1869.0, + 1403.0, + 1869.0, + 1403.0, + 1906.0, + 294.0, + 1906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1900.0, + 863.0, + 1900.0, + 863.0, + 1938.0, + 320.0, + 1938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1930.0, + 1405.0, + 1930.0, + 1405.0, + 1973.0, + 294.0, + 1973.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 1961.0, + 1109.0, + 1961.0, + 1109.0, + 2001.0, + 318.0, + 2001.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 11, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 294, + 228, + 1408, + 228, + 1408, + 2021, + 294, + 2021 + ], + "score": 0.927 + }, + { + "category_id": 2, + "poly": [ + 299, + 75, + 816, + 75, + 816, + 105, + 299, + 105 + ], + "score": 0.889 + }, + { + "category_id": 2, + "poly": [ + 836, + 2088, + 865, + 2088, + 865, + 2113, + 836, + 2113 + ], + "score": 0.842 + }, + { + "category_id": 15, + "poly": [ + 295.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 295.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 870.0, + 2085.0, + 870.0, + 2123.0, + 832.0, + 2123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 227.0, + 1406.0, + 227.0, + 1406.0, + 267.0, + 292.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 261.0, + 1404.0, + 261.0, + 1404.0, + 298.0, + 322.0, + 298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 288.0, + 977.0, + 288.0, + 977.0, + 330.0, + 318.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 322.0, + 1408.0, + 322.0, + 1408.0, + 365.0, + 292.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 355.0, + 1202.0, + 355.0, + 1202.0, + 393.0, + 320.0, + 393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 383.0, + 1408.0, + 383.0, + 1408.0, + 428.0, + 290.0, + 428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 419.0, + 1406.0, + 419.0, + 1406.0, + 456.0, + 322.0, + 456.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 450.0, + 1109.0, + 450.0, + 1109.0, + 488.0, + 320.0, + 488.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 478.0, + 1408.0, + 478.0, + 1408.0, + 523.0, + 292.0, + 523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 511.0, + 731.0, + 511.0, + 731.0, + 549.0, + 318.0, + 549.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 545.0, + 1408.0, + 545.0, + 1408.0, + 584.0, + 292.0, + 584.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 574.0, + 1408.0, + 574.0, + 1408.0, + 616.0, + 320.0, + 616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 604.0, + 703.0, + 604.0, + 703.0, + 643.0, + 318.0, + 643.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 639.0, + 1406.0, + 639.0, + 1406.0, + 677.0, + 296.0, + 677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 667.0, + 1408.0, + 667.0, + 1408.0, + 708.0, + 318.0, + 708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 699.0, + 1001.0, + 699.0, + 1001.0, + 740.0, + 318.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 730.0, + 1406.0, + 730.0, + 1406.0, + 776.0, + 292.0, + 776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 762.0, + 1188.0, + 762.0, + 1188.0, + 803.0, + 316.0, + 803.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 795.0, + 1404.0, + 795.0, + 1404.0, + 833.0, + 296.0, + 833.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 829.0, + 1151.0, + 829.0, + 1151.0, + 866.0, + 320.0, + 866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 860.0, + 1404.0, + 860.0, + 1404.0, + 898.0, + 294.0, + 898.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 892.0, + 1406.0, + 892.0, + 1406.0, + 929.0, + 322.0, + 929.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 923.0, + 1406.0, + 923.0, + 1406.0, + 961.0, + 322.0, + 961.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 951.0, + 617.0, + 951.0, + 617.0, + 990.0, + 318.0, + 990.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 986.0, + 1404.0, + 986.0, + 1404.0, + 1024.0, + 296.0, + 1024.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1012.0, + 1406.0, + 1012.0, + 1406.0, + 1057.0, + 320.0, + 1057.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1044.0, + 1406.0, + 1044.0, + 1406.0, + 1087.0, + 320.0, + 1087.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1081.0, + 1408.0, + 1081.0, + 1408.0, + 1119.0, + 294.0, + 1119.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1109.0, + 1408.0, + 1109.0, + 1408.0, + 1152.0, + 320.0, + 1152.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1144.0, + 538.0, + 1144.0, + 538.0, + 1176.0, + 324.0, + 1176.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1170.0, + 1404.0, + 1170.0, + 1404.0, + 1215.0, + 292.0, + 1215.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1203.0, + 881.0, + 1203.0, + 881.0, + 1239.0, + 320.0, + 1239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1237.0, + 1402.0, + 1237.0, + 1402.0, + 1274.0, + 296.0, + 1274.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1268.0, + 1404.0, + 1268.0, + 1404.0, + 1306.0, + 320.0, + 1306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1300.0, + 1408.0, + 1300.0, + 1408.0, + 1337.0, + 294.0, + 1337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1332.0, + 1370.0, + 1332.0, + 1370.0, + 1369.0, + 320.0, + 1369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1365.0, + 1406.0, + 1365.0, + 1406.0, + 1403.0, + 296.0, + 1403.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1395.0, + 1404.0, + 1395.0, + 1404.0, + 1432.0, + 322.0, + 1432.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 1422.0, + 1406.0, + 1422.0, + 1406.0, + 1468.0, + 318.0, + 1468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1456.0, + 679.0, + 1456.0, + 679.0, + 1491.0, + 320.0, + 1491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1489.0, + 1406.0, + 1489.0, + 1406.0, + 1527.0, + 294.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1521.0, + 1408.0, + 1521.0, + 1408.0, + 1558.0, + 322.0, + 1558.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1548.0, + 1125.0, + 1548.0, + 1125.0, + 1592.0, + 320.0, + 1592.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1584.0, + 1406.0, + 1584.0, + 1406.0, + 1621.0, + 294.0, + 1621.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 1612.0, + 1408.0, + 1612.0, + 1408.0, + 1653.0, + 318.0, + 1653.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1645.0, + 396.0, + 1645.0, + 396.0, + 1679.0, + 324.0, + 1679.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1671.0, + 1406.0, + 1671.0, + 1406.0, + 1722.0, + 290.0, + 1722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 1710.0, + 1045.0, + 1710.0, + 1045.0, + 1748.0, + 320.0, + 1748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1742.0, + 1404.0, + 1742.0, + 1404.0, + 1779.0, + 294.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1773.0, + 1406.0, + 1773.0, + 1406.0, + 1811.0, + 322.0, + 1811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1803.0, + 1296.0, + 1803.0, + 1296.0, + 1840.0, + 322.0, + 1840.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1832.0, + 1406.0, + 1832.0, + 1406.0, + 1874.0, + 294.0, + 1874.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1868.0, + 1406.0, + 1868.0, + 1406.0, + 1905.0, + 322.0, + 1905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 326.0, + 1897.0, + 909.0, + 1897.0, + 909.0, + 1935.0, + 326.0, + 1935.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1929.0, + 1404.0, + 1929.0, + 1404.0, + 1966.0, + 296.0, + 1966.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1961.0, + 1404.0, + 1961.0, + 1404.0, + 1998.0, + 322.0, + 1998.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1992.0, + 484.0, + 1992.0, + 484.0, + 2024.0, + 324.0, + 2024.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 12, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 294, + 231, + 1408, + 231, + 1408, + 703, + 294, + 703 + ], + "score": 0.9 + }, + { + "category_id": 2, + "poly": [ + 300, + 75, + 815, + 75, + 815, + 105, + 300, + 105 + ], + "score": 0.877 + }, + { + "category_id": 2, + "poly": [ + 836, + 2088, + 865, + 2088, + 865, + 2112, + 836, + 2112 + ], + "score": 0.837 + }, + { + "category_id": 15, + "poly": [ + 296.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 296.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2084.0, + 871.0, + 2084.0, + 871.0, + 2122.0, + 832.0, + 2122.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 230.0, + 1405.0, + 230.0, + 1405.0, + 265.0, + 295.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 261.0, + 1005.0, + 261.0, + 1005.0, + 297.0, + 324.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 296.0, + 1400.0, + 296.0, + 1400.0, + 328.0, + 297.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 325.0, + 1042.0, + 325.0, + 1042.0, + 361.0, + 320.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 355.0, + 1405.0, + 355.0, + 1405.0, + 396.0, + 293.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 389.0, + 1405.0, + 389.0, + 1405.0, + 425.0, + 322.0, + 425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 327.0, + 420.0, + 448.0, + 420.0, + 448.0, + 449.0, + 327.0, + 449.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 450.0, + 1405.0, + 450.0, + 1405.0, + 490.0, + 293.0, + 490.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 481.0, + 1405.0, + 481.0, + 1405.0, + 519.0, + 322.0, + 519.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 512.0, + 918.0, + 512.0, + 918.0, + 551.0, + 320.0, + 551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 546.0, + 1405.0, + 546.0, + 1405.0, + 582.0, + 295.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 577.0, + 1405.0, + 577.0, + 1405.0, + 613.0, + 322.0, + 613.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 325.0, + 609.0, + 482.0, + 609.0, + 482.0, + 638.0, + 325.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 638.0, + 1407.0, + 638.0, + 1407.0, + 679.0, + 293.0, + 679.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 321.0, + 670.0, + 1281.0, + 670.0, + 1281.0, + 706.0, + 321.0, + 706.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 13, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 296, + 960, + 1405, + 960, + 1405, + 1275, + 296, + 1275 + ], + "score": 0.981 + }, + { + "category_id": 5, + "poly": [ + 521, + 1540, + 1177, + 1540, + 1177, + 1766, + 521, + 1766 + ], + "score": 0.979, + "html": "
NaturalFGSMPGD20CW20
Madry83.5656.6951.9251.00
CausalAdv-M80.4257.9854.4452.51
CausalAdv-M*80.0357.4752.9852.72
TRADES81.3957.2553.6451.39
CausalAdv-T81.2258.9754.5552.95
CausalAdv-T*79.9458.4654.1252.28
" + }, + { + "category_id": 1, + "poly": [ + 297, + 295, + 1405, + 295, + 1405, + 424, + 297, + 424 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 1918, + 1404, + 1918, + 1404, + 2037, + 298, + 2037 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 296, + 764, + 1405, + 764, + 1405, + 950, + 296, + 950 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 296, + 494, + 1406, + 494, + 1406, + 611, + 296, + 611 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 299, + 1287, + 1405, + 1287, + 1405, + 1382, + 299, + 1382 + ], + "score": 0.97 + }, + { + "category_id": 2, + "poly": [ + 299, + 74, + 816, + 74, + 816, + 105, + 299, + 105 + ], + "score": 0.905 + }, + { + "category_id": 6, + "poly": [ + 294, + 1452, + 1398, + 1452, + 1398, + 1516, + 294, + 1516 + ], + "score": 0.903 + }, + { + "category_id": 8, + "poly": [ + 297, + 630, + 1402, + 630, + 1402, + 718, + 297, + 718 + ], + "score": 0.896 + }, + { + "category_id": 0, + "poly": [ + 298, + 227, + 993, + 227, + 993, + 263, + 298, + 263 + ], + "score": 0.883 + }, + { + "category_id": 9, + "poly": [ + 1351, + 716, + 1400, + 716, + 1400, + 746, + 1351, + 746 + ], + "score": 0.876 + }, + { + "category_id": 0, + "poly": [ + 286, + 1852, + 1305, + 1852, + 1305, + 1889, + 286, + 1889 + ], + "score": 0.865 + }, + { + "category_id": 2, + "poly": [ + 836, + 2087, + 865, + 2087, + 865, + 2113, + 836, + 2113 + ], + "score": 0.864 + }, + { + "category_id": 8, + "poly": [ + 341, + 436, + 1310, + 436, + 1310, + 483, + 341, + 483 + ], + "score": 0.86 + }, + { + "category_id": 9, + "poly": [ + 1352, + 444, + 1400, + 444, + 1400, + 476, + 1352, + 476 + ], + "score": 0.859 + }, + { + "category_id": 13, + "poly": [ + 601, + 528, + 867, + 528, + 867, + 583, + 601, + 583 + ], + "score": 0.96, + "latex": "\\begin{array} { r } { d ( P , Q ) \\leq \\sqrt { \\frac { K L ( P | | Q ) } { 2 } } } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 809, + 1134, + 928, + 1134, + 928, + 1172, + 809, + 1172 + ], + "score": 0.93, + "latex": "P _ { \\theta } \\left( Y | \\tilde { X } \\right)" + }, + { + "category_id": 13, + "poly": [ + 674, + 1022, + 799, + 1022, + 799, + 1061, + 674, + 1061 + ], + "score": 0.93, + "latex": "Q _ { \\theta } \\left( Y | \\tilde { X } \\right)" + }, + { + "category_id": 13, + "poly": [ + 1276, + 1095, + 1401, + 1095, + 1401, + 1136, + 1276, + 1136 + ], + "score": 0.93, + "latex": "Q _ { \\theta } \\left( Y | \\tilde { X } \\right)" + }, + { + "category_id": 13, + "poly": [ + 640, + 1962, + 938, + 1962, + 938, + 2005, + 640, + 2005 + ], + "score": 0.93, + "latex": "\\hat { s } \\left( \\tilde { X } \\right) \\sim \\mathcal { N } \\left( \\mu \\left( \\tilde { X } \\right) , \\sigma ^ { 2 } I \\right)" + }, + { + "category_id": 13, + "poly": [ + 374, + 495, + 495, + 495, + 495, + 530, + 374, + 530 + ], + "score": 0.93, + "latex": "Q _ { \\theta } ( Y | X )" + }, + { + "category_id": 13, + "poly": [ + 886, + 1095, + 1010, + 1095, + 1010, + 1135, + 886, + 1135 + ], + "score": 0.93, + "latex": "Q _ { \\theta } \\left( Y | \\tilde { X } \\right)" + }, + { + "category_id": 13, + "poly": [ + 774, + 848, + 914, + 848, + 914, + 883, + 774, + 883 + ], + "score": 0.93, + "latex": "K L \\left( P | | Q \\right)" + }, + { + "category_id": 13, + "poly": [ + 353, + 1022, + 474, + 1022, + 474, + 1060, + 353, + 1060 + ], + "score": 0.93, + "latex": "P _ { \\theta } \\left( Y | \\tilde { X } \\right)" + }, + { + "category_id": 13, + "poly": [ + 1265, + 798, + 1400, + 798, + 1400, + 853, + 1265, + 853 + ], + "score": 0.93, + "latex": "\\sqrt { \\frac { K L ( P | | Q ) } { 2 } }" + }, + { + "category_id": 13, + "poly": [ + 1131, + 496, + 1232, + 496, + 1232, + 529, + 1131, + 529 + ], + "score": 0.92, + "latex": "h \\left( X ; \\pmb { \\theta } \\right)" + }, + { + "category_id": 13, + "poly": [ + 713, + 1099, + 812, + 1099, + 812, + 1133, + 713, + 1133 + ], + "score": 0.92, + "latex": "h \\left( X ; \\theta \\right)" + }, + { + "category_id": 13, + "poly": [ + 368, + 799, + 503, + 799, + 503, + 851, + 368, + 851 + ], + "score": 0.92, + "latex": "\\sqrt { \\frac { K L ( P | | Q ) } { 2 } }" + }, + { + "category_id": 13, + "poly": [ + 574, + 810, + 714, + 810, + 714, + 846, + 574, + 846 + ], + "score": 0.92, + "latex": "K L \\left( P | | Q \\right)" + }, + { + "category_id": 13, + "poly": [ + 664, + 1206, + 1007, + 1206, + 1007, + 1247, + 664, + 1247 + ], + "score": 0.92, + "latex": "K L \\left( P _ { \\pmb { \\theta } } \\left( Y | \\tilde { X } \\right) | | Q _ { \\pmb { \\theta } } \\left( Y | \\tilde { X } \\right) \\right)" + }, + { + "category_id": 13, + "poly": [ + 756, + 328, + 1043, + 328, + 1043, + 367, + 756, + 367 + ], + "score": 0.92, + "latex": "d \\left( P \\left( Y | X \\right) , P _ { \\theta } \\left( Y | \\tilde { X } \\right) \\right)" + }, + { + "category_id": 13, + "poly": [ + 619, + 2003, + 687, + 2003, + 687, + 2035, + 619, + 2035 + ], + "score": 0.92, + "latex": "( x , y )" + }, + { + "category_id": 13, + "poly": [ + 664, + 1918, + 1160, + 1918, + 1160, + 1961, + 664, + 1961 + ], + "score": 0.91, + "latex": "\\mathbb { E } _ { \\hat { s } \\left( \\tilde { X } \\right) \\sim \\mathcal { N } \\left( \\mu \\left( \\tilde { X } \\right) , \\sigma ^ { 2 } I \\right) } C E \\left( g \\left( \\hat { s } \\left( \\tilde { X } \\right) ; W _ { g } \\right) , Y \\right)" + }, + { + "category_id": 13, + "poly": [ + 363, + 1059, + 703, + 1059, + 703, + 1099, + 363, + 1099 + ], + "score": 0.91, + "latex": "K L \\left( P _ { \\theta } \\left( Y | \\tilde { X } \\right) | | Q _ { \\theta } \\left( Y | \\tilde { X } \\right) \\right)" + }, + { + "category_id": 13, + "poly": [ + 297, + 1169, + 627, + 1169, + 627, + 1210, + 297, + 1210 + ], + "score": 0.91, + "latex": "K L \\left( P \\left( Y | X \\right) | | Q _ { \\theta } \\left( Y | \\tilde { X } \\right) \\right)" + }, + { + "category_id": 13, + "poly": [ + 1026, + 880, + 1403, + 880, + 1403, + 920, + 1026, + 920 + ], + "score": 0.9, + "latex": "K L \\left( \\bar { P } _ { \\theta } \\left( \\bar { Y } | \\tilde { X } \\right) | | Q _ { \\theta } \\left( \\bar { Y } | \\tilde { X } \\right) \\right) +" + }, + { + "category_id": 13, + "poly": [ + 347, + 851, + 487, + 851, + 487, + 884, + 347, + 884 + ], + "score": 0.9, + "latex": "K L \\left( P | | Q \\right)" + }, + { + "category_id": 14, + "poly": [ + 310, + 626, + 1410, + 626, + 1410, + 721, + 310, + 721 + ], + "score": 0.9, + "latex": "l \\bigl ( P \\left( Y | X \\right) , P _ { \\theta } \\left( Y | \\tilde { X } \\right) \\bigr ) \\leq \\sqrt { \\frac { K L \\left( P _ { \\theta } \\left( Y | \\tilde { X } \\right) | | Q _ { \\theta } \\left( Y | \\tilde { X } \\right) \\right) } { 2 } } + \\sqrt { \\frac { K L \\left( P \\left( Y | X \\right) | | Q _ { \\theta } \\left( Y | X \\right) \\right) } { 2 } } ." + }, + { + "category_id": 14, + "poly": [ + 341, + 436, + 1308, + 436, + 1308, + 481, + 341, + 481 + ], + "score": 0.9, + "latex": "d \\big ( P \\left( Y | X \\right) , P _ { \\theta } \\left( Y | \\tilde { X } \\right) \\big ) \\leq d \\big ( P _ { \\theta } \\left( Y | \\tilde { X } \\right) , Q _ { \\theta } \\left( Y | \\tilde { X } \\right) \\big ) + d \\left( P \\left( Y | X \\right) , Q _ { \\theta } \\left( Y | X \\right) \\right) ," + }, + { + "category_id": 13, + "poly": [ + 1276, + 1452, + 1394, + 1452, + 1394, + 1486, + 1276, + 1486 + ], + "score": 0.9, + "latex": "\\epsilon = 8 / 2 5 5" + }, + { + "category_id": 13, + "poly": [ + 296, + 914, + 648, + 914, + 648, + 950, + 296, + 950 + ], + "score": 0.89, + "latex": "\\gamma K L \\left( P \\left( Y | X \\right) | | Q _ { \\theta } \\left( Y | X \\right) \\right)" + }, + { + "category_id": 13, + "poly": [ + 954, + 542, + 1002, + 542, + 1002, + 571, + 954, + 571 + ], + "score": 0.86, + "latex": "K L" + }, + { + "category_id": 13, + "poly": [ + 734, + 920, + 755, + 920, + 755, + 948, + 734, + 948 + ], + "score": 0.81, + "latex": "\\gamma" + }, + { + "category_id": 13, + "poly": [ + 374, + 364, + 393, + 364, + 393, + 390, + 374, + 390 + ], + "score": 0.81, + "latex": "d" + }, + { + "category_id": 13, + "poly": [ + 653, + 1455, + 696, + 1455, + 696, + 1484, + 653, + 1484 + ], + "score": 0.76, + "latex": "( \\% )" + }, + { + "category_id": 13, + "poly": [ + 1275, + 1027, + 1294, + 1027, + 1294, + 1055, + 1275, + 1055 + ], + "score": 0.72, + "latex": "\\theta" + }, + { + "category_id": 13, + "poly": [ + 1067, + 1027, + 1085, + 1027, + 1085, + 1054, + 1067, + 1054 + ], + "score": 0.67, + "latex": "\\theta" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 72.0, + 817.0, + 72.0, + 817.0, + 108.0, + 295.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1453.0, + 652.0, + 1453.0, + 652.0, + 1485.0, + 295.0, + 1485.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 697.0, + 1453.0, + 1275.0, + 1453.0, + 1275.0, + 1485.0, + 697.0, + 1485.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1395.0, + 1453.0, + 1402.0, + 1453.0, + 1402.0, + 1485.0, + 1395.0, + 1485.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1478.0, + 891.0, + 1478.0, + 891.0, + 1520.0, + 292.0, + 1520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 224.0, + 997.0, + 224.0, + 997.0, + 271.0, + 294.0, + 271.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1851.0, + 1310.0, + 1851.0, + 1310.0, + 1895.0, + 291.0, + 1895.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 870.0, + 2085.0, + 870.0, + 2124.0, + 832.0, + 2124.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 959.0, + 1405.0, + 959.0, + 1405.0, + 997.0, + 293.0, + 997.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 993.0, + 1402.0, + 993.0, + 1402.0, + 1025.0, + 296.0, + 1025.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1023.0, + 352.0, + 1023.0, + 352.0, + 1063.0, + 294.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 475.0, + 1023.0, + 673.0, + 1023.0, + 673.0, + 1063.0, + 475.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 800.0, + 1023.0, + 1066.0, + 1023.0, + 1066.0, + 1063.0, + 800.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1086.0, + 1023.0, + 1274.0, + 1023.0, + 1274.0, + 1063.0, + 1086.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1295.0, + 1023.0, + 1405.0, + 1023.0, + 1405.0, + 1063.0, + 1295.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1062.0, + 362.0, + 1062.0, + 362.0, + 1098.0, + 295.0, + 1098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 704.0, + 1062.0, + 1403.0, + 1062.0, + 1403.0, + 1098.0, + 704.0, + 1098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1097.0, + 712.0, + 1097.0, + 712.0, + 1137.0, + 294.0, + 1137.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 813.0, + 1097.0, + 885.0, + 1097.0, + 885.0, + 1137.0, + 813.0, + 1137.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1011.0, + 1097.0, + 1275.0, + 1097.0, + 1275.0, + 1137.0, + 1011.0, + 1137.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1132.0, + 808.0, + 1132.0, + 808.0, + 1176.0, + 293.0, + 1176.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 929.0, + 1132.0, + 1405.0, + 1132.0, + 1405.0, + 1176.0, + 929.0, + 1176.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 628.0, + 1173.0, + 1403.0, + 1173.0, + 1403.0, + 1210.0, + 628.0, + 1210.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1207.0, + 663.0, + 1207.0, + 663.0, + 1251.0, + 291.0, + 1251.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1008.0, + 1207.0, + 1407.0, + 1207.0, + 1407.0, + 1251.0, + 1008.0, + 1251.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1239.0, + 786.0, + 1239.0, + 786.0, + 1276.0, + 294.0, + 1276.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 297.0, + 1403.0, + 297.0, + 1403.0, + 331.0, + 297.0, + 331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 330.0, + 755.0, + 330.0, + 755.0, + 370.0, + 292.0, + 370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1044.0, + 330.0, + 1406.0, + 330.0, + 1406.0, + 370.0, + 1044.0, + 370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 363.0, + 373.0, + 363.0, + 373.0, + 396.0, + 294.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 363.0, + 1405.0, + 363.0, + 1405.0, + 396.0, + 394.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 389.0, + 496.0, + 389.0, + 496.0, + 426.0, + 294.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1916.0, + 663.0, + 1916.0, + 663.0, + 1967.0, + 293.0, + 1967.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1161.0, + 1916.0, + 1408.0, + 1916.0, + 1408.0, + 1967.0, + 1161.0, + 1967.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1961.0, + 639.0, + 1961.0, + 639.0, + 2004.0, + 294.0, + 2004.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 939.0, + 1961.0, + 1404.0, + 1961.0, + 1404.0, + 2004.0, + 939.0, + 2004.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2001.0, + 618.0, + 2001.0, + 618.0, + 2039.0, + 294.0, + 2039.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 688.0, + 2001.0, + 797.0, + 2001.0, + 797.0, + 2039.0, + 688.0, + 2039.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 765.0, + 1403.0, + 765.0, + 1403.0, + 803.0, + 294.0, + 803.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 809.0, + 367.0, + 809.0, + 367.0, + 849.0, + 295.0, + 849.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 504.0, + 798.0, + 573.0, + 798.0, + 573.0, + 853.0, + 504.0, + 853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 715.0, + 798.0, + 1264.0, + 798.0, + 1264.0, + 853.0, + 715.0, + 853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 848.0, + 346.0, + 848.0, + 346.0, + 888.0, + 293.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 488.0, + 848.0, + 773.0, + 848.0, + 773.0, + 888.0, + 488.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 915.0, + 848.0, + 1406.0, + 848.0, + 1406.0, + 888.0, + 915.0, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 881.0, + 1025.0, + 881.0, + 1025.0, + 921.0, + 294.0, + 921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 649.0, + 916.0, + 733.0, + 916.0, + 733.0, + 951.0, + 649.0, + 951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 916.0, + 1077.0, + 916.0, + 1077.0, + 951.0, + 756.0, + 951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 494.0, + 373.0, + 494.0, + 373.0, + 528.0, + 296.0, + 528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 496.0, + 494.0, + 1130.0, + 494.0, + 1130.0, + 528.0, + 496.0, + 528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1233.0, + 494.0, + 1403.0, + 494.0, + 1403.0, + 528.0, + 1233.0, + 528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 538.0, + 600.0, + 538.0, + 600.0, + 580.0, + 294.0, + 580.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 868.0, + 525.0, + 953.0, + 525.0, + 953.0, + 586.0, + 868.0, + 586.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1003.0, + 525.0, + 1406.0, + 525.0, + 1406.0, + 586.0, + 1003.0, + 586.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 577.0, + 691.0, + 577.0, + 691.0, + 611.0, + 295.0, + 611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1289.0, + 1403.0, + 1289.0, + 1403.0, + 1320.0, + 297.0, + 1320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1317.0, + 1402.0, + 1317.0, + 1402.0, + 1352.0, + 295.0, + 1352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1347.0, + 1008.0, + 1347.0, + 1008.0, + 1383.0, + 294.0, + 1383.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 14, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 450, + 1581, + 1246, + 1581, + 1246, + 1807, + 450, + 1807 + ], + "score": 0.98, + "html": "
NaturalFGSMPGD20CW20
Madry83.5656.6951.9251.00
CausalAdv-M+Linear80.6857.1853.3651.41
CausalAdv-M+Non-linear80.4257.9854.4452.51
TRADES81.3957.2553.6451.39
CausalAdv-T+Linear80.3158.4354.3152.25
CausalAdv-T+Non-linear81.2258.9754.5552.95
" + }, + { + "category_id": 1, + "poly": [ + 297, + 1281, + 1404, + 1281, + 1404, + 1437, + 297, + 1437 + ], + "score": 0.978 + }, + { + "category_id": 8, + "poly": [ + 526, + 247, + 1175, + 247, + 1175, + 963, + 526, + 963 + ], + "score": 0.966 + }, + { + "category_id": 8, + "poly": [ + 426, + 1153, + 1269, + 1153, + 1269, + 1256, + 426, + 1256 + ], + "score": 0.958 + }, + { + "category_id": 8, + "poly": [ + 650, + 1043, + 1047, + 1043, + 1047, + 1093, + 650, + 1093 + ], + "score": 0.946 + }, + { + "category_id": 1, + "poly": [ + 298, + 967, + 1401, + 967, + 1401, + 1031, + 298, + 1031 + ], + "score": 0.945 + }, + { + "category_id": 1, + "poly": [ + 297, + 1971, + 1397, + 1971, + 1397, + 2035, + 297, + 2035 + ], + "score": 0.943 + }, + { + "category_id": 0, + "poly": [ + 299, + 1865, + 1231, + 1865, + 1231, + 1938, + 299, + 1938 + ], + "score": 0.932 + }, + { + "category_id": 1, + "poly": [ + 296, + 1101, + 998, + 1101, + 998, + 1141, + 296, + 1141 + ], + "score": 0.927 + }, + { + "category_id": 2, + "poly": [ + 298, + 75, + 815, + 75, + 815, + 105, + 298, + 105 + ], + "score": 0.9 + }, + { + "category_id": 6, + "poly": [ + 297, + 1463, + 1400, + 1463, + 1400, + 1557, + 297, + 1557 + ], + "score": 0.898 + }, + { + "category_id": 9, + "poly": [ + 1352, + 1059, + 1400, + 1059, + 1400, + 1090, + 1352, + 1090 + ], + "score": 0.895 + }, + { + "category_id": 9, + "poly": [ + 1351, + 1184, + 1400, + 1184, + 1400, + 1215, + 1351, + 1215 + ], + "score": 0.888 + }, + { + "category_id": 9, + "poly": [ + 1350, + 591, + 1402, + 591, + 1402, + 623, + 1350, + 623 + ], + "score": 0.884 + }, + { + "category_id": 2, + "poly": [ + 836, + 2088, + 865, + 2088, + 865, + 2113, + 836, + 2113 + ], + "score": 0.861 + }, + { + "category_id": 14, + "poly": [ + 426, + 1150, + 1270, + 1150, + 1270, + 1258, + 426, + 1258 + ], + "score": 0.93, + "latex": "P ( ( Y = y | \\overline { { g } } ( \\boldsymbol { \\mu } ( \\boldsymbol { x } ) ; W _ { g } ) ) \\triangleq \\frac { W _ { g , y } ^ { \\top } \\mu ( \\boldsymbol { x } ) } { \\sum _ { j = 1 } e ^ { W _ { g , j } ^ { \\top } \\mu ( \\boldsymbol { x } ) + \\frac { \\sigma ^ { 2 } } { 2 } ( W _ { g , j } - W _ { g , y } ) \\top } ( W _ { g , j } - W _ { g , y } ) } ." + }, + { + "category_id": 13, + "poly": [ + 668, + 1102, + 831, + 1102, + 831, + 1141, + 668, + 1141 + ], + "score": 0.93, + "latex": "\\overline { { g } } \\left( \\mu \\left( x \\right) ; W _ { g } \\right)" + }, + { + "category_id": 14, + "poly": [ + 524, + 248, + 1175, + 248, + 1175, + 966, + 524, + 966 + ], + "score": 0.93, + "latex": "\\begin{array} { r l } & { \\frac { 3 } { 2 } \\Delta \\phi _ { \\phi } \\varepsilon \\partial \\phi _ { \\phi } ^ { \\star } ( x , y ) + \\varepsilon \\mathcal { F } \\mathcal { F } \\mathcal { F } \\mathcal { F } ( \\phi \\xi ( x ; x ; \\theta ) , \\theta ) } \\\\ & { = 2 e _ { - \\phi , x \\phi , y } \\frac { 1 } { \\varepsilon ( x , y ) } e _ { x , y } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } } \\\\ & { \\quad \\times \\log \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) \\sin ( x , y ) + \\varepsilon \\mathcal { F } ( x , y ) } \\frac { 1 } { \\varepsilon ( x , y ) } } \\\\ & { \\quad - \\log \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) \\sin ( x , y ) + \\varepsilon \\mathcal { F } ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } } \\\\ & { \\quad - \\log \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) \\sin ( x , y ) + \\varepsilon \\mathcal { F } ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } } \\\\ & { \\quad - \\log \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) + \\varepsilon \\mathcal { F } ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } } \\\\ & { \\quad - \\log \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) + \\varepsilon \\mathcal { F } ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } , \\ \\exp _ { ( ( \\phi , x ) ) } \\ , \\ } \\\\ & { \\quad - \\log \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } } \\\\ & { \\quad - \\log \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } } \\\\ & { \\quad \\times \\log \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } \\frac { \\varepsilon ( x , y ) } { \\varepsilon ( x , y ) } } \\\\ & \\quad \\ \\end{array}" + }, + { + "category_id": 14, + "poly": [ + 649, + 1042, + 1050, + 1042, + 1050, + 1093, + 649, + 1093 + ], + "score": 0.93, + "latex": "\\mathbb { E } e ^ { t X } = e ^ { t \\mu + \\frac { 1 } { 2 } { \\sigma } ^ { 2 } t ^ { 2 } } , X \\sim { \\mathcal { N } } ( \\mu , { \\sigma } ^ { 2 } ) ." + }, + { + "category_id": 13, + "poly": [ + 957, + 968, + 1190, + 968, + 1190, + 1001, + 957, + 1001 + ], + "score": 0.92, + "latex": "\\mathbb { E } \\log ( X ) \\le \\log \\mathbb { E } X" + }, + { + "category_id": 13, + "poly": [ + 1276, + 1464, + 1394, + 1464, + 1394, + 1497, + 1276, + 1497 + ], + "score": 0.9, + "latex": "\\epsilon = 8 / 2 5 5" + }, + { + "category_id": 13, + "poly": [ + 492, + 2004, + 519, + 2004, + 519, + 2030, + 492, + 2030 + ], + "score": 0.81, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 664, + 2004, + 691, + 2004, + 691, + 2030, + 664, + 2030 + ], + "score": 0.81, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 465, + 1289, + 483, + 1289, + 483, + 1315, + 465, + 1315 + ], + "score": 0.8, + "latex": "g" + }, + { + "category_id": 13, + "poly": [ + 752, + 2004, + 776, + 2004, + 776, + 2030, + 752, + 2030 + ], + "score": 0.8, + "latex": "R" + }, + { + "category_id": 13, + "poly": [ + 1128, + 1974, + 1153, + 1974, + 1153, + 2001, + 1128, + 2001 + ], + "score": 0.79, + "latex": "R" + }, + { + "category_id": 13, + "poly": [ + 1334, + 1319, + 1352, + 1319, + 1352, + 1346, + 1334, + 1346 + ], + "score": 0.79, + "latex": "g" + }, + { + "category_id": 13, + "poly": [ + 683, + 1499, + 702, + 1499, + 702, + 1527, + 683, + 1527 + ], + "score": 0.79, + "latex": "g" + }, + { + "category_id": 13, + "poly": [ + 421, + 1529, + 440, + 1529, + 440, + 1557, + 421, + 1557 + ], + "score": 0.78, + "latex": "g" + }, + { + "category_id": 13, + "poly": [ + 737, + 1349, + 755, + 1349, + 755, + 1375, + 737, + 1375 + ], + "score": 0.77, + "latex": "g" + }, + { + "category_id": 13, + "poly": [ + 653, + 1466, + 695, + 1466, + 695, + 1494, + 653, + 1494 + ], + "score": 0.76, + "latex": "( \\% )" + }, + { + "category_id": 13, + "poly": [ + 1257, + 1378, + 1276, + 1378, + 1276, + 1407, + 1257, + 1407 + ], + "score": 0.73, + "latex": "g" + }, + { + "category_id": 13, + "poly": [ + 364, + 1496, + 396, + 1496, + 396, + 1523, + 364, + 1523 + ], + "score": 0.61, + "latex": "^ { 6 6 } +" + }, + { + "category_id": 13, + "poly": [ + 1159, + 1496, + 1189, + 1496, + 1189, + 1523, + 1159, + 1523 + ], + "score": 0.35, + "latex": "^ { 6 6 } +" + }, + { + "category_id": 13, + "poly": [ + 1155, + 1495, + 1240, + 1495, + 1240, + 1524, + 1155, + 1524 + ], + "score": 0.29, + "latex": "^ { 6 6 } + \\mathrm { N o n }" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1865.0, + 1236.0, + 1865.0, + 1236.0, + 1906.0, + 295.0, + 1906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1910.0, + 582.0, + 1910.0, + 582.0, + 1942.0, + 354.0, + 1942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 70.0, + 818.0, + 70.0, + 818.0, + 110.0, + 294.0, + 110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1463.0, + 652.0, + 1463.0, + 652.0, + 1497.0, + 293.0, + 1497.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 696.0, + 1463.0, + 1275.0, + 1463.0, + 1275.0, + 1497.0, + 696.0, + 1497.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1395.0, + 1463.0, + 1402.0, + 1463.0, + 1402.0, + 1497.0, + 1395.0, + 1497.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1494.0, + 363.0, + 1494.0, + 363.0, + 1528.0, + 295.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 397.0, + 1494.0, + 682.0, + 1494.0, + 682.0, + 1528.0, + 397.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 703.0, + 1494.0, + 1154.0, + 1494.0, + 1154.0, + 1528.0, + 703.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1241.0, + 1494.0, + 1405.0, + 1494.0, + 1405.0, + 1528.0, + 1241.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1526.0, + 420.0, + 1526.0, + 420.0, + 1556.0, + 296.0, + 1556.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 441.0, + 1526.0, + 855.0, + 1526.0, + 855.0, + 1556.0, + 441.0, + 1556.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 871.0, + 2085.0, + 871.0, + 2123.0, + 832.0, + 2123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1282.0, + 464.0, + 1282.0, + 464.0, + 1315.0, + 296.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 484.0, + 1282.0, + 1402.0, + 1282.0, + 1402.0, + 1315.0, + 484.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1312.0, + 1333.0, + 1312.0, + 1333.0, + 1350.0, + 295.0, + 1350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1353.0, + 1312.0, + 1406.0, + 1312.0, + 1406.0, + 1350.0, + 1353.0, + 1350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1344.0, + 736.0, + 1344.0, + 736.0, + 1378.0, + 296.0, + 1378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 1344.0, + 1405.0, + 1344.0, + 1405.0, + 1378.0, + 756.0, + 1378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1371.0, + 1256.0, + 1371.0, + 1256.0, + 1410.0, + 294.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1277.0, + 1371.0, + 1406.0, + 1371.0, + 1406.0, + 1410.0, + 1277.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1404.0, + 1301.0, + 1404.0, + 1301.0, + 1438.0, + 296.0, + 1438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 966.0, + 956.0, + 966.0, + 956.0, + 1005.0, + 294.0, + 1005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1191.0, + 966.0, + 1406.0, + 966.0, + 1406.0, + 1005.0, + 1191.0, + 1005.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 999.0, + 942.0, + 999.0, + 942.0, + 1031.0, + 297.0, + 1031.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1971.0, + 1127.0, + 1971.0, + 1127.0, + 2007.0, + 296.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1154.0, + 1971.0, + 1402.0, + 1971.0, + 1402.0, + 2007.0, + 1154.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 2000.0, + 491.0, + 2000.0, + 491.0, + 2040.0, + 293.0, + 2040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 520.0, + 2000.0, + 663.0, + 2000.0, + 663.0, + 2040.0, + 520.0, + 2040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 692.0, + 2000.0, + 751.0, + 2000.0, + 751.0, + 2040.0, + 692.0, + 2040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 777.0, + 2000.0, + 942.0, + 2000.0, + 942.0, + 2040.0, + 777.0, + 2040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1096.0, + 667.0, + 1096.0, + 667.0, + 1146.0, + 291.0, + 1146.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 1096.0, + 1000.0, + 1096.0, + 1000.0, + 1146.0, + 832.0, + 1146.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 15, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1187, + 1405, + 1187, + 1405, + 1563, + 298, + 1563 + ], + "score": 0.985 + }, + { + "category_id": 5, + "poly": [ + 326, + 1764, + 1373, + 1764, + 1373, + 1941, + 326, + 1941 + ], + "score": 0.979, + "html": "
MethodBest checkpointLast checkpoint
NaturalFGSMPGD-20CW-20NaturalFGSMPGD-20CW-20
Madry86.6359.4853.6553.5886.6057.0749.2349.46
CausalAdv-M85.2461.2255.1755.6885.6160.0851.7652.59
TRADES84.3260.9456.6954.8784.8659.9452.0452.39
CausalAdv-T84.1961.6257.3655.7584.3561.5755.1555.23
" + }, + { + "category_id": 1, + "poly": [ + 295, + 225, + 1406, + 225, + 1406, + 557, + 295, + 557 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 297, + 1018, + 1404, + 1018, + 1404, + 1174, + 297, + 1174 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 299, + 588, + 1400, + 588, + 1400, + 653, + 299, + 653 + ], + "score": 0.937 + }, + { + "category_id": 8, + "poly": [ + 316, + 655, + 1388, + 655, + 1388, + 852, + 316, + 852 + ], + "score": 0.933 + }, + { + "category_id": 0, + "poly": [ + 297, + 1602, + 1009, + 1602, + 1009, + 1641, + 297, + 1641 + ], + "score": 0.93 + }, + { + "category_id": 2, + "poly": [ + 299, + 75, + 815, + 75, + 815, + 105, + 299, + 105 + ], + "score": 0.9 + }, + { + "category_id": 0, + "poly": [ + 294, + 956, + 1359, + 956, + 1359, + 989, + 294, + 989 + ], + "score": 0.898 + }, + { + "category_id": 1, + "poly": [ + 297, + 1971, + 1403, + 1971, + 1403, + 2035, + 297, + 2035 + ], + "score": 0.871 + }, + { + "category_id": 2, + "poly": [ + 836, + 2087, + 864, + 2087, + 864, + 2112, + 836, + 2112 + ], + "score": 0.858 + }, + { + "category_id": 6, + "poly": [ + 296, + 1673, + 1398, + 1673, + 1398, + 1740, + 296, + 1740 + ], + "score": 0.792 + }, + { + "category_id": 9, + "poly": [ + 1352, + 849, + 1400, + 849, + 1400, + 878, + 1352, + 878 + ], + "score": 0.727 + }, + { + "category_id": 2, + "poly": [ + 1374, + 882, + 1402, + 882, + 1402, + 907, + 1374, + 907 + ], + "score": 0.188 + }, + { + "category_id": 1, + "poly": [ + 296, + 1673, + 1398, + 1673, + 1398, + 1740, + 296, + 1740 + ], + "score": 0.122 + }, + { + "category_id": 13, + "poly": [ + 368, + 1341, + 471, + 1341, + 471, + 1377, + 368, + 1377 + ], + "score": 0.95, + "latex": "2 \\times { 1 0 } ^ { - 4 }" + }, + { + "category_id": 13, + "poly": [ + 684, + 1112, + 796, + 1112, + 796, + 1145, + 684, + 1145 + ], + "score": 0.93, + "latex": "\\eta = \\epsilon / 1 0" + }, + { + "category_id": 13, + "poly": [ + 1223, + 1407, + 1346, + 1407, + 1346, + 1440, + 1223, + 1440 + ], + "score": 0.92, + "latex": "\\eta = 2 / 2 5 5" + }, + { + "category_id": 13, + "poly": [ + 1056, + 393, + 1175, + 393, + 1175, + 426, + 1056, + 426 + ], + "score": 0.92, + "latex": "\\mu \\left( R ; W _ { s } \\right)" + }, + { + "category_id": 13, + "poly": [ + 577, + 494, + 792, + 494, + 792, + 527, + 577, + 527 + ], + "score": 0.92, + "latex": "\\langle a , b \\rangle _ { M } = \\langle a , M b \\rangle" + }, + { + "category_id": 13, + "poly": [ + 805, + 1408, + 924, + 1408, + 924, + 1440, + 805, + 1440 + ], + "score": 0.91, + "latex": "\\epsilon = 8 / 2 5 5" + }, + { + "category_id": 14, + "poly": [ + 313, + 650, + 1383, + 650, + 1383, + 854, + 313, + 854 + ], + "score": 0.91, + "latex": "\\begin{array} { r l } & { \\ker ( W _ { s } ) ^ { \\perp } \\perp _ { M } \\ker ( W _ { c } ) ^ { \\perp } \\Longleftrightarrow \\operatorname { i m } ( W _ { s } ^ { \\top } ) \\perp _ { M } \\operatorname { i m } ( W _ { c } ^ { \\top } ) \\Longleftrightarrow \\langle W _ { s } ^ { \\top } { \\boldsymbol a } , W _ { c } ^ { \\top } b \\rangle _ { M } = 0 \\forall { \\boldsymbol a } , { \\boldsymbol b } } \\\\ & { \\Longleftrightarrow \\langle W _ { s } ^ { \\top } { \\boldsymbol a } , M W _ { c } ^ { \\top } { \\boldsymbol b } \\rangle = 0 \\forall { \\boldsymbol a } , { \\boldsymbol b } \\Longleftrightarrow W _ { s } M W _ { c } ^ { \\top } = 0 \\Longleftrightarrow \\mathbb { E } _ { R } W _ { s } \\left( R - m \\right) \\left( R - m \\right) ^ { \\top } W _ { c } ^ { \\top } = 0 } \\\\ & { \\Longleftrightarrow \\mathbb { E } _ { R , n } W _ { s } \\Big ( R + \\Sigma ^ { \\frac { 1 } { 2 } } n - m \\Big ) ( R - m ) ^ { \\top } W _ { c } ^ { \\top } = 0 \\Longleftrightarrow C o v \\left( \\hat { s } \\left( R \\right) , \\hat { c } \\left( R \\right) \\right) = 0 } \\\\ & { \\Longleftrightarrow \\hat { c } ( R ) \\perp \\hat { s } \\left( R \\right) } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 298, + 1112, + 422, + 1112, + 422, + 1145, + 298, + 1145 + ], + "score": 0.91, + "latex": "\\epsilon = 8 / 2 5 5" + }, + { + "category_id": 13, + "poly": [ + 428, + 224, + 520, + 224, + 520, + 259, + 428, + 259 + ], + "score": 0.91, + "latex": "R \\in \\mathbb { R } ^ { d }" + }, + { + "category_id": 13, + "poly": [ + 1164, + 292, + 1321, + 292, + 1321, + 324, + 1164, + 324 + ], + "score": 0.91, + "latex": "{ \\hat { c } } ( R ) = W _ { c } R _ { }" + }, + { + "category_id": 13, + "poly": [ + 477, + 1052, + 580, + 1052, + 580, + 1084, + 477, + 1084 + ], + "score": 0.9, + "latex": "\\eta = 0 . 0 1" + }, + { + "category_id": 13, + "poly": [ + 916, + 1052, + 1010, + 1052, + 1010, + 1080, + 916, + 1080 + ], + "score": 0.9, + "latex": "K = 4 0" + }, + { + "category_id": 13, + "poly": [ + 570, + 1439, + 661, + 1439, + 661, + 1467, + 570, + 1467 + ], + "score": 0.9, + "latex": "K = 1 0" + }, + { + "category_id": 13, + "poly": [ + 870, + 592, + 910, + 592, + 910, + 622, + 870, + 622 + ], + "score": 0.9, + "latex": "W _ { s }" + }, + { + "category_id": 13, + "poly": [ + 298, + 525, + 407, + 525, + 407, + 556, + 298, + 556 + ], + "score": 0.9, + "latex": "A \\perp _ { M } B" + }, + { + "category_id": 13, + "poly": [ + 1138, + 1113, + 1235, + 1113, + 1235, + 1141, + 1138, + 1141 + ], + "score": 0.9, + "latex": "K = 2 0" + }, + { + "category_id": 13, + "poly": [ + 346, + 423, + 655, + 423, + 655, + 466, + 346, + 466 + ], + "score": 0.89, + "latex": "\\hat { s } \\left( R \\right) = W _ { s } R + \\Sigma \\left( Y \\right) ^ { \\frac { 1 } { 2 } } n" + }, + { + "category_id": 13, + "poly": [ + 390, + 621, + 430, + 621, + 430, + 652, + 390, + 652 + ], + "score": 0.89, + "latex": "W _ { c }" + }, + { + "category_id": 13, + "poly": [ + 1226, + 1022, + 1311, + 1022, + 1311, + 1050, + 1226, + 1050 + ], + "score": 0.89, + "latex": "\\epsilon = 0 . 3" + }, + { + "category_id": 13, + "poly": [ + 755, + 362, + 796, + 362, + 796, + 394, + 755, + 394 + ], + "score": 0.88, + "latex": "W _ { s }" + }, + { + "category_id": 13, + "poly": [ + 328, + 465, + 368, + 465, + 368, + 495, + 328, + 495 + ], + "score": 0.88, + "latex": "W _ { c }" + }, + { + "category_id": 13, + "poly": [ + 297, + 352, + 668, + 352, + 668, + 395, + 297, + 395 + ], + "score": 0.88, + "latex": "\\hat { s } \\left( R \\right) = \\mu \\left( R ; W _ { s } \\right) + \\bar { \\Sigma } \\left( Y \\right) ^ { \\frac { 1 } { 2 } } \\bar { n } ," + }, + { + "category_id": 13, + "poly": [ + 823, + 434, + 863, + 434, + 863, + 464, + 823, + 464 + ], + "score": 0.87, + "latex": "W _ { s }" + }, + { + "category_id": 13, + "poly": [ + 298, + 323, + 338, + 323, + 338, + 354, + 298, + 354 + ], + "score": 0.86, + "latex": "W _ { c }" + }, + { + "category_id": 13, + "poly": [ + 682, + 528, + 707, + 528, + 707, + 552, + 682, + 552 + ], + "score": 0.83, + "latex": "B" + }, + { + "category_id": 13, + "poly": [ + 297, + 1469, + 318, + 1469, + 318, + 1501, + 297, + 1501 + ], + "score": 0.83, + "latex": "\\beta" + }, + { + "category_id": 13, + "poly": [ + 693, + 1470, + 713, + 1470, + 713, + 1496, + 693, + 1496 + ], + "score": 0.79, + "latex": "\\lambda" + }, + { + "category_id": 13, + "poly": [ + 668, + 1501, + 689, + 1501, + 689, + 1527, + 668, + 1527 + ], + "score": 0.74, + "latex": "\\lambda" + }, + { + "category_id": 13, + "poly": [ + 1066, + 262, + 1099, + 262, + 1099, + 289, + 1066, + 289 + ], + "score": 0.73, + "latex": "M" + }, + { + "category_id": 13, + "poly": [ + 654, + 1677, + 697, + 1677, + 697, + 1706, + 654, + 1706 + ], + "score": 0.73, + "latex": "( \\% )" + }, + { + "category_id": 13, + "poly": [ + 606, + 528, + 630, + 528, + 630, + 552, + 606, + 552 + ], + "score": 0.7, + "latex": "A" + }, + { + "category_id": 13, + "poly": [ + 347, + 399, + 369, + 399, + 369, + 420, + 347, + 420 + ], + "score": 0.69, + "latex": "\\textbf { \\em n }" + }, + { + "category_id": 13, + "poly": [ + 714, + 324, + 739, + 324, + 739, + 350, + 714, + 350 + ], + "score": 0.69, + "latex": "R" + }, + { + "category_id": 13, + "poly": [ + 1081, + 293, + 1105, + 293, + 1105, + 320, + 1081, + 320 + ], + "score": 0.69, + "latex": "R" + }, + { + "category_id": 13, + "poly": [ + 349, + 263, + 373, + 263, + 373, + 289, + 349, + 289 + ], + "score": 0.69, + "latex": "R" + }, + { + "category_id": 13, + "poly": [ + 801, + 465, + 865, + 465, + 865, + 496, + 801, + 496 + ], + "score": 0.6, + "latex": "{ \\hat { c } } \\left( R \\right)" + }, + { + "category_id": 13, + "poly": [ + 1153, + 495, + 1185, + 495, + 1185, + 522, + 1153, + 522 + ], + "score": 0.57, + "latex": "M" + }, + { + "category_id": 13, + "poly": [ + 801, + 464, + 972, + 464, + 972, + 496, + 801, + 496 + ], + "score": 0.56, + "latex": "\\hat { c } \\left( R \\right) \\perp \\perp \\hat { s } \\left( R \\right)" + }, + { + "category_id": 13, + "poly": [ + 1059, + 464, + 1084, + 464, + 1084, + 492, + 1059, + 492 + ], + "score": 0.39, + "latex": "\\perp" + }, + { + "category_id": 13, + "poly": [ + 1277, + 232, + 1301, + 232, + 1301, + 258, + 1277, + 258 + ], + "score": 0.38, + "latex": "R" + }, + { + "category_id": 13, + "poly": [ + 906, + 464, + 971, + 464, + 971, + 496, + 906, + 496 + ], + "score": 0.3, + "latex": "{ \\hat { s } } \\left( R \\right)" + }, + { + "category_id": 13, + "poly": [ + 924, + 232, + 942, + 232, + 942, + 258, + 924, + 258 + ], + "score": 0.26, + "latex": "d" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1603.0, + 1009.0, + 1603.0, + 1009.0, + 1644.0, + 294.0, + 1644.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 295.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 952.0, + 1360.0, + 952.0, + 1360.0, + 993.0, + 294.0, + 993.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 869.0, + 2085.0, + 869.0, + 2124.0, + 832.0, + 2124.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1674.0, + 653.0, + 1674.0, + 653.0, + 1707.0, + 297.0, + 1707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 698.0, + 1674.0, + 1399.0, + 1674.0, + 1399.0, + 1707.0, + 698.0, + 1707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1703.0, + 1160.0, + 1703.0, + 1160.0, + 1742.0, + 295.0, + 1742.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1378.0, + 884.0, + 1402.0, + 884.0, + 1402.0, + 910.0, + 1378.0, + 910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1188.0, + 1403.0, + 1188.0, + 1403.0, + 1223.0, + 296.0, + 1223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1220.0, + 1405.0, + 1220.0, + 1405.0, + 1255.0, + 296.0, + 1255.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1250.0, + 1403.0, + 1250.0, + 1403.0, + 1285.0, + 295.0, + 1285.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1281.0, + 1403.0, + 1281.0, + 1403.0, + 1313.0, + 293.0, + 1313.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1309.0, + 1405.0, + 1309.0, + 1405.0, + 1345.0, + 295.0, + 1345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1344.0, + 367.0, + 1344.0, + 367.0, + 1381.0, + 295.0, + 1381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 472.0, + 1344.0, + 1405.0, + 1344.0, + 1405.0, + 1381.0, + 472.0, + 1381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1377.0, + 1405.0, + 1377.0, + 1405.0, + 1412.0, + 295.0, + 1412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1406.0, + 804.0, + 1406.0, + 804.0, + 1441.0, + 293.0, + 1441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 925.0, + 1406.0, + 1222.0, + 1406.0, + 1222.0, + 1441.0, + 925.0, + 1441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1347.0, + 1406.0, + 1406.0, + 1406.0, + 1406.0, + 1441.0, + 1347.0, + 1441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1436.0, + 569.0, + 1436.0, + 569.0, + 1473.0, + 293.0, + 1473.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 662.0, + 1436.0, + 1405.0, + 1436.0, + 1405.0, + 1473.0, + 662.0, + 1473.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1467.0, + 296.0, + 1467.0, + 296.0, + 1503.0, + 293.0, + 1503.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 319.0, + 1467.0, + 692.0, + 1467.0, + 692.0, + 1503.0, + 319.0, + 1503.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 714.0, + 1467.0, + 1405.0, + 1467.0, + 1405.0, + 1503.0, + 714.0, + 1503.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1499.0, + 667.0, + 1499.0, + 667.0, + 1534.0, + 295.0, + 1534.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 690.0, + 1499.0, + 1406.0, + 1499.0, + 1406.0, + 1534.0, + 690.0, + 1534.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1528.0, + 444.0, + 1528.0, + 444.0, + 1566.0, + 292.0, + 1566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 224.0, + 427.0, + 224.0, + 427.0, + 268.0, + 293.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 521.0, + 224.0, + 923.0, + 224.0, + 923.0, + 268.0, + 521.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 943.0, + 224.0, + 1276.0, + 224.0, + 1276.0, + 268.0, + 943.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1302.0, + 224.0, + 1406.0, + 224.0, + 1406.0, + 268.0, + 1302.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 260.0, + 348.0, + 260.0, + 348.0, + 297.0, + 294.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 260.0, + 1065.0, + 260.0, + 1065.0, + 297.0, + 374.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 260.0, + 1404.0, + 260.0, + 1404.0, + 297.0, + 1100.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 292.0, + 1080.0, + 292.0, + 1080.0, + 326.0, + 295.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1106.0, + 292.0, + 1163.0, + 292.0, + 1163.0, + 326.0, + 1106.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1322.0, + 292.0, + 1404.0, + 292.0, + 1404.0, + 326.0, + 1322.0, + 326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 320.0, + 297.0, + 320.0, + 297.0, + 359.0, + 294.0, + 359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 339.0, + 320.0, + 713.0, + 320.0, + 713.0, + 359.0, + 339.0, + 359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 740.0, + 320.0, + 1404.0, + 320.0, + 1404.0, + 359.0, + 740.0, + 359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 359.0, + 296.0, + 359.0, + 296.0, + 400.0, + 293.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 669.0, + 359.0, + 754.0, + 359.0, + 754.0, + 400.0, + 669.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 797.0, + 359.0, + 1404.0, + 359.0, + 1404.0, + 400.0, + 797.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 388.0, + 346.0, + 388.0, + 346.0, + 430.0, + 292.0, + 430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 370.0, + 388.0, + 1055.0, + 388.0, + 1055.0, + 430.0, + 370.0, + 430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1176.0, + 388.0, + 1404.0, + 388.0, + 1404.0, + 430.0, + 1176.0, + 430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 429.0, + 345.0, + 429.0, + 345.0, + 471.0, + 292.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 656.0, + 429.0, + 822.0, + 429.0, + 822.0, + 471.0, + 656.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 864.0, + 429.0, + 1406.0, + 429.0, + 1406.0, + 471.0, + 864.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 461.0, + 327.0, + 461.0, + 327.0, + 499.0, + 294.0, + 499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 461.0, + 800.0, + 461.0, + 800.0, + 499.0, + 369.0, + 499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 973.0, + 461.0, + 1058.0, + 461.0, + 1058.0, + 499.0, + 973.0, + 499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1085.0, + 461.0, + 1406.0, + 461.0, + 1406.0, + 499.0, + 1085.0, + 499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 493.0, + 576.0, + 493.0, + 576.0, + 531.0, + 292.0, + 531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.0, + 493.0, + 1152.0, + 493.0, + 1152.0, + 531.0, + 793.0, + 531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1186.0, + 493.0, + 1406.0, + 493.0, + 1406.0, + 531.0, + 1186.0, + 531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 525.0, + 297.0, + 525.0, + 297.0, + 559.0, + 294.0, + 559.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 408.0, + 525.0, + 605.0, + 525.0, + 605.0, + 559.0, + 408.0, + 559.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 631.0, + 525.0, + 681.0, + 525.0, + 681.0, + 559.0, + 631.0, + 559.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 708.0, + 525.0, + 924.0, + 525.0, + 924.0, + 559.0, + 708.0, + 559.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1020.0, + 1225.0, + 1020.0, + 1225.0, + 1053.0, + 295.0, + 1053.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1312.0, + 1020.0, + 1404.0, + 1020.0, + 1404.0, + 1053.0, + 1312.0, + 1053.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1049.0, + 476.0, + 1049.0, + 476.0, + 1087.0, + 294.0, + 1087.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 581.0, + 1049.0, + 915.0, + 1049.0, + 915.0, + 1087.0, + 581.0, + 1087.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1011.0, + 1049.0, + 1405.0, + 1049.0, + 1405.0, + 1087.0, + 1011.0, + 1087.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1080.0, + 1404.0, + 1080.0, + 1404.0, + 1113.0, + 295.0, + 1113.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1111.0, + 297.0, + 1111.0, + 297.0, + 1148.0, + 291.0, + 1148.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 423.0, + 1111.0, + 683.0, + 1111.0, + 683.0, + 1148.0, + 423.0, + 1148.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 797.0, + 1111.0, + 1137.0, + 1111.0, + 1137.0, + 1148.0, + 797.0, + 1148.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1236.0, + 1111.0, + 1406.0, + 1111.0, + 1406.0, + 1148.0, + 1236.0, + 1148.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1141.0, + 388.0, + 1141.0, + 388.0, + 1177.0, + 294.0, + 1177.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 587.0, + 869.0, + 587.0, + 869.0, + 627.0, + 294.0, + 627.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 911.0, + 587.0, + 1401.0, + 587.0, + 1401.0, + 627.0, + 911.0, + 627.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 619.0, + 389.0, + 619.0, + 389.0, + 655.0, + 294.0, + 655.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 431.0, + 619.0, + 547.0, + 619.0, + 547.0, + 655.0, + 431.0, + 655.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1967.0, + 1405.0, + 1967.0, + 1405.0, + 2009.0, + 292.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 2002.0, + 1403.0, + 2002.0, + 1403.0, + 2034.0, + 297.0, + 2034.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1674.0, + 653.0, + 1674.0, + 653.0, + 1707.0, + 297.0, + 1707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 698.0, + 1674.0, + 1399.0, + 1674.0, + 1399.0, + 1707.0, + 698.0, + 1707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1703.0, + 1160.0, + 1703.0, + 1160.0, + 1742.0, + 295.0, + 1742.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 16, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1116, + 1404, + 1116, + 1404, + 1364, + 297, + 1364 + ], + "score": 0.983 + }, + { + "category_id": 5, + "poly": [ + 312, + 658, + 1381, + 658, + 1381, + 851, + 312, + 851 + ], + "score": 0.981, + "html": "
Methodt1t2t3CIFAR-10CIFAR-100
FGSMPGD-20CW-20FGSMPGD-20CW-20
Madry56.6951.9251.0056.6951.9251.00
CausalAdv-M专√57.9854.4452.5157.9854.4452.51
TRADES√√57.2553.6451.3957.2553.6451.39
CausalAdv-T√√√58.9754.5552.9558.9754.5552.95
" + }, + { + "category_id": 1, + "poly": [ + 297, + 886, + 1404, + 886, + 1404, + 1102, + 297, + 1102 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 297, + 228, + 1404, + 228, + 1404, + 415, + 297, + 415 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 300, + 1941, + 1402, + 1941, + 1402, + 2033, + 300, + 2033 + ], + "score": 0.973 + }, + { + "category_id": 3, + "poly": [ + 566, + 1444, + 1096, + 1444, + 1096, + 1847, + 566, + 1847 + ], + "score": 0.971 + }, + { + "category_id": 4, + "poly": [ + 304, + 1873, + 1378, + 1873, + 1378, + 1907, + 304, + 1907 + ], + "score": 0.917 + }, + { + "category_id": 6, + "poly": [ + 298, + 539, + 1404, + 539, + 1404, + 634, + 298, + 634 + ], + "score": 0.907 + }, + { + "category_id": 2, + "poly": [ + 299, + 75, + 815, + 75, + 815, + 104, + 299, + 104 + ], + "score": 0.901 + }, + { + "category_id": 0, + "poly": [ + 300, + 461, + 606, + 461, + 606, + 495, + 300, + 495 + ], + "score": 0.894 + }, + { + "category_id": 2, + "poly": [ + 836, + 2088, + 864, + 2088, + 864, + 2112, + 836, + 2112 + ], + "score": 0.859 + }, + { + "category_id": 13, + "poly": [ + 887, + 1238, + 1360, + 1238, + 1360, + 1273, + 887, + 1273 + ], + "score": 0.92, + "latex": "K L \\left( P \\left( Y | h \\left( x , s \\right) \\right) | | P \\left( Y | h \\left( \\dot { x } _ { a d v } , s \\right) \\right) \\right)" + }, + { + "category_id": 13, + "poly": [ + 1255, + 1873, + 1372, + 1873, + 1372, + 1907, + 1255, + 1907 + ], + "score": 0.9, + "latex": "\\epsilon = 8 / 2 5 5" + }, + { + "category_id": 13, + "poly": [ + 835, + 573, + 862, + 573, + 862, + 602, + 835, + 602 + ], + "score": 0.87, + "latex": "t _ { 3 }" + }, + { + "category_id": 13, + "poly": [ + 1143, + 981, + 1187, + 981, + 1187, + 1011, + 1143, + 1011 + ], + "score": 0.83, + "latex": "( t _ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 749, + 573, + 775, + 573, + 775, + 602, + 749, + 602 + ], + "score": 0.81, + "latex": "t _ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 711, + 573, + 737, + 573, + 737, + 602, + 711, + 602 + ], + "score": 0.81, + "latex": "t _ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 298, + 1011, + 341, + 1011, + 341, + 1041, + 298, + 1041 + ], + "score": 0.8, + "latex": "( t _ { 3 } )" + }, + { + "category_id": 13, + "poly": [ + 1077, + 1042, + 1120, + 1042, + 1120, + 1072, + 1077, + 1072 + ], + "score": 0.79, + "latex": "( t _ { 3 } )" + }, + { + "category_id": 13, + "poly": [ + 601, + 542, + 645, + 542, + 645, + 571, + 601, + 571 + ], + "score": 0.77, + "latex": "( \\% )" + }, + { + "category_id": 13, + "poly": [ + 691, + 1876, + 734, + 1876, + 734, + 1905, + 691, + 1905 + ], + "score": 0.71, + "latex": "( \\% )" + }, + { + "category_id": 13, + "poly": [ + 425, + 1276, + 445, + 1276, + 445, + 1298, + 425, + 1298 + ], + "score": 0.7, + "latex": "x" + }, + { + "category_id": 15, + "poly": [ + 949.0, + 1453.0, + 1023.0, + 1453.0, + 1023.0, + 1492.0, + 949.0, + 1492.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 593.0, + 1480.0, + 620.0, + 1480.0, + 620.0, + 1504.0, + 593.0, + 1504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 949.0, + 1484.0, + 1082.0, + 1484.0, + 1082.0, + 1516.0, + 949.0, + 1516.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 590.0, + 1522.0, + 627.0, + 1522.0, + 627.0, + 1551.0, + 590.0, + 1551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 563.0, + 1557.0, + 602.0, + 1557.0, + 602.0, + 1693.0, + 563.0, + 1693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 590.0, + 1564.0, + 621.0, + 1564.0, + 621.0, + 1597.0, + 590.0, + 1597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 1654.0, + 619.0, + 1654.0, + 619.0, + 1684.0, + 592.0, + 1684.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 1699.0, + 621.0, + 1699.0, + 621.0, + 1726.0, + 592.0, + 1726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 593.0, + 1744.0, + 620.0, + 1744.0, + 620.0, + 1768.0, + 593.0, + 1768.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 603.0, + 1792.0, + 618.0, + 1792.0, + 618.0, + 1809.0, + 603.0, + 1809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 646.0, + 1802.0, + 796.0, + 1802.0, + 796.0, + 1827.0, + 646.0, + 1827.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 800.0, + 1804.0, + 834.0, + 1804.0, + 834.0, + 1824.0, + 800.0, + 1824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 843.0, + 1804.0, + 873.0, + 1804.0, + 873.0, + 1825.0, + 843.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 881.0, + 1804.0, + 913.0, + 1804.0, + 913.0, + 1825.0, + 881.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 921.0, + 1804.0, + 952.0, + 1804.0, + 952.0, + 1825.0, + 921.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 957.0, + 1801.0, + 1072.0, + 1801.0, + 1072.0, + 1826.0, + 957.0, + 1826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 786.0, + 1819.0, + 929.0, + 1819.0, + 929.0, + 1850.0, + 786.0, + 1850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 1870.0, + 690.0, + 1870.0, + 690.0, + 1911.0, + 314.0, + 1911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 735.0, + 1870.0, + 1254.0, + 1870.0, + 1254.0, + 1911.0, + 735.0, + 1911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1373.0, + 1870.0, + 1383.0, + 1870.0, + 1383.0, + 1911.0, + 1373.0, + 1911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 539.0, + 600.0, + 539.0, + 600.0, + 573.0, + 294.0, + 573.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 646.0, + 539.0, + 1404.0, + 539.0, + 1404.0, + 573.0, + 646.0, + 573.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 569.0, + 710.0, + 569.0, + 710.0, + 604.0, + 294.0, + 604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 738.0, + 569.0, + 748.0, + 569.0, + 748.0, + 604.0, + 738.0, + 604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 776.0, + 569.0, + 834.0, + 569.0, + 834.0, + 604.0, + 776.0, + 604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 863.0, + 569.0, + 1402.0, + 569.0, + 1402.0, + 604.0, + 863.0, + 604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 601.0, + 1396.0, + 601.0, + 1396.0, + 635.0, + 296.0, + 635.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 72.0, + 817.0, + 72.0, + 817.0, + 108.0, + 296.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 458.0, + 611.0, + 458.0, + 611.0, + 501.0, + 293.0, + 501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 870.0, + 2085.0, + 870.0, + 2123.0, + 832.0, + 2123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1114.0, + 1404.0, + 1114.0, + 1404.0, + 1153.0, + 294.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1148.0, + 1406.0, + 1148.0, + 1406.0, + 1182.0, + 295.0, + 1182.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1178.0, + 1404.0, + 1178.0, + 1404.0, + 1212.0, + 294.0, + 1212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1210.0, + 1402.0, + 1210.0, + 1402.0, + 1244.0, + 294.0, + 1244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1236.0, + 886.0, + 1236.0, + 886.0, + 1276.0, + 291.0, + 1276.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1361.0, + 1236.0, + 1406.0, + 1236.0, + 1406.0, + 1276.0, + 1361.0, + 1276.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1268.0, + 424.0, + 1268.0, + 424.0, + 1304.0, + 292.0, + 1304.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 446.0, + 1268.0, + 1405.0, + 1268.0, + 1405.0, + 1304.0, + 446.0, + 1304.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1300.0, + 1404.0, + 1300.0, + 1404.0, + 1334.0, + 295.0, + 1334.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1331.0, + 1405.0, + 1331.0, + 1405.0, + 1364.0, + 294.0, + 1364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 888.0, + 1405.0, + 888.0, + 1405.0, + 922.0, + 296.0, + 922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 917.0, + 1404.0, + 917.0, + 1404.0, + 952.0, + 295.0, + 952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 943.0, + 1405.0, + 943.0, + 1405.0, + 987.0, + 291.0, + 987.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 979.0, + 1142.0, + 979.0, + 1142.0, + 1014.0, + 294.0, + 1014.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1188.0, + 979.0, + 1405.0, + 979.0, + 1405.0, + 1014.0, + 1188.0, + 1014.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1006.0, + 297.0, + 1006.0, + 297.0, + 1047.0, + 291.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 342.0, + 1006.0, + 1405.0, + 1006.0, + 1405.0, + 1047.0, + 342.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1037.0, + 1076.0, + 1037.0, + 1076.0, + 1078.0, + 292.0, + 1078.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1121.0, + 1037.0, + 1406.0, + 1037.0, + 1406.0, + 1078.0, + 1121.0, + 1078.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1069.0, + 430.0, + 1069.0, + 430.0, + 1104.0, + 293.0, + 1104.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 231.0, + 1402.0, + 231.0, + 1402.0, + 263.0, + 296.0, + 263.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 260.0, + 1404.0, + 260.0, + 1404.0, + 296.0, + 295.0, + 296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 291.0, + 1405.0, + 291.0, + 1405.0, + 325.0, + 294.0, + 325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 320.0, + 1405.0, + 320.0, + 1405.0, + 357.0, + 294.0, + 357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 352.0, + 1402.0, + 352.0, + 1402.0, + 384.0, + 296.0, + 384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 384.0, + 962.0, + 384.0, + 962.0, + 416.0, + 295.0, + 416.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1940.0, + 1406.0, + 1940.0, + 1406.0, + 1979.0, + 293.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1971.0, + 1404.0, + 1971.0, + 1404.0, + 2007.0, + 294.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 2001.0, + 807.0, + 2001.0, + 807.0, + 2036.0, + 295.0, + 2036.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 17, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 298, + 851, + 1396, + 851, + 1396, + 1249, + 298, + 1249 + ], + "score": 0.984, + "html": "
CIFAR10CIFAR100
NaturalFGSMPGD20CW20NaturalFGSMPGD20CW20
Madry83.5656.6951.9251.0055.9828.3925.1524.04
CausalAdv-M入 = 0.282.1757.1353.2851.8054.6829.8527.3225.88
CausalAdv-M入= 0.580.8357.2853.6551.9754.0729.7627.6225.44
CausalAdv-M入= 1.080.4257.9854.4452.5152.6929.7727.8026.04
CausalAdv-M入= 1.580.1757.5353.2451.4052.3730.7428.4926.71
CausalAdv-M入 = 2.080.3558.2953.1652.7148.4030.0728.5225.97
TRADES81.3957.2553.6451.3953.8529.0427.9124.09
CausalAdv-T入= 0.281.7258.2654.0651.9053.8329.7628.0524.49
CausalAdv-T入= 0.581.2258.9754.5552.9553.9130.1928.1124.90
CausalAdv-T入= 1.079.6558.4854.4552.8953.1730.6628.5725.74
CausalAdv-T入= 1.578.0957.4253.6651.3852.2230.4928.1725.48
CausalAdv-T入= 2.074.4255.2952.0750.0450.4030.2228.1425.49
" + }, + { + "category_id": 1, + "poly": [ + 298, + 1499, + 1404, + 1499, + 1404, + 1805, + 298, + 1805 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 298, + 1819, + 1402, + 1819, + 1402, + 2034, + 298, + 2034 + ], + "score": 0.978 + }, + { + "category_id": 5, + "poly": [ + 305, + 467, + 1390, + 467, + 1390, + 600, + 305, + 600 + ], + "score": 0.975, + "html": "
CIFAR-10CIFAR-100
NaturalFGSMPGD20CW20NaturalFGSMPGD20CW20
TRADES81.3957.2553.6451.3953.8529.0427.9124.09
CausalAdv-T81.7258.2654.0651.9053.9130.1928.1124.90
" + }, + { + "category_id": 1, + "poly": [ + 299, + 229, + 1401, + 229, + 1401, + 353, + 299, + 353 + ], + "score": 0.971 + }, + { + "category_id": 1, + "poly": [ + 301, + 1420, + 1399, + 1420, + 1399, + 1484, + 301, + 1484 + ], + "score": 0.951 + }, + { + "category_id": 1, + "poly": [ + 298, + 674, + 1401, + 674, + 1401, + 736, + 298, + 736 + ], + "score": 0.943 + }, + { + "category_id": 6, + "poly": [ + 299, + 379, + 1399, + 379, + 1399, + 442, + 299, + 442 + ], + "score": 0.921 + }, + { + "category_id": 2, + "poly": [ + 300, + 76, + 815, + 76, + 815, + 104, + 300, + 104 + ], + "score": 0.9 + }, + { + "category_id": 6, + "poly": [ + 301, + 762, + 1400, + 762, + 1400, + 824, + 301, + 824 + ], + "score": 0.893 + }, + { + "category_id": 0, + "poly": [ + 297, + 1356, + 1056, + 1356, + 1056, + 1388, + 297, + 1388 + ], + "score": 0.89 + }, + { + "category_id": 2, + "poly": [ + 837, + 2088, + 865, + 2088, + 865, + 2112, + 837, + 2112 + ], + "score": 0.844 + }, + { + "category_id": 7, + "poly": [ + 297, + 1277, + 1257, + 1277, + 1257, + 1310, + 297, + 1310 + ], + "score": 0.384 + }, + { + "category_id": 1, + "poly": [ + 297, + 1277, + 1257, + 1277, + 1257, + 1310, + 297, + 1310 + ], + "score": 0.213 + }, + { + "category_id": 13, + "poly": [ + 460, + 1105, + 542, + 1105, + 542, + 1130, + 460, + 1130 + ], + "score": 0.73, + "latex": "\\lambda = 0 . 2" + }, + { + "category_id": 13, + "poly": [ + 601, + 382, + 645, + 382, + 645, + 411, + 601, + 411 + ], + "score": 0.71, + "latex": "( \\% )" + }, + { + "category_id": 13, + "poly": [ + 463, + 938, + 546, + 938, + 546, + 963, + 463, + 963 + ], + "score": 0.7, + "latex": "\\lambda = 0 . 2" + }, + { + "category_id": 13, + "poly": [ + 601, + 765, + 644, + 765, + 644, + 794, + 601, + 794 + ], + "score": 0.69, + "latex": "( \\% )" + }, + { + "category_id": 13, + "poly": [ + 460, + 1216, + 542, + 1216, + 542, + 1242, + 460, + 1242 + ], + "score": 0.55, + "latex": "\\lambda = 2 . 0" + }, + { + "category_id": 13, + "poly": [ + 460, + 1134, + 543, + 1134, + 543, + 1158, + 460, + 1158 + ], + "score": 0.47, + "latex": "\\lambda = 0 . 5" + }, + { + "category_id": 13, + "poly": [ + 463, + 1049, + 545, + 1049, + 545, + 1074, + 463, + 1074 + ], + "score": 0.46, + "latex": "\\lambda = 2 . 0" + }, + { + "category_id": 13, + "poly": [ + 461, + 1189, + 542, + 1189, + 542, + 1213, + 461, + 1213 + ], + "score": 0.4, + "latex": "\\lambda = 1 . 5" + }, + { + "category_id": 13, + "poly": [ + 464, + 966, + 546, + 966, + 546, + 990, + 464, + 990 + ], + "score": 0.4, + "latex": "\\lambda = 0 . 5" + }, + { + "category_id": 13, + "poly": [ + 463, + 994, + 546, + 994, + 546, + 1019, + 463, + 1019 + ], + "score": 0.36, + "latex": "\\lambda = 1 . 0" + }, + { + "category_id": 13, + "poly": [ + 461, + 1161, + 543, + 1161, + 543, + 1185, + 461, + 1185 + ], + "score": 0.36, + "latex": "\\lambda = 1 . 0" + }, + { + "category_id": 13, + "poly": [ + 463, + 1022, + 545, + 1022, + 545, + 1046, + 463, + 1046 + ], + "score": 0.34, + "latex": "\\lambda = 1 . 5" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 377.0, + 600.0, + 377.0, + 600.0, + 413.0, + 295.0, + 413.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 646.0, + 377.0, + 1403.0, + 377.0, + 1403.0, + 413.0, + 646.0, + 413.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 409.0, + 452.0, + 409.0, + 452.0, + 442.0, + 296.0, + 442.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 296.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 761.0, + 600.0, + 761.0, + 600.0, + 797.0, + 296.0, + 797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 761.0, + 1404.0, + 761.0, + 1404.0, + 797.0, + 645.0, + 797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 792.0, + 450.0, + 792.0, + 450.0, + 825.0, + 297.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1350.0, + 1059.0, + 1350.0, + 1059.0, + 1397.0, + 295.0, + 1397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2085.0, + 869.0, + 2085.0, + 869.0, + 2124.0, + 832.0, + 2124.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1271.0, + 1264.0, + 1271.0, + 1264.0, + 1317.0, + 292.0, + 1317.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1497.0, + 1406.0, + 1497.0, + 1406.0, + 1531.0, + 296.0, + 1531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1528.0, + 1404.0, + 1528.0, + 1404.0, + 1562.0, + 293.0, + 1562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1559.0, + 1405.0, + 1559.0, + 1405.0, + 1595.0, + 294.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1589.0, + 1404.0, + 1589.0, + 1404.0, + 1624.0, + 294.0, + 1624.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1618.0, + 1404.0, + 1618.0, + 1404.0, + 1655.0, + 296.0, + 1655.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1648.0, + 1405.0, + 1648.0, + 1405.0, + 1686.0, + 293.0, + 1686.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1682.0, + 1404.0, + 1682.0, + 1404.0, + 1714.0, + 296.0, + 1714.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1712.0, + 1404.0, + 1712.0, + 1404.0, + 1747.0, + 294.0, + 1747.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1742.0, + 1405.0, + 1742.0, + 1405.0, + 1778.0, + 294.0, + 1778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1771.0, + 1183.0, + 1771.0, + 1183.0, + 1809.0, + 293.0, + 1809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1816.0, + 1402.0, + 1816.0, + 1402.0, + 1855.0, + 293.0, + 1855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1848.0, + 1406.0, + 1848.0, + 1406.0, + 1883.0, + 293.0, + 1883.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1881.0, + 1401.0, + 1881.0, + 1401.0, + 1915.0, + 294.0, + 1915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1909.0, + 1406.0, + 1909.0, + 1406.0, + 1945.0, + 294.0, + 1945.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1941.0, + 1403.0, + 1941.0, + 1403.0, + 1972.0, + 297.0, + 1972.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1973.0, + 1404.0, + 1973.0, + 1404.0, + 2004.0, + 297.0, + 2004.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 2000.0, + 1404.0, + 2000.0, + 1404.0, + 2036.0, + 293.0, + 2036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 229.0, + 1406.0, + 229.0, + 1406.0, + 265.0, + 294.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 261.0, + 1403.0, + 261.0, + 1403.0, + 293.0, + 295.0, + 293.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 292.0, + 1405.0, + 292.0, + 1405.0, + 325.0, + 295.0, + 325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 319.0, + 1405.0, + 319.0, + 1405.0, + 358.0, + 293.0, + 358.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1419.0, + 1404.0, + 1419.0, + 1404.0, + 1457.0, + 295.0, + 1457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1453.0, + 563.0, + 1453.0, + 563.0, + 1485.0, + 297.0, + 1485.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 672.0, + 1403.0, + 672.0, + 1403.0, + 708.0, + 296.0, + 708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 706.0, + 1403.0, + 706.0, + 1403.0, + 738.0, + 297.0, + 738.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1271.0, + 1264.0, + 1271.0, + 1264.0, + 1317.0, + 292.0, + 1317.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 18, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 299, + 307, + 1403, + 307, + 1403, + 672, + 299, + 672 + ], + "score": 0.985 + }, + { + "category_id": 1, + "poly": [ + 298, + 1079, + 1403, + 1079, + 1403, + 1567, + 298, + 1567 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 299, + 787, + 1403, + 787, + 1403, + 1064, + 299, + 1064 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 295, + 229, + 1399, + 229, + 1399, + 292, + 295, + 292 + ], + "score": 0.952 + }, + { + "category_id": 2, + "poly": [ + 300, + 76, + 815, + 76, + 815, + 104, + 300, + 104 + ], + "score": 0.893 + }, + { + "category_id": 0, + "poly": [ + 298, + 722, + 984, + 722, + 984, + 754, + 298, + 754 + ], + "score": 0.891 + }, + { + "category_id": 2, + "poly": [ + 835, + 2088, + 865, + 2088, + 865, + 2112, + 835, + 2112 + ], + "score": 0.845 + }, + { + "category_id": 13, + "poly": [ + 936, + 462, + 963, + 462, + 963, + 488, + 936, + 488 + ], + "score": 0.79, + "latex": "X" + }, + { + "category_id": 13, + "poly": [ + 1044, + 461, + 1069, + 461, + 1069, + 488, + 1044, + 488 + ], + "score": 0.71, + "latex": "Y" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 73.0, + 817.0, + 73.0, + 817.0, + 108.0, + 296.0, + 108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 716.0, + 988.0, + 716.0, + 988.0, + 762.0, + 293.0, + 762.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 829.0, + 2084.0, + 871.0, + 2084.0, + 871.0, + 2123.0, + 829.0, + 2123.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 307.0, + 1404.0, + 307.0, + 1404.0, + 343.0, + 294.0, + 343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 341.0, + 1404.0, + 341.0, + 1404.0, + 371.0, + 295.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 368.0, + 1405.0, + 368.0, + 1405.0, + 402.0, + 295.0, + 402.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 400.0, + 1405.0, + 400.0, + 1405.0, + 434.0, + 295.0, + 434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 430.0, + 1403.0, + 430.0, + 1403.0, + 464.0, + 295.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 460.0, + 935.0, + 460.0, + 935.0, + 494.0, + 295.0, + 494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 460.0, + 1043.0, + 460.0, + 1043.0, + 494.0, + 964.0, + 494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1070.0, + 460.0, + 1404.0, + 460.0, + 1404.0, + 494.0, + 1070.0, + 494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 489.0, + 1405.0, + 489.0, + 1405.0, + 525.0, + 294.0, + 525.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 521.0, + 1405.0, + 521.0, + 1405.0, + 555.0, + 294.0, + 555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 554.0, + 1403.0, + 554.0, + 1403.0, + 584.0, + 294.0, + 584.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 582.0, + 1402.0, + 582.0, + 1402.0, + 615.0, + 294.0, + 615.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 612.0, + 1405.0, + 612.0, + 1405.0, + 647.0, + 293.0, + 647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 644.0, + 885.0, + 644.0, + 885.0, + 675.0, + 295.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1074.0, + 1404.0, + 1074.0, + 1404.0, + 1116.0, + 295.0, + 1116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1109.0, + 1404.0, + 1109.0, + 1404.0, + 1143.0, + 294.0, + 1143.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1140.0, + 1405.0, + 1140.0, + 1405.0, + 1174.0, + 293.0, + 1174.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1169.0, + 1405.0, + 1169.0, + 1405.0, + 1205.0, + 293.0, + 1205.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1199.0, + 1405.0, + 1199.0, + 1405.0, + 1236.0, + 293.0, + 1236.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1234.0, + 1404.0, + 1234.0, + 1404.0, + 1264.0, + 294.0, + 1264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1262.0, + 1404.0, + 1262.0, + 1404.0, + 1297.0, + 294.0, + 1297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1291.0, + 1405.0, + 1291.0, + 1405.0, + 1329.0, + 293.0, + 1329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1323.0, + 1404.0, + 1323.0, + 1404.0, + 1357.0, + 294.0, + 1357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1350.0, + 1405.0, + 1350.0, + 1405.0, + 1390.0, + 292.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1383.0, + 1403.0, + 1383.0, + 1403.0, + 1419.0, + 292.0, + 1419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1411.0, + 1407.0, + 1411.0, + 1407.0, + 1449.0, + 294.0, + 1449.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1446.0, + 1403.0, + 1446.0, + 1403.0, + 1476.0, + 294.0, + 1476.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1471.0, + 1404.0, + 1471.0, + 1404.0, + 1512.0, + 293.0, + 1512.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1503.0, + 1404.0, + 1503.0, + 1404.0, + 1539.0, + 293.0, + 1539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1535.0, + 393.0, + 1535.0, + 393.0, + 1570.0, + 294.0, + 1570.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 788.0, + 1405.0, + 788.0, + 1405.0, + 821.0, + 298.0, + 821.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 818.0, + 1407.0, + 818.0, + 1407.0, + 855.0, + 295.0, + 855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 847.0, + 1405.0, + 847.0, + 1405.0, + 885.0, + 294.0, + 885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 877.0, + 1407.0, + 877.0, + 1407.0, + 916.0, + 293.0, + 916.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 911.0, + 1404.0, + 911.0, + 1404.0, + 944.0, + 294.0, + 944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 943.0, + 1404.0, + 943.0, + 1404.0, + 975.0, + 295.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 969.0, + 1403.0, + 969.0, + 1403.0, + 1007.0, + 293.0, + 1007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1000.0, + 1405.0, + 1000.0, + 1405.0, + 1037.0, + 294.0, + 1037.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1030.0, + 1173.0, + 1030.0, + 1173.0, + 1070.0, + 293.0, + 1070.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 229.0, + 1404.0, + 229.0, + 1404.0, + 264.0, + 294.0, + 264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 261.0, + 986.0, + 261.0, + 986.0, + 296.0, + 295.0, + 296.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 19, + "width": 1700, + "height": 2200 + } + } +] \ No newline at end of file diff --git a/parse/dev/gUL6zYN4Uaf/images/02deee97d444a824b1a8c0f7ca0a163db9c295bc62b512b6a23eddad3a9d7a0a.jpg b/parse/dev/gUL6zYN4Uaf/images/02deee97d444a824b1a8c0f7ca0a163db9c295bc62b512b6a23eddad3a9d7a0a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..68ccb667f3bb8f72cce8f34eb6b34435e45512b2 --- /dev/null +++ b/parse/dev/gUL6zYN4Uaf/images/02deee97d444a824b1a8c0f7ca0a163db9c295bc62b512b6a23eddad3a9d7a0a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9ee41848dc5d67de92f1b43306b626037fffb7d1ba1529b1a3218dd2694e1998 +size 184966 diff --git a/parse/dev/gUL6zYN4Uaf/images/593e0d4024a8680b497d6699ab53f0ba3319ac04c8ebeec412f54e125d566d91.jpg b/parse/dev/gUL6zYN4Uaf/images/593e0d4024a8680b497d6699ab53f0ba3319ac04c8ebeec412f54e125d566d91.jpg new file mode 100644 index 0000000000000000000000000000000000000000..014d10d9b9fd86e9f64f4de137e8fae49fb830b2 --- /dev/null +++ b/parse/dev/gUL6zYN4Uaf/images/593e0d4024a8680b497d6699ab53f0ba3319ac04c8ebeec412f54e125d566d91.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79a1b75993384ce861303b2fafcfe532f5a5de6d9869cbd36ff0e123d172fcf1 +size 98938 diff --git a/parse/dev/gUL6zYN4Uaf/images/69fecbdad2104df29d8b19d7421f64913a764c6a0ae908c57b09500d5adfb2db.jpg b/parse/dev/gUL6zYN4Uaf/images/69fecbdad2104df29d8b19d7421f64913a764c6a0ae908c57b09500d5adfb2db.jpg new file mode 100644 index 0000000000000000000000000000000000000000..83a682f5e08500844d7713e1be1ee163a7b4227c --- /dev/null +++ b/parse/dev/gUL6zYN4Uaf/images/69fecbdad2104df29d8b19d7421f64913a764c6a0ae908c57b09500d5adfb2db.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:310063a16ca823eab6e156efa99ec6ba9d5fa36f81a35e3d030ac1ea40f54b11 +size 21759 diff --git a/parse/dev/gUL6zYN4Uaf/images/8652aa75413eb04c6ffcc03b86f67680e2146ba66c46a8cbd43317c52e73a82b.jpg b/parse/dev/gUL6zYN4Uaf/images/8652aa75413eb04c6ffcc03b86f67680e2146ba66c46a8cbd43317c52e73a82b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c539115bcd6bc6d06e194d37558a73079a52a3dc --- /dev/null +++ b/parse/dev/gUL6zYN4Uaf/images/8652aa75413eb04c6ffcc03b86f67680e2146ba66c46a8cbd43317c52e73a82b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:729de4a22dc59f0f45d94c1c947c53f4c583dc59b00bb4e46d434e60992a749c +size 46907 diff --git a/parse/dev/gUL6zYN4Uaf/images/8cd26bba684725bbd056708d3bb0ebc1f9e045294e428ee8fedf0e0349287421.jpg b/parse/dev/gUL6zYN4Uaf/images/8cd26bba684725bbd056708d3bb0ebc1f9e045294e428ee8fedf0e0349287421.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4444c1f52bafecefb850f25c438d8f9aea49a0ef --- /dev/null +++ b/parse/dev/gUL6zYN4Uaf/images/8cd26bba684725bbd056708d3bb0ebc1f9e045294e428ee8fedf0e0349287421.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e051fb988df10e2359a48427220b939b5cd22ee8623d12d8efb0e23a6d4cedf5 +size 57145 diff --git a/parse/dev/gUL6zYN4Uaf/images/a4766458a574b146bb88a1266de27caa7767f910aa5be43a59409aa9dbb36521.jpg b/parse/dev/gUL6zYN4Uaf/images/a4766458a574b146bb88a1266de27caa7767f910aa5be43a59409aa9dbb36521.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a9a1555666e141c582fecea6a2d64b6984cdec2a --- /dev/null +++ b/parse/dev/gUL6zYN4Uaf/images/a4766458a574b146bb88a1266de27caa7767f910aa5be43a59409aa9dbb36521.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2024e0a1bc3f611ee3a09d5a9fede21fdc4ca5f3e813b38a692de13c1e4a9db1 +size 60481 diff --git a/parse/dev/gUL6zYN4Uaf/images/b23886d1b8814b578cd3b263c70c77d0ba602f6d67f23df5270ae143a982f994.jpg b/parse/dev/gUL6zYN4Uaf/images/b23886d1b8814b578cd3b263c70c77d0ba602f6d67f23df5270ae143a982f994.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9476da6ec1a9c764a46cdc63738d4e072fa279a3 --- /dev/null +++ b/parse/dev/gUL6zYN4Uaf/images/b23886d1b8814b578cd3b263c70c77d0ba602f6d67f23df5270ae143a982f994.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b4bd8d006771164defaa5524e93e3cc8c397fe5990cc05dd1b50fba5c967ee4c +size 86610 diff --git a/parse/dev/hfjbX1UKNx/images/073fcca357c7e77469d11d2f82ce6b3a265318f8aee808301177c6698d26e7e0.jpg b/parse/dev/hfjbX1UKNx/images/073fcca357c7e77469d11d2f82ce6b3a265318f8aee808301177c6698d26e7e0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b21070096b73ca1f82a84f82f8a8078bda4537c9 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/073fcca357c7e77469d11d2f82ce6b3a265318f8aee808301177c6698d26e7e0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:818a1c36b88bd7f0287ef4b36d97092562c6ca29e4ec225b4e5a2121e66b5f10 +size 11479 diff --git a/parse/dev/hfjbX1UKNx/images/0bca6aba1ca36ebbb0c7124a9bb85a1a81d32d5384cba972725278bd850147cc.jpg b/parse/dev/hfjbX1UKNx/images/0bca6aba1ca36ebbb0c7124a9bb85a1a81d32d5384cba972725278bd850147cc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9464d0e33f9f2b19b8aaf11d1571ec0bbe8f4586 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/0bca6aba1ca36ebbb0c7124a9bb85a1a81d32d5384cba972725278bd850147cc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:64931ed46d811f6f480fa927ccb257108e78f227df2dbd678553c9908daf79bd +size 7010 diff --git a/parse/dev/hfjbX1UKNx/images/0f243534d8382138320dfebe6ae4b9a59b7f129b0805adfc394073e66f98a960.jpg b/parse/dev/hfjbX1UKNx/images/0f243534d8382138320dfebe6ae4b9a59b7f129b0805adfc394073e66f98a960.jpg new file mode 100644 index 0000000000000000000000000000000000000000..754b1ec5c7d79abf3e65f15061efd031992cfcf7 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/0f243534d8382138320dfebe6ae4b9a59b7f129b0805adfc394073e66f98a960.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f2c460344624601935cab765b7a63de45550c3f2ecee1c1eaf990d12027e5fa +size 6920 diff --git a/parse/dev/hfjbX1UKNx/images/16357c2ca103f8683974a8dfd088e2d78dc36e4194f97243e27e498e6d1d8cfd.jpg b/parse/dev/hfjbX1UKNx/images/16357c2ca103f8683974a8dfd088e2d78dc36e4194f97243e27e498e6d1d8cfd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..55513c53fc223988c3a2ea68663b2c69e6dc14e6 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/16357c2ca103f8683974a8dfd088e2d78dc36e4194f97243e27e498e6d1d8cfd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76876dc7495ffd6dbe621f58841b0f1b66fd167ef0a3851bb9821049fb09dec6 +size 18024 diff --git a/parse/dev/hfjbX1UKNx/images/1ae2ae9134ff47774c93d42a3136c130bb28f4877ffba668d80492582a246bf8.jpg b/parse/dev/hfjbX1UKNx/images/1ae2ae9134ff47774c93d42a3136c130bb28f4877ffba668d80492582a246bf8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5f53dbe8c39a0724c7f111f870a5e22b098043c8 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/1ae2ae9134ff47774c93d42a3136c130bb28f4877ffba668d80492582a246bf8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74445f13f379bac16f218b8a93234e4d83917352d6f3a3ef00c7e2317dbcfc50 +size 27762 diff --git a/parse/dev/hfjbX1UKNx/images/22850b01953cc756c9b382836a62d9fc4125b7b98a2fcdce15a45f4e28ba18eb.jpg b/parse/dev/hfjbX1UKNx/images/22850b01953cc756c9b382836a62d9fc4125b7b98a2fcdce15a45f4e28ba18eb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..243a3e2dd500920fb61c0362431fef8fe0c734d8 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/22850b01953cc756c9b382836a62d9fc4125b7b98a2fcdce15a45f4e28ba18eb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc3ca5321b42a8069a0c3952632a57540bfd58a039d188113c2eeb81d0ffecfb +size 20838 diff --git a/parse/dev/hfjbX1UKNx/images/348be3be5132a0fba91c0dba2bfa0565c5fcf55c7ab795b63bbff658b780208c.jpg b/parse/dev/hfjbX1UKNx/images/348be3be5132a0fba91c0dba2bfa0565c5fcf55c7ab795b63bbff658b780208c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..96313a8cec43ba1fa26876d07e2ef6a3f0587ff3 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/348be3be5132a0fba91c0dba2bfa0565c5fcf55c7ab795b63bbff658b780208c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b0413d41e5b9a19d13e3ecc3e617a6b879a145682deb7f50dfe16d14fdcd5c8 +size 27456 diff --git a/parse/dev/hfjbX1UKNx/images/378bd2d666f735dee7ed1fe82defcc1afb6cb8cba85de209da2b1109e494d50b.jpg b/parse/dev/hfjbX1UKNx/images/378bd2d666f735dee7ed1fe82defcc1afb6cb8cba85de209da2b1109e494d50b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bed7f921d07be7cf9fc7ad98e0ecbbb54b18ea0c --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/378bd2d666f735dee7ed1fe82defcc1afb6cb8cba85de209da2b1109e494d50b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02a599033e8566c86f430e78d744ddadb76dae943ad77b2766192735358c1ef9 +size 147346 diff --git a/parse/dev/hfjbX1UKNx/images/3dd820e16a898714cd46097c38e6109ab335e31d831499d645f11262676724ea.jpg b/parse/dev/hfjbX1UKNx/images/3dd820e16a898714cd46097c38e6109ab335e31d831499d645f11262676724ea.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c1958b9b790fe5dd315099b67e59489793ef541c --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/3dd820e16a898714cd46097c38e6109ab335e31d831499d645f11262676724ea.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:abc422dc2dec5157de01e6fbae8a7d3a670cd160c014a186b38b094a9a403e46 +size 138385 diff --git a/parse/dev/hfjbX1UKNx/images/3f599628a9af883be973bd4185680798181d47044763be8c03f5a5d110f44476.jpg b/parse/dev/hfjbX1UKNx/images/3f599628a9af883be973bd4185680798181d47044763be8c03f5a5d110f44476.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9505c828afbbf53c240a1ea9b7b366fd5dfa70ca --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/3f599628a9af883be973bd4185680798181d47044763be8c03f5a5d110f44476.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85e662fbd2a859763bb7da7475ea7894e2b2fb48aaa73364699654cd1b846288 +size 21925 diff --git a/parse/dev/hfjbX1UKNx/images/57c222060c6d94f7e907c0cfdc9f65c1f49631d29a5b83e6252c9fb804f77ae5.jpg b/parse/dev/hfjbX1UKNx/images/57c222060c6d94f7e907c0cfdc9f65c1f49631d29a5b83e6252c9fb804f77ae5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2e9fd4357e40c23a4476cb2df7e51c4fdab0476f --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/57c222060c6d94f7e907c0cfdc9f65c1f49631d29a5b83e6252c9fb804f77ae5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:08d05d53da4456c1cea148376b6d8de1807a850c7928d7bc2484f7f6469ea51b +size 266990 diff --git a/parse/dev/hfjbX1UKNx/images/5ffcfc6c80e4149d05c8d5f7aa3a5208f64cc4fc55b90baa0c261e94ae257b68.jpg b/parse/dev/hfjbX1UKNx/images/5ffcfc6c80e4149d05c8d5f7aa3a5208f64cc4fc55b90baa0c261e94ae257b68.jpg new file mode 100644 index 0000000000000000000000000000000000000000..69171aaafc3bf4172c19a07a9807cf395e15a5fb --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/5ffcfc6c80e4149d05c8d5f7aa3a5208f64cc4fc55b90baa0c261e94ae257b68.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c73fff538a749619af0fcf1cbe80eb23eb1a9ba4ddc278d85698341c57577dd3 +size 93681 diff --git a/parse/dev/hfjbX1UKNx/images/74af58776c660138cdf98a485471f5f73320c4efbe5967bf337e15882cef1163.jpg b/parse/dev/hfjbX1UKNx/images/74af58776c660138cdf98a485471f5f73320c4efbe5967bf337e15882cef1163.jpg new file mode 100644 index 0000000000000000000000000000000000000000..79e9e0e40a57e422e519bebba6be8384f4854c7f --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/74af58776c660138cdf98a485471f5f73320c4efbe5967bf337e15882cef1163.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b415fbe9d7e3c5baf43adbd5d664790cd4aec764cb7aaf4c17bff5bdb6f86887 +size 152485 diff --git a/parse/dev/hfjbX1UKNx/images/79a0f52ae8f6533ff7f24d374b52e2988c8e0b086e59b2ac9bfbd4b236c92657.jpg b/parse/dev/hfjbX1UKNx/images/79a0f52ae8f6533ff7f24d374b52e2988c8e0b086e59b2ac9bfbd4b236c92657.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b9a68e227b61bf9f119e514e7f5a6966fdecd889 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/79a0f52ae8f6533ff7f24d374b52e2988c8e0b086e59b2ac9bfbd4b236c92657.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef23a3aab3942c262e85f04528cc1e756ae1a465d80f7a6b64e600e5339da4a3 +size 102506 diff --git a/parse/dev/hfjbX1UKNx/images/7aefa5075678a2ee7ce63ebed88504607d7fea05fe886cc1a21a48147f48713d.jpg b/parse/dev/hfjbX1UKNx/images/7aefa5075678a2ee7ce63ebed88504607d7fea05fe886cc1a21a48147f48713d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1f066fb753cf40dadb6a07373ad616003e020332 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/7aefa5075678a2ee7ce63ebed88504607d7fea05fe886cc1a21a48147f48713d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f31ea25aa65a0b663ab955e3b61ec089a4ecf14af723bba77763ff49e972c04 +size 8093 diff --git a/parse/dev/hfjbX1UKNx/images/7cace8220f8c6e96e2d7f9482ffc55a2ccb340ae00506d38d3bd57687b5e8327.jpg b/parse/dev/hfjbX1UKNx/images/7cace8220f8c6e96e2d7f9482ffc55a2ccb340ae00506d38d3bd57687b5e8327.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b8817555536abd511d5767dbbac7bc5685e004e9 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/7cace8220f8c6e96e2d7f9482ffc55a2ccb340ae00506d38d3bd57687b5e8327.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fbb7835096d19812fe485fd6ff58e9e8a3b01ad0154a6269a849ebdc813a949b +size 57770 diff --git a/parse/dev/hfjbX1UKNx/images/853a1ca9346ee640065f64c3b8d0231630f3dcb428a7f4aa505a7881dccbe5db.jpg b/parse/dev/hfjbX1UKNx/images/853a1ca9346ee640065f64c3b8d0231630f3dcb428a7f4aa505a7881dccbe5db.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1e6878f7a54dad91383eea8b83fa2f55b02342f1 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/853a1ca9346ee640065f64c3b8d0231630f3dcb428a7f4aa505a7881dccbe5db.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c38a67f6122fa1ff8b1b87fa80f2128bc9b9d5bf6a5e1453dea2d15e3c605c40 +size 23425 diff --git a/parse/dev/hfjbX1UKNx/images/9591dbe887cc010016deade135d0463bbbf67a59900b182cbf98c6d317787c1b.jpg b/parse/dev/hfjbX1UKNx/images/9591dbe887cc010016deade135d0463bbbf67a59900b182cbf98c6d317787c1b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ac21450fea9e2f33c92c49eebc195d7aa3499008 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/9591dbe887cc010016deade135d0463bbbf67a59900b182cbf98c6d317787c1b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8405a0d487748bfd0deeeed421a5a59eca9c166c49f367afb462c5c4bf93684b +size 7273 diff --git a/parse/dev/hfjbX1UKNx/images/a7a225bb641204d4940938736547008a4c090a7b7402942eac6d13a0370ddc0c.jpg b/parse/dev/hfjbX1UKNx/images/a7a225bb641204d4940938736547008a4c090a7b7402942eac6d13a0370ddc0c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ab7bbfedaf766d6021f130d131c411ce52ab9bb9 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/a7a225bb641204d4940938736547008a4c090a7b7402942eac6d13a0370ddc0c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fbd0c0ae3dfc97500454efc0ee801f0b4cbb5a0e3b76734b43c0b36402f02896 +size 60448 diff --git a/parse/dev/hfjbX1UKNx/images/b379a38114d11b8f123a1286df260d3c71eb0eacf64c42d9ee6b2ebf3de1127b.jpg b/parse/dev/hfjbX1UKNx/images/b379a38114d11b8f123a1286df260d3c71eb0eacf64c42d9ee6b2ebf3de1127b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..82c3ad177b7d5540ad9a74a3fe6916f208e6cdb6 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/b379a38114d11b8f123a1286df260d3c71eb0eacf64c42d9ee6b2ebf3de1127b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3784551cf9d2db79400459a5443e0fd323bac13355d291d1d2bc201313e92faa +size 3001 diff --git a/parse/dev/hfjbX1UKNx/images/b64bd0f94f453c9dc766dc20830b15ef390dff0c17347d156c8c756bb56e4d8c.jpg b/parse/dev/hfjbX1UKNx/images/b64bd0f94f453c9dc766dc20830b15ef390dff0c17347d156c8c756bb56e4d8c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f8f2663b6ff655b54054f1fab5bf54cb9a54ca95 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/b64bd0f94f453c9dc766dc20830b15ef390dff0c17347d156c8c756bb56e4d8c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9f26a7f671a75a8cd1ccda15f490e7458e2d0ae534bacf55afa7f51445461f2 +size 3082 diff --git a/parse/dev/hfjbX1UKNx/images/b96f31c372f7662decf86a5dec13666b47e30f39324299299ff7961e64fc55fd.jpg b/parse/dev/hfjbX1UKNx/images/b96f31c372f7662decf86a5dec13666b47e30f39324299299ff7961e64fc55fd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..96f7856eac9e1078d2d3f0bc3f91cb93e9d0c2aa --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/b96f31c372f7662decf86a5dec13666b47e30f39324299299ff7961e64fc55fd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e4085b621f11d3e39e13a29404e6bb48ef4ba53d31b8879bacfe1ca6077f026 +size 2838 diff --git a/parse/dev/hfjbX1UKNx/images/ba2e8be3f9bf42e8c4addfd2a3d0db64c9acaa55274beee8f00e4ba281489f9d.jpg b/parse/dev/hfjbX1UKNx/images/ba2e8be3f9bf42e8c4addfd2a3d0db64c9acaa55274beee8f00e4ba281489f9d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d2f159f7b5943eb9684f549e1a4c4b1b7e2ea6c3 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/ba2e8be3f9bf42e8c4addfd2a3d0db64c9acaa55274beee8f00e4ba281489f9d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ccd6c966ec51af397d409c399097e51286b535fd309fff522417b4d5804ab73 +size 83833 diff --git a/parse/dev/hfjbX1UKNx/images/c0f3c692f24bb99058ccfe8b7392864dc7d19a757e1a94d847cc7323e7b13f98.jpg b/parse/dev/hfjbX1UKNx/images/c0f3c692f24bb99058ccfe8b7392864dc7d19a757e1a94d847cc7323e7b13f98.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8c1af3bfd13cc612b633602ee96619f4440d29dd --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/c0f3c692f24bb99058ccfe8b7392864dc7d19a757e1a94d847cc7323e7b13f98.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35c4ddef5ec774baad97e49f094cb419890fb8e27252c9c1db97398fac5f4e97 +size 34226 diff --git a/parse/dev/hfjbX1UKNx/images/c5af9719714ac28d03e5d7f85ff2375d010ca6ad128262b820b75d39189feb67.jpg b/parse/dev/hfjbX1UKNx/images/c5af9719714ac28d03e5d7f85ff2375d010ca6ad128262b820b75d39189feb67.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fd371ac9e7d9e96f8e2bb655ea2fc4f772ad5e01 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/c5af9719714ac28d03e5d7f85ff2375d010ca6ad128262b820b75d39189feb67.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad03a1488f514e6ec8c1445f687c25886dd0982a0c4f3ddb2a03175b7f8c7551 +size 78222 diff --git a/parse/dev/hfjbX1UKNx/images/c648b4d1bf76c86f33d9fdc47b25846a8d5dcef65826954ffdd3bd910cef4ac0.jpg b/parse/dev/hfjbX1UKNx/images/c648b4d1bf76c86f33d9fdc47b25846a8d5dcef65826954ffdd3bd910cef4ac0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..23f3777af521bcedf7ecc881599fd373b9813bb4 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/c648b4d1bf76c86f33d9fdc47b25846a8d5dcef65826954ffdd3bd910cef4ac0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e03da899a2a570ba22c0ffc8ef2a353bf7a9481b775060b60557034402fca1fb +size 5588 diff --git a/parse/dev/hfjbX1UKNx/images/c7f418a32c5eb4404718034733328a8ec3d9aa4c88f70f1f334cfbe47a425dc8.jpg b/parse/dev/hfjbX1UKNx/images/c7f418a32c5eb4404718034733328a8ec3d9aa4c88f70f1f334cfbe47a425dc8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9af6636f84d9458fd66a26d5ef13d5f4f7c35686 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/c7f418a32c5eb4404718034733328a8ec3d9aa4c88f70f1f334cfbe47a425dc8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c0e5f583be44dc6cd291ba50ccab3f7c9c33504ca24d63489896b9e860262393 +size 131039 diff --git a/parse/dev/hfjbX1UKNx/images/c7f4cd09ace9549f8a9642691118c51d19ef63ddcd8e74b9c114e11a5bd110b7.jpg b/parse/dev/hfjbX1UKNx/images/c7f4cd09ace9549f8a9642691118c51d19ef63ddcd8e74b9c114e11a5bd110b7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c6e44c1df5b2b798d10ecf21488b884cea62314d --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/c7f4cd09ace9549f8a9642691118c51d19ef63ddcd8e74b9c114e11a5bd110b7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:242be9f736c6d062555619a4b1d3f6f30304a8665e00023c007dbfb5b4960dbe +size 13357 diff --git a/parse/dev/hfjbX1UKNx/images/cf7b52087523ecae395c10843e31f1b8b6f1c60ae8376b17b299f8a6da736c63.jpg b/parse/dev/hfjbX1UKNx/images/cf7b52087523ecae395c10843e31f1b8b6f1c60ae8376b17b299f8a6da736c63.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4f854cd8c57f8eee9ff24e7a11c097cb29f916d1 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/cf7b52087523ecae395c10843e31f1b8b6f1c60ae8376b17b299f8a6da736c63.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3c83e16f2849b80ddd55a90fd787a275ebab12b29a94d1b644d8bebb7ffadba3 +size 28013 diff --git a/parse/dev/hfjbX1UKNx/images/d7937d406a392481f61d84f5fc832df0a60dafc72e3e33010e52da2f76e07e33.jpg b/parse/dev/hfjbX1UKNx/images/d7937d406a392481f61d84f5fc832df0a60dafc72e3e33010e52da2f76e07e33.jpg new file mode 100644 index 0000000000000000000000000000000000000000..76c08ef65c7d491294348d969b013d0d82e60b17 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/d7937d406a392481f61d84f5fc832df0a60dafc72e3e33010e52da2f76e07e33.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b7fc43e518d45f96d376b1c77b4f9184b58466d36eceb93e6a52299caf5c701a +size 63298 diff --git a/parse/dev/hfjbX1UKNx/images/db51d54976b7f8d9430c23dc4be404d0f163d21b659567e32d0783d1dc9de498.jpg b/parse/dev/hfjbX1UKNx/images/db51d54976b7f8d9430c23dc4be404d0f163d21b659567e32d0783d1dc9de498.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8ad0b3b471dab1f648d39180c0bd2401f58bd532 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/db51d54976b7f8d9430c23dc4be404d0f163d21b659567e32d0783d1dc9de498.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:529ea4125fca0fa508f30b6c8f31b60418786f5583cda16745e0d6634ee10e1f +size 7702 diff --git a/parse/dev/hfjbX1UKNx/images/dcf71f0e43760949e40aca281e0e433f0625d71bd135dcbec193c92ba0385d31.jpg b/parse/dev/hfjbX1UKNx/images/dcf71f0e43760949e40aca281e0e433f0625d71bd135dcbec193c92ba0385d31.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4adaab00b38b1aa12a4463498031f4458a3aec42 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/dcf71f0e43760949e40aca281e0e433f0625d71bd135dcbec193c92ba0385d31.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c42b5a5bec786b0b46a0f170b49a2fbe6db93e3d9ec6964f09249490ea57ca5 +size 73959 diff --git a/parse/dev/hfjbX1UKNx/images/dd58ac1307f3a731726afd7a50457bad6baf092b1113030b83a973507cab67b5.jpg b/parse/dev/hfjbX1UKNx/images/dd58ac1307f3a731726afd7a50457bad6baf092b1113030b83a973507cab67b5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c0a3ea62ff5dd97c9f543bed545e81285249f0ac --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/dd58ac1307f3a731726afd7a50457bad6baf092b1113030b83a973507cab67b5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:69d9b2105ffa237a0e271933334430415941d00802cc744451518f9078c1b02a +size 76465 diff --git a/parse/dev/hfjbX1UKNx/images/e2d2d850b3e5db6de5a79a50590c3b04b24f4b03e2dac88107f85d8879dea0f6.jpg b/parse/dev/hfjbX1UKNx/images/e2d2d850b3e5db6de5a79a50590c3b04b24f4b03e2dac88107f85d8879dea0f6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b2b89aedd8fb3b855f2ac9576bdde372c9e44c23 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/e2d2d850b3e5db6de5a79a50590c3b04b24f4b03e2dac88107f85d8879dea0f6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2cb9da75a313bd7f7352022d013f33105fdf9a8e0430509fb585660b50e75b9e +size 20923 diff --git a/parse/dev/hfjbX1UKNx/images/e33210a6cd961747ac565bb9a39ba41dc9cf739fc965bbcf1a36112afebc47ee.jpg b/parse/dev/hfjbX1UKNx/images/e33210a6cd961747ac565bb9a39ba41dc9cf739fc965bbcf1a36112afebc47ee.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dd965564dc322dca369018a78f362f9edbef35d9 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/e33210a6cd961747ac565bb9a39ba41dc9cf739fc965bbcf1a36112afebc47ee.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:165ff924597ade806d6f63d84c26b913b6396bc810e068055e16147a50daf354 +size 25818 diff --git a/parse/dev/hfjbX1UKNx/images/e6479eeb466d314a87fb51caae1fd194e7396931976ae14c6eb9771bf420f50b.jpg b/parse/dev/hfjbX1UKNx/images/e6479eeb466d314a87fb51caae1fd194e7396931976ae14c6eb9771bf420f50b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d4d200038591107b0526fa27c921304413591b23 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/e6479eeb466d314a87fb51caae1fd194e7396931976ae14c6eb9771bf420f50b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f1c9e073b2b0055887a4d98e960d60cc212d4f2e1c1ee4762455a73dd4339bd +size 38615 diff --git a/parse/dev/hfjbX1UKNx/images/e9e88cc07a578e6a373d54707532061f893fa988ea1e74e9671600c943562552.jpg b/parse/dev/hfjbX1UKNx/images/e9e88cc07a578e6a373d54707532061f893fa988ea1e74e9671600c943562552.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1dafbde015ca69a22fe95e52aca9e91dba3f9b12 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/e9e88cc07a578e6a373d54707532061f893fa988ea1e74e9671600c943562552.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f864b65398fa97888f6441fce43cdeb56ac11316bd33e351054dcb6c64c3f44b +size 55967 diff --git a/parse/dev/hfjbX1UKNx/images/f6b594ffe8a902f2ffe713ae764a2b08e51fcd98bf5ed21936c739bf175d5c09.jpg b/parse/dev/hfjbX1UKNx/images/f6b594ffe8a902f2ffe713ae764a2b08e51fcd98bf5ed21936c739bf175d5c09.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a7c1313a73cfaef0953cbb5f39cf8f90d8e6cbea --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/f6b594ffe8a902f2ffe713ae764a2b08e51fcd98bf5ed21936c739bf175d5c09.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27ae6209aae65d2c0d4259cd6619fd495f24db8ca8fb9008f4b7fb274fac0998 +size 110217 diff --git a/parse/dev/hfjbX1UKNx/images/ffbc180026cb43647ae3567235e03e31df62f3668830d7ebfbc46a688e3ca531.jpg b/parse/dev/hfjbX1UKNx/images/ffbc180026cb43647ae3567235e03e31df62f3668830d7ebfbc46a688e3ca531.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f0ed689f0098bf13d3733c2b49f290705359aba8 --- /dev/null +++ b/parse/dev/hfjbX1UKNx/images/ffbc180026cb43647ae3567235e03e31df62f3668830d7ebfbc46a688e3ca531.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8de23f0f247e455b334cd69dbe562c42f89021669385bad0efd1ee065d839fa2 +size 14563 diff --git a/parse/dev/jDIlzSU8wJ/jDIlzSU8wJ_middle.json b/parse/dev/jDIlzSU8wJ/jDIlzSU8wJ_middle.json new file mode 100644 index 0000000000000000000000000000000000000000..7be40d3d09bc0207de7cac569ed85f8c1b3c8ed4 --- /dev/null +++ b/parse/dev/jDIlzSU8wJ/jDIlzSU8wJ_middle.json @@ -0,0 +1,54803 @@ +{ + "pdf_info": [ + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 113, + 97, + 499, + 137 + ], + "lines": [ + { + "bbox": [ + 111, + 97, + 500, + 118 + ], + "spans": [ + { + "bbox": [ + 111, + 97, + 500, + 118 + ], + "score": 1.0, + "content": "The Surprising Effectiveness of Diffusion Models for", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 130, + 117, + 480, + 139 + ], + "spans": [ + { + "bbox": [ + 130, + 117, + 480, + 139 + ], + "score": 1.0, + "content": "Optical Flow and Monocular Depth Estimation", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 162, + 178, + 451, + 205 + ], + "lines": [ + { + "bbox": [ + 161, + 178, + 452, + 194 + ], + "spans": [ + { + "bbox": [ + 161, + 178, + 452, + 194 + ], + "score": 1.0, + "content": "Saurabh Saxena, Charles Herrmann, Junhwa Hur, Abhishek Kar,", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 198, + 192, + 416, + 206 + ], + "spans": [ + { + "bbox": [ + 198, + 192, + 416, + 206 + ], + "score": 1.0, + "content": "Mohammad Norouzi, Deqing Sun, David J. Fleet∗", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 119, + 210, + 490, + 221 + ], + "lines": [ + { + "bbox": [ + 118, + 210, + 492, + 223 + ], + "spans": [ + { + "bbox": [ + 118, + 210, + 492, + 223 + ], + "score": 1.0, + "content": "{srbs,irwinherrmann,junhwahur,abhiskar,deqingsun,davidfleet}@google.com", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 224, + 226, + 388, + 237 + ], + "lines": [ + { + "bbox": [ + 222, + 225, + 389, + 240 + ], + "spans": [ + { + "bbox": [ + 222, + 225, + 389, + 240 + ], + "score": 1.0, + "content": "Google DeepMind and Google Research", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "title", + "bbox": [ + 283, + 266, + 328, + 279 + ], + "lines": [ + { + "bbox": [ + 282, + 266, + 330, + 280 + ], + "spans": [ + { + "bbox": [ + 282, + 266, + 330, + 280 + ], + "score": 1.0, + "content": "Abstract", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 143, + 283, + 469, + 468 + ], + "lines": [ + { + "bbox": [ + 141, + 282, + 469, + 295 + ], + "spans": [ + { + "bbox": [ + 141, + 282, + 469, + 295 + ], + "score": 1.0, + "content": "Denoising diffusion probabilistic models have transformed image generation with", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 141, + 293, + 469, + 307 + ], + "spans": [ + { + "bbox": [ + 141, + 293, + 469, + 307 + ], + "score": 1.0, + "content": "their impressive fidelity and diversity. We show that they also excel in estimating", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 141, + 304, + 470, + 318 + ], + "spans": [ + { + "bbox": [ + 141, + 304, + 470, + 318 + ], + "score": 1.0, + "content": "optical flow and monocular depth, surprisingly, without task-specific architectures", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 141, + 315, + 470, + 328 + ], + "spans": [ + { + "bbox": [ + 141, + 315, + 470, + 328 + ], + "score": 1.0, + "content": "and loss functions that are predominant for these tasks. Compared to the point", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 326, + 470, + 338 + ], + "spans": [ + { + "bbox": [ + 141, + 326, + 470, + 338 + ], + "score": 1.0, + "content": "estimates of conventional regression-based methods, diffusion models also enable", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 337, + 470, + 350 + ], + "spans": [ + { + "bbox": [ + 141, + 337, + 470, + 350 + ], + "score": 1.0, + "content": "Monte Carlo inference, e.g., capturing uncertainty and ambiguity in flow and depth.", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 348, + 470, + 360 + ], + "spans": [ + { + "bbox": [ + 141, + 348, + 470, + 360 + ], + "score": 1.0, + "content": "With self-supervised pre-training, the combined use of synthetic and real data for", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 359, + 469, + 372 + ], + "spans": [ + { + "bbox": [ + 141, + 359, + 469, + 372 + ], + "score": 1.0, + "content": "supervised training, and technical innovations (infilling and step-unrolled denoising", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 369, + 469, + 383 + ], + "spans": [ + { + "bbox": [ + 141, + 369, + 469, + 383 + ], + "score": 1.0, + "content": "diffusion training) to handle noisy-incomplete training data, and a simple form", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 380, + 469, + 393 + ], + "spans": [ + { + "bbox": [ + 141, + 380, + 469, + 393 + ], + "score": 1.0, + "content": "of coarse-to-fine refinement, one can train state-of-the-art diffusion models for", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 142, + 392, + 469, + 403 + ], + "spans": [ + { + "bbox": [ + 142, + 392, + 469, + 403 + ], + "score": 1.0, + "content": "depth and optical flow estimation. Extensive experiments focus on quantitative", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 402, + 470, + 415 + ], + "spans": [ + { + "bbox": [ + 141, + 402, + 470, + 415 + ], + "score": 1.0, + "content": "performance against benchmarks, ablations, and the model’s ability to capture", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 414, + 470, + 425 + ], + "spans": [ + { + "bbox": [ + 141, + 414, + 470, + 425 + ], + "score": 1.0, + "content": "uncertainty and multimodality, and impute missing values. Our model, DDVM", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 141, + 425, + 470, + 437 + ], + "spans": [ + { + "bbox": [ + 141, + 425, + 470, + 437 + ], + "score": 1.0, + "content": "(Denoising Diffusion Vision Model), obtains a state-of-the-art relative depth error", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 142, + 435, + 469, + 447 + ], + "spans": [ + { + "bbox": [ + 142, + 436, + 414, + 447 + ], + "score": 1.0, + "content": "of 0.074 on the indoor NYU benchmark and an Fl-all outlier rate of", + "type": "text" + }, + { + "bbox": [ + 414, + 435, + 441, + 446 + ], + "score": 0.87, + "content": "3 . 2 6 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 441, + 436, + 469, + 447 + ], + "score": 1.0, + "content": "on the", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 141, + 446, + 470, + 458 + ], + "spans": [ + { + "bbox": [ + 141, + 446, + 295, + 458 + ], + "score": 1.0, + "content": "KITTI optical flow benchmark, about", + "type": "text" + }, + { + "bbox": [ + 295, + 446, + 315, + 457 + ], + "score": 0.88, + "content": "2 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 446, + 470, + 458 + ], + "score": 1.0, + "content": "better than the best published method.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 141, + 457, + 328, + 469 + ], + "spans": [ + { + "bbox": [ + 141, + 457, + 328, + 469 + ], + "score": 1.0, + "content": "For an overview see diffusion-vision.github.io", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 15 + }, + { + "type": "title", + "bbox": [ + 107, + 488, + 190, + 501 + ], + "lines": [ + { + "bbox": [ + 105, + 487, + 192, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 192, + 504 + ], + "score": 1.0, + "content": "1 Introduction", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 107, + 510, + 505, + 609 + ], + "lines": [ + { + "bbox": [ + 105, + 509, + 507, + 523 + ], + "spans": [ + { + "bbox": [ + 105, + 509, + 507, + 523 + ], + "score": 1.0, + "content": "Diffusion models have emerged as powerful generative models for high fidelity image synthesis,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "score": 1.0, + "content": "capturing rich knowledge about the visual world [21, 48, 55, 62]. However, at first glance, it is", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 532, + 505, + 544 + ], + "spans": [ + { + "bbox": [ + 106, + 532, + 505, + 544 + ], + "score": 1.0, + "content": "unclear whether these models can be as effective on many classical computer vision tasks. For", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "score": 1.0, + "content": "example, consider two dense vision estimation tasks, namely, optical flow, which estimates frame-to-", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 554, + 506, + 567 + ], + "spans": [ + { + "bbox": [ + 106, + 554, + 506, + 567 + ], + "score": 1.0, + "content": "frame correspondences, and monocular depth perception, which makes depth predictions based on a", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "score": 1.0, + "content": "single image. Both tasks are usually treated as regression problems and addressed with specialized", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 575, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 575, + 505, + 588 + ], + "score": 1.0, + "content": "architectures and task-specific loss functions, e.g., cost volumes, feature warps, or suitable losses", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 586, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 505, + 600 + ], + "score": 1.0, + "content": "for depth. Without these specialized components or the regression framework, general generative", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 597, + 483, + 610 + ], + "spans": [ + { + "bbox": [ + 106, + 597, + 483, + 610 + ], + "score": 1.0, + "content": "techniques may be ill-equipped and vulnerable to both generalization and performance issues.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 614, + 505, + 669 + ], + "lines": [ + { + "bbox": [ + 106, + 614, + 505, + 626 + ], + "spans": [ + { + "bbox": [ + 106, + 614, + 505, + 626 + ], + "score": 1.0, + "content": "In this paper, we show that these concerns, while valid, can be addressed and that, surprisingly, a", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "score": 1.0, + "content": "generic, conventional diffusion model for image to image translation works impressively well on both", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 636, + 505, + 648 + ], + "spans": [ + { + "bbox": [ + 106, + 636, + 505, + 648 + ], + "score": 1.0, + "content": "tasks, often outperforming the state of the art. In addition, diffusion models provide valuable benefits", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 646, + 505, + 659 + ], + "spans": [ + { + "bbox": [ + 105, + 646, + 505, + 659 + ], + "score": 1.0, + "content": "over networks trained with regression; in particular, diffusion allows for approximate inference with", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 657, + 436, + 670 + ], + "spans": [ + { + "bbox": [ + 106, + 657, + 436, + 670 + ], + "score": 1.0, + "content": "multi-modal distributions, capturing uncertainty and ambiguity (e.g. see Figure 1).", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 36 + }, + { + "type": "text", + "bbox": [ + 109, + 673, + 502, + 696 + ], + "lines": [ + { + "bbox": [ + 106, + 672, + 505, + 687 + ], + "spans": [ + { + "bbox": [ + 106, + 672, + 505, + 687 + ], + "score": 1.0, + "content": "One key barrier to training useful diffusion models for monocular depth and optical flow inference", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 685, + 504, + 697 + ], + "spans": [ + { + "bbox": [ + 106, + 685, + 504, + 697 + ], + "score": 1.0, + "content": "concerns the amount and quality of available training data. Given the limited availability of labelled", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5 + } + ], + "page_idx": 0, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 118, + 703, + 393, + 713 + ], + "lines": [ + { + "bbox": [ + 119, + 702, + 394, + 714 + ], + "spans": [ + { + "bbox": [ + 119, + 702, + 394, + 714 + ], + "score": 1.0, + "content": "∗DF is also affiliated with the University of Toronto and the Vector Institute.", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 107, + 731, + 385, + 742 + ], + "lines": [ + { + "bbox": [ + 105, + 730, + 386, + 743 + ], + "spans": [ + { + "bbox": [ + 105, + 730, + 386, + 743 + ], + "score": 1.0, + "content": "37th Conference on Neural Information Processing Systems (NeurIPS 2023).", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 113, + 97, + 499, + 137 + ], + "lines": [ + { + "bbox": [ + 111, + 97, + 500, + 118 + ], + "spans": [ + { + "bbox": [ + 111, + 97, + 500, + 118 + ], + "score": 1.0, + "content": "The Surprising Effectiveness of Diffusion Models for", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 130, + 117, + 480, + 139 + ], + "spans": [ + { + "bbox": [ + 130, + 117, + 480, + 139 + ], + "score": 1.0, + "content": "Optical Flow and Monocular Depth Estimation", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 162, + 178, + 451, + 205 + ], + "lines": [ + { + "bbox": [ + 161, + 178, + 452, + 194 + ], + "spans": [ + { + "bbox": [ + 161, + 178, + 452, + 194 + ], + "score": 1.0, + "content": "Saurabh Saxena, Charles Herrmann, Junhwa Hur, Abhishek Kar,", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 198, + 192, + 416, + 206 + ], + "spans": [ + { + "bbox": [ + 198, + 192, + 416, + 206 + ], + "score": 1.0, + "content": "Mohammad Norouzi, Deqing Sun, David J. Fleet∗", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5, + "bbox_fs": [ + 161, + 178, + 452, + 206 + ] + }, + { + "type": "text", + "bbox": [ + 119, + 210, + 490, + 221 + ], + "lines": [ + { + "bbox": [ + 118, + 210, + 492, + 223 + ], + "spans": [ + { + "bbox": [ + 118, + 210, + 492, + 223 + ], + "score": 1.0, + "content": "{srbs,irwinherrmann,junhwahur,abhiskar,deqingsun,davidfleet}@google.com", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4, + "bbox_fs": [ + 118, + 210, + 492, + 223 + ] + }, + { + "type": "text", + "bbox": [ + 224, + 226, + 388, + 237 + ], + "lines": [ + { + "bbox": [ + 222, + 225, + 389, + 240 + ], + "spans": [ + { + "bbox": [ + 222, + 225, + 389, + 240 + ], + "score": 1.0, + "content": "Google DeepMind and Google Research", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5, + "bbox_fs": [ + 222, + 225, + 389, + 240 + ] + }, + { + "type": "title", + "bbox": [ + 283, + 266, + 328, + 279 + ], + "lines": [ + { + "bbox": [ + 282, + 266, + 330, + 280 + ], + "spans": [ + { + "bbox": [ + 282, + 266, + 330, + 280 + ], + "score": 1.0, + "content": "Abstract", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 143, + 283, + 469, + 468 + ], + "lines": [ + { + "bbox": [ + 141, + 282, + 469, + 295 + ], + "spans": [ + { + "bbox": [ + 141, + 282, + 469, + 295 + ], + "score": 1.0, + "content": "Denoising diffusion probabilistic models have transformed image generation with", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 141, + 293, + 469, + 307 + ], + "spans": [ + { + "bbox": [ + 141, + 293, + 469, + 307 + ], + "score": 1.0, + "content": "their impressive fidelity and diversity. We show that they also excel in estimating", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 141, + 304, + 470, + 318 + ], + "spans": [ + { + "bbox": [ + 141, + 304, + 470, + 318 + ], + "score": 1.0, + "content": "optical flow and monocular depth, surprisingly, without task-specific architectures", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 141, + 315, + 470, + 328 + ], + "spans": [ + { + "bbox": [ + 141, + 315, + 470, + 328 + ], + "score": 1.0, + "content": "and loss functions that are predominant for these tasks. Compared to the point", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 326, + 470, + 338 + ], + "spans": [ + { + "bbox": [ + 141, + 326, + 470, + 338 + ], + "score": 1.0, + "content": "estimates of conventional regression-based methods, diffusion models also enable", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 337, + 470, + 350 + ], + "spans": [ + { + "bbox": [ + 141, + 337, + 470, + 350 + ], + "score": 1.0, + "content": "Monte Carlo inference, e.g., capturing uncertainty and ambiguity in flow and depth.", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 348, + 470, + 360 + ], + "spans": [ + { + "bbox": [ + 141, + 348, + 470, + 360 + ], + "score": 1.0, + "content": "With self-supervised pre-training, the combined use of synthetic and real data for", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 359, + 469, + 372 + ], + "spans": [ + { + "bbox": [ + 141, + 359, + 469, + 372 + ], + "score": 1.0, + "content": "supervised training, and technical innovations (infilling and step-unrolled denoising", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 369, + 469, + 383 + ], + "spans": [ + { + "bbox": [ + 141, + 369, + 469, + 383 + ], + "score": 1.0, + "content": "diffusion training) to handle noisy-incomplete training data, and a simple form", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 380, + 469, + 393 + ], + "spans": [ + { + "bbox": [ + 141, + 380, + 469, + 393 + ], + "score": 1.0, + "content": "of coarse-to-fine refinement, one can train state-of-the-art diffusion models for", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 142, + 392, + 469, + 403 + ], + "spans": [ + { + "bbox": [ + 142, + 392, + 469, + 403 + ], + "score": 1.0, + "content": "depth and optical flow estimation. Extensive experiments focus on quantitative", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 402, + 470, + 415 + ], + "spans": [ + { + "bbox": [ + 141, + 402, + 470, + 415 + ], + "score": 1.0, + "content": "performance against benchmarks, ablations, and the model’s ability to capture", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 414, + 470, + 425 + ], + "spans": [ + { + "bbox": [ + 141, + 414, + 470, + 425 + ], + "score": 1.0, + "content": "uncertainty and multimodality, and impute missing values. Our model, DDVM", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 141, + 425, + 470, + 437 + ], + "spans": [ + { + "bbox": [ + 141, + 425, + 470, + 437 + ], + "score": 1.0, + "content": "(Denoising Diffusion Vision Model), obtains a state-of-the-art relative depth error", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 142, + 435, + 469, + 447 + ], + "spans": [ + { + "bbox": [ + 142, + 436, + 414, + 447 + ], + "score": 1.0, + "content": "of 0.074 on the indoor NYU benchmark and an Fl-all outlier rate of", + "type": "text" + }, + { + "bbox": [ + 414, + 435, + 441, + 446 + ], + "score": 0.87, + "content": "3 . 2 6 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 441, + 436, + 469, + 447 + ], + "score": 1.0, + "content": "on the", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 141, + 446, + 470, + 458 + ], + "spans": [ + { + "bbox": [ + 141, + 446, + 295, + 458 + ], + "score": 1.0, + "content": "KITTI optical flow benchmark, about", + "type": "text" + }, + { + "bbox": [ + 295, + 446, + 315, + 457 + ], + "score": 0.88, + "content": "2 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 446, + 470, + 458 + ], + "score": 1.0, + "content": "better than the best published method.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 141, + 457, + 328, + 469 + ], + "spans": [ + { + "bbox": [ + 141, + 457, + 328, + 469 + ], + "score": 1.0, + "content": "For an overview see diffusion-vision.github.io", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 15, + "bbox_fs": [ + 141, + 282, + 470, + 469 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 488, + 190, + 501 + ], + "lines": [ + { + "bbox": [ + 105, + 487, + 192, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 192, + 504 + ], + "score": 1.0, + "content": "1 Introduction", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 107, + 510, + 505, + 609 + ], + "lines": [ + { + "bbox": [ + 105, + 509, + 507, + 523 + ], + "spans": [ + { + "bbox": [ + 105, + 509, + 507, + 523 + ], + "score": 1.0, + "content": "Diffusion models have emerged as powerful generative models for high fidelity image synthesis,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 521, + 506, + 534 + ], + "score": 1.0, + "content": "capturing rich knowledge about the visual world [21, 48, 55, 62]. However, at first glance, it is", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 532, + 505, + 544 + ], + "spans": [ + { + "bbox": [ + 106, + 532, + 505, + 544 + ], + "score": 1.0, + "content": "unclear whether these models can be as effective on many classical computer vision tasks. For", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "score": 1.0, + "content": "example, consider two dense vision estimation tasks, namely, optical flow, which estimates frame-to-", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 554, + 506, + 567 + ], + "spans": [ + { + "bbox": [ + 106, + 554, + 506, + 567 + ], + "score": 1.0, + "content": "frame correspondences, and monocular depth perception, which makes depth predictions based on a", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "score": 1.0, + "content": "single image. Both tasks are usually treated as regression problems and addressed with specialized", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 575, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 575, + 505, + 588 + ], + "score": 1.0, + "content": "architectures and task-specific loss functions, e.g., cost volumes, feature warps, or suitable losses", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 586, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 505, + 600 + ], + "score": 1.0, + "content": "for depth. Without these specialized components or the regression framework, general generative", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 597, + 483, + 610 + ], + "spans": [ + { + "bbox": [ + 106, + 597, + 483, + 610 + ], + "score": 1.0, + "content": "techniques may be ill-equipped and vulnerable to both generalization and performance issues.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 29, + "bbox_fs": [ + 105, + 509, + 507, + 610 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 614, + 505, + 669 + ], + "lines": [ + { + "bbox": [ + 106, + 614, + 505, + 626 + ], + "spans": [ + { + "bbox": [ + 106, + 614, + 505, + 626 + ], + "score": 1.0, + "content": "In this paper, we show that these concerns, while valid, can be addressed and that, surprisingly, a", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "score": 1.0, + "content": "generic, conventional diffusion model for image to image translation works impressively well on both", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 636, + 505, + 648 + ], + "spans": [ + { + "bbox": [ + 106, + 636, + 505, + 648 + ], + "score": 1.0, + "content": "tasks, often outperforming the state of the art. In addition, diffusion models provide valuable benefits", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 646, + 505, + 659 + ], + "spans": [ + { + "bbox": [ + 105, + 646, + 505, + 659 + ], + "score": 1.0, + "content": "over networks trained with regression; in particular, diffusion allows for approximate inference with", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 657, + 436, + 670 + ], + "spans": [ + { + "bbox": [ + 106, + 657, + 436, + 670 + ], + "score": 1.0, + "content": "multi-modal distributions, capturing uncertainty and ambiguity (e.g. see Figure 1).", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 36, + "bbox_fs": [ + 105, + 614, + 505, + 670 + ] + }, + { + "type": "text", + "bbox": [ + 109, + 673, + 502, + 696 + ], + "lines": [ + { + "bbox": [ + 106, + 672, + 505, + 687 + ], + "spans": [ + { + "bbox": [ + 106, + 672, + 505, + 687 + ], + "score": 1.0, + "content": "One key barrier to training useful diffusion models for monocular depth and optical flow inference", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 685, + 504, + 697 + ], + "spans": [ + { + "bbox": [ + 106, + 685, + 504, + 697 + ], + "score": 1.0, + "content": "concerns the amount and quality of available training data. Given the limited availability of labelled", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 295, + 506, + 311 + ], + "spans": [ + { + "bbox": [ + 105, + 295, + 506, + 311 + ], + "score": 1.0, + "content": "training data, we propose a training pipeline comprising multi-task self-supervised pre-training", + "type": "text", + "cross_page": true + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 308, + 505, + 321 + ], + "spans": [ + { + "bbox": [ + 105, + 308, + 505, + 321 + ], + "score": 1.0, + "content": "followed by supervised pre-training using a combination of real and synthetic data. Multi-task", + "type": "text", + "cross_page": true + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 318, + 505, + 331 + ], + "spans": [ + { + "bbox": [ + 105, + 318, + 505, + 331 + ], + "score": 1.0, + "content": "self-supervised pre-training leverages the strong performance of diffusion models on tasks like", + "type": "text", + "cross_page": true + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 330, + 505, + 343 + ], + "spans": [ + { + "bbox": [ + 105, + 330, + 505, + 343 + ], + "score": 1.0, + "content": "colorization and inpainting [e.g., 54]. We also find that supervised (pre-)training with a combination", + "type": "text", + "cross_page": true + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 340, + 400, + 354 + ], + "spans": [ + { + "bbox": [ + 105, + 340, + 400, + 354 + ], + "score": 1.0, + "content": "of real and large-scale synthetic data improves performance significantly.", + "type": "text", + "cross_page": true + } + ], + "index": 13 + } + ], + "index": 39.5, + "bbox_fs": [ + 106, + 672, + 505, + 697 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 69, + 502, + 214 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 69, + 502, + 214 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 69, + 502, + 214 + ], + "spans": [ + { + "bbox": [ + 106, + 69, + 502, + 214 + ], + "score": 0.966, + "type": "image", + "image_path": "4753793252e60e934fa0c425781c56b83be628669ba296b8de50760137a9b122.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 69, + 502, + 117.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 117.33333333333334, + 502, + 165.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 165.66666666666669, + 502, + 214.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 220, + 505, + 281 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 220, + 505, + 231 + ], + "spans": [ + { + "bbox": [ + 105, + 220, + 505, + 231 + ], + "score": 1.0, + "content": "Figure 1: Examples of multi-modal prediction on depth (NYU) and optical flow (Sintel and KITTI). Each", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 230, + 506, + 243 + ], + "spans": [ + { + "bbox": [ + 105, + 230, + 506, + 243 + ], + "score": 1.0, + "content": "row shows an input image (or overlayed image pair for optical flow), a variance heat map from 8 samples,", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 239, + 505, + 252 + ], + "spans": [ + { + "bbox": [ + 105, + 239, + 505, + 252 + ], + "score": 1.0, + "content": "and 3 individual samples. Our model captures multi-modality in uncertain/ambiguous cases, such as reflective", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 104, + 250, + 507, + 262 + ], + "spans": [ + { + "bbox": [ + 104, + 250, + 507, + 262 + ], + "score": 1.0, + "content": "(e.g. mirror on NYU), transparent (e.g. vehicle window on KITTI), and translucent (e.g. fog on Sintel) regions.", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 260, + 506, + 272 + ], + "spans": [ + { + "bbox": [ + 105, + 260, + 506, + 272 + ], + "score": 1.0, + "content": "High variance also occurs at object boundaries, which are often challenging cases for optical flow, and also", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 270, + 505, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 270, + 505, + 282 + ], + "score": 1.0, + "content": "partially originate from noisy ground truth measurements for depth. See Fig. 8, 9, 10 and 11 for more examples.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 5.5 + } + ], + "index": 3.25 + }, + { + "type": "text", + "bbox": [ + 107, + 297, + 505, + 352 + ], + "lines": [ + { + "bbox": [ + 105, + 295, + 506, + 311 + ], + "spans": [ + { + "bbox": [ + 105, + 295, + 506, + 311 + ], + "score": 1.0, + "content": "training data, we propose a training pipeline comprising multi-task self-supervised pre-training", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 308, + 505, + 321 + ], + "spans": [ + { + "bbox": [ + 105, + 308, + 505, + 321 + ], + "score": 1.0, + "content": "followed by supervised pre-training using a combination of real and synthetic data. Multi-task", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 318, + 505, + 331 + ], + "spans": [ + { + "bbox": [ + 105, + 318, + 505, + 331 + ], + "score": 1.0, + "content": "self-supervised pre-training leverages the strong performance of diffusion models on tasks like", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 330, + 505, + 343 + ], + "spans": [ + { + "bbox": [ + 105, + 330, + 505, + 343 + ], + "score": 1.0, + "content": "colorization and inpainting [e.g., 54]. We also find that supervised (pre-)training with a combination", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 340, + 400, + 354 + ], + "spans": [ + { + "bbox": [ + 105, + 340, + 400, + 354 + ], + "score": 1.0, + "content": "of real and large-scale synthetic data improves performance significantly.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 107, + 357, + 505, + 434 + ], + "lines": [ + { + "bbox": [ + 105, + 357, + 505, + 370 + ], + "spans": [ + { + "bbox": [ + 105, + 357, + 505, + 370 + ], + "score": 1.0, + "content": "A further issue concerns the fact that many existing real datasets for depth and optical flow have", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 368, + 505, + 380 + ], + "spans": [ + { + "bbox": [ + 105, + 368, + 505, + 380 + ], + "score": 1.0, + "content": "noisy and incomplete ground truth annotations. This presents a challenge for the conventional", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 379, + 505, + 391 + ], + "spans": [ + { + "bbox": [ + 105, + 379, + 505, + 391 + ], + "score": 1.0, + "content": "training framework and iterative sampling in diffusion models, leading to a problematic distribution", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 390, + 505, + 402 + ], + "spans": [ + { + "bbox": [ + 105, + 390, + 472, + 402 + ], + "score": 1.0, + "content": "shift between training and inference. To mitigate these issues we propose the use of an", + "type": "text" + }, + { + "bbox": [ + 472, + 390, + 484, + 401 + ], + "score": 0.87, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 485, + 390, + 505, + 402 + ], + "score": 1.0, + "content": "loss", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 401, + 506, + 413 + ], + "spans": [ + { + "bbox": [ + 105, + 401, + 506, + 413 + ], + "score": 1.0, + "content": "for robustness, infilling missing depth values during training, and the introduction of step-unrolled", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 412, + 506, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 412, + 506, + 424 + ], + "score": 1.0, + "content": "denoising diffusion. These elements of the model are shown through ablations to be important for", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 423, + 235, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 423, + 235, + 434 + ], + "score": 1.0, + "content": "both depth and flow estimation.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 17 + }, + { + "type": "text", + "bbox": [ + 107, + 439, + 239, + 450 + ], + "lines": [ + { + "bbox": [ + 106, + 438, + 240, + 451 + ], + "spans": [ + { + "bbox": [ + 106, + 438, + 240, + 451 + ], + "score": 1.0, + "content": "Our contributions are as follows:", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 21 + }, + { + "type": "text", + "bbox": [ + 108, + 455, + 506, + 613 + ], + "lines": [ + { + "bbox": [ + 108, + 456, + 505, + 467 + ], + "spans": [ + { + "bbox": [ + 108, + 456, + 505, + 467 + ], + "score": 1.0, + "content": "1. We formulate optical flow and monocular depth estimation as image to image translation with", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 119, + 466, + 475, + 479 + ], + "spans": [ + { + "bbox": [ + 119, + 466, + 475, + 479 + ], + "score": 1.0, + "content": "generative diffusion models, without specialized loss functions and model architectures.", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 108, + 477, + 506, + 492 + ], + "spans": [ + { + "bbox": [ + 108, + 479, + 118, + 489 + ], + "score": 1.0, + "content": "2.", + "type": "text" + }, + { + "bbox": [ + 118, + 477, + 506, + 492 + ], + "score": 1.0, + "content": "We identify and propose solutions to several important issues w.r.t data. For both tasks, to mitigate", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 119, + 487, + 507, + 503 + ], + "spans": [ + { + "bbox": [ + 119, + 487, + 507, + 503 + ], + "score": 1.0, + "content": "distribution shift between training and inference with noisy, incomplete data, we propose infilling,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 119, + 501, + 506, + 514 + ], + "spans": [ + { + "bbox": [ + 119, + 501, + 208, + 514 + ], + "score": 1.0, + "content": "step-unrolling, and an", + "type": "text" + }, + { + "bbox": [ + 208, + 501, + 221, + 511 + ], + "score": 0.88, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 221, + 501, + 506, + 514 + ], + "score": 1.0, + "content": "loss during training. For flow, to improve generalization, we introduce a", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 119, + 511, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 119, + 511, + 505, + 524 + ], + "score": 1.0, + "content": "new dataset mixture for pre-training, yielding a RAFT [74] baseline that outperforms all published", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 120, + 523, + 443, + 534 + ], + "spans": [ + { + "bbox": [ + 120, + 523, + 443, + 534 + ], + "score": 1.0, + "content": "methods in zero-shot performance on the Sintel and KITTI training benchmarks.", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 108, + 534, + 505, + 548 + ], + "spans": [ + { + "bbox": [ + 108, + 536, + 118, + 546 + ], + "score": 1.0, + "content": "3.", + "type": "text" + }, + { + "bbox": [ + 119, + 534, + 505, + 548 + ], + "score": 1.0, + "content": "Our diffusion models is competitive with or surpasses SOTA for both tasks. For monocular", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 120, + 545, + 505, + 559 + ], + "spans": [ + { + "bbox": [ + 120, + 545, + 505, + 559 + ], + "score": 1.0, + "content": "depth estimation we achieve a SOTA relative error of 0.074 on the NYU dataset and perform", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 119, + 556, + 506, + 570 + ], + "spans": [ + { + "bbox": [ + 119, + 556, + 506, + 570 + ], + "score": 1.0, + "content": "competitively on KITTI. For flow, diffusion surpasses the stronger RAFT baseline by a large", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 119, + 567, + 505, + 579 + ], + "spans": [ + { + "bbox": [ + 119, + 567, + 448, + 579 + ], + "score": 1.0, + "content": "margin in pre-training and our fine-tuned model achieves an Fl-all outlier rate of", + "type": "text" + }, + { + "bbox": [ + 449, + 567, + 476, + 578 + ], + "score": 0.85, + "content": "3 . 2 6 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 477, + 567, + 505, + 579 + ], + "score": 1.0, + "content": "on the", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 119, + 578, + 443, + 591 + ], + "spans": [ + { + "bbox": [ + 119, + 578, + 242, + 591 + ], + "score": 1.0, + "content": "public KITTI test benchmark,", + "type": "text" + }, + { + "bbox": [ + 242, + 578, + 270, + 589 + ], + "score": 0.89, + "content": "{ \\sim } 2 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 578, + 443, + 591 + ], + "score": 1.0, + "content": "lower than the best published method [70].", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 108, + 590, + 506, + 603 + ], + "spans": [ + { + "bbox": [ + 108, + 590, + 506, + 603 + ], + "score": 1.0, + "content": "4. Our diffusion model is also shown to capture flow and depth uncertainty, and the iterative", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 121, + 602, + 438, + 614 + ], + "spans": [ + { + "bbox": [ + 121, + 602, + 438, + 614 + ], + "score": 1.0, + "content": "denoising process enables zero-shot, coarse-to-fine refinement, and imputation.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 28.5 + }, + { + "type": "title", + "bbox": [ + 107, + 628, + 194, + 641 + ], + "lines": [ + { + "bbox": [ + 105, + 627, + 196, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 196, + 643 + ], + "score": 1.0, + "content": "2 Related work", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 36 + }, + { + "type": "text", + "bbox": [ + 108, + 650, + 504, + 673 + ], + "lines": [ + { + "bbox": [ + 106, + 650, + 504, + 663 + ], + "spans": [ + { + "bbox": [ + 106, + 650, + 504, + 663 + ], + "score": 1.0, + "content": "Optical flow and depth estimation have been extensively studied. Here we briefly review only the", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 662, + 436, + 673 + ], + "spans": [ + { + "bbox": [ + 106, + 662, + 436, + 673 + ], + "score": 1.0, + "content": "most relevant work, and refer the interested readers to the references cited therein.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 37.5 + }, + { + "type": "text", + "bbox": [ + 107, + 678, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 678, + 505, + 691 + ], + "spans": [ + { + "bbox": [ + 106, + 678, + 505, + 691 + ], + "score": 1.0, + "content": "Optical flow. The predominant approach to optical flow is regression-based, with a focus on", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 688, + 507, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 507, + 702 + ], + "score": 1.0, + "content": "specialized network architectures to exploit domain knowledge, e.g., cost volume construction [14,", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 700, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 700, + 506, + 712 + ], + "score": 1.0, + "content": "22, 23, 38, 68, 81, 83, 85], coarse-to-fine estimation [68, 77, 82], occlusion handling [24, 27, 67],", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "score": 1.0, + "content": "or iterative refinement [25, 26, 74], as evidenced by public benchmark datasets [4, 44]. Some", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 40.5 + } + ], + "page_idx": 1, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 330, + 71, + 443, + 80 + ], + "lines": [] + }, + { + "type": "discarded", + "bbox": [ + 302, + 741, + 309, + 750 + ], + "lines": [ + { + "bbox": [ + 301, + 740, + 310, + 753 + ], + "spans": [ + { + "bbox": [ + 301, + 740, + 310, + 753 + ], + "score": 1.0, + "content": "2", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 69, + 502, + 214 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 69, + 502, + 214 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 69, + 502, + 214 + ], + "spans": [ + { + "bbox": [ + 106, + 69, + 502, + 214 + ], + "score": 0.966, + "type": "image", + "image_path": "4753793252e60e934fa0c425781c56b83be628669ba296b8de50760137a9b122.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 69, + 502, + 117.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 117.33333333333334, + 502, + 165.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 165.66666666666669, + 502, + 214.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 220, + 505, + 281 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 220, + 505, + 231 + ], + "spans": [ + { + "bbox": [ + 105, + 220, + 505, + 231 + ], + "score": 1.0, + "content": "Figure 1: Examples of multi-modal prediction on depth (NYU) and optical flow (Sintel and KITTI). Each", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 230, + 506, + 243 + ], + "spans": [ + { + "bbox": [ + 105, + 230, + 506, + 243 + ], + "score": 1.0, + "content": "row shows an input image (or overlayed image pair for optical flow), a variance heat map from 8 samples,", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 239, + 505, + 252 + ], + "spans": [ + { + "bbox": [ + 105, + 239, + 505, + 252 + ], + "score": 1.0, + "content": "and 3 individual samples. Our model captures multi-modality in uncertain/ambiguous cases, such as reflective", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 104, + 250, + 507, + 262 + ], + "spans": [ + { + "bbox": [ + 104, + 250, + 507, + 262 + ], + "score": 1.0, + "content": "(e.g. mirror on NYU), transparent (e.g. vehicle window on KITTI), and translucent (e.g. fog on Sintel) regions.", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 260, + 506, + 272 + ], + "spans": [ + { + "bbox": [ + 105, + 260, + 506, + 272 + ], + "score": 1.0, + "content": "High variance also occurs at object boundaries, which are often challenging cases for optical flow, and also", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 270, + 505, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 270, + 505, + 282 + ], + "score": 1.0, + "content": "partially originate from noisy ground truth measurements for depth. See Fig. 8, 9, 10 and 11 for more examples.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 5.5 + } + ], + "index": 3.25 + }, + { + "type": "text", + "bbox": [ + 107, + 297, + 505, + 352 + ], + "lines": [], + "index": 11, + "bbox_fs": [ + 105, + 295, + 506, + 354 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 357, + 505, + 434 + ], + "lines": [ + { + "bbox": [ + 105, + 357, + 505, + 370 + ], + "spans": [ + { + "bbox": [ + 105, + 357, + 505, + 370 + ], + "score": 1.0, + "content": "A further issue concerns the fact that many existing real datasets for depth and optical flow have", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 368, + 505, + 380 + ], + "spans": [ + { + "bbox": [ + 105, + 368, + 505, + 380 + ], + "score": 1.0, + "content": "noisy and incomplete ground truth annotations. This presents a challenge for the conventional", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 379, + 505, + 391 + ], + "spans": [ + { + "bbox": [ + 105, + 379, + 505, + 391 + ], + "score": 1.0, + "content": "training framework and iterative sampling in diffusion models, leading to a problematic distribution", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 390, + 505, + 402 + ], + "spans": [ + { + "bbox": [ + 105, + 390, + 472, + 402 + ], + "score": 1.0, + "content": "shift between training and inference. To mitigate these issues we propose the use of an", + "type": "text" + }, + { + "bbox": [ + 472, + 390, + 484, + 401 + ], + "score": 0.87, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 485, + 390, + 505, + 402 + ], + "score": 1.0, + "content": "loss", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 401, + 506, + 413 + ], + "spans": [ + { + "bbox": [ + 105, + 401, + 506, + 413 + ], + "score": 1.0, + "content": "for robustness, infilling missing depth values during training, and the introduction of step-unrolled", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 412, + 506, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 412, + 506, + 424 + ], + "score": 1.0, + "content": "denoising diffusion. These elements of the model are shown through ablations to be important for", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 423, + 235, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 423, + 235, + 434 + ], + "score": 1.0, + "content": "both depth and flow estimation.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 17, + "bbox_fs": [ + 105, + 357, + 506, + 434 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 439, + 239, + 450 + ], + "lines": [ + { + "bbox": [ + 106, + 438, + 240, + 451 + ], + "spans": [ + { + "bbox": [ + 106, + 438, + 240, + 451 + ], + "score": 1.0, + "content": "Our contributions are as follows:", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 21, + "bbox_fs": [ + 106, + 438, + 240, + 451 + ] + }, + { + "type": "list", + "bbox": [ + 108, + 455, + 506, + 613 + ], + "lines": [ + { + "bbox": [ + 108, + 456, + 505, + 467 + ], + "spans": [ + { + "bbox": [ + 108, + 456, + 505, + 467 + ], + "score": 1.0, + "content": "1. We formulate optical flow and monocular depth estimation as image to image translation with", + "type": "text" + } + ], + "index": 22, + "is_list_start_line": true + }, + { + "bbox": [ + 119, + 466, + 475, + 479 + ], + "spans": [ + { + "bbox": [ + 119, + 466, + 475, + 479 + ], + "score": 1.0, + "content": "generative diffusion models, without specialized loss functions and model architectures.", + "type": "text" + } + ], + "index": 23, + "is_list_end_line": true + }, + { + "bbox": [ + 108, + 477, + 506, + 492 + ], + "spans": [ + { + "bbox": [ + 108, + 479, + 118, + 489 + ], + "score": 1.0, + "content": "2.", + "type": "text" + }, + { + "bbox": [ + 118, + 477, + 506, + 492 + ], + "score": 1.0, + "content": "We identify and propose solutions to several important issues w.r.t data. For both tasks, to mitigate", + "type": "text" + } + ], + "index": 24, + "is_list_start_line": true + }, + { + "bbox": [ + 119, + 487, + 507, + 503 + ], + "spans": [ + { + "bbox": [ + 119, + 487, + 507, + 503 + ], + "score": 1.0, + "content": "distribution shift between training and inference with noisy, incomplete data, we propose infilling,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 119, + 501, + 506, + 514 + ], + "spans": [ + { + "bbox": [ + 119, + 501, + 208, + 514 + ], + "score": 1.0, + "content": "step-unrolling, and an", + "type": "text" + }, + { + "bbox": [ + 208, + 501, + 221, + 511 + ], + "score": 0.88, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 221, + 501, + 506, + 514 + ], + "score": 1.0, + "content": "loss during training. For flow, to improve generalization, we introduce a", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 119, + 511, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 119, + 511, + 505, + 524 + ], + "score": 1.0, + "content": "new dataset mixture for pre-training, yielding a RAFT [74] baseline that outperforms all published", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 120, + 523, + 443, + 534 + ], + "spans": [ + { + "bbox": [ + 120, + 523, + 443, + 534 + ], + "score": 1.0, + "content": "methods in zero-shot performance on the Sintel and KITTI training benchmarks.", + "type": "text" + } + ], + "index": 28, + "is_list_end_line": true + }, + { + "bbox": [ + 108, + 534, + 505, + 548 + ], + "spans": [ + { + "bbox": [ + 108, + 536, + 118, + 546 + ], + "score": 1.0, + "content": "3.", + "type": "text" + }, + { + "bbox": [ + 119, + 534, + 505, + 548 + ], + "score": 1.0, + "content": "Our diffusion models is competitive with or surpasses SOTA for both tasks. For monocular", + "type": "text" + } + ], + "index": 29, + "is_list_start_line": true + }, + { + "bbox": [ + 120, + 545, + 505, + 559 + ], + "spans": [ + { + "bbox": [ + 120, + 545, + 505, + 559 + ], + "score": 1.0, + "content": "depth estimation we achieve a SOTA relative error of 0.074 on the NYU dataset and perform", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 119, + 556, + 506, + 570 + ], + "spans": [ + { + "bbox": [ + 119, + 556, + 506, + 570 + ], + "score": 1.0, + "content": "competitively on KITTI. For flow, diffusion surpasses the stronger RAFT baseline by a large", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 119, + 567, + 505, + 579 + ], + "spans": [ + { + "bbox": [ + 119, + 567, + 448, + 579 + ], + "score": 1.0, + "content": "margin in pre-training and our fine-tuned model achieves an Fl-all outlier rate of", + "type": "text" + }, + { + "bbox": [ + 449, + 567, + 476, + 578 + ], + "score": 0.85, + "content": "3 . 2 6 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 477, + 567, + 505, + 579 + ], + "score": 1.0, + "content": "on the", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 119, + 578, + 443, + 591 + ], + "spans": [ + { + "bbox": [ + 119, + 578, + 242, + 591 + ], + "score": 1.0, + "content": "public KITTI test benchmark,", + "type": "text" + }, + { + "bbox": [ + 242, + 578, + 270, + 589 + ], + "score": 0.89, + "content": "{ \\sim } 2 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 578, + 443, + 591 + ], + "score": 1.0, + "content": "lower than the best published method [70].", + "type": "text" + } + ], + "index": 33, + "is_list_end_line": true + }, + { + "bbox": [ + 108, + 590, + 506, + 603 + ], + "spans": [ + { + "bbox": [ + 108, + 590, + 506, + 603 + ], + "score": 1.0, + "content": "4. Our diffusion model is also shown to capture flow and depth uncertainty, and the iterative", + "type": "text" + } + ], + "index": 34, + "is_list_start_line": true + }, + { + "bbox": [ + 121, + 602, + 438, + 614 + ], + "spans": [ + { + "bbox": [ + 121, + 602, + 438, + 614 + ], + "score": 1.0, + "content": "denoising process enables zero-shot, coarse-to-fine refinement, and imputation.", + "type": "text" + } + ], + "index": 35, + "is_list_end_line": true + } + ], + "index": 28.5, + "bbox_fs": [ + 108, + 456, + 507, + 614 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 628, + 194, + 641 + ], + "lines": [ + { + "bbox": [ + 105, + 627, + 196, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 196, + 643 + ], + "score": 1.0, + "content": "2 Related work", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 36 + }, + { + "type": "text", + "bbox": [ + 108, + 650, + 504, + 673 + ], + "lines": [ + { + "bbox": [ + 106, + 650, + 504, + 663 + ], + "spans": [ + { + "bbox": [ + 106, + 650, + 504, + 663 + ], + "score": 1.0, + "content": "Optical flow and depth estimation have been extensively studied. Here we briefly review only the", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 662, + 436, + 673 + ], + "spans": [ + { + "bbox": [ + 106, + 662, + 436, + 673 + ], + "score": 1.0, + "content": "most relevant work, and refer the interested readers to the references cited therein.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 37.5, + "bbox_fs": [ + 106, + 650, + 504, + 673 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 678, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 678, + 505, + 691 + ], + "spans": [ + { + "bbox": [ + 106, + 678, + 505, + 691 + ], + "score": 1.0, + "content": "Optical flow. The predominant approach to optical flow is regression-based, with a focus on", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 688, + 507, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 507, + 702 + ], + "score": 1.0, + "content": "specialized network architectures to exploit domain knowledge, e.g., cost volume construction [14,", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 700, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 700, + 506, + 712 + ], + "score": 1.0, + "content": "22, 23, 38, 68, 81, 83, 85], coarse-to-fine estimation [68, 77, 82], occlusion handling [24, 27, 67],", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "score": 1.0, + "content": "or iterative refinement [25, 26, 74], as evidenced by public benchmark datasets [4, 44]. Some", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 302, + 505, + 314 + ], + "spans": [ + { + "bbox": [ + 106, + 302, + 505, + 314 + ], + "score": 1.0, + "content": "recent work has also advocated for generic architectures: Perceiver IO [28] introduces a generic", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 312, + 506, + 327 + ], + "spans": [ + { + "bbox": [ + 105, + 312, + 506, + 327 + ], + "score": 1.0, + "content": "transformer-based model that works for any modality, including optical flow and language modeling.", + "type": "text", + "cross_page": true + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 324, + 506, + 337 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 506, + 337 + ], + "score": 1.0, + "content": "Regression-based methods, however, only give a single prediction of the optical flow and do not", + "type": "text", + "cross_page": true + } + ], + "index": 9 + }, + { + "bbox": [ + 104, + 334, + 506, + 349 + ], + "spans": [ + { + "bbox": [ + 104, + 334, + 506, + 349 + ], + "score": 1.0, + "content": "readily capture uncertainty or ambiguity in the flow. Our work introduces a surprisingly simple,", + "type": "text", + "cross_page": true + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 345, + 392, + 359 + ], + "spans": [ + { + "bbox": [ + 105, + 345, + 392, + 359 + ], + "score": 1.0, + "content": "generic architecture for optical flow using a denoising diffusion model.", + "type": "text", + "cross_page": true + } + ], + "index": 11 + } + ], + "index": 40.5, + "bbox_fs": [ + 105, + 678, + 507, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 110, + 76, + 497, + 239 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 110, + 76, + 497, + 239 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 110, + 76, + 497, + 239 + ], + "spans": [ + { + "bbox": [ + 110, + 76, + 497, + 239 + ], + "score": 0.973, + "type": "image", + "image_path": "4f646787799dc6d6809ae8af0c1f1a8181e23bad548bba803e62712afccf0d98.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 110, + 76, + 497, + 130.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 110, + 130.33333333333334, + 497, + 184.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 110, + 184.66666666666669, + 497, + 239.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 253, + 505, + 293 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 252, + 506, + 265 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 506, + 265 + ], + "score": 1.0, + "content": "Figure 2: Training architecture. Given ground truth flow/depth, we first infill missing values using interpolation.", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 262, + 506, + 275 + ], + "spans": [ + { + "bbox": [ + 105, + 262, + 506, + 275 + ], + "score": 1.0, + "content": "Then, we add noise to the label map and train a neural network to model the conditional distribution of the noise", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 271, + 505, + 285 + ], + "spans": [ + { + "bbox": [ + 105, + 271, + 505, + 285 + ], + "score": 1.0, + "content": "given the RGB image(s), noisy label, and time step. One can optionally unroll the denoising step(s) during", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 281, + 455, + 297 + ], + "spans": [ + { + "bbox": [ + 105, + 281, + 441, + 297 + ], + "score": 1.0, + "content": "training (with stop gradient) to bridge the distribution gap between training and inference for", + "type": "text" + }, + { + "bbox": [ + 442, + 285, + 450, + 293 + ], + "score": 0.83, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 281, + 455, + 297 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "text", + "bbox": [ + 106, + 302, + 506, + 357 + ], + "lines": [ + { + "bbox": [ + 106, + 302, + 505, + 314 + ], + "spans": [ + { + "bbox": [ + 106, + 302, + 505, + 314 + ], + "score": 1.0, + "content": "recent work has also advocated for generic architectures: Perceiver IO [28] introduces a generic", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 312, + 506, + 327 + ], + "spans": [ + { + "bbox": [ + 105, + 312, + 506, + 327 + ], + "score": 1.0, + "content": "transformer-based model that works for any modality, including optical flow and language modeling.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 324, + 506, + 337 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 506, + 337 + ], + "score": 1.0, + "content": "Regression-based methods, however, only give a single prediction of the optical flow and do not", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 104, + 334, + 506, + 349 + ], + "spans": [ + { + "bbox": [ + 104, + 334, + 506, + 349 + ], + "score": 1.0, + "content": "readily capture uncertainty or ambiguity in the flow. Our work introduces a surprisingly simple,", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 345, + 392, + 359 + ], + "spans": [ + { + "bbox": [ + 105, + 345, + 392, + 359 + ], + "score": 1.0, + "content": "generic architecture for optical flow using a denoising diffusion model.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 107, + 362, + 503, + 385 + ], + "lines": [ + { + "bbox": [ + 105, + 361, + 505, + 375 + ], + "spans": [ + { + "bbox": [ + 105, + 361, + 505, + 375 + ], + "score": 1.0, + "content": "We find that this generic generative model is surprisingly effective for optical flow, recovering fine", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 373, + 461, + 386 + ], + "spans": [ + { + "bbox": [ + 106, + 373, + 461, + 386 + ], + "score": 1.0, + "content": "details on motion boundaries, while capturing multi-modality of the motion distribution.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12.5 + }, + { + "type": "text", + "bbox": [ + 106, + 389, + 505, + 456 + ], + "lines": [ + { + "bbox": [ + 105, + 389, + 506, + 402 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 506, + 402 + ], + "score": 1.0, + "content": "Monocular depth. Monocular depth estimation has been a long-standing problem in computer vision", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 401, + 505, + 412 + ], + "spans": [ + { + "bbox": [ + 106, + 401, + 505, + 412 + ], + "score": 1.0, + "content": "[58, 59] with recent progress focusing on specialized loss functions and architectures [1, 5, 15, 31]", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 410, + 505, + 423 + ], + "spans": [ + { + "bbox": [ + 106, + 410, + 505, + 423 + ], + "score": 1.0, + "content": "such as the use of multi-scale networks [12, 13], adaptive binning [3, 35] and weighted scale-shift", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 423, + 505, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 423, + 505, + 434 + ], + "score": 1.0, + "content": "invariant losses [13]. Large-scale in-domain pre-training has also been effective for depth estimation", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 432, + 506, + 446 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 506, + 446 + ], + "score": 1.0, + "content": "[49, 50, 52], which we find to be the case here as well. We build on this rich literature, but with a", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 444, + 419, + 457 + ], + "spans": [ + { + "bbox": [ + 106, + 444, + 419, + 457 + ], + "score": 1.0, + "content": "simple, generic architecture, leveraging recent advances in generative models.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 16.5 + }, + { + "type": "text", + "bbox": [ + 106, + 460, + 505, + 527 + ], + "lines": [ + { + "bbox": [ + 105, + 459, + 506, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 506, + 473 + ], + "score": 1.0, + "content": "Diffusion models. Diffusion models are latent-variable generative models trained to transform a", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "sample of a Gaussian noise into a sample from a data distribution [21, 62]. They comprise a forward", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 104, + 482, + 506, + 495 + ], + "spans": [ + { + "bbox": [ + 104, + 482, + 359, + 495 + ], + "score": 1.0, + "content": "process that gradually annihilates data by adding noise, as ‘time’", + "type": "text" + }, + { + "bbox": [ + 359, + 483, + 365, + 492 + ], + "score": 0.47, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 365, + 482, + 506, + 495 + ], + "score": 1.0, + "content": "increases from 0 to 1, and a learned", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 104, + 493, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 104, + 493, + 481, + 506 + ], + "score": 1.0, + "content": "generative process that reverses the forward process, starting from a sample of random noise at", + "type": "text" + }, + { + "bbox": [ + 482, + 493, + 505, + 504 + ], + "score": 0.88, + "content": "t = 1", + "type": "inline_equation" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 504, + 506, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 504, + 339, + 516 + ], + "score": 1.0, + "content": "and incrementally adding structure (attenuating noise) as", + "type": "text" + }, + { + "bbox": [ + 339, + 505, + 344, + 514 + ], + "score": 0.63, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 504, + 506, + 516 + ], + "score": 1.0, + "content": "decreases to 0. A conditional diffusion", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 515, + 442, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 442, + 528 + ], + "score": 1.0, + "content": "model conditions the steps of the reverse process (e.g., on labels, text, or an image).", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 22.5 + }, + { + "type": "text", + "bbox": [ + 107, + 531, + 505, + 586 + ], + "lines": [ + { + "bbox": [ + 106, + 531, + 505, + 544 + ], + "spans": [ + { + "bbox": [ + 106, + 531, + 291, + 544 + ], + "score": 1.0, + "content": "Central to the model is a denoising network", + "type": "text" + }, + { + "bbox": [ + 291, + 532, + 302, + 543 + ], + "score": 0.86, + "content": "f _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 302, + 531, + 459, + 544 + ], + "score": 1.0, + "content": "that is trained to take a noisy sample", + "type": "text" + }, + { + "bbox": [ + 459, + 533, + 469, + 543 + ], + "score": 0.83, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 469, + 531, + 505, + 544 + ], + "score": 1.0, + "content": "at some", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 542, + 505, + 554 + ], + "spans": [ + { + "bbox": [ + 106, + 542, + 147, + 554 + ], + "score": 1.0, + "content": "time-step", + "type": "text" + }, + { + "bbox": [ + 147, + 543, + 152, + 552 + ], + "score": 0.67, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 153, + 542, + 291, + 554 + ], + "score": 1.0, + "content": ", along with a conditioning signal", + "type": "text" + }, + { + "bbox": [ + 292, + 545, + 299, + 552 + ], + "score": 0.77, + "content": "x", + "type": "inline_equation" + }, + { + "bbox": [ + 299, + 542, + 505, + 554 + ], + "score": 1.0, + "content": ", and predict a less noisy sample. Using Gaussian", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 553, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 553, + 505, + 567 + ], + "score": 1.0, + "content": "noise in the forward process, one can express the training objective over the sequence of transitions", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 120, + 577 + ], + "score": 1.0, + "content": "(as", + "type": "text" + }, + { + "bbox": [ + 120, + 565, + 126, + 574 + ], + "score": 0.71, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 564, + 505, + 577 + ], + "score": 1.0, + "content": "slowly decreases) as a sum of non-linear regression objectives, with the L2 loss (here with the", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 107, + 576, + 189, + 587 + ], + "spans": [ + { + "bbox": [ + 107, + 577, + 112, + 585 + ], + "score": 0.67, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 112, + 576, + 189, + 587 + ], + "score": 1.0, + "content": "-parameterization):", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 28 + }, + { + "type": "interline_equation", + "bbox": [ + 202, + 588, + 409, + 626 + ], + "lines": [ + { + "bbox": [ + 202, + 588, + 409, + 626 + ], + "spans": [ + { + "bbox": [ + 202, + 588, + 409, + 626 + ], + "score": 0.93, + "content": "\\mathbb { E } _ { ( \\pmb { x } , \\pmb { y } ) } \\mathbb { E } _ { ( t , \\epsilon ) } \\bigg \\| f _ { \\theta } ( \\pmb { x } , \\underbrace { \\sqrt { \\gamma _ { t } } \\pmb { y } + \\sqrt { 1 - \\gamma _ { t } } \\pmb { \\epsilon } } _ { \\pmb { y _ { t } } } , t ) - \\pmb { \\epsilon } \\bigg \\| _ { 2 } ^ { 2 }", + "type": "interline_equation", + "image_path": "f16794dc92c133b4267a99986ce44c6bf54d08e321f0f07879b449286dcb8122.jpg" + } + ] + } + ], + "index": 31.5, + "virtual_lines": [ + { + "bbox": [ + 202, + 588, + 409, + 607.0 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 202, + 607.0, + 409, + 626.0 + ], + "spans": [], + "index": 32 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 628, + 505, + 662 + ], + "lines": [ + { + "bbox": [ + 105, + 627, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 134, + 642 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 135, + 629, + 189, + 641 + ], + "score": 0.78, + "content": "\\epsilon \\sim \\mathcal { N } ( 0 , I )", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 627, + 194, + 642 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 194, + 629, + 245, + 641 + ], + "score": 0.83, + "content": "t \\sim \\mathcal { U } ( 0 , 1 )", + "type": "inline_equation" + }, + { + "bbox": [ + 246, + 627, + 297, + 642 + ], + "score": 1.0, + "content": ", and where", + "type": "text" + }, + { + "bbox": [ + 297, + 629, + 330, + 641 + ], + "score": 0.93, + "content": "\\gamma _ { t } ~ > ~ 0", + "type": "inline_equation" + }, + { + "bbox": [ + 330, + 627, + 506, + 642 + ], + "score": 1.0, + "content": "is computed with a pre-determined noise", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 639, + 505, + 652 + ], + "spans": [ + { + "bbox": [ + 105, + 639, + 409, + 652 + ], + "score": 1.0, + "content": "schedule. For inference (i.e., sampling), one draws a random noise sample", + "type": "text" + }, + { + "bbox": [ + 410, + 641, + 421, + 651 + ], + "score": 0.85, + "content": "\\mathbf { \\pmb { y } } _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 421, + 639, + 505, + 652 + ], + "score": 1.0, + "content": ", and then iteratively", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 651, + 489, + 663 + ], + "spans": [ + { + "bbox": [ + 106, + 651, + 126, + 663 + ], + "score": 1.0, + "content": "uses", + "type": "text" + }, + { + "bbox": [ + 126, + 651, + 137, + 662 + ], + "score": 0.87, + "content": "f _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 651, + 433, + 663 + ], + "score": 1.0, + "content": "to estimate the noise, from which one can compute the next latent sample", + "type": "text" + }, + { + "bbox": [ + 433, + 652, + 444, + 662 + ], + "score": 0.85, + "content": "\\mathbf { \\eta } _ { \\mathbf { { 3 } } \\mathbf { { ' } } \\mathbf { { 3 } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 444, + 651, + 462, + 663 + ], + "score": 1.0, + "content": ", for", + "type": "text" + }, + { + "bbox": [ + 462, + 651, + 485, + 661 + ], + "score": 0.88, + "content": "s < t", + "type": "inline_equation" + }, + { + "bbox": [ + 485, + 651, + 489, + 663 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 34 + }, + { + "type": "text", + "bbox": [ + 107, + 667, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 667, + 506, + 679 + ], + "spans": [ + { + "bbox": [ + 106, + 667, + 506, + 679 + ], + "score": 1.0, + "content": "Self-supervised pre-training. Prior work has shown that self-supervised tasks such as coloriza-", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 678, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 678, + 506, + 690 + ], + "score": 1.0, + "content": "tion [33, 86] and masked prediction [80] serve as effective pre-training for downstream vision tasks.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 688, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 506, + 702 + ], + "score": 1.0, + "content": "Our work also confirms the benefit of self-supervised pre-training [54] for diffusion-based image-to-", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 699, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 505, + 712 + ], + "score": 1.0, + "content": "image translation, by establishing a new SOTA on optical flow and monocular depth estimation while", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 711, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 506, + 723 + ], + "score": 1.0, + "content": "also representing multi-modality and supporting zero-shot coarse-to-fine refinement and imputation.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 38 + } + ], + "page_idx": 2, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 741, + 309, + 750 + ], + "lines": [ + { + "bbox": [ + 301, + 740, + 310, + 752 + ], + "spans": [ + { + "bbox": [ + 301, + 740, + 310, + 752 + ], + "score": 1.0, + "content": "3", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 400, + 74, + 449, + 80 + ], + "lines": [ + { + "bbox": [ + 399, + 73, + 450, + 81 + ], + "spans": [ + { + "bbox": [ + 399, + 73, + 450, + 81 + ], + "score": 1.0, + "content": "GT label with holes", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 110, + 76, + 497, + 239 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 110, + 76, + 497, + 239 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 110, + 76, + 497, + 239 + ], + "spans": [ + { + "bbox": [ + 110, + 76, + 497, + 239 + ], + "score": 0.973, + "type": "image", + "image_path": "4f646787799dc6d6809ae8af0c1f1a8181e23bad548bba803e62712afccf0d98.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 110, + 76, + 497, + 130.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 110, + 130.33333333333334, + 497, + 184.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 110, + 184.66666666666669, + 497, + 239.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 253, + 505, + 293 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 252, + 506, + 265 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 506, + 265 + ], + "score": 1.0, + "content": "Figure 2: Training architecture. Given ground truth flow/depth, we first infill missing values using interpolation.", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 262, + 506, + 275 + ], + "spans": [ + { + "bbox": [ + 105, + 262, + 506, + 275 + ], + "score": 1.0, + "content": "Then, we add noise to the label map and train a neural network to model the conditional distribution of the noise", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 271, + 505, + 285 + ], + "spans": [ + { + "bbox": [ + 105, + 271, + 505, + 285 + ], + "score": 1.0, + "content": "given the RGB image(s), noisy label, and time step. One can optionally unroll the denoising step(s) during", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 281, + 455, + 297 + ], + "spans": [ + { + "bbox": [ + 105, + 281, + 441, + 297 + ], + "score": 1.0, + "content": "training (with stop gradient) to bridge the distribution gap between training and inference for", + "type": "text" + }, + { + "bbox": [ + 442, + 285, + 450, + 293 + ], + "score": 0.83, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 281, + 455, + 297 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "text", + "bbox": [ + 106, + 302, + 506, + 357 + ], + "lines": [], + "index": 9, + "bbox_fs": [ + 104, + 302, + 506, + 359 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 362, + 503, + 385 + ], + "lines": [ + { + "bbox": [ + 105, + 361, + 505, + 375 + ], + "spans": [ + { + "bbox": [ + 105, + 361, + 505, + 375 + ], + "score": 1.0, + "content": "We find that this generic generative model is surprisingly effective for optical flow, recovering fine", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 373, + 461, + 386 + ], + "spans": [ + { + "bbox": [ + 106, + 373, + 461, + 386 + ], + "score": 1.0, + "content": "details on motion boundaries, while capturing multi-modality of the motion distribution.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12.5, + "bbox_fs": [ + 105, + 361, + 505, + 386 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 389, + 505, + 456 + ], + "lines": [ + { + "bbox": [ + 105, + 389, + 506, + 402 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 506, + 402 + ], + "score": 1.0, + "content": "Monocular depth. Monocular depth estimation has been a long-standing problem in computer vision", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 401, + 505, + 412 + ], + "spans": [ + { + "bbox": [ + 106, + 401, + 505, + 412 + ], + "score": 1.0, + "content": "[58, 59] with recent progress focusing on specialized loss functions and architectures [1, 5, 15, 31]", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 410, + 505, + 423 + ], + "spans": [ + { + "bbox": [ + 106, + 410, + 505, + 423 + ], + "score": 1.0, + "content": "such as the use of multi-scale networks [12, 13], adaptive binning [3, 35] and weighted scale-shift", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 423, + 505, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 423, + 505, + 434 + ], + "score": 1.0, + "content": "invariant losses [13]. Large-scale in-domain pre-training has also been effective for depth estimation", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 432, + 506, + 446 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 506, + 446 + ], + "score": 1.0, + "content": "[49, 50, 52], which we find to be the case here as well. We build on this rich literature, but with a", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 444, + 419, + 457 + ], + "spans": [ + { + "bbox": [ + 106, + 444, + 419, + 457 + ], + "score": 1.0, + "content": "simple, generic architecture, leveraging recent advances in generative models.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 16.5, + "bbox_fs": [ + 105, + 389, + 506, + 457 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 460, + 505, + 527 + ], + "lines": [ + { + "bbox": [ + 105, + 459, + 506, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 506, + 473 + ], + "score": 1.0, + "content": "Diffusion models. Diffusion models are latent-variable generative models trained to transform a", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "sample of a Gaussian noise into a sample from a data distribution [21, 62]. They comprise a forward", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 104, + 482, + 506, + 495 + ], + "spans": [ + { + "bbox": [ + 104, + 482, + 359, + 495 + ], + "score": 1.0, + "content": "process that gradually annihilates data by adding noise, as ‘time’", + "type": "text" + }, + { + "bbox": [ + 359, + 483, + 365, + 492 + ], + "score": 0.47, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 365, + 482, + 506, + 495 + ], + "score": 1.0, + "content": "increases from 0 to 1, and a learned", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 104, + 493, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 104, + 493, + 481, + 506 + ], + "score": 1.0, + "content": "generative process that reverses the forward process, starting from a sample of random noise at", + "type": "text" + }, + { + "bbox": [ + 482, + 493, + 505, + 504 + ], + "score": 0.88, + "content": "t = 1", + "type": "inline_equation" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 504, + 506, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 504, + 339, + 516 + ], + "score": 1.0, + "content": "and incrementally adding structure (attenuating noise) as", + "type": "text" + }, + { + "bbox": [ + 339, + 505, + 344, + 514 + ], + "score": 0.63, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 504, + 506, + 516 + ], + "score": 1.0, + "content": "decreases to 0. A conditional diffusion", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 515, + 442, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 442, + 528 + ], + "score": 1.0, + "content": "model conditions the steps of the reverse process (e.g., on labels, text, or an image).", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 22.5, + "bbox_fs": [ + 104, + 459, + 506, + 528 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 531, + 505, + 586 + ], + "lines": [ + { + "bbox": [ + 106, + 531, + 505, + 544 + ], + "spans": [ + { + "bbox": [ + 106, + 531, + 291, + 544 + ], + "score": 1.0, + "content": "Central to the model is a denoising network", + "type": "text" + }, + { + "bbox": [ + 291, + 532, + 302, + 543 + ], + "score": 0.86, + "content": "f _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 302, + 531, + 459, + 544 + ], + "score": 1.0, + "content": "that is trained to take a noisy sample", + "type": "text" + }, + { + "bbox": [ + 459, + 533, + 469, + 543 + ], + "score": 0.83, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 469, + 531, + 505, + 544 + ], + "score": 1.0, + "content": "at some", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 542, + 505, + 554 + ], + "spans": [ + { + "bbox": [ + 106, + 542, + 147, + 554 + ], + "score": 1.0, + "content": "time-step", + "type": "text" + }, + { + "bbox": [ + 147, + 543, + 152, + 552 + ], + "score": 0.67, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 153, + 542, + 291, + 554 + ], + "score": 1.0, + "content": ", along with a conditioning signal", + "type": "text" + }, + { + "bbox": [ + 292, + 545, + 299, + 552 + ], + "score": 0.77, + "content": "x", + "type": "inline_equation" + }, + { + "bbox": [ + 299, + 542, + 505, + 554 + ], + "score": 1.0, + "content": ", and predict a less noisy sample. Using Gaussian", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 553, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 553, + 505, + 567 + ], + "score": 1.0, + "content": "noise in the forward process, one can express the training objective over the sequence of transitions", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 120, + 577 + ], + "score": 1.0, + "content": "(as", + "type": "text" + }, + { + "bbox": [ + 120, + 565, + 126, + 574 + ], + "score": 0.71, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 564, + 505, + 577 + ], + "score": 1.0, + "content": "slowly decreases) as a sum of non-linear regression objectives, with the L2 loss (here with the", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 107, + 576, + 189, + 587 + ], + "spans": [ + { + "bbox": [ + 107, + 577, + 112, + 585 + ], + "score": 0.67, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 112, + 576, + 189, + 587 + ], + "score": 1.0, + "content": "-parameterization):", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 28, + "bbox_fs": [ + 105, + 531, + 505, + 587 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 202, + 588, + 409, + 626 + ], + "lines": [ + { + "bbox": [ + 202, + 588, + 409, + 626 + ], + "spans": [ + { + "bbox": [ + 202, + 588, + 409, + 626 + ], + "score": 0.93, + "content": "\\mathbb { E } _ { ( \\pmb { x } , \\pmb { y } ) } \\mathbb { E } _ { ( t , \\epsilon ) } \\bigg \\| f _ { \\theta } ( \\pmb { x } , \\underbrace { \\sqrt { \\gamma _ { t } } \\pmb { y } + \\sqrt { 1 - \\gamma _ { t } } \\pmb { \\epsilon } } _ { \\pmb { y _ { t } } } , t ) - \\pmb { \\epsilon } \\bigg \\| _ { 2 } ^ { 2 }", + "type": "interline_equation", + "image_path": "f16794dc92c133b4267a99986ce44c6bf54d08e321f0f07879b449286dcb8122.jpg" + } + ] + } + ], + "index": 31.5, + "virtual_lines": [ + { + "bbox": [ + 202, + 588, + 409, + 607.0 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 202, + 607.0, + 409, + 626.0 + ], + "spans": [], + "index": 32 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 628, + 505, + 662 + ], + "lines": [ + { + "bbox": [ + 105, + 627, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 134, + 642 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 135, + 629, + 189, + 641 + ], + "score": 0.78, + "content": "\\epsilon \\sim \\mathcal { N } ( 0 , I )", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 627, + 194, + 642 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 194, + 629, + 245, + 641 + ], + "score": 0.83, + "content": "t \\sim \\mathcal { U } ( 0 , 1 )", + "type": "inline_equation" + }, + { + "bbox": [ + 246, + 627, + 297, + 642 + ], + "score": 1.0, + "content": ", and where", + "type": "text" + }, + { + "bbox": [ + 297, + 629, + 330, + 641 + ], + "score": 0.93, + "content": "\\gamma _ { t } ~ > ~ 0", + "type": "inline_equation" + }, + { + "bbox": [ + 330, + 627, + 506, + 642 + ], + "score": 1.0, + "content": "is computed with a pre-determined noise", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 639, + 505, + 652 + ], + "spans": [ + { + "bbox": [ + 105, + 639, + 409, + 652 + ], + "score": 1.0, + "content": "schedule. For inference (i.e., sampling), one draws a random noise sample", + "type": "text" + }, + { + "bbox": [ + 410, + 641, + 421, + 651 + ], + "score": 0.85, + "content": "\\mathbf { \\pmb { y } } _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 421, + 639, + 505, + 652 + ], + "score": 1.0, + "content": ", and then iteratively", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 651, + 489, + 663 + ], + "spans": [ + { + "bbox": [ + 106, + 651, + 126, + 663 + ], + "score": 1.0, + "content": "uses", + "type": "text" + }, + { + "bbox": [ + 126, + 651, + 137, + 662 + ], + "score": 0.87, + "content": "f _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 651, + 433, + 663 + ], + "score": 1.0, + "content": "to estimate the noise, from which one can compute the next latent sample", + "type": "text" + }, + { + "bbox": [ + 433, + 652, + 444, + 662 + ], + "score": 0.85, + "content": "\\mathbf { \\eta } _ { \\mathbf { { 3 } } \\mathbf { { ' } } \\mathbf { { 3 } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 444, + 651, + 462, + 663 + ], + "score": 1.0, + "content": ", for", + "type": "text" + }, + { + "bbox": [ + 462, + 651, + 485, + 661 + ], + "score": 0.88, + "content": "s < t", + "type": "inline_equation" + }, + { + "bbox": [ + 485, + 651, + 489, + 663 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 34, + "bbox_fs": [ + 105, + 627, + 506, + 663 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 667, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 667, + 506, + 679 + ], + "spans": [ + { + "bbox": [ + 106, + 667, + 506, + 679 + ], + "score": 1.0, + "content": "Self-supervised pre-training. Prior work has shown that self-supervised tasks such as coloriza-", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 678, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 678, + 506, + 690 + ], + "score": 1.0, + "content": "tion [33, 86] and masked prediction [80] serve as effective pre-training for downstream vision tasks.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 688, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 506, + 702 + ], + "score": 1.0, + "content": "Our work also confirms the benefit of self-supervised pre-training [54] for diffusion-based image-to-", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 699, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 505, + 712 + ], + "score": 1.0, + "content": "image translation, by establishing a new SOTA on optical flow and monocular depth estimation while", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 711, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 506, + 723 + ], + "score": 1.0, + "content": "also representing multi-modality and supporting zero-shot coarse-to-fine refinement and imputation.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 38, + "bbox_fs": [ + 105, + 667, + 506, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 68, + 504, + 195 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 68, + 504, + 195 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 68, + 504, + 195 + ], + "spans": [ + { + "bbox": [ + 107, + 68, + 504, + 195 + ], + "score": 0.97, + "type": "image", + "image_path": "aa11abb1d3f1a145b38f0348a48d1dcdd3d0b462501e59d1b0010a478ed6a6df.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 68, + 504, + 110.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 110.33333333333334, + 504, + 152.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 152.66666666666669, + 504, + 195.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 203, + 506, + 244 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 203, + 505, + 215 + ], + "spans": [ + { + "bbox": [ + 106, + 203, + 505, + 215 + ], + "score": 1.0, + "content": "Figure 3: Effects of adding synthetic datasets in pretraining. Diffusion models trained only with AutoFlow", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 213, + 506, + 225 + ], + "spans": [ + { + "bbox": [ + 106, + 213, + 506, + 225 + ], + "score": 1.0, + "content": "(AF) tend to provide very coarse flow estimates and can hallucinate shapes. The addition of FlyingThings (FT),", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 222, + 506, + 236 + ], + "spans": [ + { + "bbox": [ + 105, + 222, + 506, + 236 + ], + "score": 1.0, + "content": "Kubric (KU), and TartanAir (TA) remove the AF-induced bias toward polgonal-shaped regions, and significantly", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 234, + 414, + 245 + ], + "spans": [ + { + "bbox": [ + 106, + 234, + 414, + 245 + ], + "score": 1.0, + "content": "improve flow quality on fine detail, e.g. trees, thin structures, and motion boundaries.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "title", + "bbox": [ + 107, + 261, + 221, + 275 + ], + "lines": [ + { + "bbox": [ + 104, + 259, + 223, + 277 + ], + "spans": [ + { + "bbox": [ + 104, + 259, + 223, + 277 + ], + "score": 1.0, + "content": "3 Model Framework", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 107, + 286, + 505, + 342 + ], + "lines": [ + { + "bbox": [ + 106, + 285, + 506, + 299 + ], + "spans": [ + { + "bbox": [ + 106, + 285, + 506, + 299 + ], + "score": 1.0, + "content": "In contrast to the conventional monocular depth and optical flow methods, with rich usage of spe-", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 297, + 505, + 310 + ], + "spans": [ + { + "bbox": [ + 106, + 297, + 505, + 310 + ], + "score": 1.0, + "content": "cialized domain knowledge on their architecture designs, we introduce simple, generic architectures", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 308, + 505, + 320 + ], + "spans": [ + { + "bbox": [ + 106, + 308, + 505, + 320 + ], + "score": 1.0, + "content": "and loss functions. We replace the inductive bias in state-of-the-art architectures and losses with a", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 319, + 506, + 333 + ], + "spans": [ + { + "bbox": [ + 105, + 319, + 506, + 333 + ], + "score": 1.0, + "content": "powerful generative model along with a combination of self-supervised pre-training and supervised", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 330, + 266, + 342 + ], + "spans": [ + { + "bbox": [ + 106, + 330, + 266, + 342 + ], + "score": 1.0, + "content": "training on both real and synthetic data.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 10 + }, + { + "type": "text", + "bbox": [ + 107, + 346, + 505, + 402 + ], + "lines": [ + { + "bbox": [ + 105, + 346, + 506, + 360 + ], + "spans": [ + { + "bbox": [ + 105, + 346, + 506, + 360 + ], + "score": 1.0, + "content": "The denoising diffusion model (Figure 2) takes a noisy version of the target map (i.e., a depth or", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 356, + 505, + 371 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 505, + 371 + ], + "score": 1.0, + "content": "flow) as input, along with the conditioning signal (one RGB image for depth and two RGB images", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 367, + 505, + 381 + ], + "spans": [ + { + "bbox": [ + 105, + 367, + 505, + 381 + ], + "score": 1.0, + "content": "for flow). The denoiser effectively provides a noise-free estimate of the target map (i.e., ignoring the", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 378, + 507, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 507, + 394 + ], + "score": 1.0, + "content": "specific loss parameterization used). The training loss penalizes residual error in the denoised map,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 390, + 486, + 403 + ], + "spans": [ + { + "bbox": [ + 106, + 390, + 486, + 403 + ], + "score": 1.0, + "content": "which is quite distinct from typical image reconstruction losses used in optical flow estimation.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 15 + }, + { + "type": "title", + "bbox": [ + 107, + 412, + 327, + 424 + ], + "lines": [ + { + "bbox": [ + 105, + 411, + 329, + 426 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 329, + 426 + ], + "score": 1.0, + "content": "3.1 Synthetic pre-training data and generalization", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 107, + 428, + 505, + 494 + ], + "lines": [ + { + "bbox": [ + 106, + 428, + 506, + 441 + ], + "spans": [ + { + "bbox": [ + 106, + 428, + 506, + 441 + ], + "score": 1.0, + "content": "Given that we train these models with a generic denoising objective, without task-specific inductive", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 439, + 505, + 451 + ], + "spans": [ + { + "bbox": [ + 106, + 439, + 505, + 451 + ], + "score": 1.0, + "content": "biases in the form of specialized architectures, the choice of training data becomes critical. Below", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "spans": [ + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "score": 1.0, + "content": "we discuss the datasets used and their contributions in detail. Because training data with annotated", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "score": 1.0, + "content": "ground truth is limited for many dense vision tasks, here we make extensive use of synthetic data in", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 471, + 506, + 485 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 485 + ], + "score": 1.0, + "content": "the hope that the geometric properties acquired from synthetic data during training will transfer to", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 482, + 286, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 286, + 496 + ], + "score": 1.0, + "content": "different domains, including natural images.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21.5 + }, + { + "type": "text", + "bbox": [ + 107, + 499, + 505, + 565 + ], + "lines": [ + { + "bbox": [ + 105, + 499, + 505, + 512 + ], + "spans": [ + { + "bbox": [ + 105, + 499, + 505, + 512 + ], + "score": 1.0, + "content": "AutoFlow [69] has recently emerged as a powerful synthetic dataset for training flow models. We", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 510, + 506, + 523 + ], + "spans": [ + { + "bbox": [ + 105, + 510, + 506, + 523 + ], + "score": 1.0, + "content": "were surprised to find that training on AutoFlow alone is insufficient, as the diffusion model appears to", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 520, + 507, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 520, + 507, + 534 + ], + "score": 1.0, + "content": "devote a significant fraction of its representation capacity to represent the shapes of AutoFlow regions,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 532, + 506, + 543 + ], + "spans": [ + { + "bbox": [ + 105, + 532, + 506, + 543 + ], + "score": 1.0, + "content": "rather than solving for correspondence. As a result, models trained on AutoFlow alone exhibit a", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 543, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 105, + 543, + 505, + 555 + ], + "score": 1.0, + "content": "strong bias to generate flow fields with polygonal shaped regions, much like those in AutoFlow, often", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 554, + 440, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 554, + 440, + 567 + ], + "score": 1.0, + "content": "ignoring the shapes of boundaries in the two-frame RGB inputs (e.g. see Figure 3).", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 107, + 570, + 505, + 636 + ], + "lines": [ + { + "bbox": [ + 105, + 569, + 506, + 584 + ], + "spans": [ + { + "bbox": [ + 105, + 569, + 506, + 584 + ], + "score": 1.0, + "content": "To mitigate bias induced by AutoFlow in training, we further mix in three synthetic datasets during", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 582, + 505, + 593 + ], + "spans": [ + { + "bbox": [ + 106, + 582, + 505, + 593 + ], + "score": 1.0, + "content": "training, namely, FlyingThings3D [40], Kubric [19] and TartanAir [76]. Given a model pre-trained on", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 592, + 506, + 605 + ], + "spans": [ + { + "bbox": [ + 105, + 592, + 506, + 605 + ], + "score": 1.0, + "content": "AutoFlow, for compute efficiency, we use a greedy mixing strategy where we fix the relative ratio of", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 603, + 506, + 615 + ], + "spans": [ + { + "bbox": [ + 105, + 603, + 506, + 615 + ], + "score": 1.0, + "content": "the previous mixture and tune the proportion of the newly added dataset. We leave further exploration", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 613, + 506, + 627 + ], + "spans": [ + { + "bbox": [ + 105, + 613, + 506, + 627 + ], + "score": 1.0, + "content": "of an optimal mixing strategy to future work. Zero-shot testing of the model on Sintel and KITTI", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "score": 1.0, + "content": "(see Table 1 and Fig. 3) shows substantial performance gains with each additional synthetic dataset.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 33.5 + }, + { + "type": "text", + "bbox": [ + 107, + 641, + 505, + 675 + ], + "lines": [ + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "score": 1.0, + "content": "We find that pre-training is similarly important for depth estimation (see Table 7). We learn separate", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 652, + 505, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 652, + 505, + 664 + ], + "score": 1.0, + "content": "indoor and outdoor models. For the indoor model we pre-train on a mix of ScanNet [9] and SceneNet", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 662, + 435, + 676 + ], + "spans": [ + { + "bbox": [ + 105, + 662, + 435, + 676 + ], + "score": 1.0, + "content": "RGB-D [41]. The outdoor model is pre-trained on the Waymo Open Dataset [71].", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38 + }, + { + "type": "title", + "bbox": [ + 106, + 685, + 379, + 697 + ], + "lines": [ + { + "bbox": [ + 105, + 683, + 380, + 699 + ], + "spans": [ + { + "bbox": [ + 105, + 683, + 380, + 699 + ], + "score": 1.0, + "content": "3.2 Real data: Challenges with noisy, incomplete ground truth", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 40 + }, + { + "type": "text", + "bbox": [ + 107, + 700, + 506, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 699, + 505, + 713 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 713 + ], + "score": 1.0, + "content": "Ground truth annotations for real-world depth or flow data are often sparse and noisy, due to highly", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "score": 1.0, + "content": "reflective surfaces, light absorbing surfaces [65], dynamic objects [43], etc. While regression-based", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41.5 + } + ], + "page_idx": 3, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 742, + 308, + 750 + ], + "lines": [ + { + "bbox": [ + 301, + 741, + 310, + 752 + ], + "spans": [ + { + "bbox": [ + 301, + 741, + 310, + 752 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 11, + "width": 9 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 68, + 504, + 195 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 68, + 504, + 195 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 68, + 504, + 195 + ], + "spans": [ + { + "bbox": [ + 107, + 68, + 504, + 195 + ], + "score": 0.97, + "type": "image", + "image_path": "aa11abb1d3f1a145b38f0348a48d1dcdd3d0b462501e59d1b0010a478ed6a6df.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 68, + 504, + 110.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 110.33333333333334, + 504, + 152.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 152.66666666666669, + 504, + 195.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 203, + 506, + 244 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 203, + 505, + 215 + ], + "spans": [ + { + "bbox": [ + 106, + 203, + 505, + 215 + ], + "score": 1.0, + "content": "Figure 3: Effects of adding synthetic datasets in pretraining. Diffusion models trained only with AutoFlow", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 213, + 506, + 225 + ], + "spans": [ + { + "bbox": [ + 106, + 213, + 506, + 225 + ], + "score": 1.0, + "content": "(AF) tend to provide very coarse flow estimates and can hallucinate shapes. The addition of FlyingThings (FT),", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 222, + 506, + 236 + ], + "spans": [ + { + "bbox": [ + 105, + 222, + 506, + 236 + ], + "score": 1.0, + "content": "Kubric (KU), and TartanAir (TA) remove the AF-induced bias toward polgonal-shaped regions, and significantly", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 234, + 414, + 245 + ], + "spans": [ + { + "bbox": [ + 106, + 234, + 414, + 245 + ], + "score": 1.0, + "content": "improve flow quality on fine detail, e.g. trees, thin structures, and motion boundaries.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "title", + "bbox": [ + 107, + 261, + 221, + 275 + ], + "lines": [ + { + "bbox": [ + 104, + 259, + 223, + 277 + ], + "spans": [ + { + "bbox": [ + 104, + 259, + 223, + 277 + ], + "score": 1.0, + "content": "3 Model Framework", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 107, + 286, + 505, + 342 + ], + "lines": [ + { + "bbox": [ + 106, + 285, + 506, + 299 + ], + "spans": [ + { + "bbox": [ + 106, + 285, + 506, + 299 + ], + "score": 1.0, + "content": "In contrast to the conventional monocular depth and optical flow methods, with rich usage of spe-", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 297, + 505, + 310 + ], + "spans": [ + { + "bbox": [ + 106, + 297, + 505, + 310 + ], + "score": 1.0, + "content": "cialized domain knowledge on their architecture designs, we introduce simple, generic architectures", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 308, + 505, + 320 + ], + "spans": [ + { + "bbox": [ + 106, + 308, + 505, + 320 + ], + "score": 1.0, + "content": "and loss functions. We replace the inductive bias in state-of-the-art architectures and losses with a", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 319, + 506, + 333 + ], + "spans": [ + { + "bbox": [ + 105, + 319, + 506, + 333 + ], + "score": 1.0, + "content": "powerful generative model along with a combination of self-supervised pre-training and supervised", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 330, + 266, + 342 + ], + "spans": [ + { + "bbox": [ + 106, + 330, + 266, + 342 + ], + "score": 1.0, + "content": "training on both real and synthetic data.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 10, + "bbox_fs": [ + 105, + 285, + 506, + 342 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 346, + 505, + 402 + ], + "lines": [ + { + "bbox": [ + 105, + 346, + 506, + 360 + ], + "spans": [ + { + "bbox": [ + 105, + 346, + 506, + 360 + ], + "score": 1.0, + "content": "The denoising diffusion model (Figure 2) takes a noisy version of the target map (i.e., a depth or", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 356, + 505, + 371 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 505, + 371 + ], + "score": 1.0, + "content": "flow) as input, along with the conditioning signal (one RGB image for depth and two RGB images", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 367, + 505, + 381 + ], + "spans": [ + { + "bbox": [ + 105, + 367, + 505, + 381 + ], + "score": 1.0, + "content": "for flow). The denoiser effectively provides a noise-free estimate of the target map (i.e., ignoring the", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 378, + 507, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 507, + 394 + ], + "score": 1.0, + "content": "specific loss parameterization used). The training loss penalizes residual error in the denoised map,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 390, + 486, + 403 + ], + "spans": [ + { + "bbox": [ + 106, + 390, + 486, + 403 + ], + "score": 1.0, + "content": "which is quite distinct from typical image reconstruction losses used in optical flow estimation.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 15, + "bbox_fs": [ + 105, + 346, + 507, + 403 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 412, + 327, + 424 + ], + "lines": [ + { + "bbox": [ + 105, + 411, + 329, + 426 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 329, + 426 + ], + "score": 1.0, + "content": "3.1 Synthetic pre-training data and generalization", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 107, + 428, + 505, + 494 + ], + "lines": [ + { + "bbox": [ + 106, + 428, + 506, + 441 + ], + "spans": [ + { + "bbox": [ + 106, + 428, + 506, + 441 + ], + "score": 1.0, + "content": "Given that we train these models with a generic denoising objective, without task-specific inductive", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 439, + 505, + 451 + ], + "spans": [ + { + "bbox": [ + 106, + 439, + 505, + 451 + ], + "score": 1.0, + "content": "biases in the form of specialized architectures, the choice of training data becomes critical. Below", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "spans": [ + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "score": 1.0, + "content": "we discuss the datasets used and their contributions in detail. Because training data with annotated", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "score": 1.0, + "content": "ground truth is limited for many dense vision tasks, here we make extensive use of synthetic data in", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 471, + 506, + 485 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 485 + ], + "score": 1.0, + "content": "the hope that the geometric properties acquired from synthetic data during training will transfer to", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 482, + 286, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 286, + 496 + ], + "score": 1.0, + "content": "different domains, including natural images.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21.5, + "bbox_fs": [ + 105, + 428, + 506, + 496 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 499, + 505, + 565 + ], + "lines": [ + { + "bbox": [ + 105, + 499, + 505, + 512 + ], + "spans": [ + { + "bbox": [ + 105, + 499, + 505, + 512 + ], + "score": 1.0, + "content": "AutoFlow [69] has recently emerged as a powerful synthetic dataset for training flow models. We", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 510, + 506, + 523 + ], + "spans": [ + { + "bbox": [ + 105, + 510, + 506, + 523 + ], + "score": 1.0, + "content": "were surprised to find that training on AutoFlow alone is insufficient, as the diffusion model appears to", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 520, + 507, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 520, + 507, + 534 + ], + "score": 1.0, + "content": "devote a significant fraction of its representation capacity to represent the shapes of AutoFlow regions,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 532, + 506, + 543 + ], + "spans": [ + { + "bbox": [ + 105, + 532, + 506, + 543 + ], + "score": 1.0, + "content": "rather than solving for correspondence. As a result, models trained on AutoFlow alone exhibit a", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 543, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 105, + 543, + 505, + 555 + ], + "score": 1.0, + "content": "strong bias to generate flow fields with polygonal shaped regions, much like those in AutoFlow, often", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 554, + 440, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 554, + 440, + 567 + ], + "score": 1.0, + "content": "ignoring the shapes of boundaries in the two-frame RGB inputs (e.g. see Figure 3).", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 27.5, + "bbox_fs": [ + 105, + 499, + 507, + 567 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 570, + 505, + 636 + ], + "lines": [ + { + "bbox": [ + 105, + 569, + 506, + 584 + ], + "spans": [ + { + "bbox": [ + 105, + 569, + 506, + 584 + ], + "score": 1.0, + "content": "To mitigate bias induced by AutoFlow in training, we further mix in three synthetic datasets during", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 582, + 505, + 593 + ], + "spans": [ + { + "bbox": [ + 106, + 582, + 505, + 593 + ], + "score": 1.0, + "content": "training, namely, FlyingThings3D [40], Kubric [19] and TartanAir [76]. Given a model pre-trained on", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 592, + 506, + 605 + ], + "spans": [ + { + "bbox": [ + 105, + 592, + 506, + 605 + ], + "score": 1.0, + "content": "AutoFlow, for compute efficiency, we use a greedy mixing strategy where we fix the relative ratio of", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 603, + 506, + 615 + ], + "spans": [ + { + "bbox": [ + 105, + 603, + 506, + 615 + ], + "score": 1.0, + "content": "the previous mixture and tune the proportion of the newly added dataset. We leave further exploration", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 613, + 506, + 627 + ], + "spans": [ + { + "bbox": [ + 105, + 613, + 506, + 627 + ], + "score": 1.0, + "content": "of an optimal mixing strategy to future work. Zero-shot testing of the model on Sintel and KITTI", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "score": 1.0, + "content": "(see Table 1 and Fig. 3) shows substantial performance gains with each additional synthetic dataset.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 33.5, + "bbox_fs": [ + 105, + 569, + 506, + 637 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 641, + 505, + 675 + ], + "lines": [ + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "score": 1.0, + "content": "We find that pre-training is similarly important for depth estimation (see Table 7). We learn separate", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 652, + 505, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 652, + 505, + 664 + ], + "score": 1.0, + "content": "indoor and outdoor models. For the indoor model we pre-train on a mix of ScanNet [9] and SceneNet", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 662, + 435, + 676 + ], + "spans": [ + { + "bbox": [ + 105, + 662, + 435, + 676 + ], + "score": 1.0, + "content": "RGB-D [41]. The outdoor model is pre-trained on the Waymo Open Dataset [71].", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38, + "bbox_fs": [ + 105, + 641, + 505, + 676 + ] + }, + { + "type": "title", + "bbox": [ + 106, + 685, + 379, + 697 + ], + "lines": [ + { + "bbox": [ + 105, + 683, + 380, + 699 + ], + "spans": [ + { + "bbox": [ + 105, + 683, + 380, + 699 + ], + "score": 1.0, + "content": "3.2 Real data: Challenges with noisy, incomplete ground truth", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 40 + }, + { + "type": "text", + "bbox": [ + 107, + 700, + 506, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 699, + 505, + 713 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 713 + ], + "score": 1.0, + "content": "Ground truth annotations for real-world depth or flow data are often sparse and noisy, due to highly", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "score": 1.0, + "content": "reflective surfaces, light absorbing surfaces [65], dynamic objects [43], etc. While regression-based", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41.5, + "bbox_fs": [ + 106, + 699, + 505, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 72, + 409, + 84 + ], + "lines": [ + { + "bbox": [ + 105, + 69, + 411, + 88 + ], + "spans": [ + { + "bbox": [ + 105, + 69, + 411, + 88 + ], + "score": 1.0, + "content": "Algorithm 1 Denoising diffusion train step with infilling and step unrolling", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 109, + 87, + 452, + 214 + ], + "lines": [ + { + "bbox": [ + 109, + 86, + 450, + 99 + ], + "spans": [ + { + "bbox": [ + 109, + 86, + 121, + 99 + ], + "score": 1.0, + "content": "1:", + "type": "text" + }, + { + "bbox": [ + 122, + 89, + 141, + 97 + ], + "score": 0.76, + "content": "x \\gets", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 86, + 218, + 99 + ], + "score": 1.0, + "content": "conditioning images,", + "type": "text" + }, + { + "bbox": [ + 218, + 89, + 236, + 98 + ], + "score": 0.83, + "content": "y ", + "type": "inline_equation" + }, + { + "bbox": [ + 237, + 86, + 306, + 99 + ], + "score": 1.0, + "content": "flow or depth map,", + "type": "text" + }, + { + "bbox": [ + 307, + 88, + 342, + 97 + ], + "score": 0.59, + "content": "m a s k \\gets", + "type": "inline_equation" + }, + { + "bbox": [ + 343, + 86, + 450, + 99 + ], + "score": 1.0, + "content": "binary mask of known values", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 109, + 96, + 216, + 110 + ], + "spans": [ + { + "bbox": [ + 109, + 96, + 121, + 110 + ], + "score": 1.0, + "content": "2:", + "type": "text" + }, + { + "bbox": [ + 121, + 98, + 216, + 109 + ], + "score": 0.6, + "content": "t \\sim U ( 0 , 1 ) , \\epsilon \\sim N ( 0 , 1 )", + "type": "inline_equation" + } + ], + "index": 2 + }, + { + "bbox": [ + 109, + 107, + 253, + 119 + ], + "spans": [ + { + "bbox": [ + 109, + 107, + 121, + 119 + ], + "score": 1.0, + "content": "3:", + "type": "text" + }, + { + "bbox": [ + 122, + 109, + 136, + 118 + ], + "score": 0.34, + "content": "y =", + "type": "inline_equation" + }, + { + "bbox": [ + 136, + 107, + 253, + 119 + ], + "score": 1.0, + "content": "fill_holes_with_interpolation(y)√", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 108, + 117, + 228, + 130 + ], + "spans": [ + { + "bbox": [ + 108, + 117, + 121, + 130 + ], + "score": 1.0, + "content": "4:", + "type": "text" + }, + { + "bbox": [ + 122, + 118, + 228, + 129 + ], + "score": 0.81, + "content": "y _ { t } = \\sqrt { \\gamma _ { t } } * y + \\sqrt { 1 - \\gamma _ { t } } * \\epsilon", + "type": "inline_equation" + } + ], + "index": 4 + }, + { + "bbox": [ + 109, + 126, + 192, + 139 + ], + "spans": [ + { + "bbox": [ + 109, + 126, + 192, + 139 + ], + "score": 1.0, + "content": "5: if unroll_step then", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 109, + 138, + 262, + 149 + ], + "spans": [ + { + "bbox": [ + 109, + 138, + 120, + 148 + ], + "score": 1.0, + "content": "6:", + "type": "text" + }, + { + "bbox": [ + 134, + 138, + 262, + 149 + ], + "score": 1.0, + "content": "\u000fpred = stop_gradient(fθ(x, yt, t))√", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 109, + 147, + 276, + 159 + ], + "spans": [ + { + "bbox": [ + 109, + 147, + 120, + 158 + ], + "score": 1.0, + "content": "7:", + "type": "text" + }, + { + "bbox": [ + 133, + 147, + 276, + 159 + ], + "score": 1.0, + "content": "ypred = (yt − 1 − γt ∗ \u000fpred)/ γt", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 109, + 158, + 257, + 169 + ], + "spans": [ + { + "bbox": [ + 109, + 158, + 120, + 169 + ], + "score": 1.0, + "content": "8:", + "type": "text" + }, + { + "bbox": [ + 135, + 159, + 257, + 169 + ], + "score": 0.41, + "content": "y _ { t } = \\sqrt { \\gamma _ { t } } * y _ { p r e d } + \\sqrt { 1 - \\gamma _ { t } } * \\epsilon", + "type": "inline_equation", + "image_path": "fed88e2d3ff7d7171dcda51d651b62f5db22ba5fe002d59d56b38f820083f86f.jpg" + } + ], + "index": 8 + }, + { + "bbox": [ + 108, + 169, + 245, + 182 + ], + "spans": [ + { + "bbox": [ + 108, + 169, + 120, + 181 + ], + "score": 1.0, + "content": "9:", + "type": "text" + }, + { + "bbox": [ + 135, + 170, + 245, + 182 + ], + "score": 0.58, + "content": "\\epsilon = ( y _ { t } - { \\sqrt { \\gamma _ { t } } } * y ) / { \\sqrt { 1 - \\gamma _ { t } } }", + "type": "inline_equation" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 179, + 149, + 191 + ], + "spans": [ + { + "bbox": [ + 106, + 179, + 149, + 191 + ], + "score": 1.0, + "content": "10: end if", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 189, + 195, + 202 + ], + "spans": [ + { + "bbox": [ + 106, + 189, + 122, + 202 + ], + "score": 1.0, + "content": "11:", + "type": "text" + }, + { + "bbox": [ + 122, + 190, + 195, + 201 + ], + "score": 0.71, + "content": "\\epsilon _ { p r e d } = f _ { \\theta } ( x , y _ { t } , t )", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 199, + 269, + 212 + ], + "spans": [ + { + "bbox": [ + 106, + 199, + 138, + 212 + ], + "score": 1.0, + "content": "12: loss", + "type": "text" + }, + { + "bbox": [ + 138, + 202, + 146, + 209 + ], + "score": 0.49, + "content": "=", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 199, + 269, + 212 + ], + "score": 1.0, + "content": "reduce_mean(|\u000f − \u000fpred|[mask])", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 106, + 236, + 505, + 378 + ], + "lines": [ + { + "bbox": [ + 105, + 234, + 506, + 250 + ], + "spans": [ + { + "bbox": [ + 105, + 234, + 506, + 250 + ], + "score": 1.0, + "content": "methods can simply compute the loss on pixels with valid ground truth, corruption of the training", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 246, + 505, + 260 + ], + "spans": [ + { + "bbox": [ + 105, + 246, + 505, + 260 + ], + "score": 1.0, + "content": "data is more challenging for diffusion models. Diffusion models perform inference through iterative", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 257, + 505, + 270 + ], + "spans": [ + { + "bbox": [ + 105, + 257, + 219, + 270 + ], + "score": 1.0, + "content": "refinement of the target map", + "type": "text" + }, + { + "bbox": [ + 219, + 260, + 227, + 269 + ], + "score": 0.77, + "content": "\\textbf { { y } }", + "type": "inline_equation" + }, + { + "bbox": [ + 227, + 257, + 356, + 270 + ], + "score": 1.0, + "content": "conditioned on RGB image data", + "type": "text" + }, + { + "bbox": [ + 356, + 260, + 364, + 268 + ], + "score": 0.68, + "content": "_ { \\textbf { \\em x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 364, + 257, + 505, + 270 + ], + "score": 1.0, + "content": ". It starts with a sample of Gaussian", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 268, + 506, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 268, + 129, + 282 + ], + "score": 1.0, + "content": "noise", + "type": "text" + }, + { + "bbox": [ + 130, + 270, + 141, + 280 + ], + "score": 0.85, + "content": "\\mathbf { \\pmb { y } } _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 142, + 268, + 390, + 282 + ], + "score": 1.0, + "content": ", and terminates with a sample from the predictive distribution", + "type": "text" + }, + { + "bbox": [ + 390, + 268, + 427, + 281 + ], + "score": 0.93, + "content": "p ( \\pmb { y } _ { 0 } \\vert \\pmb { x } )", + "type": "inline_equation" + }, + { + "bbox": [ + 427, + 268, + 506, + 282 + ], + "score": 1.0, + "content": ". A refinement step", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 279, + 506, + 293 + ], + "spans": [ + { + "bbox": [ + 105, + 279, + 147, + 293 + ], + "score": 1.0, + "content": "from time", + "type": "text" + }, + { + "bbox": [ + 148, + 281, + 153, + 290 + ], + "score": 0.71, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 154, + 279, + 164, + 293 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 164, + 282, + 170, + 290 + ], + "score": 0.6, + "content": "s", + "type": "inline_equation" + }, + { + "bbox": [ + 171, + 279, + 194, + 293 + ], + "score": 1.0, + "content": ", with", + "type": "text" + }, + { + "bbox": [ + 194, + 281, + 213, + 290 + ], + "score": 0.88, + "content": "s < t", + "type": "inline_equation" + }, + { + "bbox": [ + 213, + 279, + 447, + 293 + ], + "score": 1.0, + "content": ", proceeds by sampling from the parameterized distribution", + "type": "text" + }, + { + "bbox": [ + 447, + 280, + 502, + 292 + ], + "score": 0.93, + "content": "p _ { \\theta } ( \\pmb { y } _ { s } \\mid \\pmb { y } _ { t } , \\pmb { x } )", + "type": "inline_equation" + }, + { + "bbox": [ + 502, + 279, + 506, + 293 + ], + "score": 1.0, + "content": ";", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 290, + 506, + 304 + ], + "spans": [ + { + "bbox": [ + 105, + 290, + 506, + 304 + ], + "score": 1.0, + "content": "i.e., each step operates on the output from the previous step. During training, however, the denoising", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 301, + 506, + 315 + ], + "spans": [ + { + "bbox": [ + 105, + 301, + 506, + 315 + ], + "score": 1.0, + "content": "steps are decoupled (see Eqn. 1), where the denoising network operates on a noisy version of the", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 311, + 506, + 327 + ], + "spans": [ + { + "bbox": [ + 105, + 311, + 506, + 327 + ], + "score": 1.0, + "content": "ground truth depth map instead of the output of the previous iteration (reminiscent of teaching forcing", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 323, + 505, + 337 + ], + "spans": [ + { + "bbox": [ + 105, + 323, + 505, + 337 + ], + "score": 1.0, + "content": "in RNN training [79]). Thus there is a distribution shift between marginals over the noisy target maps", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 334, + 506, + 347 + ], + "spans": [ + { + "bbox": [ + 105, + 334, + 506, + 347 + ], + "score": 1.0, + "content": "during training and inference, because the ground truth maps have missing annotations and heavy-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 345, + 505, + 358 + ], + "spans": [ + { + "bbox": [ + 106, + 345, + 505, + 358 + ], + "score": 1.0, + "content": "tailed sensor noise while the noisy maps obtained from the previous time step at inference time should", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 356, + 505, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 505, + 369 + ], + "score": 1.0, + "content": "not. This distribution shift has a very negative impact on model performance. Nevertheless, with the", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 367, + 484, + 380 + ], + "spans": [ + { + "bbox": [ + 106, + 367, + 484, + 380 + ], + "score": 1.0, + "content": "following modifications during training we find that the problems can be mitigated effectively.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 107, + 383, + 505, + 428 + ], + "lines": [ + { + "bbox": [ + 105, + 383, + 506, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 383, + 506, + 396 + ], + "score": 1.0, + "content": "Infilling. One way to reduce the distribution shift is to impute the missing ground truth. We explored", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 394, + 505, + 406 + ], + "spans": [ + { + "bbox": [ + 105, + 394, + 505, + 406 + ], + "score": 1.0, + "content": "several ways to do this, including simple interpolation schemes, and inference using our model", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 406, + 505, + 417 + ], + "spans": [ + { + "bbox": [ + 106, + 406, + 505, + 417 + ], + "score": 1.0, + "content": "(trained with nearest neighbor interpolation). We find that nearest neighbor interpolation is sufficient", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 417, + 470, + 428 + ], + "spans": [ + { + "bbox": [ + 105, + 417, + 470, + 428 + ], + "score": 1.0, + "content": "to impute missing values in the ground truth maps in the depth and flow field training data.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 107, + 432, + 505, + 466 + ], + "lines": [ + { + "bbox": [ + 105, + 431, + 505, + 446 + ], + "spans": [ + { + "bbox": [ + 105, + 431, + 505, + 446 + ], + "score": 1.0, + "content": "Despite the imputation of missing ground truth depth and flow values, note that the training loss is", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 443, + 505, + 456 + ], + "spans": [ + { + "bbox": [ + 105, + 443, + 505, + 456 + ], + "score": 1.0, + "content": "only computed and backpropagated from pixels with known (not infilled) ground truth depth. We", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 455, + 332, + 467 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 332, + 467 + ], + "score": 1.0, + "content": "refer to this as the masked denoising loss (see Figure 2).", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 107, + 470, + 505, + 602 + ], + "lines": [ + { + "bbox": [ + 106, + 470, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 506, + 484 + ], + "score": 1.0, + "content": "Step-unrolled denoising diffusion training. A second way to mitigate distribution shift in the", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 482, + 506, + 494 + ], + "spans": [ + { + "bbox": [ + 106, + 483, + 116, + 493 + ], + "score": 0.81, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 116, + 482, + 329, + 494 + ], + "score": 1.0, + "content": "marginals in training and inference, is to construct", + "type": "text" + }, + { + "bbox": [ + 330, + 483, + 340, + 493 + ], + "score": 0.85, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 340, + 482, + 506, + 494 + ], + "score": 1.0, + "content": "from model outputs rather than ground", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 493, + 505, + 505 + ], + "spans": [ + { + "bbox": [ + 106, + 493, + 505, + 505 + ], + "score": 1.0, + "content": "truth maps. One can do this by slightly modifying the training procedure (see Algorithm 1) to run", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 504, + 505, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 504, + 274, + 516 + ], + "score": 1.0, + "content": "one forward pass of the model and build", + "type": "text" + }, + { + "bbox": [ + 275, + 505, + 284, + 515 + ], + "score": 0.83, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 285, + 504, + 505, + 516 + ], + "score": 1.0, + "content": "by adding noise to the model’s output rather than the", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 514, + 506, + 527 + ], + "spans": [ + { + "bbox": [ + 105, + 514, + 506, + 527 + ], + "score": 1.0, + "content": "training map. We do not propagate gradients for this forward pass. This process, called step-unrolled", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 525, + 505, + 538 + ], + "spans": [ + { + "bbox": [ + 105, + 525, + 321, + 538 + ], + "score": 1.0, + "content": "denoising diffusion, slows training only marginally", + "type": "text" + }, + { + "bbox": [ + 321, + 525, + 349, + 536 + ], + "score": 0.87, + "content": "( \\sim 1 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 350, + 525, + 505, + 538 + ], + "score": 1.0, + "content": "on a TPU v4). This step-unrolling is", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 537, + 505, + 549 + ], + "spans": [ + { + "bbox": [ + 106, + 537, + 505, + 549 + ], + "score": 1.0, + "content": "akin to the predictor-corrector sampler of [63] which uses an extra Langevin step to improve the", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 547, + 506, + 559 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 228, + 559 + ], + "score": 1.0, + "content": "target marginal distribution of", + "type": "text" + }, + { + "bbox": [ + 228, + 549, + 237, + 558 + ], + "score": 0.83, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 238, + 547, + 506, + 559 + ], + "score": 1.0, + "content": ". Interestingly, the problem of training / inference distribution shift", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 558, + 504, + 570 + ], + "spans": [ + { + "bbox": [ + 106, + 558, + 504, + 570 + ], + "score": 1.0, + "content": "resembles that of exposure bias [51] in autoregressive models, for which the mismatch is caused by", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 568, + 506, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 568, + 506, + 581 + ], + "score": 1.0, + "content": "teacher forcing during training [79]. Several solutions have been proposed for this problem in the", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 579, + 506, + 593 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 506, + 593 + ], + "score": 1.0, + "content": "literature [2, 32, 84]. Step-unrolled denoising diffusion also closely resembles the approach in [57]", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 591, + 282, + 604 + ], + "spans": [ + { + "bbox": [ + 106, + 591, + 282, + 604 + ], + "score": 1.0, + "content": "for training denoising autoencoders on text.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 38.5 + }, + { + "type": "text", + "bbox": [ + 107, + 607, + 505, + 684 + ], + "lines": [ + { + "bbox": [ + 106, + 607, + 505, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 607, + 505, + 619 + ], + "score": 1.0, + "content": "We only perform step-unrolled denoising diffusion during model fine-tuning. Early in training the", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "score": 1.0, + "content": "denoising predictions are inaccurate, so the latent marginals over the noisy target maps will be closer", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 629, + 505, + 641 + ], + "spans": [ + { + "bbox": [ + 105, + 629, + 505, + 641 + ], + "score": 1.0, + "content": "to the desired true marginals than those produced by adding noise to denoiser network outputs. One", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 640, + 506, + 653 + ], + "spans": [ + { + "bbox": [ + 105, + 640, + 506, + 653 + ], + "score": 1.0, + "content": "might consider the use of a curriculum for gradually introducing step-unrolled denoising diffusion in", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 650, + 505, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 650, + 505, + 663 + ], + "score": 1.0, + "content": "the later stages of supervised pre-training, but this introduces additional hyper-parameters, so we", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 662, + 506, + 675 + ], + "spans": [ + { + "bbox": [ + 105, + 662, + 506, + 675 + ], + "score": 1.0, + "content": "simply invoke step-unrolled denoising diffusion during fine-tuning, and leave an exploration of", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 673, + 206, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 673, + 206, + 684 + ], + "score": 1.0, + "content": "curricula to future work.", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 48 + }, + { + "type": "text", + "bbox": [ + 108, + 689, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 688, + 505, + 702 + ], + "spans": [ + { + "bbox": [ + 106, + 689, + 119, + 700 + ], + "score": 0.86, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 119, + 688, + 226, + 702 + ], + "score": 1.0, + "content": "denoiser loss. While the", + "type": "text" + }, + { + "bbox": [ + 226, + 689, + 238, + 700 + ], + "score": 0.88, + "content": "L _ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 688, + 505, + 702 + ], + "score": 1.0, + "content": "loss in Eqn. 1 is ideal for Gaussian noise and noise-free ground", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 106, + 700, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 700, + 505, + 712 + ], + "score": 1.0, + "content": "truth maps, in practice, real ground truth depth and flow fields are noisy and heavy tailed; e.g., for", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 106, + 710, + 505, + 724 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 724 + ], + "score": 1.0, + "content": "distant objects, near object boundaries, and near pixels with missing annotations. We hypothesize", + "type": "text" + } + ], + "index": 54 + } + ], + "index": 53 + } + ], + "page_idx": 4, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 741, + 309, + 750 + ], + "lines": [ + { + "bbox": [ + 301, + 740, + 310, + 753 + ], + "spans": [ + { + "bbox": [ + 301, + 740, + 310, + 753 + ], + "score": 1.0, + "content": "5", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 72, + 409, + 84 + ], + "lines": [ + { + "bbox": [ + 105, + 69, + 411, + 88 + ], + "spans": [ + { + "bbox": [ + 105, + 69, + 411, + 88 + ], + "score": 1.0, + "content": "Algorithm 1 Denoising diffusion train step with infilling and step unrolling", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "index", + "bbox": [ + 109, + 87, + 452, + 214 + ], + "lines": [ + { + "bbox": [ + 109, + 86, + 450, + 99 + ], + "spans": [ + { + "bbox": [ + 109, + 86, + 121, + 99 + ], + "score": 1.0, + "content": "1:", + "type": "text" + }, + { + "bbox": [ + 122, + 89, + 141, + 97 + ], + "score": 0.76, + "content": "x \\gets", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 86, + 218, + 99 + ], + "score": 1.0, + "content": "conditioning images,", + "type": "text" + }, + { + "bbox": [ + 218, + 89, + 236, + 98 + ], + "score": 0.83, + "content": "y ", + "type": "inline_equation" + }, + { + "bbox": [ + 237, + 86, + 306, + 99 + ], + "score": 1.0, + "content": "flow or depth map,", + "type": "text" + }, + { + "bbox": [ + 307, + 88, + 342, + 97 + ], + "score": 0.59, + "content": "m a s k \\gets", + "type": "inline_equation" + }, + { + "bbox": [ + 343, + 86, + 450, + 99 + ], + "score": 1.0, + "content": "binary mask of known values", + "type": "text" + } + ], + "index": 1, + "is_list_start_line": true + }, + { + "bbox": [ + 109, + 96, + 216, + 110 + ], + "spans": [ + { + "bbox": [ + 109, + 96, + 121, + 110 + ], + "score": 1.0, + "content": "2:", + "type": "text" + }, + { + "bbox": [ + 121, + 98, + 216, + 109 + ], + "score": 0.6, + "content": "t \\sim U ( 0 , 1 ) , \\epsilon \\sim N ( 0 , 1 )", + "type": "inline_equation" + } + ], + "index": 2, + "is_list_start_line": true + }, + { + "bbox": [ + 109, + 107, + 253, + 119 + ], + "spans": [ + { + "bbox": [ + 109, + 107, + 121, + 119 + ], + "score": 1.0, + "content": "3:", + "type": "text" + }, + { + "bbox": [ + 122, + 109, + 136, + 118 + ], + "score": 0.34, + "content": "y =", + "type": "inline_equation" + }, + { + "bbox": [ + 136, + 107, + 253, + 119 + ], + "score": 1.0, + "content": "fill_holes_with_interpolation(y)√", + "type": "text" + } + ], + "index": 3, + "is_list_start_line": true + }, + { + "bbox": [ + 108, + 117, + 228, + 130 + ], + "spans": [ + { + "bbox": [ + 108, + 117, + 121, + 130 + ], + "score": 1.0, + "content": "4:", + "type": "text" + }, + { + "bbox": [ + 122, + 118, + 228, + 129 + ], + "score": 0.81, + "content": "y _ { t } = \\sqrt { \\gamma _ { t } } * y + \\sqrt { 1 - \\gamma _ { t } } * \\epsilon", + "type": "inline_equation" + } + ], + "index": 4, + "is_list_start_line": true + }, + { + "bbox": [ + 109, + 126, + 192, + 139 + ], + "spans": [ + { + "bbox": [ + 109, + 126, + 192, + 139 + ], + "score": 1.0, + "content": "5: if unroll_step then", + "type": "text" + } + ], + "index": 5, + "is_list_start_line": true + }, + { + "bbox": [ + 109, + 138, + 262, + 149 + ], + "spans": [ + { + "bbox": [ + 109, + 138, + 120, + 148 + ], + "score": 1.0, + "content": "6:", + "type": "text" + }, + { + "bbox": [ + 134, + 138, + 262, + 149 + ], + "score": 1.0, + "content": "\u000fpred = stop_gradient(fθ(x, yt, t))√", + "type": "text" + } + ], + "index": 6, + "is_list_start_line": true + }, + { + "bbox": [ + 109, + 147, + 276, + 159 + ], + "spans": [ + { + "bbox": [ + 109, + 147, + 120, + 158 + ], + "score": 1.0, + "content": "7:", + "type": "text" + }, + { + "bbox": [ + 133, + 147, + 276, + 159 + ], + "score": 1.0, + "content": "ypred = (yt − 1 − γt ∗ \u000fpred)/ γt", + "type": "text" + } + ], + "index": 7, + "is_list_start_line": true + }, + { + "bbox": [ + 109, + 158, + 257, + 169 + ], + "spans": [ + { + "bbox": [ + 109, + 158, + 120, + 169 + ], + "score": 1.0, + "content": "8:", + "type": "text" + }, + { + "bbox": [ + 135, + 159, + 257, + 169 + ], + "score": 0.41, + "content": "y _ { t } = \\sqrt { \\gamma _ { t } } * y _ { p r e d } + \\sqrt { 1 - \\gamma _ { t } } * \\epsilon", + "type": "inline_equation", + "image_path": "fed88e2d3ff7d7171dcda51d651b62f5db22ba5fe002d59d56b38f820083f86f.jpg" + } + ], + "index": 8, + "is_list_start_line": true + }, + { + "bbox": [ + 108, + 169, + 245, + 182 + ], + "spans": [ + { + "bbox": [ + 108, + 169, + 120, + 181 + ], + "score": 1.0, + "content": "9:", + "type": "text" + }, + { + "bbox": [ + 135, + 170, + 245, + 182 + ], + "score": 0.58, + "content": "\\epsilon = ( y _ { t } - { \\sqrt { \\gamma _ { t } } } * y ) / { \\sqrt { 1 - \\gamma _ { t } } }", + "type": "inline_equation" + } + ], + "index": 9, + "is_list_start_line": true + }, + { + "bbox": [ + 106, + 179, + 149, + 191 + ], + "spans": [ + { + "bbox": [ + 106, + 179, + 149, + 191 + ], + "score": 1.0, + "content": "10: end if", + "type": "text" + } + ], + "index": 10, + "is_list_start_line": true + }, + { + "bbox": [ + 106, + 189, + 195, + 202 + ], + "spans": [ + { + "bbox": [ + 106, + 189, + 122, + 202 + ], + "score": 1.0, + "content": "11:", + "type": "text" + }, + { + "bbox": [ + 122, + 190, + 195, + 201 + ], + "score": 0.71, + "content": "\\epsilon _ { p r e d } = f _ { \\theta } ( x , y _ { t } , t )", + "type": "inline_equation" + } + ], + "index": 11, + "is_list_start_line": true + }, + { + "bbox": [ + 106, + 199, + 269, + 212 + ], + "spans": [ + { + "bbox": [ + 106, + 199, + 138, + 212 + ], + "score": 1.0, + "content": "12: loss", + "type": "text" + }, + { + "bbox": [ + 138, + 202, + 146, + 209 + ], + "score": 0.49, + "content": "=", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 199, + 269, + 212 + ], + "score": 1.0, + "content": "reduce_mean(|\u000f − \u000fpred|[mask])", + "type": "text" + } + ], + "index": 12, + "is_list_start_line": true + } + ], + "index": 6.5, + "bbox_fs": [ + 106, + 86, + 450, + 212 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 236, + 505, + 378 + ], + "lines": [ + { + "bbox": [ + 105, + 234, + 506, + 250 + ], + "spans": [ + { + "bbox": [ + 105, + 234, + 506, + 250 + ], + "score": 1.0, + "content": "methods can simply compute the loss on pixels with valid ground truth, corruption of the training", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 246, + 505, + 260 + ], + "spans": [ + { + "bbox": [ + 105, + 246, + 505, + 260 + ], + "score": 1.0, + "content": "data is more challenging for diffusion models. Diffusion models perform inference through iterative", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 257, + 505, + 270 + ], + "spans": [ + { + "bbox": [ + 105, + 257, + 219, + 270 + ], + "score": 1.0, + "content": "refinement of the target map", + "type": "text" + }, + { + "bbox": [ + 219, + 260, + 227, + 269 + ], + "score": 0.77, + "content": "\\textbf { { y } }", + "type": "inline_equation" + }, + { + "bbox": [ + 227, + 257, + 356, + 270 + ], + "score": 1.0, + "content": "conditioned on RGB image data", + "type": "text" + }, + { + "bbox": [ + 356, + 260, + 364, + 268 + ], + "score": 0.68, + "content": "_ { \\textbf { \\em x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 364, + 257, + 505, + 270 + ], + "score": 1.0, + "content": ". It starts with a sample of Gaussian", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 268, + 506, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 268, + 129, + 282 + ], + "score": 1.0, + "content": "noise", + "type": "text" + }, + { + "bbox": [ + 130, + 270, + 141, + 280 + ], + "score": 0.85, + "content": "\\mathbf { \\pmb { y } } _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 142, + 268, + 390, + 282 + ], + "score": 1.0, + "content": ", and terminates with a sample from the predictive distribution", + "type": "text" + }, + { + "bbox": [ + 390, + 268, + 427, + 281 + ], + "score": 0.93, + "content": "p ( \\pmb { y } _ { 0 } \\vert \\pmb { x } )", + "type": "inline_equation" + }, + { + "bbox": [ + 427, + 268, + 506, + 282 + ], + "score": 1.0, + "content": ". A refinement step", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 279, + 506, + 293 + ], + "spans": [ + { + "bbox": [ + 105, + 279, + 147, + 293 + ], + "score": 1.0, + "content": "from time", + "type": "text" + }, + { + "bbox": [ + 148, + 281, + 153, + 290 + ], + "score": 0.71, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 154, + 279, + 164, + 293 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 164, + 282, + 170, + 290 + ], + "score": 0.6, + "content": "s", + "type": "inline_equation" + }, + { + "bbox": [ + 171, + 279, + 194, + 293 + ], + "score": 1.0, + "content": ", with", + "type": "text" + }, + { + "bbox": [ + 194, + 281, + 213, + 290 + ], + "score": 0.88, + "content": "s < t", + "type": "inline_equation" + }, + { + "bbox": [ + 213, + 279, + 447, + 293 + ], + "score": 1.0, + "content": ", proceeds by sampling from the parameterized distribution", + "type": "text" + }, + { + "bbox": [ + 447, + 280, + 502, + 292 + ], + "score": 0.93, + "content": "p _ { \\theta } ( \\pmb { y } _ { s } \\mid \\pmb { y } _ { t } , \\pmb { x } )", + "type": "inline_equation" + }, + { + "bbox": [ + 502, + 279, + 506, + 293 + ], + "score": 1.0, + "content": ";", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 290, + 506, + 304 + ], + "spans": [ + { + "bbox": [ + 105, + 290, + 506, + 304 + ], + "score": 1.0, + "content": "i.e., each step operates on the output from the previous step. During training, however, the denoising", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 301, + 506, + 315 + ], + "spans": [ + { + "bbox": [ + 105, + 301, + 506, + 315 + ], + "score": 1.0, + "content": "steps are decoupled (see Eqn. 1), where the denoising network operates on a noisy version of the", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 311, + 506, + 327 + ], + "spans": [ + { + "bbox": [ + 105, + 311, + 506, + 327 + ], + "score": 1.0, + "content": "ground truth depth map instead of the output of the previous iteration (reminiscent of teaching forcing", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 323, + 505, + 337 + ], + "spans": [ + { + "bbox": [ + 105, + 323, + 505, + 337 + ], + "score": 1.0, + "content": "in RNN training [79]). Thus there is a distribution shift between marginals over the noisy target maps", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 334, + 506, + 347 + ], + "spans": [ + { + "bbox": [ + 105, + 334, + 506, + 347 + ], + "score": 1.0, + "content": "during training and inference, because the ground truth maps have missing annotations and heavy-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 345, + 505, + 358 + ], + "spans": [ + { + "bbox": [ + 106, + 345, + 505, + 358 + ], + "score": 1.0, + "content": "tailed sensor noise while the noisy maps obtained from the previous time step at inference time should", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 356, + 505, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 505, + 369 + ], + "score": 1.0, + "content": "not. This distribution shift has a very negative impact on model performance. Nevertheless, with the", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 367, + 484, + 380 + ], + "spans": [ + { + "bbox": [ + 106, + 367, + 484, + 380 + ], + "score": 1.0, + "content": "following modifications during training we find that the problems can be mitigated effectively.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 19, + "bbox_fs": [ + 105, + 234, + 506, + 380 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 383, + 505, + 428 + ], + "lines": [ + { + "bbox": [ + 105, + 383, + 506, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 383, + 506, + 396 + ], + "score": 1.0, + "content": "Infilling. One way to reduce the distribution shift is to impute the missing ground truth. We explored", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 394, + 505, + 406 + ], + "spans": [ + { + "bbox": [ + 105, + 394, + 505, + 406 + ], + "score": 1.0, + "content": "several ways to do this, including simple interpolation schemes, and inference using our model", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 406, + 505, + 417 + ], + "spans": [ + { + "bbox": [ + 106, + 406, + 505, + 417 + ], + "score": 1.0, + "content": "(trained with nearest neighbor interpolation). We find that nearest neighbor interpolation is sufficient", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 417, + 470, + 428 + ], + "spans": [ + { + "bbox": [ + 105, + 417, + 470, + 428 + ], + "score": 1.0, + "content": "to impute missing values in the ground truth maps in the depth and flow field training data.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 27.5, + "bbox_fs": [ + 105, + 383, + 506, + 428 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 432, + 505, + 466 + ], + "lines": [ + { + "bbox": [ + 105, + 431, + 505, + 446 + ], + "spans": [ + { + "bbox": [ + 105, + 431, + 505, + 446 + ], + "score": 1.0, + "content": "Despite the imputation of missing ground truth depth and flow values, note that the training loss is", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 443, + 505, + 456 + ], + "spans": [ + { + "bbox": [ + 105, + 443, + 505, + 456 + ], + "score": 1.0, + "content": "only computed and backpropagated from pixels with known (not infilled) ground truth depth. We", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 455, + 332, + 467 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 332, + 467 + ], + "score": 1.0, + "content": "refer to this as the masked denoising loss (see Figure 2).", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31, + "bbox_fs": [ + 105, + 431, + 505, + 467 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 470, + 505, + 602 + ], + "lines": [ + { + "bbox": [ + 106, + 470, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 506, + 484 + ], + "score": 1.0, + "content": "Step-unrolled denoising diffusion training. A second way to mitigate distribution shift in the", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 482, + 506, + 494 + ], + "spans": [ + { + "bbox": [ + 106, + 483, + 116, + 493 + ], + "score": 0.81, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 116, + 482, + 329, + 494 + ], + "score": 1.0, + "content": "marginals in training and inference, is to construct", + "type": "text" + }, + { + "bbox": [ + 330, + 483, + 340, + 493 + ], + "score": 0.85, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 340, + 482, + 506, + 494 + ], + "score": 1.0, + "content": "from model outputs rather than ground", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 493, + 505, + 505 + ], + "spans": [ + { + "bbox": [ + 106, + 493, + 505, + 505 + ], + "score": 1.0, + "content": "truth maps. One can do this by slightly modifying the training procedure (see Algorithm 1) to run", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 504, + 505, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 504, + 274, + 516 + ], + "score": 1.0, + "content": "one forward pass of the model and build", + "type": "text" + }, + { + "bbox": [ + 275, + 505, + 284, + 515 + ], + "score": 0.83, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 285, + 504, + 505, + 516 + ], + "score": 1.0, + "content": "by adding noise to the model’s output rather than the", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 514, + 506, + 527 + ], + "spans": [ + { + "bbox": [ + 105, + 514, + 506, + 527 + ], + "score": 1.0, + "content": "training map. We do not propagate gradients for this forward pass. This process, called step-unrolled", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 525, + 505, + 538 + ], + "spans": [ + { + "bbox": [ + 105, + 525, + 321, + 538 + ], + "score": 1.0, + "content": "denoising diffusion, slows training only marginally", + "type": "text" + }, + { + "bbox": [ + 321, + 525, + 349, + 536 + ], + "score": 0.87, + "content": "( \\sim 1 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 350, + 525, + 505, + 538 + ], + "score": 1.0, + "content": "on a TPU v4). This step-unrolling is", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 537, + 505, + 549 + ], + "spans": [ + { + "bbox": [ + 106, + 537, + 505, + 549 + ], + "score": 1.0, + "content": "akin to the predictor-corrector sampler of [63] which uses an extra Langevin step to improve the", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 547, + 506, + 559 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 228, + 559 + ], + "score": 1.0, + "content": "target marginal distribution of", + "type": "text" + }, + { + "bbox": [ + 228, + 549, + 237, + 558 + ], + "score": 0.83, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 238, + 547, + 506, + 559 + ], + "score": 1.0, + "content": ". Interestingly, the problem of training / inference distribution shift", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 558, + 504, + 570 + ], + "spans": [ + { + "bbox": [ + 106, + 558, + 504, + 570 + ], + "score": 1.0, + "content": "resembles that of exposure bias [51] in autoregressive models, for which the mismatch is caused by", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 568, + 506, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 568, + 506, + 581 + ], + "score": 1.0, + "content": "teacher forcing during training [79]. Several solutions have been proposed for this problem in the", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 579, + 506, + 593 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 506, + 593 + ], + "score": 1.0, + "content": "literature [2, 32, 84]. Step-unrolled denoising diffusion also closely resembles the approach in [57]", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 591, + 282, + 604 + ], + "spans": [ + { + "bbox": [ + 106, + 591, + 282, + 604 + ], + "score": 1.0, + "content": "for training denoising autoencoders on text.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 38.5, + "bbox_fs": [ + 105, + 470, + 506, + 604 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 607, + 505, + 684 + ], + "lines": [ + { + "bbox": [ + 106, + 607, + 505, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 607, + 505, + 619 + ], + "score": 1.0, + "content": "We only perform step-unrolled denoising diffusion during model fine-tuning. Early in training the", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "score": 1.0, + "content": "denoising predictions are inaccurate, so the latent marginals over the noisy target maps will be closer", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 629, + 505, + 641 + ], + "spans": [ + { + "bbox": [ + 105, + 629, + 505, + 641 + ], + "score": 1.0, + "content": "to the desired true marginals than those produced by adding noise to denoiser network outputs. One", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 640, + 506, + 653 + ], + "spans": [ + { + "bbox": [ + 105, + 640, + 506, + 653 + ], + "score": 1.0, + "content": "might consider the use of a curriculum for gradually introducing step-unrolled denoising diffusion in", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 650, + 505, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 650, + 505, + 663 + ], + "score": 1.0, + "content": "the later stages of supervised pre-training, but this introduces additional hyper-parameters, so we", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 662, + 506, + 675 + ], + "spans": [ + { + "bbox": [ + 105, + 662, + 506, + 675 + ], + "score": 1.0, + "content": "simply invoke step-unrolled denoising diffusion during fine-tuning, and leave an exploration of", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 673, + 206, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 673, + 206, + 684 + ], + "score": 1.0, + "content": "curricula to future work.", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 48, + "bbox_fs": [ + 105, + 607, + 506, + 684 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 689, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 688, + 505, + 702 + ], + "spans": [ + { + "bbox": [ + 106, + 689, + 119, + 700 + ], + "score": 0.86, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 119, + 688, + 226, + 702 + ], + "score": 1.0, + "content": "denoiser loss. While the", + "type": "text" + }, + { + "bbox": [ + 226, + 689, + 238, + 700 + ], + "score": 0.88, + "content": "L _ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 688, + 505, + 702 + ], + "score": 1.0, + "content": "loss in Eqn. 1 is ideal for Gaussian noise and noise-free ground", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 106, + 700, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 700, + 505, + 712 + ], + "score": 1.0, + "content": "truth maps, in practice, real ground truth depth and flow fields are noisy and heavy tailed; e.g., for", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 106, + 710, + 505, + 724 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 724 + ], + "score": 1.0, + "content": "distant objects, near object boundaries, and near pixels with missing annotations. We hypothesize", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 105, + 356, + 506, + 371 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 247, + 371 + ], + "score": 1.0, + "content": "that the robustness afforded by the", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 247, + 358, + 259, + 369 + ], + "score": 0.88, + "content": "L _ { 1 }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 260, + 356, + 506, + 371 + ], + "score": 1.0, + "content": "loss may therefore be useful in training the neural denoising", + "type": "text", + "cross_page": true + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 369, + 505, + 381 + ], + "spans": [ + { + "bbox": [ + 106, + 369, + 505, + 381 + ], + "score": 1.0, + "content": "network. (See Tables 11 and 12 in the supplementary material for an ablation of the loss function for", + "type": "text", + "cross_page": true + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 380, + 226, + 392 + ], + "spans": [ + { + "bbox": [ + 106, + 380, + 226, + 392 + ], + "score": 1.0, + "content": "monocular depth estimation.)", + "type": "text", + "cross_page": true + } + ], + "index": 7 + } + ], + "index": 53, + "bbox_fs": [ + 106, + 688, + 505, + 724 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 68, + 504, + 315 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 68, + 504, + 315 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 68, + 504, + 315 + ], + "spans": [ + { + "bbox": [ + 107, + 68, + 504, + 315 + ], + "score": 0.968, + "type": "image", + "image_path": "d14637fbc0cc84db48eb726d5c2b32c9adae0c3d94b7f2bacb17aa9a739c9a3a.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 68, + 504, + 150.33333333333331 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 150.33333333333331, + 504, + 232.66666666666663 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 232.66666666666663, + 504, + 314.99999999999994 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 321, + 503, + 342 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 321, + 505, + 334 + ], + "spans": [ + { + "bbox": [ + 106, + 321, + 505, + 334 + ], + "score": 1.0, + "content": "Figure 4: Visual results comparing RAFT with our method after pretraining. Note that our method does", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 331, + 291, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 331, + 291, + 343 + ], + "score": 1.0, + "content": "much better on fine details and ambiguous regions.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "text", + "bbox": [ + 107, + 357, + 505, + 391 + ], + "lines": [ + { + "bbox": [ + 105, + 356, + 506, + 371 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 247, + 371 + ], + "score": 1.0, + "content": "that the robustness afforded by the", + "type": "text" + }, + { + "bbox": [ + 247, + 358, + 259, + 369 + ], + "score": 0.88, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 260, + 356, + 506, + 371 + ], + "score": 1.0, + "content": "loss may therefore be useful in training the neural denoising", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 369, + 505, + 381 + ], + "spans": [ + { + "bbox": [ + 106, + 369, + 505, + 381 + ], + "score": 1.0, + "content": "network. (See Tables 11 and 12 in the supplementary material for an ablation of the loss function for", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 380, + 226, + 392 + ], + "spans": [ + { + "bbox": [ + 106, + 380, + 226, + 392 + ], + "score": 1.0, + "content": "monocular depth estimation.)", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6 + }, + { + "type": "title", + "bbox": [ + 107, + 405, + 239, + 416 + ], + "lines": [ + { + "bbox": [ + 105, + 404, + 240, + 418 + ], + "spans": [ + { + "bbox": [ + 105, + 404, + 240, + 418 + ], + "score": 1.0, + "content": "3.3 Coarse-to-fine refinement", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 107, + 423, + 505, + 554 + ], + "lines": [ + { + "bbox": [ + 105, + 421, + 505, + 437 + ], + "spans": [ + { + "bbox": [ + 105, + 421, + 505, + 437 + ], + "score": 1.0, + "content": "Training high resolution diffusion models is often slow and memory intensive but increasing the image", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 433, + 505, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 433, + 505, + 447 + ], + "score": 1.0, + "content": "resolution of the model has been shown to improve performance on vision tasks [18]. One simple", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 443, + 506, + 458 + ], + "spans": [ + { + "bbox": [ + 105, + 443, + 506, + 458 + ], + "score": 1.0, + "content": "solution, yielding high-resolution output without increasing the training cost, is to perform inference", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 455, + 505, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 505, + 468 + ], + "score": 1.0, + "content": "in a coarse-to-fine manner, first estimating flow over the entire field of view at low resolution, and then", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 466, + 505, + 478 + ], + "spans": [ + { + "bbox": [ + 105, + 466, + 505, + 478 + ], + "score": 1.0, + "content": "refining the estimates in a patch-wise manner. For refinement, we first up-sample the low-resolution", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 478, + 505, + 489 + ], + "spans": [ + { + "bbox": [ + 105, + 478, + 505, + 489 + ], + "score": 1.0, + "content": "map to the target resolution using bicubic interpolation. Patches are cropped from the up-scaled", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 488, + 505, + 501 + ], + "spans": [ + { + "bbox": [ + 105, + 488, + 163, + 501 + ], + "score": 1.0, + "content": "map, denoted", + "type": "text" + }, + { + "bbox": [ + 163, + 490, + 169, + 498 + ], + "score": 0.72, + "content": "z", + "type": "inline_equation" + }, + { + "bbox": [ + 170, + 488, + 505, + 501 + ], + "score": 1.0, + "content": ", along with the corresponding RGB inputs. Then we run diffusion model inference", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 499, + 505, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 499, + 171, + 513 + ], + "score": 1.0, + "content": "starting at time", + "type": "text" + }, + { + "bbox": [ + 171, + 500, + 179, + 509 + ], + "score": 0.82, + "content": "t ^ { \\prime }", + "type": "inline_equation" + }, + { + "bbox": [ + 179, + 499, + 254, + 513 + ], + "score": 1.0, + "content": "with a noisy map", + "type": "text" + }, + { + "bbox": [ + 254, + 499, + 387, + 512 + ], + "score": 0.92, + "content": "y _ { t ^ { \\prime } } \\sim \\mathcal { N } ( y _ { t ^ { \\prime } } ; \\sqrt { \\gamma _ { t ^ { \\prime } } } z , ( 1 - \\gamma _ { t ^ { \\prime } } ) I )", + "type": "inline_equation" + }, + { + "bbox": [ + 388, + 499, + 455, + 513 + ], + "score": 1.0, + "content": ". For simplicity,", + "type": "text" + }, + { + "bbox": [ + 456, + 499, + 464, + 510 + ], + "score": 0.81, + "content": "t ^ { \\prime }", + "type": "inline_equation" + }, + { + "bbox": [ + 464, + 499, + 505, + 513 + ], + "score": 1.0, + "content": "is a fixed", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 509, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 105, + 509, + 505, + 524 + ], + "score": 1.0, + "content": "hyper-parameter, set based on a validation set. This process is carried out for multiple overlapping", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 521, + 505, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 521, + 505, + 534 + ], + "score": 1.0, + "content": "patches. Following Perceiver IO [28], the patch estimates are merged using weighted masks with", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 532, + 505, + 544 + ], + "spans": [ + { + "bbox": [ + 105, + 532, + 505, + 544 + ], + "score": 1.0, + "content": "lower weight near the patch boundaries since predictions at boundaries are more prone to errors. (See", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 543, + 227, + 554 + ], + "spans": [ + { + "bbox": [ + 106, + 543, + 227, + 554 + ], + "score": 1.0, + "content": "Section H.5 for more details.)", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 14.5 + }, + { + "type": "title", + "bbox": [ + 107, + 570, + 191, + 584 + ], + "lines": [ + { + "bbox": [ + 104, + 568, + 193, + 588 + ], + "spans": [ + { + "bbox": [ + 104, + 568, + 193, + 588 + ], + "score": 1.0, + "content": "4 Experiments", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 21 + }, + { + "type": "text", + "bbox": [ + 107, + 596, + 505, + 662 + ], + "lines": [ + { + "bbox": [ + 106, + 596, + 505, + 608 + ], + "spans": [ + { + "bbox": [ + 106, + 596, + 505, + 608 + ], + "score": 1.0, + "content": "As our denoiser backbone, we adopt the Efficient UNet architecture [55], pretrained with Palette", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 606, + 505, + 620 + ], + "spans": [ + { + "bbox": [ + 105, + 606, + 505, + 620 + ], + "score": 1.0, + "content": "[54] style self-supervised pretraining, and slightly modified to have the appropriate input and output", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 617, + 505, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 505, + 631 + ], + "score": 1.0, + "content": "channels for each task. Since diffusion models expect inputs and generate outputs in the range", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 107, + 628, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 107, + 629, + 141, + 641 + ], + "score": 0.81, + "content": "[ - 1 . , 1 . ]", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 628, + 506, + 642 + ], + "score": 1.0, + "content": ", we normalize depths using max depth of 10 meters and 80 meters respectively for the indoor", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 640, + 505, + 651 + ], + "spans": [ + { + "bbox": [ + 106, + 640, + 505, + 651 + ], + "score": 1.0, + "content": "and outdoor models. We normalize the flow using the height and width of the ground truth. Refer to", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 650, + 468, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 650, + 468, + 663 + ], + "score": 1.0, + "content": "Section H for more details on the architecture, augmentations and other hyper-parameters.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 24.5 + }, + { + "type": "text", + "bbox": [ + 107, + 667, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 667, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 667, + 447, + 678 + ], + "score": 1.0, + "content": "Optical flow. We pre-train on the mixture described in Section 3.1 at a resolution of", + "type": "text" + }, + { + "bbox": [ + 447, + 667, + 487, + 678 + ], + "score": 0.89, + "content": "3 2 0 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 487, + 667, + 505, + 678 + ], + "score": 1.0, + "content": "and", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 677, + 505, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 505, + 691 + ], + "score": 1.0, + "content": "report zero-shot results on the widely used Sintel [4] and KITTI [44] datasets. We further fine-tune", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 687, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 702 + ], + "score": 1.0, + "content": "this model on the standard mixture consisting of AutoFlow [69], FlyingThings [40], VIPER [53],", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 699, + 506, + 713 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 306, + 713 + ], + "score": 1.0, + "content": "HD1K [30], Sintel and KITTI at a resolution of", + "type": "text" + }, + { + "bbox": [ + 306, + 700, + 346, + 711 + ], + "score": 0.89, + "content": "3 2 0 \\times 7 6 8", + "type": "inline_equation" + }, + { + "bbox": [ + 347, + 699, + 506, + 713 + ], + "score": 1.0, + "content": "and report results on the test set from", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 710, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 723 + ], + "score": 1.0, + "content": "the public benchmark. We use a standard average end-point error (AEPE) metric that calculates L2", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 30 + } + ], + "page_idx": 5, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 742, + 308, + 750 + ], + "lines": [ + { + "bbox": [ + 302, + 741, + 309, + 752 + ], + "spans": [ + { + "bbox": [ + 302, + 741, + 309, + 752 + ], + "score": 1.0, + "content": "6", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 68, + 504, + 315 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 68, + 504, + 315 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 68, + 504, + 315 + ], + "spans": [ + { + "bbox": [ + 107, + 68, + 504, + 315 + ], + "score": 0.968, + "type": "image", + "image_path": "d14637fbc0cc84db48eb726d5c2b32c9adae0c3d94b7f2bacb17aa9a739c9a3a.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 68, + 504, + 150.33333333333331 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 150.33333333333331, + 504, + 232.66666666666663 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 232.66666666666663, + 504, + 314.99999999999994 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 321, + 503, + 342 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 321, + 505, + 334 + ], + "spans": [ + { + "bbox": [ + 106, + 321, + 505, + 334 + ], + "score": 1.0, + "content": "Figure 4: Visual results comparing RAFT with our method after pretraining. Note that our method does", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 331, + 291, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 331, + 291, + 343 + ], + "score": 1.0, + "content": "much better on fine details and ambiguous regions.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "text", + "bbox": [ + 107, + 357, + 505, + 391 + ], + "lines": [], + "index": 6, + "bbox_fs": [ + 105, + 356, + 506, + 392 + ], + "lines_deleted": true + }, + { + "type": "title", + "bbox": [ + 107, + 405, + 239, + 416 + ], + "lines": [ + { + "bbox": [ + 105, + 404, + 240, + 418 + ], + "spans": [ + { + "bbox": [ + 105, + 404, + 240, + 418 + ], + "score": 1.0, + "content": "3.3 Coarse-to-fine refinement", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 107, + 423, + 505, + 554 + ], + "lines": [ + { + "bbox": [ + 105, + 421, + 505, + 437 + ], + "spans": [ + { + "bbox": [ + 105, + 421, + 505, + 437 + ], + "score": 1.0, + "content": "Training high resolution diffusion models is often slow and memory intensive but increasing the image", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 433, + 505, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 433, + 505, + 447 + ], + "score": 1.0, + "content": "resolution of the model has been shown to improve performance on vision tasks [18]. One simple", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 443, + 506, + 458 + ], + "spans": [ + { + "bbox": [ + 105, + 443, + 506, + 458 + ], + "score": 1.0, + "content": "solution, yielding high-resolution output without increasing the training cost, is to perform inference", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 455, + 505, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 505, + 468 + ], + "score": 1.0, + "content": "in a coarse-to-fine manner, first estimating flow over the entire field of view at low resolution, and then", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 466, + 505, + 478 + ], + "spans": [ + { + "bbox": [ + 105, + 466, + 505, + 478 + ], + "score": 1.0, + "content": "refining the estimates in a patch-wise manner. For refinement, we first up-sample the low-resolution", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 478, + 505, + 489 + ], + "spans": [ + { + "bbox": [ + 105, + 478, + 505, + 489 + ], + "score": 1.0, + "content": "map to the target resolution using bicubic interpolation. Patches are cropped from the up-scaled", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 488, + 505, + 501 + ], + "spans": [ + { + "bbox": [ + 105, + 488, + 163, + 501 + ], + "score": 1.0, + "content": "map, denoted", + "type": "text" + }, + { + "bbox": [ + 163, + 490, + 169, + 498 + ], + "score": 0.72, + "content": "z", + "type": "inline_equation" + }, + { + "bbox": [ + 170, + 488, + 505, + 501 + ], + "score": 1.0, + "content": ", along with the corresponding RGB inputs. Then we run diffusion model inference", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 499, + 505, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 499, + 171, + 513 + ], + "score": 1.0, + "content": "starting at time", + "type": "text" + }, + { + "bbox": [ + 171, + 500, + 179, + 509 + ], + "score": 0.82, + "content": "t ^ { \\prime }", + "type": "inline_equation" + }, + { + "bbox": [ + 179, + 499, + 254, + 513 + ], + "score": 1.0, + "content": "with a noisy map", + "type": "text" + }, + { + "bbox": [ + 254, + 499, + 387, + 512 + ], + "score": 0.92, + "content": "y _ { t ^ { \\prime } } \\sim \\mathcal { N } ( y _ { t ^ { \\prime } } ; \\sqrt { \\gamma _ { t ^ { \\prime } } } z , ( 1 - \\gamma _ { t ^ { \\prime } } ) I )", + "type": "inline_equation" + }, + { + "bbox": [ + 388, + 499, + 455, + 513 + ], + "score": 1.0, + "content": ". For simplicity,", + "type": "text" + }, + { + "bbox": [ + 456, + 499, + 464, + 510 + ], + "score": 0.81, + "content": "t ^ { \\prime }", + "type": "inline_equation" + }, + { + "bbox": [ + 464, + 499, + 505, + 513 + ], + "score": 1.0, + "content": "is a fixed", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 509, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 105, + 509, + 505, + 524 + ], + "score": 1.0, + "content": "hyper-parameter, set based on a validation set. This process is carried out for multiple overlapping", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 521, + 505, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 521, + 505, + 534 + ], + "score": 1.0, + "content": "patches. Following Perceiver IO [28], the patch estimates are merged using weighted masks with", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 532, + 505, + 544 + ], + "spans": [ + { + "bbox": [ + 105, + 532, + 505, + 544 + ], + "score": 1.0, + "content": "lower weight near the patch boundaries since predictions at boundaries are more prone to errors. (See", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 543, + 227, + 554 + ], + "spans": [ + { + "bbox": [ + 106, + 543, + 227, + 554 + ], + "score": 1.0, + "content": "Section H.5 for more details.)", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 14.5, + "bbox_fs": [ + 105, + 421, + 506, + 554 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 570, + 191, + 584 + ], + "lines": [ + { + "bbox": [ + 104, + 568, + 193, + 588 + ], + "spans": [ + { + "bbox": [ + 104, + 568, + 193, + 588 + ], + "score": 1.0, + "content": "4 Experiments", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 21 + }, + { + "type": "text", + "bbox": [ + 107, + 596, + 505, + 662 + ], + "lines": [ + { + "bbox": [ + 106, + 596, + 505, + 608 + ], + "spans": [ + { + "bbox": [ + 106, + 596, + 505, + 608 + ], + "score": 1.0, + "content": "As our denoiser backbone, we adopt the Efficient UNet architecture [55], pretrained with Palette", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 606, + 505, + 620 + ], + "spans": [ + { + "bbox": [ + 105, + 606, + 505, + 620 + ], + "score": 1.0, + "content": "[54] style self-supervised pretraining, and slightly modified to have the appropriate input and output", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 617, + 505, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 505, + 631 + ], + "score": 1.0, + "content": "channels for each task. Since diffusion models expect inputs and generate outputs in the range", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 107, + 628, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 107, + 629, + 141, + 641 + ], + "score": 0.81, + "content": "[ - 1 . , 1 . ]", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 628, + 506, + 642 + ], + "score": 1.0, + "content": ", we normalize depths using max depth of 10 meters and 80 meters respectively for the indoor", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 640, + 505, + 651 + ], + "spans": [ + { + "bbox": [ + 106, + 640, + 505, + 651 + ], + "score": 1.0, + "content": "and outdoor models. We normalize the flow using the height and width of the ground truth. Refer to", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 650, + 468, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 650, + 468, + 663 + ], + "score": 1.0, + "content": "Section H for more details on the architecture, augmentations and other hyper-parameters.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 24.5, + "bbox_fs": [ + 105, + 596, + 506, + 663 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 667, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 667, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 667, + 447, + 678 + ], + "score": 1.0, + "content": "Optical flow. We pre-train on the mixture described in Section 3.1 at a resolution of", + "type": "text" + }, + { + "bbox": [ + 447, + 667, + 487, + 678 + ], + "score": 0.89, + "content": "3 2 0 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 487, + 667, + 505, + 678 + ], + "score": 1.0, + "content": "and", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 677, + 505, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 505, + 691 + ], + "score": 1.0, + "content": "report zero-shot results on the widely used Sintel [4] and KITTI [44] datasets. We further fine-tune", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 687, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 702 + ], + "score": 1.0, + "content": "this model on the standard mixture consisting of AutoFlow [69], FlyingThings [40], VIPER [53],", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 699, + 506, + 713 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 306, + 713 + ], + "score": 1.0, + "content": "HD1K [30], Sintel and KITTI at a resolution of", + "type": "text" + }, + { + "bbox": [ + 306, + 700, + 346, + 711 + ], + "score": 0.89, + "content": "3 2 0 \\times 7 6 8", + "type": "inline_equation" + }, + { + "bbox": [ + 347, + 699, + 506, + 713 + ], + "score": 1.0, + "content": "and report results on the test set from", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 710, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 723 + ], + "score": 1.0, + "content": "the public benchmark. We use a standard average end-point error (AEPE) metric that calculates L2", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 455, + 506, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 506, + 468 + ], + "score": 1.0, + "content": "distance between ground truth and prediction. On KITTI, we additionally use the outlier rate, Fl-all,", + "type": "text", + "cross_page": true + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 467, + 506, + 479 + ], + "spans": [ + { + "bbox": [ + 106, + 467, + 239, + 479 + ], + "score": 1.0, + "content": "which reports the outlier ratio in", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 240, + 467, + 250, + 477 + ], + "score": 0.85, + "content": "\\%", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 250, + 467, + 506, + 479 + ], + "score": 1.0, + "content": "among all pixels with valid ground truth, where an estimate is", + "type": "text", + "cross_page": true + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 478, + 437, + 490 + ], + "spans": [ + { + "bbox": [ + 105, + 478, + 329, + 490 + ], + "score": 1.0, + "content": "considered as an outlier if its error exceeds 3 pixels and", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 330, + 478, + 345, + 488 + ], + "score": 0.87, + "content": "5 \\%", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 345, + 478, + 437, + 490 + ], + "score": 1.0, + "content": "w.r.t. the ground truth.", + "type": "text", + "cross_page": true + } + ], + "index": 31 + } + ], + "index": 30, + "bbox_fs": [ + 105, + 667, + 506, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 67, + 505, + 240 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 67, + 505, + 240 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 67, + 505, + 240 + ], + "spans": [ + { + "bbox": [ + 107, + 67, + 505, + 240 + ], + "score": 0.96, + "type": "image", + "image_path": "e4d6c5005bb47f266c2f4f71dc972c3f7e2c3803ce41a7e60f9eb15e36eb3af7.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 67, + 505, + 124.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 124.66666666666666, + 505, + 182.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 182.33333333333331, + 505, + 239.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 249, + 503, + 270 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 248, + 505, + 261 + ], + "spans": [ + { + "bbox": [ + 105, + 248, + 505, + 261 + ], + "score": 1.0, + "content": "Figure 5: Visual results comparing RAFT with our method after finetuning. Ours does much better on fine", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 259, + 220, + 271 + ], + "spans": [ + { + "bbox": [ + 105, + 259, + 220, + 271 + ], + "score": 1.0, + "content": "details and ambiguous regions.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "table", + "bbox": [ + 106, + 351, + 331, + 433 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 286, + 327, + 346 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 287, + 326, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 287, + 326, + 298 + ], + "score": 1.0, + "content": "Table 1: Zero-shot optical flow estimation results on Sintel", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 296, + 327, + 308 + ], + "spans": [ + { + "bbox": [ + 105, + 296, + 327, + 308 + ], + "score": 1.0, + "content": "and KITTI. We provide a new RAFT baseline using our", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 307, + 327, + 318 + ], + "spans": [ + { + "bbox": [ + 106, + 307, + 327, + 318 + ], + "score": 1.0, + "content": "proposed pre-training mixture and substantially improve the", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 317, + 327, + 328 + ], + "spans": [ + { + "bbox": [ + 105, + 317, + 327, + 328 + ], + "score": 1.0, + "content": "accuracy over the original. Our diffusion model outperforms", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 327, + 327, + 337 + ], + "spans": [ + { + "bbox": [ + 106, + 327, + 327, + 337 + ], + "score": 1.0, + "content": "even this much stronger baseline and achieves state-of-the-art", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 337, + 263, + 346 + ], + "spans": [ + { + "bbox": [ + 106, + 337, + 263, + 346 + ], + "score": 1.0, + "content": "zero-shot results on Sintel.final and KITTI.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 7.5 + }, + { + "type": "table_body", + "bbox": [ + 106, + 351, + 331, + 433 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 351, + 331, + 433 + ], + "spans": [ + { + "bbox": [ + 106, + 351, + 331, + 433 + ], + "score": 0.977, + "html": "
ModelDatasetSintel.clean Sintel.finalKITTI
AEPEAEPEFl-all
FlowFormer RAFTChairs-→Things1.012.404.0914.72%
Chairs-→Things1.682.805.921
Perceiver IO RAFTAutoFlow1.812.424.98-
AutoFlow1.742.414.1813.41%
RAFT(ours)AF→AF+FT+KU+TA1.272.282.719.16%
DDVM(ours) AF-→AF+FT+KU+TA1.242.002.197.58%
", + "type": "table", + "image_path": "cded639441183179955a6aba42079d7c1357b21fa461a5f4fa42756628a0f7fe.jpg" + } + ] + } + ], + "index": 22.0, + "virtual_lines": [ + { + "bbox": [ + 106, + 351, + 331, + 364.6666666666667 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 106, + 364.6666666666667, + 331, + 378.33333333333337 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 106, + 378.33333333333337, + 331, + 392.00000000000006 + ], + "spans": [], + "index": 21 + }, + { + "bbox": [ + 106, + 392.00000000000006, + 331, + 405.66666666666674 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 106, + 405.66666666666674, + 331, + 419.3333333333334 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 106, + 419.3333333333334, + 331, + 433.0000000000001 + ], + "spans": [], + "index": 27 + } + ] + } + ], + "index": 14.75 + }, + { + "type": "table", + "bbox": [ + 344, + 363, + 504, + 433 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 344, + 286, + 505, + 357 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 343, + 286, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 343, + 286, + 505, + 298 + ], + "score": 1.0, + "content": "Table 2: Optical flow finetuning evaluation", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 343, + 297, + 505, + 308 + ], + "spans": [ + { + "bbox": [ + 343, + 297, + 459, + 308 + ], + "score": 1.0, + "content": "on public benchmark datasets", + "type": "text" + }, + { + "bbox": [ + 459, + 297, + 491, + 307 + ], + "score": 0.65, + "content": "( \\mathrm { A E P E } \\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 491, + 297, + 505, + 308 + ], + "score": 1.0, + "content": "for", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 343, + 307, + 505, + 317 + ], + "spans": [ + { + "bbox": [ + 343, + 307, + 505, + 317 + ], + "score": 1.0, + "content": "Sintel and Fl-all↓ for KITTI). Bold indicates", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 343, + 315, + 507, + 327 + ], + "spans": [ + { + "bbox": [ + 343, + 315, + 436, + 327 + ], + "score": 1.0, + "content": "the best and underline the", + "type": "text" + }, + { + "bbox": [ + 437, + 316, + 449, + 326 + ], + "score": 0.87, + "content": "2 ^ { \\mathrm { n d } }", + "type": "inline_equation" + }, + { + "bbox": [ + 449, + 315, + 507, + 327 + ], + "score": 1.0, + "content": "-best. § uses ex-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 343, + 326, + 506, + 338 + ], + "spans": [ + { + "bbox": [ + 343, + 326, + 506, + 338 + ], + "score": 1.0, + "content": "tra datasets (AutoFlow and VIPER) on top of", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 343, + 335, + 505, + 347 + ], + "spans": [ + { + "bbox": [ + 343, + 335, + 505, + 347 + ], + "score": 1.0, + "content": "defaults (FlyingThings, HD1K, KITTI, and", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 344, + 347, + 472, + 357 + ], + "spans": [ + { + "bbox": [ + 344, + 347, + 472, + 357 + ], + "score": 1.0, + "content": "Sintel). ∗uses warm start on Sintel.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 14 + }, + { + "type": "table_body", + "bbox": [ + 344, + 363, + 504, + 433 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 344, + 363, + 504, + 433 + ], + "spans": [ + { + "bbox": [ + 344, + 363, + 504, + 433 + ], + "score": 0.975, + "html": "
MethodSintel.cleanSintel.finalKITTI
SKFlow [72]*1.302.264.84%
CRAFT [66]*1.442.424.79%
FlowFormer [23]1.142.184.68%
RAFT-OCTC [29]*1.512.574.33%
RAFT-it [70]$1.552.904.31%
DDVM (ours)$1.752.483.26%
", + "type": "table", + "image_path": "1a0c1cc34d71e3177749b62627c99338c09494c68c5d32ead929543042c0f6aa.jpg" + } + ] + } + ], + "index": 24, + "virtual_lines": [ + { + "bbox": [ + 344, + 363, + 504, + 377.0 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 344, + 377.0, + 504, + 391.0 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 344, + 391.0, + 504, + 405.0 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 344, + 405.0, + 504, + 419.0 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 344, + 419.0, + 504, + 433.0 + ], + "spans": [], + "index": 28 + } + ] + } + ], + "index": 19.0 + }, + { + "type": "text", + "bbox": [ + 107, + 456, + 505, + 489 + ], + "lines": [ + { + "bbox": [ + 105, + 455, + 506, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 506, + 468 + ], + "score": 1.0, + "content": "distance between ground truth and prediction. On KITTI, we additionally use the outlier rate, Fl-all,", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 467, + 506, + 479 + ], + "spans": [ + { + "bbox": [ + 106, + 467, + 239, + 479 + ], + "score": 1.0, + "content": "which reports the outlier ratio in", + "type": "text" + }, + { + "bbox": [ + 240, + 467, + 250, + 477 + ], + "score": 0.85, + "content": "\\%", + "type": "inline_equation" + }, + { + "bbox": [ + 250, + 467, + 506, + 479 + ], + "score": 1.0, + "content": "among all pixels with valid ground truth, where an estimate is", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 478, + 437, + 490 + ], + "spans": [ + { + "bbox": [ + 105, + 478, + 329, + 490 + ], + "score": 1.0, + "content": "considered as an outlier if its error exceeds 3 pixels and", + "type": "text" + }, + { + "bbox": [ + 330, + 478, + 345, + 488 + ], + "score": 0.87, + "content": "5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 478, + 437, + 490 + ], + "score": 1.0, + "content": "w.r.t. the ground truth.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 107, + 494, + 505, + 549 + ], + "lines": [ + { + "bbox": [ + 105, + 493, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 506, + 507 + ], + "score": 1.0, + "content": "Depth. We separately pre-train indoor and outdoor models on the respective pre-training datasets", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 504, + 504, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 504, + 504, + 516 + ], + "score": 1.0, + "content": "described in Section 3.1. The indoor depth model is then finetuned and evaluated on the NYU depth", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 516, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 505, + 528 + ], + "score": 1.0, + "content": "v2 dataset [61] and the outdoor model on the KITTI depth dataset [17]. We follow the standard", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 526, + 505, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 505, + 540 + ], + "score": 1.0, + "content": "evaluation protocol used in prior work [35]. For both NYU depth v2 and KITTI, we report the", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 537, + 492, + 551 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 443, + 551 + ], + "score": 1.0, + "content": "absolute relative error (REL), root mean squared error (RMS) and accuracy metrics", + "type": "text" + }, + { + "bbox": [ + 443, + 538, + 487, + 549 + ], + "score": 0.87, + "content": "( \\delta _ { 1 } < 1 . 2 5 )", + "type": "inline_equation" + }, + { + "bbox": [ + 488, + 537, + 492, + 551 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34 + }, + { + "type": "title", + "bbox": [ + 108, + 565, + 278, + 577 + ], + "lines": [ + { + "bbox": [ + 106, + 565, + 279, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 279, + 578 + ], + "score": 1.0, + "content": "4.1 Evaluation on benchmark datasets", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 106, + 585, + 505, + 651 + ], + "lines": [ + { + "bbox": [ + 105, + 585, + 506, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 506, + 597 + ], + "score": 1.0, + "content": "Depth. Table 3 reports the results on NYU depth v2 and KITTI (see Section D for more detailed", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 596, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 596, + 506, + 609 + ], + "score": 1.0, + "content": "results and Section B for qualitative comparison with DPT on NYU). We achieve a state-of-the-art", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 606, + 505, + 620 + ], + "spans": [ + { + "bbox": [ + 105, + 606, + 505, + 620 + ], + "score": 1.0, + "content": "absolute relative error of 0.074 on NYU depth v2. On KITTI, our method performs competitively", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "score": 1.0, + "content": "with prior work. We report results with averaging depth maps from one or more samples. Note that", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 629, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 106, + 629, + 506, + 642 + ], + "score": 1.0, + "content": "most prior works use post processing that averages two samples, one from the input image, and the", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 640, + 312, + 651 + ], + "spans": [ + { + "bbox": [ + 106, + 640, + 312, + 651 + ], + "score": 1.0, + "content": "other based on its reflection about the vertical axis.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 40.5 + }, + { + "type": "text", + "bbox": [ + 107, + 656, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 656, + 505, + 668 + ], + "spans": [ + { + "bbox": [ + 106, + 656, + 505, + 668 + ], + "score": 1.0, + "content": "Flow. Table 1 reports the zero-shot results of our model on Sintel and KITTI Train datasets where", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 667, + 507, + 680 + ], + "spans": [ + { + "bbox": [ + 105, + 667, + 507, + 680 + ], + "score": 1.0, + "content": "ground truth are provided. The model is trained on our newly proposed pre-training mixtures (Aut-", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 677, + 506, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 506, + 691 + ], + "score": 1.0, + "content": "oFlow (AF), FlyingThings (FT), Kubric (KU), and TartanAir (TA)). We report results by averaging 8", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 688, + 507, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 507, + 701 + ], + "score": 1.0, + "content": "samples at a coarse resolution and then refining them to the full resolution as described in Section 3.3.", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 700, + 507, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 507, + 712 + ], + "score": 1.0, + "content": "For a fair comparison, we re-train RAFT on this pre-training mixture; this new RAFT model signif-", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 710, + 506, + 724 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 506, + 724 + ], + "score": 1.0, + "content": "icantly outperforms the original RAFT model. And our diffusion model outperforms the stronger", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 46.5 + } + ], + "page_idx": 6, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 741, + 309, + 750 + ], + "lines": [ + { + "bbox": [ + 302, + 741, + 309, + 752 + ], + "spans": [ + { + "bbox": [ + 302, + 741, + 309, + 752 + ], + "score": 1.0, + "content": "7", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 67, + 505, + 240 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 67, + 505, + 240 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 67, + 505, + 240 + ], + "spans": [ + { + "bbox": [ + 107, + 67, + 505, + 240 + ], + "score": 0.96, + "type": "image", + "image_path": "e4d6c5005bb47f266c2f4f71dc972c3f7e2c3803ce41a7e60f9eb15e36eb3af7.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 67, + 505, + 124.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 124.66666666666666, + 505, + 182.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 182.33333333333331, + 505, + 239.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 249, + 503, + 270 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 248, + 505, + 261 + ], + "spans": [ + { + "bbox": [ + 105, + 248, + 505, + 261 + ], + "score": 1.0, + "content": "Figure 5: Visual results comparing RAFT with our method after finetuning. Ours does much better on fine", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 259, + 220, + 271 + ], + "spans": [ + { + "bbox": [ + 105, + 259, + 220, + 271 + ], + "score": 1.0, + "content": "details and ambiguous regions.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "table", + "bbox": [ + 106, + 351, + 331, + 433 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 286, + 327, + 346 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 287, + 326, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 287, + 326, + 298 + ], + "score": 1.0, + "content": "Table 1: Zero-shot optical flow estimation results on Sintel", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 296, + 327, + 308 + ], + "spans": [ + { + "bbox": [ + 105, + 296, + 327, + 308 + ], + "score": 1.0, + "content": "and KITTI. We provide a new RAFT baseline using our", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 307, + 327, + 318 + ], + "spans": [ + { + "bbox": [ + 106, + 307, + 327, + 318 + ], + "score": 1.0, + "content": "proposed pre-training mixture and substantially improve the", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 317, + 327, + 328 + ], + "spans": [ + { + "bbox": [ + 105, + 317, + 327, + 328 + ], + "score": 1.0, + "content": "accuracy over the original. Our diffusion model outperforms", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 327, + 327, + 337 + ], + "spans": [ + { + "bbox": [ + 106, + 327, + 327, + 337 + ], + "score": 1.0, + "content": "even this much stronger baseline and achieves state-of-the-art", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 337, + 263, + 346 + ], + "spans": [ + { + "bbox": [ + 106, + 337, + 263, + 346 + ], + "score": 1.0, + "content": "zero-shot results on Sintel.final and KITTI.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 7.5 + }, + { + "type": "table_body", + "bbox": [ + 106, + 351, + 331, + 433 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 351, + 331, + 433 + ], + "spans": [ + { + "bbox": [ + 106, + 351, + 331, + 433 + ], + "score": 0.977, + "html": "
ModelDatasetSintel.clean Sintel.finalKITTI
AEPEAEPEFl-all
FlowFormer RAFTChairs-→Things1.012.404.0914.72%
Chairs-→Things1.682.805.921
Perceiver IO RAFTAutoFlow1.812.424.98-
AutoFlow1.742.414.1813.41%
RAFT(ours)AF→AF+FT+KU+TA1.272.282.719.16%
DDVM(ours) AF-→AF+FT+KU+TA1.242.002.197.58%
", + "type": "table", + "image_path": "cded639441183179955a6aba42079d7c1357b21fa461a5f4fa42756628a0f7fe.jpg" + } + ] + } + ], + "index": 22.0, + "virtual_lines": [ + { + "bbox": [ + 106, + 351, + 331, + 364.6666666666667 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 106, + 364.6666666666667, + 331, + 378.33333333333337 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 106, + 378.33333333333337, + 331, + 392.00000000000006 + ], + "spans": [], + "index": 21 + }, + { + "bbox": [ + 106, + 392.00000000000006, + 331, + 405.66666666666674 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 106, + 405.66666666666674, + 331, + 419.3333333333334 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 106, + 419.3333333333334, + 331, + 433.0000000000001 + ], + "spans": [], + "index": 27 + } + ] + } + ], + "index": 14.75 + }, + { + "type": "table", + "bbox": [ + 344, + 363, + 504, + 433 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 344, + 286, + 505, + 357 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 343, + 286, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 343, + 286, + 505, + 298 + ], + "score": 1.0, + "content": "Table 2: Optical flow finetuning evaluation", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 343, + 297, + 505, + 308 + ], + "spans": [ + { + "bbox": [ + 343, + 297, + 459, + 308 + ], + "score": 1.0, + "content": "on public benchmark datasets", + "type": "text" + }, + { + "bbox": [ + 459, + 297, + 491, + 307 + ], + "score": 0.65, + "content": "( \\mathrm { A E P E } \\downarrow", + "type": "inline_equation" + }, + { + "bbox": [ + 491, + 297, + 505, + 308 + ], + "score": 1.0, + "content": "for", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 343, + 307, + 505, + 317 + ], + "spans": [ + { + "bbox": [ + 343, + 307, + 505, + 317 + ], + "score": 1.0, + "content": "Sintel and Fl-all↓ for KITTI). Bold indicates", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 343, + 315, + 507, + 327 + ], + "spans": [ + { + "bbox": [ + 343, + 315, + 436, + 327 + ], + "score": 1.0, + "content": "the best and underline the", + "type": "text" + }, + { + "bbox": [ + 437, + 316, + 449, + 326 + ], + "score": 0.87, + "content": "2 ^ { \\mathrm { n d } }", + "type": "inline_equation" + }, + { + "bbox": [ + 449, + 315, + 507, + 327 + ], + "score": 1.0, + "content": "-best. § uses ex-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 343, + 326, + 506, + 338 + ], + "spans": [ + { + "bbox": [ + 343, + 326, + 506, + 338 + ], + "score": 1.0, + "content": "tra datasets (AutoFlow and VIPER) on top of", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 343, + 335, + 505, + 347 + ], + "spans": [ + { + "bbox": [ + 343, + 335, + 505, + 347 + ], + "score": 1.0, + "content": "defaults (FlyingThings, HD1K, KITTI, and", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 344, + 347, + 472, + 357 + ], + "spans": [ + { + "bbox": [ + 344, + 347, + 472, + 357 + ], + "score": 1.0, + "content": "Sintel). ∗uses warm start on Sintel.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 14 + }, + { + "type": "table_body", + "bbox": [ + 344, + 363, + 504, + 433 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 344, + 363, + 504, + 433 + ], + "spans": [ + { + "bbox": [ + 344, + 363, + 504, + 433 + ], + "score": 0.975, + "html": "
MethodSintel.cleanSintel.finalKITTI
SKFlow [72]*1.302.264.84%
CRAFT [66]*1.442.424.79%
FlowFormer [23]1.142.184.68%
RAFT-OCTC [29]*1.512.574.33%
RAFT-it [70]$1.552.904.31%
DDVM (ours)$1.752.483.26%
", + "type": "table", + "image_path": "1a0c1cc34d71e3177749b62627c99338c09494c68c5d32ead929543042c0f6aa.jpg" + } + ] + } + ], + "index": 24, + "virtual_lines": [ + { + "bbox": [ + 344, + 363, + 504, + 377.0 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 344, + 377.0, + 504, + 391.0 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 344, + 391.0, + 504, + 405.0 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 344, + 405.0, + 504, + 419.0 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 344, + 419.0, + 504, + 433.0 + ], + "spans": [], + "index": 28 + } + ] + } + ], + "index": 19.0 + }, + { + "type": "text", + "bbox": [ + 107, + 456, + 505, + 489 + ], + "lines": [], + "index": 30, + "bbox_fs": [ + 105, + 455, + 506, + 490 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 494, + 505, + 549 + ], + "lines": [ + { + "bbox": [ + 105, + 493, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 506, + 507 + ], + "score": 1.0, + "content": "Depth. We separately pre-train indoor and outdoor models on the respective pre-training datasets", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 504, + 504, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 504, + 504, + 516 + ], + "score": 1.0, + "content": "described in Section 3.1. The indoor depth model is then finetuned and evaluated on the NYU depth", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 516, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 505, + 528 + ], + "score": 1.0, + "content": "v2 dataset [61] and the outdoor model on the KITTI depth dataset [17]. We follow the standard", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 526, + 505, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 505, + 540 + ], + "score": 1.0, + "content": "evaluation protocol used in prior work [35]. For both NYU depth v2 and KITTI, we report the", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 537, + 492, + 551 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 443, + 551 + ], + "score": 1.0, + "content": "absolute relative error (REL), root mean squared error (RMS) and accuracy metrics", + "type": "text" + }, + { + "bbox": [ + 443, + 538, + 487, + 549 + ], + "score": 0.87, + "content": "( \\delta _ { 1 } < 1 . 2 5 )", + "type": "inline_equation" + }, + { + "bbox": [ + 488, + 537, + 492, + 551 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34, + "bbox_fs": [ + 105, + 493, + 506, + 551 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 565, + 278, + 577 + ], + "lines": [ + { + "bbox": [ + 106, + 565, + 279, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 279, + 578 + ], + "score": 1.0, + "content": "4.1 Evaluation on benchmark datasets", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 106, + 585, + 505, + 651 + ], + "lines": [ + { + "bbox": [ + 105, + 585, + 506, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 506, + 597 + ], + "score": 1.0, + "content": "Depth. Table 3 reports the results on NYU depth v2 and KITTI (see Section D for more detailed", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 596, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 596, + 506, + 609 + ], + "score": 1.0, + "content": "results and Section B for qualitative comparison with DPT on NYU). We achieve a state-of-the-art", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 606, + 505, + 620 + ], + "spans": [ + { + "bbox": [ + 105, + 606, + 505, + 620 + ], + "score": 1.0, + "content": "absolute relative error of 0.074 on NYU depth v2. On KITTI, our method performs competitively", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 506, + 631 + ], + "score": 1.0, + "content": "with prior work. We report results with averaging depth maps from one or more samples. Note that", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 629, + 506, + 642 + ], + "spans": [ + { + "bbox": [ + 106, + 629, + 506, + 642 + ], + "score": 1.0, + "content": "most prior works use post processing that averages two samples, one from the input image, and the", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 640, + 312, + 651 + ], + "spans": [ + { + "bbox": [ + 106, + 640, + 312, + 651 + ], + "score": 1.0, + "content": "other based on its reflection about the vertical axis.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 40.5, + "bbox_fs": [ + 105, + 585, + 506, + 651 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 656, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 656, + 505, + 668 + ], + "spans": [ + { + "bbox": [ + 106, + 656, + 505, + 668 + ], + "score": 1.0, + "content": "Flow. Table 1 reports the zero-shot results of our model on Sintel and KITTI Train datasets where", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 667, + 507, + 680 + ], + "spans": [ + { + "bbox": [ + 105, + 667, + 507, + 680 + ], + "score": 1.0, + "content": "ground truth are provided. The model is trained on our newly proposed pre-training mixtures (Aut-", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 677, + 506, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 506, + 691 + ], + "score": 1.0, + "content": "oFlow (AF), FlyingThings (FT), Kubric (KU), and TartanAir (TA)). We report results by averaging 8", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 688, + 507, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 507, + 701 + ], + "score": 1.0, + "content": "samples at a coarse resolution and then refining them to the full resolution as described in Section 3.3.", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 700, + 507, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 507, + 712 + ], + "score": 1.0, + "content": "For a fair comparison, we re-train RAFT on this pre-training mixture; this new RAFT model signif-", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 710, + 506, + 724 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 506, + 724 + ], + "score": 1.0, + "content": "icantly outperforms the original RAFT model. And our diffusion model outperforms the stronger", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 46.5, + "bbox_fs": [ + 105, + 656, + 507, + 724 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 151, + 112, + 459, + 249 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 107, + 77, + 505, + 108 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 77, + 505, + 89 + ], + "spans": [ + { + "bbox": [ + 105, + 77, + 316, + 89 + ], + "score": 1.0, + "content": "Table 3: Performance comparison on the NYU-Depth-", + "type": "text" + }, + { + "bbox": [ + 316, + 78, + 326, + 87 + ], + "score": 0.3, + "content": "\\mathbf { \\nabla } \\cdot \\mathbf { v } _ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 327, + 77, + 411, + 89 + ], + "score": 1.0, + "content": "and KITTI datasets.", + "type": "text" + }, + { + "bbox": [ + 411, + 78, + 420, + 87 + ], + "score": 0.46, + "content": "\\top", + "type": "inline_equation" + }, + { + "bbox": [ + 420, + 77, + 505, + 89 + ], + "score": 1.0, + "content": "indicates method uses", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 87, + 505, + 99 + ], + "spans": [ + { + "bbox": [ + 105, + 87, + 338, + 99 + ], + "score": 1.0, + "content": "unsupervised pretraining, †indicates supervised pretraining and", + "type": "text" + }, + { + "bbox": [ + 339, + 88, + 345, + 98 + ], + "score": 0.84, + "content": "^ \\ddag", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 87, + 505, + 99 + ], + "score": 1.0, + "content": "indicates use of auxilliary supervised depth", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 97, + 483, + 109 + ], + "spans": [ + { + "bbox": [ + 105, + 97, + 351, + 109 + ], + "score": 1.0, + "content": "data. Best / second best results are bolded / underlined respectively.", + "type": "text" + }, + { + "bbox": [ + 352, + 98, + 360, + 108 + ], + "score": 0.41, + "content": "\\downarrow \\colon", + "type": "inline_equation" + }, + { + "bbox": [ + 361, + 97, + 415, + 109 + ], + "score": 1.0, + "content": "lower is better", + "type": "text" + }, + { + "bbox": [ + 415, + 98, + 421, + 108 + ], + "score": 0.55, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 422, + 97, + 483, + 109 + ], + "score": 1.0, + "content": ": higher is better.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "table_body", + "bbox": [ + 151, + 112, + 459, + 249 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 151, + 112, + 459, + 249 + ], + "spans": [ + { + "bbox": [ + 151, + 112, + 459, + 249 + ], + "score": 0.978, + "html": "
MethodArchitectureNYU-Depth-v2KITTI
81↑REL↓RMS↓81↑REL↓RMS↓
TransDepth [87]Res-50+ViT-B†0.900 0.1060.3650.9560.0642.755
DPT[50]Res-50+ViT-Bt‡0.904 0.1100.3570.959 0.0622.573
BTS [34]DenseNet-161t0.885 0.1100.3920.9560.0592.756
AdaBins [3]E-B5+Mini-ViT†0.903 0.1030.3640.9640.0582.360
BinsFormer[35]Swin-Large†0.925 0.0940.3300.9740.0522.098
PixelFormer [1]Swin-Larget0.929 0.0900.3220.9760.0512.081
MIM [80]SwinV2-LT0.949 0.0830.2870.9770.0501.966
AiT-P [46]SwinV2-LT0.953 0.0760.279111
DDVM samples=2samples=1 Efficient U-NetT#0.9440.0750.3240.9640.0562.700
Efficient U-NetT#0.9440.0740.3190.9650.0552.660
samples=4 Efficient U-NetT#0.9460.0740.3150.9650.0552.613
", + "type": "table", + "image_path": "ec54fd732f5ffac595043875ae4da5583b107492aac54978d86f52ded6b143d2.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 151, + 112, + 459, + 157.66666666666666 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 151, + 157.66666666666666, + 459, + 203.33333333333331 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 151, + 203.33333333333331, + 459, + 248.99999999999997 + ], + "spans": [], + "index": 5 + } + ] + } + ], + "index": 2.5 + }, + { + "type": "table", + "bbox": [ + 106, + 288, + 333, + 353 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 256, + 330, + 286 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 255, + 331, + 268 + ], + "spans": [ + { + "bbox": [ + 105, + 255, + 331, + 268 + ], + "score": 1.0, + "content": "Table 4: Ablation on infilling and step-unrolling. Without", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 267, + 331, + 277 + ], + "spans": [ + { + "bbox": [ + 106, + 267, + 331, + 277 + ], + "score": 1.0, + "content": "either one, performance deteriorates. Without both, optical", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 276, + 234, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 276, + 234, + 287 + ], + "score": 1.0, + "content": "flow models fail to train on KITTI.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "table_body", + "bbox": [ + 106, + 288, + 333, + 353 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 288, + 333, + 353 + ], + "spans": [ + { + "bbox": [ + 106, + 288, + 333, + 353 + ], + "score": 0.979, + "html": "
NYU valKITTI valKITTI val
RELRMSRELRMSAEPEFl-all
Baseline0.0790.3310.2223.770--
Step-unroll0.0760.3240.0852.8441.846.16%
Infill0.0770.3380.0572.7441.535.24%
Step-unroll & infill0.0750.3240.0562.7001.474.74%
", + "type": "table", + "image_path": "43b822f41f7dcd11784589130aa60439a317a6efca5f986fcf60cb64904eced2.jpg" + } + ] + } + ], + "index": 11, + "virtual_lines": [ + { + "bbox": [ + 106, + 288, + 333, + 301.0 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 106, + 301.0, + 333, + 314.0 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 106, + 314.0, + 333, + 327.0 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 106, + 327.0, + 333, + 340.0 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 106, + 340.0, + 333, + 353.0 + ], + "spans": [], + "index": 13 + } + ] + } + ], + "index": 9.0 + }, + { + "type": "table", + "bbox": [ + 343, + 303, + 502, + 353 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 342, + 256, + 503, + 297 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 341, + 255, + 505, + 267 + ], + "spans": [ + { + "bbox": [ + 341, + 255, + 505, + 267 + ], + "score": 1.0, + "content": "Table 5: Coarse-to-fine refinement im-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 340, + 266, + 504, + 277 + ], + "spans": [ + { + "bbox": [ + 340, + 266, + 504, + 277 + ], + "score": 1.0, + "content": "proves zero-shot optical flow estimation re-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 341, + 276, + 504, + 287 + ], + "spans": [ + { + "bbox": [ + 341, + 276, + 504, + 287 + ], + "score": 1.0, + "content": "sults on Sintel and KITTI, along with the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 341, + 286, + 503, + 297 + ], + "spans": [ + { + "bbox": [ + 341, + 286, + 503, + 297 + ], + "score": 1.0, + "content": "qualitative improvements shown in Figure 6.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 15.5 + }, + { + "type": "table_body", + "bbox": [ + 343, + 303, + 502, + 353 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 343, + 303, + 502, + 353 + ], + "spans": [ + { + "bbox": [ + 343, + 303, + 502, + 353 + ], + "score": 0.972, + "html": "
Coarse-to-fineSintel.clean Sintel.finalKITTI
refinementAEPEAEPEAEPE Fl-all
Without1.422.122.35 8.65%
With1.242.002.19 7.58%
", + "type": "table", + "image_path": "139baafdaf8362ea1d3401e261f20ffe9f92d24a7990a33c32f0163a3d4f78ac.jpg" + } + ] + } + ], + "index": 19, + "virtual_lines": [ + { + "bbox": [ + 343, + 303, + 502, + 319.6666666666667 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 343, + 319.6666666666667, + 502, + 336.33333333333337 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 343, + 336.33333333333337, + 502, + 353.00000000000006 + ], + "spans": [], + "index": 20 + } + ] + } + ], + "index": 17.25 + }, + { + "type": "text", + "bbox": [ + 106, + 371, + 504, + 393 + ], + "lines": [ + { + "bbox": [ + 106, + 370, + 505, + 383 + ], + "spans": [ + { + "bbox": [ + 106, + 370, + 505, + 383 + ], + "score": 1.0, + "content": "RAFT baseline. It achieves the state-of-the-art zero-shot results on both the challenging Sintel Final", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 381, + 188, + 393 + ], + "spans": [ + { + "bbox": [ + 106, + 381, + 188, + 393 + ], + "score": 1.0, + "content": "and KITTI datasets.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5 + }, + { + "type": "text", + "bbox": [ + 108, + 398, + 503, + 431 + ], + "lines": [ + { + "bbox": [ + 105, + 398, + 505, + 411 + ], + "spans": [ + { + "bbox": [ + 105, + 398, + 505, + 411 + ], + "score": 1.0, + "content": "Figure 4 provides a qualitative comparison of pre-trained models. Our method demonstrates finer", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 409, + 505, + 421 + ], + "spans": [ + { + "bbox": [ + 106, + 409, + 505, + 421 + ], + "score": 1.0, + "content": "details on both object and motion boundaries. Especially on KITTI, our model recovers fine details", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 420, + 439, + 432 + ], + "spans": [ + { + "bbox": [ + 106, + 420, + 439, + 432 + ], + "score": 1.0, + "content": "remarkably well, e.g. on trees and its layered motion between tree and background.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 107, + 436, + 505, + 567 + ], + "lines": [ + { + "bbox": [ + 105, + 435, + 507, + 450 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 507, + 450 + ], + "score": 1.0, + "content": "We further finetune our model on the mixture of the following datasets, AutoFlow, FlyingThings,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 447, + 505, + 460 + ], + "spans": [ + { + "bbox": [ + 105, + 447, + 505, + 460 + ], + "score": 1.0, + "content": "HD1K, KITTI, Sintel, and VIPER. Table 2 reports the comparison to state-of-the-art optical flow", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 458, + 505, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 458, + 505, + 470 + ], + "score": 1.0, + "content": "methods on public benchmark datasets, Sintel and KITTI. On KITTI, our method outperforms all", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 469, + 506, + 482 + ], + "spans": [ + { + "bbox": [ + 105, + 469, + 506, + 482 + ], + "score": 1.0, + "content": "existing optical flow methods by a substantial margin (even most scene flow methods that use stereo", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 480, + 505, + 492 + ], + "spans": [ + { + "bbox": [ + 105, + 480, + 505, + 492 + ], + "score": 1.0, + "content": "inputs), and sets the new state of the art. On the challenging Sintel final, our method is competitive", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 491, + 506, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 491, + 506, + 504 + ], + "score": 1.0, + "content": "with other state of the art models. Except for methods using warm-start strategies, our method is", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 501, + 506, + 515 + ], + "spans": [ + { + "bbox": [ + 106, + 501, + 506, + 515 + ], + "score": 1.0, + "content": "only behind FlowFormer which adopts strong domain knowledge on optical flow (e.g. cost volume,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 513, + 506, + 526 + ], + "spans": [ + { + "bbox": [ + 105, + 513, + 506, + 526 + ], + "score": 1.0, + "content": "iterative refinement, or attention layers for larger context) unlike our generic model. Interestingly, we", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 523, + 506, + 536 + ], + "spans": [ + { + "bbox": [ + 105, + 523, + 506, + 536 + ], + "score": 1.0, + "content": "find that our model outperforms FlowFormer on 11/12 Sintel test sequences and our overall worse", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 535, + 505, + 547 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 505, + 547 + ], + "score": 1.0, + "content": "performance can be attributed to a much higher AEPE on a single (possibly out-of-distribution) test", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 545, + 504, + 557 + ], + "spans": [ + { + "bbox": [ + 105, + 545, + 504, + 557 + ], + "score": 1.0, + "content": "sequence. We discuss this in more detail in Section 5. On KITTI, our diffusion model outperforms", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 556, + 272, + 569 + ], + "spans": [ + { + "bbox": [ + 105, + 556, + 231, + 569 + ], + "score": 1.0, + "content": "FlowFormer by a large margin", + "type": "text" + }, + { + "bbox": [ + 231, + 556, + 268, + 568 + ], + "score": 0.85, + "content": "( 3 0 . 3 4 \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 268, + 556, + 272, + 569 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 31.5 + }, + { + "type": "title", + "bbox": [ + 107, + 582, + 193, + 594 + ], + "lines": [ + { + "bbox": [ + 105, + 579, + 194, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 194, + 597 + ], + "score": 1.0, + "content": "4.2 Ablation study", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 106, + 601, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 602, + 505, + 614 + ], + "spans": [ + { + "bbox": [ + 106, + 602, + 505, + 614 + ], + "score": 1.0, + "content": "Infilling and step-unrolling. We study the effect of infilling and step-unrolling in Table 4. For depth,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 612, + 506, + 626 + ], + "spans": [ + { + "bbox": [ + 105, + 612, + 506, + 626 + ], + "score": 1.0, + "content": "we report results for fine-tuning our pre-trained model on the NYU and KITTI datasets with the same", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 623, + 505, + 636 + ], + "spans": [ + { + "bbox": [ + 105, + 623, + 505, + 636 + ], + "score": 1.0, + "content": "resolution and augmentations as our best results. For flow, we fine-tune on the KITTI train set alone", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 635, + 506, + 647 + ], + "spans": [ + { + "bbox": [ + 106, + 635, + 506, + 647 + ], + "score": 1.0, + "content": "(with nearest neighbor resizing to the target resolution being the only augmentation) at a resolution of", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 644, + 505, + 658 + ], + "spans": [ + { + "bbox": [ + 106, + 645, + 145, + 656 + ], + "score": 0.88, + "content": "3 2 0 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 644, + 505, + 658 + ], + "score": 1.0, + "content": "and report metrics on the KITTI val set [39]. We report results with a single sample and no", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 104, + 655, + 505, + 670 + ], + "spans": [ + { + "bbox": [ + 104, + 655, + 505, + 670 + ], + "score": 1.0, + "content": "coarse-to-fine refinement. We find that training on raw sparse data without infilling and step unrolling", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 667, + 505, + 680 + ], + "spans": [ + { + "bbox": [ + 105, + 667, + 505, + 680 + ], + "score": 1.0, + "content": "leads to poor results, especially on KITTI where the ground truth is quite sparse. Step-unrolling", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 678, + 506, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 678, + 506, + 691 + ], + "score": 1.0, + "content": "helps to stabilize training without requiring any extra data pre-processing. However, we find that", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 689, + 505, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 689, + 505, + 702 + ], + "score": 1.0, + "content": "most gains come from interpolating missing values in the sparse labels. Infilling and step-unrolling", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 700, + 505, + 714 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 505, + 714 + ], + "score": 1.0, + "content": "compose well as our best results use both; infilling (being an approximation) does not completely", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 106, + 711, + 368, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 368, + 723 + ], + "score": 1.0, + "content": "bridge the training-inference distribution shift of the noisy latent.", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 44 + } + ], + "page_idx": 7, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 742, + 308, + 750 + ], + "lines": [ + { + "bbox": [ + 301, + 740, + 310, + 752 + ], + "spans": [ + { + "bbox": [ + 301, + 740, + 310, + 752 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 12, + "width": 9 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 151, + 112, + 459, + 249 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 107, + 77, + 505, + 108 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 77, + 505, + 89 + ], + "spans": [ + { + "bbox": [ + 105, + 77, + 316, + 89 + ], + "score": 1.0, + "content": "Table 3: Performance comparison on the NYU-Depth-", + "type": "text" + }, + { + "bbox": [ + 316, + 78, + 326, + 87 + ], + "score": 0.3, + "content": "\\mathbf { \\nabla } \\cdot \\mathbf { v } _ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 327, + 77, + 411, + 89 + ], + "score": 1.0, + "content": "and KITTI datasets.", + "type": "text" + }, + { + "bbox": [ + 411, + 78, + 420, + 87 + ], + "score": 0.46, + "content": "\\top", + "type": "inline_equation" + }, + { + "bbox": [ + 420, + 77, + 505, + 89 + ], + "score": 1.0, + "content": "indicates method uses", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 87, + 505, + 99 + ], + "spans": [ + { + "bbox": [ + 105, + 87, + 338, + 99 + ], + "score": 1.0, + "content": "unsupervised pretraining, †indicates supervised pretraining and", + "type": "text" + }, + { + "bbox": [ + 339, + 88, + 345, + 98 + ], + "score": 0.84, + "content": "^ \\ddag", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 87, + 505, + 99 + ], + "score": 1.0, + "content": "indicates use of auxilliary supervised depth", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 97, + 483, + 109 + ], + "spans": [ + { + "bbox": [ + 105, + 97, + 351, + 109 + ], + "score": 1.0, + "content": "data. Best / second best results are bolded / underlined respectively.", + "type": "text" + }, + { + "bbox": [ + 352, + 98, + 360, + 108 + ], + "score": 0.41, + "content": "\\downarrow \\colon", + "type": "inline_equation" + }, + { + "bbox": [ + 361, + 97, + 415, + 109 + ], + "score": 1.0, + "content": "lower is better", + "type": "text" + }, + { + "bbox": [ + 415, + 98, + 421, + 108 + ], + "score": 0.55, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 422, + 97, + 483, + 109 + ], + "score": 1.0, + "content": ": higher is better.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "table_body", + "bbox": [ + 151, + 112, + 459, + 249 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 151, + 112, + 459, + 249 + ], + "spans": [ + { + "bbox": [ + 151, + 112, + 459, + 249 + ], + "score": 0.978, + "html": "
MethodArchitectureNYU-Depth-v2KITTI
81↑REL↓RMS↓81↑REL↓RMS↓
TransDepth [87]Res-50+ViT-B†0.900 0.1060.3650.9560.0642.755
DPT[50]Res-50+ViT-Bt‡0.904 0.1100.3570.959 0.0622.573
BTS [34]DenseNet-161t0.885 0.1100.3920.9560.0592.756
AdaBins [3]E-B5+Mini-ViT†0.903 0.1030.3640.9640.0582.360
BinsFormer[35]Swin-Large†0.925 0.0940.3300.9740.0522.098
PixelFormer [1]Swin-Larget0.929 0.0900.3220.9760.0512.081
MIM [80]SwinV2-LT0.949 0.0830.2870.9770.0501.966
AiT-P [46]SwinV2-LT0.953 0.0760.279111
DDVM samples=2samples=1 Efficient U-NetT#0.9440.0750.3240.9640.0562.700
Efficient U-NetT#0.9440.0740.3190.9650.0552.660
samples=4 Efficient U-NetT#0.9460.0740.3150.9650.0552.613
", + "type": "table", + "image_path": "ec54fd732f5ffac595043875ae4da5583b107492aac54978d86f52ded6b143d2.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 151, + 112, + 459, + 157.66666666666666 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 151, + 157.66666666666666, + 459, + 203.33333333333331 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 151, + 203.33333333333331, + 459, + 248.99999999999997 + ], + "spans": [], + "index": 5 + } + ] + } + ], + "index": 2.5 + }, + { + "type": "table", + "bbox": [ + 106, + 288, + 333, + 353 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 256, + 330, + 286 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 255, + 331, + 268 + ], + "spans": [ + { + "bbox": [ + 105, + 255, + 331, + 268 + ], + "score": 1.0, + "content": "Table 4: Ablation on infilling and step-unrolling. Without", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 267, + 331, + 277 + ], + "spans": [ + { + "bbox": [ + 106, + 267, + 331, + 277 + ], + "score": 1.0, + "content": "either one, performance deteriorates. Without both, optical", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 276, + 234, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 276, + 234, + 287 + ], + "score": 1.0, + "content": "flow models fail to train on KITTI.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "table_body", + "bbox": [ + 106, + 288, + 333, + 353 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 288, + 333, + 353 + ], + "spans": [ + { + "bbox": [ + 106, + 288, + 333, + 353 + ], + "score": 0.979, + "html": "
NYU valKITTI valKITTI val
RELRMSRELRMSAEPEFl-all
Baseline0.0790.3310.2223.770--
Step-unroll0.0760.3240.0852.8441.846.16%
Infill0.0770.3380.0572.7441.535.24%
Step-unroll & infill0.0750.3240.0562.7001.474.74%
", + "type": "table", + "image_path": "43b822f41f7dcd11784589130aa60439a317a6efca5f986fcf60cb64904eced2.jpg" + } + ] + } + ], + "index": 11, + "virtual_lines": [ + { + "bbox": [ + 106, + 288, + 333, + 301.0 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 106, + 301.0, + 333, + 314.0 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 106, + 314.0, + 333, + 327.0 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 106, + 327.0, + 333, + 340.0 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 106, + 340.0, + 333, + 353.0 + ], + "spans": [], + "index": 13 + } + ] + } + ], + "index": 9.0 + }, + { + "type": "table", + "bbox": [ + 343, + 303, + 502, + 353 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 342, + 256, + 503, + 297 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 341, + 255, + 505, + 267 + ], + "spans": [ + { + "bbox": [ + 341, + 255, + 505, + 267 + ], + "score": 1.0, + "content": "Table 5: Coarse-to-fine refinement im-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 340, + 266, + 504, + 277 + ], + "spans": [ + { + "bbox": [ + 340, + 266, + 504, + 277 + ], + "score": 1.0, + "content": "proves zero-shot optical flow estimation re-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 341, + 276, + 504, + 287 + ], + "spans": [ + { + "bbox": [ + 341, + 276, + 504, + 287 + ], + "score": 1.0, + "content": "sults on Sintel and KITTI, along with the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 341, + 286, + 503, + 297 + ], + "spans": [ + { + "bbox": [ + 341, + 286, + 503, + 297 + ], + "score": 1.0, + "content": "qualitative improvements shown in Figure 6.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 15.5 + }, + { + "type": "table_body", + "bbox": [ + 343, + 303, + 502, + 353 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 343, + 303, + 502, + 353 + ], + "spans": [ + { + "bbox": [ + 343, + 303, + 502, + 353 + ], + "score": 0.972, + "html": "
Coarse-to-fineSintel.clean Sintel.finalKITTI
refinementAEPEAEPEAEPE Fl-all
Without1.422.122.35 8.65%
With1.242.002.19 7.58%
", + "type": "table", + "image_path": "139baafdaf8362ea1d3401e261f20ffe9f92d24a7990a33c32f0163a3d4f78ac.jpg" + } + ] + } + ], + "index": 19, + "virtual_lines": [ + { + "bbox": [ + 343, + 303, + 502, + 319.6666666666667 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 343, + 319.6666666666667, + 502, + 336.33333333333337 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 343, + 336.33333333333337, + 502, + 353.00000000000006 + ], + "spans": [], + "index": 20 + } + ] + } + ], + "index": 17.25 + }, + { + "type": "text", + "bbox": [ + 106, + 371, + 504, + 393 + ], + "lines": [ + { + "bbox": [ + 106, + 370, + 505, + 383 + ], + "spans": [ + { + "bbox": [ + 106, + 370, + 505, + 383 + ], + "score": 1.0, + "content": "RAFT baseline. It achieves the state-of-the-art zero-shot results on both the challenging Sintel Final", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 381, + 188, + 393 + ], + "spans": [ + { + "bbox": [ + 106, + 381, + 188, + 393 + ], + "score": 1.0, + "content": "and KITTI datasets.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5, + "bbox_fs": [ + 106, + 370, + 505, + 393 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 398, + 503, + 431 + ], + "lines": [ + { + "bbox": [ + 105, + 398, + 505, + 411 + ], + "spans": [ + { + "bbox": [ + 105, + 398, + 505, + 411 + ], + "score": 1.0, + "content": "Figure 4 provides a qualitative comparison of pre-trained models. Our method demonstrates finer", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 409, + 505, + 421 + ], + "spans": [ + { + "bbox": [ + 106, + 409, + 505, + 421 + ], + "score": 1.0, + "content": "details on both object and motion boundaries. Especially on KITTI, our model recovers fine details", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 420, + 439, + 432 + ], + "spans": [ + { + "bbox": [ + 106, + 420, + 439, + 432 + ], + "score": 1.0, + "content": "remarkably well, e.g. on trees and its layered motion between tree and background.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24, + "bbox_fs": [ + 105, + 398, + 505, + 432 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 436, + 505, + 567 + ], + "lines": [ + { + "bbox": [ + 105, + 435, + 507, + 450 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 507, + 450 + ], + "score": 1.0, + "content": "We further finetune our model on the mixture of the following datasets, AutoFlow, FlyingThings,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 447, + 505, + 460 + ], + "spans": [ + { + "bbox": [ + 105, + 447, + 505, + 460 + ], + "score": 1.0, + "content": "HD1K, KITTI, Sintel, and VIPER. Table 2 reports the comparison to state-of-the-art optical flow", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 458, + 505, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 458, + 505, + 470 + ], + "score": 1.0, + "content": "methods on public benchmark datasets, Sintel and KITTI. On KITTI, our method outperforms all", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 469, + 506, + 482 + ], + "spans": [ + { + "bbox": [ + 105, + 469, + 506, + 482 + ], + "score": 1.0, + "content": "existing optical flow methods by a substantial margin (even most scene flow methods that use stereo", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 480, + 505, + 492 + ], + "spans": [ + { + "bbox": [ + 105, + 480, + 505, + 492 + ], + "score": 1.0, + "content": "inputs), and sets the new state of the art. On the challenging Sintel final, our method is competitive", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 491, + 506, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 491, + 506, + 504 + ], + "score": 1.0, + "content": "with other state of the art models. Except for methods using warm-start strategies, our method is", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 501, + 506, + 515 + ], + "spans": [ + { + "bbox": [ + 106, + 501, + 506, + 515 + ], + "score": 1.0, + "content": "only behind FlowFormer which adopts strong domain knowledge on optical flow (e.g. cost volume,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 513, + 506, + 526 + ], + "spans": [ + { + "bbox": [ + 105, + 513, + 506, + 526 + ], + "score": 1.0, + "content": "iterative refinement, or attention layers for larger context) unlike our generic model. Interestingly, we", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 523, + 506, + 536 + ], + "spans": [ + { + "bbox": [ + 105, + 523, + 506, + 536 + ], + "score": 1.0, + "content": "find that our model outperforms FlowFormer on 11/12 Sintel test sequences and our overall worse", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 535, + 505, + 547 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 505, + 547 + ], + "score": 1.0, + "content": "performance can be attributed to a much higher AEPE on a single (possibly out-of-distribution) test", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 545, + 504, + 557 + ], + "spans": [ + { + "bbox": [ + 105, + 545, + 504, + 557 + ], + "score": 1.0, + "content": "sequence. We discuss this in more detail in Section 5. On KITTI, our diffusion model outperforms", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 556, + 272, + 569 + ], + "spans": [ + { + "bbox": [ + 105, + 556, + 231, + 569 + ], + "score": 1.0, + "content": "FlowFormer by a large margin", + "type": "text" + }, + { + "bbox": [ + 231, + 556, + 268, + 568 + ], + "score": 0.85, + "content": "( 3 0 . 3 4 \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 268, + 556, + 272, + 569 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 31.5, + "bbox_fs": [ + 105, + 435, + 507, + 569 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 582, + 193, + 594 + ], + "lines": [ + { + "bbox": [ + 105, + 579, + 194, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 194, + 597 + ], + "score": 1.0, + "content": "4.2 Ablation study", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 106, + 601, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 602, + 505, + 614 + ], + "spans": [ + { + "bbox": [ + 106, + 602, + 505, + 614 + ], + "score": 1.0, + "content": "Infilling and step-unrolling. We study the effect of infilling and step-unrolling in Table 4. For depth,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 612, + 506, + 626 + ], + "spans": [ + { + "bbox": [ + 105, + 612, + 506, + 626 + ], + "score": 1.0, + "content": "we report results for fine-tuning our pre-trained model on the NYU and KITTI datasets with the same", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 623, + 505, + 636 + ], + "spans": [ + { + "bbox": [ + 105, + 623, + 505, + 636 + ], + "score": 1.0, + "content": "resolution and augmentations as our best results. For flow, we fine-tune on the KITTI train set alone", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 635, + 506, + 647 + ], + "spans": [ + { + "bbox": [ + 106, + 635, + 506, + 647 + ], + "score": 1.0, + "content": "(with nearest neighbor resizing to the target resolution being the only augmentation) at a resolution of", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 644, + 505, + 658 + ], + "spans": [ + { + "bbox": [ + 106, + 645, + 145, + 656 + ], + "score": 0.88, + "content": "3 2 0 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 644, + 505, + 658 + ], + "score": 1.0, + "content": "and report metrics on the KITTI val set [39]. We report results with a single sample and no", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 104, + 655, + 505, + 670 + ], + "spans": [ + { + "bbox": [ + 104, + 655, + 505, + 670 + ], + "score": 1.0, + "content": "coarse-to-fine refinement. We find that training on raw sparse data without infilling and step unrolling", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 667, + 505, + 680 + ], + "spans": [ + { + "bbox": [ + 105, + 667, + 505, + 680 + ], + "score": 1.0, + "content": "leads to poor results, especially on KITTI where the ground truth is quite sparse. Step-unrolling", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 678, + 506, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 678, + 506, + 691 + ], + "score": 1.0, + "content": "helps to stabilize training without requiring any extra data pre-processing. However, we find that", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 689, + 505, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 689, + 505, + 702 + ], + "score": 1.0, + "content": "most gains come from interpolating missing values in the sparse labels. Infilling and step-unrolling", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 700, + 505, + 714 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 505, + 714 + ], + "score": 1.0, + "content": "compose well as our best results use both; infilling (being an approximation) does not completely", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 106, + 711, + 368, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 368, + 723 + ], + "score": 1.0, + "content": "bridge the training-inference distribution shift of the noisy latent.", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 44, + "bbox_fs": [ + 104, + 602, + 506, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 97, + 505, + 258 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 97, + 505, + 258 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 97, + 505, + 258 + ], + "spans": [ + { + "bbox": [ + 106, + 97, + 505, + 258 + ], + "score": 0.676, + "type": "image", + "image_path": "d17094c18838f6bebf79c88f82e740c22f32a82e794d2989adf607790e7cce37.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 97, + 505, + 150.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 150.66666666666666, + 505, + 204.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 204.33333333333331, + 505, + 258.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 266, + 504, + 286 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 265, + 505, + 278 + ], + "spans": [ + { + "bbox": [ + 105, + 265, + 505, + 278 + ], + "score": 1.0, + "content": "Figure 6: Visual results with and without coarse-to-fine refinement. For our pretrained model, refinement", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 276, + 313, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 276, + 313, + 287 + ], + "score": 1.0, + "content": "helps correct wrong flow and adds details to correct flow.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "text", + "bbox": [ + 107, + 310, + 504, + 343 + ], + "lines": [ + { + "bbox": [ + 106, + 310, + 505, + 322 + ], + "spans": [ + { + "bbox": [ + 106, + 310, + 505, + 322 + ], + "score": 1.0, + "content": "Coarse-to-fine refinement. Figure 6 shows that coarse-to-fine refinement (Section 3.3) substantially", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 321, + 505, + 332 + ], + "spans": [ + { + "bbox": [ + 105, + 321, + 505, + 332 + ], + "score": 1.0, + "content": "improves fine-grained details in estimated optical flow fields. It also improves the metrics for zero-shot", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 332, + 390, + 343 + ], + "spans": [ + { + "bbox": [ + 105, + 332, + 390, + 343 + ], + "score": 1.0, + "content": "optical flow estimation on both KITTI and Sintel, as shown in Table 5.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 107, + 348, + 505, + 436 + ], + "lines": [ + { + "bbox": [ + 105, + 348, + 505, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 348, + 505, + 361 + ], + "score": 1.0, + "content": "Datasets. When using different mixtures of datasets for pretraining, we find that diffusion models", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "score": 1.0, + "content": "sometimes capture region boundaries and shape at the expense of local textural variation (eg see", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 370, + 505, + 382 + ], + "spans": [ + { + "bbox": [ + 106, + 370, + 505, + 382 + ], + "score": 1.0, + "content": "Figure 3). The model trained solely on AutoFlow tends to provide very coarse flow, and mimics the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 381, + 505, + 393 + ], + "spans": [ + { + "bbox": [ + 106, + 381, + 505, + 393 + ], + "score": 1.0, + "content": "object shapes found in AutoFlow. The addition of FlyingThings, Kubric, and TartanAir removes this", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 392, + 506, + 405 + ], + "spans": [ + { + "bbox": [ + 105, + 392, + 506, + 405 + ], + "score": 1.0, + "content": "hallucination and significantly improves the fine details in the flow estimates (eg, shadows, trees,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 403, + 506, + 415 + ], + "spans": [ + { + "bbox": [ + 106, + 403, + 453, + 415 + ], + "score": 1.0, + "content": "thin structure, and motion boundaries) together with a substantial boost in accuracy", + "type": "text" + }, + { + "bbox": [ + 454, + 403, + 464, + 414 + ], + "score": 0.58, + "content": "_ { c f }", + "type": "inline_equation" + }, + { + "bbox": [ + 465, + 403, + 506, + 415 + ], + "score": 1.0, + "content": ". Table 6).", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 412, + 506, + 427 + ], + "spans": [ + { + "bbox": [ + 105, + 412, + 506, + 427 + ], + "score": 1.0, + "content": "Similarly, we find that mixing SceneNet RGB-D [41], a synthetic dataset, along with ScanNet [9]", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 425, + 465, + 436 + ], + "spans": [ + { + "bbox": [ + 105, + 425, + 465, + 436 + ], + "score": 1.0, + "content": "provides a performance boost for fine-tuning results on NYU depth v2, shown in Table 7.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 11.5 + }, + { + "type": "title", + "bbox": [ + 108, + 448, + 306, + 459 + ], + "lines": [ + { + "bbox": [ + 106, + 447, + 307, + 461 + ], + "spans": [ + { + "bbox": [ + 106, + 447, + 307, + 461 + ], + "score": 1.0, + "content": "4.3 Interesting properties of diffusion models", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 107, + 465, + 505, + 531 + ], + "lines": [ + { + "bbox": [ + 106, + 466, + 505, + 477 + ], + "spans": [ + { + "bbox": [ + 106, + 466, + 505, + 477 + ], + "score": 1.0, + "content": "Multimodality. One strength of diffusion models is their ability to capture complex multimodal", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 476, + 505, + 489 + ], + "spans": [ + { + "bbox": [ + 105, + 476, + 505, + 489 + ], + "score": 1.0, + "content": "distributions. This can be effective in representing uncertainty, especially where there may exist", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 487, + 505, + 500 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 505, + 500 + ], + "score": 1.0, + "content": "natural ambiguities and thus multiple predictions, e.g. in cases of transparent, translucent, or reflective", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 498, + 505, + 511 + ], + "spans": [ + { + "bbox": [ + 105, + 498, + 505, + 511 + ], + "score": 1.0, + "content": "cases. Figure 1 presents multiple samples on the NYU, KITTI, and Sintel datasets, showing that our", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 510, + 505, + 521 + ], + "spans": [ + { + "bbox": [ + 106, + 510, + 505, + 521 + ], + "score": 1.0, + "content": "model captures multimodality and provides plausible samples when ambiguities exist. More details", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 521, + 271, + 532 + ], + "spans": [ + { + "bbox": [ + 106, + 521, + 271, + 532 + ], + "score": 1.0, + "content": "and examples are available in Section A.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 19.5 + }, + { + "type": "text", + "bbox": [ + 107, + 536, + 505, + 646 + ], + "lines": [ + { + "bbox": [ + 105, + 535, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 477, + 550 + ], + "score": 1.0, + "content": "Imputation of missing labels. A diffusion model trained to model the conditional distribution", + "type": "text" + }, + { + "bbox": [ + 477, + 536, + 505, + 548 + ], + "score": 0.91, + "content": "p ( y | x )", + "type": "inline_equation" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 547, + 507, + 561 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 273, + 561 + ], + "score": 1.0, + "content": "can be zero-shot leveraged to sample from", + "type": "text" + }, + { + "bbox": [ + 274, + 547, + 336, + 560 + ], + "score": 0.92, + "content": "p ( y | x , y _ { p a r t i a l } )", + "type": "inline_equation" + }, + { + "bbox": [ + 336, + 547, + 363, + 561 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 363, + 549, + 395, + 560 + ], + "score": 0.83, + "content": "y _ { p a r t i a l }", + "type": "inline_equation" + }, + { + "bbox": [ + 396, + 547, + 507, + 561 + ], + "score": 1.0, + "content": "is the partially known label.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 557, + 506, + 572 + ], + "spans": [ + { + "bbox": [ + 105, + 557, + 506, + 572 + ], + "score": 1.0, + "content": "One approach for doing this, known as the replacement method for conditional inference [63], is to", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 568, + 506, + 583 + ], + "spans": [ + { + "bbox": [ + 105, + 568, + 252, + 583 + ], + "score": 1.0, + "content": "replace the known portion the latent", + "type": "text" + }, + { + "bbox": [ + 252, + 571, + 262, + 581 + ], + "score": 0.85, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 262, + 568, + 506, + 583 + ], + "score": 1.0, + "content": "at each inference step with the noisy latent built by applying", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 579, + 506, + 593 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 506, + 593 + ], + "score": 1.0, + "content": "the forward process to the known label. We qualitatively study the results of leveraging replacement", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 590, + 506, + 604 + ], + "spans": [ + { + "bbox": [ + 105, + 590, + 506, + 604 + ], + "score": 1.0, + "content": "guidance for depth completion and find it to be surprisingly effective. We illustrate this by building a", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 601, + 506, + 615 + ], + "spans": [ + { + "bbox": [ + 105, + 601, + 506, + 615 + ], + "score": 1.0, + "content": "pipeline for iteratively generating 3D scenes (conditioned on a text prompt) as shown in Figure 7 by", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 613, + 505, + 626 + ], + "spans": [ + { + "bbox": [ + 106, + 613, + 505, + 626 + ], + "score": 1.0, + "content": "leveraging existing models for text-to-image generation and text-conditional image inpainting. While", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "score": 1.0, + "content": "a more thorough evaluation of depth completion and novel view synthesis against existing methods is", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 635, + 502, + 648 + ], + "spans": [ + { + "bbox": [ + 105, + 635, + 502, + 648 + ], + "score": 1.0, + "content": "warranted, we leave that exploration to future work. (See Section C for more details and examples.)", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 27.5 + }, + { + "type": "title", + "bbox": [ + 107, + 663, + 185, + 676 + ], + "lines": [ + { + "bbox": [ + 104, + 661, + 187, + 678 + ], + "spans": [ + { + "bbox": [ + 104, + 661, + 187, + 678 + ], + "score": 1.0, + "content": "5 Limitations", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 33 + }, + { + "type": "text", + "bbox": [ + 108, + 689, + 504, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 688, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 506, + 702 + ], + "score": 1.0, + "content": "Latency. We adopt standard practices from image-generation models, leading to larger models and", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 700, + 506, + 713 + ], + "spans": [ + { + "bbox": [ + 106, + 700, + 506, + 713 + ], + "score": 1.0, + "content": "slower running times than RAFT. However, we are excited by the recent progress on progressive", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 711, + 494, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 494, + 723 + ], + "score": 1.0, + "content": "distillation [42, 56] and consistency models [64] to improve inference speed in diffusion models.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35 + } + ], + "page_idx": 8, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 742, + 308, + 750 + ], + "lines": [ + { + "bbox": [ + 302, + 741, + 309, + 752 + ], + "spans": [ + { + "bbox": [ + 302, + 741, + 309, + 752 + ], + "score": 1.0, + "content": "9", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 97, + 505, + 258 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 97, + 505, + 258 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 97, + 505, + 258 + ], + "spans": [ + { + "bbox": [ + 106, + 97, + 505, + 258 + ], + "score": 0.676, + "type": "image", + "image_path": "d17094c18838f6bebf79c88f82e740c22f32a82e794d2989adf607790e7cce37.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 97, + 505, + 150.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 150.66666666666666, + 505, + 204.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 204.33333333333331, + 505, + 258.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 266, + 504, + 286 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 265, + 505, + 278 + ], + "spans": [ + { + "bbox": [ + 105, + 265, + 505, + 278 + ], + "score": 1.0, + "content": "Figure 6: Visual results with and without coarse-to-fine refinement. For our pretrained model, refinement", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 276, + 313, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 276, + 313, + 287 + ], + "score": 1.0, + "content": "helps correct wrong flow and adds details to correct flow.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "text", + "bbox": [ + 107, + 310, + 504, + 343 + ], + "lines": [ + { + "bbox": [ + 106, + 310, + 505, + 322 + ], + "spans": [ + { + "bbox": [ + 106, + 310, + 505, + 322 + ], + "score": 1.0, + "content": "Coarse-to-fine refinement. Figure 6 shows that coarse-to-fine refinement (Section 3.3) substantially", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 321, + 505, + 332 + ], + "spans": [ + { + "bbox": [ + 105, + 321, + 505, + 332 + ], + "score": 1.0, + "content": "improves fine-grained details in estimated optical flow fields. It also improves the metrics for zero-shot", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 332, + 390, + 343 + ], + "spans": [ + { + "bbox": [ + 105, + 332, + 390, + 343 + ], + "score": 1.0, + "content": "optical flow estimation on both KITTI and Sintel, as shown in Table 5.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6, + "bbox_fs": [ + 105, + 310, + 505, + 343 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 348, + 505, + 436 + ], + "lines": [ + { + "bbox": [ + 105, + 348, + 505, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 348, + 505, + 361 + ], + "score": 1.0, + "content": "Datasets. When using different mixtures of datasets for pretraining, we find that diffusion models", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "score": 1.0, + "content": "sometimes capture region boundaries and shape at the expense of local textural variation (eg see", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 370, + 505, + 382 + ], + "spans": [ + { + "bbox": [ + 106, + 370, + 505, + 382 + ], + "score": 1.0, + "content": "Figure 3). The model trained solely on AutoFlow tends to provide very coarse flow, and mimics the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 381, + 505, + 393 + ], + "spans": [ + { + "bbox": [ + 106, + 381, + 505, + 393 + ], + "score": 1.0, + "content": "object shapes found in AutoFlow. The addition of FlyingThings, Kubric, and TartanAir removes this", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 392, + 506, + 405 + ], + "spans": [ + { + "bbox": [ + 105, + 392, + 506, + 405 + ], + "score": 1.0, + "content": "hallucination and significantly improves the fine details in the flow estimates (eg, shadows, trees,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 403, + 506, + 415 + ], + "spans": [ + { + "bbox": [ + 106, + 403, + 453, + 415 + ], + "score": 1.0, + "content": "thin structure, and motion boundaries) together with a substantial boost in accuracy", + "type": "text" + }, + { + "bbox": [ + 454, + 403, + 464, + 414 + ], + "score": 0.58, + "content": "_ { c f }", + "type": "inline_equation" + }, + { + "bbox": [ + 465, + 403, + 506, + 415 + ], + "score": 1.0, + "content": ". Table 6).", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 412, + 506, + 427 + ], + "spans": [ + { + "bbox": [ + 105, + 412, + 506, + 427 + ], + "score": 1.0, + "content": "Similarly, we find that mixing SceneNet RGB-D [41], a synthetic dataset, along with ScanNet [9]", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 425, + 465, + 436 + ], + "spans": [ + { + "bbox": [ + 105, + 425, + 465, + 436 + ], + "score": 1.0, + "content": "provides a performance boost for fine-tuning results on NYU depth v2, shown in Table 7.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 11.5, + "bbox_fs": [ + 105, + 348, + 506, + 436 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 448, + 306, + 459 + ], + "lines": [ + { + "bbox": [ + 106, + 447, + 307, + 461 + ], + "spans": [ + { + "bbox": [ + 106, + 447, + 307, + 461 + ], + "score": 1.0, + "content": "4.3 Interesting properties of diffusion models", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 107, + 465, + 505, + 531 + ], + "lines": [ + { + "bbox": [ + 106, + 466, + 505, + 477 + ], + "spans": [ + { + "bbox": [ + 106, + 466, + 505, + 477 + ], + "score": 1.0, + "content": "Multimodality. One strength of diffusion models is their ability to capture complex multimodal", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 476, + 505, + 489 + ], + "spans": [ + { + "bbox": [ + 105, + 476, + 505, + 489 + ], + "score": 1.0, + "content": "distributions. This can be effective in representing uncertainty, especially where there may exist", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 487, + 505, + 500 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 505, + 500 + ], + "score": 1.0, + "content": "natural ambiguities and thus multiple predictions, e.g. in cases of transparent, translucent, or reflective", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 498, + 505, + 511 + ], + "spans": [ + { + "bbox": [ + 105, + 498, + 505, + 511 + ], + "score": 1.0, + "content": "cases. Figure 1 presents multiple samples on the NYU, KITTI, and Sintel datasets, showing that our", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 510, + 505, + 521 + ], + "spans": [ + { + "bbox": [ + 106, + 510, + 505, + 521 + ], + "score": 1.0, + "content": "model captures multimodality and provides plausible samples when ambiguities exist. More details", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 521, + 271, + 532 + ], + "spans": [ + { + "bbox": [ + 106, + 521, + 271, + 532 + ], + "score": 1.0, + "content": "and examples are available in Section A.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 19.5, + "bbox_fs": [ + 105, + 466, + 505, + 532 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 536, + 505, + 646 + ], + "lines": [ + { + "bbox": [ + 105, + 535, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 477, + 550 + ], + "score": 1.0, + "content": "Imputation of missing labels. A diffusion model trained to model the conditional distribution", + "type": "text" + }, + { + "bbox": [ + 477, + 536, + 505, + 548 + ], + "score": 0.91, + "content": "p ( y | x )", + "type": "inline_equation" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 547, + 507, + 561 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 273, + 561 + ], + "score": 1.0, + "content": "can be zero-shot leveraged to sample from", + "type": "text" + }, + { + "bbox": [ + 274, + 547, + 336, + 560 + ], + "score": 0.92, + "content": "p ( y | x , y _ { p a r t i a l } )", + "type": "inline_equation" + }, + { + "bbox": [ + 336, + 547, + 363, + 561 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 363, + 549, + 395, + 560 + ], + "score": 0.83, + "content": "y _ { p a r t i a l }", + "type": "inline_equation" + }, + { + "bbox": [ + 396, + 547, + 507, + 561 + ], + "score": 1.0, + "content": "is the partially known label.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 557, + 506, + 572 + ], + "spans": [ + { + "bbox": [ + 105, + 557, + 506, + 572 + ], + "score": 1.0, + "content": "One approach for doing this, known as the replacement method for conditional inference [63], is to", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 568, + 506, + 583 + ], + "spans": [ + { + "bbox": [ + 105, + 568, + 252, + 583 + ], + "score": 1.0, + "content": "replace the known portion the latent", + "type": "text" + }, + { + "bbox": [ + 252, + 571, + 262, + 581 + ], + "score": 0.85, + "content": "y _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 262, + 568, + 506, + 583 + ], + "score": 1.0, + "content": "at each inference step with the noisy latent built by applying", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 579, + 506, + 593 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 506, + 593 + ], + "score": 1.0, + "content": "the forward process to the known label. We qualitatively study the results of leveraging replacement", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 590, + 506, + 604 + ], + "spans": [ + { + "bbox": [ + 105, + 590, + 506, + 604 + ], + "score": 1.0, + "content": "guidance for depth completion and find it to be surprisingly effective. We illustrate this by building a", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 601, + 506, + 615 + ], + "spans": [ + { + "bbox": [ + 105, + 601, + 506, + 615 + ], + "score": 1.0, + "content": "pipeline for iteratively generating 3D scenes (conditioned on a text prompt) as shown in Figure 7 by", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 613, + 505, + 626 + ], + "spans": [ + { + "bbox": [ + 106, + 613, + 505, + 626 + ], + "score": 1.0, + "content": "leveraging existing models for text-to-image generation and text-conditional image inpainting. While", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "spans": [ + { + "bbox": [ + 105, + 624, + 505, + 637 + ], + "score": 1.0, + "content": "a more thorough evaluation of depth completion and novel view synthesis against existing methods is", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 635, + 502, + 648 + ], + "spans": [ + { + "bbox": [ + 105, + 635, + 502, + 648 + ], + "score": 1.0, + "content": "warranted, we leave that exploration to future work. (See Section C for more details and examples.)", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 27.5, + "bbox_fs": [ + 105, + 535, + 507, + 648 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 663, + 185, + 676 + ], + "lines": [ + { + "bbox": [ + 104, + 661, + 187, + 678 + ], + "spans": [ + { + "bbox": [ + 104, + 661, + 187, + 678 + ], + "score": 1.0, + "content": "5 Limitations", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 33 + }, + { + "type": "text", + "bbox": [ + 108, + 689, + 504, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 688, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 506, + 702 + ], + "score": 1.0, + "content": "Latency. We adopt standard practices from image-generation models, leading to larger models and", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 700, + 506, + 713 + ], + "spans": [ + { + "bbox": [ + 106, + 700, + 506, + 713 + ], + "score": 1.0, + "content": "slower running times than RAFT. However, we are excited by the recent progress on progressive", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 711, + 494, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 494, + 723 + ], + "score": 1.0, + "content": "distillation [42, 56] and consistency models [64] to improve inference speed in diffusion models.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35, + "bbox_fs": [ + 106, + 688, + 506, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 109, + 72, + 501, + 182 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 109, + 72, + 501, + 182 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 109, + 72, + 501, + 182 + ], + "spans": [ + { + "bbox": [ + 109, + 72, + 501, + 182 + ], + "score": 0.967, + "type": "image", + "image_path": "01253d8e088f9b0c7e2f798a6acd974225d4c1b6913503fb0f8e02b7c815bbe1.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 109, + 72, + 501, + 108.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 109, + 108.66666666666666, + 501, + 145.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 109, + 145.33333333333331, + 501, + 181.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 191, + 505, + 252 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 191, + 505, + 203 + ], + "spans": [ + { + "bbox": [ + 106, + 191, + 505, + 203 + ], + "score": 1.0, + "content": "Figure 7: Application of zero-shot depth completion with our model by incorporating it into an iterative 3D", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 201, + 506, + 213 + ], + "spans": [ + { + "bbox": [ + 105, + 201, + 506, + 213 + ], + "score": 1.0, + "content": "scene generation pipeline. Starting with a initial image (optionally generated from a text-to-image model), we", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 211, + 506, + 223 + ], + "spans": [ + { + "bbox": [ + 105, + 211, + 506, + 223 + ], + "score": 1.0, + "content": "sample an image-only conditioned depth map using our model. The image-depth pair is added to a point cloud.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 220, + 506, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 220, + 506, + 233 + ], + "score": 1.0, + "content": "We then iteratively render images and depth maps (with holes) from this point cloud by moving the camera. We", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "spans": [ + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "score": 1.0, + "content": "then fill image holes using an existing image inpainter (optionally text conditioned), and then use our model", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 241, + 504, + 252 + ], + "spans": [ + { + "bbox": [ + 106, + 241, + 504, + 252 + ], + "score": 1.0, + "content": "with replacement guidance to impute missing depths (conditioned on the filled RGB image and known depth).", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 5.5 + } + ], + "index": 3.25 + }, + { + "type": "text", + "bbox": [ + 107, + 271, + 504, + 305 + ], + "lines": [ + { + "bbox": [ + 105, + 271, + 505, + 283 + ], + "spans": [ + { + "bbox": [ + 105, + 271, + 505, + 283 + ], + "score": 1.0, + "content": "Sintel fine-tuning. Under the zero-shot setting, our method achieves state-of-the-art results on both", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 281, + 505, + 295 + ], + "spans": [ + { + "bbox": [ + 105, + 281, + 505, + 295 + ], + "score": 1.0, + "content": "Sintel Final and KITTI. Under the fine-tuning setting, ours is state-of-the-art on KITTI but is behind", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 291, + 506, + 307 + ], + "spans": [ + { + "bbox": [ + 105, + 291, + 506, + 307 + ], + "score": 1.0, + "content": "FlowFormer [23] on Sintel Final. We discuss several possible reasons for why this may be the case.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + }, + { + "type": "text", + "bbox": [ + 106, + 315, + 505, + 453 + ], + "lines": [ + { + "bbox": [ + 106, + 315, + 505, + 329 + ], + "spans": [ + { + "bbox": [ + 106, + 315, + 505, + 329 + ], + "score": 1.0, + "content": "• We follow the fine-tuning procedure in [70]. While their zero-shot RAFT results are comparable", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 115, + 326, + 505, + 339 + ], + "spans": [ + { + "bbox": [ + 115, + 326, + 505, + 339 + ], + "score": 1.0, + "content": "to FlowFormer on Sintel and KITTI, the fine-tuned RAFT-it is significantly better on KITTI but", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 337, + 506, + 350 + ], + "spans": [ + { + "bbox": [ + 115, + 337, + 506, + 350 + ], + "score": 1.0, + "content": "less accurate on Sintel than FlowFormer. It is possible that the fine-tuning procedure (e.g. dataset", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 116, + 349, + 445, + 360 + ], + "spans": [ + { + "bbox": [ + 116, + 349, + 445, + 360 + ], + "score": 1.0, + "content": "mixture or augmentations) developed in [70] is more suited for KITTI than Sintel.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 111, + 365, + 505, + 378 + ], + "spans": [ + { + "bbox": [ + 111, + 365, + 505, + 378 + ], + "score": 1.0, + "content": "Another possible reason is that there is substantial domain gap between the training and test data on", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 375, + 505, + 388 + ], + "spans": [ + { + "bbox": [ + 115, + 375, + 505, + 388 + ], + "score": 1.0, + "content": "Sintel than KITTI. On Sintel test, there is a particular sequence “Ambush 1”, where the girl’s right", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 116, + 387, + 506, + 399 + ], + "spans": [ + { + "bbox": [ + 116, + 387, + 506, + 399 + ], + "score": 1.0, + "content": "arm moves out of the image boundary. Our method has an AEPE close to 30 while FlowFormer", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 397, + 505, + 410 + ], + "spans": [ + { + "bbox": [ + 115, + 397, + 505, + 410 + ], + "score": 1.0, + "content": "has lower than 10. It is likely that the attention on the cost volume mechanism by FlowFormer can", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 116, + 410, + 506, + 421 + ], + "spans": [ + { + "bbox": [ + 116, + 410, + 506, + 421 + ], + "score": 1.0, + "content": "better reason about the motion globally and handles this particular sequence well. This particular", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 116, + 420, + 506, + 431 + ], + "spans": [ + { + "bbox": [ + 116, + 420, + 506, + 431 + ], + "score": 1.0, + "content": "sequence may account for the major difference in the overall results; among 12 available results on", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 430, + 507, + 442 + ], + "spans": [ + { + "bbox": [ + 115, + 430, + 507, + 442 + ], + "score": 1.0, + "content": "the Sintel website, ours has lower AEPE on 11 sequences but a higher AEPE on the “Ambush 1”", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 116, + 442, + 469, + 454 + ], + "spans": [ + { + "bbox": [ + 116, + 442, + 469, + 454 + ], + "score": 1.0, + "content": "sequence, as shown in Table 8. Figure 16 in the appendix further provides visualization.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 17.5 + }, + { + "type": "table", + "bbox": [ + 221, + 479, + 388, + 595 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 111, + 466, + 496, + 477 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 112, + 465, + 497, + 479 + ], + "spans": [ + { + "bbox": [ + 112, + 465, + 497, + 479 + ], + "score": 1.0, + "content": "Table 8: Average end-point error (AEPE) on 12 Sintel test sequences available from the public website.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "table_body", + "bbox": [ + 221, + 479, + 388, + 595 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 221, + 479, + 388, + 595 + ], + "spans": [ + { + "bbox": [ + 221, + 479, + 388, + 595 + ], + "score": 0.978, + "html": "
SequenceOursFlowFormer [23]
Perturbed Market 30.7870.869
Perturbed Shaman 10.2190.252
Ambush 129.338.141
Ambush 32.8552.973
Bamboo 30.4150.577
Cave 32.0422.352
Market 10.7191.174
Market 45.5178.768
Mountain 20.1760.518
Temple 10.4520.612
Tiger0.4130.596
Wall1.6391.723
", + "type": "table", + "image_path": "5449ce339e0581e45d376b37e0fb8989e6655e38a2d63a4122a6f075b5b0c474.jpg" + } + ] + } + ], + "index": 29, + "virtual_lines": [ + { + "bbox": [ + 221, + 479, + 388, + 491.8888888888889 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 221, + 491.8888888888889, + 388, + 504.7777777777778 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 221, + 504.7777777777778, + 388, + 517.6666666666667 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 221, + 517.6666666666667, + 388, + 530.5555555555557 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 221, + 530.5555555555557, + 388, + 543.4444444444446 + ], + "spans": [], + "index": 29 + }, + { + "bbox": [ + 221, + 543.4444444444446, + 388, + 556.3333333333335 + ], + "spans": [], + "index": 30 + }, + { + "bbox": [ + 221, + 556.3333333333335, + 388, + 569.2222222222224 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 221, + 569.2222222222224, + 388, + 582.1111111111113 + ], + "spans": [], + "index": 32 + }, + { + "bbox": [ + 221, + 582.1111111111113, + 388, + 595.0000000000002 + ], + "spans": [], + "index": 33 + } + ] + } + ], + "index": 26.5 + }, + { + "type": "title", + "bbox": [ + 106, + 615, + 183, + 629 + ], + "lines": [ + { + "bbox": [ + 104, + 614, + 185, + 632 + ], + "spans": [ + { + "bbox": [ + 104, + 614, + 185, + 632 + ], + "score": 1.0, + "content": "6 Conclusion", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 34 + }, + { + "type": "text", + "bbox": [ + 106, + 634, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 635, + 505, + 646 + ], + "spans": [ + { + "bbox": [ + 106, + 635, + 505, + 646 + ], + "score": 1.0, + "content": "We introduced a simple denoising diffusion model for monocular depth and optical flow estimation", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 646, + 506, + 658 + ], + "spans": [ + { + "bbox": [ + 106, + 646, + 506, + 658 + ], + "score": 1.0, + "content": "using an image-to-image translation framework. Our generative approach obtains state-of-the-art", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 656, + 505, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 656, + 505, + 668 + ], + "score": 1.0, + "content": "results without task-specific architectures or loss functions. In particular, our model achieves an Fl-all", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 667, + 504, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 667, + 140, + 679 + ], + "score": 1.0, + "content": "score of", + "type": "text" + }, + { + "bbox": [ + 141, + 667, + 167, + 678 + ], + "score": 0.88, + "content": "3 . 2 6 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 168, + 667, + 235, + 679 + ], + "score": 1.0, + "content": "on KITTI, about", + "type": "text" + }, + { + "bbox": [ + 236, + 667, + 255, + 678 + ], + "score": 0.88, + "content": "2 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 256, + 667, + 504, + 679 + ], + "score": 1.0, + "content": "better than the best published method [70]. Further, our model", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "score": 1.0, + "content": "captures the multi-modality and uncertainty through multiple samples from the posterior. It also", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 689, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 106, + 689, + 505, + 701 + ], + "score": 1.0, + "content": "allows imputation of missing values, which enables iterative generation of 3D scenes conditioned on", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 700, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 505, + 712 + ], + "score": 1.0, + "content": "a text prompt. Our work suggests that diffusion models could be a simple and generic framework for", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 710, + 375, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 375, + 723 + ], + "score": 1.0, + "content": "dense vision tasks, and we hope to see more work in this direction.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 38.5 + } + ], + "page_idx": 9, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "10", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 109, + 72, + 501, + 182 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 109, + 72, + 501, + 182 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 109, + 72, + 501, + 182 + ], + "spans": [ + { + "bbox": [ + 109, + 72, + 501, + 182 + ], + "score": 0.967, + "type": "image", + "image_path": "01253d8e088f9b0c7e2f798a6acd974225d4c1b6913503fb0f8e02b7c815bbe1.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 109, + 72, + 501, + 108.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 109, + 108.66666666666666, + 501, + 145.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 109, + 145.33333333333331, + 501, + 181.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 191, + 505, + 252 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 191, + 505, + 203 + ], + "spans": [ + { + "bbox": [ + 106, + 191, + 505, + 203 + ], + "score": 1.0, + "content": "Figure 7: Application of zero-shot depth completion with our model by incorporating it into an iterative 3D", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 201, + 506, + 213 + ], + "spans": [ + { + "bbox": [ + 105, + 201, + 506, + 213 + ], + "score": 1.0, + "content": "scene generation pipeline. Starting with a initial image (optionally generated from a text-to-image model), we", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 211, + 506, + 223 + ], + "spans": [ + { + "bbox": [ + 105, + 211, + 506, + 223 + ], + "score": 1.0, + "content": "sample an image-only conditioned depth map using our model. The image-depth pair is added to a point cloud.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 220, + 506, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 220, + 506, + 233 + ], + "score": 1.0, + "content": "We then iteratively render images and depth maps (with holes) from this point cloud by moving the camera. We", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "spans": [ + { + "bbox": [ + 105, + 230, + 506, + 244 + ], + "score": 1.0, + "content": "then fill image holes using an existing image inpainter (optionally text conditioned), and then use our model", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 241, + 504, + 252 + ], + "spans": [ + { + "bbox": [ + 106, + 241, + 504, + 252 + ], + "score": 1.0, + "content": "with replacement guidance to impute missing depths (conditioned on the filled RGB image and known depth).", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 5.5 + } + ], + "index": 3.25 + }, + { + "type": "text", + "bbox": [ + 107, + 271, + 504, + 305 + ], + "lines": [ + { + "bbox": [ + 105, + 271, + 505, + 283 + ], + "spans": [ + { + "bbox": [ + 105, + 271, + 505, + 283 + ], + "score": 1.0, + "content": "Sintel fine-tuning. Under the zero-shot setting, our method achieves state-of-the-art results on both", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 281, + 505, + 295 + ], + "spans": [ + { + "bbox": [ + 105, + 281, + 505, + 295 + ], + "score": 1.0, + "content": "Sintel Final and KITTI. Under the fine-tuning setting, ours is state-of-the-art on KITTI but is behind", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 291, + 506, + 307 + ], + "spans": [ + { + "bbox": [ + 105, + 291, + 506, + 307 + ], + "score": 1.0, + "content": "FlowFormer [23] on Sintel Final. We discuss several possible reasons for why this may be the case.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10, + "bbox_fs": [ + 105, + 271, + 506, + 307 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 315, + 505, + 453 + ], + "lines": [ + { + "bbox": [ + 106, + 315, + 505, + 329 + ], + "spans": [ + { + "bbox": [ + 106, + 315, + 505, + 329 + ], + "score": 1.0, + "content": "• We follow the fine-tuning procedure in [70]. While their zero-shot RAFT results are comparable", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 115, + 326, + 505, + 339 + ], + "spans": [ + { + "bbox": [ + 115, + 326, + 505, + 339 + ], + "score": 1.0, + "content": "to FlowFormer on Sintel and KITTI, the fine-tuned RAFT-it is significantly better on KITTI but", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 337, + 506, + 350 + ], + "spans": [ + { + "bbox": [ + 115, + 337, + 506, + 350 + ], + "score": 1.0, + "content": "less accurate on Sintel than FlowFormer. It is possible that the fine-tuning procedure (e.g. dataset", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 116, + 349, + 445, + 360 + ], + "spans": [ + { + "bbox": [ + 116, + 349, + 445, + 360 + ], + "score": 1.0, + "content": "mixture or augmentations) developed in [70] is more suited for KITTI than Sintel.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 111, + 365, + 505, + 378 + ], + "spans": [ + { + "bbox": [ + 111, + 365, + 505, + 378 + ], + "score": 1.0, + "content": "Another possible reason is that there is substantial domain gap between the training and test data on", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 375, + 505, + 388 + ], + "spans": [ + { + "bbox": [ + 115, + 375, + 505, + 388 + ], + "score": 1.0, + "content": "Sintel than KITTI. On Sintel test, there is a particular sequence “Ambush 1”, where the girl’s right", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 116, + 387, + 506, + 399 + ], + "spans": [ + { + "bbox": [ + 116, + 387, + 506, + 399 + ], + "score": 1.0, + "content": "arm moves out of the image boundary. Our method has an AEPE close to 30 while FlowFormer", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 397, + 505, + 410 + ], + "spans": [ + { + "bbox": [ + 115, + 397, + 505, + 410 + ], + "score": 1.0, + "content": "has lower than 10. It is likely that the attention on the cost volume mechanism by FlowFormer can", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 116, + 410, + 506, + 421 + ], + "spans": [ + { + "bbox": [ + 116, + 410, + 506, + 421 + ], + "score": 1.0, + "content": "better reason about the motion globally and handles this particular sequence well. This particular", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 116, + 420, + 506, + 431 + ], + "spans": [ + { + "bbox": [ + 116, + 420, + 506, + 431 + ], + "score": 1.0, + "content": "sequence may account for the major difference in the overall results; among 12 available results on", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 430, + 507, + 442 + ], + "spans": [ + { + "bbox": [ + 115, + 430, + 507, + 442 + ], + "score": 1.0, + "content": "the Sintel website, ours has lower AEPE on 11 sequences but a higher AEPE on the “Ambush 1”", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 116, + 442, + 469, + 454 + ], + "spans": [ + { + "bbox": [ + 116, + 442, + 469, + 454 + ], + "score": 1.0, + "content": "sequence, as shown in Table 8. Figure 16 in the appendix further provides visualization.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 17.5, + "bbox_fs": [ + 106, + 315, + 507, + 454 + ] + }, + { + "type": "table", + "bbox": [ + 221, + 479, + 388, + 595 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 111, + 466, + 496, + 477 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 112, + 465, + 497, + 479 + ], + "spans": [ + { + "bbox": [ + 112, + 465, + 497, + 479 + ], + "score": 1.0, + "content": "Table 8: Average end-point error (AEPE) on 12 Sintel test sequences available from the public website.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "table_body", + "bbox": [ + 221, + 479, + 388, + 595 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 221, + 479, + 388, + 595 + ], + "spans": [ + { + "bbox": [ + 221, + 479, + 388, + 595 + ], + "score": 0.978, + "html": "
SequenceOursFlowFormer [23]
Perturbed Market 30.7870.869
Perturbed Shaman 10.2190.252
Ambush 129.338.141
Ambush 32.8552.973
Bamboo 30.4150.577
Cave 32.0422.352
Market 10.7191.174
Market 45.5178.768
Mountain 20.1760.518
Temple 10.4520.612
Tiger0.4130.596
Wall1.6391.723
", + "type": "table", + "image_path": "5449ce339e0581e45d376b37e0fb8989e6655e38a2d63a4122a6f075b5b0c474.jpg" + } + ] + } + ], + "index": 29, + "virtual_lines": [ + { + "bbox": [ + 221, + 479, + 388, + 491.8888888888889 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 221, + 491.8888888888889, + 388, + 504.7777777777778 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 221, + 504.7777777777778, + 388, + 517.6666666666667 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 221, + 517.6666666666667, + 388, + 530.5555555555557 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 221, + 530.5555555555557, + 388, + 543.4444444444446 + ], + "spans": [], + "index": 29 + }, + { + "bbox": [ + 221, + 543.4444444444446, + 388, + 556.3333333333335 + ], + "spans": [], + "index": 30 + }, + { + "bbox": [ + 221, + 556.3333333333335, + 388, + 569.2222222222224 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 221, + 569.2222222222224, + 388, + 582.1111111111113 + ], + "spans": [], + "index": 32 + }, + { + "bbox": [ + 221, + 582.1111111111113, + 388, + 595.0000000000002 + ], + "spans": [], + "index": 33 + } + ] + } + ], + "index": 26.5 + }, + { + "type": "title", + "bbox": [ + 106, + 615, + 183, + 629 + ], + "lines": [ + { + "bbox": [ + 104, + 614, + 185, + 632 + ], + "spans": [ + { + "bbox": [ + 104, + 614, + 185, + 632 + ], + "score": 1.0, + "content": "6 Conclusion", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 34 + }, + { + "type": "text", + "bbox": [ + 106, + 634, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 635, + 505, + 646 + ], + "spans": [ + { + "bbox": [ + 106, + 635, + 505, + 646 + ], + "score": 1.0, + "content": "We introduced a simple denoising diffusion model for monocular depth and optical flow estimation", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 646, + 506, + 658 + ], + "spans": [ + { + "bbox": [ + 106, + 646, + 506, + 658 + ], + "score": 1.0, + "content": "using an image-to-image translation framework. Our generative approach obtains state-of-the-art", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 656, + 505, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 656, + 505, + 668 + ], + "score": 1.0, + "content": "results without task-specific architectures or loss functions. In particular, our model achieves an Fl-all", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 667, + 504, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 667, + 140, + 679 + ], + "score": 1.0, + "content": "score of", + "type": "text" + }, + { + "bbox": [ + 141, + 667, + 167, + 678 + ], + "score": 0.88, + "content": "3 . 2 6 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 168, + 667, + 235, + 679 + ], + "score": 1.0, + "content": "on KITTI, about", + "type": "text" + }, + { + "bbox": [ + 236, + 667, + 255, + 678 + ], + "score": 0.88, + "content": "2 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 256, + 667, + 504, + 679 + ], + "score": 1.0, + "content": "better than the best published method [70]. Further, our model", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "score": 1.0, + "content": "captures the multi-modality and uncertainty through multiple samples from the posterior. It also", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 689, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 106, + 689, + 505, + 701 + ], + "score": 1.0, + "content": "allows imputation of missing values, which enables iterative generation of 3D scenes conditioned on", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 700, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 505, + 712 + ], + "score": 1.0, + "content": "a text prompt. Our work suggests that diffusion models could be a simple and generic framework for", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 710, + 375, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 375, + 723 + ], + "score": 1.0, + "content": "dense vision tasks, and we hope to see more work in this direction.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 38.5, + "bbox_fs": [ + 105, + 635, + 506, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 73, + 190, + 84 + ], + "lines": [ + { + "bbox": [ + 106, + 72, + 192, + 86 + ], + "spans": [ + { + "bbox": [ + 106, + 72, + 192, + 86 + ], + "score": 1.0, + "content": "Acknowledgements", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 92, + 505, + 126 + ], + "lines": [ + { + "bbox": [ + 105, + 92, + 505, + 104 + ], + "spans": [ + { + "bbox": [ + 105, + 92, + 505, + 104 + ], + "score": 1.0, + "content": "We thank Ting Chen, Daniel Watson, Hugo Larochelle and the rest of Google DeepMind for feedback", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 102, + 507, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 102, + 507, + 117 + ], + "score": 1.0, + "content": "on this work. Thanks to Klaus Greff and Andrea Tagliasacchi for their help with the Kubric generator,", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 115, + 343, + 126 + ], + "spans": [ + { + "bbox": [ + 106, + 115, + 343, + 126 + ], + "score": 1.0, + "content": "and to Chitwan Saharia for help training the Palette model.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2 + }, + { + "type": "title", + "bbox": [ + 107, + 144, + 163, + 157 + ], + "lines": [ + { + "bbox": [ + 106, + 143, + 165, + 159 + ], + "spans": [ + { + "bbox": [ + 106, + 143, + 165, + 159 + ], + "score": 1.0, + "content": "References", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 107, + 163, + 507, + 725 + ], + "lines": [ + { + "bbox": [ + 111, + 163, + 505, + 177 + ], + "spans": [ + { + "bbox": [ + 111, + 163, + 505, + 177 + ], + "score": 1.0, + "content": "[1] Ashutosh Agarwal and Chetan Arora. Attention Attention Everywhere: Monocular depth", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 125, + 173, + 319, + 189 + ], + "spans": [ + { + "bbox": [ + 125, + 173, + 319, + 189 + ], + "score": 1.0, + "content": "prediction with skip attention. In WACV, 2023.", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 110, + 191, + 506, + 208 + ], + "spans": [ + { + "bbox": [ + 110, + 191, + 506, + 208 + ], + "score": 1.0, + "content": "[2] Samy Bengio, Oriol Vinyals, Navdeep Jaitly, and Noam Shazeer. Scheduled sampling for", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 126, + 204, + 391, + 219 + ], + "spans": [ + { + "bbox": [ + 126, + 204, + 391, + 219 + ], + "score": 1.0, + "content": "sequence prediction with recurrent neural networks. NIPS, 2015.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 109, + 221, + 507, + 239 + ], + "spans": [ + { + "bbox": [ + 109, + 221, + 507, + 239 + ], + "score": 1.0, + "content": "[3] Shariq Farooq Bhat, Ibraheem Alhashim, and Peter Wonka. AdaBins: Depth estimation using", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 126, + 232, + 329, + 249 + ], + "spans": [ + { + "bbox": [ + 126, + 232, + 329, + 249 + ], + "score": 1.0, + "content": "adaptive bins. In CVPR, pages 4009–4018, 2021.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 109, + 250, + 507, + 268 + ], + "spans": [ + { + "bbox": [ + 109, + 250, + 507, + 268 + ], + "score": 1.0, + "content": "[4] Daniel J. Butler, Jonas Wulff, Garrett B. Stanley, and Michael J. Black. A naturalistic open", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 126, + 262, + 426, + 277 + ], + "spans": [ + { + "bbox": [ + 126, + 262, + 426, + 277 + ], + "score": 1.0, + "content": "source movie for optical flow evaluation. In ECCV, pages 611–625, 2012.", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 110, + 281, + 507, + 297 + ], + "spans": [ + { + "bbox": [ + 110, + 281, + 507, + 297 + ], + "score": 1.0, + "content": "[5] Yuanzhouhan Cao, Zifeng Wu, and Chunhua Shen. Estimating depth from monocular images", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 126, + 292, + 507, + 308 + ], + "spans": [ + { + "bbox": [ + 126, + 292, + 507, + 308 + ], + "score": 1.0, + "content": "as classification using deep fully convolutional residual networks. IEEE T-CSVT, 28(11):", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 127, + 302, + 204, + 318 + ], + "spans": [ + { + "bbox": [ + 127, + 302, + 204, + 318 + ], + "score": 1.0, + "content": "3174–3182, 2017.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 111, + 323, + 506, + 336 + ], + "spans": [ + { + "bbox": [ + 111, + 323, + 506, + 336 + ], + "score": 1.0, + "content": "[6] Angel X. Chang, Thomas Funkhouser, Leonidas Guibas, Pat Hanrahan, Qixing Huang, Zimo", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 126, + 333, + 507, + 347 + ], + "spans": [ + { + "bbox": [ + 126, + 333, + 507, + 347 + ], + "score": 1.0, + "content": "Li, Silvio Savarese, Manolis Savva, Shuran Song, Hao Su, Jianxiong Xiao, Li Yi, and Fisher Yu.", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 126, + 343, + 447, + 360 + ], + "spans": [ + { + "bbox": [ + 126, + 343, + 447, + 360 + ], + "score": 1.0, + "content": "ShapeNet: An information-rich 3D model repository. arXiv:1512.03012, 2015.", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 110, + 362, + 506, + 378 + ], + "spans": [ + { + "bbox": [ + 110, + 362, + 506, + 378 + ], + "score": 1.0, + "content": "[7] Ting Chen, Lala Li, Saurabh Saxena, Geoffrey Hinton, and David J. Fleet. A generalist", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 126, + 373, + 435, + 389 + ], + "spans": [ + { + "bbox": [ + 126, + 373, + 435, + 389 + ], + "score": 1.0, + "content": "framework for panoptic segmentation of images and videos. In ICCV, 2023.", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 109, + 389, + 507, + 411 + ], + "spans": [ + { + "bbox": [ + 109, + 389, + 507, + 411 + ], + "score": 1.0, + "content": "[8] Ting Chen, Ruixiang Zhang, and Geoffrey Hinton. Analog bits: Generating discrete data using", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 126, + 402, + 355, + 418 + ], + "spans": [ + { + "bbox": [ + 126, + 402, + 355, + 418 + ], + "score": 1.0, + "content": "diffusion models with self-conditioning. In ICLR, 2023.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 111, + 422, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 111, + 422, + 505, + 436 + ], + "score": 1.0, + "content": "[9] Angela Dai, Angel X. Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, and Matthias", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 127, + 434, + 492, + 447 + ], + "spans": [ + { + "bbox": [ + 127, + 434, + 492, + 447 + ], + "score": 1.0, + "content": "Nießner. ScanNet: Richly-annotated 3D reconstructions of indoor scenes. In CVPR, 2017.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 452, + 506, + 467 + ], + "spans": [ + { + "bbox": [ + 105, + 452, + 506, + 467 + ], + "score": 1.0, + "content": "[10] Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. ImageNet: A large-scale", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 126, + 462, + 376, + 478 + ], + "spans": [ + { + "bbox": [ + 126, + 462, + 376, + 478 + ], + "score": 1.0, + "content": "hierarchical image database. In CVPR, pages 248–255, 2009.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 481, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 481, + 506, + 496 + ], + "score": 1.0, + "content": "[11] Prafulla Dhariwal and Alex Nichol. Diffusion models beat GANs on image synthesis. In", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 126, + 492, + 191, + 506 + ], + "spans": [ + { + "bbox": [ + 126, + 492, + 191, + 506 + ], + "score": 1.0, + "content": "NeurIPS, 2022.", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 511, + 507, + 526 + ], + "spans": [ + { + "bbox": [ + 105, + 511, + 507, + 526 + ], + "score": 1.0, + "content": "[12] David Eigen and Rob Fergus. Predicting depth, surface normals and semantic labels with a", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 126, + 521, + 461, + 536 + ], + "spans": [ + { + "bbox": [ + 126, + 521, + 461, + 536 + ], + "score": 1.0, + "content": "common multi-scale convolutional architecture. In ICCV, pages 2650–2658, 2015.", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 541, + 506, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 541, + 506, + 556 + ], + "score": 1.0, + "content": "[13] David Eigen, Christian Puhrsch, and Rob Fergus. Depth map prediction from a single image", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 126, + 551, + 376, + 566 + ], + "spans": [ + { + "bbox": [ + 126, + 551, + 376, + 566 + ], + "score": 1.0, + "content": "using a multi-scale deep network. In NIPS, volume 27, 2014.", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 570, + 507, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 570, + 507, + 586 + ], + "score": 1.0, + "content": "[14] Philipp Fischer, Alexey Dosovitskiy, Eddy Ilg, Philip Häusser, Caner Hazirbas, Vladimir Golkov,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 126, + 581, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 126, + 581, + 506, + 596 + ], + "score": 1.0, + "content": "Patrick van der Smagt, Daniel Cremers, and Thomas Brox. FlowNet: Learning optical flow", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 128, + 592, + 386, + 606 + ], + "spans": [ + { + "bbox": [ + 128, + 592, + 386, + 606 + ], + "score": 1.0, + "content": "with convolutional networks. In ICCV, pages 2758–2766, 2015.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 610, + 506, + 626 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 506, + 626 + ], + "score": 1.0, + "content": "[15] Huan Fu, Mingming Gong, Chaohui Wang, Kayhan Batmanghelich, and Dacheng Tao. Deep", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 128, + 622, + 507, + 636 + ], + "spans": [ + { + "bbox": [ + 128, + 622, + 507, + 636 + ], + "score": 1.0, + "content": "ordinal regression network for monocular depth estimation. In CVPR, pages 2002–2011, 2018.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 640, + 506, + 656 + ], + "spans": [ + { + "bbox": [ + 105, + 640, + 506, + 656 + ], + "score": 1.0, + "content": "[16] Ravi Garg, Vijay Kumar Bg, Gustavo Carneiro, and Ian Reid. Unsupervised CNN for single", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 127, + 651, + 452, + 666 + ], + "spans": [ + { + "bbox": [ + 127, + 651, + 452, + 666 + ], + "score": 1.0, + "content": "view depth estimation: Geometry to the rescue. In ECCV, pages 740–756, 2016.", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 669, + 507, + 685 + ], + "spans": [ + { + "bbox": [ + 105, + 669, + 507, + 685 + ], + "score": 1.0, + "content": "[17] Andreas Geiger, Philip Lenz, Christoph Stiller, and Raquel Urtasun. Vision meets Robotics:", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 128, + 681, + 507, + 695 + ], + "spans": [ + { + "bbox": [ + 128, + 681, + 507, + 695 + ], + "score": 1.0, + "content": "The KITTI dataset. The International Journal of Robotics Research, 32(11):1231–1237, 2013.", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 104, + 698, + 507, + 717 + ], + "spans": [ + { + "bbox": [ + 104, + 698, + 507, + 717 + ], + "score": 1.0, + "content": "[18] Clément Godard, Oisin Mac Aodha, Michael Firman, and Gabriel J. Brostow. Digging into", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 127, + 711, + 447, + 725 + ], + "spans": [ + { + "bbox": [ + 127, + 711, + 447, + 725 + ], + "score": 1.0, + "content": "self-supervised monocular depth estimation. In ICCV, pages 3828–3838, 2019.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 24 + } + ], + "page_idx": 10, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 310, + 750 + ], + "lines": [ + { + "bbox": [ + 299, + 740, + 312, + 755 + ], + "spans": [ + { + "bbox": [ + 299, + 740, + 312, + 755 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 73, + 190, + 84 + ], + "lines": [ + { + "bbox": [ + 106, + 72, + 192, + 86 + ], + "spans": [ + { + "bbox": [ + 106, + 72, + 192, + 86 + ], + "score": 1.0, + "content": "Acknowledgements", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 92, + 505, + 126 + ], + "lines": [ + { + "bbox": [ + 105, + 92, + 505, + 104 + ], + "spans": [ + { + "bbox": [ + 105, + 92, + 505, + 104 + ], + "score": 1.0, + "content": "We thank Ting Chen, Daniel Watson, Hugo Larochelle and the rest of Google DeepMind for feedback", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 102, + 507, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 102, + 507, + 117 + ], + "score": 1.0, + "content": "on this work. Thanks to Klaus Greff and Andrea Tagliasacchi for their help with the Kubric generator,", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 115, + 343, + 126 + ], + "spans": [ + { + "bbox": [ + 106, + 115, + 343, + 126 + ], + "score": 1.0, + "content": "and to Chitwan Saharia for help training the Palette model.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2, + "bbox_fs": [ + 105, + 92, + 507, + 126 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 144, + 163, + 157 + ], + "lines": [ + { + "bbox": [ + 106, + 143, + 165, + 159 + ], + "spans": [ + { + "bbox": [ + 106, + 143, + 165, + 159 + ], + "score": 1.0, + "content": "References", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "list", + "bbox": [ + 107, + 163, + 507, + 725 + ], + "lines": [ + { + "bbox": [ + 111, + 163, + 505, + 177 + ], + "spans": [ + { + "bbox": [ + 111, + 163, + 505, + 177 + ], + "score": 1.0, + "content": "[1] Ashutosh Agarwal and Chetan Arora. Attention Attention Everywhere: Monocular depth", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 125, + 173, + 319, + 189 + ], + "spans": [ + { + "bbox": [ + 125, + 173, + 319, + 189 + ], + "score": 1.0, + "content": "prediction with skip attention. In WACV, 2023.", + "type": "text" + } + ], + "index": 6, + "is_list_end_line": true + }, + { + "bbox": [ + 110, + 191, + 506, + 208 + ], + "spans": [ + { + "bbox": [ + 110, + 191, + 506, + 208 + ], + "score": 1.0, + "content": "[2] Samy Bengio, Oriol Vinyals, Navdeep Jaitly, and Noam Shazeer. Scheduled sampling for", + "type": "text" + } + ], + "index": 7, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 204, + 391, + 219 + ], + "spans": [ + { + "bbox": [ + 126, + 204, + 391, + 219 + ], + "score": 1.0, + "content": "sequence prediction with recurrent neural networks. NIPS, 2015.", + "type": "text" + } + ], + "index": 8, + "is_list_end_line": true + }, + { + "bbox": [ + 109, + 221, + 507, + 239 + ], + "spans": [ + { + "bbox": [ + 109, + 221, + 507, + 239 + ], + "score": 1.0, + "content": "[3] Shariq Farooq Bhat, Ibraheem Alhashim, and Peter Wonka. AdaBins: Depth estimation using", + "type": "text" + } + ], + "index": 9, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 232, + 329, + 249 + ], + "spans": [ + { + "bbox": [ + 126, + 232, + 329, + 249 + ], + "score": 1.0, + "content": "adaptive bins. In CVPR, pages 4009–4018, 2021.", + "type": "text" + } + ], + "index": 10, + "is_list_end_line": true + }, + { + "bbox": [ + 109, + 250, + 507, + 268 + ], + "spans": [ + { + "bbox": [ + 109, + 250, + 507, + 268 + ], + "score": 1.0, + "content": "[4] Daniel J. Butler, Jonas Wulff, Garrett B. Stanley, and Michael J. Black. A naturalistic open", + "type": "text" + } + ], + "index": 11, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 262, + 426, + 277 + ], + "spans": [ + { + "bbox": [ + 126, + 262, + 426, + 277 + ], + "score": 1.0, + "content": "source movie for optical flow evaluation. In ECCV, pages 611–625, 2012.", + "type": "text" + } + ], + "index": 12, + "is_list_end_line": true + }, + { + "bbox": [ + 110, + 281, + 507, + 297 + ], + "spans": [ + { + "bbox": [ + 110, + 281, + 507, + 297 + ], + "score": 1.0, + "content": "[5] Yuanzhouhan Cao, Zifeng Wu, and Chunhua Shen. Estimating depth from monocular images", + "type": "text" + } + ], + "index": 13, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 292, + 507, + 308 + ], + "spans": [ + { + "bbox": [ + 126, + 292, + 507, + 308 + ], + "score": 1.0, + "content": "as classification using deep fully convolutional residual networks. IEEE T-CSVT, 28(11):", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 127, + 302, + 204, + 318 + ], + "spans": [ + { + "bbox": [ + 127, + 302, + 204, + 318 + ], + "score": 1.0, + "content": "3174–3182, 2017.", + "type": "text" + } + ], + "index": 15, + "is_list_end_line": true + }, + { + "bbox": [ + 111, + 323, + 506, + 336 + ], + "spans": [ + { + "bbox": [ + 111, + 323, + 506, + 336 + ], + "score": 1.0, + "content": "[6] Angel X. Chang, Thomas Funkhouser, Leonidas Guibas, Pat Hanrahan, Qixing Huang, Zimo", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 126, + 333, + 507, + 347 + ], + "spans": [ + { + "bbox": [ + 126, + 333, + 507, + 347 + ], + "score": 1.0, + "content": "Li, Silvio Savarese, Manolis Savva, Shuran Song, Hao Su, Jianxiong Xiao, Li Yi, and Fisher Yu.", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 126, + 343, + 447, + 360 + ], + "spans": [ + { + "bbox": [ + 126, + 343, + 447, + 360 + ], + "score": 1.0, + "content": "ShapeNet: An information-rich 3D model repository. arXiv:1512.03012, 2015.", + "type": "text" + } + ], + "index": 18, + "is_list_end_line": true + }, + { + "bbox": [ + 110, + 362, + 506, + 378 + ], + "spans": [ + { + "bbox": [ + 110, + 362, + 506, + 378 + ], + "score": 1.0, + "content": "[7] Ting Chen, Lala Li, Saurabh Saxena, Geoffrey Hinton, and David J. Fleet. A generalist", + "type": "text" + } + ], + "index": 19, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 373, + 435, + 389 + ], + "spans": [ + { + "bbox": [ + 126, + 373, + 435, + 389 + ], + "score": 1.0, + "content": "framework for panoptic segmentation of images and videos. In ICCV, 2023.", + "type": "text" + } + ], + "index": 20, + "is_list_end_line": true + }, + { + "bbox": [ + 109, + 389, + 507, + 411 + ], + "spans": [ + { + "bbox": [ + 109, + 389, + 507, + 411 + ], + "score": 1.0, + "content": "[8] Ting Chen, Ruixiang Zhang, and Geoffrey Hinton. Analog bits: Generating discrete data using", + "type": "text" + } + ], + "index": 21, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 402, + 355, + 418 + ], + "spans": [ + { + "bbox": [ + 126, + 402, + 355, + 418 + ], + "score": 1.0, + "content": "diffusion models with self-conditioning. In ICLR, 2023.", + "type": "text" + } + ], + "index": 22, + "is_list_end_line": true + }, + { + "bbox": [ + 111, + 422, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 111, + 422, + 505, + 436 + ], + "score": 1.0, + "content": "[9] Angela Dai, Angel X. Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, and Matthias", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 127, + 434, + 492, + 447 + ], + "spans": [ + { + "bbox": [ + 127, + 434, + 492, + 447 + ], + "score": 1.0, + "content": "Nießner. ScanNet: Richly-annotated 3D reconstructions of indoor scenes. In CVPR, 2017.", + "type": "text" + } + ], + "index": 24, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 452, + 506, + 467 + ], + "spans": [ + { + "bbox": [ + 105, + 452, + 506, + 467 + ], + "score": 1.0, + "content": "[10] Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. ImageNet: A large-scale", + "type": "text" + } + ], + "index": 25, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 462, + 376, + 478 + ], + "spans": [ + { + "bbox": [ + 126, + 462, + 376, + 478 + ], + "score": 1.0, + "content": "hierarchical image database. In CVPR, pages 248–255, 2009.", + "type": "text" + } + ], + "index": 26, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 481, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 481, + 506, + 496 + ], + "score": 1.0, + "content": "[11] Prafulla Dhariwal and Alex Nichol. Diffusion models beat GANs on image synthesis. In", + "type": "text" + } + ], + "index": 27, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 492, + 191, + 506 + ], + "spans": [ + { + "bbox": [ + 126, + 492, + 191, + 506 + ], + "score": 1.0, + "content": "NeurIPS, 2022.", + "type": "text" + } + ], + "index": 28, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 511, + 507, + 526 + ], + "spans": [ + { + "bbox": [ + 105, + 511, + 507, + 526 + ], + "score": 1.0, + "content": "[12] David Eigen and Rob Fergus. Predicting depth, surface normals and semantic labels with a", + "type": "text" + } + ], + "index": 29, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 521, + 461, + 536 + ], + "spans": [ + { + "bbox": [ + 126, + 521, + 461, + 536 + ], + "score": 1.0, + "content": "common multi-scale convolutional architecture. In ICCV, pages 2650–2658, 2015.", + "type": "text" + } + ], + "index": 30, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 541, + 506, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 541, + 506, + 556 + ], + "score": 1.0, + "content": "[13] David Eigen, Christian Puhrsch, and Rob Fergus. Depth map prediction from a single image", + "type": "text" + } + ], + "index": 31, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 551, + 376, + 566 + ], + "spans": [ + { + "bbox": [ + 126, + 551, + 376, + 566 + ], + "score": 1.0, + "content": "using a multi-scale deep network. In NIPS, volume 27, 2014.", + "type": "text" + } + ], + "index": 32, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 570, + 507, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 570, + 507, + 586 + ], + "score": 1.0, + "content": "[14] Philipp Fischer, Alexey Dosovitskiy, Eddy Ilg, Philip Häusser, Caner Hazirbas, Vladimir Golkov,", + "type": "text" + } + ], + "index": 33, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 581, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 126, + 581, + 506, + 596 + ], + "score": 1.0, + "content": "Patrick van der Smagt, Daniel Cremers, and Thomas Brox. FlowNet: Learning optical flow", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 128, + 592, + 386, + 606 + ], + "spans": [ + { + "bbox": [ + 128, + 592, + 386, + 606 + ], + "score": 1.0, + "content": "with convolutional networks. In ICCV, pages 2758–2766, 2015.", + "type": "text" + } + ], + "index": 35, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 610, + 506, + 626 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 506, + 626 + ], + "score": 1.0, + "content": "[15] Huan Fu, Mingming Gong, Chaohui Wang, Kayhan Batmanghelich, and Dacheng Tao. Deep", + "type": "text" + } + ], + "index": 36, + "is_list_start_line": true + }, + { + "bbox": [ + 128, + 622, + 507, + 636 + ], + "spans": [ + { + "bbox": [ + 128, + 622, + 507, + 636 + ], + "score": 1.0, + "content": "ordinal regression network for monocular depth estimation. In CVPR, pages 2002–2011, 2018.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 640, + 506, + 656 + ], + "spans": [ + { + "bbox": [ + 105, + 640, + 506, + 656 + ], + "score": 1.0, + "content": "[16] Ravi Garg, Vijay Kumar Bg, Gustavo Carneiro, and Ian Reid. Unsupervised CNN for single", + "type": "text" + } + ], + "index": 38, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 651, + 452, + 666 + ], + "spans": [ + { + "bbox": [ + 127, + 651, + 452, + 666 + ], + "score": 1.0, + "content": "view depth estimation: Geometry to the rescue. In ECCV, pages 740–756, 2016.", + "type": "text" + } + ], + "index": 39, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 669, + 507, + 685 + ], + "spans": [ + { + "bbox": [ + 105, + 669, + 507, + 685 + ], + "score": 1.0, + "content": "[17] Andreas Geiger, Philip Lenz, Christoph Stiller, and Raquel Urtasun. Vision meets Robotics:", + "type": "text" + } + ], + "index": 40, + "is_list_start_line": true + }, + { + "bbox": [ + 128, + 681, + 507, + 695 + ], + "spans": [ + { + "bbox": [ + 128, + 681, + 507, + 695 + ], + "score": 1.0, + "content": "The KITTI dataset. The International Journal of Robotics Research, 32(11):1231–1237, 2013.", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 104, + 698, + 507, + 717 + ], + "spans": [ + { + "bbox": [ + 104, + 698, + 507, + 717 + ], + "score": 1.0, + "content": "[18] Clément Godard, Oisin Mac Aodha, Michael Firman, and Gabriel J. Brostow. Digging into", + "type": "text" + } + ], + "index": 42, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 711, + 447, + 725 + ], + "spans": [ + { + "bbox": [ + 127, + 711, + 447, + 725 + ], + "score": 1.0, + "content": "self-supervised monocular depth estimation. In ICCV, pages 3828–3838, 2019.", + "type": "text" + } + ], + "index": 43, + "is_list_end_line": true + } + ], + "index": 24, + "bbox_fs": [ + 104, + 163, + 507, + 725 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 72, + 507, + 150 + ], + "lines": [ + { + "bbox": [ + 105, + 72, + 507, + 85 + ], + "spans": [ + { + "bbox": [ + 105, + 72, + 507, + 85 + ], + "score": 1.0, + "content": "[19] Klaus Greff, Francois Belletti, Lucas Beyer, Carl Doersch, Yilun Du, Daniel Duckworth, David J.", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 126, + 83, + 506, + 96 + ], + "spans": [ + { + "bbox": [ + 126, + 83, + 506, + 96 + ], + "score": 1.0, + "content": "Fleet, Dan Gnanapragasam, Florian Golemo, Charles Herrmann, Thomas Kipf, Abhijit Kundu,", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 127, + 94, + 506, + 108 + ], + "spans": [ + { + "bbox": [ + 127, + 94, + 506, + 108 + ], + "score": 1.0, + "content": "Dmitry Lagun, Issam Laradji, Hsueh-Ti (Derek) Liu, Henning Meyer, Yishu Miao, Derek", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 126, + 104, + 507, + 119 + ], + "spans": [ + { + "bbox": [ + 126, + 104, + 507, + 119 + ], + "score": 1.0, + "content": "Nowrouzezahrai, Cengiz Oztireli, Etienne Pot, Noha Radwan, Daniel Rebain, Sara Sabour,", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 127, + 115, + 506, + 129 + ], + "spans": [ + { + "bbox": [ + 127, + 115, + 506, + 129 + ], + "score": 1.0, + "content": "Mehdi S. M. Sajjadi, Matan Sela, Vincent Sitzmann, Austin Stone, Deqing Sun, Suhani Vora,", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 127, + 127, + 506, + 140 + ], + "spans": [ + { + "bbox": [ + 127, + 127, + 506, + 140 + ], + "score": 1.0, + "content": "Ziyu Wang, Tianhao Wu, Kwang Moo Yi, Fangcheng Zhong, and Andrea Tagliasacchi. Kubric:", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 127, + 138, + 406, + 150 + ], + "spans": [ + { + "bbox": [ + 127, + 138, + 406, + 150 + ], + "score": 1.0, + "content": "A scalable dataset generator. In CVPR, pages 3749–3761, June 2022.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 105, + 158, + 504, + 181 + ], + "lines": [ + { + "bbox": [ + 106, + 158, + 506, + 171 + ], + "spans": [ + { + "bbox": [ + 106, + 158, + 506, + 171 + ], + "score": 1.0, + "content": "[20] Ankur Handa, Viorica Patraucean, Vijay Badrinarayanan, Simon Stent, and Roberto Cipolla.", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 128, + 169, + 507, + 182 + ], + "spans": [ + { + "bbox": [ + 128, + 169, + 507, + 182 + ], + "score": 1.0, + "content": "Understanding real world indoor scenes with synthetic data. In CVPR, pages 4077–4085, 2016.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 106, + 189, + 505, + 212 + ], + "lines": [ + { + "bbox": [ + 106, + 187, + 507, + 202 + ], + "spans": [ + { + "bbox": [ + 106, + 187, + 507, + 202 + ], + "score": 1.0, + "content": "[21] Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising Diffusion Probabilistic Models. NeurIPS,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 126, + 199, + 153, + 212 + ], + "spans": [ + { + "bbox": [ + 126, + 199, + 153, + 212 + ], + "score": 1.0, + "content": "2020.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "text", + "bbox": [ + 107, + 219, + 505, + 253 + ], + "lines": [ + { + "bbox": [ + 105, + 220, + 506, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 220, + 506, + 233 + ], + "score": 1.0, + "content": "[22] Asmaa Hosni, Christoph Rhemann, Michael Bleyer, Carsten Rother, and Margrit Gelautz. Fast", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 127, + 231, + 506, + 243 + ], + "spans": [ + { + "bbox": [ + 127, + 231, + 506, + 243 + ], + "score": 1.0, + "content": "cost-volume filtering for visual correspondence and beyond. IEEE T-PAMI, 35(2):504–511,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 126, + 241, + 153, + 253 + ], + "spans": [ + { + "bbox": [ + 126, + 241, + 153, + 253 + ], + "score": 1.0, + "content": "2012.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12 + }, + { + "type": "text", + "bbox": [ + 107, + 261, + 504, + 296 + ], + "lines": [ + { + "bbox": [ + 105, + 261, + 506, + 276 + ], + "spans": [ + { + "bbox": [ + 105, + 261, + 506, + 276 + ], + "score": 1.0, + "content": "[23] Zhaoyang Huang, Xiaoyu Shi, Chao Zhang, Qiang Wang, Ka Chun Cheung, Hongwei Qin,", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 127, + 272, + 506, + 286 + ], + "spans": [ + { + "bbox": [ + 127, + 272, + 506, + 286 + ], + "score": 1.0, + "content": "Jifeng Dai, and Hongsheng Li. FlowFormer: A transformer architecture for optical flow. In", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 128, + 284, + 248, + 297 + ], + "spans": [ + { + "bbox": [ + 128, + 284, + 248, + 297 + ], + "score": 1.0, + "content": "ECCV, pages 668–685, 2022.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 106, + 304, + 504, + 327 + ], + "lines": [ + { + "bbox": [ + 106, + 303, + 505, + 318 + ], + "spans": [ + { + "bbox": [ + 106, + 303, + 505, + 318 + ], + "score": 1.0, + "content": "[24] Junhwa Hur and Stefan Roth. MirrorFlow: Exploiting symmetries in joint optical flow and", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 127, + 314, + 345, + 327 + ], + "spans": [ + { + "bbox": [ + 127, + 314, + 345, + 327 + ], + "score": 1.0, + "content": "occlusion estimation. In ICCV, pages 312–321, 2017.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5 + }, + { + "type": "text", + "bbox": [ + 106, + 335, + 504, + 358 + ], + "lines": [ + { + "bbox": [ + 106, + 334, + 506, + 348 + ], + "spans": [ + { + "bbox": [ + 106, + 334, + 506, + 348 + ], + "score": 1.0, + "content": "[25] Junhwa Hur and Stefan Roth. Iterative residual refinement for joint optical flow and occlusion", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 127, + 346, + 317, + 358 + ], + "spans": [ + { + "bbox": [ + 127, + 346, + 317, + 358 + ], + "score": 1.0, + "content": "estimation. In CVPR, pages 5754–5763, 2019.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5 + }, + { + "type": "text", + "bbox": [ + 107, + 366, + 505, + 399 + ], + "lines": [ + { + "bbox": [ + 105, + 366, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 105, + 366, + 505, + 379 + ], + "score": 1.0, + "content": "[26] Eddy Ilg, Nikolaus Mayer, Tonmoy Saikia, Margret Keuper, Alexey Dosovitskiy, and Thomas", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 126, + 376, + 505, + 390 + ], + "spans": [ + { + "bbox": [ + 126, + 376, + 505, + 390 + ], + "score": 1.0, + "content": "Brox. FlowNet 2.0: Evolution of optical flow estimation with deep networks. In CVPR, pages", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 128, + 388, + 203, + 399 + ], + "spans": [ + { + "bbox": [ + 128, + 388, + 203, + 399 + ], + "score": 1.0, + "content": "2462–2470, 2017.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 106, + 407, + 505, + 430 + ], + "lines": [ + { + "bbox": [ + 106, + 407, + 507, + 421 + ], + "spans": [ + { + "bbox": [ + 106, + 407, + 507, + 421 + ], + "score": 1.0, + "content": "[27] Serdar Ince and Janusz Konrad. Occlusion-aware optical flow estimation. IEEE T-IP, 17(8):", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 127, + 419, + 202, + 430 + ], + "spans": [ + { + "bbox": [ + 127, + 419, + 202, + 430 + ], + "score": 1.0, + "content": "1443–1451, 2008.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24.5 + }, + { + "type": "text", + "bbox": [ + 107, + 438, + 504, + 473 + ], + "lines": [ + { + "bbox": [ + 105, + 437, + 506, + 453 + ], + "spans": [ + { + "bbox": [ + 105, + 437, + 506, + 453 + ], + "score": 1.0, + "content": "[28] Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 127, + 450, + 506, + 462 + ], + "spans": [ + { + "bbox": [ + 127, + 450, + 506, + 462 + ], + "score": 1.0, + "content": "David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, et al. Perceiver", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 127, + 461, + 426, + 473 + ], + "spans": [ + { + "bbox": [ + 127, + 461, + 426, + 473 + ], + "score": 1.0, + "content": "IO: A general architecture for structured inputs & outputs. In ICLR, 2022.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 105, + 480, + 504, + 503 + ], + "lines": [ + { + "bbox": [ + 106, + 480, + 505, + 493 + ], + "spans": [ + { + "bbox": [ + 106, + 480, + 505, + 493 + ], + "score": 1.0, + "content": "[29] Jisoo Jeong, Jamie Lin, Fatih Porikli, and Nojun Kwak. Imposing consistency for optical flow", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 127, + 492, + 241, + 503 + ], + "spans": [ + { + "bbox": [ + 127, + 492, + 241, + 503 + ], + "score": 1.0, + "content": "estimation. In CVPR, 2022.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5 + }, + { + "type": "text", + "bbox": [ + 106, + 511, + 506, + 557 + ], + "lines": [ + { + "bbox": [ + 105, + 511, + 506, + 524 + ], + "spans": [ + { + "bbox": [ + 105, + 511, + 506, + 524 + ], + "score": 1.0, + "content": "[30] Daniel Kondermann, Rahul Nair, Katrin Honauer, Karsten Krispin, Jonas Andrulis, Alexander", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 126, + 521, + 507, + 536 + ], + "spans": [ + { + "bbox": [ + 126, + 521, + 507, + 536 + ], + "score": 1.0, + "content": "Brock, Burkhard Gussefeld, Mohsen Rahimimoghaddam, Sabine Hofmann, Claus Brenner,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 127, + 534, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 127, + 534, + 505, + 546 + ], + "score": 1.0, + "content": "et al. The HCI Benchmark Suite: Stereo and flow ground truth with uncertainties for urban", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 126, + 545, + 380, + 557 + ], + "spans": [ + { + "bbox": [ + 126, + 545, + 380, + 557 + ], + "score": 1.0, + "content": "autonomous driving. In CVPR Workshops, pages 19–28, 2016.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 32.5 + }, + { + "type": "text", + "bbox": [ + 105, + 564, + 505, + 598 + ], + "lines": [ + { + "bbox": [ + 105, + 564, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 506, + 577 + ], + "score": 1.0, + "content": "[31] Iro Laina, Christian Rupprecht, Vasileios Belagiannis, Federico Tombari, and Nassir Navab.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 126, + 574, + 507, + 588 + ], + "spans": [ + { + "bbox": [ + 126, + 574, + 507, + 588 + ], + "score": 1.0, + "content": "Deeper depth prediction with fully convolutional residual networks. In 3DV, pages 239–248,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 126, + 585, + 155, + 599 + ], + "spans": [ + { + "bbox": [ + 126, + 585, + 155, + 599 + ], + "score": 1.0, + "content": "2016.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36 + }, + { + "type": "text", + "bbox": [ + 105, + 606, + 505, + 630 + ], + "lines": [ + { + "bbox": [ + 106, + 606, + 506, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 606, + 506, + 619 + ], + "score": 1.0, + "content": "[32] Alex Lamb, Anirudh Goyal, Ying Zhang, Saizheng Zhang, Aaron Courville, and Yoshua Bengio.", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 127, + 618, + 470, + 631 + ], + "spans": [ + { + "bbox": [ + 127, + 618, + 470, + 631 + ], + "score": 1.0, + "content": "Professor Forcing: A new algorithm for training recurrent networks. NIPS, 29, 2016.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5 + }, + { + "type": "text", + "bbox": [ + 106, + 637, + 505, + 661 + ], + "lines": [ + { + "bbox": [ + 106, + 637, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 106, + 637, + 505, + 650 + ], + "score": 1.0, + "content": "[33] Gustav Larsson, Michael Maire, and Gregory Shakhnarovich. Learning representations for", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 127, + 648, + 355, + 661 + ], + "spans": [ + { + "bbox": [ + 127, + 648, + 355, + 661 + ], + "score": 1.0, + "content": "automatic colorization. In ECCV, pages 577–593, 2016.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 40.5 + }, + { + "type": "text", + "bbox": [ + 105, + 668, + 504, + 691 + ], + "lines": [ + { + "bbox": [ + 106, + 668, + 505, + 681 + ], + "spans": [ + { + "bbox": [ + 106, + 668, + 505, + 681 + ], + "score": 1.0, + "content": "[34] Jin Han Lee, Myung-Kyu Han, Dong Wook Ko, and Il Hong Suh. From big to small: Multi-scale", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 128, + 679, + 450, + 692 + ], + "spans": [ + { + "bbox": [ + 128, + 679, + 450, + 692 + ], + "score": 1.0, + "content": "local planar guidance for monocular depth estimation. arXiv:1907.10326, 2019.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42.5 + }, + { + "type": "text", + "bbox": [ + 107, + 700, + 502, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 700, + 504, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 700, + 504, + 712 + ], + "score": 1.0, + "content": "[35] Zhenyu Li, Xuyang Wang, Xianming Liu, and Junjun Jiang. BinsFormer: Revisiting adaptive", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 127, + 710, + 378, + 722 + ], + "spans": [ + { + "bbox": [ + 127, + 710, + 378, + 722 + ], + "score": 1.0, + "content": "bins for monocular depth estimation. arxiv.2204.00987, 2022.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44.5 + } + ], + "page_idx": 11, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "12", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 72, + 507, + 150 + ], + "lines": [ + { + "bbox": [ + 105, + 72, + 507, + 85 + ], + "spans": [ + { + "bbox": [ + 105, + 72, + 507, + 85 + ], + "score": 1.0, + "content": "[19] Klaus Greff, Francois Belletti, Lucas Beyer, Carl Doersch, Yilun Du, Daniel Duckworth, David J.", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 126, + 83, + 506, + 96 + ], + "spans": [ + { + "bbox": [ + 126, + 83, + 506, + 96 + ], + "score": 1.0, + "content": "Fleet, Dan Gnanapragasam, Florian Golemo, Charles Herrmann, Thomas Kipf, Abhijit Kundu,", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 127, + 94, + 506, + 108 + ], + "spans": [ + { + "bbox": [ + 127, + 94, + 506, + 108 + ], + "score": 1.0, + "content": "Dmitry Lagun, Issam Laradji, Hsueh-Ti (Derek) Liu, Henning Meyer, Yishu Miao, Derek", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 126, + 104, + 507, + 119 + ], + "spans": [ + { + "bbox": [ + 126, + 104, + 507, + 119 + ], + "score": 1.0, + "content": "Nowrouzezahrai, Cengiz Oztireli, Etienne Pot, Noha Radwan, Daniel Rebain, Sara Sabour,", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 127, + 115, + 506, + 129 + ], + "spans": [ + { + "bbox": [ + 127, + 115, + 506, + 129 + ], + "score": 1.0, + "content": "Mehdi S. M. Sajjadi, Matan Sela, Vincent Sitzmann, Austin Stone, Deqing Sun, Suhani Vora,", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 127, + 127, + 506, + 140 + ], + "spans": [ + { + "bbox": [ + 127, + 127, + 506, + 140 + ], + "score": 1.0, + "content": "Ziyu Wang, Tianhao Wu, Kwang Moo Yi, Fangcheng Zhong, and Andrea Tagliasacchi. Kubric:", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 127, + 138, + 406, + 150 + ], + "spans": [ + { + "bbox": [ + 127, + 138, + 406, + 150 + ], + "score": 1.0, + "content": "A scalable dataset generator. In CVPR, pages 3749–3761, June 2022.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 3, + "bbox_fs": [ + 105, + 72, + 507, + 150 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 158, + 504, + 181 + ], + "lines": [ + { + "bbox": [ + 106, + 158, + 506, + 171 + ], + "spans": [ + { + "bbox": [ + 106, + 158, + 506, + 171 + ], + "score": 1.0, + "content": "[20] Ankur Handa, Viorica Patraucean, Vijay Badrinarayanan, Simon Stent, and Roberto Cipolla.", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 128, + 169, + 507, + 182 + ], + "spans": [ + { + "bbox": [ + 128, + 169, + 507, + 182 + ], + "score": 1.0, + "content": "Understanding real world indoor scenes with synthetic data. In CVPR, pages 4077–4085, 2016.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5, + "bbox_fs": [ + 106, + 158, + 507, + 182 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 189, + 505, + 212 + ], + "lines": [ + { + "bbox": [ + 106, + 187, + 507, + 202 + ], + "spans": [ + { + "bbox": [ + 106, + 187, + 507, + 202 + ], + "score": 1.0, + "content": "[21] Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising Diffusion Probabilistic Models. NeurIPS,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 126, + 199, + 153, + 212 + ], + "spans": [ + { + "bbox": [ + 126, + 199, + 153, + 212 + ], + "score": 1.0, + "content": "2020.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5, + "bbox_fs": [ + 106, + 187, + 507, + 212 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 219, + 505, + 253 + ], + "lines": [ + { + "bbox": [ + 105, + 220, + 506, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 220, + 506, + 233 + ], + "score": 1.0, + "content": "[22] Asmaa Hosni, Christoph Rhemann, Michael Bleyer, Carsten Rother, and Margrit Gelautz. Fast", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 127, + 231, + 506, + 243 + ], + "spans": [ + { + "bbox": [ + 127, + 231, + 506, + 243 + ], + "score": 1.0, + "content": "cost-volume filtering for visual correspondence and beyond. IEEE T-PAMI, 35(2):504–511,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 126, + 241, + 153, + 253 + ], + "spans": [ + { + "bbox": [ + 126, + 241, + 153, + 253 + ], + "score": 1.0, + "content": "2012.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12, + "bbox_fs": [ + 105, + 220, + 506, + 253 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 261, + 504, + 296 + ], + "lines": [ + { + "bbox": [ + 105, + 261, + 506, + 276 + ], + "spans": [ + { + "bbox": [ + 105, + 261, + 506, + 276 + ], + "score": 1.0, + "content": "[23] Zhaoyang Huang, Xiaoyu Shi, Chao Zhang, Qiang Wang, Ka Chun Cheung, Hongwei Qin,", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 127, + 272, + 506, + 286 + ], + "spans": [ + { + "bbox": [ + 127, + 272, + 506, + 286 + ], + "score": 1.0, + "content": "Jifeng Dai, and Hongsheng Li. FlowFormer: A transformer architecture for optical flow. In", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 128, + 284, + 248, + 297 + ], + "spans": [ + { + "bbox": [ + 128, + 284, + 248, + 297 + ], + "score": 1.0, + "content": "ECCV, pages 668–685, 2022.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15, + "bbox_fs": [ + 105, + 261, + 506, + 297 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 304, + 504, + 327 + ], + "lines": [ + { + "bbox": [ + 106, + 303, + 505, + 318 + ], + "spans": [ + { + "bbox": [ + 106, + 303, + 505, + 318 + ], + "score": 1.0, + "content": "[24] Junhwa Hur and Stefan Roth. MirrorFlow: Exploiting symmetries in joint optical flow and", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 127, + 314, + 345, + 327 + ], + "spans": [ + { + "bbox": [ + 127, + 314, + 345, + 327 + ], + "score": 1.0, + "content": "occlusion estimation. In ICCV, pages 312–321, 2017.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5, + "bbox_fs": [ + 106, + 303, + 505, + 327 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 335, + 504, + 358 + ], + "lines": [ + { + "bbox": [ + 106, + 334, + 506, + 348 + ], + "spans": [ + { + "bbox": [ + 106, + 334, + 506, + 348 + ], + "score": 1.0, + "content": "[25] Junhwa Hur and Stefan Roth. Iterative residual refinement for joint optical flow and occlusion", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 127, + 346, + 317, + 358 + ], + "spans": [ + { + "bbox": [ + 127, + 346, + 317, + 358 + ], + "score": 1.0, + "content": "estimation. In CVPR, pages 5754–5763, 2019.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5, + "bbox_fs": [ + 106, + 334, + 506, + 358 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 366, + 505, + 399 + ], + "lines": [ + { + "bbox": [ + 105, + 366, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 105, + 366, + 505, + 379 + ], + "score": 1.0, + "content": "[26] Eddy Ilg, Nikolaus Mayer, Tonmoy Saikia, Margret Keuper, Alexey Dosovitskiy, and Thomas", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 126, + 376, + 505, + 390 + ], + "spans": [ + { + "bbox": [ + 126, + 376, + 505, + 390 + ], + "score": 1.0, + "content": "Brox. FlowNet 2.0: Evolution of optical flow estimation with deep networks. In CVPR, pages", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 128, + 388, + 203, + 399 + ], + "spans": [ + { + "bbox": [ + 128, + 388, + 203, + 399 + ], + "score": 1.0, + "content": "2462–2470, 2017.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22, + "bbox_fs": [ + 105, + 366, + 505, + 399 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 407, + 505, + 430 + ], + "lines": [ + { + "bbox": [ + 106, + 407, + 507, + 421 + ], + "spans": [ + { + "bbox": [ + 106, + 407, + 507, + 421 + ], + "score": 1.0, + "content": "[27] Serdar Ince and Janusz Konrad. Occlusion-aware optical flow estimation. IEEE T-IP, 17(8):", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 127, + 419, + 202, + 430 + ], + "spans": [ + { + "bbox": [ + 127, + 419, + 202, + 430 + ], + "score": 1.0, + "content": "1443–1451, 2008.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24.5, + "bbox_fs": [ + 106, + 407, + 507, + 430 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 438, + 504, + 473 + ], + "lines": [ + { + "bbox": [ + 105, + 437, + 506, + 453 + ], + "spans": [ + { + "bbox": [ + 105, + 437, + 506, + 453 + ], + "score": 1.0, + "content": "[28] Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 127, + 450, + 506, + 462 + ], + "spans": [ + { + "bbox": [ + 127, + 450, + 506, + 462 + ], + "score": 1.0, + "content": "David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, et al. Perceiver", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 127, + 461, + 426, + 473 + ], + "spans": [ + { + "bbox": [ + 127, + 461, + 426, + 473 + ], + "score": 1.0, + "content": "IO: A general architecture for structured inputs & outputs. In ICLR, 2022.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27, + "bbox_fs": [ + 105, + 437, + 506, + 473 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 480, + 504, + 503 + ], + "lines": [ + { + "bbox": [ + 106, + 480, + 505, + 493 + ], + "spans": [ + { + "bbox": [ + 106, + 480, + 505, + 493 + ], + "score": 1.0, + "content": "[29] Jisoo Jeong, Jamie Lin, Fatih Porikli, and Nojun Kwak. Imposing consistency for optical flow", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 127, + 492, + 241, + 503 + ], + "spans": [ + { + "bbox": [ + 127, + 492, + 241, + 503 + ], + "score": 1.0, + "content": "estimation. In CVPR, 2022.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5, + "bbox_fs": [ + 106, + 480, + 505, + 503 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 511, + 506, + 557 + ], + "lines": [ + { + "bbox": [ + 105, + 511, + 506, + 524 + ], + "spans": [ + { + "bbox": [ + 105, + 511, + 506, + 524 + ], + "score": 1.0, + "content": "[30] Daniel Kondermann, Rahul Nair, Katrin Honauer, Karsten Krispin, Jonas Andrulis, Alexander", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 126, + 521, + 507, + 536 + ], + "spans": [ + { + "bbox": [ + 126, + 521, + 507, + 536 + ], + "score": 1.0, + "content": "Brock, Burkhard Gussefeld, Mohsen Rahimimoghaddam, Sabine Hofmann, Claus Brenner,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 127, + 534, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 127, + 534, + 505, + 546 + ], + "score": 1.0, + "content": "et al. The HCI Benchmark Suite: Stereo and flow ground truth with uncertainties for urban", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 126, + 545, + 380, + 557 + ], + "spans": [ + { + "bbox": [ + 126, + 545, + 380, + 557 + ], + "score": 1.0, + "content": "autonomous driving. In CVPR Workshops, pages 19–28, 2016.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 32.5, + "bbox_fs": [ + 105, + 511, + 507, + 557 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 564, + 505, + 598 + ], + "lines": [ + { + "bbox": [ + 105, + 564, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 506, + 577 + ], + "score": 1.0, + "content": "[31] Iro Laina, Christian Rupprecht, Vasileios Belagiannis, Federico Tombari, and Nassir Navab.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 126, + 574, + 507, + 588 + ], + "spans": [ + { + "bbox": [ + 126, + 574, + 507, + 588 + ], + "score": 1.0, + "content": "Deeper depth prediction with fully convolutional residual networks. In 3DV, pages 239–248,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 126, + 585, + 155, + 599 + ], + "spans": [ + { + "bbox": [ + 126, + 585, + 155, + 599 + ], + "score": 1.0, + "content": "2016.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36, + "bbox_fs": [ + 105, + 564, + 507, + 599 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 606, + 505, + 630 + ], + "lines": [ + { + "bbox": [ + 106, + 606, + 506, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 606, + 506, + 619 + ], + "score": 1.0, + "content": "[32] Alex Lamb, Anirudh Goyal, Ying Zhang, Saizheng Zhang, Aaron Courville, and Yoshua Bengio.", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 127, + 618, + 470, + 631 + ], + "spans": [ + { + "bbox": [ + 127, + 618, + 470, + 631 + ], + "score": 1.0, + "content": "Professor Forcing: A new algorithm for training recurrent networks. NIPS, 29, 2016.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5, + "bbox_fs": [ + 106, + 606, + 506, + 631 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 637, + 505, + 661 + ], + "lines": [ + { + "bbox": [ + 106, + 637, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 106, + 637, + 505, + 650 + ], + "score": 1.0, + "content": "[33] Gustav Larsson, Michael Maire, and Gregory Shakhnarovich. Learning representations for", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 127, + 648, + 355, + 661 + ], + "spans": [ + { + "bbox": [ + 127, + 648, + 355, + 661 + ], + "score": 1.0, + "content": "automatic colorization. In ECCV, pages 577–593, 2016.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 40.5, + "bbox_fs": [ + 106, + 637, + 505, + 661 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 668, + 504, + 691 + ], + "lines": [ + { + "bbox": [ + 106, + 668, + 505, + 681 + ], + "spans": [ + { + "bbox": [ + 106, + 668, + 505, + 681 + ], + "score": 1.0, + "content": "[34] Jin Han Lee, Myung-Kyu Han, Dong Wook Ko, and Il Hong Suh. From big to small: Multi-scale", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 128, + 679, + 450, + 692 + ], + "spans": [ + { + "bbox": [ + 128, + 679, + 450, + 692 + ], + "score": 1.0, + "content": "local planar guidance for monocular depth estimation. arXiv:1907.10326, 2019.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42.5, + "bbox_fs": [ + 106, + 668, + 505, + 692 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 700, + 502, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 700, + 504, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 700, + 504, + 712 + ], + "score": 1.0, + "content": "[35] Zhenyu Li, Xuyang Wang, Xianming Liu, and Junjun Jiang. BinsFormer: Revisiting adaptive", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 127, + 710, + 378, + 722 + ], + "spans": [ + { + "bbox": [ + 127, + 710, + 378, + 722 + ], + "score": 1.0, + "content": "bins for monocular depth estimation. arxiv.2204.00987, 2022.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44.5, + "bbox_fs": [ + 106, + 700, + 504, + 722 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 38, + 508, + 724 + ], + "lines": [ + { + "bbox": [ + 104, + 70, + 507, + 88 + ], + "spans": [ + { + "bbox": [ + 104, + 70, + 507, + 88 + ], + "score": 1.0, + "content": "[36] Orly Liba, Longqi Cai, Yun-Ta Tsai, Elad Eban, Yair Movshovitz-Attias, Yael Pritch, Huizhong", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 127, + 83, + 506, + 98 + ], + "spans": [ + { + "bbox": [ + 127, + 83, + 506, + 98 + ], + "score": 1.0, + "content": "Chen, and Jonathan T. Barron. Sky Optimization: Semantically aware image processing of", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 126, + 94, + 389, + 108 + ], + "spans": [ + { + "bbox": [ + 126, + 94, + 389, + 108 + ], + "score": 1.0, + "content": "skies in low-light photography. In CVPR Workshops, June 2020.", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 113, + 506, + 127 + ], + "spans": [ + { + "bbox": [ + 105, + 113, + 506, + 127 + ], + "score": 1.0, + "content": "[37] Andrew Liu, Richard Tucker, Varun Jampani, Ameesh Makadia, Noah Snavely, and Angjoo", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 128, + 126, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 128, + 126, + 505, + 138 + ], + "score": 1.0, + "content": "Kanazawa. Infinite Nature: Perpetual view generation of natural scenes from a single image. In", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 127, + 136, + 181, + 148 + ], + "spans": [ + { + "bbox": [ + 127, + 136, + 181, + 148 + ], + "score": 1.0, + "content": "ICCV, 2021.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 154, + 506, + 168 + ], + "spans": [ + { + "bbox": [ + 105, + 154, + 506, + 168 + ], + "score": 1.0, + "content": "[38] Ao Luo, Fan Yang, Kunming Luo, Xin Li, Haoqiang Fan, and Shuaicheng Liu. Learning optical", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 126, + 164, + 412, + 180 + ], + "spans": [ + { + "bbox": [ + 126, + 164, + 412, + 180 + ], + "score": 1.0, + "content": "flow with adaptive graph reasoning. In AAAI, pages 1890–1898, 2022.", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 184, + 507, + 199 + ], + "spans": [ + { + "bbox": [ + 106, + 184, + 507, + 199 + ], + "score": 1.0, + "content": "[39] Zhaoyang Lv, Kihwan Kim, Alejandro Troccoli, Deqing Sun, James M. Rehg, and Jan Kautz.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 126, + 196, + 507, + 210 + ], + "spans": [ + { + "bbox": [ + 126, + 196, + 507, + 210 + ], + "score": 1.0, + "content": "Learning rigidity in dynamic scenes with a moving camera for 3D motion field estimation. In", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 126, + 206, + 249, + 221 + ], + "spans": [ + { + "bbox": [ + 126, + 206, + 249, + 221 + ], + "score": 1.0, + "content": "ECCV, pages 468–484, 2018.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 225, + 507, + 239 + ], + "spans": [ + { + "bbox": [ + 106, + 225, + 507, + 239 + ], + "score": 1.0, + "content": "[40] Nikolaus Mayer, Eddy Ilg, Philip Hausser, Philipp Fischer, Daniel Cremers, Alexey Dosovitskiy,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 126, + 235, + 507, + 250 + ], + "spans": [ + { + "bbox": [ + 126, + 235, + 507, + 250 + ], + "score": 1.0, + "content": "and Thomas Brox. A large dataset to train convolutional networks for disparity, optical flow,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 127, + 248, + 303, + 259 + ], + "spans": [ + { + "bbox": [ + 127, + 248, + 303, + 259 + ], + "score": 1.0, + "content": "and scene flow estimation. In CVPR, 2016.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 103, + 265, + 507, + 281 + ], + "spans": [ + { + "bbox": [ + 103, + 265, + 507, + 281 + ], + "score": 1.0, + "content": "[41] John McCormac, Ankur Handa, Stefan Leutenegger, and Andrew J. Davison. SceneNet RGB-D:", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 126, + 276, + 507, + 290 + ], + "spans": [ + { + "bbox": [ + 126, + 276, + 507, + 290 + ], + "score": 1.0, + "content": "Can 5M synthetic images beat generic imagenet pre-training on indoor segmentation? In ICCV,", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 127, + 287, + 155, + 303 + ], + "spans": [ + { + "bbox": [ + 127, + 287, + 155, + 303 + ], + "score": 1.0, + "content": "2017.", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 308, + 506, + 320 + ], + "spans": [ + { + "bbox": [ + 106, + 308, + 506, + 320 + ], + "score": 1.0, + "content": "[42] Chenlin Meng, Ruiqi Gao, Diederik P. Kingma, Stefano Ermon, Jonathan Ho, and Tim Salimans.", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 126, + 317, + 507, + 332 + ], + "spans": [ + { + "bbox": [ + 126, + 317, + 507, + 332 + ], + "score": 1.0, + "content": "On distillation of guided diffusion models. In NeurIPS 2022 Workshop on Score-Based Methods,", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 126, + 328, + 155, + 343 + ], + "spans": [ + { + "bbox": [ + 126, + 328, + 155, + 343 + ], + "score": 1.0, + "content": "2022.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 348, + 506, + 362 + ], + "spans": [ + { + "bbox": [ + 105, + 348, + 506, + 362 + ], + "score": 1.0, + "content": "[43] Moritz Menze, Christian Heipke, and Andreas Geiger. Joint 3D estimation of vehicles and", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 126, + 359, + 424, + 373 + ], + "spans": [ + { + "bbox": [ + 126, + 359, + 424, + 373 + ], + "score": 1.0, + "content": "scene flow. In ISPRS Workshop on Image Sequence Analysis (ISA), 2015.", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 378, + 507, + 392 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 507, + 392 + ], + "score": 1.0, + "content": "[44] Moritz Menze, Christian Heipke, and Andreas Geiger. Object scene flow. ISPRS Journal of", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 127, + 389, + 339, + 402 + ], + "spans": [ + { + "bbox": [ + 127, + 389, + 339, + 402 + ], + "score": 1.0, + "content": "Photogrammetry and Remote Sensing (JPRS), 2018.", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 408, + 506, + 422 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 506, + 422 + ], + "score": 1.0, + "content": "[45] Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 126, + 419, + 304, + 433 + ], + "spans": [ + { + "bbox": [ + 126, + 419, + 304, + 433 + ], + "score": 1.0, + "content": "models. In ICML, pages 8162–8171, 2021.", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 438, + 507, + 452 + ], + "spans": [ + { + "bbox": [ + 105, + 438, + 507, + 452 + ], + "score": 1.0, + "content": "[46] Jia Ning, Chen Li, Zheng Zhang, Zigang Geng, Qi Dai, Kun He, and Han Hu. All in Tokens:", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 128, + 450, + 402, + 462 + ], + "spans": [ + { + "bbox": [ + 128, + 450, + 402, + 462 + ], + "score": 1.0, + "content": "Unifying output space of visual tasks via soft token. In ICCV, 2023.", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 104, + 466, + 507, + 483 + ], + "spans": [ + { + "bbox": [ + 104, + 466, + 507, + 483 + ], + "score": 1.0, + "content": "[47] Ethan Perez, Florian Strub, Harm de Vries, Vincent Dumoulin, and Aaron C. Courville. FiLm:", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 126, + 478, + 398, + 493 + ], + "spans": [ + { + "bbox": [ + 126, + 478, + 398, + 493 + ], + "score": 1.0, + "content": "Visual reasoning with a general conditioning layer. In AAAI, 2018.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 498, + 505, + 511 + ], + "spans": [ + { + "bbox": [ + 106, + 498, + 505, + 511 + ], + "score": 1.0, + "content": "[48] Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 127, + 510, + 436, + 522 + ], + "spans": [ + { + "bbox": [ + 127, + 510, + 436, + 522 + ], + "score": 1.0, + "content": "text-conditional image generation with clip latents. arXiv:2204.06125, 2022.", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 528, + 506, + 542 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 506, + 542 + ], + "score": 1.0, + "content": "[49] René Ranftl, Katrin Lasinger, David Hafner, Konrad Schindler, and Vladlen Koltun. Towards", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 127, + 540, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 127, + 540, + 505, + 552 + ], + "score": 1.0, + "content": "robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer. IEEE", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 127, + 550, + 262, + 563 + ], + "spans": [ + { + "bbox": [ + 127, + 550, + 262, + 563 + ], + "score": 1.0, + "content": "T-PAMI, 44(3):1623–1637, 2020.", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 569, + 506, + 581 + ], + "spans": [ + { + "bbox": [ + 106, + 569, + 506, + 581 + ], + "score": 1.0, + "content": "[50] René Ranftl, Alexey Bochkovskiy, and Vladlen Koltun. Vision transformers for dense prediction.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 127, + 581, + 277, + 592 + ], + "spans": [ + { + "bbox": [ + 127, + 581, + 277, + 592 + ], + "score": 1.0, + "content": "In ICCV, pages 12179–12188, 2021.", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 599, + 505, + 612 + ], + "spans": [ + { + "bbox": [ + 106, + 599, + 505, + 612 + ], + "score": 1.0, + "content": "[51] Marc’Aurelio Ranzato, Sumit Chopra, Michael Auli, and Wojciech Zaremba. Sequence level", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 128, + 611, + 352, + 623 + ], + "spans": [ + { + "bbox": [ + 128, + 611, + 352, + 623 + ], + "score": 1.0, + "content": "training with recurrent neural networks. In ICLR, 2016.", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 103, + 627, + 507, + 645 + ], + "spans": [ + { + "bbox": [ + 103, + 627, + 507, + 645 + ], + "score": 1.0, + "content": "[52] Zhongzheng Ren and Yong Jae Lee. Cross-Domain self-supervised multi-task feature learning", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 126, + 640, + 294, + 653 + ], + "spans": [ + { + "bbox": [ + 126, + 640, + 294, + 653 + ], + "score": 1.0, + "content": "using synthetic imagery. In CVPR, 2018.", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 658, + 507, + 672 + ], + "spans": [ + { + "bbox": [ + 105, + 658, + 507, + 672 + ], + "score": 1.0, + "content": "[53] Stephan R. Richter, Zeeshan Hayder, and Vladlen Koltun. Playing for benchmarks. In ICCV,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 125, + 669, + 229, + 682 + ], + "spans": [ + { + "bbox": [ + 125, + 669, + 229, + 682 + ], + "score": 1.0, + "content": "pages 2213–2222, 2017.", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 688, + 507, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 507, + 702 + ], + "score": 1.0, + "content": "[54] Chitwan Saharia, William Chan, Huiwen Chang, Chris A. Lee, Jonathan Ho, Tim Salimans,", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 126, + 699, + 507, + 714 + ], + "spans": [ + { + "bbox": [ + 126, + 699, + 507, + 714 + ], + "score": 1.0, + "content": "David J. Fleet, and Mohammad Norouzi. Palette: Image-to-Image Diffusion Models. In", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 126, + 710, + 206, + 724 + ], + "spans": [ + { + "bbox": [ + 126, + 710, + 206, + 724 + ], + "score": 1.0, + "content": "SIGGRAPH, 2022.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 22.5 + } + ], + "page_idx": 12, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "13", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "list", + "bbox": [ + 105, + 38, + 508, + 724 + ], + "lines": [ + { + "bbox": [ + 104, + 70, + 507, + 88 + ], + "spans": [ + { + "bbox": [ + 104, + 70, + 507, + 88 + ], + "score": 1.0, + "content": "[36] Orly Liba, Longqi Cai, Yun-Ta Tsai, Elad Eban, Yair Movshovitz-Attias, Yael Pritch, Huizhong", + "type": "text" + } + ], + "index": 0, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 83, + 506, + 98 + ], + "spans": [ + { + "bbox": [ + 127, + 83, + 506, + 98 + ], + "score": 1.0, + "content": "Chen, and Jonathan T. Barron. Sky Optimization: Semantically aware image processing of", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 126, + 94, + 389, + 108 + ], + "spans": [ + { + "bbox": [ + 126, + 94, + 389, + 108 + ], + "score": 1.0, + "content": "skies in low-light photography. In CVPR Workshops, June 2020.", + "type": "text" + } + ], + "index": 2, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 113, + 506, + 127 + ], + "spans": [ + { + "bbox": [ + 105, + 113, + 506, + 127 + ], + "score": 1.0, + "content": "[37] Andrew Liu, Richard Tucker, Varun Jampani, Ameesh Makadia, Noah Snavely, and Angjoo", + "type": "text" + } + ], + "index": 3, + "is_list_start_line": true + }, + { + "bbox": [ + 128, + 126, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 128, + 126, + 505, + 138 + ], + "score": 1.0, + "content": "Kanazawa. Infinite Nature: Perpetual view generation of natural scenes from a single image. In", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 127, + 136, + 181, + 148 + ], + "spans": [ + { + "bbox": [ + 127, + 136, + 181, + 148 + ], + "score": 1.0, + "content": "ICCV, 2021.", + "type": "text" + } + ], + "index": 5, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 154, + 506, + 168 + ], + "spans": [ + { + "bbox": [ + 105, + 154, + 506, + 168 + ], + "score": 1.0, + "content": "[38] Ao Luo, Fan Yang, Kunming Luo, Xin Li, Haoqiang Fan, and Shuaicheng Liu. Learning optical", + "type": "text" + } + ], + "index": 6, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 164, + 412, + 180 + ], + "spans": [ + { + "bbox": [ + 126, + 164, + 412, + 180 + ], + "score": 1.0, + "content": "flow with adaptive graph reasoning. In AAAI, pages 1890–1898, 2022.", + "type": "text" + } + ], + "index": 7, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 184, + 507, + 199 + ], + "spans": [ + { + "bbox": [ + 106, + 184, + 507, + 199 + ], + "score": 1.0, + "content": "[39] Zhaoyang Lv, Kihwan Kim, Alejandro Troccoli, Deqing Sun, James M. Rehg, and Jan Kautz.", + "type": "text" + } + ], + "index": 8, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 196, + 507, + 210 + ], + "spans": [ + { + "bbox": [ + 126, + 196, + 507, + 210 + ], + "score": 1.0, + "content": "Learning rigidity in dynamic scenes with a moving camera for 3D motion field estimation. In", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 126, + 206, + 249, + 221 + ], + "spans": [ + { + "bbox": [ + 126, + 206, + 249, + 221 + ], + "score": 1.0, + "content": "ECCV, pages 468–484, 2018.", + "type": "text" + } + ], + "index": 10, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 225, + 507, + 239 + ], + "spans": [ + { + "bbox": [ + 106, + 225, + 507, + 239 + ], + "score": 1.0, + "content": "[40] Nikolaus Mayer, Eddy Ilg, Philip Hausser, Philipp Fischer, Daniel Cremers, Alexey Dosovitskiy,", + "type": "text" + } + ], + "index": 11, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 235, + 507, + 250 + ], + "spans": [ + { + "bbox": [ + 126, + 235, + 507, + 250 + ], + "score": 1.0, + "content": "and Thomas Brox. A large dataset to train convolutional networks for disparity, optical flow,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 127, + 248, + 303, + 259 + ], + "spans": [ + { + "bbox": [ + 127, + 248, + 303, + 259 + ], + "score": 1.0, + "content": "and scene flow estimation. In CVPR, 2016.", + "type": "text" + } + ], + "index": 13, + "is_list_end_line": true + }, + { + "bbox": [ + 103, + 265, + 507, + 281 + ], + "spans": [ + { + "bbox": [ + 103, + 265, + 507, + 281 + ], + "score": 1.0, + "content": "[41] John McCormac, Ankur Handa, Stefan Leutenegger, and Andrew J. Davison. SceneNet RGB-D:", + "type": "text" + } + ], + "index": 14, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 276, + 507, + 290 + ], + "spans": [ + { + "bbox": [ + 126, + 276, + 507, + 290 + ], + "score": 1.0, + "content": "Can 5M synthetic images beat generic imagenet pre-training on indoor segmentation? In ICCV,", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 127, + 287, + 155, + 303 + ], + "spans": [ + { + "bbox": [ + 127, + 287, + 155, + 303 + ], + "score": 1.0, + "content": "2017.", + "type": "text" + } + ], + "index": 16, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 308, + 506, + 320 + ], + "spans": [ + { + "bbox": [ + 106, + 308, + 506, + 320 + ], + "score": 1.0, + "content": "[42] Chenlin Meng, Ruiqi Gao, Diederik P. Kingma, Stefano Ermon, Jonathan Ho, and Tim Salimans.", + "type": "text" + } + ], + "index": 17, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 317, + 507, + 332 + ], + "spans": [ + { + "bbox": [ + 126, + 317, + 507, + 332 + ], + "score": 1.0, + "content": "On distillation of guided diffusion models. In NeurIPS 2022 Workshop on Score-Based Methods,", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 126, + 328, + 155, + 343 + ], + "spans": [ + { + "bbox": [ + 126, + 328, + 155, + 343 + ], + "score": 1.0, + "content": "2022.", + "type": "text" + } + ], + "index": 19, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 348, + 506, + 362 + ], + "spans": [ + { + "bbox": [ + 105, + 348, + 506, + 362 + ], + "score": 1.0, + "content": "[43] Moritz Menze, Christian Heipke, and Andreas Geiger. Joint 3D estimation of vehicles and", + "type": "text" + } + ], + "index": 20, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 359, + 424, + 373 + ], + "spans": [ + { + "bbox": [ + 126, + 359, + 424, + 373 + ], + "score": 1.0, + "content": "scene flow. In ISPRS Workshop on Image Sequence Analysis (ISA), 2015.", + "type": "text" + } + ], + "index": 21, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 378, + 507, + 392 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 507, + 392 + ], + "score": 1.0, + "content": "[44] Moritz Menze, Christian Heipke, and Andreas Geiger. Object scene flow. ISPRS Journal of", + "type": "text" + } + ], + "index": 22, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 389, + 339, + 402 + ], + "spans": [ + { + "bbox": [ + 127, + 389, + 339, + 402 + ], + "score": 1.0, + "content": "Photogrammetry and Remote Sensing (JPRS), 2018.", + "type": "text" + } + ], + "index": 23, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 408, + 506, + 422 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 506, + 422 + ], + "score": 1.0, + "content": "[45] Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic", + "type": "text" + } + ], + "index": 24, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 419, + 304, + 433 + ], + "spans": [ + { + "bbox": [ + 126, + 419, + 304, + 433 + ], + "score": 1.0, + "content": "models. In ICML, pages 8162–8171, 2021.", + "type": "text" + } + ], + "index": 25, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 438, + 507, + 452 + ], + "spans": [ + { + "bbox": [ + 105, + 438, + 507, + 452 + ], + "score": 1.0, + "content": "[46] Jia Ning, Chen Li, Zheng Zhang, Zigang Geng, Qi Dai, Kun He, and Han Hu. All in Tokens:", + "type": "text" + } + ], + "index": 26, + "is_list_start_line": true + }, + { + "bbox": [ + 128, + 450, + 402, + 462 + ], + "spans": [ + { + "bbox": [ + 128, + 450, + 402, + 462 + ], + "score": 1.0, + "content": "Unifying output space of visual tasks via soft token. In ICCV, 2023.", + "type": "text" + } + ], + "index": 27, + "is_list_end_line": true + }, + { + "bbox": [ + 104, + 466, + 507, + 483 + ], + "spans": [ + { + "bbox": [ + 104, + 466, + 507, + 483 + ], + "score": 1.0, + "content": "[47] Ethan Perez, Florian Strub, Harm de Vries, Vincent Dumoulin, and Aaron C. Courville. FiLm:", + "type": "text" + } + ], + "index": 28, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 478, + 398, + 493 + ], + "spans": [ + { + "bbox": [ + 126, + 478, + 398, + 493 + ], + "score": 1.0, + "content": "Visual reasoning with a general conditioning layer. In AAAI, 2018.", + "type": "text" + } + ], + "index": 29, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 498, + 505, + 511 + ], + "spans": [ + { + "bbox": [ + 106, + 498, + 505, + 511 + ], + "score": 1.0, + "content": "[48] Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical", + "type": "text" + } + ], + "index": 30, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 510, + 436, + 522 + ], + "spans": [ + { + "bbox": [ + 127, + 510, + 436, + 522 + ], + "score": 1.0, + "content": "text-conditional image generation with clip latents. arXiv:2204.06125, 2022.", + "type": "text" + } + ], + "index": 31, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 528, + 506, + 542 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 506, + 542 + ], + "score": 1.0, + "content": "[49] René Ranftl, Katrin Lasinger, David Hafner, Konrad Schindler, and Vladlen Koltun. Towards", + "type": "text" + } + ], + "index": 32, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 540, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 127, + 540, + 505, + 552 + ], + "score": 1.0, + "content": "robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer. IEEE", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 127, + 550, + 262, + 563 + ], + "spans": [ + { + "bbox": [ + 127, + 550, + 262, + 563 + ], + "score": 1.0, + "content": "T-PAMI, 44(3):1623–1637, 2020.", + "type": "text" + } + ], + "index": 34, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 569, + 506, + 581 + ], + "spans": [ + { + "bbox": [ + 106, + 569, + 506, + 581 + ], + "score": 1.0, + "content": "[50] René Ranftl, Alexey Bochkovskiy, and Vladlen Koltun. Vision transformers for dense prediction.", + "type": "text" + } + ], + "index": 35, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 581, + 277, + 592 + ], + "spans": [ + { + "bbox": [ + 127, + 581, + 277, + 592 + ], + "score": 1.0, + "content": "In ICCV, pages 12179–12188, 2021.", + "type": "text" + } + ], + "index": 36, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 599, + 505, + 612 + ], + "spans": [ + { + "bbox": [ + 106, + 599, + 505, + 612 + ], + "score": 1.0, + "content": "[51] Marc’Aurelio Ranzato, Sumit Chopra, Michael Auli, and Wojciech Zaremba. Sequence level", + "type": "text" + } + ], + "index": 37, + "is_list_start_line": true + }, + { + "bbox": [ + 128, + 611, + 352, + 623 + ], + "spans": [ + { + "bbox": [ + 128, + 611, + 352, + 623 + ], + "score": 1.0, + "content": "training with recurrent neural networks. In ICLR, 2016.", + "type": "text" + } + ], + "index": 38, + "is_list_end_line": true + }, + { + "bbox": [ + 103, + 627, + 507, + 645 + ], + "spans": [ + { + "bbox": [ + 103, + 627, + 507, + 645 + ], + "score": 1.0, + "content": "[52] Zhongzheng Ren and Yong Jae Lee. Cross-Domain self-supervised multi-task feature learning", + "type": "text" + } + ], + "index": 39, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 640, + 294, + 653 + ], + "spans": [ + { + "bbox": [ + 126, + 640, + 294, + 653 + ], + "score": 1.0, + "content": "using synthetic imagery. In CVPR, 2018.", + "type": "text" + } + ], + "index": 40, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 658, + 507, + 672 + ], + "spans": [ + { + "bbox": [ + 105, + 658, + 507, + 672 + ], + "score": 1.0, + "content": "[53] Stephan R. Richter, Zeeshan Hayder, and Vladlen Koltun. Playing for benchmarks. In ICCV,", + "type": "text" + } + ], + "index": 41, + "is_list_start_line": true + }, + { + "bbox": [ + 125, + 669, + 229, + 682 + ], + "spans": [ + { + "bbox": [ + 125, + 669, + 229, + 682 + ], + "score": 1.0, + "content": "pages 2213–2222, 2017.", + "type": "text" + } + ], + "index": 42, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 688, + 507, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 507, + 702 + ], + "score": 1.0, + "content": "[54] Chitwan Saharia, William Chan, Huiwen Chang, Chris A. Lee, Jonathan Ho, Tim Salimans,", + "type": "text" + } + ], + "index": 43, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 699, + 507, + 714 + ], + "spans": [ + { + "bbox": [ + 126, + 699, + 507, + 714 + ], + "score": 1.0, + "content": "David J. Fleet, and Mohammad Norouzi. Palette: Image-to-Image Diffusion Models. In", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 126, + 710, + 206, + 724 + ], + "spans": [ + { + "bbox": [ + 126, + 710, + 206, + 724 + ], + "score": 1.0, + "content": "SIGGRAPH, 2022.", + "type": "text" + } + ], + "index": 45, + "is_list_end_line": true + }, + { + "bbox": [ + 104, + 72, + 506, + 85 + ], + "spans": [ + { + "bbox": [ + 104, + 72, + 506, + 85 + ], + "score": 1.0, + "content": "[55] Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily Denton, Seyed", + "type": "text", + "cross_page": true + } + ], + "index": 0, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 83, + 506, + 97 + ], + "spans": [ + { + "bbox": [ + 127, + 83, + 506, + 97 + ], + "score": 1.0, + "content": "Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, S. Sara Mahdavi, Rapha Gontijo Lopes, Tim", + "type": "text", + "cross_page": true + } + ], + "index": 1 + }, + { + "bbox": [ + 125, + 91, + 507, + 110 + ], + "spans": [ + { + "bbox": [ + 125, + 91, + 507, + 110 + ], + "score": 1.0, + "content": "Salimans, Jonathan Ho, David J. Fleet, and Mohammad Norouzi. Photorealistic Text-to-Image", + "type": "text", + "cross_page": true + } + ], + "index": 2 + }, + { + "bbox": [ + 128, + 106, + 425, + 118 + ], + "spans": [ + { + "bbox": [ + 128, + 106, + 425, + 118 + ], + "score": 1.0, + "content": "Diffusion Models with Deep Language Understanding. In NeurIPS, 2022.", + "type": "text", + "cross_page": true + } + ], + "index": 3, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 125, + 507, + 139 + ], + "spans": [ + { + "bbox": [ + 105, + 125, + 507, + 139 + ], + "score": 1.0, + "content": "[56] Tim Salimans and Jonathan Ho. Progressive distillation for fast sampling of diffusion models.", + "type": "text", + "cross_page": true + } + ], + "index": 4, + "is_list_start_line": true + }, + { + "bbox": [ + 128, + 136, + 191, + 149 + ], + "spans": [ + { + "bbox": [ + 128, + 136, + 191, + 149 + ], + "score": 1.0, + "content": "In ICLR, 2022.", + "type": "text", + "cross_page": true + } + ], + "index": 5, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 155, + 507, + 170 + ], + "spans": [ + { + "bbox": [ + 105, + 155, + 507, + 170 + ], + "score": 1.0, + "content": "[57] Nikolay Savinov, Junyoung Chung, Mikolaj Binkowski, Erich Elsen, and Aaron van den Oord.", + "type": "text", + "cross_page": true + } + ], + "index": 6, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 167, + 422, + 180 + ], + "spans": [ + { + "bbox": [ + 127, + 167, + 422, + 180 + ], + "score": 1.0, + "content": "Step-unrolled denoising autoencoders for text generation. In ICLR, 2022.", + "type": "text", + "cross_page": true + } + ], + "index": 7, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 186, + 507, + 200 + ], + "spans": [ + { + "bbox": [ + 105, + 186, + 507, + 200 + ], + "score": 1.0, + "content": "[58] Ashutosh Saxena, Sung Chung, and Andrew Ng. Learning depth from single monocular images.", + "type": "text", + "cross_page": true + } + ], + "index": 8, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 196, + 181, + 211 + ], + "spans": [ + { + "bbox": [ + 126, + 196, + 181, + 211 + ], + "score": 1.0, + "content": "NIPS, 2005.", + "type": "text", + "cross_page": true + } + ], + "index": 9, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 216, + 507, + 230 + ], + "spans": [ + { + "bbox": [ + 105, + 216, + 507, + 230 + ], + "score": 1.0, + "content": "[59] Ashutosh Saxena, Min Sun, and Andrew Y. Ng. Make3D: Learning 3D scene structure from a", + "type": "text", + "cross_page": true + } + ], + "index": 10, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 227, + 352, + 241 + ], + "spans": [ + { + "bbox": [ + 126, + 227, + 352, + 241 + ], + "score": 1.0, + "content": "single still image. IEEE T-PAMI, 31(5):824–840, 2009.", + "type": "text", + "cross_page": true + } + ], + "index": 11, + "is_list_end_line": true + }, + { + "bbox": [ + 103, + 245, + 507, + 262 + ], + "spans": [ + { + "bbox": [ + 103, + 245, + 507, + 262 + ], + "score": 1.0, + "content": "[60] Meng-Li Shih, Shih-Yang Su, Johannes Kopf, and Jia-Bin Huang. 3D photography using", + "type": "text", + "cross_page": true + } + ], + "index": 12, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 258, + 357, + 272 + ], + "spans": [ + { + "bbox": [ + 126, + 258, + 357, + 272 + ], + "score": 1.0, + "content": "context-aware layered depth inpainting. In CVPR, 2020.", + "type": "text", + "cross_page": true + } + ], + "index": 13, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "score": 1.0, + "content": "[61] Nathan Silberman, Derek Hoiem, Pushmeet Kohli, and Rob Fergus. Indoor segmentation and", + "type": "text", + "cross_page": true + } + ], + "index": 14, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 288, + 418, + 302 + ], + "spans": [ + { + "bbox": [ + 126, + 288, + 418, + 302 + ], + "score": 1.0, + "content": "support inference from RGBD images. In ECCV, pages 746–760, 2012.", + "type": "text", + "cross_page": true + } + ], + "index": 15, + "is_list_end_line": true + }, + { + "bbox": [ + 104, + 307, + 507, + 322 + ], + "spans": [ + { + "bbox": [ + 104, + 307, + 507, + 322 + ], + "score": 1.0, + "content": "[62] Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsuper-", + "type": "text", + "cross_page": true + } + ], + "index": 16, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 318, + 487, + 332 + ], + "spans": [ + { + "bbox": [ + 126, + 318, + 487, + 332 + ], + "score": 1.0, + "content": "vised learning using nonequilibrium thermodynamics. In ICML, pages 2256–2265, 2015.", + "type": "text", + "cross_page": true + } + ], + "index": 17, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 339, + 505, + 351 + ], + "spans": [ + { + "bbox": [ + 106, + 339, + 505, + 351 + ], + "score": 1.0, + "content": "[63] Yang Song, Jascha Sohl-Dickstein, Diederik P. Kingma, Abhishek Kumar, Stefano Ermon, and", + "type": "text", + "cross_page": true + } + ], + "index": 18, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 348, + 507, + 363 + ], + "spans": [ + { + "bbox": [ + 126, + 348, + 507, + 363 + ], + "score": 1.0, + "content": "Ben Poole. Score-based generative modeling through stochastic differential equations. In ICLR,", + "type": "text", + "cross_page": true + } + ], + "index": 19 + }, + { + "bbox": [ + 126, + 359, + 155, + 374 + ], + "spans": [ + { + "bbox": [ + 126, + 359, + 155, + 374 + ], + "score": 1.0, + "content": "2021.", + "type": "text", + "cross_page": true + } + ], + "index": 20, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 379, + 507, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 379, + 507, + 394 + ], + "score": 1.0, + "content": "[64] Yang Song, Prafulla Dhariwal, Mark Chen, and Ilya Sutskever. Consistency models. In ICML,", + "type": "text", + "cross_page": true + } + ], + "index": 21, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 390, + 155, + 405 + ], + "spans": [ + { + "bbox": [ + 126, + 390, + 155, + 405 + ], + "score": 1.0, + "content": "2023.", + "type": "text", + "cross_page": true + } + ], + "index": 22, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 410, + 506, + 424 + ], + "spans": [ + { + "bbox": [ + 106, + 410, + 506, + 424 + ], + "score": 1.0, + "content": "[65] Martin Stommel, Michael Beetz, and Weiliang Xu. Inpainting of missing values in the kinect", + "type": "text", + "cross_page": true + } + ], + "index": 23, + "is_list_start_line": true + }, + { + "bbox": [ + 128, + 421, + 505, + 433 + ], + "spans": [ + { + "bbox": [ + 128, + 421, + 505, + 433 + ], + "score": 1.0, + "content": "sensor’s depth maps based on background estimates. IEEE Sensors Journal, 14(4):1107–1116,", + "type": "text", + "cross_page": true + } + ], + "index": 24 + }, + { + "bbox": [ + 127, + 430, + 155, + 446 + ], + "spans": [ + { + "bbox": [ + 127, + 430, + 155, + 446 + ], + "score": 1.0, + "content": "2014.", + "type": "text", + "cross_page": true + } + ], + "index": 25, + "is_list_end_line": true + }, + { + "bbox": [ + 104, + 451, + 506, + 466 + ], + "spans": [ + { + "bbox": [ + 104, + 451, + 506, + 466 + ], + "score": 1.0, + "content": "[66] Xiuchao Sui, Shaohua Li, Xue Geng, Yan Wu, Xinxing Xu, Yong Liu, Rick Goh, and Hongyuan", + "type": "text", + "cross_page": true + } + ], + "index": 26, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 460, + 507, + 478 + ], + "spans": [ + { + "bbox": [ + 126, + 460, + 507, + 478 + ], + "score": 1.0, + "content": "Zhu. CRAFT: Cross-attentional flow transformer for robust optical flow. In CVPR, pages", + "type": "text", + "cross_page": true + } + ], + "index": 27 + }, + { + "bbox": [ + 126, + 473, + 211, + 486 + ], + "spans": [ + { + "bbox": [ + 126, + 473, + 211, + 486 + ], + "score": 1.0, + "content": "17602–17611, 2022.", + "type": "text", + "cross_page": true + } + ], + "index": 28, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 492, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 506, + 507 + ], + "score": 1.0, + "content": "[67] Deqing Sun, Erik B. Sudderth, and Michael J. Black. Layered segmentation and optical flow", + "type": "text", + "cross_page": true + } + ], + "index": 29, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 504, + 358, + 518 + ], + "spans": [ + { + "bbox": [ + 126, + 504, + 358, + 518 + ], + "score": 1.0, + "content": "estimation over time. In CVPR, pages 1768–1775, 2012.", + "type": "text", + "cross_page": true + } + ], + "index": 30, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 523, + 506, + 537 + ], + "spans": [ + { + "bbox": [ + 106, + 523, + 506, + 537 + ], + "score": 1.0, + "content": "[68] Deqing Sun, Xiaodong Yang, Ming-Yu Liu, and Jan Kautz. PWC-Net: CNNs for optical flow", + "type": "text", + "cross_page": true + } + ], + "index": 31, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 534, + 440, + 549 + ], + "spans": [ + { + "bbox": [ + 127, + 534, + 440, + 549 + ], + "score": 1.0, + "content": "using pyramid, warping, and cost volume. In CVPR, pages 8934–8943, 2018.", + "type": "text", + "cross_page": true + } + ], + "index": 32, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 553, + 506, + 568 + ], + "spans": [ + { + "bbox": [ + 106, + 553, + 506, + 568 + ], + "score": 1.0, + "content": "[69] Deqing Sun, Daniel Vlasic, Charles Herrmann, Varun Jampani, Michael Krainin, Huiwen", + "type": "text", + "cross_page": true + } + ], + "index": 33, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 563, + 507, + 581 + ], + "spans": [ + { + "bbox": [ + 126, + 563, + 507, + 581 + ], + "score": 1.0, + "content": "Chang, Ramin Zabih, William T. Freeman, and Ce Liu. AutoFlow: Learning a better training", + "type": "text", + "cross_page": true + } + ], + "index": 34 + }, + { + "bbox": [ + 126, + 575, + 360, + 590 + ], + "spans": [ + { + "bbox": [ + 126, + 575, + 360, + 590 + ], + "score": 1.0, + "content": "set for optical flow. In CVPR, pages 10093–10102, 2021.", + "type": "text", + "cross_page": true + } + ], + "index": 35, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 595, + 507, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 595, + 507, + 609 + ], + "score": 1.0, + "content": "[70] Deqing Sun, Charles Herrmann, Fitsum Reda, Michael Rubinstein, David J. Fleet, and William T.", + "type": "text", + "cross_page": true + } + ], + "index": 36, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 605, + 507, + 620 + ], + "spans": [ + { + "bbox": [ + 126, + 605, + 507, + 620 + ], + "score": 1.0, + "content": "Freeman. Disentangling architecture and training for optical flow. In ECCV, pages 165–182,", + "type": "text", + "cross_page": true + } + ], + "index": 37 + }, + { + "bbox": [ + 127, + 617, + 154, + 631 + ], + "spans": [ + { + "bbox": [ + 127, + 617, + 154, + 631 + ], + "score": 1.0, + "content": "2022.", + "type": "text", + "cross_page": true + } + ], + "index": 38, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 636, + 507, + 650 + ], + "spans": [ + { + "bbox": [ + 105, + 636, + 507, + 650 + ], + "score": 1.0, + "content": "[71] Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aurelien Chouard, Vijaysai Patnaik, Paul Tsui,", + "type": "text", + "cross_page": true + } + ], + "index": 39, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 646, + 506, + 662 + ], + "spans": [ + { + "bbox": [ + 126, + 646, + 506, + 662 + ], + "score": 1.0, + "content": "James Guo, Yin Zhou, Yuning Chai, Benjamin Caine, Vijay Vasudevan, Wei Han, Jiquan Ngiam,", + "type": "text", + "cross_page": true + } + ], + "index": 40 + }, + { + "bbox": [ + 126, + 657, + 507, + 674 + ], + "spans": [ + { + "bbox": [ + 126, + 657, + 507, + 674 + ], + "score": 1.0, + "content": "Hang Zhao, Aleksei Timofeev, Scott Ettinger, Maxim Krivokon, Amy Gao, Aditya Joshi, Sheng", + "type": "text", + "cross_page": true + } + ], + "index": 41 + }, + { + "bbox": [ + 128, + 670, + 506, + 682 + ], + "spans": [ + { + "bbox": [ + 128, + 670, + 506, + 682 + ], + "score": 1.0, + "content": "Zhao, Shuyang Cheng, Yu Zhang, Jonathon Shlens, Zhifeng Chen, and Dragomir Anguelov.", + "type": "text", + "cross_page": true + } + ], + "index": 42 + }, + { + "bbox": [ + 126, + 680, + 483, + 694 + ], + "spans": [ + { + "bbox": [ + 126, + 680, + 483, + 694 + ], + "score": 1.0, + "content": "Scalability in perception for autonomous driving: Waymo open dataset. In CVPR, 2020.", + "type": "text", + "cross_page": true + } + ], + "index": 43, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 700, + 505, + 714 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 505, + 714 + ], + "score": 1.0, + "content": "[72] Shangkun Sun, Yuanqi Chen, Yu Zhu, Guodong Guo, and Ge Li. SKFlow: Learning optical", + "type": "text", + "cross_page": true + } + ], + "index": 44, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 711, + 302, + 725 + ], + "spans": [ + { + "bbox": [ + 127, + 711, + 302, + 725 + ], + "score": 1.0, + "content": "flow with super kernels. In NeurIPS, 2022.", + "type": "text", + "cross_page": true + } + ], + "index": 45, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 72, + 505, + 87 + ], + "spans": [ + { + "bbox": [ + 105, + 72, + 505, + 87 + ], + "score": 1.0, + "content": "[73] Mingxing Tan and Quoc Le. EfficientNet: Rethinking model scaling for convolutional neural", + "type": "text", + "cross_page": true + } + ], + "index": 0, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 84, + 311, + 97 + ], + "spans": [ + { + "bbox": [ + 127, + 84, + 311, + 97 + ], + "score": 1.0, + "content": "networks. In ICML, pages 6105–6114, 2019.", + "type": "text", + "cross_page": true + } + ], + "index": 1, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 102, + 506, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 102, + 506, + 117 + ], + "score": 1.0, + "content": "[74] Zachary Teed and Jia Deng. RAFT: Recurrent all-pairs field transforms for optical flow. In", + "type": "text", + "cross_page": true + } + ], + "index": 2, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 113, + 249, + 128 + ], + "spans": [ + { + "bbox": [ + 127, + 113, + 249, + 128 + ], + "score": 1.0, + "content": "ECCV, pages 402–419, 2020.", + "type": "text", + "cross_page": true + } + ], + "index": 3, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 132, + 506, + 147 + ], + "spans": [ + { + "bbox": [ + 105, + 132, + 506, + 147 + ], + "score": 1.0, + "content": "[75] Su Wang, Chitwan Saharia, Ceslee Montgomery, Jordi Pont-Tuset, Shai Noy, Stefano Pellegrini,", + "type": "text", + "cross_page": true + } + ], + "index": 4, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 143, + 506, + 157 + ], + "spans": [ + { + "bbox": [ + 127, + 143, + 506, + 157 + ], + "score": 1.0, + "content": "Yasumasa Onoe, Sarah Laszlo, David J. Fleet, Radu Soricut, Jason Baldridge, Mohammad", + "type": "text", + "cross_page": true + } + ], + "index": 5 + }, + { + "bbox": [ + 127, + 154, + 506, + 168 + ], + "spans": [ + { + "bbox": [ + 127, + 154, + 506, + 168 + ], + "score": 1.0, + "content": "Norouzi, Peter Anderson, and William Chan. Imagen Editor and EditBench: Advancing and", + "type": "text", + "cross_page": true + } + ], + "index": 6 + }, + { + "bbox": [ + 126, + 165, + 358, + 179 + ], + "spans": [ + { + "bbox": [ + 126, + 165, + 358, + 179 + ], + "score": 1.0, + "content": "evaluating text-guided image inpainting. In CVPR, 2023.", + "type": "text", + "cross_page": true + } + ], + "index": 7, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 184, + 506, + 198 + ], + "spans": [ + { + "bbox": [ + 105, + 184, + 506, + 198 + ], + "score": 1.0, + "content": "[76] Wenshan Wang, Delong Zhu, Xiangwei Wang, Yaoyu Hu, Yuheng Qiu, Chen Wang, Yafei Hu,", + "type": "text", + "cross_page": true + } + ], + "index": 8, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 195, + 507, + 209 + ], + "spans": [ + { + "bbox": [ + 127, + 195, + 507, + 209 + ], + "score": 1.0, + "content": "Ashish Kapoor, and Sebastian Scherer. TartanAir: A dataset to push the limits of visual SLAM.", + "type": "text", + "cross_page": true + } + ], + "index": 9 + }, + { + "bbox": [ + 127, + 206, + 190, + 219 + ], + "spans": [ + { + "bbox": [ + 127, + 206, + 190, + 219 + ], + "score": 1.0, + "content": "In IROS, 2020.", + "type": "text", + "cross_page": true + } + ], + "index": 10, + "is_list_end_line": true + }, + { + "bbox": [ + 104, + 223, + 506, + 240 + ], + "spans": [ + { + "bbox": [ + 104, + 223, + 506, + 240 + ], + "score": 1.0, + "content": "[77] Philippe Weinzaepfel, Jerome Revaud, Zaid Harchaoui, and Cordelia Schmid. DeepFlow: Large", + "type": "text", + "cross_page": true + } + ], + "index": 11, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 235, + 456, + 249 + ], + "spans": [ + { + "bbox": [ + 126, + 235, + 456, + 249 + ], + "score": 1.0, + "content": "displacement optical flow with deep matching. In ICCV, pages 1385–1392, 2013.", + "type": "text", + "cross_page": true + } + ], + "index": 12, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 254, + 505, + 269 + ], + "spans": [ + { + "bbox": [ + 105, + 254, + 505, + 269 + ], + "score": 1.0, + "content": "[78] Olivia Wiles, Georgia Gkioxari, Richard Szeliski, and Justin Johnson. SynSin: End-to-end view", + "type": "text", + "cross_page": true + } + ], + "index": 13, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 264, + 319, + 280 + ], + "spans": [ + { + "bbox": [ + 126, + 264, + 319, + 280 + ], + "score": 1.0, + "content": "synthesis from a single image. In CVPR, 2020.", + "type": "text", + "cross_page": true + } + ], + "index": 14, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 284, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 284, + 505, + 298 + ], + "score": 1.0, + "content": "[79] Ronald J. Williams and David Zipser. A learning algorithm for continually running fully", + "type": "text", + "cross_page": true + } + ], + "index": 15, + "is_list_start_line": true + }, + { + "bbox": [ + 125, + 295, + 408, + 309 + ], + "spans": [ + { + "bbox": [ + 125, + 295, + 408, + 309 + ], + "score": 1.0, + "content": "recurrent neural networks. Neural Computation, 1(2):270–280, 1989.", + "type": "text", + "cross_page": true + } + ], + "index": 16, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 314, + 505, + 328 + ], + "spans": [ + { + "bbox": [ + 105, + 314, + 505, + 328 + ], + "score": 1.0, + "content": "[80] Zhenda Xie, Zigang Geng, Jingcheng Hu, Zheng Zhang, Han Hu, and Yue Cao. Revealing the", + "type": "text", + "cross_page": true + } + ], + "index": 17, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 325, + 357, + 338 + ], + "spans": [ + { + "bbox": [ + 127, + 325, + 357, + 338 + ], + "score": 1.0, + "content": "dark secrets of masked image modeling. In CVPR, 2023.", + "type": "text", + "cross_page": true + } + ], + "index": 18, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 343, + 506, + 359 + ], + "spans": [ + { + "bbox": [ + 105, + 343, + 506, + 359 + ], + "score": 1.0, + "content": "[81] Haofei Xu, Jing Zhang, Jianfei Cai, Hamid Rezatofighi, and Dacheng Tao. GMFlow: Learning", + "type": "text", + "cross_page": true + } + ], + "index": 19, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 354, + 405, + 369 + ], + "spans": [ + { + "bbox": [ + 126, + 354, + 405, + 369 + ], + "score": 1.0, + "content": "optical flow via global matching. In CVPR, pages 8121–8130, 2022.", + "type": "text", + "cross_page": true + } + ], + "index": 20, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 372, + 506, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 372, + 506, + 387 + ], + "score": 1.0, + "content": "[82] Li Xu, Jiaya Jia, and Yasuyuki Matsushita. Motion detail preserving optical flow estimation.", + "type": "text", + "cross_page": true + } + ], + "index": 21, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 384, + 286, + 397 + ], + "spans": [ + { + "bbox": [ + 127, + 384, + 286, + 397 + ], + "score": 1.0, + "content": "IEEE T-PAMI, 34(9):1744–1757, 2011.", + "type": "text", + "cross_page": true + } + ], + "index": 22, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 403, + 507, + 417 + ], + "spans": [ + { + "bbox": [ + 105, + 403, + 507, + 417 + ], + "score": 1.0, + "content": "[83] Gengshan Yang and Deva Ramanan. Volumetric correspondence networks for optical flow.", + "type": "text", + "cross_page": true + } + ], + "index": 23, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 414, + 193, + 427 + ], + "spans": [ + { + "bbox": [ + 127, + 414, + 193, + 427 + ], + "score": 1.0, + "content": "NeurIPS, 2019.", + "type": "text", + "cross_page": true + } + ], + "index": 24, + "is_list_end_line": true + }, + { + "bbox": [ + 104, + 432, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 104, + 432, + 506, + 448 + ], + "score": 1.0, + "content": "[84] Lantao Yu, Weinan Zhang, Jun Wang, and Yong Yu. SeqGAN: Sequence generative adversarial", + "type": "text", + "cross_page": true + } + ], + "index": 25, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 444, + 295, + 458 + ], + "spans": [ + { + "bbox": [ + 126, + 444, + 295, + 458 + ], + "score": 1.0, + "content": "nets with policy gradient. In AAAI, 2017.", + "type": "text", + "cross_page": true + } + ], + "index": 26, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 462, + 505, + 477 + ], + "spans": [ + { + "bbox": [ + 105, + 462, + 505, + 477 + ], + "score": 1.0, + "content": "[85] Feihu Zhang, Oliver J. Woodford, Victor Adrian Prisacariu, and Philip H.S. Torr. Separable", + "type": "text", + "cross_page": true + } + ], + "index": 27, + "is_list_start_line": true + }, + { + "bbox": [ + 126, + 473, + 507, + 488 + ], + "spans": [ + { + "bbox": [ + 126, + 473, + 507, + 488 + ], + "score": 1.0, + "content": "flow: Learning motion cost volumes for optical flow estimation. In ICCV, pages 10807–10817,", + "type": "text", + "cross_page": true + } + ], + "index": 28 + }, + { + "bbox": [ + 126, + 483, + 155, + 498 + ], + "spans": [ + { + "bbox": [ + 126, + 483, + 155, + 498 + ], + "score": 1.0, + "content": "2021.", + "type": "text", + "cross_page": true + } + ], + "index": 29, + "is_list_end_line": true + }, + { + "bbox": [ + 104, + 502, + 506, + 519 + ], + "spans": [ + { + "bbox": [ + 104, + 502, + 506, + 519 + ], + "score": 1.0, + "content": "[86] Richard Zhang, Phillip Isola, and Alexei A. Efros. Colorful image colorization. In ECCV, pages", + "type": "text", + "cross_page": true + } + ], + "index": 30, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 515, + 192, + 528 + ], + "spans": [ + { + "bbox": [ + 127, + 515, + 192, + 528 + ], + "score": 1.0, + "content": "649–666, 2016.", + "type": "text", + "cross_page": true + } + ], + "index": 31, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 533, + 506, + 547 + ], + "spans": [ + { + "bbox": [ + 106, + 533, + 506, + 547 + ], + "score": 1.0, + "content": "[87] Jiawei Zhao, Ke Yan, Yifan Zhao, Xiaowei Guo, Feiyue Huang, and Jia Li. Transformer-based", + "type": "text", + "cross_page": true + } + ], + "index": 32, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 544, + 474, + 559 + ], + "spans": [ + { + "bbox": [ + 127, + 544, + 474, + 559 + ], + "score": 1.0, + "content": "dual relation graph for multi-label image recognition. In ICCV, pages 163–172, 2021.", + "type": "text", + "cross_page": true + } + ], + "index": 33, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 563, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 106, + 563, + 506, + 577 + ], + "score": 1.0, + "content": "[88] Bolei Zhou, Agata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba. Places: A 10", + "type": "text", + "cross_page": true + } + ], + "index": 34, + "is_list_start_line": true + }, + { + "bbox": [ + 127, + 574, + 398, + 587 + ], + "spans": [ + { + "bbox": [ + 127, + 574, + 398, + 587 + ], + "score": 1.0, + "content": "million image database for scene recognition. IEEE T-PAMI, 2017.", + "type": "text", + "cross_page": true + } + ], + "index": 35, + "is_list_end_line": true + } + ], + "index": 22.5, + "bbox_fs": [ + 103, + 70, + 507, + 724 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 30, + 507, + 723 + ], + "lines": [ + { + "bbox": [ + 104, + 72, + 506, + 85 + ], + "spans": [ + { + "bbox": [ + 104, + 72, + 506, + 85 + ], + "score": 1.0, + "content": "[55] Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily Denton, Seyed", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 127, + 83, + 506, + 97 + ], + "spans": [ + { + "bbox": [ + 127, + 83, + 506, + 97 + ], + "score": 1.0, + "content": "Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, S. Sara Mahdavi, Rapha Gontijo Lopes, Tim", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 125, + 91, + 507, + 110 + ], + "spans": [ + { + "bbox": [ + 125, + 91, + 507, + 110 + ], + "score": 1.0, + "content": "Salimans, Jonathan Ho, David J. Fleet, and Mohammad Norouzi. Photorealistic Text-to-Image", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 128, + 106, + 425, + 118 + ], + "spans": [ + { + "bbox": [ + 128, + 106, + 425, + 118 + ], + "score": 1.0, + "content": "Diffusion Models with Deep Language Understanding. In NeurIPS, 2022.", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 125, + 507, + 139 + ], + "spans": [ + { + "bbox": [ + 105, + 125, + 507, + 139 + ], + "score": 1.0, + "content": "[56] Tim Salimans and Jonathan Ho. Progressive distillation for fast sampling of diffusion models.", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 128, + 136, + 191, + 149 + ], + "spans": [ + { + "bbox": [ + 128, + 136, + 191, + 149 + ], + "score": 1.0, + "content": "In ICLR, 2022.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 155, + 507, + 170 + ], + "spans": [ + { + "bbox": [ + 105, + 155, + 507, + 170 + ], + "score": 1.0, + "content": "[57] Nikolay Savinov, Junyoung Chung, Mikolaj Binkowski, Erich Elsen, and Aaron van den Oord.", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 127, + 167, + 422, + 180 + ], + "spans": [ + { + "bbox": [ + 127, + 167, + 422, + 180 + ], + "score": 1.0, + "content": "Step-unrolled denoising autoencoders for text generation. In ICLR, 2022.", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 186, + 507, + 200 + ], + "spans": [ + { + "bbox": [ + 105, + 186, + 507, + 200 + ], + "score": 1.0, + "content": "[58] Ashutosh Saxena, Sung Chung, and Andrew Ng. Learning depth from single monocular images.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 126, + 196, + 181, + 211 + ], + "spans": [ + { + "bbox": [ + 126, + 196, + 181, + 211 + ], + "score": 1.0, + "content": "NIPS, 2005.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 216, + 507, + 230 + ], + "spans": [ + { + "bbox": [ + 105, + 216, + 507, + 230 + ], + "score": 1.0, + "content": "[59] Ashutosh Saxena, Min Sun, and Andrew Y. Ng. Make3D: Learning 3D scene structure from a", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 126, + 227, + 352, + 241 + ], + "spans": [ + { + "bbox": [ + 126, + 227, + 352, + 241 + ], + "score": 1.0, + "content": "single still image. IEEE T-PAMI, 31(5):824–840, 2009.", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 103, + 245, + 507, + 262 + ], + "spans": [ + { + "bbox": [ + 103, + 245, + 507, + 262 + ], + "score": 1.0, + "content": "[60] Meng-Li Shih, Shih-Yang Su, Johannes Kopf, and Jia-Bin Huang. 3D photography using", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 126, + 258, + 357, + 272 + ], + "spans": [ + { + "bbox": [ + 126, + 258, + 357, + 272 + ], + "score": 1.0, + "content": "context-aware layered depth inpainting. In CVPR, 2020.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "score": 1.0, + "content": "[61] Nathan Silberman, Derek Hoiem, Pushmeet Kohli, and Rob Fergus. Indoor segmentation and", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 126, + 288, + 418, + 302 + ], + "spans": [ + { + "bbox": [ + 126, + 288, + 418, + 302 + ], + "score": 1.0, + "content": "support inference from RGBD images. In ECCV, pages 746–760, 2012.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 104, + 307, + 507, + 322 + ], + "spans": [ + { + "bbox": [ + 104, + 307, + 507, + 322 + ], + "score": 1.0, + "content": "[62] Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsuper-", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 126, + 318, + 487, + 332 + ], + "spans": [ + { + "bbox": [ + 126, + 318, + 487, + 332 + ], + "score": 1.0, + "content": "vised learning using nonequilibrium thermodynamics. In ICML, pages 2256–2265, 2015.", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 339, + 505, + 351 + ], + "spans": [ + { + "bbox": [ + 106, + 339, + 505, + 351 + ], + "score": 1.0, + "content": "[63] Yang Song, Jascha Sohl-Dickstein, Diederik P. Kingma, Abhishek Kumar, Stefano Ermon, and", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 126, + 348, + 507, + 363 + ], + "spans": [ + { + "bbox": [ + 126, + 348, + 507, + 363 + ], + "score": 1.0, + "content": "Ben Poole. Score-based generative modeling through stochastic differential equations. In ICLR,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 126, + 359, + 155, + 374 + ], + "spans": [ + { + "bbox": [ + 126, + 359, + 155, + 374 + ], + "score": 1.0, + "content": "2021.", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 379, + 507, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 379, + 507, + 394 + ], + "score": 1.0, + "content": "[64] Yang Song, Prafulla Dhariwal, Mark Chen, and Ilya Sutskever. Consistency models. In ICML,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 126, + 390, + 155, + 405 + ], + "spans": [ + { + "bbox": [ + 126, + 390, + 155, + 405 + ], + "score": 1.0, + "content": "2023.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 410, + 506, + 424 + ], + "spans": [ + { + "bbox": [ + 106, + 410, + 506, + 424 + ], + "score": 1.0, + "content": "[65] Martin Stommel, Michael Beetz, and Weiliang Xu. Inpainting of missing values in the kinect", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 128, + 421, + 505, + 433 + ], + "spans": [ + { + "bbox": [ + 128, + 421, + 505, + 433 + ], + "score": 1.0, + "content": "sensor’s depth maps based on background estimates. IEEE Sensors Journal, 14(4):1107–1116,", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 127, + 430, + 155, + 446 + ], + "spans": [ + { + "bbox": [ + 127, + 430, + 155, + 446 + ], + "score": 1.0, + "content": "2014.", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 104, + 451, + 506, + 466 + ], + "spans": [ + { + "bbox": [ + 104, + 451, + 506, + 466 + ], + "score": 1.0, + "content": "[66] Xiuchao Sui, Shaohua Li, Xue Geng, Yan Wu, Xinxing Xu, Yong Liu, Rick Goh, and Hongyuan", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 126, + 460, + 507, + 478 + ], + "spans": [ + { + "bbox": [ + 126, + 460, + 507, + 478 + ], + "score": 1.0, + "content": "Zhu. CRAFT: Cross-attentional flow transformer for robust optical flow. In CVPR, pages", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 126, + 473, + 211, + 486 + ], + "spans": [ + { + "bbox": [ + 126, + 473, + 211, + 486 + ], + "score": 1.0, + "content": "17602–17611, 2022.", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 492, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 506, + 507 + ], + "score": 1.0, + "content": "[67] Deqing Sun, Erik B. Sudderth, and Michael J. Black. Layered segmentation and optical flow", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 126, + 504, + 358, + 518 + ], + "spans": [ + { + "bbox": [ + 126, + 504, + 358, + 518 + ], + "score": 1.0, + "content": "estimation over time. In CVPR, pages 1768–1775, 2012.", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 523, + 506, + 537 + ], + "spans": [ + { + "bbox": [ + 106, + 523, + 506, + 537 + ], + "score": 1.0, + "content": "[68] Deqing Sun, Xiaodong Yang, Ming-Yu Liu, and Jan Kautz. PWC-Net: CNNs for optical flow", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 127, + 534, + 440, + 549 + ], + "spans": [ + { + "bbox": [ + 127, + 534, + 440, + 549 + ], + "score": 1.0, + "content": "using pyramid, warping, and cost volume. In CVPR, pages 8934–8943, 2018.", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 553, + 506, + 568 + ], + "spans": [ + { + "bbox": [ + 106, + 553, + 506, + 568 + ], + "score": 1.0, + "content": "[69] Deqing Sun, Daniel Vlasic, Charles Herrmann, Varun Jampani, Michael Krainin, Huiwen", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 126, + 563, + 507, + 581 + ], + "spans": [ + { + "bbox": [ + 126, + 563, + 507, + 581 + ], + "score": 1.0, + "content": "Chang, Ramin Zabih, William T. Freeman, and Ce Liu. AutoFlow: Learning a better training", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 126, + 575, + 360, + 590 + ], + "spans": [ + { + "bbox": [ + 126, + 575, + 360, + 590 + ], + "score": 1.0, + "content": "set for optical flow. In CVPR, pages 10093–10102, 2021.", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 595, + 507, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 595, + 507, + 609 + ], + "score": 1.0, + "content": "[70] Deqing Sun, Charles Herrmann, Fitsum Reda, Michael Rubinstein, David J. Fleet, and William T.", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 126, + 605, + 507, + 620 + ], + "spans": [ + { + "bbox": [ + 126, + 605, + 507, + 620 + ], + "score": 1.0, + "content": "Freeman. Disentangling architecture and training for optical flow. In ECCV, pages 165–182,", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 127, + 617, + 154, + 631 + ], + "spans": [ + { + "bbox": [ + 127, + 617, + 154, + 631 + ], + "score": 1.0, + "content": "2022.", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 636, + 507, + 650 + ], + "spans": [ + { + "bbox": [ + 105, + 636, + 507, + 650 + ], + "score": 1.0, + "content": "[71] Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aurelien Chouard, Vijaysai Patnaik, Paul Tsui,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 126, + 646, + 506, + 662 + ], + "spans": [ + { + "bbox": [ + 126, + 646, + 506, + 662 + ], + "score": 1.0, + "content": "James Guo, Yin Zhou, Yuning Chai, Benjamin Caine, Vijay Vasudevan, Wei Han, Jiquan Ngiam,", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 126, + 657, + 507, + 674 + ], + "spans": [ + { + "bbox": [ + 126, + 657, + 507, + 674 + ], + "score": 1.0, + "content": "Hang Zhao, Aleksei Timofeev, Scott Ettinger, Maxim Krivokon, Amy Gao, Aditya Joshi, Sheng", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 128, + 670, + 506, + 682 + ], + "spans": [ + { + "bbox": [ + 128, + 670, + 506, + 682 + ], + "score": 1.0, + "content": "Zhao, Shuyang Cheng, Yu Zhang, Jonathon Shlens, Zhifeng Chen, and Dragomir Anguelov.", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 126, + 680, + 483, + 694 + ], + "spans": [ + { + "bbox": [ + 126, + 680, + 483, + 694 + ], + "score": 1.0, + "content": "Scalability in perception for autonomous driving: Waymo open dataset. In CVPR, 2020.", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 700, + 505, + 714 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 505, + 714 + ], + "score": 1.0, + "content": "[72] Shangkun Sun, Yuanqi Chen, Yu Zhu, Guodong Guo, and Ge Li. SKFlow: Learning optical", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 127, + 711, + 302, + 725 + ], + "spans": [ + { + "bbox": [ + 127, + 711, + 302, + 725 + ], + "score": 1.0, + "content": "flow with super kernels. In NeurIPS, 2022.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 22.5 + } + ], + "page_idx": 13, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "14", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "list", + "bbox": [ + 105, + 30, + 507, + 723 + ], + "lines": [], + "index": 22.5, + "bbox_fs": [ + 103, + 72, + 507, + 725 + ], + "lines_deleted": true + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 72, + 507, + 591 + ], + "lines": [ + { + "bbox": [ + 105, + 72, + 505, + 87 + ], + "spans": [ + { + "bbox": [ + 105, + 72, + 505, + 87 + ], + "score": 1.0, + "content": "[73] Mingxing Tan and Quoc Le. EfficientNet: Rethinking model scaling for convolutional neural", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 127, + 84, + 311, + 97 + ], + "spans": [ + { + "bbox": [ + 127, + 84, + 311, + 97 + ], + "score": 1.0, + "content": "networks. In ICML, pages 6105–6114, 2019.", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 102, + 506, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 102, + 506, + 117 + ], + "score": 1.0, + "content": "[74] Zachary Teed and Jia Deng. RAFT: Recurrent all-pairs field transforms for optical flow. In", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 127, + 113, + 249, + 128 + ], + "spans": [ + { + "bbox": [ + 127, + 113, + 249, + 128 + ], + "score": 1.0, + "content": "ECCV, pages 402–419, 2020.", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 132, + 506, + 147 + ], + "spans": [ + { + "bbox": [ + 105, + 132, + 506, + 147 + ], + "score": 1.0, + "content": "[75] Su Wang, Chitwan Saharia, Ceslee Montgomery, Jordi Pont-Tuset, Shai Noy, Stefano Pellegrini,", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 127, + 143, + 506, + 157 + ], + "spans": [ + { + "bbox": [ + 127, + 143, + 506, + 157 + ], + "score": 1.0, + "content": "Yasumasa Onoe, Sarah Laszlo, David J. Fleet, Radu Soricut, Jason Baldridge, Mohammad", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 127, + 154, + 506, + 168 + ], + "spans": [ + { + "bbox": [ + 127, + 154, + 506, + 168 + ], + "score": 1.0, + "content": "Norouzi, Peter Anderson, and William Chan. Imagen Editor and EditBench: Advancing and", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 126, + 165, + 358, + 179 + ], + "spans": [ + { + "bbox": [ + 126, + 165, + 358, + 179 + ], + "score": 1.0, + "content": "evaluating text-guided image inpainting. In CVPR, 2023.", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 184, + 506, + 198 + ], + "spans": [ + { + "bbox": [ + 105, + 184, + 506, + 198 + ], + "score": 1.0, + "content": "[76] Wenshan Wang, Delong Zhu, Xiangwei Wang, Yaoyu Hu, Yuheng Qiu, Chen Wang, Yafei Hu,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 127, + 195, + 507, + 209 + ], + "spans": [ + { + "bbox": [ + 127, + 195, + 507, + 209 + ], + "score": 1.0, + "content": "Ashish Kapoor, and Sebastian Scherer. TartanAir: A dataset to push the limits of visual SLAM.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 127, + 206, + 190, + 219 + ], + "spans": [ + { + "bbox": [ + 127, + 206, + 190, + 219 + ], + "score": 1.0, + "content": "In IROS, 2020.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 104, + 223, + 506, + 240 + ], + "spans": [ + { + "bbox": [ + 104, + 223, + 506, + 240 + ], + "score": 1.0, + "content": "[77] Philippe Weinzaepfel, Jerome Revaud, Zaid Harchaoui, and Cordelia Schmid. DeepFlow: Large", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 126, + 235, + 456, + 249 + ], + "spans": [ + { + "bbox": [ + 126, + 235, + 456, + 249 + ], + "score": 1.0, + "content": "displacement optical flow with deep matching. In ICCV, pages 1385–1392, 2013.", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 254, + 505, + 269 + ], + "spans": [ + { + "bbox": [ + 105, + 254, + 505, + 269 + ], + "score": 1.0, + "content": "[78] Olivia Wiles, Georgia Gkioxari, Richard Szeliski, and Justin Johnson. SynSin: End-to-end view", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 126, + 264, + 319, + 280 + ], + "spans": [ + { + "bbox": [ + 126, + 264, + 319, + 280 + ], + "score": 1.0, + "content": "synthesis from a single image. In CVPR, 2020.", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 284, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 284, + 505, + 298 + ], + "score": 1.0, + "content": "[79] Ronald J. Williams and David Zipser. A learning algorithm for continually running fully", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 125, + 295, + 408, + 309 + ], + "spans": [ + { + "bbox": [ + 125, + 295, + 408, + 309 + ], + "score": 1.0, + "content": "recurrent neural networks. Neural Computation, 1(2):270–280, 1989.", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 314, + 505, + 328 + ], + "spans": [ + { + "bbox": [ + 105, + 314, + 505, + 328 + ], + "score": 1.0, + "content": "[80] Zhenda Xie, Zigang Geng, Jingcheng Hu, Zheng Zhang, Han Hu, and Yue Cao. Revealing the", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 127, + 325, + 357, + 338 + ], + "spans": [ + { + "bbox": [ + 127, + 325, + 357, + 338 + ], + "score": 1.0, + "content": "dark secrets of masked image modeling. In CVPR, 2023.", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 343, + 506, + 359 + ], + "spans": [ + { + "bbox": [ + 105, + 343, + 506, + 359 + ], + "score": 1.0, + "content": "[81] Haofei Xu, Jing Zhang, Jianfei Cai, Hamid Rezatofighi, and Dacheng Tao. GMFlow: Learning", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 126, + 354, + 405, + 369 + ], + "spans": [ + { + "bbox": [ + 126, + 354, + 405, + 369 + ], + "score": 1.0, + "content": "optical flow via global matching. In CVPR, pages 8121–8130, 2022.", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 372, + 506, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 372, + 506, + 387 + ], + "score": 1.0, + "content": "[82] Li Xu, Jiaya Jia, and Yasuyuki Matsushita. Motion detail preserving optical flow estimation.", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 127, + 384, + 286, + 397 + ], + "spans": [ + { + "bbox": [ + 127, + 384, + 286, + 397 + ], + "score": 1.0, + "content": "IEEE T-PAMI, 34(9):1744–1757, 2011.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 403, + 507, + 417 + ], + "spans": [ + { + "bbox": [ + 105, + 403, + 507, + 417 + ], + "score": 1.0, + "content": "[83] Gengshan Yang and Deva Ramanan. Volumetric correspondence networks for optical flow.", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 127, + 414, + 193, + 427 + ], + "spans": [ + { + "bbox": [ + 127, + 414, + 193, + 427 + ], + "score": 1.0, + "content": "NeurIPS, 2019.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 104, + 432, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 104, + 432, + 506, + 448 + ], + "score": 1.0, + "content": "[84] Lantao Yu, Weinan Zhang, Jun Wang, and Yong Yu. SeqGAN: Sequence generative adversarial", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 126, + 444, + 295, + 458 + ], + "spans": [ + { + "bbox": [ + 126, + 444, + 295, + 458 + ], + "score": 1.0, + "content": "nets with policy gradient. In AAAI, 2017.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 462, + 505, + 477 + ], + "spans": [ + { + "bbox": [ + 105, + 462, + 505, + 477 + ], + "score": 1.0, + "content": "[85] Feihu Zhang, Oliver J. Woodford, Victor Adrian Prisacariu, and Philip H.S. Torr. Separable", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 126, + 473, + 507, + 488 + ], + "spans": [ + { + "bbox": [ + 126, + 473, + 507, + 488 + ], + "score": 1.0, + "content": "flow: Learning motion cost volumes for optical flow estimation. In ICCV, pages 10807–10817,", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 126, + 483, + 155, + 498 + ], + "spans": [ + { + "bbox": [ + 126, + 483, + 155, + 498 + ], + "score": 1.0, + "content": "2021.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 104, + 502, + 506, + 519 + ], + "spans": [ + { + "bbox": [ + 104, + 502, + 506, + 519 + ], + "score": 1.0, + "content": "[86] Richard Zhang, Phillip Isola, and Alexei A. Efros. Colorful image colorization. In ECCV, pages", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 127, + 515, + 192, + 528 + ], + "spans": [ + { + "bbox": [ + 127, + 515, + 192, + 528 + ], + "score": 1.0, + "content": "649–666, 2016.", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 533, + 506, + 547 + ], + "spans": [ + { + "bbox": [ + 106, + 533, + 506, + 547 + ], + "score": 1.0, + "content": "[87] Jiawei Zhao, Ke Yan, Yifan Zhao, Xiaowei Guo, Feiyue Huang, and Jia Li. Transformer-based", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 127, + 544, + 474, + 559 + ], + "spans": [ + { + "bbox": [ + 127, + 544, + 474, + 559 + ], + "score": 1.0, + "content": "dual relation graph for multi-label image recognition. In ICCV, pages 163–172, 2021.", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 563, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 106, + 563, + 506, + 577 + ], + "score": 1.0, + "content": "[88] Bolei Zhou, Agata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba. Places: A 10", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 127, + 574, + 398, + 587 + ], + "spans": [ + { + "bbox": [ + 127, + 574, + 398, + 587 + ], + "score": 1.0, + "content": "million image database for scene recognition. IEEE T-PAMI, 2017.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 17.5 + } + ], + "page_idx": 14, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "15", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "list", + "bbox": [ + 105, + 72, + 507, + 591 + ], + "lines": [], + "index": 17.5, + "bbox_fs": [ + 104, + 72, + 507, + 587 + ], + "lines_deleted": true + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 71, + 244, + 84 + ], + "lines": [ + { + "bbox": [ + 105, + 69, + 245, + 87 + ], + "spans": [ + { + "bbox": [ + 105, + 69, + 245, + 87 + ], + "score": 1.0, + "content": "A Multimodal prediction", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 106, + 95, + 505, + 183 + ], + "lines": [ + { + "bbox": [ + 105, + 95, + 506, + 109 + ], + "spans": [ + { + "bbox": [ + 105, + 95, + 506, + 109 + ], + "score": 1.0, + "content": "We provide more qualitative examples for multimodal prediction. Figures 8 and 9 illustrate multi-", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 108, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 106, + 108, + 505, + 118 + ], + "score": 1.0, + "content": "modal depth predictions on NYU and KITTI respectively. Multimodality of the posterior distribution", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 118, + 505, + 130 + ], + "spans": [ + { + "bbox": [ + 106, + 118, + 505, + 130 + ], + "score": 1.0, + "content": "exists in regions where there are multiple plausible predictions. For example, this includes reflective", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 129, + 505, + 141 + ], + "spans": [ + { + "bbox": [ + 106, + 129, + 505, + 141 + ], + "score": 1.0, + "content": "and transparent surfaces (mirrors and glass surfaces in rows 1 to 5 of Figure 8 and windows of cars in", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 139, + 506, + 152 + ], + "spans": [ + { + "bbox": [ + 105, + 139, + 506, + 152 + ], + "score": 1.0, + "content": "Figure 9). We further find that the model captures uncertainty in depth estimates in the vicinity of", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 149, + 505, + 164 + ], + "spans": [ + { + "bbox": [ + 105, + 149, + 505, + 164 + ], + "score": 1.0, + "content": "object boundaries, some of which arise due to noise in ground truth measurements in the training", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 161, + 505, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 161, + 505, + 174 + ], + "score": 1.0, + "content": "data. This can be observed at the boundaries of cars in Figure 9 and around edges of objects in Figure", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 173, + 262, + 184 + ], + "spans": [ + { + "bbox": [ + 106, + 173, + 262, + 184 + ], + "score": 1.0, + "content": "8 (most clearly visible in the last row).", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 4.5 + }, + { + "type": "text", + "bbox": [ + 106, + 189, + 505, + 232 + ], + "lines": [ + { + "bbox": [ + 105, + 187, + 505, + 202 + ], + "spans": [ + { + "bbox": [ + 105, + 187, + 505, + 202 + ], + "score": 1.0, + "content": "Figure 10 illustrates different samples on KITTI from the optical flow diffusion model, also capturing", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 200, + 505, + 212 + ], + "spans": [ + { + "bbox": [ + 106, + 200, + 505, + 212 + ], + "score": 1.0, + "content": "multiple modes of the predictive posterior. Multimodality exists on transparent surfaces and near", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 210, + 505, + 222 + ], + "spans": [ + { + "bbox": [ + 105, + 210, + 505, + 222 + ], + "score": 1.0, + "content": "occlusions. As shown in Figure 11, on Sintel, multimodality also exists on occluded or out-of-bounds", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 221, + 298, + 234 + ], + "spans": [ + { + "bbox": [ + 105, + 221, + 298, + 234 + ], + "score": 1.0, + "content": "pixels where multiple predictions are plausible.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 10.5 + }, + { + "type": "image", + "bbox": [ + 108, + 241, + 504, + 609 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 241, + 504, + 609 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 241, + 504, + 609 + ], + "spans": [ + { + "bbox": [ + 108, + 241, + 504, + 609 + ], + "score": 0.971, + "type": "image", + "image_path": "3c99984f3d20cd7326a7c031f4bdfa98268c4ae79149f0f348f4dbd850bbaef3.jpg" + } + ] + } + ], + "index": 14, + "virtual_lines": [ + { + "bbox": [ + 108, + 241, + 504, + 363.6666666666667 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 108, + 363.6666666666667, + 504, + 486.33333333333337 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 108, + 486.33333333333337, + 504, + 609.0 + ], + "spans": [], + "index": 15 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 621, + 505, + 662 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 622, + 505, + 632 + ], + "spans": [ + { + "bbox": [ + 106, + 622, + 505, + 632 + ], + "score": 1.0, + "content": "Figure 8: Qualitative examples of multimodal estimation on the NYU depth dataset. Our model is able", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 632, + 505, + 642 + ], + "spans": [ + { + "bbox": [ + 106, + 632, + 505, + 642 + ], + "score": 1.0, + "content": "to output multiple plausible depth maps where ambiguity exists. Rows 1 to 5 show transparent or reflective", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "score": 1.0, + "content": "surfaces where two answers exist. In all samples (specially the last row) we observe the model’s ability to", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 650, + 408, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 650, + 408, + 663 + ], + "score": 1.0, + "content": "capture uncertainty in depth near object boundaries (see the areas of high variance).", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 17.5 + } + ], + "index": 15.75 + } + ], + "page_idx": 15, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 742, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 14 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 71, + 244, + 84 + ], + "lines": [ + { + "bbox": [ + 105, + 69, + 245, + 87 + ], + "spans": [ + { + "bbox": [ + 105, + 69, + 245, + 87 + ], + "score": 1.0, + "content": "A Multimodal prediction", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 106, + 95, + 505, + 183 + ], + "lines": [ + { + "bbox": [ + 105, + 95, + 506, + 109 + ], + "spans": [ + { + "bbox": [ + 105, + 95, + 506, + 109 + ], + "score": 1.0, + "content": "We provide more qualitative examples for multimodal prediction. Figures 8 and 9 illustrate multi-", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 108, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 106, + 108, + 505, + 118 + ], + "score": 1.0, + "content": "modal depth predictions on NYU and KITTI respectively. Multimodality of the posterior distribution", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 118, + 505, + 130 + ], + "spans": [ + { + "bbox": [ + 106, + 118, + 505, + 130 + ], + "score": 1.0, + "content": "exists in regions where there are multiple plausible predictions. For example, this includes reflective", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 129, + 505, + 141 + ], + "spans": [ + { + "bbox": [ + 106, + 129, + 505, + 141 + ], + "score": 1.0, + "content": "and transparent surfaces (mirrors and glass surfaces in rows 1 to 5 of Figure 8 and windows of cars in", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 139, + 506, + 152 + ], + "spans": [ + { + "bbox": [ + 105, + 139, + 506, + 152 + ], + "score": 1.0, + "content": "Figure 9). We further find that the model captures uncertainty in depth estimates in the vicinity of", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 149, + 505, + 164 + ], + "spans": [ + { + "bbox": [ + 105, + 149, + 505, + 164 + ], + "score": 1.0, + "content": "object boundaries, some of which arise due to noise in ground truth measurements in the training", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 161, + 505, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 161, + 505, + 174 + ], + "score": 1.0, + "content": "data. This can be observed at the boundaries of cars in Figure 9 and around edges of objects in Figure", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 173, + 262, + 184 + ], + "spans": [ + { + "bbox": [ + 106, + 173, + 262, + 184 + ], + "score": 1.0, + "content": "8 (most clearly visible in the last row).", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 4.5, + "bbox_fs": [ + 105, + 95, + 506, + 184 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 189, + 505, + 232 + ], + "lines": [ + { + "bbox": [ + 105, + 187, + 505, + 202 + ], + "spans": [ + { + "bbox": [ + 105, + 187, + 505, + 202 + ], + "score": 1.0, + "content": "Figure 10 illustrates different samples on KITTI from the optical flow diffusion model, also capturing", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 200, + 505, + 212 + ], + "spans": [ + { + "bbox": [ + 106, + 200, + 505, + 212 + ], + "score": 1.0, + "content": "multiple modes of the predictive posterior. Multimodality exists on transparent surfaces and near", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 210, + 505, + 222 + ], + "spans": [ + { + "bbox": [ + 105, + 210, + 505, + 222 + ], + "score": 1.0, + "content": "occlusions. As shown in Figure 11, on Sintel, multimodality also exists on occluded or out-of-bounds", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 221, + 298, + 234 + ], + "spans": [ + { + "bbox": [ + 105, + 221, + 298, + 234 + ], + "score": 1.0, + "content": "pixels where multiple predictions are plausible.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 10.5, + "bbox_fs": [ + 105, + 187, + 505, + 234 + ] + }, + { + "type": "image", + "bbox": [ + 108, + 241, + 504, + 609 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 241, + 504, + 609 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 241, + 504, + 609 + ], + "spans": [ + { + "bbox": [ + 108, + 241, + 504, + 609 + ], + "score": 0.971, + "type": "image", + "image_path": "3c99984f3d20cd7326a7c031f4bdfa98268c4ae79149f0f348f4dbd850bbaef3.jpg" + } + ] + } + ], + "index": 14, + "virtual_lines": [ + { + "bbox": [ + 108, + 241, + 504, + 363.6666666666667 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 108, + 363.6666666666667, + 504, + 486.33333333333337 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 108, + 486.33333333333337, + 504, + 609.0 + ], + "spans": [], + "index": 15 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 621, + 505, + 662 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 622, + 505, + 632 + ], + "spans": [ + { + "bbox": [ + 106, + 622, + 505, + 632 + ], + "score": 1.0, + "content": "Figure 8: Qualitative examples of multimodal estimation on the NYU depth dataset. Our model is able", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 632, + 505, + 642 + ], + "spans": [ + { + "bbox": [ + 106, + 632, + 505, + 642 + ], + "score": 1.0, + "content": "to output multiple plausible depth maps where ambiguity exists. Rows 1 to 5 show transparent or reflective", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "score": 1.0, + "content": "surfaces where two answers exist. In all samples (specially the last row) we observe the model’s ability to", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 650, + 408, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 650, + 408, + 663 + ], + "score": 1.0, + "content": "capture uncertainty in depth near object boundaries (see the areas of high variance).", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 17.5 + } + ], + "index": 15.75 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 124, + 506, + 318 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 124, + 506, + 318 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 124, + 506, + 318 + ], + "spans": [ + { + "bbox": [ + 107, + 124, + 506, + 318 + ], + "score": 0.969, + "type": "image", + "image_path": "20353338bc9ad67dbe37e7016a060d30dd2847aa34644787bdd18a5a93a98c3b.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 124, + 506, + 188.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 188.66666666666669, + 506, + 253.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 253.33333333333337, + 506, + 318.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 326, + 505, + 358 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 326, + 506, + 339 + ], + "spans": [ + { + "bbox": [ + 105, + 326, + 506, + 339 + ], + "score": 1.0, + "content": "Figure 9: Qualitative examples of multimodal depth estimation on KITTI. Our model is able to predict", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 337, + 505, + 349 + ], + "spans": [ + { + "bbox": [ + 106, + 337, + 505, + 349 + ], + "score": 1.0, + "content": "multimodal samples, especially windows of cars and object boundaries. Please refer to the areas with high", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 347, + 142, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 142, + 358 + ], + "score": 1.0, + "content": "variance.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 107, + 457, + 505, + 633 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 457, + 505, + 633 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 107, + 457, + 505, + 633 + ], + "spans": [ + { + "bbox": [ + 107, + 457, + 505, + 633 + ], + "score": 0.967, + "type": "image", + "image_path": "2543a5fefa1c19c633a723491ca593dbdb37d51278322de13d91110f59a153c8.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 107, + 457, + 505, + 515.6666666666666 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 107, + 515.6666666666666, + 505, + 574.3333333333333 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 107, + 574.3333333333333, + 505, + 632.9999999999999 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 641, + 505, + 672 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 641, + 506, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 506, + 654 + ], + "score": 1.0, + "content": "Figure 10: Qualitative examples of multimodal optical flow estimation on KITTI. Multimodality exists on", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 652, + 505, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 652, + 505, + 663 + ], + "score": 1.0, + "content": "transparent surfaces (e.g., windows of cars) and shadows where our model estimates layered motion in different", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 662, + 204, + 673 + ], + "spans": [ + { + "bbox": [ + 105, + 662, + 204, + 673 + ], + "score": 1.0, + "content": "samples (see the last row).", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + } + ], + "index": 8.5 + } + ], + "page_idx": 16, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 299, + 740, + 312, + 754 + ], + "spans": [ + { + "bbox": [ + 299, + 740, + 312, + 754 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 124, + 506, + 318 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 124, + 506, + 318 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 124, + 506, + 318 + ], + "spans": [ + { + "bbox": [ + 107, + 124, + 506, + 318 + ], + "score": 0.969, + "type": "image", + "image_path": "20353338bc9ad67dbe37e7016a060d30dd2847aa34644787bdd18a5a93a98c3b.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 124, + 506, + 188.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 188.66666666666669, + 506, + 253.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 253.33333333333337, + 506, + 318.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 326, + 505, + 358 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 326, + 506, + 339 + ], + "spans": [ + { + "bbox": [ + 105, + 326, + 506, + 339 + ], + "score": 1.0, + "content": "Figure 9: Qualitative examples of multimodal depth estimation on KITTI. Our model is able to predict", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 337, + 505, + 349 + ], + "spans": [ + { + "bbox": [ + 106, + 337, + 505, + 349 + ], + "score": 1.0, + "content": "multimodal samples, especially windows of cars and object boundaries. Please refer to the areas with high", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 347, + 142, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 142, + 358 + ], + "score": 1.0, + "content": "variance.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 107, + 457, + 505, + 633 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 457, + 505, + 633 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 107, + 457, + 505, + 633 + ], + "spans": [ + { + "bbox": [ + 107, + 457, + 505, + 633 + ], + "score": 0.967, + "type": "image", + "image_path": "2543a5fefa1c19c633a723491ca593dbdb37d51278322de13d91110f59a153c8.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 107, + 457, + 505, + 515.6666666666666 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 107, + 515.6666666666666, + 505, + 574.3333333333333 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 107, + 574.3333333333333, + 505, + 632.9999999999999 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 641, + 505, + 672 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 641, + 506, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 506, + 654 + ], + "score": 1.0, + "content": "Figure 10: Qualitative examples of multimodal optical flow estimation on KITTI. Multimodality exists on", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 652, + 505, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 652, + 505, + 663 + ], + "score": 1.0, + "content": "transparent surfaces (e.g., windows of cars) and shadows where our model estimates layered motion in different", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 662, + 204, + 673 + ], + "spans": [ + { + "bbox": [ + 105, + 662, + 204, + 673 + ], + "score": 1.0, + "content": "samples (see the last row).", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + } + ], + "index": 8.5 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 108, + 77, + 507, + 424 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 77, + 507, + 424 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 77, + 507, + 424 + ], + "spans": [ + { + "bbox": [ + 108, + 77, + 507, + 424 + ], + "score": 0.926, + "type": "image", + "image_path": "f98ca7c4cec2c4aca1b203cd36940f0f3003785ebdf421e5dc6fe911345933e0.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 108, + 77, + 507, + 192.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 108, + 192.66666666666669, + 507, + 308.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 308.33333333333337, + 507, + 424.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 430, + 504, + 451 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 430, + 505, + 442 + ], + "spans": [ + { + "bbox": [ + 106, + 430, + 505, + 442 + ], + "score": 1.0, + "content": "Figure 11: Qualitative examples of multimodal optical flow estimation on Sintel. Multimodality also exists", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 441, + 335, + 452 + ], + "spans": [ + { + "bbox": [ + 105, + 441, + 335, + 452 + ], + "score": 1.0, + "content": "on examples with challenging occlusion or out-of-bound cases.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "title", + "bbox": [ + 107, + 489, + 401, + 503 + ], + "lines": [ + { + "bbox": [ + 104, + 488, + 403, + 506 + ], + "spans": [ + { + "bbox": [ + 104, + 488, + 403, + 506 + ], + "score": 1.0, + "content": "B Qualitative comparison of depth estimation with DPT", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 108, + 507, + 505, + 540 + ], + "lines": [ + { + "bbox": [ + 105, + 506, + 505, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 505, + 519 + ], + "score": 1.0, + "content": "Figure 12 provides a qualitative comparison of our model with DPT-Hybrid [50] finetuned on the", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 518, + 505, + 530 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 505, + 530 + ], + "score": 1.0, + "content": "NYU depth v2 [61] dataset. The depth estimates of our diffusion model are more accurate both on", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 529, + 401, + 541 + ], + "spans": [ + { + "bbox": [ + 106, + 529, + 401, + 541 + ], + "score": 1.0, + "content": "coarse-scale scene structure (walls, floors, etc.) and on individual objects.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "title", + "bbox": [ + 107, + 555, + 372, + 569 + ], + "lines": [ + { + "bbox": [ + 104, + 553, + 374, + 573 + ], + "spans": [ + { + "bbox": [ + 104, + 553, + 374, + 573 + ], + "score": 1.0, + "content": "C More samples for zero-shot imputation of depth", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 106, + 573, + 505, + 671 + ], + "lines": [ + { + "bbox": [ + 105, + 572, + 505, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 505, + 586 + ], + "score": 1.0, + "content": "Figure 13 provides samples generated using our iterative text-to-3D pipeline. We note that such", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 584, + 506, + 597 + ], + "spans": [ + { + "bbox": [ + 106, + 584, + 506, + 597 + ], + "score": 1.0, + "content": "pipelines for iteratively generating 3D scenes have been previously proposed in literature [37, 60, 78].", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 592, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 104, + 592, + 506, + 609 + ], + "score": 1.0, + "content": "However, these methods explicitly learn networks to refine the color [37, 60, 78] and the depth map", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 606, + 505, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 606, + 505, + 619 + ], + "score": 1.0, + "content": "[37, 60]. In contrast, we propose leveraging the text-conditioned image prior from existing large scale", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 616, + 506, + 630 + ], + "spans": [ + { + "bbox": [ + 105, + 616, + 506, + 630 + ], + "score": 1.0, + "content": "text-to-image [55] and text-conditional image completion [75] models, and use our depth estimation", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 627, + 506, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 506, + 640 + ], + "score": 1.0, + "content": "model zero-shot for depth completion. One caveat with our current approach of using the replacement", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 639, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 106, + 639, + 505, + 650 + ], + "score": 1.0, + "content": "method for conditional inference [63] for imputing depth, is that it does not enable one to fix errors in", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 647, + 507, + 664 + ], + "spans": [ + { + "bbox": [ + 105, + 647, + 507, + 664 + ], + "score": 1.0, + "content": "the depth predicted in the previous step. One approach to fix artifacts would be by noising-denoising,", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 659, + 494, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 659, + 494, + 672 + ], + "score": 1.0, + "content": "like that used for coarse-to-fine refinement. We leave further exploration into this to future work.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 14 + }, + { + "type": "title", + "bbox": [ + 108, + 686, + 357, + 699 + ], + "lines": [ + { + "bbox": [ + 104, + 685, + 352, + 703 + ], + "spans": [ + { + "bbox": [ + 104, + 685, + 352, + 703 + ], + "score": 1.0, + "content": "D Complete depth results on NYU and KITTI", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 106, + 700, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "score": 1.0, + "content": "Tables 9 and 10 provide detailed results on the val set of NYU depth v2 and KITTI depth datasets.", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 710, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 505, + 723 + ], + "score": 1.0, + "content": "We follow the standard evaluation protocol used in prior work [35]. For both the NYU depth v2", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20.5 + } + ], + "page_idx": 17, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 742, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 14 + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 283, + 70, + 434, + 79 + ], + "lines": [ + { + "bbox": [ + 282, + 68, + 435, + 81 + ], + "spans": [ + { + "bbox": [ + 282, + 68, + 435, + 81 + ], + "score": 1.0, + "content": "Variance heat map & multimodal samples", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 133, + 70, + 181, + 79 + ], + "lines": [ + { + "bbox": [ + 132, + 69, + 183, + 81 + ], + "spans": [ + { + "bbox": [ + 132, + 69, + 183, + 81 + ], + "score": 1.0, + "content": "Input images", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 108, + 77, + 507, + 424 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 77, + 507, + 424 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 77, + 507, + 424 + ], + "spans": [ + { + "bbox": [ + 108, + 77, + 507, + 424 + ], + "score": 0.926, + "type": "image", + "image_path": "f98ca7c4cec2c4aca1b203cd36940f0f3003785ebdf421e5dc6fe911345933e0.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 108, + 77, + 507, + 192.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 108, + 192.66666666666669, + 507, + 308.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 308.33333333333337, + 507, + 424.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 430, + 504, + 451 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 430, + 505, + 442 + ], + "spans": [ + { + "bbox": [ + 106, + 430, + 505, + 442 + ], + "score": 1.0, + "content": "Figure 11: Qualitative examples of multimodal optical flow estimation on Sintel. Multimodality also exists", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 441, + 335, + 452 + ], + "spans": [ + { + "bbox": [ + 105, + 441, + 335, + 452 + ], + "score": 1.0, + "content": "on examples with challenging occlusion or out-of-bound cases.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "title", + "bbox": [ + 107, + 489, + 401, + 503 + ], + "lines": [ + { + "bbox": [ + 104, + 488, + 403, + 506 + ], + "spans": [ + { + "bbox": [ + 104, + 488, + 403, + 506 + ], + "score": 1.0, + "content": "B Qualitative comparison of depth estimation with DPT", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 108, + 507, + 505, + 540 + ], + "lines": [ + { + "bbox": [ + 105, + 506, + 505, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 505, + 519 + ], + "score": 1.0, + "content": "Figure 12 provides a qualitative comparison of our model with DPT-Hybrid [50] finetuned on the", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 518, + 505, + 530 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 505, + 530 + ], + "score": 1.0, + "content": "NYU depth v2 [61] dataset. The depth estimates of our diffusion model are more accurate both on", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 529, + 401, + 541 + ], + "spans": [ + { + "bbox": [ + 106, + 529, + 401, + 541 + ], + "score": 1.0, + "content": "coarse-scale scene structure (walls, floors, etc.) and on individual objects.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7, + "bbox_fs": [ + 105, + 506, + 505, + 541 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 555, + 372, + 569 + ], + "lines": [ + { + "bbox": [ + 104, + 553, + 374, + 573 + ], + "spans": [ + { + "bbox": [ + 104, + 553, + 374, + 573 + ], + "score": 1.0, + "content": "C More samples for zero-shot imputation of depth", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 106, + 573, + 505, + 671 + ], + "lines": [ + { + "bbox": [ + 105, + 572, + 505, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 505, + 586 + ], + "score": 1.0, + "content": "Figure 13 provides samples generated using our iterative text-to-3D pipeline. We note that such", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 584, + 506, + 597 + ], + "spans": [ + { + "bbox": [ + 106, + 584, + 506, + 597 + ], + "score": 1.0, + "content": "pipelines for iteratively generating 3D scenes have been previously proposed in literature [37, 60, 78].", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 592, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 104, + 592, + 506, + 609 + ], + "score": 1.0, + "content": "However, these methods explicitly learn networks to refine the color [37, 60, 78] and the depth map", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 606, + 505, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 606, + 505, + 619 + ], + "score": 1.0, + "content": "[37, 60]. In contrast, we propose leveraging the text-conditioned image prior from existing large scale", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 616, + 506, + 630 + ], + "spans": [ + { + "bbox": [ + 105, + 616, + 506, + 630 + ], + "score": 1.0, + "content": "text-to-image [55] and text-conditional image completion [75] models, and use our depth estimation", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 627, + 506, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 506, + 640 + ], + "score": 1.0, + "content": "model zero-shot for depth completion. One caveat with our current approach of using the replacement", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 639, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 106, + 639, + 505, + 650 + ], + "score": 1.0, + "content": "method for conditional inference [63] for imputing depth, is that it does not enable one to fix errors in", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 647, + 507, + 664 + ], + "spans": [ + { + "bbox": [ + 105, + 647, + 507, + 664 + ], + "score": 1.0, + "content": "the depth predicted in the previous step. One approach to fix artifacts would be by noising-denoising,", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 659, + 494, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 659, + 494, + 672 + ], + "score": 1.0, + "content": "like that used for coarse-to-fine refinement. We leave further exploration into this to future work.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 14, + "bbox_fs": [ + 104, + 572, + 507, + 672 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 686, + 357, + 699 + ], + "lines": [ + { + "bbox": [ + 104, + 685, + 352, + 703 + ], + "spans": [ + { + "bbox": [ + 104, + 685, + 352, + 703 + ], + "score": 1.0, + "content": "D Complete depth results on NYU and KITTI", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 106, + 700, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 712 + ], + "score": 1.0, + "content": "Tables 9 and 10 provide detailed results on the val set of NYU depth v2 and KITTI depth datasets.", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 710, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 505, + 723 + ], + "score": 1.0, + "content": "We follow the standard evaluation protocol used in prior work [35]. For both the NYU depth v2", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 403, + 506, + 416 + ], + "spans": [ + { + "bbox": [ + 106, + 403, + 506, + 416 + ], + "score": 1.0, + "content": "and KITTI datasets we report the absolute relative error (REL), root mean squared error (RMS) and", + "type": "text", + "cross_page": true + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 414, + 505, + 426 + ], + "spans": [ + { + "bbox": [ + 105, + 414, + 180, + 426 + ], + "score": 1.0, + "content": "accuracy metrics", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 181, + 414, + 225, + 425 + ], + "score": 0.91, + "content": "\\cdot \\delta _ { i } < 1 . 2 5 ^ { i }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 225, + 414, + 240, + 426 + ], + "score": 1.0, + "content": "for", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 240, + 415, + 282, + 426 + ], + "score": 0.88, + "content": "i \\in { 1 , 2 , 3 }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 283, + 414, + 505, + 426 + ], + "score": 1.0, + "content": "). For NYU we also report absolute error of log depths", + "type": "text", + "cross_page": true + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 426, + 506, + 438 + ], + "spans": [ + { + "bbox": [ + 106, + 426, + 135, + 437 + ], + "score": 0.84, + "content": "( l o g _ { 1 0 } )", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 135, + 426, + 506, + 438 + ], + "score": 1.0, + "content": ". For KITTI we additionally report the squared relative error (Sq-rel) and root mean squared", + "type": "text", + "cross_page": true + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 437, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 106, + 437, + 506, + 448 + ], + "score": 1.0, + "content": "error of log depths (RMS log). The predicted depth is up-sampled to the full resolution using bilinear", + "type": "text", + "cross_page": true + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 447, + 505, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 447, + 505, + 459 + ], + "score": 1.0, + "content": "interpolation before evaluation. For the indoor model we evaluate on the cropped region proposed by", + "type": "text", + "cross_page": true + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 458, + 493, + 471 + ], + "spans": [ + { + "bbox": [ + 106, + 458, + 493, + 471 + ], + "score": 1.0, + "content": "[13] and for the outdoor model the cropped region proposed by [16] as is standard in prior work.", + "type": "text", + "cross_page": true + } + ], + "index": 17 + } + ], + "index": 20.5, + "bbox_fs": [ + 105, + 699, + 506, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 108, + 71, + 506, + 665 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 71, + 506, + 665 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 71, + 506, + 665 + ], + "spans": [ + { + "bbox": [ + 108, + 71, + 506, + 665 + ], + "score": 0.961, + "type": "image", + "image_path": "5748f207563fdd78a3d6643decb1aa45b6e357689949509adf5b048ec218d0db.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 108, + 71, + 506, + 269.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 108, + 269.0, + 506, + 467.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 467.0, + 506, + 665.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 671, + 504, + 701 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 670, + 505, + 682 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 505, + 682 + ], + "score": 1.0, + "content": "Figure 12: Qualitative comparison of our model with DPT-Hybrid [50] (fine-tuned on NYU) on the NYU", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 680, + 505, + 692 + ], + "spans": [ + { + "bbox": [ + 105, + 680, + 505, + 692 + ], + "score": 1.0, + "content": "depth v2 val set. Our method infers better depth for both scene structure (walls, floors, etc.) and individual", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 691, + 327, + 703 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 327, + 703 + ], + "score": 1.0, + "content": "objects. Specific differences are highlighted with red arrows.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + } + ], + "page_idx": 18, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 299, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "19", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 108, + 71, + 506, + 665 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 71, + 506, + 665 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 71, + 506, + 665 + ], + "spans": [ + { + "bbox": [ + 108, + 71, + 506, + 665 + ], + "score": 0.961, + "type": "image", + "image_path": "5748f207563fdd78a3d6643decb1aa45b6e357689949509adf5b048ec218d0db.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 108, + 71, + 506, + 269.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 108, + 269.0, + 506, + 467.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 467.0, + 506, + 665.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 671, + 504, + 701 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 670, + 505, + 682 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 505, + 682 + ], + "score": 1.0, + "content": "Figure 12: Qualitative comparison of our model with DPT-Hybrid [50] (fine-tuned on NYU) on the NYU", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 680, + 505, + 692 + ], + "spans": [ + { + "bbox": [ + 105, + 680, + 505, + 692 + ], + "score": 1.0, + "content": "depth v2 val set. Our method infers better depth for both scene structure (walls, floors, etc.) and individual", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 691, + 327, + 703 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 327, + 703 + ], + "score": 1.0, + "content": "objects. Specific differences are highlighted with red arrows.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 128, + 31, + 484, + 635 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 128, + 31, + 484, + 635 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 128, + 31, + 484, + 634 + ], + "spans": [ + { + "bbox": [ + 128, + 31, + 484, + 634 + ], + "score": 0.93, + "type": "image", + "image_path": "091fb86ab1ddcb3861322b43f9a4d8fa56945cf2655378ebd0f5a2b5a4f74023.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 128, + 31, + 484, + 232.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 128, + 232.33333333333334, + 484, + 433.6666666666667 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 128, + 433.6666666666667, + 484, + 635.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 646, + 506, + 696 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 646, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 106, + 646, + 506, + 657 + ], + "score": 1.0, + "content": "Figure 13: Text-to-3D samples. Given a text prompt, an image is first generated using Imagen [55] (first row of", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 656, + 504, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 656, + 504, + 666 + ], + "score": 1.0, + "content": "first column), after which depth is estimated (second row of first column). Subsequently the camera is moved", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 666, + 505, + 677 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 505, + 677 + ], + "score": 1.0, + "content": "to reveal new parts of the scene which are infilled using an image completion model and our model (which", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 676, + 505, + 687 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 505, + 687 + ], + "score": 1.0, + "content": "conditions on both the incomplete depth map and the filled image). At each step, newly generated RGBD points", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 685, + 388, + 697 + ], + "spans": [ + { + "bbox": [ + 105, + 685, + 388, + 697 + ], + "score": 1.0, + "content": "are added to a global point cloud which is visualized in the rightmost column.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5 + } + ], + "index": 3.0 + } + ], + "page_idx": 19, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 298, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 298, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 128, + 31, + 484, + 635 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 128, + 31, + 484, + 635 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 128, + 31, + 484, + 634 + ], + "spans": [ + { + "bbox": [ + 128, + 31, + 484, + 634 + ], + "score": 0.93, + "type": "image", + "image_path": "091fb86ab1ddcb3861322b43f9a4d8fa56945cf2655378ebd0f5a2b5a4f74023.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 128, + 31, + 484, + 232.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 128, + 232.33333333333334, + 484, + 433.6666666666667 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 128, + 433.6666666666667, + 484, + 635.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 646, + 506, + 696 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 646, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 106, + 646, + 506, + 657 + ], + "score": 1.0, + "content": "Figure 13: Text-to-3D samples. Given a text prompt, an image is first generated using Imagen [55] (first row of", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 656, + 504, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 656, + 504, + 666 + ], + "score": 1.0, + "content": "first column), after which depth is estimated (second row of first column). Subsequently the camera is moved", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 666, + 505, + 677 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 505, + 677 + ], + "score": 1.0, + "content": "to reveal new parts of the scene which are infilled using an image completion model and our model (which", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 676, + 505, + 687 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 505, + 687 + ], + "score": 1.0, + "content": "conditions on both the incomplete depth map and the filled image). At each step, newly generated RGBD points", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 685, + 388, + 697 + ], + "spans": [ + { + "bbox": [ + 105, + 685, + 388, + 697 + ], + "score": 1.0, + "content": "are added to a global point cloud which is visualized in the rightmost column.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5 + } + ], + "index": 3.0 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 109, + 107, + 501, + 222 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 70, + 506, + 101 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 70, + 505, + 81 + ], + "spans": [ + { + "bbox": [ + 106, + 70, + 323, + 81 + ], + "score": 1.0, + "content": "Table 9: Comparison of performance on the NYU-Depth-", + "type": "text" + }, + { + "bbox": [ + 323, + 70, + 333, + 80 + ], + "score": 0.28, + "content": "\\mathbf { v } 2", + "type": "inline_equation" + }, + { + "bbox": [ + 334, + 70, + 366, + 81 + ], + "score": 1.0, + "content": "dataset.", + "type": "text" + }, + { + "bbox": [ + 366, + 70, + 375, + 80 + ], + "score": 0.55, + "content": "\\top", + "type": "inline_equation" + }, + { + "bbox": [ + 375, + 70, + 505, + 81 + ], + "score": 1.0, + "content": "indicates method uses unsupervised", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 81, + 506, + 91 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 151, + 91 + ], + "score": 1.0, + "content": "pretraining,", + "type": "text" + }, + { + "bbox": [ + 151, + 81, + 158, + 91 + ], + "score": 0.75, + "content": "^ \\dagger", + "type": "inline_equation" + }, + { + "bbox": [ + 158, + 81, + 293, + 91 + ], + "score": 1.0, + "content": "indicates supervised pretraining and", + "type": "text" + }, + { + "bbox": [ + 293, + 81, + 300, + 91 + ], + "score": 0.82, + "content": "^ \\ddag", + "type": "inline_equation" + }, + { + "bbox": [ + 300, + 81, + 506, + 91 + ], + "score": 1.0, + "content": "indicates use of auxiliary supervised depth data. Best /", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 91, + 454, + 102 + ], + "spans": [ + { + "bbox": [ + 106, + 91, + 308, + 102 + ], + "score": 1.0, + "content": "second best results are bolded / underlined respectively.", + "type": "text" + }, + { + "bbox": [ + 308, + 91, + 316, + 101 + ], + "score": 0.48, + "content": "\\downarrow :", + "type": "inline_equation" + }, + { + "bbox": [ + 316, + 91, + 386, + 102 + ], + "score": 1.0, + "content": "lower is better and", + "type": "text" + }, + { + "bbox": [ + 387, + 91, + 393, + 101 + ], + "score": 0.65, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 393, + 91, + 454, + 102 + ], + "score": 1.0, + "content": ": higher is better.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "table_body", + "bbox": [ + 109, + 107, + 501, + 222 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 109, + 107, + 501, + 222 + ], + "spans": [ + { + "bbox": [ + 109, + 107, + 501, + 222 + ], + "score": 0.983, + "html": "
MethodArchitecture81↑&↑8↑REL↓RMS↓log10
TransDepth [87]Res-50+ViT-Bt0.9000.9830.9960.1060.3650.045
DPT[50]Res-50+ViT-B†#0.9040.9880.9980.1100.3570.045
AdaBins [3]E-B5+Mini-ViTt0.9030.9840.9970.1030.3640.044
BinsFormer [35]Swin-Larget0.9250.9890.9970.0940.3300.040
PixelFormer [1]Swin-Large†0.9290.9910.9980.0900.3220.039
MIM[80]SwinV2-LT0.9490.9940.9990.0830.2870.035
AiT-P [46]SwinV2-LT0.9530.9930.9990.0760.2790.033
DDVMsamples=1Efficient U-NetT#0.9440.9860.9950.0750.3240.032
samples=2Efficient U-NetT‡0.9440.9870.9960.0740.3190.032
samples=4Efficient U-Net T‡0.9460.9870.9960.0740.3150.032
", + "type": "table", + "image_path": "7bbd209b9c7c8bf744dab6d05aa10b33ba7576ba9029ab2193927a3f4eb99f54.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 109, + 107, + 501, + 145.33333333333334 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 109, + 145.33333333333334, + 501, + 183.66666666666669 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 109, + 183.66666666666669, + 501, + 222.00000000000003 + ], + "spans": [], + "index": 5 + } + ] + } + ], + "index": 2.5 + }, + { + "type": "table", + "bbox": [ + 109, + 268, + 501, + 383 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 107, + 232, + 505, + 263 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 231, + 506, + 244 + ], + "spans": [ + { + "bbox": [ + 105, + 231, + 337, + 244 + ], + "score": 1.0, + "content": "Table 10: Comparison of performance on the KITTI dataset.", + "type": "text" + }, + { + "bbox": [ + 337, + 232, + 345, + 241 + ], + "score": 0.26, + "content": "\\top", + "type": "inline_equation" + }, + { + "bbox": [ + 346, + 231, + 506, + 244 + ], + "score": 1.0, + "content": "indicates method uses unsupervised pretrain-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 242, + 506, + 253 + ], + "spans": [ + { + "bbox": [ + 105, + 242, + 122, + 253 + ], + "score": 1.0, + "content": "ing,", + "type": "text" + }, + { + "bbox": [ + 122, + 243, + 128, + 253 + ], + "score": 0.77, + "content": "^ \\dagger", + "type": "inline_equation" + }, + { + "bbox": [ + 128, + 242, + 258, + 253 + ], + "score": 1.0, + "content": "indicates supervised pretraining and", + "type": "text" + }, + { + "bbox": [ + 258, + 243, + 264, + 253 + ], + "score": 0.81, + "content": "^ \\ddag", + "type": "inline_equation" + }, + { + "bbox": [ + 265, + 242, + 506, + 253 + ], + "score": 1.0, + "content": "indicates use of auxiliary supervised depth data. Best / second best", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 252, + 505, + 263 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 260, + 263 + ], + "score": 1.0, + "content": "results are bolded / underlined respectively.", + "type": "text" + }, + { + "bbox": [ + 261, + 253, + 269, + 262 + ], + "score": 0.27, + "content": "\\downarrow \\colon", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 252, + 336, + 263 + ], + "score": 1.0, + "content": "lower is better and", + "type": "text" + }, + { + "bbox": [ + 337, + 252, + 343, + 262 + ], + "score": 0.5, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 344, + 252, + 505, + 263 + ], + "score": 1.0, + "content": ": higher is better. E-B5: EfficientNet-B5 [73].", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "table_body", + "bbox": [ + 109, + 268, + 501, + 383 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 109, + 268, + 501, + 383 + ], + "spans": [ + { + "bbox": [ + 109, + 268, + 501, + 383 + ], + "score": 0.983, + "html": "
MethodBackbone81↑82个8↑REL↓Sq-rel ↓RMS↓RMS log↓
BTS [34]DenseNet-161†0.9560.9930.9980.0590.2452.7560.096
TransDepth [87]ResNet-50+ViT-Bt0.9560.9940.9990.0640.2522.7550.098
DPT [50]ResNet-50+ViT-Bt‡0.9590.9950.9990.06212.5730.092
AdaBins [3]E-B5+mini-ViTt0.9640.9950.9990.0580.1902.3600.088
BinsFormer [35]Swin-Larget0.9740.9970.9990.0520.1512.0980.079
PixelFormer [1]Swin-Larget0.9760.9970.9990.0510.1492.0810.077
MIM [80]SwinV2-LT0.9770.9981.0000.0500.1391.9660.075
samples=1 DDVM samples=2 samples=4Efficient U-NetT#0.9640.9940.9980.0560.3392.7000.091
Efficient U-NetT#0.9650.9940.9980.0550.3252.6600.090
Efficient U-NetT‡0.9650.9940.9980.0550.2922.6130.089
", + "type": "table", + "image_path": "670a51ac368f4fb133605b8a6f8d1ead069ab67b32ebed7234af22e605d1870c.jpg" + } + ] + } + ], + "index": 10, + "virtual_lines": [ + { + "bbox": [ + 109, + 268, + 501, + 306.3333333333333 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 109, + 306.3333333333333, + 501, + 344.66666666666663 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 109, + 344.66666666666663, + 501, + 382.99999999999994 + ], + "spans": [], + "index": 11 + } + ] + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 106, + 403, + 505, + 470 + ], + "lines": [ + { + "bbox": [ + 106, + 403, + 506, + 416 + ], + "spans": [ + { + "bbox": [ + 106, + 403, + 506, + 416 + ], + "score": 1.0, + "content": "and KITTI datasets we report the absolute relative error (REL), root mean squared error (RMS) and", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 414, + 505, + 426 + ], + "spans": [ + { + "bbox": [ + 105, + 414, + 180, + 426 + ], + "score": 1.0, + "content": "accuracy metrics", + "type": "text" + }, + { + "bbox": [ + 181, + 414, + 225, + 425 + ], + "score": 0.91, + "content": "\\cdot \\delta _ { i } < 1 . 2 5 ^ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 414, + 240, + 426 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 240, + 415, + 282, + 426 + ], + "score": 0.88, + "content": "i \\in { 1 , 2 , 3 }", + "type": "inline_equation" + }, + { + "bbox": [ + 283, + 414, + 505, + 426 + ], + "score": 1.0, + "content": "). For NYU we also report absolute error of log depths", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 426, + 506, + 438 + ], + "spans": [ + { + "bbox": [ + 106, + 426, + 135, + 437 + ], + "score": 0.84, + "content": "( l o g _ { 1 0 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 135, + 426, + 506, + 438 + ], + "score": 1.0, + "content": ". For KITTI we additionally report the squared relative error (Sq-rel) and root mean squared", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 437, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 106, + 437, + 506, + 448 + ], + "score": 1.0, + "content": "error of log depths (RMS log). The predicted depth is up-sampled to the full resolution using bilinear", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 447, + 505, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 447, + 505, + 459 + ], + "score": 1.0, + "content": "interpolation before evaluation. For the indoor model we evaluate on the cropped region proposed by", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 458, + 493, + 471 + ], + "spans": [ + { + "bbox": [ + 106, + 458, + 493, + 471 + ], + "score": 1.0, + "content": "[13] and for the outdoor model the cropped region proposed by [16] as is standard in prior work.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 14.5 + }, + { + "type": "title", + "bbox": [ + 107, + 484, + 177, + 498 + ], + "lines": [ + { + "bbox": [ + 104, + 483, + 179, + 500 + ], + "spans": [ + { + "bbox": [ + 104, + 483, + 179, + 500 + ], + "score": 1.0, + "content": "E Ablations", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 107, + 504, + 505, + 570 + ], + "lines": [ + { + "bbox": [ + 105, + 503, + 506, + 516 + ], + "spans": [ + { + "bbox": [ + 105, + 503, + 229, + 516 + ], + "score": 1.0, + "content": "Tables 11 and 12 show that an", + "type": "text" + }, + { + "bbox": [ + 229, + 505, + 241, + 515 + ], + "score": 0.88, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 242, + 503, + 506, + 516 + ], + "score": 1.0, + "content": "loss in training the diffusion model performs much better than an", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 107, + 515, + 506, + 528 + ], + "spans": [ + { + "bbox": [ + 107, + 515, + 119, + 526 + ], + "score": 0.87, + "content": "L _ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 119, + 515, + 506, + 528 + ], + "score": 1.0, + "content": "loss for monocular depth estimation on NYU and KITTI. Tables 13 and 14 show the effectiveness", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 526, + 505, + 538 + ], + "spans": [ + { + "bbox": [ + 106, + 526, + 505, + 538 + ], + "score": 1.0, + "content": "of Palette-style [54] self-supervised pretraining for monocular depth estimation on NYU and KITTI", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 537, + 505, + 549 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 549 + ], + "score": 1.0, + "content": "respectively. All results use a single sample. Because these findings are reasonable and expected to", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 547, + 506, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 506, + 560 + ], + "score": 1.0, + "content": "generalize to other dense vision tasks, we do not further ablate them for optical flow estimation for", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 559, + 186, + 572 + ], + "spans": [ + { + "bbox": [ + 106, + 559, + 186, + 572 + ], + "score": 1.0, + "content": "compute efficiency.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21.5 + }, + { + "type": "table", + "bbox": [ + 189, + 594, + 418, + 637 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 162, + 580, + 448, + 591 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 160, + 578, + 450, + 592 + ], + "spans": [ + { + "bbox": [ + 160, + 578, + 450, + 592 + ], + "score": 1.0, + "content": "Table 11: Ablation for the choice of loss function on the NYU depth v2 dataset.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "table_body", + "bbox": [ + 189, + 594, + 418, + 637 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 189, + 594, + 418, + 637 + ], + "spans": [ + { + "bbox": [ + 189, + 594, + 418, + 637 + ], + "score": 0.968, + "html": "
81个8↑8↑REL↓RMS↓log10↓
L20.9320.9810.9940.0850.3490.037
L10.9440.9860.9950.0750.3240.032
", + "type": "table", + "image_path": "e69ca46ab39302668467ea923046c4efe4bc28e75b1aada2afc93ad652c3aabb.jpg" + } + ] + } + ], + "index": 27, + "virtual_lines": [ + { + "bbox": [ + 189, + 594, + 418, + 608.3333333333334 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 189, + 608.3333333333334, + 418, + 622.6666666666667 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 189, + 622.6666666666667, + 418, + 637.0000000000001 + ], + "spans": [], + "index": 28 + } + ] + } + ], + "index": 26.0 + }, + { + "type": "table", + "bbox": [ + 150, + 668, + 458, + 711 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 175, + 654, + 435, + 664 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 175, + 653, + 435, + 665 + ], + "spans": [ + { + "bbox": [ + 175, + 653, + 435, + 665 + ], + "score": 1.0, + "content": "Table 12: Ablation for the choice of loss function on the KITTI dataset.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "table_body", + "bbox": [ + 150, + 668, + 458, + 711 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 150, + 668, + 458, + 711 + ], + "spans": [ + { + "bbox": [ + 150, + 668, + 458, + 711 + ], + "score": 0.976, + "html": "
81↑82↑83↑REL↓Sq-rel↓RMS↓RMS log↓
L20.9540.9930.9980.0650.3212.7730.099
L10.9640.9940.9980.0560.3392.7000.091
", + "type": "table", + "image_path": "e7b691ad4eda65dcc2fd46f162b06525afcc496ad3050f834634206f6d6746c0.jpg" + } + ] + } + ], + "index": 31, + "virtual_lines": [ + { + "bbox": [ + 150, + 668, + 458, + 682.3333333333334 + ], + "spans": [], + "index": 30 + }, + { + "bbox": [ + 150, + 682.3333333333334, + 458, + 696.6666666666667 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 150, + 696.6666666666667, + 458, + 711.0000000000001 + ], + "spans": [], + "index": 32 + } + ] + } + ], + "index": 30.0 + } + ], + "page_idx": 20, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 310, + 750 + ], + "lines": [ + { + "bbox": [ + 298, + 740, + 312, + 754 + ], + "spans": [ + { + "bbox": [ + 298, + 740, + 312, + 754 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 14 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 109, + 107, + 501, + 222 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 70, + 506, + 101 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 70, + 505, + 81 + ], + "spans": [ + { + "bbox": [ + 106, + 70, + 323, + 81 + ], + "score": 1.0, + "content": "Table 9: Comparison of performance on the NYU-Depth-", + "type": "text" + }, + { + "bbox": [ + 323, + 70, + 333, + 80 + ], + "score": 0.28, + "content": "\\mathbf { v } 2", + "type": "inline_equation" + }, + { + "bbox": [ + 334, + 70, + 366, + 81 + ], + "score": 1.0, + "content": "dataset.", + "type": "text" + }, + { + "bbox": [ + 366, + 70, + 375, + 80 + ], + "score": 0.55, + "content": "\\top", + "type": "inline_equation" + }, + { + "bbox": [ + 375, + 70, + 505, + 81 + ], + "score": 1.0, + "content": "indicates method uses unsupervised", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 81, + 506, + 91 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 151, + 91 + ], + "score": 1.0, + "content": "pretraining,", + "type": "text" + }, + { + "bbox": [ + 151, + 81, + 158, + 91 + ], + "score": 0.75, + "content": "^ \\dagger", + "type": "inline_equation" + }, + { + "bbox": [ + 158, + 81, + 293, + 91 + ], + "score": 1.0, + "content": "indicates supervised pretraining and", + "type": "text" + }, + { + "bbox": [ + 293, + 81, + 300, + 91 + ], + "score": 0.82, + "content": "^ \\ddag", + "type": "inline_equation" + }, + { + "bbox": [ + 300, + 81, + 506, + 91 + ], + "score": 1.0, + "content": "indicates use of auxiliary supervised depth data. Best /", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 91, + 454, + 102 + ], + "spans": [ + { + "bbox": [ + 106, + 91, + 308, + 102 + ], + "score": 1.0, + "content": "second best results are bolded / underlined respectively.", + "type": "text" + }, + { + "bbox": [ + 308, + 91, + 316, + 101 + ], + "score": 0.48, + "content": "\\downarrow :", + "type": "inline_equation" + }, + { + "bbox": [ + 316, + 91, + 386, + 102 + ], + "score": 1.0, + "content": "lower is better and", + "type": "text" + }, + { + "bbox": [ + 387, + 91, + 393, + 101 + ], + "score": 0.65, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 393, + 91, + 454, + 102 + ], + "score": 1.0, + "content": ": higher is better.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "table_body", + "bbox": [ + 109, + 107, + 501, + 222 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 109, + 107, + 501, + 222 + ], + "spans": [ + { + "bbox": [ + 109, + 107, + 501, + 222 + ], + "score": 0.983, + "html": "
MethodArchitecture81↑&↑8↑REL↓RMS↓log10
TransDepth [87]Res-50+ViT-Bt0.9000.9830.9960.1060.3650.045
DPT[50]Res-50+ViT-B†#0.9040.9880.9980.1100.3570.045
AdaBins [3]E-B5+Mini-ViTt0.9030.9840.9970.1030.3640.044
BinsFormer [35]Swin-Larget0.9250.9890.9970.0940.3300.040
PixelFormer [1]Swin-Large†0.9290.9910.9980.0900.3220.039
MIM[80]SwinV2-LT0.9490.9940.9990.0830.2870.035
AiT-P [46]SwinV2-LT0.9530.9930.9990.0760.2790.033
DDVMsamples=1Efficient U-NetT#0.9440.9860.9950.0750.3240.032
samples=2Efficient U-NetT‡0.9440.9870.9960.0740.3190.032
samples=4Efficient U-Net T‡0.9460.9870.9960.0740.3150.032
", + "type": "table", + "image_path": "7bbd209b9c7c8bf744dab6d05aa10b33ba7576ba9029ab2193927a3f4eb99f54.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 109, + 107, + 501, + 145.33333333333334 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 109, + 145.33333333333334, + 501, + 183.66666666666669 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 109, + 183.66666666666669, + 501, + 222.00000000000003 + ], + "spans": [], + "index": 5 + } + ] + } + ], + "index": 2.5 + }, + { + "type": "table", + "bbox": [ + 109, + 268, + 501, + 383 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 107, + 232, + 505, + 263 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 231, + 506, + 244 + ], + "spans": [ + { + "bbox": [ + 105, + 231, + 337, + 244 + ], + "score": 1.0, + "content": "Table 10: Comparison of performance on the KITTI dataset.", + "type": "text" + }, + { + "bbox": [ + 337, + 232, + 345, + 241 + ], + "score": 0.26, + "content": "\\top", + "type": "inline_equation" + }, + { + "bbox": [ + 346, + 231, + 506, + 244 + ], + "score": 1.0, + "content": "indicates method uses unsupervised pretrain-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 242, + 506, + 253 + ], + "spans": [ + { + "bbox": [ + 105, + 242, + 122, + 253 + ], + "score": 1.0, + "content": "ing,", + "type": "text" + }, + { + "bbox": [ + 122, + 243, + 128, + 253 + ], + "score": 0.77, + "content": "^ \\dagger", + "type": "inline_equation" + }, + { + "bbox": [ + 128, + 242, + 258, + 253 + ], + "score": 1.0, + "content": "indicates supervised pretraining and", + "type": "text" + }, + { + "bbox": [ + 258, + 243, + 264, + 253 + ], + "score": 0.81, + "content": "^ \\ddag", + "type": "inline_equation" + }, + { + "bbox": [ + 265, + 242, + 506, + 253 + ], + "score": 1.0, + "content": "indicates use of auxiliary supervised depth data. Best / second best", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 252, + 505, + 263 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 260, + 263 + ], + "score": 1.0, + "content": "results are bolded / underlined respectively.", + "type": "text" + }, + { + "bbox": [ + 261, + 253, + 269, + 262 + ], + "score": 0.27, + "content": "\\downarrow \\colon", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 252, + 336, + 263 + ], + "score": 1.0, + "content": "lower is better and", + "type": "text" + }, + { + "bbox": [ + 337, + 252, + 343, + 262 + ], + "score": 0.5, + "content": "\\uparrow", + "type": "inline_equation" + }, + { + "bbox": [ + 344, + 252, + 505, + 263 + ], + "score": 1.0, + "content": ": higher is better. E-B5: EfficientNet-B5 [73].", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "table_body", + "bbox": [ + 109, + 268, + 501, + 383 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 109, + 268, + 501, + 383 + ], + "spans": [ + { + "bbox": [ + 109, + 268, + 501, + 383 + ], + "score": 0.983, + "html": "
MethodBackbone81↑82个8↑REL↓Sq-rel ↓RMS↓RMS log↓
BTS [34]DenseNet-161†0.9560.9930.9980.0590.2452.7560.096
TransDepth [87]ResNet-50+ViT-Bt0.9560.9940.9990.0640.2522.7550.098
DPT [50]ResNet-50+ViT-Bt‡0.9590.9950.9990.06212.5730.092
AdaBins [3]E-B5+mini-ViTt0.9640.9950.9990.0580.1902.3600.088
BinsFormer [35]Swin-Larget0.9740.9970.9990.0520.1512.0980.079
PixelFormer [1]Swin-Larget0.9760.9970.9990.0510.1492.0810.077
MIM [80]SwinV2-LT0.9770.9981.0000.0500.1391.9660.075
samples=1 DDVM samples=2 samples=4Efficient U-NetT#0.9640.9940.9980.0560.3392.7000.091
Efficient U-NetT#0.9650.9940.9980.0550.3252.6600.090
Efficient U-NetT‡0.9650.9940.9980.0550.2922.6130.089
", + "type": "table", + "image_path": "670a51ac368f4fb133605b8a6f8d1ead069ab67b32ebed7234af22e605d1870c.jpg" + } + ] + } + ], + "index": 10, + "virtual_lines": [ + { + "bbox": [ + 109, + 268, + 501, + 306.3333333333333 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 109, + 306.3333333333333, + 501, + 344.66666666666663 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 109, + 344.66666666666663, + 501, + 382.99999999999994 + ], + "spans": [], + "index": 11 + } + ] + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 106, + 403, + 505, + 470 + ], + "lines": [], + "index": 14.5, + "bbox_fs": [ + 105, + 403, + 506, + 471 + ], + "lines_deleted": true + }, + { + "type": "title", + "bbox": [ + 107, + 484, + 177, + 498 + ], + "lines": [ + { + "bbox": [ + 104, + 483, + 179, + 500 + ], + "spans": [ + { + "bbox": [ + 104, + 483, + 179, + 500 + ], + "score": 1.0, + "content": "E Ablations", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 107, + 504, + 505, + 570 + ], + "lines": [ + { + "bbox": [ + 105, + 503, + 506, + 516 + ], + "spans": [ + { + "bbox": [ + 105, + 503, + 229, + 516 + ], + "score": 1.0, + "content": "Tables 11 and 12 show that an", + "type": "text" + }, + { + "bbox": [ + 229, + 505, + 241, + 515 + ], + "score": 0.88, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 242, + 503, + 506, + 516 + ], + "score": 1.0, + "content": "loss in training the diffusion model performs much better than an", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 107, + 515, + 506, + 528 + ], + "spans": [ + { + "bbox": [ + 107, + 515, + 119, + 526 + ], + "score": 0.87, + "content": "L _ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 119, + 515, + 506, + 528 + ], + "score": 1.0, + "content": "loss for monocular depth estimation on NYU and KITTI. Tables 13 and 14 show the effectiveness", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 526, + 505, + 538 + ], + "spans": [ + { + "bbox": [ + 106, + 526, + 505, + 538 + ], + "score": 1.0, + "content": "of Palette-style [54] self-supervised pretraining for monocular depth estimation on NYU and KITTI", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 537, + 505, + 549 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 549 + ], + "score": 1.0, + "content": "respectively. All results use a single sample. Because these findings are reasonable and expected to", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 547, + 506, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 506, + 560 + ], + "score": 1.0, + "content": "generalize to other dense vision tasks, we do not further ablate them for optical flow estimation for", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 559, + 186, + 572 + ], + "spans": [ + { + "bbox": [ + 106, + 559, + 186, + 572 + ], + "score": 1.0, + "content": "compute efficiency.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21.5, + "bbox_fs": [ + 105, + 503, + 506, + 572 + ] + }, + { + "type": "table", + "bbox": [ + 189, + 594, + 418, + 637 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 162, + 580, + 448, + 591 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 160, + 578, + 450, + 592 + ], + "spans": [ + { + "bbox": [ + 160, + 578, + 450, + 592 + ], + "score": 1.0, + "content": "Table 11: Ablation for the choice of loss function on the NYU depth v2 dataset.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "table_body", + "bbox": [ + 189, + 594, + 418, + 637 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 189, + 594, + 418, + 637 + ], + "spans": [ + { + "bbox": [ + 189, + 594, + 418, + 637 + ], + "score": 0.968, + "html": "
81个8↑8↑REL↓RMS↓log10↓
L20.9320.9810.9940.0850.3490.037
L10.9440.9860.9950.0750.3240.032
", + "type": "table", + "image_path": "e69ca46ab39302668467ea923046c4efe4bc28e75b1aada2afc93ad652c3aabb.jpg" + } + ] + } + ], + "index": 27, + "virtual_lines": [ + { + "bbox": [ + 189, + 594, + 418, + 608.3333333333334 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 189, + 608.3333333333334, + 418, + 622.6666666666667 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 189, + 622.6666666666667, + 418, + 637.0000000000001 + ], + "spans": [], + "index": 28 + } + ] + } + ], + "index": 26.0 + }, + { + "type": "table", + "bbox": [ + 150, + 668, + 458, + 711 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 175, + 654, + 435, + 664 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 175, + 653, + 435, + 665 + ], + "spans": [ + { + "bbox": [ + 175, + 653, + 435, + 665 + ], + "score": 1.0, + "content": "Table 12: Ablation for the choice of loss function on the KITTI dataset.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "table_body", + "bbox": [ + 150, + 668, + 458, + 711 + ], + "group_id": 3, + "lines": [ + { + "bbox": [ + 150, + 668, + 458, + 711 + ], + "spans": [ + { + "bbox": [ + 150, + 668, + 458, + 711 + ], + "score": 0.976, + "html": "
81↑82↑83↑REL↓Sq-rel↓RMS↓RMS log↓
L20.9540.9930.9980.0650.3212.7730.099
L10.9640.9940.9980.0560.3392.7000.091
", + "type": "table", + "image_path": "e7b691ad4eda65dcc2fd46f162b06525afcc496ad3050f834634206f6d6746c0.jpg" + } + ] + } + ], + "index": 31, + "virtual_lines": [ + { + "bbox": [ + 150, + 668, + 458, + 682.3333333333334 + ], + "spans": [], + "index": 30 + }, + { + "bbox": [ + 150, + 682.3333333333334, + 458, + 696.6666666666667 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 150, + 696.6666666666667, + 458, + 711.0000000000001 + ], + "spans": [], + "index": 32 + } + ] + } + ], + "index": 30.0 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 122, + 84, + 486, + 127 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 160, + 70, + 449, + 81 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 159, + 68, + 451, + 83 + ], + "spans": [ + { + "bbox": [ + 159, + 68, + 451, + 83 + ], + "score": 1.0, + "content": "Table 13: Ablation for self-supervised pretraining on the NYU depth v2 dataset.", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table_body", + "bbox": [ + 122, + 84, + 486, + 127 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 122, + 84, + 486, + 127 + ], + "spans": [ + { + "bbox": [ + 122, + 84, + 486, + 127 + ], + "score": 0.973, + "html": "
81↑8↑8↑REL↓RMS↓log10↓
No self-supervised pre-training0.9360.9800.9920.0810.3520.035
With self-supervised pre-training0.9440.9860.9950.0750.3240.032
", + "type": "table", + "image_path": "e8a9a17f24c2510ccd024d8734b01deffaf76f3a5725f5487e22fec36fb3f77f.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 122, + 84, + 486, + 98.33333333333333 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 122, + 98.33333333333333, + 486, + 112.66666666666666 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 122, + 112.66666666666666, + 486, + 126.99999999999999 + ], + "spans": [], + "index": 3 + } + ] + } + ], + "index": 1.0 + }, + { + "type": "table", + "bbox": [ + 124, + 152, + 485, + 195 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 163, + 138, + 446, + 149 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 163, + 136, + 447, + 151 + ], + "spans": [ + { + "bbox": [ + 163, + 136, + 447, + 151 + ], + "score": 1.0, + "content": "Table 14: Ablation for self-supervised pretraining on the KITTI depth dataset.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "table_body", + "bbox": [ + 124, + 152, + 485, + 195 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 124, + 152, + 485, + 195 + ], + "spans": [ + { + "bbox": [ + 124, + 152, + 485, + 195 + ], + "score": 0.974, + "html": "
81&28↑REL↓Sq-rel↓RMS↓RMS log↓
No self-supervised pre-training0.9520.990(0.9970.0640.389 2.9980.104
With self-supervised pre-training0.9650.9940.9980.0550.3322.6960.091
", + "type": "table", + "image_path": "76f0404b48fe4d284a3b13ffe5eb3693f46ecbc654971da4fde957257764a66e.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 124, + 152, + 485, + 166.33333333333334 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 124, + 166.33333333333334, + 485, + 180.66666666666669 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 124, + 180.66666666666669, + 485, + 195.00000000000003 + ], + "spans": [], + "index": 7 + } + ] + } + ], + "index": 5.0 + }, + { + "type": "title", + "bbox": [ + 106, + 214, + 308, + 227 + ], + "lines": [ + { + "bbox": [ + 104, + 212, + 308, + 230 + ], + "spans": [ + { + "bbox": [ + 104, + 212, + 308, + 230 + ], + "score": 1.0, + "content": "F Coarse-to-fine refinement for depth", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 106, + 233, + 505, + 288 + ], + "lines": [ + { + "bbox": [ + 106, + 234, + 505, + 245 + ], + "spans": [ + { + "bbox": [ + 106, + 234, + 505, + 245 + ], + "score": 1.0, + "content": "Figure 14 demonstrates performance of coarse-to-fine refinement on the NYU depth v2 dataset. While", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 244, + 505, + 257 + ], + "spans": [ + { + "bbox": [ + 105, + 244, + 505, + 257 + ], + "score": 1.0, + "content": "refinement improves fine-scale details in the estimated depth maps, the qualitative improvements are", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 255, + 505, + 267 + ], + "spans": [ + { + "bbox": [ + 106, + 255, + 505, + 267 + ], + "score": 1.0, + "content": "small and we do not find significant quantitative improvements. Hence the results reported in this", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 264, + 507, + 280 + ], + "spans": [ + { + "bbox": [ + 104, + 264, + 507, + 280 + ], + "score": 1.0, + "content": "work do not use coarse-to-fine refinement for depth estimation. Further work is needed to develop a", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 277, + 403, + 290 + ], + "spans": [ + { + "bbox": [ + 106, + 277, + 403, + 290 + ], + "score": 1.0, + "content": "coarse-to-fine algorithm capable of more robust gains in depth estimation.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 11 + }, + { + "type": "image", + "bbox": [ + 107, + 300, + 506, + 457 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 300, + 506, + 457 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 300, + 506, + 457 + ], + "spans": [ + { + "bbox": [ + 107, + 300, + 506, + 457 + ], + "score": 0.899, + "type": "image", + "image_path": "dd13e125d12e37e3ec51d61fa7228d17a03f316bf3785e371ea9f43c2429e3cf.jpg" + } + ] + } + ], + "index": 15, + "virtual_lines": [ + { + "bbox": [ + 107, + 300, + 506, + 352.3333333333333 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 107, + 352.3333333333333, + 506, + 404.66666666666663 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 107, + 404.66666666666663, + 506, + 456.99999999999994 + ], + "spans": [], + "index": 16 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 463, + 504, + 484 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 464, + 505, + 475 + ], + "spans": [ + { + "bbox": [ + 106, + 464, + 505, + 475 + ], + "score": 1.0, + "content": "Figure 14: Samples with coarse-to-fine refinement on the NYU depth v2 dataset. We find that refinement", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 473, + 459, + 485 + ], + "spans": [ + { + "bbox": [ + 106, + 473, + 459, + 485 + ], + "score": 1.0, + "content": "adds sharpness and detail to the depth estimation but does not provide quantitative improvements.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5 + } + ], + "index": 16.25 + }, + { + "type": "title", + "bbox": [ + 107, + 500, + 374, + 514 + ], + "lines": [ + { + "bbox": [ + 105, + 499, + 376, + 516 + ], + "spans": [ + { + "bbox": [ + 105, + 499, + 376, + 516 + ], + "score": 1.0, + "content": "G Coarse-to-fine optical flow refinement for RAFT", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 106, + 520, + 505, + 629 + ], + "lines": [ + { + "bbox": [ + 106, + 520, + 505, + 532 + ], + "spans": [ + { + "bbox": [ + 106, + 520, + 505, + 532 + ], + "score": 1.0, + "content": "For a fair comparison with optical flow estimation, we also apply our coarse-to-fine refinement", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 531, + 505, + 543 + ], + "spans": [ + { + "bbox": [ + 105, + 531, + 505, + 543 + ], + "score": 1.0, + "content": "scheme to RAFT [74], to determine whether our performance gains translate to RAFT as well. We", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 542, + 506, + 554 + ], + "spans": [ + { + "bbox": [ + 106, + 542, + 261, + 554 + ], + "score": 1.0, + "content": "first estimate flow at a low resolution,", + "type": "text" + }, + { + "bbox": [ + 261, + 542, + 304, + 552 + ], + "score": 0.9, + "content": "3 2 0 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 305, + 542, + 506, + 554 + ], + "score": 1.0, + "content": ", upsample the low-resolution flow to the original", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 552, + 506, + 565 + ], + "spans": [ + { + "bbox": [ + 105, + 552, + 324, + 565 + ], + "score": 1.0, + "content": "resolution, divide original-resolution input images into", + "type": "text" + }, + { + "bbox": [ + 325, + 553, + 348, + 563 + ], + "score": 0.9, + "content": "2 \\times 5", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 552, + 459, + 565 + ], + "score": 1.0, + "content": "overlapping patches of size", + "type": "text" + }, + { + "bbox": [ + 459, + 553, + 502, + 563 + ], + "score": 0.9, + "content": "3 2 0 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 552, + 506, + 565 + ], + "score": 1.0, + "content": ",", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 564, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 106, + 564, + 506, + 577 + ], + "score": 1.0, + "content": "then estimate flow on the cropped patches using the upsampled flow field as the initial guess for the", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 104, + 574, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 104, + 574, + 505, + 588 + ], + "score": 1.0, + "content": "recurrent refinement (12 steps in total) of RAFT [74]. After estimating flow of each patch, we merge", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 585, + 506, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 506, + 597 + ], + "score": 1.0, + "content": "them using weighted masks [28]. Table 15 reports the result. Unlike our diffusion-based method,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 596, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 106, + 596, + 506, + 609 + ], + "score": 1.0, + "content": "the coarse-to-fine scheme actually hurts the accuracy of RAFT on Sintel Clean and KITTI and only", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 608, + 506, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 506, + 619 + ], + "score": 1.0, + "content": "marginally improves the accuracy on Sintel Final. Further exploration into better approaches for", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 618, + 423, + 630 + ], + "spans": [ + { + "bbox": [ + 106, + 618, + 423, + 630 + ], + "score": 1.0, + "content": "coarse-to-fine refinement for RAFT is warranted. We leave that to future work.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 24.5 + }, + { + "type": "title", + "bbox": [ + 108, + 645, + 282, + 659 + ], + "lines": [ + { + "bbox": [ + 105, + 644, + 283, + 660 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 283, + 660 + ], + "score": 1.0, + "content": "H Training and inference details", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "title", + "bbox": [ + 107, + 665, + 187, + 676 + ], + "lines": [ + { + "bbox": [ + 105, + 663, + 189, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 663, + 189, + 678 + ], + "score": 1.0, + "content": "H.1 Architecture", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 108, + 678, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 678, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 678, + 505, + 690 + ], + "score": 1.0, + "content": "UNet. The predominant architecture for diffusion models is the U-Net developed for the DDPM", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 689, + 506, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 689, + 494, + 701 + ], + "score": 1.0, + "content": "model [21], and later improved in several respects [11, 45, 63]. Here we adapt the Efficient", + "type": "text" + }, + { + "bbox": [ + 494, + 689, + 502, + 699 + ], + "score": 0.43, + "content": "U _ { ☉ }", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 689, + 506, + 701 + ], + "score": 1.0, + "content": "-", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 700, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 505, + 712 + ], + "score": 1.0, + "content": "Net architecture that was developed for Imagen [55]. It is more efficient that the U-Nets used in prior", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "score": 1.0, + "content": "work owing to the use of fewer self-attention layers, fewer parameters and less computation at higher", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 33.5 + } + ], + "page_idx": 21, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 298, + 740, + 313, + 755 + ], + "spans": [ + { + "bbox": [ + 298, + 740, + 313, + 755 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 122, + 84, + 486, + 127 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 160, + 70, + 449, + 81 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 159, + 68, + 451, + 83 + ], + "spans": [ + { + "bbox": [ + 159, + 68, + 451, + 83 + ], + "score": 1.0, + "content": "Table 13: Ablation for self-supervised pretraining on the NYU depth v2 dataset.", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table_body", + "bbox": [ + 122, + 84, + 486, + 127 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 122, + 84, + 486, + 127 + ], + "spans": [ + { + "bbox": [ + 122, + 84, + 486, + 127 + ], + "score": 0.973, + "html": "
81↑8↑8↑REL↓RMS↓log10↓
No self-supervised pre-training0.9360.9800.9920.0810.3520.035
With self-supervised pre-training0.9440.9860.9950.0750.3240.032
", + "type": "table", + "image_path": "e8a9a17f24c2510ccd024d8734b01deffaf76f3a5725f5487e22fec36fb3f77f.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 122, + 84, + 486, + 98.33333333333333 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 122, + 98.33333333333333, + 486, + 112.66666666666666 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 122, + 112.66666666666666, + 486, + 126.99999999999999 + ], + "spans": [], + "index": 3 + } + ] + } + ], + "index": 1.0 + }, + { + "type": "table", + "bbox": [ + 124, + 152, + 485, + 195 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 163, + 138, + 446, + 149 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 163, + 136, + 447, + 151 + ], + "spans": [ + { + "bbox": [ + 163, + 136, + 447, + 151 + ], + "score": 1.0, + "content": "Table 14: Ablation for self-supervised pretraining on the KITTI depth dataset.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "table_body", + "bbox": [ + 124, + 152, + 485, + 195 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 124, + 152, + 485, + 195 + ], + "spans": [ + { + "bbox": [ + 124, + 152, + 485, + 195 + ], + "score": 0.974, + "html": "
81&28↑REL↓Sq-rel↓RMS↓RMS log↓
No self-supervised pre-training0.9520.990(0.9970.0640.389 2.9980.104
With self-supervised pre-training0.9650.9940.9980.0550.3322.6960.091
", + "type": "table", + "image_path": "76f0404b48fe4d284a3b13ffe5eb3693f46ecbc654971da4fde957257764a66e.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 124, + 152, + 485, + 166.33333333333334 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 124, + 166.33333333333334, + 485, + 180.66666666666669 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 124, + 180.66666666666669, + 485, + 195.00000000000003 + ], + "spans": [], + "index": 7 + } + ] + } + ], + "index": 5.0 + }, + { + "type": "title", + "bbox": [ + 106, + 214, + 308, + 227 + ], + "lines": [ + { + "bbox": [ + 104, + 212, + 308, + 230 + ], + "spans": [ + { + "bbox": [ + 104, + 212, + 308, + 230 + ], + "score": 1.0, + "content": "F Coarse-to-fine refinement for depth", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 106, + 233, + 505, + 288 + ], + "lines": [ + { + "bbox": [ + 106, + 234, + 505, + 245 + ], + "spans": [ + { + "bbox": [ + 106, + 234, + 505, + 245 + ], + "score": 1.0, + "content": "Figure 14 demonstrates performance of coarse-to-fine refinement on the NYU depth v2 dataset. While", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 244, + 505, + 257 + ], + "spans": [ + { + "bbox": [ + 105, + 244, + 505, + 257 + ], + "score": 1.0, + "content": "refinement improves fine-scale details in the estimated depth maps, the qualitative improvements are", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 255, + 505, + 267 + ], + "spans": [ + { + "bbox": [ + 106, + 255, + 505, + 267 + ], + "score": 1.0, + "content": "small and we do not find significant quantitative improvements. Hence the results reported in this", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 264, + 507, + 280 + ], + "spans": [ + { + "bbox": [ + 104, + 264, + 507, + 280 + ], + "score": 1.0, + "content": "work do not use coarse-to-fine refinement for depth estimation. Further work is needed to develop a", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 277, + 403, + 290 + ], + "spans": [ + { + "bbox": [ + 106, + 277, + 403, + 290 + ], + "score": 1.0, + "content": "coarse-to-fine algorithm capable of more robust gains in depth estimation.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 11, + "bbox_fs": [ + 104, + 234, + 507, + 290 + ] + }, + { + "type": "image", + "bbox": [ + 107, + 300, + 506, + 457 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 300, + 506, + 457 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 300, + 506, + 457 + ], + "spans": [ + { + "bbox": [ + 107, + 300, + 506, + 457 + ], + "score": 0.899, + "type": "image", + "image_path": "dd13e125d12e37e3ec51d61fa7228d17a03f316bf3785e371ea9f43c2429e3cf.jpg" + } + ] + } + ], + "index": 15, + "virtual_lines": [ + { + "bbox": [ + 107, + 300, + 506, + 352.3333333333333 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 107, + 352.3333333333333, + 506, + 404.66666666666663 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 107, + 404.66666666666663, + 506, + 456.99999999999994 + ], + "spans": [], + "index": 16 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 463, + 504, + 484 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 464, + 505, + 475 + ], + "spans": [ + { + "bbox": [ + 106, + 464, + 505, + 475 + ], + "score": 1.0, + "content": "Figure 14: Samples with coarse-to-fine refinement on the NYU depth v2 dataset. We find that refinement", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 473, + 459, + 485 + ], + "spans": [ + { + "bbox": [ + 106, + 473, + 459, + 485 + ], + "score": 1.0, + "content": "adds sharpness and detail to the depth estimation but does not provide quantitative improvements.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5 + } + ], + "index": 16.25 + }, + { + "type": "title", + "bbox": [ + 107, + 500, + 374, + 514 + ], + "lines": [ + { + "bbox": [ + 105, + 499, + 376, + 516 + ], + "spans": [ + { + "bbox": [ + 105, + 499, + 376, + 516 + ], + "score": 1.0, + "content": "G Coarse-to-fine optical flow refinement for RAFT", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 106, + 520, + 505, + 629 + ], + "lines": [ + { + "bbox": [ + 106, + 520, + 505, + 532 + ], + "spans": [ + { + "bbox": [ + 106, + 520, + 505, + 532 + ], + "score": 1.0, + "content": "For a fair comparison with optical flow estimation, we also apply our coarse-to-fine refinement", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 531, + 505, + 543 + ], + "spans": [ + { + "bbox": [ + 105, + 531, + 505, + 543 + ], + "score": 1.0, + "content": "scheme to RAFT [74], to determine whether our performance gains translate to RAFT as well. We", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 542, + 506, + 554 + ], + "spans": [ + { + "bbox": [ + 106, + 542, + 261, + 554 + ], + "score": 1.0, + "content": "first estimate flow at a low resolution,", + "type": "text" + }, + { + "bbox": [ + 261, + 542, + 304, + 552 + ], + "score": 0.9, + "content": "3 2 0 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 305, + 542, + 506, + 554 + ], + "score": 1.0, + "content": ", upsample the low-resolution flow to the original", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 552, + 506, + 565 + ], + "spans": [ + { + "bbox": [ + 105, + 552, + 324, + 565 + ], + "score": 1.0, + "content": "resolution, divide original-resolution input images into", + "type": "text" + }, + { + "bbox": [ + 325, + 553, + 348, + 563 + ], + "score": 0.9, + "content": "2 \\times 5", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 552, + 459, + 565 + ], + "score": 1.0, + "content": "overlapping patches of size", + "type": "text" + }, + { + "bbox": [ + 459, + 553, + 502, + 563 + ], + "score": 0.9, + "content": "3 2 0 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 552, + 506, + 565 + ], + "score": 1.0, + "content": ",", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 564, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 106, + 564, + 506, + 577 + ], + "score": 1.0, + "content": "then estimate flow on the cropped patches using the upsampled flow field as the initial guess for the", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 104, + 574, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 104, + 574, + 505, + 588 + ], + "score": 1.0, + "content": "recurrent refinement (12 steps in total) of RAFT [74]. After estimating flow of each patch, we merge", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 585, + 506, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 506, + 597 + ], + "score": 1.0, + "content": "them using weighted masks [28]. Table 15 reports the result. Unlike our diffusion-based method,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 596, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 106, + 596, + 506, + 609 + ], + "score": 1.0, + "content": "the coarse-to-fine scheme actually hurts the accuracy of RAFT on Sintel Clean and KITTI and only", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 608, + 506, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 506, + 619 + ], + "score": 1.0, + "content": "marginally improves the accuracy on Sintel Final. Further exploration into better approaches for", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 618, + 423, + 630 + ], + "spans": [ + { + "bbox": [ + 106, + 618, + 423, + 630 + ], + "score": 1.0, + "content": "coarse-to-fine refinement for RAFT is warranted. We leave that to future work.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 24.5, + "bbox_fs": [ + 104, + 520, + 506, + 630 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 645, + 282, + 659 + ], + "lines": [ + { + "bbox": [ + 105, + 644, + 283, + 660 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 283, + 660 + ], + "score": 1.0, + "content": "H Training and inference details", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "title", + "bbox": [ + 107, + 665, + 187, + 676 + ], + "lines": [ + { + "bbox": [ + 105, + 663, + 189, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 663, + 189, + 678 + ], + "score": 1.0, + "content": "H.1 Architecture", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 108, + 678, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 106, + 678, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 678, + 505, + 690 + ], + "score": 1.0, + "content": "UNet. The predominant architecture for diffusion models is the U-Net developed for the DDPM", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 689, + 506, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 689, + 494, + 701 + ], + "score": 1.0, + "content": "model [21], and later improved in several respects [11, 45, 63]. Here we adapt the Efficient", + "type": "text" + }, + { + "bbox": [ + 494, + 689, + 502, + 699 + ], + "score": 0.43, + "content": "U _ { ☉ }", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 689, + 506, + 701 + ], + "score": 1.0, + "content": "-", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 700, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 505, + 712 + ], + "score": 1.0, + "content": "Net architecture that was developed for Imagen [55]. It is more efficient that the U-Nets used in prior", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 505, + 723 + ], + "score": 1.0, + "content": "work owing to the use of fewer self-attention layers, fewer parameters and less computation at higher", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "score": 1.0, + "content": "resolutions, along with other adjustments that make it well suited to training medium resolution", + "type": "text", + "cross_page": true + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 400, + 179, + 413 + ], + "spans": [ + { + "bbox": [ + 106, + 400, + 179, + 413 + ], + "score": 1.0, + "content": "diffusion models.", + "type": "text", + "cross_page": true + } + ], + "index": 12 + } + ], + "index": 33.5, + "bbox_fs": [ + 105, + 678, + 506, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 146, + 97, + 466, + 140 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 70, + 505, + 91 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 70, + 505, + 82 + ], + "spans": [ + { + "bbox": [ + 106, + 70, + 505, + 82 + ], + "score": 1.0, + "content": "Table 15: Our coarse-to-fine refinement scheme marginally improves the performance of RAFT on Sintel", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 80, + 505, + 91 + ], + "spans": [ + { + "bbox": [ + 106, + 80, + 505, + 91 + ], + "score": 1.0, + "content": "Final while hurting performance on Sintel Clean and KITTI. We report the EPE on the Sintel and KITTI datasets.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 146, + 97, + 466, + 140 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 146, + 97, + 466, + 140 + ], + "spans": [ + { + "bbox": [ + 146, + 97, + 466, + 140 + ], + "score": 0.521, + "html": "
Sintel CleanSintel FinalKITTI
RAFT baseline1.272.282.71
RAFT with the coarse-to-fine refinement1.352.262.85
", + "type": "table", + "image_path": "a3fd4476e786ed849c127fcd63b14ecfecf60c25ce6050b6ab6c105cdadb9090.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 146, + 97, + 466, + 111.33333333333333 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 146, + 111.33333333333333, + 466, + 125.66666666666666 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 146, + 125.66666666666666, + 466, + 140.0 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "image", + "bbox": [ + 109, + 158, + 504, + 325 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 109, + 158, + 504, + 325 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 109, + 158, + 504, + 325 + ], + "spans": [ + { + "bbox": [ + 109, + 158, + 504, + 325 + ], + "score": 0.601, + "type": "image", + "image_path": "b89cb4a8c761c7dd9ce807b02b3ad203ba77e444dd56259afba6e66435fd486a.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 109, + 158, + 504, + 213.66666666666666 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 109, + 213.66666666666666, + 504, + 269.3333333333333 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 109, + 269.3333333333333, + 504, + 325.0 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 333, + 504, + 364 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 333, + 506, + 345 + ], + "spans": [ + { + "bbox": [ + 105, + 333, + 380, + 345 + ], + "score": 1.0, + "content": "Figure 15: Overview of the Efficient UNet architecture proposed in [55].", + "type": "text" + }, + { + "bbox": [ + 380, + 333, + 409, + 344 + ], + "score": 0.25, + "content": "C H \\_ I N", + "type": "inline_equation" + }, + { + "bbox": [ + 409, + 333, + 424, + 345 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 425, + 334, + 462, + 344 + ], + "score": 0.26, + "content": "C H \\_ O U T", + "type": "inline_equation" + }, + { + "bbox": [ + 462, + 333, + 506, + 345 + ], + "score": 1.0, + "content": "refer to the", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 344, + 505, + 354 + ], + "spans": [ + { + "bbox": [ + 106, + 344, + 505, + 354 + ], + "score": 1.0, + "content": "number of input and output channels respectively. t refers to the time embedding. FiLM refers to the modulation", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 354, + 377, + 364 + ], + "spans": [ + { + "bbox": [ + 106, + 354, + 194, + 364 + ], + "score": 1.0, + "content": "layers proposed in [47].", + "type": "text" + }, + { + "bbox": [ + 194, + 354, + 203, + 363 + ], + "score": 0.5, + "content": "N", + "type": "inline_equation" + }, + { + "bbox": [ + 203, + 354, + 291, + 364 + ], + "score": 1.0, + "content": "is the number of ResNet", + "type": "text" + }, + { + "bbox": [ + 292, + 355, + 300, + 363 + ], + "score": 0.72, + "content": "^ +", + "type": "inline_equation" + }, + { + "bbox": [ + 300, + 354, + 377, + 364 + ], + "score": 1.0, + "content": "self-attention blocks.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 106, + 390, + 504, + 412 + ], + "lines": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "score": 1.0, + "content": "resolutions, along with other adjustments that make it well suited to training medium resolution", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 400, + 179, + 413 + ], + "spans": [ + { + "bbox": [ + 106, + 400, + 179, + 413 + ], + "score": 1.0, + "content": "diffusion models.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5 + }, + { + "type": "text", + "bbox": [ + 107, + 417, + 506, + 559 + ], + "lines": [ + { + "bbox": [ + 106, + 416, + 505, + 430 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 287, + 430 + ], + "score": 1.0, + "content": "Specifically we adopt the configuration for the", + "type": "text" + }, + { + "bbox": [ + 287, + 417, + 367, + 428 + ], + "score": 0.91, + "content": "6 4 \\times 6 4 2 5 6 \\times 2 5 6", + "type": "inline_equation" + }, + { + "bbox": [ + 368, + 416, + 505, + 430 + ], + "score": 1.0, + "content": "super-resolution model (see Figure", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 428, + 506, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 428, + 506, + 441 + ], + "score": 1.0, + "content": "15 for an overview) with several changes. We drop the text cross-attention layers but preserve the self-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 104, + 437, + 506, + 453 + ], + "spans": [ + { + "bbox": [ + 104, + 437, + 506, + 453 + ], + "score": 1.0, + "content": "attention in the lowest resolution layers dblock4 and ublock4 (see Figure 15). For supervised training", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "spans": [ + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "score": 1.0, + "content": "for the flow model, we find it beneficial to additionally enable self-attention for the last-but-one", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 461, + 506, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 506, + 474 + ], + "score": 1.0, + "content": "layers dblock3 and ublock3. The number of input and output channels differ across self-supervised", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 472, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 472, + 506, + 484 + ], + "score": 1.0, + "content": "pre-training and supervised pre-training and are also different for flow and depth models. For self-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 483, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 204, + 496 + ], + "score": 1.0, + "content": "supervised pre-training", + "type": "text" + }, + { + "bbox": [ + 204, + 483, + 248, + 494 + ], + "score": 0.85, + "content": "C H \\_ I N { = } 6", + "type": "inline_equation" + }, + { + "bbox": [ + 248, + 483, + 267, + 496 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 267, + 483, + 320, + 494 + ], + "score": 0.83, + "content": "C H \\_ O U T { = } 3", + "type": "inline_equation" + }, + { + "bbox": [ + 321, + 483, + 506, + 496 + ], + "score": 1.0, + "content": "(see Figure 15) since the input consists of a", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 493, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 506, + 507 + ], + "score": 1.0, + "content": "3-channel source RGB image and a 3-channel noisy target image concatenated along the channel", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 505, + 506, + 518 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 404, + 518 + ], + "score": 1.0, + "content": "dimension and the output is a RGB image. The supervised depth model has", + "type": "text" + }, + { + "bbox": [ + 405, + 505, + 446, + 515 + ], + "score": 0.63, + "content": "C H \\_ I N { = } 4", + "type": "inline_equation" + }, + { + "bbox": [ + 446, + 505, + 498, + 518 + ], + "score": 1.0, + "content": "(RGB image", + "type": "text" + }, + { + "bbox": [ + 498, + 506, + 506, + 515 + ], + "score": 0.72, + "content": "^ +", + "type": "inline_equation" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 515, + 506, + 528 + ], + "spans": [ + { + "bbox": [ + 106, + 515, + 176, + 528 + ], + "score": 1.0, + "content": "noisy depth) and", + "type": "text" + }, + { + "bbox": [ + 176, + 516, + 228, + 527 + ], + "score": 0.86, + "content": "C H _ { - } O U T { = } I", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 515, + 389, + 528 + ], + "score": 1.0, + "content": ". The supervised optical flow model has", + "type": "text" + }, + { + "bbox": [ + 390, + 516, + 433, + 527 + ], + "score": 0.79, + "content": "C H \\_ I N { = } 8", + "type": "inline_equation" + }, + { + "bbox": [ + 433, + 515, + 497, + 528 + ], + "score": 1.0, + "content": "(2 RGB images", + "type": "text" + }, + { + "bbox": [ + 498, + 517, + 506, + 526 + ], + "score": 0.71, + "content": "^ +", + "type": "inline_equation" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 526, + 506, + 539 + ], + "spans": [ + { + "bbox": [ + 106, + 526, + 173, + 539 + ], + "score": 1.0, + "content": "noisy flow along", + "type": "text" + }, + { + "bbox": [ + 174, + 529, + 181, + 537 + ], + "score": 0.71, + "content": "x", + "type": "inline_equation" + }, + { + "bbox": [ + 181, + 526, + 198, + 539 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 198, + 528, + 205, + 538 + ], + "score": 0.75, + "content": "y", + "type": "inline_equation" + }, + { + "bbox": [ + 205, + 526, + 226, + 539 + ], + "score": 1.0, + "content": ") and", + "type": "text" + }, + { + "bbox": [ + 226, + 527, + 276, + 538 + ], + "score": 0.76, + "content": "C H _ { - } O U T { = } 2", + "type": "inline_equation" + }, + { + "bbox": [ + 277, + 526, + 506, + 539 + ], + "score": 1.0, + "content": ". Note that this means we need to reinitialize the input and", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 537, + 505, + 551 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 551 + ], + "score": 1.0, + "content": "output convolutional kernels and biases before the supervised pretraining stage. All other weights are", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 550, + 141, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 550, + 141, + 560 + ], + "score": 1.0, + "content": "re-used.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 108, + 565, + 505, + 598 + ], + "lines": [ + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 385, + 577 + ], + "score": 1.0, + "content": "Resolution. Our self-supervised model was trained at a resolution of", + "type": "text" + }, + { + "bbox": [ + 386, + 565, + 429, + 576 + ], + "score": 0.89, + "content": "2 5 6 \\times 2 5 6", + "type": "inline_equation" + }, + { + "bbox": [ + 429, + 564, + 505, + 577 + ], + "score": 1.0, + "content": ". The indoor depth", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 575, + 505, + 587 + ], + "spans": [ + { + "bbox": [ + 106, + 575, + 185, + 587 + ], + "score": 1.0, + "content": "model is trained at", + "type": "text" + }, + { + "bbox": [ + 185, + 576, + 225, + 586 + ], + "score": 0.89, + "content": "2 4 0 \\times 3 2 0", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 575, + 310, + 587 + ], + "score": 1.0, + "content": ". For Waymo we use", + "type": "text" + }, + { + "bbox": [ + 311, + 576, + 351, + 586 + ], + "score": 0.89, + "content": "2 5 6 \\times 3 8 4", + "type": "inline_equation" + }, + { + "bbox": [ + 351, + 575, + 438, + 587 + ], + "score": 1.0, + "content": "and for KITTI depth", + "type": "text" + }, + { + "bbox": [ + 438, + 576, + 477, + 586 + ], + "score": 0.87, + "content": "2 5 6 \\times 8 3 2", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 575, + 505, + 587 + ], + "score": 1.0, + "content": ". Flow", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 587, + 409, + 599 + ], + "spans": [ + { + "bbox": [ + 106, + 587, + 253, + 599 + ], + "score": 1.0, + "content": "pretraining is done at a resolution of", + "type": "text" + }, + { + "bbox": [ + 254, + 587, + 292, + 597 + ], + "score": 0.89, + "content": "3 2 0 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 293, + 587, + 365, + 599 + ], + "score": 1.0, + "content": ", and finetuning at", + "type": "text" + }, + { + "bbox": [ + 366, + 587, + 405, + 597 + ], + "score": 0.88, + "content": "3 2 0 \\times 7 6 8", + "type": "inline_equation" + }, + { + "bbox": [ + 405, + 587, + 409, + 599 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27 + }, + { + "type": "title", + "bbox": [ + 108, + 619, + 249, + 631 + ], + "lines": [ + { + "bbox": [ + 105, + 618, + 250, + 632 + ], + "spans": [ + { + "bbox": [ + 105, + 618, + 250, + 632 + ], + "score": 1.0, + "content": "H.2 Datasets and augmentation", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 640, + 505, + 673 + ], + "lines": [ + { + "bbox": [ + 105, + 639, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 105, + 639, + 505, + 653 + ], + "score": 1.0, + "content": "For unsupervised pre-training, we use the ImageNet-1K [10] and Places365 [88] datasets and train on", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 649, + 505, + 664 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 505, + 664 + ], + "score": 1.0, + "content": "the self-supervised tasks of colorization, inpainting, uncropping, and JPEG decompression, following", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 661, + 320, + 674 + ], + "spans": [ + { + "bbox": [ + 105, + 661, + 320, + 674 + ], + "score": 1.0, + "content": "[54]. Throughout, we mix datasets at the batch level.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 108, + 678, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 105, + 678, + 507, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 678, + 459, + 691 + ], + "score": 1.0, + "content": "Flow. For supervised flow pretraining we use a mix of AutoFlow (native resolution", + "type": "text" + }, + { + "bbox": [ + 460, + 678, + 501, + 689 + ], + "score": 0.85, + "content": "4 4 8 \\times 5 7 6", + "type": "inline_equation" + }, + { + "bbox": [ + 501, + 678, + 507, + 691 + ], + "score": 1.0, + "content": "),", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 689, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 689, + 166, + 701 + ], + "score": 1.0, + "content": "FlyingThings", + "type": "text" + }, + { + "bbox": [ + 166, + 689, + 210, + 700 + ], + "score": 0.87, + "content": "( 5 4 0 \\times 9 6 0 )", + "type": "inline_equation" + }, + { + "bbox": [ + 210, + 689, + 249, + 701 + ], + "score": 1.0, + "content": ", Kubric", + "type": "text" + }, + { + "bbox": [ + 250, + 689, + 293, + 700 + ], + "score": 0.87, + "content": "( 5 1 2 \\times 5 1 2 )", + "type": "inline_equation" + }, + { + "bbox": [ + 294, + 689, + 360, + 701 + ], + "score": 1.0, + "content": "and TartanAir", + "type": "text" + }, + { + "bbox": [ + 361, + 689, + 403, + 700 + ], + "score": 0.84, + "content": "( 4 8 0 \\times 6 4 0 )", + "type": "inline_equation" + }, + { + "bbox": [ + 404, + 689, + 505, + 701 + ], + "score": 1.0, + "content": "synthetic datasets. We", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 104, + 699, + 506, + 714 + ], + "spans": [ + { + "bbox": [ + 104, + 699, + 428, + 714 + ], + "score": 1.0, + "content": "finetune on the standard mixture consisting of AutoFlow, FlyingThings, Viper", + "type": "text" + }, + { + "bbox": [ + 428, + 700, + 470, + 711 + ], + "score": 0.83, + "content": "( 5 4 0 \\times 9 6 0 )", + "type": "inline_equation" + }, + { + "bbox": [ + 470, + 699, + 506, + 714 + ], + "score": 1.0, + "content": "), HD1K", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 108, + 711, + 340, + 723 + ], + "spans": [ + { + "bbox": [ + 108, + 711, + 155, + 722 + ], + "score": 0.86, + "content": "( 5 4 0 \\times 1 2 8 0 )", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 711, + 189, + 723 + ], + "score": 1.0, + "content": "), Sintel", + "type": "text" + }, + { + "bbox": [ + 189, + 711, + 235, + 722 + ], + "score": 0.86, + "content": "4 3 6 \\times 1 0 2 4 )", + "type": "inline_equation" + }, + { + "bbox": [ + 236, + 711, + 288, + 723 + ], + "score": 1.0, + "content": ", and KITTI", + "type": "text" + }, + { + "bbox": [ + 288, + 711, + 335, + 722 + ], + "score": 0.85, + "content": "( 3 7 5 \\times 1 2 4 2 )", + "type": "inline_equation" + }, + { + "bbox": [ + 335, + 711, + 340, + 723 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34.5 + } + ], + "page_idx": 22, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 298, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 298, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 146, + 97, + 466, + 140 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 70, + 505, + 91 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 70, + 505, + 82 + ], + "spans": [ + { + "bbox": [ + 106, + 70, + 505, + 82 + ], + "score": 1.0, + "content": "Table 15: Our coarse-to-fine refinement scheme marginally improves the performance of RAFT on Sintel", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 80, + 505, + 91 + ], + "spans": [ + { + "bbox": [ + 106, + 80, + 505, + 91 + ], + "score": 1.0, + "content": "Final while hurting performance on Sintel Clean and KITTI. We report the EPE on the Sintel and KITTI datasets.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table_body", + "bbox": [ + 146, + 97, + 466, + 140 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 146, + 97, + 466, + 140 + ], + "spans": [ + { + "bbox": [ + 146, + 97, + 466, + 140 + ], + "score": 0.521, + "html": "
Sintel CleanSintel FinalKITTI
RAFT baseline1.272.282.71
RAFT with the coarse-to-fine refinement1.352.262.85
", + "type": "table", + "image_path": "a3fd4476e786ed849c127fcd63b14ecfecf60c25ce6050b6ab6c105cdadb9090.jpg" + } + ] + } + ], + "index": 3, + "virtual_lines": [ + { + "bbox": [ + 146, + 97, + 466, + 111.33333333333333 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 146, + 111.33333333333333, + 466, + 125.66666666666666 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 146, + 125.66666666666666, + 466, + 140.0 + ], + "spans": [], + "index": 4 + } + ] + } + ], + "index": 1.75 + }, + { + "type": "image", + "bbox": [ + 109, + 158, + 504, + 325 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 109, + 158, + 504, + 325 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 109, + 158, + 504, + 325 + ], + "spans": [ + { + "bbox": [ + 109, + 158, + 504, + 325 + ], + "score": 0.601, + "type": "image", + "image_path": "b89cb4a8c761c7dd9ce807b02b3ad203ba77e444dd56259afba6e66435fd486a.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 109, + 158, + 504, + 213.66666666666666 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 109, + 213.66666666666666, + 504, + 269.3333333333333 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 109, + 269.3333333333333, + 504, + 325.0 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 333, + 504, + 364 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 333, + 506, + 345 + ], + "spans": [ + { + "bbox": [ + 105, + 333, + 380, + 345 + ], + "score": 1.0, + "content": "Figure 15: Overview of the Efficient UNet architecture proposed in [55].", + "type": "text" + }, + { + "bbox": [ + 380, + 333, + 409, + 344 + ], + "score": 0.25, + "content": "C H \\_ I N", + "type": "inline_equation" + }, + { + "bbox": [ + 409, + 333, + 424, + 345 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 425, + 334, + 462, + 344 + ], + "score": 0.26, + "content": "C H \\_ O U T", + "type": "inline_equation" + }, + { + "bbox": [ + 462, + 333, + 506, + 345 + ], + "score": 1.0, + "content": "refer to the", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 344, + 505, + 354 + ], + "spans": [ + { + "bbox": [ + 106, + 344, + 505, + 354 + ], + "score": 1.0, + "content": "number of input and output channels respectively. t refers to the time embedding. FiLM refers to the modulation", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 354, + 377, + 364 + ], + "spans": [ + { + "bbox": [ + 106, + 354, + 194, + 364 + ], + "score": 1.0, + "content": "layers proposed in [47].", + "type": "text" + }, + { + "bbox": [ + 194, + 354, + 203, + 363 + ], + "score": 0.5, + "content": "N", + "type": "inline_equation" + }, + { + "bbox": [ + 203, + 354, + 291, + 364 + ], + "score": 1.0, + "content": "is the number of ResNet", + "type": "text" + }, + { + "bbox": [ + 292, + 355, + 300, + 363 + ], + "score": 0.72, + "content": "^ +", + "type": "inline_equation" + }, + { + "bbox": [ + 300, + 354, + 377, + 364 + ], + "score": 1.0, + "content": "self-attention blocks.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 106, + 390, + 504, + 412 + ], + "lines": [], + "index": 11.5, + "bbox_fs": [ + 105, + 389, + 505, + 413 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 417, + 506, + 559 + ], + "lines": [ + { + "bbox": [ + 106, + 416, + 505, + 430 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 287, + 430 + ], + "score": 1.0, + "content": "Specifically we adopt the configuration for the", + "type": "text" + }, + { + "bbox": [ + 287, + 417, + 367, + 428 + ], + "score": 0.91, + "content": "6 4 \\times 6 4 2 5 6 \\times 2 5 6", + "type": "inline_equation" + }, + { + "bbox": [ + 368, + 416, + 505, + 430 + ], + "score": 1.0, + "content": "super-resolution model (see Figure", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 428, + 506, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 428, + 506, + 441 + ], + "score": 1.0, + "content": "15 for an overview) with several changes. We drop the text cross-attention layers but preserve the self-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 104, + 437, + 506, + 453 + ], + "spans": [ + { + "bbox": [ + 104, + 437, + 506, + 453 + ], + "score": 1.0, + "content": "attention in the lowest resolution layers dblock4 and ublock4 (see Figure 15). For supervised training", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "spans": [ + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "score": 1.0, + "content": "for the flow model, we find it beneficial to additionally enable self-attention for the last-but-one", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 461, + 506, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 506, + 474 + ], + "score": 1.0, + "content": "layers dblock3 and ublock3. The number of input and output channels differ across self-supervised", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 472, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 472, + 506, + 484 + ], + "score": 1.0, + "content": "pre-training and supervised pre-training and are also different for flow and depth models. For self-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 483, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 204, + 496 + ], + "score": 1.0, + "content": "supervised pre-training", + "type": "text" + }, + { + "bbox": [ + 204, + 483, + 248, + 494 + ], + "score": 0.85, + "content": "C H \\_ I N { = } 6", + "type": "inline_equation" + }, + { + "bbox": [ + 248, + 483, + 267, + 496 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 267, + 483, + 320, + 494 + ], + "score": 0.83, + "content": "C H \\_ O U T { = } 3", + "type": "inline_equation" + }, + { + "bbox": [ + 321, + 483, + 506, + 496 + ], + "score": 1.0, + "content": "(see Figure 15) since the input consists of a", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 493, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 506, + 507 + ], + "score": 1.0, + "content": "3-channel source RGB image and a 3-channel noisy target image concatenated along the channel", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 505, + 506, + 518 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 404, + 518 + ], + "score": 1.0, + "content": "dimension and the output is a RGB image. The supervised depth model has", + "type": "text" + }, + { + "bbox": [ + 405, + 505, + 446, + 515 + ], + "score": 0.63, + "content": "C H \\_ I N { = } 4", + "type": "inline_equation" + }, + { + "bbox": [ + 446, + 505, + 498, + 518 + ], + "score": 1.0, + "content": "(RGB image", + "type": "text" + }, + { + "bbox": [ + 498, + 506, + 506, + 515 + ], + "score": 0.72, + "content": "^ +", + "type": "inline_equation" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 515, + 506, + 528 + ], + "spans": [ + { + "bbox": [ + 106, + 515, + 176, + 528 + ], + "score": 1.0, + "content": "noisy depth) and", + "type": "text" + }, + { + "bbox": [ + 176, + 516, + 228, + 527 + ], + "score": 0.86, + "content": "C H _ { - } O U T { = } I", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 515, + 389, + 528 + ], + "score": 1.0, + "content": ". The supervised optical flow model has", + "type": "text" + }, + { + "bbox": [ + 390, + 516, + 433, + 527 + ], + "score": 0.79, + "content": "C H \\_ I N { = } 8", + "type": "inline_equation" + }, + { + "bbox": [ + 433, + 515, + 497, + 528 + ], + "score": 1.0, + "content": "(2 RGB images", + "type": "text" + }, + { + "bbox": [ + 498, + 517, + 506, + 526 + ], + "score": 0.71, + "content": "^ +", + "type": "inline_equation" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 526, + 506, + 539 + ], + "spans": [ + { + "bbox": [ + 106, + 526, + 173, + 539 + ], + "score": 1.0, + "content": "noisy flow along", + "type": "text" + }, + { + "bbox": [ + 174, + 529, + 181, + 537 + ], + "score": 0.71, + "content": "x", + "type": "inline_equation" + }, + { + "bbox": [ + 181, + 526, + 198, + 539 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 198, + 528, + 205, + 538 + ], + "score": 0.75, + "content": "y", + "type": "inline_equation" + }, + { + "bbox": [ + 205, + 526, + 226, + 539 + ], + "score": 1.0, + "content": ") and", + "type": "text" + }, + { + "bbox": [ + 226, + 527, + 276, + 538 + ], + "score": 0.76, + "content": "C H _ { - } O U T { = } 2", + "type": "inline_equation" + }, + { + "bbox": [ + 277, + 526, + 506, + 539 + ], + "score": 1.0, + "content": ". Note that this means we need to reinitialize the input and", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 537, + 505, + 551 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 551 + ], + "score": 1.0, + "content": "output convolutional kernels and biases before the supervised pretraining stage. All other weights are", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 550, + 141, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 550, + 141, + 560 + ], + "score": 1.0, + "content": "re-used.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 19, + "bbox_fs": [ + 104, + 416, + 506, + 560 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 565, + 505, + 598 + ], + "lines": [ + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 385, + 577 + ], + "score": 1.0, + "content": "Resolution. Our self-supervised model was trained at a resolution of", + "type": "text" + }, + { + "bbox": [ + 386, + 565, + 429, + 576 + ], + "score": 0.89, + "content": "2 5 6 \\times 2 5 6", + "type": "inline_equation" + }, + { + "bbox": [ + 429, + 564, + 505, + 577 + ], + "score": 1.0, + "content": ". The indoor depth", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 575, + 505, + 587 + ], + "spans": [ + { + "bbox": [ + 106, + 575, + 185, + 587 + ], + "score": 1.0, + "content": "model is trained at", + "type": "text" + }, + { + "bbox": [ + 185, + 576, + 225, + 586 + ], + "score": 0.89, + "content": "2 4 0 \\times 3 2 0", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 575, + 310, + 587 + ], + "score": 1.0, + "content": ". For Waymo we use", + "type": "text" + }, + { + "bbox": [ + 311, + 576, + 351, + 586 + ], + "score": 0.89, + "content": "2 5 6 \\times 3 8 4", + "type": "inline_equation" + }, + { + "bbox": [ + 351, + 575, + 438, + 587 + ], + "score": 1.0, + "content": "and for KITTI depth", + "type": "text" + }, + { + "bbox": [ + 438, + 576, + 477, + 586 + ], + "score": 0.87, + "content": "2 5 6 \\times 8 3 2", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 575, + 505, + 587 + ], + "score": 1.0, + "content": ". Flow", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 587, + 409, + 599 + ], + "spans": [ + { + "bbox": [ + 106, + 587, + 253, + 599 + ], + "score": 1.0, + "content": "pretraining is done at a resolution of", + "type": "text" + }, + { + "bbox": [ + 254, + 587, + 292, + 597 + ], + "score": 0.89, + "content": "3 2 0 \\times 4 4 8", + "type": "inline_equation" + }, + { + "bbox": [ + 293, + 587, + 365, + 599 + ], + "score": 1.0, + "content": ", and finetuning at", + "type": "text" + }, + { + "bbox": [ + 366, + 587, + 405, + 597 + ], + "score": 0.88, + "content": "3 2 0 \\times 7 6 8", + "type": "inline_equation" + }, + { + "bbox": [ + 405, + 587, + 409, + 599 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27, + "bbox_fs": [ + 105, + 564, + 505, + 599 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 619, + 249, + 631 + ], + "lines": [ + { + "bbox": [ + 105, + 618, + 250, + 632 + ], + "spans": [ + { + "bbox": [ + 105, + 618, + 250, + 632 + ], + "score": 1.0, + "content": "H.2 Datasets and augmentation", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 640, + 505, + 673 + ], + "lines": [ + { + "bbox": [ + 105, + 639, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 105, + 639, + 505, + 653 + ], + "score": 1.0, + "content": "For unsupervised pre-training, we use the ImageNet-1K [10] and Places365 [88] datasets and train on", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 649, + 505, + 664 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 505, + 664 + ], + "score": 1.0, + "content": "the self-supervised tasks of colorization, inpainting, uncropping, and JPEG decompression, following", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 661, + 320, + 674 + ], + "spans": [ + { + "bbox": [ + 105, + 661, + 320, + 674 + ], + "score": 1.0, + "content": "[54]. Throughout, we mix datasets at the batch level.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31, + "bbox_fs": [ + 105, + 639, + 505, + 674 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 678, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 105, + 678, + 507, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 678, + 459, + 691 + ], + "score": 1.0, + "content": "Flow. For supervised flow pretraining we use a mix of AutoFlow (native resolution", + "type": "text" + }, + { + "bbox": [ + 460, + 678, + 501, + 689 + ], + "score": 0.85, + "content": "4 4 8 \\times 5 7 6", + "type": "inline_equation" + }, + { + "bbox": [ + 501, + 678, + 507, + 691 + ], + "score": 1.0, + "content": "),", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 689, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 689, + 166, + 701 + ], + "score": 1.0, + "content": "FlyingThings", + "type": "text" + }, + { + "bbox": [ + 166, + 689, + 210, + 700 + ], + "score": 0.87, + "content": "( 5 4 0 \\times 9 6 0 )", + "type": "inline_equation" + }, + { + "bbox": [ + 210, + 689, + 249, + 701 + ], + "score": 1.0, + "content": ", Kubric", + "type": "text" + }, + { + "bbox": [ + 250, + 689, + 293, + 700 + ], + "score": 0.87, + "content": "( 5 1 2 \\times 5 1 2 )", + "type": "inline_equation" + }, + { + "bbox": [ + 294, + 689, + 360, + 701 + ], + "score": 1.0, + "content": "and TartanAir", + "type": "text" + }, + { + "bbox": [ + 361, + 689, + 403, + 700 + ], + "score": 0.84, + "content": "( 4 8 0 \\times 6 4 0 )", + "type": "inline_equation" + }, + { + "bbox": [ + 404, + 689, + 505, + 701 + ], + "score": 1.0, + "content": "synthetic datasets. We", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 104, + 699, + 506, + 714 + ], + "spans": [ + { + "bbox": [ + 104, + 699, + 428, + 714 + ], + "score": 1.0, + "content": "finetune on the standard mixture consisting of AutoFlow, FlyingThings, Viper", + "type": "text" + }, + { + "bbox": [ + 428, + 700, + 470, + 711 + ], + "score": 0.83, + "content": "( 5 4 0 \\times 9 6 0 )", + "type": "inline_equation" + }, + { + "bbox": [ + 470, + 699, + 506, + 714 + ], + "score": 1.0, + "content": "), HD1K", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 108, + 711, + 340, + 723 + ], + "spans": [ + { + "bbox": [ + 108, + 711, + 155, + 722 + ], + "score": 0.86, + "content": "( 5 4 0 \\times 1 2 8 0 )", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 711, + 189, + 723 + ], + "score": 1.0, + "content": "), Sintel", + "type": "text" + }, + { + "bbox": [ + 189, + 711, + 235, + 722 + ], + "score": 0.86, + "content": "4 3 6 \\times 1 0 2 4 )", + "type": "inline_equation" + }, + { + "bbox": [ + 236, + 711, + 288, + 723 + ], + "score": 1.0, + "content": ", and KITTI", + "type": "text" + }, + { + "bbox": [ + 288, + 711, + 335, + 722 + ], + "score": 0.85, + "content": "( 3 7 5 \\times 1 2 4 2 )", + "type": "inline_equation" + }, + { + "bbox": [ + 335, + 711, + 340, + 723 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34.5, + "bbox_fs": [ + 104, + 678, + 507, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 72, + 504, + 95 + ], + "lines": [ + { + "bbox": [ + 105, + 71, + 505, + 86 + ], + "spans": [ + { + "bbox": [ + 105, + 71, + 505, + 86 + ], + "score": 1.0, + "content": "We follow the same photometric and geometric augmentation schemes from [70], comprising random", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 82, + 288, + 97 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 288, + 97 + ], + "score": 1.0, + "content": "affine transformation, flipping, and cropping.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 107, + 100, + 505, + 154 + ], + "lines": [ + { + "bbox": [ + 105, + 100, + 506, + 113 + ], + "spans": [ + { + "bbox": [ + 105, + 100, + 506, + 113 + ], + "score": 1.0, + "content": "Depth. For supervised pre-training of the indoor model we mix the following datasets. ScanNet", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 110, + 505, + 124 + ], + "spans": [ + { + "bbox": [ + 105, + 110, + 180, + 124 + ], + "score": 1.0, + "content": "[9] is a dataset of", + "type": "text" + }, + { + "bbox": [ + 180, + 111, + 204, + 122 + ], + "score": 0.46, + "content": "2 . 5 \\mathbf { M }", + "type": "inline_equation" + }, + { + "bbox": [ + 204, + 110, + 505, + 124 + ], + "score": 1.0, + "content": "examples captured using a Kinect v1-like sensor. It provides depth maps", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 121, + 505, + 133 + ], + "spans": [ + { + "bbox": [ + 105, + 121, + 117, + 133 + ], + "score": 1.0, + "content": "at", + "type": "text" + }, + { + "bbox": [ + 117, + 122, + 157, + 133 + ], + "score": 0.88, + "content": "4 8 0 \\times 6 4 0", + "type": "inline_equation" + }, + { + "bbox": [ + 157, + 121, + 242, + 133 + ], + "score": 1.0, + "content": "and RGB images at", + "type": "text" + }, + { + "bbox": [ + 243, + 122, + 287, + 132 + ], + "score": 0.89, + "content": "9 6 8 \\times 1 2 9 6", + "type": "inline_equation" + }, + { + "bbox": [ + 288, + 121, + 505, + 133 + ], + "score": 1.0, + "content": ". SceneNet RGB-D [41] is a synthetic dataset of 5M", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 132, + 506, + 145 + ], + "spans": [ + { + "bbox": [ + 105, + 132, + 506, + 145 + ], + "score": 1.0, + "content": "images generated by rendering ShapeNet [6] objects in scenes from SceneNet [20] at a resolution of", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 143, + 149, + 155 + ], + "spans": [ + { + "bbox": [ + 106, + 144, + 145, + 155 + ], + "score": 0.86, + "content": "2 4 0 \\times 3 2 0", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 143, + 149, + 155 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 107, + 159, + 505, + 204 + ], + "lines": [ + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "score": 1.0, + "content": "For the outdoor model training we use the Waymo Open Dataset [71], a large-scale driving dataset", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 171, + 505, + 184 + ], + "spans": [ + { + "bbox": [ + 105, + 171, + 187, + 184 + ], + "score": 1.0, + "content": "consisting of about", + "type": "text" + }, + { + "bbox": [ + 187, + 171, + 210, + 182 + ], + "score": 0.27, + "content": "2 0 0 \\mathrm { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 210, + 171, + 505, + 184 + ], + "score": 1.0, + "content": "frames. Each frame provides RGB images from 5 cameras and LiDAR", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 182, + 505, + 194 + ], + "spans": [ + { + "bbox": [ + 105, + 182, + 505, + 194 + ], + "score": 1.0, + "content": "maps. We use the RGB images from the FRONT, FRONT_LEFT and FRONT_RIGHT cameras and", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 192, + 380, + 206 + ], + "spans": [ + { + "bbox": [ + 105, + 192, + 380, + 206 + ], + "score": 1.0, + "content": "the TOP LiDAR only to build about 600k aligned RGB depth maps.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 108, + 209, + 504, + 242 + ], + "lines": [ + { + "bbox": [ + 106, + 209, + 506, + 222 + ], + "spans": [ + { + "bbox": [ + 106, + 209, + 346, + 222 + ], + "score": 1.0, + "content": "For indoor fine-tuning and evaluation we use NYU depth", + "type": "text" + }, + { + "bbox": [ + 346, + 210, + 358, + 219 + ], + "score": 0.38, + "content": "\\nu 2", + "type": "inline_equation" + }, + { + "bbox": [ + 358, + 209, + 506, + 222 + ], + "score": 1.0, + "content": "[61], a commonly used dataset for", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 220, + 504, + 232 + ], + "spans": [ + { + "bbox": [ + 106, + 220, + 464, + 232 + ], + "score": 1.0, + "content": "evaluating indoor depth prediction models. It provides aligned image and depth maps at", + "type": "text" + }, + { + "bbox": [ + 464, + 220, + 504, + 231 + ], + "score": 0.88, + "content": "4 8 0 \\times 6 4 0", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 231, + 487, + 243 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 297, + 243 + ], + "score": 1.0, + "content": "resolution. We use the official split comprising", + "type": "text" + }, + { + "bbox": [ + 298, + 231, + 315, + 241 + ], + "score": 0.33, + "content": "5 0 \\mathrm { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 231, + 487, + 243 + ], + "score": 1.0, + "content": "images for training and 654 for evaluation.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12 + }, + { + "type": "text", + "bbox": [ + 107, + 247, + 505, + 280 + ], + "lines": [ + { + "bbox": [ + 106, + 247, + 505, + 259 + ], + "spans": [ + { + "bbox": [ + 106, + 247, + 505, + 259 + ], + "score": 1.0, + "content": "For outdoor fine-tuning and evaluation, we use KITTI [17], an outdoor driving dataset which provides", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 257, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 105, + 257, + 333, + 271 + ], + "score": 1.0, + "content": "RGB images and LiDAR scans at resolutions close to", + "type": "text" + }, + { + "bbox": [ + 333, + 258, + 378, + 269 + ], + "score": 0.89, + "content": "3 7 0 \\times 1 2 2 6", + "type": "inline_equation" + }, + { + "bbox": [ + 378, + 257, + 505, + 271 + ], + "score": 1.0, + "content": ". We use the training/test split", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 269, + 392, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 269, + 392, + 282 + ], + "score": 1.0, + "content": "proposed by [13], comprising 26k training images and 652 test images.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 107, + 285, + 505, + 319 + ], + "lines": [ + { + "bbox": [ + 105, + 284, + 506, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 284, + 506, + 298 + ], + "score": 1.0, + "content": "We use random horizontal flip data augmentation which is common in prior work. Where needed,", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 297, + 505, + 309 + ], + "spans": [ + { + "bbox": [ + 106, + 297, + 505, + 309 + ], + "score": 1.0, + "content": "images and dense depth maps are resized using bilinear interpolation to the model’s resolution for", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 307, + 376, + 321 + ], + "spans": [ + { + "bbox": [ + 105, + 307, + 376, + 321 + ], + "score": 1.0, + "content": "training and nearest neighbor interpolation is used for sparse maps.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18 + }, + { + "type": "title", + "bbox": [ + 107, + 336, + 381, + 348 + ], + "lines": [ + { + "bbox": [ + 105, + 335, + 383, + 352 + ], + "spans": [ + { + "bbox": [ + 105, + 335, + 383, + 352 + ], + "score": 1.0, + "content": "H.3 Step-unrolling and interpolation of missing depth and flow", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 356, + 505, + 389 + ], + "lines": [ + { + "bbox": [ + 105, + 355, + 505, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 355, + 505, + 369 + ], + "score": 1.0, + "content": "As discussed in Section 3.2 of the main paper, infilling and step-unrolling are used to mitigate", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 366, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 105, + 366, + 505, + 379 + ], + "score": 1.0, + "content": "distribution shift between training and inference with diffusion models. The problem arises due to", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 378, + 344, + 390 + ], + "spans": [ + { + "bbox": [ + 106, + 378, + 344, + 390 + ], + "score": 1.0, + "content": "the missing data in the training depth maps and flow fields.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 106, + 394, + 505, + 460 + ], + "lines": [ + { + "bbox": [ + 105, + 393, + 505, + 407 + ], + "spans": [ + { + "bbox": [ + 105, + 393, + 505, + 407 + ], + "score": 1.0, + "content": "Infilling. For indoor depth maps, we use nearest neighbor interpolation during training (see Section", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 404, + 505, + 417 + ], + "spans": [ + { + "bbox": [ + 105, + 404, + 505, + 417 + ], + "score": 1.0, + "content": "3.2 in the main paper). For the outdoor depth data we use nearest neighbor interpolation except for", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "score": 1.0, + "content": "sky regions, as they are often large and are much further from the camera than adjacent objects in the", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 428, + 505, + 439 + ], + "spans": [ + { + "bbox": [ + 106, + 428, + 505, + 439 + ], + "score": 1.0, + "content": "image. We use an off-the-shelf sky segmenter [36], and then set all sky pixels to be the maximum", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 437, + 506, + 451 + ], + "spans": [ + { + "bbox": [ + 105, + 437, + 193, + 451 + ], + "score": 1.0, + "content": "modeled depth (here,", + "type": "text" + }, + { + "bbox": [ + 194, + 438, + 214, + 449 + ], + "score": 0.5, + "content": "8 0 \\mathrm { m } \\mathrm { , }", + "type": "inline_equation" + }, + { + "bbox": [ + 214, + 437, + 506, + 451 + ], + "score": 1.0, + "content": "). For missing optical flow ground truth we employ a simple sequence of", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 449, + 413, + 461 + ], + "spans": [ + { + "bbox": [ + 106, + 449, + 413, + 461 + ], + "score": 1.0, + "content": "1D nearest neighbor interpolations first along rows, and then along columns.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 26.5 + }, + { + "type": "table", + "bbox": [ + 155, + 522, + 455, + 624 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 107, + 474, + 505, + 515 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 473, + 505, + 488 + ], + "spans": [ + { + "bbox": [ + 105, + 473, + 505, + 488 + ], + "score": 1.0, + "content": "Table 16: Ablation of the number of unroll steps for monocular depth estimation. Performs improves up", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 485, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 106, + 485, + 505, + 496 + ], + "score": 1.0, + "content": "to four steps of unrolling and plateaus thereafter. The models trained without infilling missing depth benefit", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 495, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 495, + 505, + 506 + ], + "score": 1.0, + "content": "more from a larger number of unroll steps, which is to be expected. Best and second best results are bolded and", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 504, + 194, + 517 + ], + "spans": [ + { + "bbox": [ + 105, + 504, + 194, + 517 + ], + "score": 1.0, + "content": "underlined respectively.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 31.5 + }, + { + "type": "table_body", + "bbox": [ + 155, + 522, + 455, + 624 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 155, + 522, + 455, + 624 + ], + "spans": [ + { + "bbox": [ + 155, + 522, + 455, + 624 + ], + "score": 0.982, + "html": "
NYUKITTI
No infillInfillNo infillInfill
Unroll stepsRELRMSRELRMSRELRMSRELRMS
00.0790.3310.0770.3380.2223.7700.0572.744
10.0760.3240.0750.3240.0852.8440.0562.700
20.0760.3170.0750.3150.0682.7990.0542.629
30.0750.3160.0740.3170.0612.7890.0542.591
40.0750.3160.0740.3150.0592.7390.0532.568
", + "type": "table", + "image_path": "578d7c0bf70f4c42b92f60b2593c6ab4ce127c04e43560ec728543a10196d35a.jpg" + } + ] + } + ], + "index": 35, + "virtual_lines": [ + { + "bbox": [ + 155, + 522, + 455, + 556.0 + ], + "spans": [], + "index": 34 + }, + { + "bbox": [ + 155, + 556.0, + 455, + 590.0 + ], + "spans": [], + "index": 35 + }, + { + "bbox": [ + 155, + 590.0, + 455, + 624.0 + ], + "spans": [], + "index": 36 + } + ] + } + ], + "index": 33.25 + }, + { + "type": "text", + "bbox": [ + 108, + 640, + 505, + 673 + ], + "lines": [ + { + "bbox": [ + 106, + 640, + 505, + 652 + ], + "spans": [ + { + "bbox": [ + 106, + 640, + 505, + 652 + ], + "score": 1.0, + "content": "Step-unrolling. By default we use a single unroll step in all results where step-unrolling is enabled.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 649, + 506, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 506, + 663 + ], + "score": 1.0, + "content": "In Table 16, we show that using multiple unroll steps can further improve performance on the task of", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 662, + 223, + 674 + ], + "spans": [ + { + "bbox": [ + 106, + 662, + 223, + 674 + ], + "score": 1.0, + "content": "monocular depth estimation.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 107, + 678, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 105, + 678, + 506, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 678, + 506, + 691 + ], + "score": 1.0, + "content": "Finally, while we use infilling and step-unrolling, there are other ways in which one might try to", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 689, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 689, + 506, + 702 + ], + "score": 1.0, + "content": "mitigate the problem. One such approach was taken by [46], which faced a similar problem when", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 700, + 506, + 713 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 506, + 713 + ], + "score": 1.0, + "content": "training a vector-quantizer on depth data. Their approach was to synthetically add more holes", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 711, + 505, + 724 + ], + "spans": [ + { + "bbox": [ + 105, + 711, + 505, + 724 + ], + "score": 1.0, + "content": "following a carefully chosen masking ratio. We prefer our approach since nearest neighbor infilling", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 41.5 + } + ], + "page_idx": 23, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 298, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 298, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 72, + 504, + 95 + ], + "lines": [ + { + "bbox": [ + 105, + 71, + 505, + 86 + ], + "spans": [ + { + "bbox": [ + 105, + 71, + 505, + 86 + ], + "score": 1.0, + "content": "We follow the same photometric and geometric augmentation schemes from [70], comprising random", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 82, + 288, + 97 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 288, + 97 + ], + "score": 1.0, + "content": "affine transformation, flipping, and cropping.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5, + "bbox_fs": [ + 105, + 71, + 505, + 97 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 100, + 505, + 154 + ], + "lines": [ + { + "bbox": [ + 105, + 100, + 506, + 113 + ], + "spans": [ + { + "bbox": [ + 105, + 100, + 506, + 113 + ], + "score": 1.0, + "content": "Depth. For supervised pre-training of the indoor model we mix the following datasets. ScanNet", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 110, + 505, + 124 + ], + "spans": [ + { + "bbox": [ + 105, + 110, + 180, + 124 + ], + "score": 1.0, + "content": "[9] is a dataset of", + "type": "text" + }, + { + "bbox": [ + 180, + 111, + 204, + 122 + ], + "score": 0.46, + "content": "2 . 5 \\mathbf { M }", + "type": "inline_equation" + }, + { + "bbox": [ + 204, + 110, + 505, + 124 + ], + "score": 1.0, + "content": "examples captured using a Kinect v1-like sensor. It provides depth maps", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 121, + 505, + 133 + ], + "spans": [ + { + "bbox": [ + 105, + 121, + 117, + 133 + ], + "score": 1.0, + "content": "at", + "type": "text" + }, + { + "bbox": [ + 117, + 122, + 157, + 133 + ], + "score": 0.88, + "content": "4 8 0 \\times 6 4 0", + "type": "inline_equation" + }, + { + "bbox": [ + 157, + 121, + 242, + 133 + ], + "score": 1.0, + "content": "and RGB images at", + "type": "text" + }, + { + "bbox": [ + 243, + 122, + 287, + 132 + ], + "score": 0.89, + "content": "9 6 8 \\times 1 2 9 6", + "type": "inline_equation" + }, + { + "bbox": [ + 288, + 121, + 505, + 133 + ], + "score": 1.0, + "content": ". SceneNet RGB-D [41] is a synthetic dataset of 5M", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 132, + 506, + 145 + ], + "spans": [ + { + "bbox": [ + 105, + 132, + 506, + 145 + ], + "score": 1.0, + "content": "images generated by rendering ShapeNet [6] objects in scenes from SceneNet [20] at a resolution of", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 143, + 149, + 155 + ], + "spans": [ + { + "bbox": [ + 106, + 144, + 145, + 155 + ], + "score": 0.86, + "content": "2 4 0 \\times 3 2 0", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 143, + 149, + 155 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4, + "bbox_fs": [ + 105, + 100, + 506, + 155 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 159, + 505, + 204 + ], + "lines": [ + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "score": 1.0, + "content": "For the outdoor model training we use the Waymo Open Dataset [71], a large-scale driving dataset", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 171, + 505, + 184 + ], + "spans": [ + { + "bbox": [ + 105, + 171, + 187, + 184 + ], + "score": 1.0, + "content": "consisting of about", + "type": "text" + }, + { + "bbox": [ + 187, + 171, + 210, + 182 + ], + "score": 0.27, + "content": "2 0 0 \\mathrm { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 210, + 171, + 505, + 184 + ], + "score": 1.0, + "content": "frames. Each frame provides RGB images from 5 cameras and LiDAR", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 182, + 505, + 194 + ], + "spans": [ + { + "bbox": [ + 105, + 182, + 505, + 194 + ], + "score": 1.0, + "content": "maps. We use the RGB images from the FRONT, FRONT_LEFT and FRONT_RIGHT cameras and", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 192, + 380, + 206 + ], + "spans": [ + { + "bbox": [ + 105, + 192, + 380, + 206 + ], + "score": 1.0, + "content": "the TOP LiDAR only to build about 600k aligned RGB depth maps.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 8.5, + "bbox_fs": [ + 105, + 159, + 506, + 206 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 209, + 504, + 242 + ], + "lines": [ + { + "bbox": [ + 106, + 209, + 506, + 222 + ], + "spans": [ + { + "bbox": [ + 106, + 209, + 346, + 222 + ], + "score": 1.0, + "content": "For indoor fine-tuning and evaluation we use NYU depth", + "type": "text" + }, + { + "bbox": [ + 346, + 210, + 358, + 219 + ], + "score": 0.38, + "content": "\\nu 2", + "type": "inline_equation" + }, + { + "bbox": [ + 358, + 209, + 506, + 222 + ], + "score": 1.0, + "content": "[61], a commonly used dataset for", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 220, + 504, + 232 + ], + "spans": [ + { + "bbox": [ + 106, + 220, + 464, + 232 + ], + "score": 1.0, + "content": "evaluating indoor depth prediction models. It provides aligned image and depth maps at", + "type": "text" + }, + { + "bbox": [ + 464, + 220, + 504, + 231 + ], + "score": 0.88, + "content": "4 8 0 \\times 6 4 0", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 231, + 487, + 243 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 297, + 243 + ], + "score": 1.0, + "content": "resolution. We use the official split comprising", + "type": "text" + }, + { + "bbox": [ + 298, + 231, + 315, + 241 + ], + "score": 0.33, + "content": "5 0 \\mathrm { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 231, + 487, + 243 + ], + "score": 1.0, + "content": "images for training and 654 for evaluation.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12, + "bbox_fs": [ + 106, + 209, + 506, + 243 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 247, + 505, + 280 + ], + "lines": [ + { + "bbox": [ + 106, + 247, + 505, + 259 + ], + "spans": [ + { + "bbox": [ + 106, + 247, + 505, + 259 + ], + "score": 1.0, + "content": "For outdoor fine-tuning and evaluation, we use KITTI [17], an outdoor driving dataset which provides", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 257, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 105, + 257, + 333, + 271 + ], + "score": 1.0, + "content": "RGB images and LiDAR scans at resolutions close to", + "type": "text" + }, + { + "bbox": [ + 333, + 258, + 378, + 269 + ], + "score": 0.89, + "content": "3 7 0 \\times 1 2 2 6", + "type": "inline_equation" + }, + { + "bbox": [ + 378, + 257, + 505, + 271 + ], + "score": 1.0, + "content": ". We use the training/test split", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 269, + 392, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 269, + 392, + 282 + ], + "score": 1.0, + "content": "proposed by [13], comprising 26k training images and 652 test images.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15, + "bbox_fs": [ + 105, + 247, + 505, + 282 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 285, + 505, + 319 + ], + "lines": [ + { + "bbox": [ + 105, + 284, + 506, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 284, + 506, + 298 + ], + "score": 1.0, + "content": "We use random horizontal flip data augmentation which is common in prior work. Where needed,", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 297, + 505, + 309 + ], + "spans": [ + { + "bbox": [ + 106, + 297, + 505, + 309 + ], + "score": 1.0, + "content": "images and dense depth maps are resized using bilinear interpolation to the model’s resolution for", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 307, + 376, + 321 + ], + "spans": [ + { + "bbox": [ + 105, + 307, + 376, + 321 + ], + "score": 1.0, + "content": "training and nearest neighbor interpolation is used for sparse maps.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18, + "bbox_fs": [ + 105, + 284, + 506, + 321 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 336, + 381, + 348 + ], + "lines": [ + { + "bbox": [ + 105, + 335, + 383, + 352 + ], + "spans": [ + { + "bbox": [ + 105, + 335, + 383, + 352 + ], + "score": 1.0, + "content": "H.3 Step-unrolling and interpolation of missing depth and flow", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 356, + 505, + 389 + ], + "lines": [ + { + "bbox": [ + 105, + 355, + 505, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 355, + 505, + 369 + ], + "score": 1.0, + "content": "As discussed in Section 3.2 of the main paper, infilling and step-unrolling are used to mitigate", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 366, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 105, + 366, + 505, + 379 + ], + "score": 1.0, + "content": "distribution shift between training and inference with diffusion models. The problem arises due to", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 378, + 344, + 390 + ], + "spans": [ + { + "bbox": [ + 106, + 378, + 344, + 390 + ], + "score": 1.0, + "content": "the missing data in the training depth maps and flow fields.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22, + "bbox_fs": [ + 105, + 355, + 505, + 390 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 394, + 505, + 460 + ], + "lines": [ + { + "bbox": [ + 105, + 393, + 505, + 407 + ], + "spans": [ + { + "bbox": [ + 105, + 393, + 505, + 407 + ], + "score": 1.0, + "content": "Infilling. For indoor depth maps, we use nearest neighbor interpolation during training (see Section", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 404, + 505, + 417 + ], + "spans": [ + { + "bbox": [ + 105, + 404, + 505, + 417 + ], + "score": 1.0, + "content": "3.2 in the main paper). For the outdoor depth data we use nearest neighbor interpolation except for", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "score": 1.0, + "content": "sky regions, as they are often large and are much further from the camera than adjacent objects in the", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 428, + 505, + 439 + ], + "spans": [ + { + "bbox": [ + 106, + 428, + 505, + 439 + ], + "score": 1.0, + "content": "image. We use an off-the-shelf sky segmenter [36], and then set all sky pixels to be the maximum", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 437, + 506, + 451 + ], + "spans": [ + { + "bbox": [ + 105, + 437, + 193, + 451 + ], + "score": 1.0, + "content": "modeled depth (here,", + "type": "text" + }, + { + "bbox": [ + 194, + 438, + 214, + 449 + ], + "score": 0.5, + "content": "8 0 \\mathrm { m } \\mathrm { , }", + "type": "inline_equation" + }, + { + "bbox": [ + 214, + 437, + 506, + 451 + ], + "score": 1.0, + "content": "). For missing optical flow ground truth we employ a simple sequence of", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 449, + 413, + 461 + ], + "spans": [ + { + "bbox": [ + 106, + 449, + 413, + 461 + ], + "score": 1.0, + "content": "1D nearest neighbor interpolations first along rows, and then along columns.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 26.5, + "bbox_fs": [ + 105, + 393, + 506, + 461 + ] + }, + { + "type": "table", + "bbox": [ + 155, + 522, + 455, + 624 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 107, + 474, + 505, + 515 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 473, + 505, + 488 + ], + "spans": [ + { + "bbox": [ + 105, + 473, + 505, + 488 + ], + "score": 1.0, + "content": "Table 16: Ablation of the number of unroll steps for monocular depth estimation. Performs improves up", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 485, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 106, + 485, + 505, + 496 + ], + "score": 1.0, + "content": "to four steps of unrolling and plateaus thereafter. The models trained without infilling missing depth benefit", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 495, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 495, + 505, + 506 + ], + "score": 1.0, + "content": "more from a larger number of unroll steps, which is to be expected. Best and second best results are bolded and", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 504, + 194, + 517 + ], + "spans": [ + { + "bbox": [ + 105, + 504, + 194, + 517 + ], + "score": 1.0, + "content": "underlined respectively.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 31.5 + }, + { + "type": "table_body", + "bbox": [ + 155, + 522, + 455, + 624 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 155, + 522, + 455, + 624 + ], + "spans": [ + { + "bbox": [ + 155, + 522, + 455, + 624 + ], + "score": 0.982, + "html": "
NYUKITTI
No infillInfillNo infillInfill
Unroll stepsRELRMSRELRMSRELRMSRELRMS
00.0790.3310.0770.3380.2223.7700.0572.744
10.0760.3240.0750.3240.0852.8440.0562.700
20.0760.3170.0750.3150.0682.7990.0542.629
30.0750.3160.0740.3170.0612.7890.0542.591
40.0750.3160.0740.3150.0592.7390.0532.568
", + "type": "table", + "image_path": "578d7c0bf70f4c42b92f60b2593c6ab4ce127c04e43560ec728543a10196d35a.jpg" + } + ] + } + ], + "index": 35, + "virtual_lines": [ + { + "bbox": [ + 155, + 522, + 455, + 556.0 + ], + "spans": [], + "index": 34 + }, + { + "bbox": [ + 155, + 556.0, + 455, + 590.0 + ], + "spans": [], + "index": 35 + }, + { + "bbox": [ + 155, + 590.0, + 455, + 624.0 + ], + "spans": [], + "index": 36 + } + ] + } + ], + "index": 33.25 + }, + { + "type": "text", + "bbox": [ + 108, + 640, + 505, + 673 + ], + "lines": [ + { + "bbox": [ + 106, + 640, + 505, + 652 + ], + "spans": [ + { + "bbox": [ + 106, + 640, + 505, + 652 + ], + "score": 1.0, + "content": "Step-unrolling. By default we use a single unroll step in all results where step-unrolling is enabled.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 649, + 506, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 506, + 663 + ], + "score": 1.0, + "content": "In Table 16, we show that using multiple unroll steps can further improve performance on the task of", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 662, + 223, + 674 + ], + "spans": [ + { + "bbox": [ + 106, + 662, + 223, + 674 + ], + "score": 1.0, + "content": "monocular depth estimation.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38, + "bbox_fs": [ + 105, + 640, + 506, + 674 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 678, + 505, + 722 + ], + "lines": [ + { + "bbox": [ + 105, + 678, + 506, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 678, + 506, + 691 + ], + "score": 1.0, + "content": "Finally, while we use infilling and step-unrolling, there are other ways in which one might try to", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 689, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 689, + 506, + 702 + ], + "score": 1.0, + "content": "mitigate the problem. One such approach was taken by [46], which faced a similar problem when", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 700, + 506, + 713 + ], + "spans": [ + { + "bbox": [ + 105, + 700, + 506, + 713 + ], + "score": 1.0, + "content": "training a vector-quantizer on depth data. Their approach was to synthetically add more holes", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 711, + 505, + 724 + ], + "spans": [ + { + "bbox": [ + 105, + 711, + 505, + 724 + ], + "score": 1.0, + "content": "following a carefully chosen masking ratio. We prefer our approach since nearest neighbor infilling", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 71, + 505, + 86 + ], + "spans": [ + { + "bbox": [ + 105, + 71, + 505, + 86 + ], + "score": 1.0, + "content": "is hyper-parameter free and step-unrolled denoising diffusion could be more generally applicable to", + "type": "text", + "cross_page": true + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 84, + 220, + 96 + ], + "spans": [ + { + "bbox": [ + 106, + 84, + 220, + 96 + ], + "score": 1.0, + "content": "other tasks with sparse data.", + "type": "text", + "cross_page": true + } + ], + "index": 1 + } + ], + "index": 41.5, + "bbox_fs": [ + 105, + 678, + 506, + 724 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 72, + 504, + 95 + ], + "lines": [ + { + "bbox": [ + 105, + 71, + 505, + 86 + ], + "spans": [ + { + "bbox": [ + 105, + 71, + 505, + 86 + ], + "score": 1.0, + "content": "is hyper-parameter free and step-unrolled denoising diffusion could be more generally applicable to", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 84, + 220, + 96 + ], + "spans": [ + { + "bbox": [ + 106, + 84, + 220, + 96 + ], + "score": 1.0, + "content": "other tasks with sparse data.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "table", + "bbox": [ + 142, + 120, + 468, + 204 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 181, + 104, + 429, + 116 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 181, + 102, + 429, + 118 + ], + "spans": [ + { + "bbox": [ + 181, + 102, + 429, + 118 + ], + "score": 1.0, + "content": "Table 17: Comparison of step-unrolling and self-conditioning [8]", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 2 + }, + { + "type": "table_body", + "bbox": [ + 142, + 120, + 468, + 204 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 142, + 120, + 468, + 204 + ], + "spans": [ + { + "bbox": [ + 142, + 120, + 468, + 204 + ], + "score": 0.982, + "html": "
NYUKITTI
No infillInfillNo infillInfill
RELRMSRELRMSRELRMSRELRMS
Baseline0.0790.3310.0770.3380.2223.7700.0572.744
Self conditioning0.0820.3350.0810.3330.2423.9400.0572.761
Step unrolling0.0760.3240.0750.3240.0852.8440.0562.700
", + "type": "table", + "image_path": "caa0c683b334777009b383a9a6fb30fc18eee9f6bf09529cc6d2cd01fc396188.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 142, + 120, + 468, + 148.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 142, + 148.0, + 468, + 176.0 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 142, + 176.0, + 468, + 204.0 + ], + "spans": [], + "index": 5 + } + ] + } + ], + "index": 3.0 + }, + { + "type": "text", + "bbox": [ + 107, + 214, + 505, + 269 + ], + "lines": [ + { + "bbox": [ + 106, + 213, + 506, + 227 + ], + "spans": [ + { + "bbox": [ + 106, + 213, + 506, + 227 + ], + "score": 1.0, + "content": "We also considered the approach of self-conditioning [8] as an alternative to step-unrolling. However,", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 225, + 505, + 238 + ], + "spans": [ + { + "bbox": [ + 105, + 225, + 505, + 238 + ], + "score": 1.0, + "content": "as we show in Table 17, we find that self-conditioning is unable to bridge the train-inference", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 235, + 505, + 249 + ], + "spans": [ + { + "bbox": [ + 105, + 235, + 505, + 249 + ], + "score": 1.0, + "content": "distribution shift of the noisy latent for the task of monocular depth estimation. This is specially", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 246, + 506, + 260 + ], + "spans": [ + { + "bbox": [ + 105, + 246, + 506, + 260 + ], + "score": 1.0, + "content": "apparent in the results for KITTI without infilling where self conditioning leads to no improvement", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 257, + 347, + 272 + ], + "spans": [ + { + "bbox": [ + 105, + 257, + 347, + 272 + ], + "score": 1.0, + "content": "whereas step-unrolling substantially improves performance.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 8 + }, + { + "type": "title", + "bbox": [ + 107, + 281, + 212, + 294 + ], + "lines": [ + { + "bbox": [ + 105, + 280, + 214, + 297 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 214, + 297 + ], + "score": 1.0, + "content": "H.4 Hyper-parameters", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 106, + 298, + 504, + 321 + ], + "lines": [ + { + "bbox": [ + 106, + 298, + 505, + 311 + ], + "spans": [ + { + "bbox": [ + 106, + 298, + 332, + 311 + ], + "score": 1.0, + "content": "Self-supervised. The self-supervised model is trained for", + "type": "text" + }, + { + "bbox": [ + 332, + 299, + 355, + 309 + ], + "score": 0.36, + "content": "2 . 8 \\mathbf { M }", + "type": "inline_equation" + }, + { + "bbox": [ + 355, + 298, + 407, + 311 + ], + "score": 1.0, + "content": "steps with an", + "type": "text" + }, + { + "bbox": [ + 407, + 299, + 420, + 310 + ], + "score": 0.88, + "content": "L _ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 420, + 298, + 505, + 311 + ], + "score": 1.0, + "content": "loss and a mini-batch", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 310, + 459, + 323 + ], + "spans": [ + { + "bbox": [ + 106, + 310, + 459, + 323 + ], + "score": 1.0, + "content": "size of 512. Other hyper-parameters are same as those in the original Palette paper [54].", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12.5 + }, + { + "type": "text", + "bbox": [ + 107, + 326, + 505, + 403 + ], + "lines": [ + { + "bbox": [ + 106, + 326, + 505, + 339 + ], + "spans": [ + { + "bbox": [ + 106, + 326, + 389, + 339 + ], + "score": 1.0, + "content": "Supervised. The supervised flow and depth models are trained with", + "type": "text" + }, + { + "bbox": [ + 389, + 327, + 402, + 337 + ], + "score": 0.88, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 402, + 326, + 505, + 339 + ], + "score": 1.0, + "content": "loss. Usually a constant", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 336, + 506, + 350 + ], + "spans": [ + { + "bbox": [ + 105, + 336, + 169, + 350 + ], + "score": 1.0, + "content": "learning rate of", + "type": "text" + }, + { + "bbox": [ + 169, + 336, + 208, + 348 + ], + "score": 0.92, + "content": "1 \\times \\dot { 1 } 0 ^ { - 4 }", + "type": "inline_equation" + }, + { + "bbox": [ + 209, + 336, + 506, + 350 + ], + "score": 1.0, + "content": "with a warm-up over 10k steps is used. However, for depth fine-tuning we", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 347, + 506, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 241, + 361 + ], + "score": 1.0, + "content": "find that a lower learning rate of", + "type": "text" + }, + { + "bbox": [ + 241, + 348, + 280, + 359 + ], + "score": 0.92, + "content": "3 \\times 1 0 ^ { - 5 }", + "type": "inline_equation" + }, + { + "bbox": [ + 281, + 347, + 506, + 361 + ], + "score": 1.0, + "content": "achieves slightly better results. All models are trained", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 359, + 505, + 371 + ], + "spans": [ + { + "bbox": [ + 105, + 359, + 505, + 371 + ], + "score": 1.0, + "content": "with a mini-batch size of 64. The indoor depth model is pre-trained for 2M steps and then fine-tuned", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 370, + 505, + 382 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 159, + 382 + ], + "score": 1.0, + "content": "on NYU for", + "type": "text" + }, + { + "bbox": [ + 159, + 370, + 176, + 380 + ], + "score": 0.37, + "content": "4 0 \\mathrm { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 177, + 370, + 383, + 382 + ], + "score": 1.0, + "content": "steps. The outdoor depth model is pre-trained for", + "type": "text" + }, + { + "bbox": [ + 383, + 370, + 407, + 380 + ], + "score": 0.36, + "content": "0 . 9 { \\bf M }", + "type": "inline_equation" + }, + { + "bbox": [ + 407, + 370, + 505, + 382 + ], + "score": 1.0, + "content": "steps and fine-tuned on", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 380, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 380, + 286, + 394 + ], + "score": 1.0, + "content": "KITTI for 40k steps. For flow, we pretrain for", + "type": "text" + }, + { + "bbox": [ + 287, + 381, + 309, + 392 + ], + "score": 0.26, + "content": "3 . 7 \\mathbf { M }", + "type": "inline_equation" + }, + { + "bbox": [ + 310, + 380, + 506, + 394 + ], + "score": 1.0, + "content": "steps, followed by finetuning for 50k steps. Other", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 392, + 378, + 404 + ], + "spans": [ + { + "bbox": [ + 105, + 392, + 378, + 404 + ], + "score": 1.0, + "content": "details, like the optimizer and the use of EMA are the same as [54].", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 17 + }, + { + "type": "title", + "bbox": [ + 107, + 415, + 174, + 427 + ], + "lines": [ + { + "bbox": [ + 105, + 414, + 175, + 429 + ], + "spans": [ + { + "bbox": [ + 105, + 414, + 175, + 429 + ], + "score": 1.0, + "content": "H.5 Inference", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 21 + }, + { + "type": "text", + "bbox": [ + 107, + 432, + 505, + 466 + ], + "lines": [ + { + "bbox": [ + 105, + 432, + 505, + 446 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 505, + 446 + ], + "score": 1.0, + "content": "Sampler. We use the DDPM ancestral sampler [21] with 128 denoising steps for monocular depth", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 444, + 505, + 456 + ], + "spans": [ + { + "bbox": [ + 106, + 444, + 505, + 456 + ], + "score": 1.0, + "content": "models and 64 steps for optical flow models. Increasing the number of denoising steps further did not", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 455, + 228, + 468 + ], + "spans": [ + { + "bbox": [ + 106, + 455, + 228, + 468 + ], + "score": 1.0, + "content": "greatly improve performance.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 23 + }, + { + "type": "text", + "bbox": [ + 107, + 471, + 505, + 504 + ], + "lines": [ + { + "bbox": [ + 106, + 470, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 255, + 484 + ], + "score": 1.0, + "content": "Coarse-to-fine refinement. We use", + "type": "text" + }, + { + "bbox": [ + 255, + 471, + 275, + 482 + ], + "score": 0.87, + "content": "2 \\times 5", + "type": "inline_equation" + }, + { + "bbox": [ + 276, + 470, + 424, + 484 + ], + "score": 1.0, + "content": "overlapping patches ({top, bottom}", + "type": "text" + }, + { + "bbox": [ + 424, + 472, + 434, + 482 + ], + "score": 0.57, + "content": "\\times", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 470, + 506, + 484 + ], + "score": 1.0, + "content": "{left, center-left,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 482, + 505, + 494 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 400, + 494 + ], + "score": 1.0, + "content": "center, center-right, right}) for coarse-to-fine refinement. For Sintel we use", + "type": "text" + }, + { + "bbox": [ + 400, + 482, + 446, + 494 + ], + "score": 0.88, + "content": "t ^ { \\prime } = 3 2 / 6 4", + "type": "inline_equation" + }, + { + "bbox": [ + 447, + 482, + 505, + 494 + ], + "score": 1.0, + "content": "and for KITTI", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 107, + 491, + 152, + 506 + ], + "spans": [ + { + "bbox": [ + 107, + 493, + 147, + 505 + ], + "score": 0.91, + "content": "t ^ { \\prime } = 8 / 6 4", + "type": "inline_equation" + }, + { + "bbox": [ + 148, + 491, + 152, + 506 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26 + }, + { + "type": "title", + "bbox": [ + 106, + 519, + 183, + 533 + ], + "lines": [ + { + "bbox": [ + 104, + 517, + 185, + 536 + ], + "spans": [ + { + "bbox": [ + 104, + 517, + 185, + 536 + ], + "score": 1.0, + "content": "I Limitations", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "table", + "bbox": [ + 106, + 608, + 506, + 687 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 540, + 505, + 601 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 541, + 504, + 551 + ], + "spans": [ + { + "bbox": [ + 106, + 541, + 504, + 551 + ], + "score": 1.0, + "content": "Table 18: Inference speed comparison of our method with DPT [50] on the indoor depth model finetuned on", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 550, + 505, + 562 + ], + "spans": [ + { + "bbox": [ + 105, + 550, + 505, + 562 + ], + "score": 1.0, + "content": "NYU. Diffusion model inference is bottlenecked by the large number of denoising steps. We show that some", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 104, + 559, + 506, + 574 + ], + "spans": [ + { + "bbox": [ + 104, + 559, + 506, + 574 + ], + "score": 1.0, + "content": "efficiency gains can be achieved by simply reducing the number of denoising steps. Our model with 24 denoising", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 570, + 506, + 582 + ], + "spans": [ + { + "bbox": [ + 105, + 570, + 305, + 582 + ], + "score": 1.0, + "content": "steps is comparable in performance to DPT while being", + "type": "text" + }, + { + "bbox": [ + 305, + 571, + 322, + 580 + ], + "score": 0.86, + "content": "{ \\sim } 5 \\mathbf { x }", + "type": "inline_equation" + }, + { + "bbox": [ + 323, + 570, + 506, + 582 + ], + "score": 1.0, + "content": "slower (modulo differences in hardware). * we use", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 581, + 506, + 592 + ], + "spans": [ + { + "bbox": [ + 106, + 581, + 310, + 592 + ], + "score": 1.0, + "content": "the step-time reported in the DPT paper at a resolution of", + "type": "text" + }, + { + "bbox": [ + 311, + 581, + 350, + 590 + ], + "score": 0.9, + "content": "3 8 4 \\times 3 8 4", + "type": "inline_equation" + }, + { + "bbox": [ + 351, + 581, + 506, + 592 + ], + "score": 1.0, + "content": ", however, the DPT performance metrics on", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 104, + 589, + 462, + 603 + ], + "spans": [ + { + "bbox": [ + 104, + 589, + 280, + 603 + ], + "score": 1.0, + "content": "NYU are with a model trained at a resolution of", + "type": "text" + }, + { + "bbox": [ + 281, + 591, + 320, + 600 + ], + "score": 0.9, + "content": "4 8 0 \\times 6 4 0", + "type": "inline_equation" + }, + { + "bbox": [ + 321, + 589, + 462, + 603 + ], + "score": 1.0, + "content": ", for which the step time will be higher.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 31.5 + }, + { + "type": "table_body", + "bbox": [ + 106, + 608, + 506, + 687 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 608, + 506, + 687 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 506, + 687 + ], + "score": 0.98, + "html": "
MethodArchitectureResolutionTotal Time [ms]Inference stepsREL↓RMS↓
DPT-HybridNvidia RTX 2080384×384*38*-0.1100.357
DDVMTPU v4240 × 320204240.1040.378
272320.0860.342
544640.0770.324
10891280.0750.324
", + "type": "table", + "image_path": "a9aed3a5b37e9fd4a75c4c175c1455e860b63da545124d03047e88c8f44b02b8.jpg" + } + ] + } + ], + "index": 36, + "virtual_lines": [ + { + "bbox": [ + 106, + 608, + 506, + 634.3333333333334 + ], + "spans": [], + "index": 35 + }, + { + "bbox": [ + 106, + 634.3333333333334, + 506, + 660.6666666666667 + ], + "spans": [], + "index": 36 + }, + { + "bbox": [ + 106, + 660.6666666666667, + 506, + 687.0000000000001 + ], + "spans": [], + "index": 37 + } + ] + } + ], + "index": 33.75 + }, + { + "type": "text", + "bbox": [ + 106, + 699, + 504, + 722 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 505, + 713 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 713 + ], + "score": 1.0, + "content": "Efficiency. Inference speed with diffusion models is a well-known issue, as multiple denoising steps", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 711, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 506, + 723 + ], + "score": 1.0, + "content": "are used to transform noise to a target signal. This can be prohibitive for vision tasks where near", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5 + } + ], + "page_idx": 24, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 298, + 740, + 313, + 754 + ], + "spans": [ + { + "bbox": [ + 298, + 740, + 313, + 754 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 72, + 504, + 95 + ], + "lines": [], + "index": 0.5, + "bbox_fs": [ + 105, + 71, + 505, + 96 + ], + "lines_deleted": true + }, + { + "type": "table", + "bbox": [ + 142, + 120, + 468, + 204 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 181, + 104, + 429, + 116 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 181, + 102, + 429, + 118 + ], + "spans": [ + { + "bbox": [ + 181, + 102, + 429, + 118 + ], + "score": 1.0, + "content": "Table 17: Comparison of step-unrolling and self-conditioning [8]", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 2 + }, + { + "type": "table_body", + "bbox": [ + 142, + 120, + 468, + 204 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 142, + 120, + 468, + 204 + ], + "spans": [ + { + "bbox": [ + 142, + 120, + 468, + 204 + ], + "score": 0.982, + "html": "
NYUKITTI
No infillInfillNo infillInfill
RELRMSRELRMSRELRMSRELRMS
Baseline0.0790.3310.0770.3380.2223.7700.0572.744
Self conditioning0.0820.3350.0810.3330.2423.9400.0572.761
Step unrolling0.0760.3240.0750.3240.0852.8440.0562.700
", + "type": "table", + "image_path": "caa0c683b334777009b383a9a6fb30fc18eee9f6bf09529cc6d2cd01fc396188.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 142, + 120, + 468, + 148.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 142, + 148.0, + 468, + 176.0 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 142, + 176.0, + 468, + 204.0 + ], + "spans": [], + "index": 5 + } + ] + } + ], + "index": 3.0 + }, + { + "type": "text", + "bbox": [ + 107, + 214, + 505, + 269 + ], + "lines": [ + { + "bbox": [ + 106, + 213, + 506, + 227 + ], + "spans": [ + { + "bbox": [ + 106, + 213, + 506, + 227 + ], + "score": 1.0, + "content": "We also considered the approach of self-conditioning [8] as an alternative to step-unrolling. However,", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 225, + 505, + 238 + ], + "spans": [ + { + "bbox": [ + 105, + 225, + 505, + 238 + ], + "score": 1.0, + "content": "as we show in Table 17, we find that self-conditioning is unable to bridge the train-inference", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 235, + 505, + 249 + ], + "spans": [ + { + "bbox": [ + 105, + 235, + 505, + 249 + ], + "score": 1.0, + "content": "distribution shift of the noisy latent for the task of monocular depth estimation. This is specially", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 246, + 506, + 260 + ], + "spans": [ + { + "bbox": [ + 105, + 246, + 506, + 260 + ], + "score": 1.0, + "content": "apparent in the results for KITTI without infilling where self conditioning leads to no improvement", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 257, + 347, + 272 + ], + "spans": [ + { + "bbox": [ + 105, + 257, + 347, + 272 + ], + "score": 1.0, + "content": "whereas step-unrolling substantially improves performance.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 8, + "bbox_fs": [ + 105, + 213, + 506, + 272 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 281, + 212, + 294 + ], + "lines": [ + { + "bbox": [ + 105, + 280, + 214, + 297 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 214, + 297 + ], + "score": 1.0, + "content": "H.4 Hyper-parameters", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 106, + 298, + 504, + 321 + ], + "lines": [ + { + "bbox": [ + 106, + 298, + 505, + 311 + ], + "spans": [ + { + "bbox": [ + 106, + 298, + 332, + 311 + ], + "score": 1.0, + "content": "Self-supervised. The self-supervised model is trained for", + "type": "text" + }, + { + "bbox": [ + 332, + 299, + 355, + 309 + ], + "score": 0.36, + "content": "2 . 8 \\mathbf { M }", + "type": "inline_equation" + }, + { + "bbox": [ + 355, + 298, + 407, + 311 + ], + "score": 1.0, + "content": "steps with an", + "type": "text" + }, + { + "bbox": [ + 407, + 299, + 420, + 310 + ], + "score": 0.88, + "content": "L _ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 420, + 298, + 505, + 311 + ], + "score": 1.0, + "content": "loss and a mini-batch", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 310, + 459, + 323 + ], + "spans": [ + { + "bbox": [ + 106, + 310, + 459, + 323 + ], + "score": 1.0, + "content": "size of 512. Other hyper-parameters are same as those in the original Palette paper [54].", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12.5, + "bbox_fs": [ + 106, + 298, + 505, + 323 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 326, + 505, + 403 + ], + "lines": [ + { + "bbox": [ + 106, + 326, + 505, + 339 + ], + "spans": [ + { + "bbox": [ + 106, + 326, + 389, + 339 + ], + "score": 1.0, + "content": "Supervised. The supervised flow and depth models are trained with", + "type": "text" + }, + { + "bbox": [ + 389, + 327, + 402, + 337 + ], + "score": 0.88, + "content": "L _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 402, + 326, + 505, + 339 + ], + "score": 1.0, + "content": "loss. Usually a constant", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 336, + 506, + 350 + ], + "spans": [ + { + "bbox": [ + 105, + 336, + 169, + 350 + ], + "score": 1.0, + "content": "learning rate of", + "type": "text" + }, + { + "bbox": [ + 169, + 336, + 208, + 348 + ], + "score": 0.92, + "content": "1 \\times \\dot { 1 } 0 ^ { - 4 }", + "type": "inline_equation" + }, + { + "bbox": [ + 209, + 336, + 506, + 350 + ], + "score": 1.0, + "content": "with a warm-up over 10k steps is used. However, for depth fine-tuning we", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 347, + 506, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 241, + 361 + ], + "score": 1.0, + "content": "find that a lower learning rate of", + "type": "text" + }, + { + "bbox": [ + 241, + 348, + 280, + 359 + ], + "score": 0.92, + "content": "3 \\times 1 0 ^ { - 5 }", + "type": "inline_equation" + }, + { + "bbox": [ + 281, + 347, + 506, + 361 + ], + "score": 1.0, + "content": "achieves slightly better results. All models are trained", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 359, + 505, + 371 + ], + "spans": [ + { + "bbox": [ + 105, + 359, + 505, + 371 + ], + "score": 1.0, + "content": "with a mini-batch size of 64. The indoor depth model is pre-trained for 2M steps and then fine-tuned", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 370, + 505, + 382 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 159, + 382 + ], + "score": 1.0, + "content": "on NYU for", + "type": "text" + }, + { + "bbox": [ + 159, + 370, + 176, + 380 + ], + "score": 0.37, + "content": "4 0 \\mathrm { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 177, + 370, + 383, + 382 + ], + "score": 1.0, + "content": "steps. The outdoor depth model is pre-trained for", + "type": "text" + }, + { + "bbox": [ + 383, + 370, + 407, + 380 + ], + "score": 0.36, + "content": "0 . 9 { \\bf M }", + "type": "inline_equation" + }, + { + "bbox": [ + 407, + 370, + 505, + 382 + ], + "score": 1.0, + "content": "steps and fine-tuned on", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 380, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 380, + 286, + 394 + ], + "score": 1.0, + "content": "KITTI for 40k steps. For flow, we pretrain for", + "type": "text" + }, + { + "bbox": [ + 287, + 381, + 309, + 392 + ], + "score": 0.26, + "content": "3 . 7 \\mathbf { M }", + "type": "inline_equation" + }, + { + "bbox": [ + 310, + 380, + 506, + 394 + ], + "score": 1.0, + "content": "steps, followed by finetuning for 50k steps. Other", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 392, + 378, + 404 + ], + "spans": [ + { + "bbox": [ + 105, + 392, + 378, + 404 + ], + "score": 1.0, + "content": "details, like the optimizer and the use of EMA are the same as [54].", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 17, + "bbox_fs": [ + 105, + 326, + 506, + 404 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 415, + 174, + 427 + ], + "lines": [ + { + "bbox": [ + 105, + 414, + 175, + 429 + ], + "spans": [ + { + "bbox": [ + 105, + 414, + 175, + 429 + ], + "score": 1.0, + "content": "H.5 Inference", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 21 + }, + { + "type": "text", + "bbox": [ + 107, + 432, + 505, + 466 + ], + "lines": [ + { + "bbox": [ + 105, + 432, + 505, + 446 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 505, + 446 + ], + "score": 1.0, + "content": "Sampler. We use the DDPM ancestral sampler [21] with 128 denoising steps for monocular depth", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 444, + 505, + 456 + ], + "spans": [ + { + "bbox": [ + 106, + 444, + 505, + 456 + ], + "score": 1.0, + "content": "models and 64 steps for optical flow models. Increasing the number of denoising steps further did not", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 455, + 228, + 468 + ], + "spans": [ + { + "bbox": [ + 106, + 455, + 228, + 468 + ], + "score": 1.0, + "content": "greatly improve performance.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 23, + "bbox_fs": [ + 105, + 432, + 505, + 468 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 471, + 505, + 504 + ], + "lines": [ + { + "bbox": [ + 106, + 470, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 255, + 484 + ], + "score": 1.0, + "content": "Coarse-to-fine refinement. We use", + "type": "text" + }, + { + "bbox": [ + 255, + 471, + 275, + 482 + ], + "score": 0.87, + "content": "2 \\times 5", + "type": "inline_equation" + }, + { + "bbox": [ + 276, + 470, + 424, + 484 + ], + "score": 1.0, + "content": "overlapping patches ({top, bottom}", + "type": "text" + }, + { + "bbox": [ + 424, + 472, + 434, + 482 + ], + "score": 0.57, + "content": "\\times", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 470, + 506, + 484 + ], + "score": 1.0, + "content": "{left, center-left,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 482, + 505, + 494 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 400, + 494 + ], + "score": 1.0, + "content": "center, center-right, right}) for coarse-to-fine refinement. For Sintel we use", + "type": "text" + }, + { + "bbox": [ + 400, + 482, + 446, + 494 + ], + "score": 0.88, + "content": "t ^ { \\prime } = 3 2 / 6 4", + "type": "inline_equation" + }, + { + "bbox": [ + 447, + 482, + 505, + 494 + ], + "score": 1.0, + "content": "and for KITTI", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 107, + 491, + 152, + 506 + ], + "spans": [ + { + "bbox": [ + 107, + 493, + 147, + 505 + ], + "score": 0.91, + "content": "t ^ { \\prime } = 8 / 6 4", + "type": "inline_equation" + }, + { + "bbox": [ + 148, + 491, + 152, + 506 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26, + "bbox_fs": [ + 105, + 470, + 506, + 506 + ] + }, + { + "type": "title", + "bbox": [ + 106, + 519, + 183, + 533 + ], + "lines": [ + { + "bbox": [ + 104, + 517, + 185, + 536 + ], + "spans": [ + { + "bbox": [ + 104, + 517, + 185, + 536 + ], + "score": 1.0, + "content": "I Limitations", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "table", + "bbox": [ + 106, + 608, + 506, + 687 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 540, + 505, + 601 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 541, + 504, + 551 + ], + "spans": [ + { + "bbox": [ + 106, + 541, + 504, + 551 + ], + "score": 1.0, + "content": "Table 18: Inference speed comparison of our method with DPT [50] on the indoor depth model finetuned on", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 550, + 505, + 562 + ], + "spans": [ + { + "bbox": [ + 105, + 550, + 505, + 562 + ], + "score": 1.0, + "content": "NYU. Diffusion model inference is bottlenecked by the large number of denoising steps. We show that some", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 104, + 559, + 506, + 574 + ], + "spans": [ + { + "bbox": [ + 104, + 559, + 506, + 574 + ], + "score": 1.0, + "content": "efficiency gains can be achieved by simply reducing the number of denoising steps. Our model with 24 denoising", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 570, + 506, + 582 + ], + "spans": [ + { + "bbox": [ + 105, + 570, + 305, + 582 + ], + "score": 1.0, + "content": "steps is comparable in performance to DPT while being", + "type": "text" + }, + { + "bbox": [ + 305, + 571, + 322, + 580 + ], + "score": 0.86, + "content": "{ \\sim } 5 \\mathbf { x }", + "type": "inline_equation" + }, + { + "bbox": [ + 323, + 570, + 506, + 582 + ], + "score": 1.0, + "content": "slower (modulo differences in hardware). * we use", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 581, + 506, + 592 + ], + "spans": [ + { + "bbox": [ + 106, + 581, + 310, + 592 + ], + "score": 1.0, + "content": "the step-time reported in the DPT paper at a resolution of", + "type": "text" + }, + { + "bbox": [ + 311, + 581, + 350, + 590 + ], + "score": 0.9, + "content": "3 8 4 \\times 3 8 4", + "type": "inline_equation" + }, + { + "bbox": [ + 351, + 581, + 506, + 592 + ], + "score": 1.0, + "content": ", however, the DPT performance metrics on", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 104, + 589, + 462, + 603 + ], + "spans": [ + { + "bbox": [ + 104, + 589, + 280, + 603 + ], + "score": 1.0, + "content": "NYU are with a model trained at a resolution of", + "type": "text" + }, + { + "bbox": [ + 281, + 591, + 320, + 600 + ], + "score": 0.9, + "content": "4 8 0 \\times 6 4 0", + "type": "inline_equation" + }, + { + "bbox": [ + 321, + 589, + 462, + 603 + ], + "score": 1.0, + "content": ", for which the step time will be higher.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 31.5 + }, + { + "type": "table_body", + "bbox": [ + 106, + 608, + 506, + 687 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 608, + 506, + 687 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 506, + 687 + ], + "score": 0.98, + "html": "
MethodArchitectureResolutionTotal Time [ms]Inference stepsREL↓RMS↓
DPT-HybridNvidia RTX 2080384×384*38*-0.1100.357
DDVMTPU v4240 × 320204240.1040.378
272320.0860.342
544640.0770.324
10891280.0750.324
", + "type": "table", + "image_path": "a9aed3a5b37e9fd4a75c4c175c1455e860b63da545124d03047e88c8f44b02b8.jpg" + } + ] + } + ], + "index": 36, + "virtual_lines": [ + { + "bbox": [ + 106, + 608, + 506, + 634.3333333333334 + ], + "spans": [], + "index": 35 + }, + { + "bbox": [ + 106, + 634.3333333333334, + 506, + 660.6666666666667 + ], + "spans": [], + "index": 36 + }, + { + "bbox": [ + 106, + 660.6666666666667, + 506, + 687.0000000000001 + ], + "spans": [], + "index": 37 + } + ] + } + ], + "index": 33.75 + }, + { + "type": "text", + "bbox": [ + 106, + 699, + 504, + 722 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 505, + 713 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 713 + ], + "score": 1.0, + "content": "Efficiency. Inference speed with diffusion models is a well-known issue, as multiple denoising steps", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 711, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 711, + 506, + 723 + ], + "score": 1.0, + "content": "are used to transform noise to a target signal. This can be prohibitive for vision tasks where near", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 73, + 505, + 84 + ], + "spans": [ + { + "bbox": [ + 106, + 73, + 505, + 84 + ], + "score": 1.0, + "content": "real-time latency is often desired. Table 18 compares the inference speed of our diffusion model for", + "type": "text", + "cross_page": true + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 82, + 506, + 98 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 394, + 98 + ], + "score": 1.0, + "content": "depth against DPT [50]. Despite having an efficient denoiser backbone (", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 394, + 84, + 415, + 94 + ], + "score": 0.7, + "content": "{ \\sim } 8 . 5", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 415, + 82, + 506, + 98 + ], + "score": 1.0, + "content": "ms per denoising step", + "type": "text", + "cross_page": true + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 94, + 506, + 107 + ], + "spans": [ + { + "bbox": [ + 105, + 94, + 506, + 107 + ], + "score": 1.0, + "content": "on a TPU v4), the diffusion model is considerably slower than DPT in total wall time. The most", + "type": "text", + "cross_page": true + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 106, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 106, + 106, + 505, + 118 + ], + "score": 1.0, + "content": "obvious way to reduce inference latency is to reduce the number of denoising steps. This can be", + "type": "text", + "cross_page": true + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 116, + 505, + 129 + ], + "spans": [ + { + "bbox": [ + 105, + 116, + 505, + 129 + ], + "score": 1.0, + "content": "done with only moderate reduction in performance. As shown in Table 18, we perform comparably", + "type": "text", + "cross_page": true + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 127, + 505, + 140 + ], + "spans": [ + { + "bbox": [ + 105, + 127, + 505, + 140 + ], + "score": 1.0, + "content": "with DPT with as few as 24 denoising steps. However, a more thorough study into optimizing the", + "type": "text", + "cross_page": true + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 138, + 506, + 150 + ], + "spans": [ + { + "bbox": [ + 105, + 138, + 506, + 150 + ], + "score": 1.0, + "content": "inference speed of these models while preserving the generation quality is warranted. With the use of", + "type": "text", + "cross_page": true + } + ], + "index": 6 + }, + { + "bbox": [ + 104, + 149, + 506, + 162 + ], + "spans": [ + { + "bbox": [ + 104, + 149, + 506, + 162 + ], + "score": 1.0, + "content": "progressive distillation [42, 56] it is likely possible to reduce latency even further, as this approach", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 159, + 505, + 173 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 505, + 173 + ], + "score": 1.0, + "content": "has been shown to successfully distill generative image models with over 1000 denoising steps into", + "type": "text", + "cross_page": true + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 169, + 208, + 184 + ], + "spans": [ + { + "bbox": [ + 105, + 169, + 208, + 184 + ], + "score": 1.0, + "content": "those with just 2-4 steps.", + "type": "text", + "cross_page": true + } + ], + "index": 9 + } + ], + "index": 38.5, + "bbox_fs": [ + 105, + 698, + 506, + 723 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 72, + 505, + 182 + ], + "lines": [ + { + "bbox": [ + 106, + 73, + 505, + 84 + ], + "spans": [ + { + "bbox": [ + 106, + 73, + 505, + 84 + ], + "score": 1.0, + "content": "real-time latency is often desired. Table 18 compares the inference speed of our diffusion model for", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 82, + 506, + 98 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 394, + 98 + ], + "score": 1.0, + "content": "depth against DPT [50]. Despite having an efficient denoiser backbone (", + "type": "text" + }, + { + "bbox": [ + 394, + 84, + 415, + 94 + ], + "score": 0.7, + "content": "{ \\sim } 8 . 5", + "type": "inline_equation" + }, + { + "bbox": [ + 415, + 82, + 506, + 98 + ], + "score": 1.0, + "content": "ms per denoising step", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 94, + 506, + 107 + ], + "spans": [ + { + "bbox": [ + 105, + 94, + 506, + 107 + ], + "score": 1.0, + "content": "on a TPU v4), the diffusion model is considerably slower than DPT in total wall time. The most", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 106, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 106, + 106, + 505, + 118 + ], + "score": 1.0, + "content": "obvious way to reduce inference latency is to reduce the number of denoising steps. This can be", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 116, + 505, + 129 + ], + "spans": [ + { + "bbox": [ + 105, + 116, + 505, + 129 + ], + "score": 1.0, + "content": "done with only moderate reduction in performance. As shown in Table 18, we perform comparably", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 127, + 505, + 140 + ], + "spans": [ + { + "bbox": [ + 105, + 127, + 505, + 140 + ], + "score": 1.0, + "content": "with DPT with as few as 24 denoising steps. However, a more thorough study into optimizing the", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 138, + 506, + 150 + ], + "spans": [ + { + "bbox": [ + 105, + 138, + 506, + 150 + ], + "score": 1.0, + "content": "inference speed of these models while preserving the generation quality is warranted. With the use of", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 104, + 149, + 506, + 162 + ], + "spans": [ + { + "bbox": [ + 104, + 149, + 506, + 162 + ], + "score": 1.0, + "content": "progressive distillation [42, 56] it is likely possible to reduce latency even further, as this approach", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 159, + 505, + 173 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 505, + 173 + ], + "score": 1.0, + "content": "has been shown to successfully distill generative image models with over 1000 denoising steps into", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 169, + 208, + 184 + ], + "spans": [ + { + "bbox": [ + 105, + 169, + 208, + 184 + ], + "score": 1.0, + "content": "those with just 2-4 steps.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 4.5 + }, + { + "type": "text", + "bbox": [ + 107, + 187, + 505, + 221 + ], + "lines": [ + { + "bbox": [ + 105, + 186, + 506, + 200 + ], + "spans": [ + { + "bbox": [ + 105, + 186, + 506, + 200 + ], + "score": 1.0, + "content": "Fine-tuning on Sintel. In Section 5 we discuss possible reasons for why our model’s superior zero-", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 198, + 505, + 210 + ], + "spans": [ + { + "bbox": [ + 106, + 198, + 505, + 210 + ], + "score": 1.0, + "content": "shot performance compared to FlowFormer [23] does not transfer to fine-tuning on Sintel. Figure 16", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 209, + 346, + 222 + ], + "spans": [ + { + "bbox": [ + 105, + 209, + 346, + 222 + ], + "score": 1.0, + "content": "provides qualitative examples to further support the claims.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11 + }, + { + "type": "image", + "bbox": [ + 110, + 231, + 498, + 591 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 110, + 231, + 498, + 591 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 110, + 231, + 498, + 591 + ], + "spans": [ + { + "bbox": [ + 110, + 231, + 498, + 591 + ], + "score": 0.976, + "type": "image", + "image_path": "1934e1d086049ece1ca2800e7b4682ae30439a2fdd359ca5085485482ba25dd4.jpg" + } + ] + } + ], + "index": 14, + "virtual_lines": [ + { + "bbox": [ + 110, + 231, + 498, + 351.0 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 110, + 351.0, + 498, + 471.0 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 110, + 471.0, + 498, + 591.0 + ], + "spans": [], + "index": 15 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 599, + 505, + 640 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 600, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 106, + 600, + 505, + 610 + ], + "score": 1.0, + "content": "Figure 16: Visual results on Sintel test. We compare with Flowformer [23] and provide flow visualization and", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 610, + 505, + 620 + ], + "spans": [ + { + "bbox": [ + 106, + 610, + 505, + 620 + ], + "score": 1.0, + "content": "an error map on each scene. On Ambush 1, FlowFormer can better predict the motion of the girl’s right arm", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 619, + 505, + 631 + ], + "spans": [ + { + "bbox": [ + 106, + 619, + 505, + 631 + ], + "score": 1.0, + "content": "that moves out of the image boundary, likely due to the global reasoning capability of attention. On Cave 3 and", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 629, + 489, + 640 + ], + "spans": [ + { + "bbox": [ + 106, + 629, + 489, + 640 + ], + "score": 1.0, + "content": "Market 1, our method provides much finer details on motion boundaries with lower end-point error (EPE).", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 17.5 + } + ], + "index": 15.75 + } + ], + "page_idx": 25, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 298, + 740, + 313, + 753 + ], + "spans": [ + { + "bbox": [ + 298, + 740, + 313, + 753 + ], + "score": 1.0, + "content": "26", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 72, + 505, + 182 + ], + "lines": [], + "index": 4.5, + "bbox_fs": [ + 104, + 73, + 506, + 184 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 187, + 505, + 221 + ], + "lines": [ + { + "bbox": [ + 105, + 186, + 506, + 200 + ], + "spans": [ + { + "bbox": [ + 105, + 186, + 506, + 200 + ], + "score": 1.0, + "content": "Fine-tuning on Sintel. In Section 5 we discuss possible reasons for why our model’s superior zero-", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 198, + 505, + 210 + ], + "spans": [ + { + "bbox": [ + 106, + 198, + 505, + 210 + ], + "score": 1.0, + "content": "shot performance compared to FlowFormer [23] does not transfer to fine-tuning on Sintel. Figure 16", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 209, + 346, + 222 + ], + "spans": [ + { + "bbox": [ + 105, + 209, + 346, + 222 + ], + "score": 1.0, + "content": "provides qualitative examples to further support the claims.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 186, + 506, + 222 + ] + }, + { + "type": "image", + "bbox": [ + 110, + 231, + 498, + 591 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 110, + 231, + 498, + 591 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 110, + 231, + 498, + 591 + ], + "spans": [ + { + "bbox": [ + 110, + 231, + 498, + 591 + ], + "score": 0.976, + "type": "image", + "image_path": "1934e1d086049ece1ca2800e7b4682ae30439a2fdd359ca5085485482ba25dd4.jpg" + } + ] + } + ], + "index": 14, + "virtual_lines": [ + { + "bbox": [ + 110, + 231, + 498, + 351.0 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 110, + 351.0, + 498, + 471.0 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 110, + 471.0, + 498, + 591.0 + ], + "spans": [], + "index": 15 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 599, + 505, + 640 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 600, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 106, + 600, + 505, + 610 + ], + "score": 1.0, + "content": "Figure 16: Visual results on Sintel test. We compare with Flowformer [23] and provide flow visualization and", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 610, + 505, + 620 + ], + "spans": [ + { + "bbox": [ + 106, + 610, + 505, + 620 + ], + "score": 1.0, + "content": "an error map on each scene. On Ambush 1, FlowFormer can better predict the motion of the girl’s right arm", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 619, + 505, + 631 + ], + "spans": [ + { + "bbox": [ + 106, + 619, + 505, + 631 + ], + "score": 1.0, + "content": "that moves out of the image boundary, likely due to the global reasoning capability of attention. On Cave 3 and", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 629, + 489, + 640 + ], + "spans": [ + { + "bbox": [ + 106, + 629, + 489, + 640 + ], + "score": 1.0, + "content": "Market 1, our method provides much finer details on motion boundaries with lower end-point error (EPE).", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 17.5 + } + ], + "index": 15.75 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 108, + 69, + 504, + 197 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 69, + 504, + 197 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 69, + 504, + 197 + ], + "spans": [ + { + "bbox": [ + 108, + 69, + 504, + 197 + ], + "score": 0.965, + "type": "image", + "image_path": "e4d3a72742ce37bcd0aba746bddd4ce120bc6b261a5eb6f1668ae4156c288559.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 108, + 69, + 504, + 111.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 108, + 111.66666666666666, + 504, + 154.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 154.33333333333331, + 504, + 196.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 203, + 504, + 234 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 203, + 505, + 214 + ], + "spans": [ + { + "bbox": [ + 106, + 203, + 505, + 214 + ], + "score": 1.0, + "content": "Figure 17: Qualitative examples of multimodal estimation on the NYU depth dataset showing examples", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 213, + 505, + 224 + ], + "spans": [ + { + "bbox": [ + 105, + 213, + 505, + 224 + ], + "score": 1.0, + "content": "where our model’s uncertainty gets captured in the multimodal posterior. In the examples above, the model", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 223, + 441, + 235 + ], + "spans": [ + { + "bbox": [ + 105, + 223, + 441, + 235 + ], + "score": 1.0, + "content": "confuses the play mat (farther away from the viewpoint) for a table (closer to the viewpoint).", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 107, + 262, + 505, + 295 + ], + "lines": [ + { + "bbox": [ + 106, + 262, + 505, + 274 + ], + "spans": [ + { + "bbox": [ + 106, + 262, + 505, + 274 + ], + "score": 1.0, + "content": "Uncertainty in depth estimation. We observe certain cases where the model is uncertain about the", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 272, + 507, + 286 + ], + "spans": [ + { + "bbox": [ + 105, + 272, + 507, + 286 + ], + "score": 1.0, + "content": "depth estimates. Interestingly, this uncertainty appears to be well captured in the predictive posterior,", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 284, + 214, + 296 + ], + "spans": [ + { + "bbox": [ + 105, + 284, + 214, + 296 + ], + "score": 1.0, + "content": "as illustrated in Figure 17.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + } + ], + "page_idx": 26, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 741, + 311, + 750 + ], + "lines": [ + { + "bbox": [ + 298, + 740, + 313, + 755 + ], + "spans": [ + { + "bbox": [ + 298, + 740, + 313, + 755 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 329, + 70, + 443, + 79 + ], + "lines": [] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 108, + 69, + 504, + 197 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 69, + 504, + 197 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 69, + 504, + 197 + ], + "spans": [ + { + "bbox": [ + 108, + 69, + 504, + 197 + ], + "score": 0.965, + "type": "image", + "image_path": "e4d3a72742ce37bcd0aba746bddd4ce120bc6b261a5eb6f1668ae4156c288559.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 108, + 69, + 504, + 111.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 108, + 111.66666666666666, + 504, + 154.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 154.33333333333331, + 504, + 196.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 203, + 504, + 234 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 203, + 505, + 214 + ], + "spans": [ + { + "bbox": [ + 106, + 203, + 505, + 214 + ], + "score": 1.0, + "content": "Figure 17: Qualitative examples of multimodal estimation on the NYU depth dataset showing examples", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 213, + 505, + 224 + ], + "spans": [ + { + "bbox": [ + 105, + 213, + 505, + 224 + ], + "score": 1.0, + "content": "where our model’s uncertainty gets captured in the multimodal posterior. In the examples above, the model", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 223, + 441, + 235 + ], + "spans": [ + { + "bbox": [ + 105, + 223, + 441, + 235 + ], + "score": 1.0, + "content": "confuses the play mat (farther away from the viewpoint) for a table (closer to the viewpoint).", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 107, + 262, + 505, + 295 + ], + "lines": [ + { + "bbox": [ + 106, + 262, + 505, + 274 + ], + "spans": [ + { + "bbox": [ + 106, + 262, + 505, + 274 + ], + "score": 1.0, + "content": "Uncertainty in depth estimation. We observe certain cases where the model is uncertain about the", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 272, + 507, + 286 + ], + "spans": [ + { + "bbox": [ + 105, + 272, + 507, + 286 + ], + "score": 1.0, + "content": "depth estimates. Interestingly, this uncertainty appears to be well captured in the predictive posterior,", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 284, + 214, + 296 + ], + "spans": [ + { + "bbox": [ + 105, + 284, + 214, + 296 + ], + "score": 1.0, + "content": "as illustrated in Figure 17.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7, + "bbox_fs": [ + 105, + 262, + 507, + 296 + ] + } + ] + } + ], + "_backend": "pipeline", + "_version_name": "2.2.2" +} \ No newline at end of file diff --git a/parse/dev/jDIlzSU8wJ/jDIlzSU8wJ_model.json b/parse/dev/jDIlzSU8wJ/jDIlzSU8wJ_model.json new file mode 100644 index 0000000000000000000000000000000000000000..d54d90c3b3cb124f156d87395f2b759549db34d4 --- /dev/null +++ b/parse/dev/jDIlzSU8wJ/jDIlzSU8wJ_model.json @@ -0,0 +1,25277 @@ +[ + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 398, + 787, + 1303, + 787, + 1303, + 1300, + 398, + 1300 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 1418, + 1405, + 1418, + 1405, + 1692, + 298, + 1692 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 299, + 1706, + 1403, + 1706, + 1403, + 1859, + 299, + 1859 + ], + "score": 0.975 + }, + { + "category_id": 0, + "poly": [ + 316, + 271, + 1388, + 271, + 1388, + 382, + 316, + 382 + ], + "score": 0.965 + }, + { + "category_id": 1, + "poly": [ + 451, + 497, + 1255, + 497, + 1255, + 570, + 451, + 570 + ], + "score": 0.929 + }, + { + "category_id": 1, + "poly": [ + 303, + 1872, + 1396, + 1872, + 1396, + 1935, + 303, + 1935 + ], + "score": 0.925 + }, + { + "category_id": 2, + "poly": [ + 330, + 1955, + 1094, + 1955, + 1094, + 1982, + 330, + 1982 + ], + "score": 0.908 + }, + { + "category_id": 0, + "poly": [ + 299, + 1357, + 530, + 1357, + 530, + 1394, + 299, + 1394 + ], + "score": 0.906 + }, + { + "category_id": 2, + "poly": [ + 299, + 2033, + 1071, + 2033, + 1071, + 2062, + 299, + 2062 + ], + "score": 0.892 + }, + { + "category_id": 0, + "poly": [ + 788, + 740, + 912, + 740, + 912, + 775, + 788, + 775 + ], + "score": 0.887 + }, + { + "category_id": 1, + "poly": [ + 331, + 584, + 1362, + 584, + 1362, + 616, + 331, + 616 + ], + "score": 0.797 + }, + { + "category_id": 1, + "poly": [ + 624, + 630, + 1078, + 630, + 1078, + 661, + 624, + 661 + ], + "score": 0.756 + }, + { + "category_id": 13, + "poly": [ + 821, + 1241, + 876, + 1241, + 876, + 1270, + 821, + 1270 + ], + "score": 0.88, + "latex": "2 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 1151, + 1211, + 1226, + 1211, + 1226, + 1240, + 1151, + 1240 + ], + "score": 0.87, + "latex": "3 . 2 6 \\%" + }, + { + "category_id": 15, + "poly": [ + 311.0, + 272.0, + 1389.0, + 272.0, + 1389.0, + 330.0, + 311.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 362.0, + 327.0, + 1335.0, + 327.0, + 1335.0, + 387.0, + 362.0, + 387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 1952.0, + 1097.0, + 1952.0, + 1097.0, + 1985.0, + 331.0, + 1985.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1353.0, + 536.0, + 1353.0, + 536.0, + 1400.0, + 292.0, + 1400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 2028.0, + 1074.0, + 2028.0, + 1074.0, + 2066.0, + 294.0, + 2066.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 784.0, + 740.0, + 918.0, + 740.0, + 918.0, + 778.0, + 784.0, + 778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 785.0, + 1304.0, + 785.0, + 1304.0, + 820.0, + 394.0, + 820.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 814.0, + 1305.0, + 814.0, + 1305.0, + 854.0, + 393.0, + 854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 846.0, + 1307.0, + 846.0, + 1307.0, + 884.0, + 393.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 876.0, + 1307.0, + 876.0, + 1307.0, + 913.0, + 394.0, + 913.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 908.0, + 1306.0, + 908.0, + 1306.0, + 940.0, + 393.0, + 940.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 937.0, + 1308.0, + 937.0, + 1308.0, + 973.0, + 393.0, + 973.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 967.0, + 1306.0, + 967.0, + 1306.0, + 1002.0, + 393.0, + 1002.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 998.0, + 1305.0, + 998.0, + 1305.0, + 1036.0, + 392.0, + 1036.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1026.0, + 1304.0, + 1026.0, + 1304.0, + 1064.0, + 394.0, + 1064.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1057.0, + 1305.0, + 1057.0, + 1305.0, + 1092.0, + 394.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 1089.0, + 1303.0, + 1089.0, + 1303.0, + 1122.0, + 395.0, + 1122.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1119.0, + 1306.0, + 1119.0, + 1306.0, + 1155.0, + 392.0, + 1155.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1150.0, + 1306.0, + 1150.0, + 1306.0, + 1183.0, + 393.0, + 1183.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1181.0, + 1306.0, + 1181.0, + 1306.0, + 1214.0, + 394.0, + 1214.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 1213.0, + 1150.0, + 1213.0, + 1150.0, + 1243.0, + 395.0, + 1243.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1227.0, + 1213.0, + 1305.0, + 1213.0, + 1305.0, + 1243.0, + 1227.0, + 1243.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1241.0, + 820.0, + 1241.0, + 820.0, + 1274.0, + 394.0, + 1274.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 877.0, + 1241.0, + 1308.0, + 1241.0, + 1308.0, + 1274.0, + 877.0, + 1274.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1271.0, + 912.0, + 1271.0, + 912.0, + 1304.0, + 394.0, + 1304.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1416.0, + 1409.0, + 1416.0, + 1409.0, + 1455.0, + 293.0, + 1455.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1449.0, + 1407.0, + 1449.0, + 1407.0, + 1484.0, + 293.0, + 1484.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1480.0, + 1405.0, + 1480.0, + 1405.0, + 1512.0, + 296.0, + 1512.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1506.0, + 1408.0, + 1506.0, + 1408.0, + 1547.0, + 292.0, + 1547.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1539.0, + 1407.0, + 1539.0, + 1407.0, + 1576.0, + 295.0, + 1576.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1569.0, + 1405.0, + 1569.0, + 1405.0, + 1604.0, + 293.0, + 1604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1599.0, + 1405.0, + 1599.0, + 1405.0, + 1636.0, + 293.0, + 1636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1628.0, + 1405.0, + 1628.0, + 1405.0, + 1667.0, + 293.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1660.0, + 1343.0, + 1660.0, + 1343.0, + 1697.0, + 295.0, + 1697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1706.0, + 1403.0, + 1706.0, + 1403.0, + 1739.0, + 295.0, + 1739.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1734.0, + 1405.0, + 1734.0, + 1405.0, + 1772.0, + 293.0, + 1772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1767.0, + 1404.0, + 1767.0, + 1404.0, + 1800.0, + 297.0, + 1800.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1797.0, + 1404.0, + 1797.0, + 1404.0, + 1831.0, + 293.0, + 1831.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1826.0, + 1213.0, + 1826.0, + 1213.0, + 1863.0, + 295.0, + 1863.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 495.0, + 1258.0, + 495.0, + 1258.0, + 539.0, + 449.0, + 539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 550.0, + 536.0, + 1158.0, + 536.0, + 1158.0, + 574.0, + 550.0, + 574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1869.0, + 1403.0, + 1869.0, + 1403.0, + 1910.0, + 297.0, + 1910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1904.0, + 1401.0, + 1904.0, + 1401.0, + 1938.0, + 296.0, + 1938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 584.0, + 1368.0, + 584.0, + 1368.0, + 622.0, + 329.0, + 622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 619.0, + 626.0, + 1081.0, + 626.0, + 1081.0, + 669.0, + 619.0, + 669.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 0, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 299, + 1884, + 1404, + 1884, + 1404, + 2007, + 299, + 2007 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 298, + 825, + 1403, + 825, + 1403, + 979, + 298, + 979 + ], + "score": 0.972 + }, + { + "category_id": 1, + "poly": [ + 298, + 993, + 1404, + 993, + 1404, + 1206, + 298, + 1206 + ], + "score": 0.972 + }, + { + "category_id": 4, + "poly": [ + 296, + 612, + 1404, + 612, + 1404, + 781, + 296, + 781 + ], + "score": 0.968 + }, + { + "category_id": 3, + "poly": [ + 296, + 194, + 1396, + 194, + 1396, + 596, + 296, + 596 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 300, + 1808, + 1400, + 1808, + 1400, + 1870, + 300, + 1870 + ], + "score": 0.951 + }, + { + "category_id": 1, + "poly": [ + 300, + 1265, + 1407, + 1265, + 1407, + 1705, + 300, + 1705 + ], + "score": 0.949 + }, + { + "category_id": 1, + "poly": [ + 299, + 1221, + 664, + 1221, + 664, + 1251, + 299, + 1251 + ], + "score": 0.923 + }, + { + "category_id": 0, + "poly": [ + 299, + 1746, + 541, + 1746, + 541, + 1783, + 299, + 1783 + ], + "score": 0.915 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 858, + 2061, + 858, + 2085, + 841, + 2085 + ], + "score": 0.723 + }, + { + "category_id": 2, + "poly": [ + 919, + 199, + 1233, + 199, + 1233, + 224, + 919, + 224 + ], + "score": 0.39 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.123 + }, + { + "category_id": 13, + "poly": [ + 674, + 1608, + 750, + 1608, + 750, + 1638, + 674, + 1638 + ], + "score": 0.89, + "latex": "{ \\sim } 2 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 580, + 1393, + 614, + 1393, + 614, + 1422, + 580, + 1422 + ], + "score": 0.88, + "latex": "L _ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 1313, + 1085, + 1347, + 1085, + 1347, + 1115, + 1313, + 1115 + ], + "score": 0.87, + "latex": "L _ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 1248, + 1577, + 1324, + 1577, + 1324, + 1606, + 1248, + 1606 + ], + "score": 0.85, + "latex": "3 . 2 6 \\%" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 612.0, + 1405.0, + 612.0, + 1405.0, + 644.0, + 294.0, + 644.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 641.0, + 1407.0, + 641.0, + 1407.0, + 675.0, + 293.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 666.0, + 1405.0, + 666.0, + 1405.0, + 701.0, + 294.0, + 701.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 697.0, + 1409.0, + 697.0, + 1409.0, + 729.0, + 291.0, + 729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 724.0, + 1406.0, + 724.0, + 1406.0, + 758.0, + 294.0, + 758.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 752.0, + 1404.0, + 752.0, + 1404.0, + 784.0, + 293.0, + 784.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 192.0, + 747.0, + 192.0, + 747.0, + 231.0, + 351.0, + 231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 915.0, + 193.0, + 1235.0, + 193.0, + 1235.0, + 228.0, + 915.0, + 228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 274.0, + 330.0, + 274.0, + 330.0, + 342.0, + 296.0, + 342.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 457.0, + 318.0, + 457.0, + 318.0, + 469.0, + 308.0, + 469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 300.0, + 521.0, + 330.0, + 521.0, + 330.0, + 588.0, + 300.0, + 588.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 644.0, + 578.0, + 652.0, + 578.0, + 652.0, + 590.0, + 644.0, + 590.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 623.0, + 572.0, + 656.0, + 572.0, + 656.0, + 585.0, + 623.0, + 585.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1743.0, + 545.0, + 1743.0, + 545.0, + 1788.0, + 292.0, + 1788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2093.0, + 839.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 917.0, + 193.0, + 1236.0, + 193.0, + 1236.0, + 229.0, + 917.0, + 229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2093.0, + 838.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1884.0, + 1405.0, + 1884.0, + 1405.0, + 1920.0, + 295.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1913.0, + 1409.0, + 1913.0, + 1409.0, + 1951.0, + 293.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1947.0, + 1408.0, + 1947.0, + 1408.0, + 1979.0, + 295.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1977.0, + 1404.0, + 1977.0, + 1404.0, + 2009.0, + 295.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 821.0, + 1407.0, + 821.0, + 1407.0, + 866.0, + 292.0, + 866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 856.0, + 1405.0, + 856.0, + 1405.0, + 893.0, + 294.0, + 893.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 885.0, + 1404.0, + 885.0, + 1404.0, + 922.0, + 293.0, + 922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 917.0, + 1405.0, + 917.0, + 1405.0, + 954.0, + 294.0, + 954.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 946.0, + 1112.0, + 946.0, + 1112.0, + 984.0, + 293.0, + 984.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 992.0, + 1405.0, + 992.0, + 1405.0, + 1030.0, + 292.0, + 1030.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1024.0, + 1405.0, + 1024.0, + 1405.0, + 1058.0, + 294.0, + 1058.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1054.0, + 1405.0, + 1054.0, + 1405.0, + 1088.0, + 294.0, + 1088.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1085.0, + 1312.0, + 1085.0, + 1312.0, + 1119.0, + 294.0, + 1119.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1348.0, + 1085.0, + 1405.0, + 1085.0, + 1405.0, + 1119.0, + 1348.0, + 1119.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1114.0, + 1406.0, + 1114.0, + 1406.0, + 1148.0, + 294.0, + 1148.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1145.0, + 1406.0, + 1145.0, + 1406.0, + 1179.0, + 294.0, + 1179.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1177.0, + 653.0, + 1177.0, + 653.0, + 1207.0, + 296.0, + 1207.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1807.0, + 1402.0, + 1807.0, + 1402.0, + 1843.0, + 296.0, + 1843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1840.0, + 1212.0, + 1840.0, + 1212.0, + 1872.0, + 296.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 1267.0, + 1403.0, + 1267.0, + 1403.0, + 1299.0, + 302.0, + 1299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 332.0, + 1296.0, + 1320.0, + 1296.0, + 1320.0, + 1333.0, + 332.0, + 1333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 1333.0, + 329.0, + 1333.0, + 329.0, + 1360.0, + 302.0, + 1360.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 330.0, + 1327.0, + 1407.0, + 1327.0, + 1407.0, + 1368.0, + 330.0, + 1368.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 1355.0, + 1409.0, + 1355.0, + 1409.0, + 1399.0, + 333.0, + 1399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 1392.0, + 579.0, + 1392.0, + 579.0, + 1428.0, + 333.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 615.0, + 1392.0, + 1407.0, + 1392.0, + 1407.0, + 1428.0, + 615.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 1421.0, + 1405.0, + 1421.0, + 1405.0, + 1458.0, + 333.0, + 1458.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 334.0, + 1453.0, + 1233.0, + 1453.0, + 1233.0, + 1485.0, + 334.0, + 1485.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 1489.0, + 329.0, + 1489.0, + 329.0, + 1517.0, + 302.0, + 1517.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 332.0, + 1486.0, + 1405.0, + 1486.0, + 1405.0, + 1523.0, + 332.0, + 1523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 335.0, + 1516.0, + 1405.0, + 1516.0, + 1405.0, + 1553.0, + 335.0, + 1553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 1545.0, + 1407.0, + 1545.0, + 1407.0, + 1585.0, + 333.0, + 1585.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 332.0, + 1576.0, + 1247.0, + 1576.0, + 1247.0, + 1611.0, + 332.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1325.0, + 1576.0, + 1405.0, + 1576.0, + 1405.0, + 1611.0, + 1325.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 1607.0, + 673.0, + 1607.0, + 673.0, + 1642.0, + 333.0, + 1642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 751.0, + 1607.0, + 1233.0, + 1607.0, + 1233.0, + 1642.0, + 751.0, + 1642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 1639.0, + 1407.0, + 1639.0, + 1407.0, + 1677.0, + 302.0, + 1677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 337.0, + 1673.0, + 1218.0, + 1673.0, + 1218.0, + 1706.0, + 337.0, + 1706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1217.0, + 668.0, + 1217.0, + 668.0, + 1254.0, + 297.0, + 1254.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 1, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1082, + 1404, + 1082, + 1404, + 1267, + 297, + 1267 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 297, + 1279, + 1404, + 1279, + 1404, + 1464, + 297, + 1464 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 298, + 1476, + 1404, + 1476, + 1404, + 1630, + 298, + 1630 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 1747, + 1403, + 1747, + 1403, + 1841, + 298, + 1841 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 1853, + 1405, + 1853, + 1405, + 2008, + 298, + 2008 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 297, + 839, + 1406, + 839, + 1406, + 994, + 297, + 994 + ], + "score": 0.976 + }, + { + "category_id": 3, + "poly": [ + 306, + 212, + 1383, + 212, + 1383, + 666, + 306, + 666 + ], + "score": 0.973 + }, + { + "category_id": 8, + "poly": [ + 562, + 1640, + 1135, + 1640, + 1135, + 1739, + 562, + 1739 + ], + "score": 0.958 + }, + { + "category_id": 1, + "poly": [ + 299, + 1006, + 1398, + 1006, + 1398, + 1070, + 299, + 1070 + ], + "score": 0.949 + }, + { + "category_id": 4, + "poly": [ + 296, + 703, + 1405, + 703, + 1405, + 816, + 296, + 816 + ], + "score": 0.924 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1664, + 1400, + 1664, + 1400, + 1694, + 1366, + 1694 + ], + "score": 0.888 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.651 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.445 + }, + { + "category_id": 2, + "poly": [ + 1112, + 206, + 1249, + 206, + 1249, + 224, + 1112, + 224 + ], + "score": 0.183 + }, + { + "category_id": 14, + "poly": [ + 563, + 1636, + 1137, + 1636, + 1137, + 1740, + 563, + 1740 + ], + "score": 0.93, + "latex": "\\mathbb { E } _ { ( \\pmb { x } , \\pmb { y } ) } \\mathbb { E } _ { ( t , \\epsilon ) } \\bigg \\| f _ { \\theta } ( \\pmb { x } , \\underbrace { \\sqrt { \\gamma _ { t } } \\pmb { y } + \\sqrt { 1 - \\gamma _ { t } } \\pmb { \\epsilon } } _ { \\pmb { y _ { t } } } , t ) - \\pmb { \\epsilon } \\bigg \\| _ { 2 } ^ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 827, + 1749, + 918, + 1749, + 918, + 1781, + 827, + 1781 + ], + "score": 0.93, + "latex": "\\gamma _ { t } ~ > ~ 0" + }, + { + "category_id": 13, + "poly": [ + 1339, + 1372, + 1403, + 1372, + 1403, + 1400, + 1339, + 1400 + ], + "score": 0.88, + "latex": "t = 1" + }, + { + "category_id": 13, + "poly": [ + 1286, + 1811, + 1348, + 1811, + 1348, + 1837, + 1286, + 1837 + ], + "score": 0.88, + "latex": "s < t" + }, + { + "category_id": 13, + "poly": [ + 352, + 1810, + 381, + 1810, + 381, + 1840, + 352, + 1840 + ], + "score": 0.87, + "latex": "f _ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 811, + 1479, + 839, + 1479, + 839, + 1509, + 811, + 1509 + ], + "score": 0.86, + "latex": "f _ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 1139, + 1783, + 1171, + 1783, + 1171, + 1811, + 1139, + 1811 + ], + "score": 0.85, + "latex": "\\mathbf { \\pmb { y } } _ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 1204, + 1813, + 1234, + 1813, + 1234, + 1841, + 1204, + 1841 + ], + "score": 0.85, + "latex": "\\mathbf { \\eta } _ { \\mathbf { { 3 } } \\mathbf { { ' } } \\mathbf { { 3 } } }" + }, + { + "category_id": 13, + "poly": [ + 541, + 1748, + 683, + 1748, + 683, + 1782, + 541, + 1782 + ], + "score": 0.83, + "latex": "t \\sim \\mathcal { U } ( 0 , 1 )" + }, + { + "category_id": 13, + "poly": [ + 1228, + 792, + 1252, + 792, + 1252, + 816, + 1228, + 816 + ], + "score": 0.83, + "latex": "y _ { t }" + }, + { + "category_id": 13, + "poly": [ + 1276, + 1483, + 1303, + 1483, + 1303, + 1509, + 1276, + 1509 + ], + "score": 0.83, + "latex": "y _ { t }" + }, + { + "category_id": 13, + "poly": [ + 375, + 1748, + 527, + 1748, + 527, + 1781, + 375, + 1781 + ], + "score": 0.78, + "latex": "\\epsilon \\sim \\mathcal { N } ( 0 , I )" + }, + { + "category_id": 13, + "poly": [ + 812, + 1514, + 831, + 1514, + 831, + 1536, + 812, + 1536 + ], + "score": 0.77, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 336, + 1572, + 350, + 1572, + 350, + 1596, + 336, + 1596 + ], + "score": 0.71, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 410, + 1511, + 424, + 1511, + 424, + 1536, + 410, + 1536 + ], + "score": 0.67, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 298, + 1605, + 312, + 1605, + 312, + 1626, + 298, + 1626 + ], + "score": 0.67, + "latex": "\\epsilon" + }, + { + "category_id": 13, + "poly": [ + 944, + 1405, + 958, + 1405, + 958, + 1429, + 944, + 1429 + ], + "score": 0.63, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 999, + 1343, + 1014, + 1343, + 1014, + 1368, + 999, + 1368 + ], + "score": 0.47, + "latex": "t" + }, + { + "category_id": 15, + "poly": [ + 1111.0, + 207.0, + 1252.0, + 207.0, + 1252.0, + 226.0, + 1111.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 236.0, + 934.0, + 236.0, + 934.0, + 261.0, + 757.0, + 261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 778.0, + 256.0, + 913.0, + 256.0, + 913.0, + 279.0, + 778.0, + 279.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 348.0, + 948.0, + 348.0, + 948.0, + 380.0, + 838.0, + 380.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 499.0, + 393.0, + 653.0, + 393.0, + 653.0, + 441.0, + 499.0, + 441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 735.0, + 400.0, + 844.0, + 400.0, + 844.0, + 434.0, + 735.0, + 434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 399.0, + 1147.0, + 399.0, + 1147.0, + 444.0, + 1100.0, + 444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1237.0, + 386.0, + 1379.0, + 386.0, + 1379.0, + 433.0, + 1237.0, + 433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 694.0, + 455.0, + 705.0, + 455.0, + 705.0, + 466.0, + 694.0, + 466.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 858.0, + 544.0, + 898.0, + 544.0, + 898.0, + 556.0, + 858.0, + 556.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 842.0, + 605.0, + 912.0, + 605.0, + 912.0, + 629.0, + 842.0, + 629.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 424.0, + 621.0, + 670.0, + 621.0, + 670.0, + 649.0, + 424.0, + 649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 847.0, + 623.0, + 906.0, + 623.0, + 906.0, + 649.0, + 847.0, + 649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 438.0, + 641.0, + 656.0, + 641.0, + 656.0, + 669.0, + 438.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 701.0, + 1407.0, + 701.0, + 1407.0, + 737.0, + 294.0, + 737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 729.0, + 1406.0, + 729.0, + 1406.0, + 765.0, + 294.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 755.0, + 1405.0, + 755.0, + 1405.0, + 794.0, + 294.0, + 794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 781.0, + 1227.0, + 781.0, + 1227.0, + 825.0, + 293.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1253.0, + 781.0, + 1265.0, + 781.0, + 1265.0, + 825.0, + 1253.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1110.0, + 203.0, + 1252.0, + 203.0, + 1252.0, + 226.0, + 1110.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1082.0, + 1406.0, + 1082.0, + 1406.0, + 1119.0, + 294.0, + 1119.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1115.0, + 1404.0, + 1115.0, + 1404.0, + 1147.0, + 296.0, + 1147.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1141.0, + 1405.0, + 1141.0, + 1405.0, + 1177.0, + 295.0, + 1177.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1176.0, + 1404.0, + 1176.0, + 1404.0, + 1208.0, + 295.0, + 1208.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1202.0, + 1407.0, + 1202.0, + 1407.0, + 1239.0, + 294.0, + 1239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1234.0, + 1166.0, + 1234.0, + 1166.0, + 1270.0, + 295.0, + 1270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1277.0, + 1408.0, + 1277.0, + 1408.0, + 1316.0, + 294.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1310.0, + 1406.0, + 1310.0, + 1406.0, + 1345.0, + 294.0, + 1345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1340.0, + 998.0, + 1340.0, + 998.0, + 1375.0, + 291.0, + 1375.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1015.0, + 1340.0, + 1407.0, + 1340.0, + 1407.0, + 1375.0, + 1015.0, + 1375.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1372.0, + 1338.0, + 1372.0, + 1338.0, + 1406.0, + 291.0, + 1406.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1404.0, + 1372.0, + 1409.0, + 1372.0, + 1409.0, + 1406.0, + 1404.0, + 1406.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1401.0, + 943.0, + 1401.0, + 943.0, + 1436.0, + 295.0, + 1436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 959.0, + 1401.0, + 1406.0, + 1401.0, + 1406.0, + 1436.0, + 959.0, + 1436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1431.0, + 1230.0, + 1431.0, + 1230.0, + 1467.0, + 294.0, + 1467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1475.0, + 810.0, + 1475.0, + 810.0, + 1513.0, + 295.0, + 1513.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 1475.0, + 1275.0, + 1475.0, + 1275.0, + 1513.0, + 840.0, + 1513.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1304.0, + 1475.0, + 1405.0, + 1475.0, + 1405.0, + 1513.0, + 1304.0, + 1513.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1508.0, + 409.0, + 1508.0, + 409.0, + 1541.0, + 296.0, + 1541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 425.0, + 1508.0, + 811.0, + 1508.0, + 811.0, + 1541.0, + 425.0, + 1541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 1508.0, + 1404.0, + 1508.0, + 1404.0, + 1541.0, + 832.0, + 1541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1538.0, + 1405.0, + 1538.0, + 1405.0, + 1575.0, + 292.0, + 1575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1568.0, + 335.0, + 1568.0, + 335.0, + 1605.0, + 293.0, + 1605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1568.0, + 1404.0, + 1568.0, + 1404.0, + 1605.0, + 351.0, + 1605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1601.0, + 297.0, + 1601.0, + 297.0, + 1632.0, + 293.0, + 1632.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 313.0, + 1601.0, + 525.0, + 1601.0, + 525.0, + 1632.0, + 313.0, + 1632.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1744.0, + 374.0, + 1744.0, + 374.0, + 1784.0, + 293.0, + 1784.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 528.0, + 1744.0, + 540.0, + 1744.0, + 540.0, + 1784.0, + 528.0, + 1784.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 684.0, + 1744.0, + 826.0, + 1744.0, + 826.0, + 1784.0, + 684.0, + 1784.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 919.0, + 1744.0, + 1407.0, + 1744.0, + 1407.0, + 1784.0, + 919.0, + 1784.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1777.0, + 1138.0, + 1777.0, + 1138.0, + 1813.0, + 294.0, + 1813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1172.0, + 1777.0, + 1403.0, + 1777.0, + 1403.0, + 1813.0, + 1172.0, + 1813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1809.0, + 351.0, + 1809.0, + 351.0, + 1843.0, + 296.0, + 1843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 382.0, + 1809.0, + 1203.0, + 1809.0, + 1203.0, + 1843.0, + 382.0, + 1843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1235.0, + 1809.0, + 1285.0, + 1809.0, + 1285.0, + 1843.0, + 1235.0, + 1843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1349.0, + 1809.0, + 1360.0, + 1809.0, + 1360.0, + 1843.0, + 1349.0, + 1843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1855.0, + 1408.0, + 1855.0, + 1408.0, + 1888.0, + 296.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1885.0, + 1407.0, + 1885.0, + 1407.0, + 1919.0, + 296.0, + 1919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1912.0, + 1407.0, + 1912.0, + 1407.0, + 1950.0, + 295.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1944.0, + 1405.0, + 1944.0, + 1405.0, + 1980.0, + 293.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1977.0, + 1407.0, + 1977.0, + 1407.0, + 2010.0, + 296.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 841.0, + 1403.0, + 841.0, + 1403.0, + 874.0, + 295.0, + 874.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 867.0, + 1408.0, + 867.0, + 1408.0, + 909.0, + 292.0, + 909.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 901.0, + 1406.0, + 901.0, + 1406.0, + 938.0, + 295.0, + 938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 928.0, + 1408.0, + 928.0, + 1408.0, + 971.0, + 291.0, + 971.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 960.0, + 1089.0, + 960.0, + 1089.0, + 998.0, + 294.0, + 998.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1005.0, + 1403.0, + 1005.0, + 1403.0, + 1043.0, + 294.0, + 1043.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1037.0, + 1282.0, + 1037.0, + 1282.0, + 1073.0, + 295.0, + 1073.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 2, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1191, + 1404, + 1191, + 1404, + 1374, + 298, + 1374 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 1388, + 1404, + 1388, + 1404, + 1571, + 298, + 1571 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 299, + 963, + 1404, + 963, + 1404, + 1117, + 299, + 1117 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 298, + 797, + 1404, + 797, + 1404, + 950, + 298, + 950 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 298, + 1585, + 1404, + 1585, + 1404, + 1769, + 298, + 1769 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 299, + 1782, + 1403, + 1782, + 1403, + 1875, + 299, + 1875 + ], + "score": 0.971 + }, + { + "category_id": 3, + "poly": [ + 298, + 190, + 1400, + 190, + 1400, + 542, + 298, + 542 + ], + "score": 0.97 + }, + { + "category_id": 4, + "poly": [ + 297, + 566, + 1406, + 566, + 1406, + 678, + 297, + 678 + ], + "score": 0.958 + }, + { + "category_id": 1, + "poly": [ + 298, + 1946, + 1407, + 1946, + 1407, + 2008, + 298, + 2008 + ], + "score": 0.945 + }, + { + "category_id": 0, + "poly": [ + 298, + 1146, + 911, + 1146, + 911, + 1179, + 298, + 1179 + ], + "score": 0.934 + }, + { + "category_id": 0, + "poly": [ + 296, + 1903, + 1055, + 1903, + 1055, + 1937, + 296, + 1937 + ], + "score": 0.928 + }, + { + "category_id": 0, + "poly": [ + 299, + 727, + 616, + 727, + 616, + 765, + 299, + 765 + ], + "score": 0.919 + }, + { + "category_id": 2, + "poly": [ + 841, + 2062, + 858, + 2062, + 858, + 2084, + 841, + 2084 + ], + "score": 0.769 + }, + { + "category_id": 15, + "poly": [ + 349.0, + 192.0, + 472.0, + 192.0, + 472.0, + 229.0, + 349.0, + 229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 587.0, + 194.0, + 674.0, + 194.0, + 674.0, + 229.0, + 587.0, + 229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 781.0, + 195.0, + 917.0, + 195.0, + 917.0, + 226.0, + 781.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 980.0, + 194.0, + 1162.0, + 194.0, + 1162.0, + 226.0, + 980.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1221.0, + 195.0, + 1360.0, + 195.0, + 1360.0, + 225.0, + 1221.0, + 225.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 330.0, + 357.0, + 455.0, + 357.0, + 455.0, + 393.0, + 330.0, + 393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 556.0, + 360.0, + 597.0, + 360.0, + 597.0, + 389.0, + 556.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 716.0, + 358.0, + 804.0, + 358.0, + 804.0, + 393.0, + 716.0, + 393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 874.0, + 357.0, + 1382.0, + 357.0, + 1382.0, + 393.0, + 874.0, + 393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 566.0, + 1404.0, + 566.0, + 1404.0, + 599.0, + 295.0, + 599.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 593.0, + 1407.0, + 593.0, + 1407.0, + 626.0, + 295.0, + 626.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 617.0, + 1406.0, + 617.0, + 1406.0, + 656.0, + 292.0, + 656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 650.0, + 1152.0, + 650.0, + 1152.0, + 683.0, + 296.0, + 683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1143.0, + 914.0, + 1143.0, + 914.0, + 1186.0, + 293.0, + 1186.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1898.0, + 1056.0, + 1898.0, + 1056.0, + 1944.0, + 292.0, + 1944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 722.0, + 621.0, + 722.0, + 621.0, + 770.0, + 290.0, + 770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2059.0, + 862.0, + 2059.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1189.0, + 1406.0, + 1189.0, + 1406.0, + 1226.0, + 295.0, + 1226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1221.0, + 1405.0, + 1221.0, + 1405.0, + 1255.0, + 295.0, + 1255.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1252.0, + 1405.0, + 1252.0, + 1405.0, + 1284.0, + 296.0, + 1284.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1283.0, + 1404.0, + 1283.0, + 1404.0, + 1318.0, + 294.0, + 1318.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1310.0, + 1406.0, + 1310.0, + 1406.0, + 1348.0, + 293.0, + 1348.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1339.0, + 795.0, + 1339.0, + 795.0, + 1379.0, + 293.0, + 1379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1387.0, + 1405.0, + 1387.0, + 1405.0, + 1423.0, + 294.0, + 1423.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1418.0, + 1406.0, + 1418.0, + 1406.0, + 1453.0, + 292.0, + 1453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1447.0, + 1409.0, + 1447.0, + 1409.0, + 1485.0, + 293.0, + 1485.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1480.0, + 1406.0, + 1480.0, + 1406.0, + 1511.0, + 294.0, + 1511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1509.0, + 1405.0, + 1509.0, + 1405.0, + 1543.0, + 293.0, + 1543.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1539.0, + 1223.0, + 1539.0, + 1223.0, + 1575.0, + 293.0, + 1575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 963.0, + 1406.0, + 963.0, + 1406.0, + 1000.0, + 294.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 991.0, + 1404.0, + 991.0, + 1404.0, + 1031.0, + 293.0, + 1031.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1021.0, + 1405.0, + 1021.0, + 1405.0, + 1061.0, + 293.0, + 1061.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1051.0, + 1410.0, + 1051.0, + 1410.0, + 1096.0, + 293.0, + 1096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1084.0, + 1352.0, + 1084.0, + 1352.0, + 1121.0, + 295.0, + 1121.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 794.0, + 1408.0, + 794.0, + 1408.0, + 833.0, + 295.0, + 833.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 825.0, + 1405.0, + 825.0, + 1405.0, + 863.0, + 295.0, + 863.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 858.0, + 1405.0, + 858.0, + 1405.0, + 891.0, + 296.0, + 891.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 887.0, + 1406.0, + 887.0, + 1406.0, + 925.0, + 292.0, + 925.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 918.0, + 740.0, + 918.0, + 740.0, + 951.0, + 297.0, + 951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1581.0, + 1406.0, + 1581.0, + 1406.0, + 1624.0, + 292.0, + 1624.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1617.0, + 1405.0, + 1617.0, + 1405.0, + 1649.0, + 296.0, + 1649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1647.0, + 1406.0, + 1647.0, + 1406.0, + 1682.0, + 293.0, + 1682.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1676.0, + 1406.0, + 1676.0, + 1406.0, + 1710.0, + 293.0, + 1710.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1705.0, + 1408.0, + 1705.0, + 1408.0, + 1744.0, + 293.0, + 1744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1735.0, + 1405.0, + 1735.0, + 1405.0, + 1771.0, + 294.0, + 1771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1783.0, + 1404.0, + 1783.0, + 1404.0, + 1816.0, + 295.0, + 1816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1812.0, + 1404.0, + 1812.0, + 1404.0, + 1845.0, + 295.0, + 1845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1841.0, + 1210.0, + 1841.0, + 1210.0, + 1880.0, + 292.0, + 1880.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1943.0, + 1404.0, + 1943.0, + 1404.0, + 1981.0, + 295.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1977.0, + 1404.0, + 1977.0, + 1404.0, + 2009.0, + 296.0, + 2009.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 3, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1308, + 1405, + 1308, + 1405, + 1673, + 298, + 1673 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 656, + 1405, + 656, + 1405, + 1052, + 297, + 1052 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 299, + 1066, + 1403, + 1066, + 1403, + 1189, + 299, + 1189 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 299, + 1202, + 1404, + 1202, + 1404, + 1295, + 299, + 1295 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 298, + 1687, + 1404, + 1687, + 1404, + 1901, + 298, + 1901 + ], + "score": 0.971 + }, + { + "category_id": 1, + "poly": [ + 301, + 1914, + 1403, + 1914, + 1403, + 2007, + 301, + 2007 + ], + "score": 0.968 + }, + { + "category_id": 1, + "poly": [ + 304, + 244, + 1256, + 244, + 1256, + 597, + 304, + 597 + ], + "score": 0.751 + }, + { + "category_id": 2, + "poly": [ + 841, + 2062, + 858, + 2062, + 858, + 2085, + 841, + 2085 + ], + "score": 0.707 + }, + { + "category_id": 0, + "poly": [ + 299, + 201, + 1138, + 201, + 1138, + 235, + 299, + 235 + ], + "score": 0.434 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.108 + }, + { + "category_id": 5, + "poly": [ + 304, + 244, + 1256, + 244, + 1256, + 597, + 304, + 597 + ], + "score": 0.107, + "html": "
2: t ~ U(0,1),∈~ N(0,1)1:x ← conditioning images,y ← flow or depth map,mask ← binary mask of known values
3:y= fill_holes_with_interpolation(y)
4:yt=√t*y+√1-t*∈
5: :if unroll_step then
6: Epred = stop_gradient(fe(x,yt,t))
7: ypred =(yt-√1-Yt*Epred)/√Yt
8: yt=√Yt*ypred+√1-γt*∈
9: ∈=(yt-√t*y)/√1-Yt
10: end if
11: Epred= fe(x,yt,t)
12:loss =reduce_mean(l∈- Epred|[mask])
" + }, + { + "category_id": 13, + "poly": [ + 1086, + 747, + 1187, + 747, + 1187, + 781, + 1086, + 781 + ], + "score": 0.93, + "latex": "p ( \\pmb { y } _ { 0 } \\vert \\pmb { x } )" + }, + { + "category_id": 13, + "poly": [ + 1243, + 778, + 1395, + 778, + 1395, + 812, + 1243, + 812 + ], + "score": 0.93, + "latex": "p _ { \\theta } ( \\pmb { y } _ { s } \\mid \\pmb { y } _ { t } , \\pmb { x } )" + }, + { + "category_id": 13, + "poly": [ + 540, + 781, + 593, + 781, + 593, + 807, + 540, + 807 + ], + "score": 0.88, + "latex": "s < t" + }, + { + "category_id": 13, + "poly": [ + 629, + 1916, + 663, + 1916, + 663, + 1946, + 629, + 1946 + ], + "score": 0.88, + "latex": "L _ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 894, + 1461, + 972, + 1461, + 972, + 1491, + 894, + 1491 + ], + "score": 0.87, + "latex": "( \\sim 1 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 297, + 1915, + 331, + 1915, + 331, + 1946, + 297, + 1946 + ], + "score": 0.86, + "latex": "L _ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 362, + 752, + 394, + 752, + 394, + 780, + 362, + 780 + ], + "score": 0.85, + "latex": "\\mathbf { \\pmb { y } } _ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 917, + 1343, + 945, + 1343, + 945, + 1371, + 917, + 1371 + ], + "score": 0.85, + "latex": "y _ { t }" + }, + { + "category_id": 13, + "poly": [ + 636, + 1525, + 661, + 1525, + 661, + 1552, + 636, + 1552 + ], + "score": 0.83, + "latex": "y _ { t }" + }, + { + "category_id": 13, + "poly": [ + 764, + 1403, + 791, + 1403, + 791, + 1431, + 764, + 1431 + ], + "score": 0.83, + "latex": "y _ { t }" + }, + { + "category_id": 13, + "poly": [ + 608, + 248, + 658, + 248, + 658, + 273, + 608, + 273 + ], + "score": 0.83, + "latex": "y " + }, + { + "category_id": 13, + "poly": [ + 339, + 328, + 635, + 328, + 635, + 360, + 339, + 360 + ], + "score": 0.81, + "latex": "y _ { t } = \\sqrt { \\gamma _ { t } } * y + \\sqrt { 1 - \\gamma _ { t } } * \\epsilon" + }, + { + "category_id": 13, + "poly": [ + 297, + 1344, + 323, + 1344, + 323, + 1371, + 297, + 1371 + ], + "score": 0.81, + "latex": "y _ { t }" + }, + { + "category_id": 13, + "poly": [ + 610, + 723, + 632, + 723, + 632, + 749, + 610, + 749 + ], + "score": 0.77, + "latex": "\\textbf { { y } }" + }, + { + "category_id": 13, + "poly": [ + 339, + 248, + 392, + 248, + 392, + 270, + 339, + 270 + ], + "score": 0.76, + "latex": "x \\gets" + }, + { + "category_id": 13, + "poly": [ + 412, + 782, + 427, + 782, + 427, + 806, + 412, + 806 + ], + "score": 0.71, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 341, + 529, + 544, + 529, + 544, + 559, + 341, + 559 + ], + "score": 0.71, + "latex": "\\epsilon _ { p r e d } = f _ { \\theta } ( x , y _ { t } , t )" + }, + { + "category_id": 13, + "poly": [ + 991, + 723, + 1012, + 723, + 1012, + 745, + 991, + 745 + ], + "score": 0.68, + "latex": "_ { \\textbf { \\em x } }" + }, + { + "category_id": 13, + "poly": [ + 458, + 785, + 474, + 785, + 474, + 806, + 458, + 806 + ], + "score": 0.6, + "latex": "s" + }, + { + "category_id": 13, + "poly": [ + 338, + 273, + 600, + 273, + 600, + 303, + 338, + 303 + ], + "score": 0.6, + "latex": "t \\sim U ( 0 , 1 ) , \\epsilon \\sim N ( 0 , 1 )" + }, + { + "category_id": 13, + "poly": [ + 853, + 245, + 952, + 245, + 952, + 272, + 853, + 272 + ], + "score": 0.59, + "latex": "m a s k \\gets" + }, + { + "category_id": 13, + "poly": [ + 377, + 474, + 682, + 474, + 682, + 507, + 377, + 507 + ], + "score": 0.58, + "latex": "\\epsilon = ( y _ { t } - { \\sqrt { \\gamma _ { t } } } * y ) / { \\sqrt { 1 - \\gamma _ { t } } }" + }, + { + "category_id": 13, + "poly": [ + 385, + 562, + 407, + 562, + 407, + 582, + 385, + 582 + ], + "score": 0.49, + "latex": "=" + }, + { + "category_id": 14, + "poly": [ + 377, + 443, + 715, + 443, + 715, + 472, + 377, + 472 + ], + "score": 0.41, + "latex": "y _ { t } = \\sqrt { \\gamma _ { t } } * y _ { p r e d } + \\sqrt { 1 - \\gamma _ { t } } * \\epsilon" + }, + { + "category_id": 13, + "poly": [ + 375, + 443, + 718, + 443, + 718, + 472, + 375, + 472 + ], + "score": 0.36, + "latex": "y _ { t } = \\sqrt { \\gamma _ { t } } * y _ { p r e d } + \\sqrt { 1 - \\gamma _ { t } } * \\epsilon" + }, + { + "category_id": 13, + "poly": [ + 339, + 304, + 379, + 304, + 379, + 329, + 339, + 329 + ], + "score": 0.34, + "latex": "y =" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2060.0, + 860.0, + 2060.0, + 860.0, + 2091.0, + 839.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 192.0, + 1144.0, + 192.0, + 1144.0, + 246.0, + 292.0, + 246.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2092.0, + 838.0, + 2092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1308.0, + 1406.0, + 1308.0, + 1406.0, + 1345.0, + 295.0, + 1345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1339.0, + 296.0, + 1339.0, + 296.0, + 1374.0, + 292.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1339.0, + 916.0, + 1339.0, + 916.0, + 1374.0, + 324.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 946.0, + 1339.0, + 1406.0, + 1339.0, + 1406.0, + 1374.0, + 946.0, + 1374.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1371.0, + 1405.0, + 1371.0, + 1405.0, + 1405.0, + 295.0, + 1405.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1400.0, + 763.0, + 1400.0, + 763.0, + 1434.0, + 295.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 792.0, + 1400.0, + 1405.0, + 1400.0, + 1405.0, + 1434.0, + 792.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1430.0, + 1406.0, + 1430.0, + 1406.0, + 1464.0, + 293.0, + 1464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1459.0, + 893.0, + 1459.0, + 893.0, + 1497.0, + 293.0, + 1497.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 973.0, + 1459.0, + 1405.0, + 1459.0, + 1405.0, + 1497.0, + 973.0, + 1497.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1492.0, + 1405.0, + 1492.0, + 1405.0, + 1526.0, + 295.0, + 1526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1520.0, + 635.0, + 1520.0, + 635.0, + 1555.0, + 292.0, + 1555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 662.0, + 1520.0, + 1406.0, + 1520.0, + 1406.0, + 1555.0, + 662.0, + 1555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1552.0, + 1402.0, + 1552.0, + 1402.0, + 1586.0, + 295.0, + 1586.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1580.0, + 1406.0, + 1580.0, + 1406.0, + 1616.0, + 292.0, + 1616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1609.0, + 1406.0, + 1609.0, + 1406.0, + 1648.0, + 292.0, + 1648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1644.0, + 786.0, + 1644.0, + 786.0, + 1678.0, + 295.0, + 1678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 652.0, + 1406.0, + 652.0, + 1406.0, + 695.0, + 292.0, + 695.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 686.0, + 1405.0, + 686.0, + 1405.0, + 724.0, + 294.0, + 724.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 716.0, + 609.0, + 716.0, + 609.0, + 752.0, + 294.0, + 752.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 633.0, + 716.0, + 990.0, + 716.0, + 990.0, + 752.0, + 633.0, + 752.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1013.0, + 716.0, + 1405.0, + 716.0, + 1405.0, + 752.0, + 1013.0, + 752.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 746.0, + 361.0, + 746.0, + 361.0, + 786.0, + 292.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 746.0, + 1085.0, + 746.0, + 1085.0, + 786.0, + 395.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1188.0, + 746.0, + 1406.0, + 746.0, + 1406.0, + 786.0, + 1188.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 776.0, + 411.0, + 776.0, + 411.0, + 815.0, + 292.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 428.0, + 776.0, + 457.0, + 776.0, + 457.0, + 815.0, + 428.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 475.0, + 776.0, + 539.0, + 776.0, + 539.0, + 815.0, + 475.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 594.0, + 776.0, + 1242.0, + 776.0, + 1242.0, + 815.0, + 594.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1396.0, + 776.0, + 1407.0, + 776.0, + 1407.0, + 815.0, + 1396.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 806.0, + 1406.0, + 806.0, + 1406.0, + 846.0, + 292.0, + 846.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 838.0, + 1406.0, + 838.0, + 1406.0, + 875.0, + 294.0, + 875.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 866.0, + 1407.0, + 866.0, + 1407.0, + 910.0, + 292.0, + 910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 898.0, + 1405.0, + 898.0, + 1405.0, + 937.0, + 292.0, + 937.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 928.0, + 1407.0, + 928.0, + 1407.0, + 966.0, + 292.0, + 966.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 960.0, + 1405.0, + 960.0, + 1405.0, + 996.0, + 295.0, + 996.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 991.0, + 1405.0, + 991.0, + 1405.0, + 1027.0, + 294.0, + 1027.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1021.0, + 1347.0, + 1021.0, + 1347.0, + 1057.0, + 295.0, + 1057.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1066.0, + 1407.0, + 1066.0, + 1407.0, + 1102.0, + 294.0, + 1102.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1097.0, + 1404.0, + 1097.0, + 1404.0, + 1130.0, + 294.0, + 1130.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1129.0, + 1405.0, + 1129.0, + 1405.0, + 1161.0, + 295.0, + 1161.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1159.0, + 1306.0, + 1159.0, + 1306.0, + 1191.0, + 294.0, + 1191.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1199.0, + 1405.0, + 1199.0, + 1405.0, + 1239.0, + 293.0, + 1239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1232.0, + 1404.0, + 1232.0, + 1404.0, + 1269.0, + 294.0, + 1269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1264.0, + 924.0, + 1264.0, + 924.0, + 1298.0, + 294.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1687.0, + 1404.0, + 1687.0, + 1404.0, + 1722.0, + 296.0, + 1722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1715.0, + 1408.0, + 1715.0, + 1408.0, + 1755.0, + 292.0, + 1755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1749.0, + 1404.0, + 1749.0, + 1404.0, + 1783.0, + 294.0, + 1783.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1778.0, + 1406.0, + 1778.0, + 1406.0, + 1816.0, + 293.0, + 1816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1807.0, + 1405.0, + 1807.0, + 1405.0, + 1844.0, + 293.0, + 1844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1841.0, + 1406.0, + 1841.0, + 1406.0, + 1875.0, + 294.0, + 1875.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1871.0, + 573.0, + 1871.0, + 573.0, + 1901.0, + 294.0, + 1901.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 332.0, + 1912.0, + 628.0, + 1912.0, + 628.0, + 1950.0, + 332.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 664.0, + 1912.0, + 1405.0, + 1912.0, + 1405.0, + 1950.0, + 664.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1945.0, + 1403.0, + 1945.0, + 1403.0, + 1979.0, + 295.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1974.0, + 1404.0, + 1974.0, + 1404.0, + 2012.0, + 295.0, + 2012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 241.0, + 338.0, + 241.0, + 338.0, + 277.0, + 304.0, + 277.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 241.0, + 607.0, + 241.0, + 607.0, + 277.0, + 393.0, + 277.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 659.0, + 241.0, + 852.0, + 241.0, + 852.0, + 277.0, + 659.0, + 277.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 953.0, + 241.0, + 1252.0, + 241.0, + 1252.0, + 277.0, + 953.0, + 277.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 268.0, + 337.0, + 268.0, + 337.0, + 306.0, + 304.0, + 306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 303.0, + 299.0, + 338.0, + 299.0, + 338.0, + 331.0, + 303.0, + 331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 380.0, + 299.0, + 704.0, + 299.0, + 704.0, + 331.0, + 380.0, + 331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 325.0, + 338.0, + 325.0, + 338.0, + 363.0, + 302.0, + 363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 636.0, + 325.0, + 641.0, + 325.0, + 641.0, + 363.0, + 636.0, + 363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 303.0, + 352.0, + 536.0, + 352.0, + 536.0, + 387.0, + 303.0, + 387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 385.0, + 335.0, + 385.0, + 335.0, + 412.0, + 304.0, + 412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 384.0, + 728.0, + 384.0, + 728.0, + 415.0, + 374.0, + 415.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 409.0, + 336.0, + 409.0, + 336.0, + 441.0, + 304.0, + 441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 372.0, + 411.0, + 767.0, + 411.0, + 767.0, + 444.0, + 372.0, + 444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 440.0, + 336.0, + 440.0, + 336.0, + 472.0, + 304.0, + 472.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 472.0, + 336.0, + 472.0, + 336.0, + 503.0, + 302.0, + 503.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 373.0, + 474.0, + 376.0, + 474.0, + 376.0, + 508.0, + 373.0, + 508.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 498.0, + 414.0, + 498.0, + 414.0, + 533.0, + 296.0, + 533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 525.0, + 340.0, + 525.0, + 340.0, + 563.0, + 295.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 545.0, + 525.0, + 550.0, + 525.0, + 550.0, + 563.0, + 545.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 553.0, + 384.0, + 553.0, + 384.0, + 589.0, + 296.0, + 589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 408.0, + 553.0, + 749.0, + 553.0, + 749.0, + 589.0, + 408.0, + 589.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 4, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1175, + 1404, + 1175, + 1404, + 1541, + 298, + 1541 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 1656, + 1403, + 1656, + 1403, + 1840, + 298, + 1840 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 299, + 994, + 1405, + 994, + 1405, + 1087, + 299, + 1087 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 299, + 1854, + 1403, + 1854, + 1403, + 2007, + 299, + 2007 + ], + "score": 0.976 + }, + { + "category_id": 3, + "poly": [ + 299, + 190, + 1400, + 190, + 1400, + 877, + 299, + 877 + ], + "score": 0.968 + }, + { + "category_id": 4, + "poly": [ + 297, + 894, + 1399, + 894, + 1399, + 952, + 297, + 952 + ], + "score": 0.932 + }, + { + "category_id": 0, + "poly": [ + 299, + 1127, + 665, + 1127, + 665, + 1158, + 299, + 1158 + ], + "score": 0.913 + }, + { + "category_id": 0, + "poly": [ + 299, + 1586, + 532, + 1586, + 532, + 1624, + 299, + 1624 + ], + "score": 0.912 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 858, + 2062, + 858, + 2085, + 840, + 2085 + ], + "score": 0.787 + }, + { + "category_id": 13, + "poly": [ + 708, + 1388, + 1077, + 1388, + 1077, + 1423, + 708, + 1423 + ], + "score": 0.92, + "latex": "y _ { t ^ { \\prime } } \\sim \\mathcal { N } ( y _ { t ^ { \\prime } } ; \\sqrt { \\gamma _ { t ^ { \\prime } } } z , ( 1 - \\gamma _ { t ^ { \\prime } } ) I )" + }, + { + "category_id": 13, + "poly": [ + 1244, + 1854, + 1354, + 1854, + 1354, + 1884, + 1244, + 1884 + ], + "score": 0.89, + "latex": "3 2 0 \\times 4 4 8" + }, + { + "category_id": 13, + "poly": [ + 852, + 1945, + 963, + 1945, + 963, + 1975, + 852, + 1975 + ], + "score": 0.89, + "latex": "3 2 0 \\times 7 6 8" + }, + { + "category_id": 13, + "poly": [ + 688, + 997, + 722, + 997, + 722, + 1026, + 688, + 1026 + ], + "score": 0.88, + "latex": "L _ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 476, + 1389, + 498, + 1389, + 498, + 1416, + 476, + 1416 + ], + "score": 0.82, + "latex": "t ^ { \\prime }" + }, + { + "category_id": 13, + "poly": [ + 1267, + 1388, + 1289, + 1388, + 1289, + 1417, + 1267, + 1417 + ], + "score": 0.81, + "latex": "t ^ { \\prime }" + }, + { + "category_id": 13, + "poly": [ + 298, + 1748, + 392, + 1748, + 392, + 1782, + 298, + 1782 + ], + "score": 0.81, + "latex": "[ - 1 . , 1 . ]" + }, + { + "category_id": 13, + "poly": [ + 454, + 1363, + 472, + 1363, + 472, + 1385, + 454, + 1385 + ], + "score": 0.72, + "latex": "z" + }, + { + "category_id": 15, + "poly": [ + 390.0, + 197.0, + 485.0, + 197.0, + 485.0, + 221.0, + 390.0, + 221.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 656.0, + 195.0, + 772.0, + 195.0, + 772.0, + 222.0, + 656.0, + 222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 957.0, + 197.0, + 1017.0, + 197.0, + 1017.0, + 222.0, + 957.0, + 222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1199.0, + 195.0, + 1327.0, + 195.0, + 1327.0, + 224.0, + 1199.0, + 224.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 398.0, + 356.0, + 410.0, + 356.0, + 410.0, + 367.0, + 398.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 648.0, + 744.0, + 659.0, + 744.0, + 659.0, + 756.0, + 648.0, + 756.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 922.0, + 745.0, + 936.0, + 745.0, + 936.0, + 755.0, + 922.0, + 755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 544.0, + 754.0, + 561.0, + 754.0, + 561.0, + 768.0, + 544.0, + 768.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 657.0, + 771.5, + 703.0, + 771.5, + 703.0, + 795.0, + 657.0, + 795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 893.0, + 1405.0, + 893.0, + 1405.0, + 928.0, + 295.0, + 928.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 920.0, + 810.0, + 920.0, + 810.0, + 955.0, + 295.0, + 955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1124.0, + 668.0, + 1124.0, + 668.0, + 1163.0, + 294.0, + 1163.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1580.0, + 538.0, + 1580.0, + 538.0, + 1634.0, + 289.0, + 1634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2060.0, + 861.0, + 2060.0, + 861.0, + 2091.0, + 840.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1171.0, + 1405.0, + 1171.0, + 1405.0, + 1214.0, + 292.0, + 1214.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1205.0, + 1404.0, + 1205.0, + 1404.0, + 1242.0, + 293.0, + 1242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1233.0, + 1406.0, + 1233.0, + 1406.0, + 1273.0, + 293.0, + 1273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1266.0, + 1404.0, + 1266.0, + 1404.0, + 1301.0, + 293.0, + 1301.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1297.0, + 1404.0, + 1297.0, + 1404.0, + 1330.0, + 293.0, + 1330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1329.0, + 1405.0, + 1329.0, + 1405.0, + 1360.0, + 293.0, + 1360.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1358.0, + 453.0, + 1358.0, + 453.0, + 1392.0, + 293.0, + 1392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 473.0, + 1358.0, + 1405.0, + 1358.0, + 1405.0, + 1392.0, + 473.0, + 1392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1387.0, + 475.0, + 1387.0, + 475.0, + 1425.0, + 293.0, + 1425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 499.0, + 1387.0, + 707.0, + 1387.0, + 707.0, + 1425.0, + 499.0, + 1425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1078.0, + 1387.0, + 1266.0, + 1387.0, + 1266.0, + 1425.0, + 1078.0, + 1425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1290.0, + 1387.0, + 1405.0, + 1387.0, + 1405.0, + 1425.0, + 1290.0, + 1425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1416.0, + 1405.0, + 1416.0, + 1405.0, + 1456.0, + 292.0, + 1456.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1448.0, + 1405.0, + 1448.0, + 1405.0, + 1485.0, + 293.0, + 1485.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1479.0, + 1405.0, + 1479.0, + 1405.0, + 1513.0, + 294.0, + 1513.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1511.0, + 632.0, + 1511.0, + 632.0, + 1541.0, + 296.0, + 1541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1658.0, + 1404.0, + 1658.0, + 1404.0, + 1690.0, + 296.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1686.0, + 1405.0, + 1686.0, + 1405.0, + 1723.0, + 293.0, + 1723.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1715.0, + 1405.0, + 1715.0, + 1405.0, + 1754.0, + 293.0, + 1754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1747.0, + 297.0, + 1747.0, + 297.0, + 1785.0, + 294.0, + 1785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1747.0, + 1407.0, + 1747.0, + 1407.0, + 1785.0, + 393.0, + 1785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1778.0, + 1404.0, + 1778.0, + 1404.0, + 1810.0, + 296.0, + 1810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1806.0, + 1302.0, + 1806.0, + 1302.0, + 1844.0, + 293.0, + 1844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 990.0, + 687.0, + 990.0, + 687.0, + 1032.0, + 293.0, + 1032.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 723.0, + 990.0, + 1406.0, + 990.0, + 1406.0, + 1032.0, + 723.0, + 1032.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1026.0, + 1405.0, + 1026.0, + 1405.0, + 1060.0, + 295.0, + 1060.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1057.0, + 628.0, + 1057.0, + 628.0, + 1091.0, + 295.0, + 1091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1853.0, + 1243.0, + 1853.0, + 1243.0, + 1886.0, + 297.0, + 1886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1355.0, + 1853.0, + 1403.0, + 1853.0, + 1403.0, + 1886.0, + 1355.0, + 1886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1883.0, + 1405.0, + 1883.0, + 1405.0, + 1920.0, + 294.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1911.0, + 1407.0, + 1911.0, + 1407.0, + 1950.0, + 293.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1942.0, + 851.0, + 1942.0, + 851.0, + 1981.0, + 293.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 1942.0, + 1407.0, + 1942.0, + 1407.0, + 1981.0, + 964.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1974.0, + 1405.0, + 1974.0, + 1405.0, + 2011.0, + 295.0, + 2011.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 5, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1625, + 1404, + 1625, + 1404, + 1809, + 297, + 1809 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 1373, + 1404, + 1373, + 1404, + 1526, + 298, + 1526 + ], + "score": 0.978 + }, + { + "category_id": 5, + "poly": [ + 297, + 976, + 920, + 976, + 920, + 1204, + 297, + 1204 + ], + "score": 0.977, + "html": "
ModelDatasetSintel.clean Sintel.finalKITTI
AEPEAEPEFl-all
FlowFormer RAFTChairs-→Things1.012.404.0914.72%
Chairs-→Things1.682.805.921
Perceiver IO RAFTAutoFlow1.812.424.98-
AutoFlow1.742.414.1813.41%
RAFT(ours)AF→AF+FT+KU+TA1.272.282.719.16%
DDVM(ours) AF-→AF+FT+KU+TA1.242.002.197.58%
" + }, + { + "category_id": 1, + "poly": [ + 299, + 1823, + 1404, + 1823, + 1404, + 2008, + 299, + 2008 + ], + "score": 0.977 + }, + { + "category_id": 5, + "poly": [ + 958, + 1010, + 1400, + 1010, + 1400, + 1205, + 958, + 1205 + ], + "score": 0.975, + "html": "
MethodSintel.cleanSintel.finalKITTI
SKFlow [72]*1.302.264.84%
CRAFT [66]*1.442.424.79%
FlowFormer [23]1.142.184.68%
RAFT-OCTC [29]*1.512.574.33%
RAFT-it [70]$1.552.904.31%
DDVM (ours)$1.752.483.26%
" + }, + { + "category_id": 1, + "poly": [ + 298, + 1267, + 1404, + 1267, + 1404, + 1359, + 298, + 1359 + ], + "score": 0.972 + }, + { + "category_id": 6, + "poly": [ + 296, + 797, + 910, + 797, + 910, + 963, + 296, + 963 + ], + "score": 0.961 + }, + { + "category_id": 3, + "poly": [ + 298, + 188, + 1405, + 188, + 1405, + 667, + 298, + 667 + ], + "score": 0.96 + }, + { + "category_id": 6, + "poly": [ + 956, + 797, + 1403, + 797, + 1403, + 992, + 956, + 992 + ], + "score": 0.931 + }, + { + "category_id": 0, + "poly": [ + 301, + 1572, + 773, + 1572, + 773, + 1603, + 301, + 1603 + ], + "score": 0.908 + }, + { + "category_id": 4, + "poly": [ + 298, + 693, + 1399, + 693, + 1399, + 751, + 298, + 751 + ], + "score": 0.906 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 858, + 2061, + 858, + 2084, + 841, + 2084 + ], + "score": 0.714 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2084, + 841, + 2084 + ], + "score": 0.109 + }, + { + "category_id": 13, + "poly": [ + 917, + 1328, + 959, + 1328, + 959, + 1358, + 917, + 1358 + ], + "score": 0.87, + "latex": "5 \\%" + }, + { + "category_id": 13, + "poly": [ + 1233, + 1495, + 1355, + 1495, + 1355, + 1526, + 1233, + 1526 + ], + "score": 0.87, + "latex": "( \\delta _ { 1 } < 1 . 2 5 )" + }, + { + "category_id": 13, + "poly": [ + 1214, + 880, + 1248, + 880, + 1248, + 906, + 1214, + 906 + ], + "score": 0.87, + "latex": "2 ^ { \\mathrm { n d } }" + }, + { + "category_id": 13, + "poly": [ + 667, + 1298, + 695, + 1298, + 695, + 1327, + 667, + 1327 + ], + "score": 0.85, + "latex": "\\%" + }, + { + "category_id": 13, + "poly": [ + 1276, + 826, + 1364, + 826, + 1364, + 854, + 1276, + 854 + ], + "score": 0.65, + "latex": "( \\mathrm { A E P E } \\downarrow" + }, + { + "category_id": 13, + "poly": [ + 473, + 1062, + 496, + 1062, + 496, + 1079, + 473, + 1079 + ], + "score": 0.52, + "latex": "" + }, + { + "category_id": 13, + "poly": [ + 474, + 1085, + 496, + 1085, + 496, + 1101, + 474, + 1101 + ], + "score": 0.36, + "latex": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 799.0, + 907.0, + 799.0, + 907.0, + 828.0, + 296.0, + 828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 824.0, + 910.0, + 824.0, + 910.0, + 858.0, + 294.0, + 858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 854.0, + 910.0, + 854.0, + 910.0, + 884.0, + 295.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 882.0, + 910.0, + 882.0, + 910.0, + 912.0, + 294.0, + 912.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 910.0, + 909.0, + 910.0, + 909.0, + 938.0, + 295.0, + 938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 937.0, + 733.0, + 937.0, + 733.0, + 963.0, + 295.0, + 963.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 378.0, + 195.0, + 499.0, + 195.0, + 499.0, + 228.0, + 378.0, + 228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 642.0, + 194.0, + 787.0, + 194.0, + 787.0, + 227.0, + 642.0, + 227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 952.0, + 194.0, + 1028.0, + 194.0, + 1028.0, + 228.0, + 952.0, + 228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1183.0, + 191.0, + 1348.0, + 191.0, + 1348.0, + 231.0, + 1183.0, + 231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 954.0, + 797.0, + 1404.0, + 797.0, + 1404.0, + 828.0, + 954.0, + 828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 955.0, + 825.0, + 1275.0, + 825.0, + 1275.0, + 856.0, + 955.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1365.0, + 825.0, + 1404.0, + 825.0, + 1404.0, + 856.0, + 1365.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 955.0, + 853.0, + 1405.0, + 853.0, + 1405.0, + 881.0, + 955.0, + 881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 954.0, + 876.0, + 1213.0, + 876.0, + 1213.0, + 911.0, + 954.0, + 911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1249.0, + 876.0, + 1409.0, + 876.0, + 1409.0, + 911.0, + 1249.0, + 911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 954.0, + 908.0, + 1406.0, + 908.0, + 1406.0, + 939.0, + 954.0, + 939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 954.0, + 933.0, + 1405.0, + 933.0, + 1405.0, + 966.0, + 954.0, + 966.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 956.0, + 964.0, + 1312.0, + 964.0, + 1312.0, + 993.0, + 956.0, + 993.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1570.0, + 775.0, + 1570.0, + 775.0, + 1607.0, + 295.0, + 1607.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 691.0, + 1404.0, + 691.0, + 1404.0, + 726.0, + 294.0, + 726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 721.0, + 612.0, + 721.0, + 612.0, + 754.0, + 294.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2059.0, + 860.0, + 2059.0, + 860.0, + 2091.0, + 839.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2059.0, + 860.0, + 2059.0, + 860.0, + 2092.0, + 840.0, + 2092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1626.0, + 1406.0, + 1626.0, + 1406.0, + 1661.0, + 294.0, + 1661.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1657.0, + 1406.0, + 1657.0, + 1406.0, + 1693.0, + 294.0, + 1693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1685.0, + 1404.0, + 1685.0, + 1404.0, + 1723.0, + 294.0, + 1723.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1716.0, + 1406.0, + 1716.0, + 1406.0, + 1754.0, + 294.0, + 1754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1749.0, + 1406.0, + 1749.0, + 1406.0, + 1784.0, + 295.0, + 1784.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1779.0, + 867.0, + 1779.0, + 867.0, + 1811.0, + 296.0, + 1811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1370.0, + 1406.0, + 1370.0, + 1406.0, + 1409.0, + 293.0, + 1409.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1402.0, + 1401.0, + 1402.0, + 1401.0, + 1435.0, + 296.0, + 1435.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1435.0, + 1404.0, + 1435.0, + 1404.0, + 1468.0, + 296.0, + 1468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1463.0, + 1405.0, + 1463.0, + 1405.0, + 1500.0, + 294.0, + 1500.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1494.0, + 1232.0, + 1494.0, + 1232.0, + 1531.0, + 294.0, + 1531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1356.0, + 1494.0, + 1369.0, + 1494.0, + 1369.0, + 1531.0, + 1356.0, + 1531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1824.0, + 1405.0, + 1824.0, + 1405.0, + 1856.0, + 295.0, + 1856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1853.0, + 1410.0, + 1853.0, + 1410.0, + 1890.0, + 292.0, + 1890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1882.0, + 1408.0, + 1882.0, + 1408.0, + 1921.0, + 293.0, + 1921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1913.0, + 1409.0, + 1913.0, + 1409.0, + 1949.0, + 293.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1945.0, + 1409.0, + 1945.0, + 1409.0, + 1980.0, + 294.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1973.0, + 1406.0, + 1973.0, + 1406.0, + 2012.0, + 294.0, + 2012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1265.0, + 1408.0, + 1265.0, + 1408.0, + 1302.0, + 294.0, + 1302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1299.0, + 666.0, + 1299.0, + 666.0, + 1332.0, + 296.0, + 1332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 696.0, + 1299.0, + 1406.0, + 1299.0, + 1406.0, + 1332.0, + 696.0, + 1332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1329.0, + 916.0, + 1329.0, + 916.0, + 1362.0, + 294.0, + 1362.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 960.0, + 1329.0, + 1214.0, + 1329.0, + 1214.0, + 1362.0, + 960.0, + 1362.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 6, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1672, + 1404, + 1672, + 1404, + 2007, + 297, + 2007 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 1212, + 1404, + 1212, + 1404, + 1577, + 298, + 1577 + ], + "score": 0.983 + }, + { + "category_id": 5, + "poly": [ + 297, + 800, + 927, + 800, + 927, + 982, + 297, + 982 + ], + "score": 0.979, + "html": "
NYU valKITTI valKITTI val
RELRMSRELRMSAEPEFl-all
Baseline0.0790.3310.2223.770--
Step-unroll0.0760.3240.0852.8441.846.16%
Infill0.0770.3380.0572.7441.535.24%
Step-unroll & infill0.0750.3240.0562.7001.474.74%
" + }, + { + "category_id": 5, + "poly": [ + 420, + 313, + 1276, + 313, + 1276, + 692, + 420, + 692 + ], + "score": 0.978, + "html": "
MethodArchitectureNYU-Depth-v2KITTI
81↑REL↓RMS↓81↑REL↓RMS↓
TransDepth [87]Res-50+ViT-B†0.900 0.1060.3650.9560.0642.755
DPT[50]Res-50+ViT-Bt‡0.904 0.1100.3570.959 0.0622.573
BTS [34]DenseNet-161t0.885 0.1100.3920.9560.0592.756
AdaBins [3]E-B5+Mini-ViT†0.903 0.1030.3640.9640.0582.360
BinsFormer[35]Swin-Large†0.925 0.0940.3300.9740.0522.098
PixelFormer [1]Swin-Larget0.929 0.0900.3220.9760.0512.081
MIM [80]SwinV2-LT0.949 0.0830.2870.9770.0501.966
AiT-P [46]SwinV2-LT0.953 0.0760.279111
DDVM samples=2samples=1 Efficient U-NetT#0.9440.0750.3240.9640.0562.700
Efficient U-NetT#0.9440.0740.3190.9650.0552.660
samples=4 Efficient U-NetT#0.9460.0740.3150.9650.0552.613
" + }, + { + "category_id": 5, + "poly": [ + 955, + 843, + 1395, + 843, + 1395, + 982, + 955, + 982 + ], + "score": 0.972, + "html": "
Coarse-to-fineSintel.clean Sintel.finalKITTI
refinementAEPEAEPEAEPE Fl-all
Without1.422.122.35 8.65%
With1.242.002.19 7.58%
" + }, + { + "category_id": 1, + "poly": [ + 302, + 1107, + 1398, + 1107, + 1398, + 1198, + 302, + 1198 + ], + "score": 0.961 + }, + { + "category_id": 6, + "poly": [ + 299, + 216, + 1403, + 216, + 1403, + 301, + 299, + 301 + ], + "score": 0.955 + }, + { + "category_id": 1, + "poly": [ + 297, + 1031, + 1401, + 1031, + 1401, + 1092, + 297, + 1092 + ], + "score": 0.947 + }, + { + "category_id": 6, + "poly": [ + 296, + 713, + 919, + 713, + 919, + 796, + 296, + 796 + ], + "score": 0.947 + }, + { + "category_id": 6, + "poly": [ + 951, + 713, + 1399, + 713, + 1399, + 826, + 951, + 826 + ], + "score": 0.944 + }, + { + "category_id": 0, + "poly": [ + 298, + 1618, + 537, + 1618, + 537, + 1651, + 298, + 1651 + ], + "score": 0.906 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 858, + 2062, + 858, + 2085, + 840, + 2085 + ], + "score": 0.804 + }, + { + "category_id": 13, + "poly": [ + 297, + 1794, + 404, + 1794, + 404, + 1823, + 297, + 1823 + ], + "score": 0.88, + "latex": "3 2 0 \\times 4 4 8" + }, + { + "category_id": 13, + "poly": [ + 644, + 1546, + 745, + 1546, + 745, + 1578, + 644, + 1578 + ], + "score": 0.85, + "latex": "( 3 0 . 3 4 \\% )" + }, + { + "category_id": 13, + "poly": [ + 942, + 245, + 959, + 245, + 959, + 273, + 942, + 273 + ], + "score": 0.84, + "latex": "^ \\ddag" + }, + { + "category_id": 13, + "poly": [ + 1154, + 273, + 1172, + 273, + 1172, + 301, + 1154, + 301 + ], + "score": 0.55, + "latex": "\\uparrow" + }, + { + "category_id": 13, + "poly": [ + 1143, + 217, + 1168, + 217, + 1168, + 242, + 1143, + 242 + ], + "score": 0.46, + "latex": "\\top" + }, + { + "category_id": 13, + "poly": [ + 978, + 273, + 1002, + 273, + 1002, + 301, + 978, + 301 + ], + "score": 0.41, + "latex": "\\downarrow \\colon" + }, + { + "category_id": 13, + "poly": [ + 879, + 217, + 908, + 217, + 908, + 243, + 879, + 243 + ], + "score": 0.3, + "latex": "\\mathbf { \\nabla } \\cdot \\mathbf { v } _ { 2 }" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 216.0, + 878.0, + 216.0, + 878.0, + 248.0, + 294.0, + 248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 909.0, + 216.0, + 1142.0, + 216.0, + 1142.0, + 248.0, + 909.0, + 248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1169.0, + 216.0, + 1405.0, + 216.0, + 1405.0, + 248.0, + 1169.0, + 248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 242.0, + 941.0, + 242.0, + 941.0, + 275.0, + 294.0, + 275.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 960.0, + 242.0, + 1403.0, + 242.0, + 1403.0, + 275.0, + 960.0, + 275.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 270.0, + 977.0, + 270.0, + 977.0, + 305.0, + 293.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1003.0, + 270.0, + 1153.0, + 270.0, + 1153.0, + 305.0, + 1003.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1173.0, + 270.0, + 1343.0, + 270.0, + 1343.0, + 305.0, + 1173.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 711.0, + 922.0, + 711.0, + 922.0, + 745.0, + 294.0, + 745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 742.0, + 920.0, + 742.0, + 920.0, + 772.0, + 297.0, + 772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 769.0, + 652.0, + 769.0, + 652.0, + 798.0, + 296.0, + 798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 948.0, + 711.0, + 1403.0, + 711.0, + 1403.0, + 743.0, + 948.0, + 743.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 947.0, + 741.0, + 1402.0, + 741.0, + 1402.0, + 771.0, + 947.0, + 771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 949.0, + 769.0, + 1400.0, + 769.0, + 1400.0, + 798.0, + 949.0, + 798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 949.0, + 797.0, + 1399.0, + 797.0, + 1399.0, + 826.0, + 949.0, + 826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1611.0, + 541.0, + 1611.0, + 541.0, + 1661.0, + 292.0, + 1661.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2090.0, + 838.0, + 2090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1673.0, + 1405.0, + 1673.0, + 1405.0, + 1707.0, + 295.0, + 1707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1701.0, + 1406.0, + 1701.0, + 1406.0, + 1739.0, + 294.0, + 1739.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1733.0, + 1405.0, + 1733.0, + 1405.0, + 1767.0, + 292.0, + 1767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1764.0, + 1406.0, + 1764.0, + 1406.0, + 1799.0, + 295.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 1791.0, + 1404.0, + 1791.0, + 1404.0, + 1828.0, + 405.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1821.0, + 1405.0, + 1821.0, + 1405.0, + 1862.0, + 291.0, + 1862.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1853.0, + 1404.0, + 1853.0, + 1404.0, + 1890.0, + 292.0, + 1890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1884.0, + 1407.0, + 1884.0, + 1407.0, + 1922.0, + 292.0, + 1922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1914.0, + 1404.0, + 1914.0, + 1404.0, + 1952.0, + 294.0, + 1952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1946.0, + 1405.0, + 1946.0, + 1405.0, + 1984.0, + 294.0, + 1984.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1977.0, + 1024.0, + 1977.0, + 1024.0, + 2011.0, + 295.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1209.0, + 1409.0, + 1209.0, + 1409.0, + 1250.0, + 292.0, + 1250.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1242.0, + 1405.0, + 1242.0, + 1405.0, + 1278.0, + 293.0, + 1278.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1273.0, + 1405.0, + 1273.0, + 1405.0, + 1307.0, + 294.0, + 1307.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1305.0, + 1406.0, + 1305.0, + 1406.0, + 1339.0, + 294.0, + 1339.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1335.0, + 1404.0, + 1335.0, + 1404.0, + 1369.0, + 294.0, + 1369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1364.0, + 1408.0, + 1364.0, + 1408.0, + 1402.0, + 293.0, + 1402.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1393.0, + 1408.0, + 1393.0, + 1408.0, + 1431.0, + 295.0, + 1431.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1425.0, + 1406.0, + 1425.0, + 1406.0, + 1462.0, + 293.0, + 1462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1454.0, + 1406.0, + 1454.0, + 1406.0, + 1490.0, + 293.0, + 1490.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1487.0, + 1405.0, + 1487.0, + 1405.0, + 1520.0, + 294.0, + 1520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1516.0, + 1402.0, + 1516.0, + 1402.0, + 1549.0, + 294.0, + 1549.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1545.0, + 643.0, + 1545.0, + 643.0, + 1581.0, + 293.0, + 1581.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 746.0, + 1545.0, + 758.0, + 1545.0, + 758.0, + 1581.0, + 746.0, + 1581.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1106.0, + 1404.0, + 1106.0, + 1404.0, + 1142.0, + 294.0, + 1142.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1138.0, + 1403.0, + 1138.0, + 1403.0, + 1172.0, + 297.0, + 1172.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1168.0, + 1222.0, + 1168.0, + 1222.0, + 1201.0, + 296.0, + 1201.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1029.0, + 1403.0, + 1029.0, + 1403.0, + 1064.0, + 295.0, + 1064.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1059.0, + 524.0, + 1059.0, + 524.0, + 1094.0, + 295.0, + 1094.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 7, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1293, + 1403, + 1293, + 1403, + 1476, + 298, + 1476 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 298, + 1491, + 1405, + 1491, + 1405, + 1796, + 298, + 1796 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 298, + 968, + 1405, + 968, + 1405, + 1212, + 298, + 1212 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 300, + 1914, + 1402, + 1914, + 1402, + 2007, + 300, + 2007 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 298, + 862, + 1400, + 862, + 1400, + 953, + 298, + 953 + ], + "score": 0.956 + }, + { + "category_id": 4, + "poly": [ + 296, + 741, + 1401, + 741, + 1401, + 797, + 296, + 797 + ], + "score": 0.942 + }, + { + "category_id": 6, + "poly": [ + 967, + 196, + 1404, + 196, + 1404, + 282, + 967, + 282 + ], + "score": 0.941 + }, + { + "category_id": 6, + "poly": [ + 295, + 197, + 921, + 197, + 921, + 254, + 295, + 254 + ], + "score": 0.923 + }, + { + "category_id": 0, + "poly": [ + 298, + 1842, + 515, + 1842, + 515, + 1879, + 298, + 1879 + ], + "score": 0.901 + }, + { + "category_id": 0, + "poly": [ + 301, + 1245, + 850, + 1245, + 850, + 1277, + 301, + 1277 + ], + "score": 0.867 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 858, + 2062, + 858, + 2084, + 840, + 2084 + ], + "score": 0.792 + }, + { + "category_id": 3, + "poly": [ + 295, + 271, + 1403, + 271, + 1403, + 719, + 295, + 719 + ], + "score": 0.676 + }, + { + "category_id": 13, + "poly": [ + 762, + 1522, + 934, + 1522, + 934, + 1556, + 762, + 1556 + ], + "score": 0.92, + "latex": "p ( y | x , y _ { p a r t i a l } )" + }, + { + "category_id": 13, + "poly": [ + 1327, + 1491, + 1403, + 1491, + 1403, + 1524, + 1327, + 1524 + ], + "score": 0.91, + "latex": "p ( y | x )" + }, + { + "category_id": 13, + "poly": [ + 701, + 1587, + 728, + 1587, + 728, + 1614, + 701, + 1614 + ], + "score": 0.85, + "latex": "y _ { t }" + }, + { + "category_id": 13, + "poly": [ + 1011, + 1525, + 1099, + 1525, + 1099, + 1556, + 1011, + 1556 + ], + "score": 0.83, + "latex": "y _ { p a r t i a l }" + }, + { + "category_id": 13, + "poly": [ + 1262, + 1121, + 1291, + 1121, + 1291, + 1151, + 1262, + 1151 + ], + "score": 0.58, + "latex": "_ { c f }" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 737.0, + 1405.0, + 737.0, + 1405.0, + 773.0, + 294.0, + 773.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 769.0, + 872.0, + 769.0, + 872.0, + 799.0, + 295.0, + 799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 194.0, + 1403.0, + 194.0, + 1403.0, + 227.0, + 964.0, + 227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 966.0, + 224.0, + 1407.0, + 224.0, + 1407.0, + 254.0, + 966.0, + 254.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 254.0, + 1387.0, + 254.0, + 1387.0, + 282.0, + 964.0, + 282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 195.0, + 921.0, + 195.0, + 921.0, + 226.0, + 295.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 225.0, + 915.0, + 225.0, + 915.0, + 254.0, + 295.0, + 254.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1838.0, + 520.0, + 1838.0, + 520.0, + 1885.0, + 291.0, + 1885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1243.0, + 855.0, + 1243.0, + 855.0, + 1283.0, + 295.0, + 1283.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2060.0, + 860.0, + 2060.0, + 860.0, + 2090.0, + 839.0, + 2090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 305.0, + 280.0, + 377.0, + 280.0, + 377.0, + 308.0, + 305.0, + 308.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 510.0, + 277.0, + 615.0, + 277.0, + 615.0, + 308.0, + 510.0, + 308.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 626.0, + 281.0, + 722.0, + 281.0, + 722.0, + 305.0, + 626.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 734.0, + 280.0, + 799.0, + 280.0, + 799.0, + 308.0, + 734.0, + 308.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 811.0, + 281.0, + 919.0, + 281.0, + 919.0, + 305.0, + 811.0, + 305.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 305.0, + 311.0, + 434.0, + 311.0, + 434.0, + 347.0, + 305.0, + 347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 536.0, + 313.0, + 586.0, + 313.0, + 586.0, + 343.0, + 536.0, + 343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 649.0, + 314.0, + 698.0, + 314.0, + 698.0, + 345.0, + 649.0, + 345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 742.0, + 313.0, + 791.0, + 313.0, + 791.0, + 343.0, + 742.0, + 343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 829.0, + 314.0, + 902.0, + 314.0, + 902.0, + 343.0, + 829.0, + 343.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 983.0, + 300.0, + 1056.0, + 300.0, + 1056.0, + 332.0, + 983.0, + 332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1255.0, + 301.0, + 1308.0, + 301.0, + 1308.0, + 330.0, + 1255.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1331.0, + 301.0, + 1386.0, + 301.0, + 1386.0, + 330.0, + 1331.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 305.0, + 337.0, + 425.0, + 337.0, + 425.0, + 365.0, + 305.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 336.0, + 587.0, + 336.0, + 587.0, + 367.0, + 538.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 649.0, + 336.0, + 699.0, + 336.0, + 699.0, + 367.0, + 649.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 741.0, + 336.0, + 791.0, + 336.0, + 791.0, + 367.0, + 741.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 830.0, + 337.0, + 902.0, + 337.0, + 902.0, + 365.0, + 830.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 984.0, + 337.0, + 1136.0, + 337.0, + 1136.0, + 365.0, + 984.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1252.0, + 336.0, + 1311.0, + 336.0, + 1311.0, + 365.0, + 1252.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1328.0, + 336.0, + 1388.0, + 336.0, + 1388.0, + 365.0, + 1328.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 305.0, + 359.0, + 462.0, + 359.0, + 462.0, + 385.0, + 305.0, + 385.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 358.0, + 587.0, + 358.0, + 587.0, + 389.0, + 537.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 649.0, + 358.0, + 699.0, + 358.0, + 699.0, + 389.0, + 649.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 742.0, + 358.0, + 792.0, + 358.0, + 792.0, + 389.0, + 742.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 833.0, + 358.0, + 898.0, + 358.0, + 898.0, + 387.0, + 833.0, + 387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 982.0, + 359.0, + 1062.0, + 359.0, + 1062.0, + 387.0, + 982.0, + 387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1254.0, + 360.0, + 1308.0, + 360.0, + 1308.0, + 386.0, + 1254.0, + 386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1330.0, + 358.0, + 1389.0, + 358.0, + 1389.0, + 387.0, + 1330.0, + 387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 385.0, + 495.0, + 385.0, + 495.0, + 408.0, + 308.0, + 408.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 537.0, + 381.0, + 586.0, + 381.0, + 586.0, + 412.0, + 537.0, + 412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 651.0, + 382.0, + 698.0, + 382.0, + 698.0, + 409.0, + 651.0, + 409.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 742.0, + 381.0, + 791.0, + 381.0, + 791.0, + 412.0, + 742.0, + 412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 831.0, + 380.0, + 899.0, + 380.0, + 899.0, + 413.0, + 831.0, + 413.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 982.0, + 381.0, + 1224.0, + 381.0, + 1224.0, + 412.0, + 982.0, + 412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1252.0, + 381.0, + 1311.0, + 381.0, + 1311.0, + 411.0, + 1252.0, + 411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1330.0, + 381.0, + 1389.0, + 381.0, + 1389.0, + 412.0, + 1330.0, + 412.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 390.0, + 440.0, + 485.0, + 440.0, + 485.0, + 464.0, + 390.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 655.0, + 438.0, + 771.0, + 438.0, + 771.0, + 466.0, + 655.0, + 466.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 437.0, + 1013.0, + 437.0, + 1013.0, + 466.0, + 964.0, + 466.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1225.0, + 438.0, + 1299.0, + 438.0, + 1299.0, + 466.0, + 1225.0, + 466.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 963.0, + 260.0, + 1086.0, + 260.0, + 1086.0, + 283.5, + 963.0, + 283.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1082.0, + 263.0, + 1172.0, + 263.0, + 1172.0, + 281.5, + 1082.0, + 281.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1295.0, + 1403.0, + 1295.0, + 1403.0, + 1326.0, + 296.0, + 1326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1323.0, + 1405.0, + 1323.0, + 1405.0, + 1361.0, + 294.0, + 1361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1355.0, + 1404.0, + 1355.0, + 1404.0, + 1391.0, + 294.0, + 1391.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1386.0, + 1405.0, + 1386.0, + 1405.0, + 1421.0, + 294.0, + 1421.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1417.0, + 1404.0, + 1417.0, + 1404.0, + 1449.0, + 296.0, + 1449.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1448.0, + 754.0, + 1448.0, + 754.0, + 1479.0, + 296.0, + 1479.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1488.0, + 1326.0, + 1488.0, + 1326.0, + 1528.0, + 292.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1404.0, + 1488.0, + 1407.0, + 1488.0, + 1407.0, + 1528.0, + 1404.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1521.0, + 761.0, + 1521.0, + 761.0, + 1560.0, + 293.0, + 1560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 935.0, + 1521.0, + 1010.0, + 1521.0, + 1010.0, + 1560.0, + 935.0, + 1560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 1521.0, + 1410.0, + 1521.0, + 1410.0, + 1560.0, + 1100.0, + 1560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1549.0, + 1408.0, + 1549.0, + 1408.0, + 1589.0, + 292.0, + 1589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1579.0, + 700.0, + 1579.0, + 700.0, + 1622.0, + 292.0, + 1622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 729.0, + 1579.0, + 1406.0, + 1579.0, + 1406.0, + 1622.0, + 729.0, + 1622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1611.0, + 1406.0, + 1611.0, + 1406.0, + 1649.0, + 293.0, + 1649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1641.0, + 1407.0, + 1641.0, + 1407.0, + 1680.0, + 293.0, + 1680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1671.0, + 1406.0, + 1671.0, + 1406.0, + 1711.0, + 292.0, + 1711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1704.0, + 1405.0, + 1704.0, + 1405.0, + 1740.0, + 295.0, + 1740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1735.0, + 1405.0, + 1735.0, + 1405.0, + 1770.0, + 292.0, + 1770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1764.0, + 1397.0, + 1764.0, + 1397.0, + 1800.0, + 292.0, + 1800.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 967.0, + 1405.0, + 967.0, + 1405.0, + 1004.0, + 293.0, + 1004.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 997.0, + 1407.0, + 997.0, + 1407.0, + 1035.0, + 293.0, + 1035.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1029.0, + 1403.0, + 1029.0, + 1403.0, + 1063.0, + 295.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1060.0, + 1405.0, + 1060.0, + 1405.0, + 1094.0, + 295.0, + 1094.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1089.0, + 1408.0, + 1089.0, + 1408.0, + 1125.0, + 293.0, + 1125.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1120.0, + 1261.0, + 1120.0, + 1261.0, + 1153.0, + 295.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1292.0, + 1120.0, + 1408.0, + 1120.0, + 1408.0, + 1153.0, + 1292.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1147.0, + 1406.0, + 1147.0, + 1406.0, + 1187.0, + 293.0, + 1187.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1182.0, + 1294.0, + 1182.0, + 1294.0, + 1213.0, + 293.0, + 1213.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1913.0, + 1406.0, + 1913.0, + 1406.0, + 1951.0, + 295.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1945.0, + 1406.0, + 1945.0, + 1406.0, + 1981.0, + 295.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1975.0, + 1374.0, + 1975.0, + 1374.0, + 2010.0, + 295.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 862.0, + 1404.0, + 862.0, + 1404.0, + 896.0, + 297.0, + 896.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 893.0, + 1405.0, + 893.0, + 1405.0, + 924.0, + 293.0, + 924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 923.0, + 1084.0, + 923.0, + 1084.0, + 954.0, + 294.0, + 954.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 8, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1762, + 1404, + 1762, + 1404, + 2007, + 297, + 2007 + ], + "score": 0.98 + }, + { + "category_id": 5, + "poly": [ + 614, + 1332, + 1078, + 1332, + 1078, + 1655, + 614, + 1655 + ], + "score": 0.978, + "html": "
SequenceOursFlowFormer [23]
Perturbed Market 30.7870.869
Perturbed Shaman 10.2190.252
Ambush 129.338.141
Ambush 32.8552.973
Bamboo 30.4150.577
Cave 32.0422.352
Market 10.7191.174
Market 45.5178.768
Mountain 20.1760.518
Temple 10.4520.612
Tiger0.4130.596
Wall1.6391.723
" + }, + { + "category_id": 3, + "poly": [ + 304, + 202, + 1394, + 202, + 1394, + 508, + 304, + 508 + ], + "score": 0.967 + }, + { + "category_id": 1, + "poly": [ + 299, + 755, + 1401, + 755, + 1401, + 848, + 299, + 848 + ], + "score": 0.965 + }, + { + "category_id": 4, + "poly": [ + 297, + 531, + 1405, + 531, + 1405, + 700, + 297, + 700 + ], + "score": 0.964 + }, + { + "category_id": 0, + "poly": [ + 297, + 1711, + 510, + 1711, + 510, + 1748, + 297, + 1748 + ], + "score": 0.912 + }, + { + "category_id": 6, + "poly": [ + 310, + 1295, + 1378, + 1295, + 1378, + 1326, + 310, + 1326 + ], + "score": 0.89 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.856 + }, + { + "category_id": 1, + "poly": [ + 296, + 877, + 1405, + 877, + 1405, + 1260, + 296, + 1260 + ], + "score": 0.702 + }, + { + "category_id": 13, + "poly": [ + 656, + 1854, + 711, + 1854, + 711, + 1884, + 656, + 1884 + ], + "score": 0.88, + "latex": "2 5 \\%" + }, + { + "category_id": 13, + "poly": [ + 392, + 1854, + 466, + 1854, + 466, + 1884, + 392, + 1884 + ], + "score": 0.88, + "latex": "3 . 2 6 \\%" + }, + { + "category_id": 15, + "poly": [ + 301.0, + 200.0, + 429.0, + 200.0, + 429.0, + 224.0, + 301.0, + 224.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 227.0, + 403.0, + 227.0, + 403.0, + 247.0, + 320.0, + 247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1180.0, + 226.0, + 1235.0, + 226.0, + 1235.0, + 247.0, + 1180.0, + 247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1179.0, + 242.0, + 1233.0, + 242.0, + 1233.0, + 267.0, + 1179.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 938.0, + 256.0, + 1041.0, + 256.0, + 1041.0, + 280.0, + 938.0, + 280.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 338.0, + 285.0, + 438.0, + 285.0, + 438.0, + 327.0, + 338.0, + 327.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 278.0, + 661.0, + 278.0, + 661.0, + 337.0, + 570.0, + 337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 928.0, + 275.0, + 1052.0, + 275.0, + 1052.0, + 316.0, + 928.0, + 316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1077.0, + 299.0, + 1109.0, + 299.0, + 1109.0, + 316.0, + 1077.0, + 316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1116.0, + 299.0, + 1168.0, + 299.0, + 1168.0, + 316.0, + 1116.0, + 316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1308.0, + 303.0, + 1327.0, + 303.0, + 1327.0, + 312.0, + 1308.0, + 312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 435.0, + 331.0, + 493.0, + 331.0, + 493.0, + 358.0, + 435.0, + 358.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 508.0, + 330.0, + 565.0, + 330.0, + 565.0, + 361.0, + 508.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1231.0, + 356.0, + 1285.0, + 356.0, + 1285.0, + 378.0, + 1231.0, + 378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1301.0, + 358.0, + 1392.0, + 358.0, + 1392.0, + 382.0, + 1301.0, + 382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 456.0, + 449.0, + 548.0, + 449.0, + 548.0, + 469.0, + 456.0, + 469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 797.0, + 446.0, + 938.0, + 446.0, + 938.0, + 476.0, + 797.0, + 476.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1306.0, + 488.0, + 1336.0, + 488.0, + 1336.0, + 501.0, + 1306.0, + 501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 463.0, + 296.0, + 536.0, + 296.0, + 536.0, + 317.5, + 463.0, + 317.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 910.0, + 413.5, + 917.0, + 413.5, + 917.0, + 426.5, + 910.0, + 426.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 531.0, + 1405.0, + 531.0, + 1405.0, + 565.0, + 296.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 559.0, + 1406.0, + 559.0, + 1406.0, + 593.0, + 294.0, + 593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 587.0, + 1408.0, + 587.0, + 1408.0, + 620.0, + 292.0, + 620.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 613.0, + 1406.0, + 613.0, + 1406.0, + 648.0, + 294.0, + 648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 641.0, + 1407.0, + 641.0, + 1407.0, + 678.0, + 294.0, + 678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 672.0, + 1401.0, + 672.0, + 1401.0, + 702.0, + 296.0, + 702.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1706.0, + 514.0, + 1706.0, + 514.0, + 1756.0, + 291.0, + 1756.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 312.0, + 1293.0, + 1383.0, + 1293.0, + 1383.0, + 1331.0, + 312.0, + 1331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1764.0, + 1404.0, + 1764.0, + 1404.0, + 1797.0, + 295.0, + 1797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1795.0, + 1406.0, + 1795.0, + 1406.0, + 1828.0, + 295.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1823.0, + 1405.0, + 1823.0, + 1405.0, + 1857.0, + 292.0, + 1857.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1855.0, + 391.0, + 1855.0, + 391.0, + 1888.0, + 294.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 467.0, + 1855.0, + 655.0, + 1855.0, + 655.0, + 1888.0, + 467.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 712.0, + 1855.0, + 1402.0, + 1855.0, + 1402.0, + 1888.0, + 712.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1883.0, + 1405.0, + 1883.0, + 1405.0, + 1919.0, + 293.0, + 1919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1916.0, + 1404.0, + 1916.0, + 1404.0, + 1949.0, + 295.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1947.0, + 1404.0, + 1947.0, + 1404.0, + 1980.0, + 292.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1974.0, + 1042.0, + 1974.0, + 1042.0, + 2009.0, + 294.0, + 2009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 754.0, + 1405.0, + 754.0, + 1405.0, + 788.0, + 294.0, + 788.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 783.0, + 1405.0, + 783.0, + 1405.0, + 821.0, + 294.0, + 821.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 811.0, + 1406.0, + 811.0, + 1406.0, + 855.0, + 293.0, + 855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 876.0, + 1405.0, + 876.0, + 1405.0, + 914.0, + 296.0, + 914.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 907.0, + 1405.0, + 907.0, + 1405.0, + 944.0, + 320.0, + 944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 320.0, + 938.0, + 1406.0, + 938.0, + 1406.0, + 974.0, + 320.0, + 974.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 972.0, + 1238.0, + 972.0, + 1238.0, + 1002.0, + 323.0, + 1002.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 310.0, + 1014.0, + 1405.0, + 1014.0, + 1405.0, + 1050.0, + 310.0, + 1050.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1043.0, + 1405.0, + 1043.0, + 1405.0, + 1080.0, + 322.0, + 1080.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1077.0, + 1406.0, + 1077.0, + 1406.0, + 1109.0, + 323.0, + 1109.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1104.0, + 1405.0, + 1104.0, + 1405.0, + 1141.0, + 322.0, + 1141.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 1139.0, + 1406.0, + 1139.0, + 1406.0, + 1170.0, + 324.0, + 1170.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1169.0, + 1406.0, + 1169.0, + 1406.0, + 1198.0, + 323.0, + 1198.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 322.0, + 1196.0, + 1410.0, + 1196.0, + 1410.0, + 1229.0, + 322.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 323.0, + 1228.0, + 1304.0, + 1228.0, + 1304.0, + 1263.0, + 323.0, + 1263.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 9, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 299, + 455, + 1409, + 455, + 1409, + 2015, + 299, + 2015 + ], + "score": 0.961 + }, + { + "category_id": 1, + "poly": [ + 298, + 258, + 1404, + 258, + 1404, + 351, + 298, + 351 + ], + "score": 0.948 + }, + { + "category_id": 0, + "poly": [ + 299, + 204, + 529, + 204, + 529, + 235, + 299, + 235 + ], + "score": 0.916 + }, + { + "category_id": 0, + "poly": [ + 298, + 401, + 455, + 401, + 455, + 437, + 298, + 437 + ], + "score": 0.9 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 863, + 2061, + 863, + 2086, + 836, + 2086 + ], + "score": 0.55 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 863, + 2061, + 863, + 2086, + 836, + 2086 + ], + "score": 0.55 + }, + { + "category_id": 15, + "poly": [ + 295.0, + 201.0, + 534.0, + 201.0, + 534.0, + 240.0, + 295.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 399.0, + 460.0, + 399.0, + 460.0, + 442.0, + 295.0, + 442.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2098.0, + 832.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2098.0, + 832.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 309.0, + 455.0, + 1404.0, + 455.0, + 1404.0, + 493.0, + 309.0, + 493.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 483.0, + 887.0, + 483.0, + 887.0, + 526.0, + 349.0, + 526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 533.0, + 1407.0, + 533.0, + 1407.0, + 580.0, + 306.0, + 580.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 568.0, + 1088.0, + 568.0, + 1088.0, + 609.0, + 350.0, + 609.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 614.0, + 1409.0, + 614.0, + 1409.0, + 666.0, + 304.0, + 666.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 645.0, + 914.0, + 645.0, + 914.0, + 692.0, + 351.0, + 692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 695.0, + 1409.0, + 695.0, + 1409.0, + 747.0, + 304.0, + 747.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 730.0, + 1186.0, + 730.0, + 1186.0, + 772.0, + 352.0, + 772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 307.0, + 782.0, + 1409.0, + 782.0, + 1409.0, + 825.0, + 307.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 813.0, + 1409.0, + 813.0, + 1409.0, + 856.0, + 350.0, + 856.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 841.0, + 567.0, + 841.0, + 567.0, + 884.0, + 353.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 309.0, + 898.0, + 1406.0, + 898.0, + 1406.0, + 936.0, + 309.0, + 936.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 927.0, + 1409.0, + 927.0, + 1409.0, + 965.0, + 352.0, + 965.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 955.0, + 1242.0, + 955.0, + 1242.0, + 1000.0, + 350.0, + 1000.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 1008.0, + 1407.0, + 1008.0, + 1407.0, + 1052.0, + 306.0, + 1052.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1038.0, + 1211.0, + 1038.0, + 1211.0, + 1081.0, + 352.0, + 1081.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 1083.0, + 1409.0, + 1083.0, + 1409.0, + 1142.0, + 304.0, + 1142.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1117.0, + 988.0, + 1117.0, + 988.0, + 1162.0, + 350.0, + 1162.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 309.0, + 1174.0, + 1404.0, + 1174.0, + 1404.0, + 1213.0, + 309.0, + 1213.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1206.0, + 1368.0, + 1206.0, + 1368.0, + 1244.0, + 354.0, + 1244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1256.0, + 1407.0, + 1256.0, + 1407.0, + 1299.0, + 294.0, + 1299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1285.0, + 1046.0, + 1285.0, + 1046.0, + 1328.0, + 352.0, + 1328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1337.0, + 1407.0, + 1337.0, + 1407.0, + 1380.0, + 292.0, + 1380.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1368.0, + 533.0, + 1368.0, + 533.0, + 1406.0, + 352.0, + 1406.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1420.0, + 1409.0, + 1420.0, + 1409.0, + 1463.0, + 294.0, + 1463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1449.0, + 1282.0, + 1449.0, + 1282.0, + 1491.0, + 350.0, + 1491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1503.0, + 1406.0, + 1503.0, + 1406.0, + 1546.0, + 294.0, + 1546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1532.0, + 1045.0, + 1532.0, + 1045.0, + 1574.0, + 352.0, + 1574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1584.0, + 1411.0, + 1584.0, + 1411.0, + 1628.0, + 294.0, + 1628.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1614.0, + 1407.0, + 1614.0, + 1407.0, + 1657.0, + 352.0, + 1657.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1647.0, + 1074.0, + 1647.0, + 1074.0, + 1685.0, + 356.0, + 1685.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1697.0, + 1407.0, + 1697.0, + 1407.0, + 1740.0, + 294.0, + 1740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1730.0, + 1409.0, + 1730.0, + 1409.0, + 1768.0, + 356.0, + 1768.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1780.0, + 1407.0, + 1780.0, + 1407.0, + 1823.0, + 294.0, + 1823.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1809.0, + 1256.0, + 1809.0, + 1256.0, + 1852.0, + 354.0, + 1852.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1861.0, + 1411.0, + 1861.0, + 1411.0, + 1904.0, + 294.0, + 1904.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1894.0, + 1409.0, + 1894.0, + 1409.0, + 1932.0, + 356.0, + 1932.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1939.0, + 1409.0, + 1939.0, + 1409.0, + 1992.0, + 290.0, + 1992.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1977.0, + 1242.0, + 1977.0, + 1242.0, + 2015.0, + 354.0, + 2015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 257.0, + 1405.0, + 257.0, + 1405.0, + 291.0, + 294.0, + 291.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 285.0, + 1409.0, + 285.0, + 1409.0, + 325.0, + 292.0, + 325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 321.0, + 953.0, + 321.0, + 953.0, + 352.0, + 296.0, + 352.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 10, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.79 + }, + { + "category_id": 1, + "poly": [ + 296, + 931, + 1402, + 931, + 1402, + 995, + 296, + 995 + ], + "score": 0.566 + }, + { + "category_id": 1, + "poly": [ + 296, + 1132, + 1404, + 1132, + 1404, + 1197, + 296, + 1197 + ], + "score": 0.549 + }, + { + "category_id": 1, + "poly": [ + 297, + 845, + 1400, + 845, + 1400, + 910, + 297, + 910 + ], + "score": 0.535 + }, + { + "category_id": 1, + "poly": [ + 299, + 1017, + 1403, + 1017, + 1403, + 1111, + 299, + 1111 + ], + "score": 0.533 + }, + { + "category_id": 1, + "poly": [ + 294, + 1569, + 1405, + 1569, + 1405, + 1662, + 294, + 1662 + ], + "score": 0.529 + }, + { + "category_id": 1, + "poly": [ + 293, + 1335, + 1402, + 1335, + 1402, + 1399, + 293, + 1399 + ], + "score": 0.518 + }, + { + "category_id": 1, + "poly": [ + 299, + 727, + 1402, + 727, + 1402, + 823, + 299, + 823 + ], + "score": 0.509 + }, + { + "category_id": 1, + "poly": [ + 298, + 611, + 1403, + 611, + 1403, + 705, + 298, + 705 + ], + "score": 0.507 + }, + { + "category_id": 1, + "poly": [ + 299, + 1218, + 1401, + 1218, + 1401, + 1315, + 299, + 1315 + ], + "score": 0.507 + }, + { + "category_id": 1, + "poly": [ + 292, + 1684, + 1405, + 1684, + 1405, + 1752, + 292, + 1752 + ], + "score": 0.485 + }, + { + "category_id": 1, + "poly": [ + 296, + 525, + 1404, + 525, + 1404, + 589, + 296, + 589 + ], + "score": 0.458 + }, + { + "category_id": 1, + "poly": [ + 294, + 1858, + 1401, + 1858, + 1401, + 1922, + 294, + 1922 + ], + "score": 0.445 + }, + { + "category_id": 1, + "poly": [ + 299, + 1945, + 1397, + 1945, + 1397, + 2008, + 299, + 2008 + ], + "score": 0.44 + }, + { + "category_id": 1, + "poly": [ + 295, + 1771, + 1403, + 1771, + 1403, + 1837, + 295, + 1837 + ], + "score": 0.431 + }, + { + "category_id": 1, + "poly": [ + 294, + 439, + 1402, + 439, + 1402, + 505, + 294, + 505 + ], + "score": 0.406 + }, + { + "category_id": 1, + "poly": [ + 296, + 1421, + 1407, + 1421, + 1407, + 1548, + 296, + 1548 + ], + "score": 0.358 + }, + { + "category_id": 1, + "poly": [ + 299, + 202, + 1409, + 202, + 1409, + 417, + 299, + 417 + ], + "score": 0.272 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.158 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 928.0, + 1406.0, + 928.0, + 1406.0, + 969.0, + 295.0, + 969.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 962.0, + 882.0, + 962.0, + 882.0, + 996.0, + 354.0, + 996.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1131.0, + 1409.0, + 1131.0, + 1409.0, + 1171.0, + 295.0, + 1171.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1164.0, + 562.0, + 1164.0, + 562.0, + 1197.0, + 354.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 844.0, + 1405.0, + 844.0, + 1405.0, + 884.0, + 296.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 874.0, + 960.0, + 874.0, + 960.0, + 910.0, + 355.0, + 910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1017.0, + 1403.0, + 1017.0, + 1403.0, + 1054.0, + 294.0, + 1054.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1045.0, + 1405.0, + 1045.0, + 1405.0, + 1086.0, + 352.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1079.0, + 565.0, + 1079.0, + 565.0, + 1110.0, + 356.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1568.0, + 1408.0, + 1568.0, + 1408.0, + 1603.0, + 293.0, + 1603.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1597.0, + 1410.0, + 1597.0, + 1410.0, + 1636.0, + 351.0, + 1636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1626.0, + 431.0, + 1626.0, + 431.0, + 1664.0, + 351.0, + 1664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1336.0, + 1404.0, + 1336.0, + 1404.0, + 1372.0, + 297.0, + 1372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1367.0, + 670.0, + 1367.0, + 670.0, + 1399.0, + 355.0, + 1399.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 726.0, + 1407.0, + 726.0, + 1407.0, + 768.0, + 294.0, + 768.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 757.0, + 1407.0, + 757.0, + 1407.0, + 796.0, + 353.0, + 796.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 790.0, + 690.0, + 790.0, + 690.0, + 825.0, + 356.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 612.0, + 1407.0, + 612.0, + 1407.0, + 648.0, + 294.0, + 648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 643.0, + 1408.0, + 643.0, + 1408.0, + 677.0, + 353.0, + 677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 671.0, + 426.0, + 671.0, + 426.0, + 705.0, + 351.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1216.0, + 1408.0, + 1216.0, + 1408.0, + 1259.0, + 294.0, + 1259.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1252.0, + 1406.0, + 1252.0, + 1406.0, + 1286.0, + 355.0, + 1286.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1281.0, + 1184.0, + 1281.0, + 1184.0, + 1316.0, + 355.0, + 1316.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1685.0, + 1408.0, + 1685.0, + 1408.0, + 1722.0, + 296.0, + 1722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1717.0, + 1306.0, + 1717.0, + 1306.0, + 1753.0, + 354.0, + 1753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 522.0, + 1409.0, + 522.0, + 1409.0, + 562.0, + 295.0, + 562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 553.0, + 426.0, + 553.0, + 426.0, + 589.0, + 350.0, + 589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1858.0, + 1404.0, + 1858.0, + 1404.0, + 1894.0, + 297.0, + 1894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1887.0, + 1251.0, + 1887.0, + 1251.0, + 1923.0, + 356.0, + 1923.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1945.0, + 1402.0, + 1945.0, + 1402.0, + 1980.0, + 296.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1974.0, + 1051.0, + 1974.0, + 1051.0, + 2008.0, + 354.0, + 2008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1770.0, + 1404.0, + 1770.0, + 1404.0, + 1807.0, + 297.0, + 1807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1801.0, + 987.0, + 1801.0, + 987.0, + 1838.0, + 354.0, + 1838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 440.0, + 1407.0, + 440.0, + 1407.0, + 476.0, + 297.0, + 476.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 471.0, + 1409.0, + 471.0, + 1409.0, + 507.0, + 356.0, + 507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1422.0, + 1407.0, + 1422.0, + 1407.0, + 1457.0, + 294.0, + 1457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1449.0, + 1409.0, + 1449.0, + 1409.0, + 1490.0, + 352.0, + 1490.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1486.0, + 1405.0, + 1486.0, + 1405.0, + 1519.0, + 354.0, + 1519.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1515.0, + 1058.0, + 1515.0, + 1058.0, + 1549.0, + 352.0, + 1549.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 201.0, + 1410.0, + 201.0, + 1410.0, + 238.0, + 293.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 232.0, + 1407.0, + 232.0, + 1407.0, + 269.0, + 352.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 262.0, + 1407.0, + 262.0, + 1407.0, + 300.0, + 354.0, + 300.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 291.0, + 1409.0, + 291.0, + 1409.0, + 332.0, + 352.0, + 332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 322.0, + 1407.0, + 322.0, + 1407.0, + 360.0, + 354.0, + 360.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 355.0, + 1407.0, + 355.0, + 1407.0, + 389.0, + 354.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 385.0, + 1130.0, + 385.0, + 1130.0, + 419.0, + 354.0, + 419.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 11, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.749 + }, + { + "category_id": 1, + "poly": [ + 294, + 108, + 1412, + 108, + 1412, + 2013, + 294, + 2013 + ], + "score": 0.72 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.318 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 196.0, + 1409.0, + 196.0, + 1409.0, + 246.0, + 289.0, + 246.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 233.0, + 1407.0, + 233.0, + 1407.0, + 273.0, + 354.0, + 273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 263.0, + 1081.0, + 263.0, + 1081.0, + 302.0, + 352.0, + 302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 315.0, + 1407.0, + 315.0, + 1407.0, + 355.0, + 293.0, + 355.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 350.0, + 1405.0, + 350.0, + 1405.0, + 384.0, + 356.0, + 384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 380.0, + 504.0, + 380.0, + 504.0, + 413.0, + 354.0, + 413.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 430.0, + 1407.0, + 430.0, + 1407.0, + 469.0, + 293.0, + 469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 457.0, + 1145.0, + 457.0, + 1145.0, + 501.0, + 352.0, + 501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 513.0, + 1411.0, + 513.0, + 1411.0, + 553.0, + 295.0, + 553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 545.0, + 1409.0, + 545.0, + 1409.0, + 584.0, + 352.0, + 584.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 574.0, + 692.0, + 574.0, + 692.0, + 614.0, + 352.0, + 614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 626.0, + 1409.0, + 626.0, + 1409.0, + 666.0, + 295.0, + 666.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 655.0, + 1411.0, + 655.0, + 1411.0, + 697.0, + 350.0, + 697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 689.0, + 842.0, + 689.0, + 842.0, + 722.0, + 354.0, + 722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 737.0, + 1411.0, + 737.0, + 1411.0, + 781.0, + 288.0, + 781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 768.0, + 1409.0, + 768.0, + 1409.0, + 808.0, + 352.0, + 808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 799.0, + 431.0, + 799.0, + 431.0, + 842.0, + 354.0, + 842.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 856.0, + 1407.0, + 856.0, + 1407.0, + 889.0, + 295.0, + 889.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 881.0, + 1411.0, + 881.0, + 1411.0, + 923.0, + 352.0, + 923.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 912.0, + 432.0, + 912.0, + 432.0, + 954.0, + 352.0, + 954.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 967.0, + 1407.0, + 967.0, + 1407.0, + 1006.0, + 293.0, + 1006.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 998.0, + 1179.0, + 998.0, + 1179.0, + 1038.0, + 352.0, + 1038.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1050.0, + 1409.0, + 1050.0, + 1409.0, + 1090.0, + 293.0, + 1090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1083.0, + 944.0, + 1083.0, + 944.0, + 1117.0, + 354.0, + 1117.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1134.0, + 1407.0, + 1134.0, + 1407.0, + 1173.0, + 293.0, + 1173.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1165.0, + 845.0, + 1165.0, + 845.0, + 1205.0, + 352.0, + 1205.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1217.0, + 1409.0, + 1217.0, + 1409.0, + 1257.0, + 293.0, + 1257.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1251.0, + 1117.0, + 1251.0, + 1117.0, + 1284.0, + 356.0, + 1284.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1297.0, + 1411.0, + 1297.0, + 1411.0, + 1342.0, + 291.0, + 1342.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1328.0, + 1107.0, + 1328.0, + 1107.0, + 1370.0, + 352.0, + 1370.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1386.0, + 1405.0, + 1386.0, + 1405.0, + 1420.0, + 295.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1418.0, + 1212.0, + 1418.0, + 1212.0, + 1451.0, + 354.0, + 1451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1468.0, + 1407.0, + 1468.0, + 1407.0, + 1507.0, + 293.0, + 1507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1501.0, + 1405.0, + 1501.0, + 1405.0, + 1535.0, + 354.0, + 1535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1530.0, + 730.0, + 1530.0, + 730.0, + 1564.0, + 354.0, + 1564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1583.0, + 1407.0, + 1583.0, + 1407.0, + 1616.0, + 295.0, + 1616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1614.0, + 771.0, + 1614.0, + 771.0, + 1647.0, + 354.0, + 1647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1666.0, + 1403.0, + 1666.0, + 1403.0, + 1700.0, + 295.0, + 1700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1698.0, + 980.0, + 1698.0, + 980.0, + 1731.0, + 356.0, + 1731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 1743.0, + 1409.0, + 1743.0, + 1409.0, + 1792.0, + 288.0, + 1792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1779.0, + 819.0, + 1779.0, + 819.0, + 1814.0, + 350.0, + 1814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1829.0, + 1409.0, + 1829.0, + 1409.0, + 1869.0, + 293.0, + 1869.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1861.0, + 637.0, + 1861.0, + 637.0, + 1896.0, + 349.0, + 1896.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1913.0, + 1411.0, + 1913.0, + 1411.0, + 1952.0, + 293.0, + 1952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1942.0, + 1409.0, + 1942.0, + 1409.0, + 1984.0, + 352.0, + 1984.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1973.0, + 574.0, + 1973.0, + 574.0, + 2013.0, + 352.0, + 2013.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 12, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.828 + }, + { + "category_id": 1, + "poly": [ + 294, + 86, + 1410, + 86, + 1410, + 2011, + 294, + 2011 + ], + "score": 0.65 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2095.0, + 832.0, + 2095.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 201.0, + 1407.0, + 201.0, + 1407.0, + 238.0, + 290.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 232.0, + 1407.0, + 232.0, + 1407.0, + 272.0, + 354.0, + 272.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 255.0, + 1409.0, + 255.0, + 1409.0, + 308.0, + 348.0, + 308.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 295.0, + 1183.0, + 295.0, + 1183.0, + 329.0, + 356.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 348.0, + 1409.0, + 348.0, + 1409.0, + 388.0, + 293.0, + 388.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 380.0, + 531.0, + 380.0, + 531.0, + 414.0, + 356.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 433.0, + 1409.0, + 433.0, + 1409.0, + 473.0, + 293.0, + 473.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 466.0, + 1173.0, + 466.0, + 1173.0, + 500.0, + 354.0, + 500.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 517.0, + 1409.0, + 517.0, + 1409.0, + 557.0, + 293.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 546.0, + 504.0, + 546.0, + 504.0, + 587.0, + 350.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 601.0, + 1409.0, + 601.0, + 1409.0, + 641.0, + 293.0, + 641.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 631.0, + 979.0, + 631.0, + 979.0, + 671.0, + 352.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 681.0, + 1409.0, + 681.0, + 1409.0, + 730.0, + 288.0, + 730.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 717.0, + 993.0, + 717.0, + 993.0, + 757.0, + 352.0, + 757.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 770.0, + 1407.0, + 770.0, + 1407.0, + 810.0, + 293.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 802.0, + 1162.0, + 802.0, + 1162.0, + 840.0, + 350.0, + 840.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 854.0, + 1409.0, + 854.0, + 1409.0, + 897.0, + 290.0, + 897.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 886.0, + 1354.0, + 886.0, + 1354.0, + 924.0, + 350.0, + 924.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 943.0, + 1405.0, + 943.0, + 1405.0, + 977.0, + 295.0, + 977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 968.0, + 1409.0, + 968.0, + 1409.0, + 1011.0, + 350.0, + 1011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 998.0, + 432.0, + 998.0, + 432.0, + 1040.0, + 352.0, + 1040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1055.0, + 1409.0, + 1055.0, + 1409.0, + 1095.0, + 293.0, + 1095.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1084.0, + 432.0, + 1084.0, + 432.0, + 1127.0, + 352.0, + 1127.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1139.0, + 1407.0, + 1139.0, + 1407.0, + 1179.0, + 295.0, + 1179.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1171.0, + 1405.0, + 1171.0, + 1405.0, + 1205.0, + 356.0, + 1205.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1197.0, + 431.0, + 1197.0, + 431.0, + 1239.0, + 354.0, + 1239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1253.0, + 1407.0, + 1253.0, + 1407.0, + 1295.0, + 290.0, + 1295.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1278.0, + 1409.0, + 1278.0, + 1409.0, + 1329.0, + 350.0, + 1329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1314.0, + 588.0, + 1314.0, + 588.0, + 1352.0, + 352.0, + 1352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1369.0, + 1407.0, + 1369.0, + 1407.0, + 1409.0, + 293.0, + 1409.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1401.0, + 996.0, + 1401.0, + 996.0, + 1441.0, + 352.0, + 1441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1453.0, + 1407.0, + 1453.0, + 1407.0, + 1494.0, + 295.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1485.0, + 1224.0, + 1485.0, + 1224.0, + 1525.0, + 354.0, + 1525.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1538.0, + 1407.0, + 1538.0, + 1407.0, + 1578.0, + 295.0, + 1578.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1565.0, + 1409.0, + 1565.0, + 1409.0, + 1614.0, + 350.0, + 1614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1599.0, + 1000.0, + 1599.0, + 1000.0, + 1639.0, + 352.0, + 1639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1654.0, + 1411.0, + 1654.0, + 1411.0, + 1694.0, + 293.0, + 1694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1683.0, + 1409.0, + 1683.0, + 1409.0, + 1723.0, + 352.0, + 1723.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1715.0, + 428.0, + 1715.0, + 428.0, + 1753.0, + 354.0, + 1753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1768.0, + 1409.0, + 1768.0, + 1409.0, + 1808.0, + 293.0, + 1808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1797.0, + 1407.0, + 1797.0, + 1407.0, + 1840.0, + 352.0, + 1840.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1825.0, + 1409.0, + 1825.0, + 1409.0, + 1873.0, + 350.0, + 1873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1863.0, + 1407.0, + 1863.0, + 1407.0, + 1896.0, + 356.0, + 1896.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1890.0, + 1342.0, + 1890.0, + 1342.0, + 1930.0, + 352.0, + 1930.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1945.0, + 1405.0, + 1945.0, + 1405.0, + 1985.0, + 293.0, + 1985.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1975.0, + 841.0, + 1975.0, + 841.0, + 2015.0, + 354.0, + 2015.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 13, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 293, + 202, + 1409, + 202, + 1409, + 1643, + 293, + 1643 + ], + "score": 0.821 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.707 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.385 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 202.0, + 1405.0, + 202.0, + 1405.0, + 242.0, + 294.0, + 242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 235.0, + 864.0, + 235.0, + 864.0, + 271.0, + 354.0, + 271.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 285.0, + 1407.0, + 285.0, + 1407.0, + 325.0, + 294.0, + 325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 316.0, + 693.0, + 316.0, + 693.0, + 356.0, + 354.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 369.0, + 1408.0, + 369.0, + 1408.0, + 409.0, + 294.0, + 409.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 398.0, + 1407.0, + 398.0, + 1407.0, + 438.0, + 355.0, + 438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 428.0, + 1407.0, + 428.0, + 1407.0, + 468.0, + 354.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 459.0, + 997.0, + 459.0, + 997.0, + 499.0, + 352.0, + 499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 512.0, + 1408.0, + 512.0, + 1408.0, + 552.0, + 294.0, + 552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 542.0, + 1410.0, + 542.0, + 1410.0, + 582.0, + 354.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 573.0, + 530.0, + 573.0, + 530.0, + 609.0, + 354.0, + 609.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 622.0, + 1407.0, + 622.0, + 1407.0, + 669.0, + 291.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 654.0, + 1269.0, + 654.0, + 1269.0, + 693.0, + 352.0, + 693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 707.0, + 1405.0, + 707.0, + 1405.0, + 748.0, + 294.0, + 748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 736.0, + 887.0, + 736.0, + 887.0, + 780.0, + 352.0, + 780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 789.0, + 1405.0, + 789.0, + 1405.0, + 829.0, + 294.0, + 829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 820.0, + 1134.0, + 820.0, + 1134.0, + 860.0, + 349.0, + 860.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 873.0, + 1405.0, + 873.0, + 1405.0, + 913.0, + 294.0, + 913.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 905.0, + 993.0, + 905.0, + 993.0, + 940.0, + 355.0, + 940.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 953.0, + 1407.0, + 953.0, + 1407.0, + 998.0, + 292.0, + 998.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 985.0, + 1125.0, + 985.0, + 1125.0, + 1027.0, + 352.0, + 1027.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1036.0, + 1408.0, + 1036.0, + 1408.0, + 1077.0, + 294.0, + 1077.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1069.0, + 797.0, + 1069.0, + 797.0, + 1104.0, + 354.0, + 1104.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1120.0, + 1410.0, + 1120.0, + 1410.0, + 1160.0, + 294.0, + 1160.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1152.0, + 537.0, + 1152.0, + 537.0, + 1187.0, + 354.0, + 1187.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1200.0, + 1408.0, + 1200.0, + 1408.0, + 1247.0, + 291.0, + 1247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1234.0, + 822.0, + 1234.0, + 822.0, + 1274.0, + 352.0, + 1274.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1285.0, + 1404.0, + 1285.0, + 1404.0, + 1325.0, + 294.0, + 1325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1314.0, + 1410.0, + 1314.0, + 1410.0, + 1358.0, + 351.0, + 1358.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1342.0, + 431.0, + 1342.0, + 431.0, + 1385.0, + 352.0, + 1385.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1396.0, + 1407.0, + 1396.0, + 1407.0, + 1443.0, + 291.0, + 1443.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1431.0, + 536.0, + 1431.0, + 536.0, + 1467.0, + 355.0, + 1467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1481.0, + 1407.0, + 1481.0, + 1407.0, + 1521.0, + 295.0, + 1521.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1513.0, + 1317.0, + 1513.0, + 1317.0, + 1553.0, + 354.0, + 1553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1565.0, + 1408.0, + 1565.0, + 1408.0, + 1605.0, + 295.0, + 1605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1597.0, + 1107.0, + 1597.0, + 1107.0, + 1632.0, + 354.0, + 1632.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 14, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 266, + 1405, + 266, + 1405, + 511, + 297, + 511 + ], + "score": 0.979 + }, + { + "category_id": 3, + "poly": [ + 302, + 670, + 1400, + 670, + 1400, + 1694, + 302, + 1694 + ], + "score": 0.971 + }, + { + "category_id": 4, + "poly": [ + 298, + 1726, + 1405, + 1726, + 1405, + 1839, + 298, + 1839 + ], + "score": 0.96 + }, + { + "category_id": 1, + "poly": [ + 297, + 525, + 1405, + 525, + 1405, + 647, + 297, + 647 + ], + "score": 0.927 + }, + { + "category_id": 0, + "poly": [ + 300, + 198, + 680, + 198, + 680, + 236, + 300, + 236 + ], + "score": 0.912 + }, + { + "category_id": 2, + "poly": [ + 836, + 2062, + 865, + 2062, + 865, + 2085, + 836, + 2085 + ], + "score": 0.847 + }, + { + "category_id": 15, + "poly": [ + 342.0, + 670.0, + 478.0, + 670.0, + 478.0, + 709.0, + 342.0, + 709.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 670.0, + 731.0, + 670.0, + 731.0, + 707.0, + 532.0, + 707.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 913.0, + 673.0, + 1235.0, + 673.0, + 1235.0, + 704.0, + 913.0, + 704.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1728.0, + 1405.0, + 1728.0, + 1405.0, + 1757.0, + 296.0, + 1757.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1756.0, + 1403.0, + 1756.0, + 1403.0, + 1785.0, + 296.0, + 1785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1781.0, + 1403.0, + 1781.0, + 1403.0, + 1816.0, + 295.0, + 1816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1808.0, + 1136.0, + 1808.0, + 1136.0, + 1844.0, + 293.0, + 1844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 193.0, + 683.0, + 193.0, + 683.0, + 244.0, + 292.0, + 244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 871.0, + 2058.0, + 871.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 266.0, + 1408.0, + 266.0, + 1408.0, + 303.0, + 294.0, + 303.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 300.0, + 1403.0, + 300.0, + 1403.0, + 330.0, + 296.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 329.0, + 1403.0, + 329.0, + 1403.0, + 362.0, + 295.0, + 362.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 359.0, + 1405.0, + 359.0, + 1405.0, + 392.0, + 295.0, + 392.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 387.0, + 1407.0, + 387.0, + 1407.0, + 424.0, + 294.0, + 424.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 415.0, + 1405.0, + 415.0, + 1405.0, + 457.0, + 292.0, + 457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 448.0, + 1403.0, + 448.0, + 1403.0, + 484.0, + 292.0, + 484.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 482.0, + 729.0, + 482.0, + 729.0, + 512.0, + 295.0, + 512.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 522.0, + 1405.0, + 522.0, + 1405.0, + 562.0, + 294.0, + 562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 557.0, + 1405.0, + 557.0, + 1405.0, + 589.0, + 295.0, + 589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 586.0, + 1403.0, + 586.0, + 1403.0, + 618.0, + 294.0, + 618.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 616.0, + 828.0, + 616.0, + 828.0, + 651.0, + 292.0, + 651.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 15, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 3, + "poly": [ + 299, + 346, + 1406, + 346, + 1406, + 886, + 299, + 886 + ], + "score": 0.969 + }, + { + "category_id": 3, + "poly": [ + 299, + 1271, + 1405, + 1271, + 1405, + 1759, + 299, + 1759 + ], + "score": 0.967 + }, + { + "category_id": 4, + "poly": [ + 298, + 908, + 1405, + 908, + 1405, + 995, + 298, + 995 + ], + "score": 0.957 + }, + { + "category_id": 4, + "poly": [ + 298, + 1783, + 1404, + 1783, + 1404, + 1868, + 298, + 1868 + ], + "score": 0.956 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 864, + 2061, + 864, + 2085, + 836, + 2085 + ], + "score": 0.856 + }, + { + "category_id": 15, + "poly": [ + 343.0, + 342.0, + 479.0, + 342.0, + 479.0, + 383.0, + 343.0, + 383.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 341.0, + 732.0, + 341.0, + 732.0, + 381.0, + 532.0, + 381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 913.0, + 344.0, + 1235.0, + 344.0, + 1235.0, + 380.0, + 913.0, + 380.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 389.0, + 720.0, + 400.0, + 720.0, + 400.0, + 732.0, + 389.0, + 732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 601.0, + 830.0, + 612.0, + 830.0, + 612.0, + 841.0, + 601.0, + 841.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 316.0, + 1270.0, + 733.0, + 1270.0, + 733.0, + 1312.0, + 316.0, + 1312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 914.0, + 1274.0, + 1234.0, + 1274.0, + 1234.0, + 1309.0, + 914.0, + 1309.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 907.0, + 1406.0, + 907.0, + 1406.0, + 942.0, + 292.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 937.0, + 1403.0, + 937.0, + 1403.0, + 970.0, + 295.0, + 970.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 964.0, + 395.0, + 964.0, + 395.0, + 997.0, + 293.0, + 997.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1782.0, + 1406.0, + 1782.0, + 1406.0, + 1817.0, + 293.0, + 1817.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1812.0, + 1405.0, + 1812.0, + 1405.0, + 1844.0, + 294.0, + 1844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1839.0, + 568.0, + 1839.0, + 568.0, + 1872.0, + 294.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 869.0, + 2058.0, + 869.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 16, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1592, + 1405, + 1592, + 1405, + 1866, + 297, + 1866 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 300, + 1409, + 1404, + 1409, + 1404, + 1501, + 300, + 1501 + ], + "score": 0.957 + }, + { + "category_id": 1, + "poly": [ + 296, + 1947, + 1403, + 1947, + 1403, + 2007, + 296, + 2007 + ], + "score": 0.936 + }, + { + "category_id": 0, + "poly": [ + 298, + 1360, + 1116, + 1360, + 1116, + 1399, + 298, + 1399 + ], + "score": 0.934 + }, + { + "category_id": 0, + "poly": [ + 300, + 1908, + 994, + 1908, + 994, + 1944, + 300, + 1944 + ], + "score": 0.932 + }, + { + "category_id": 3, + "poly": [ + 300, + 214, + 1410, + 214, + 1410, + 1180, + 300, + 1180 + ], + "score": 0.926 + }, + { + "category_id": 0, + "poly": [ + 298, + 1544, + 1036, + 1544, + 1036, + 1581, + 298, + 1581 + ], + "score": 0.926 + }, + { + "category_id": 4, + "poly": [ + 296, + 1195, + 1401, + 1195, + 1401, + 1254, + 296, + 1254 + ], + "score": 0.925 + }, + { + "category_id": 2, + "poly": [ + 836, + 2062, + 864, + 2062, + 864, + 2085, + 836, + 2085 + ], + "score": 0.844 + }, + { + "category_id": 3, + "poly": [ + 300, + 216, + 1410, + 216, + 1410, + 1181, + 300, + 1181 + ], + "score": 0.761 + }, + { + "category_id": 2, + "poly": [ + 788, + 197, + 1208, + 197, + 1208, + 221, + 788, + 221 + ], + "score": 0.673 + }, + { + "category_id": 2, + "poly": [ + 372, + 197, + 504, + 197, + 504, + 220, + 372, + 220 + ], + "score": 0.194 + }, + { + "category_id": 6, + "poly": [ + 788, + 197, + 1208, + 197, + 1208, + 221, + 788, + 221 + ], + "score": 0.098 + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1357.0, + 1120.0, + 1357.0, + 1120.0, + 1408.0, + 290.0, + 1408.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1905.0, + 980.0, + 1905.0, + 980.0, + 1953.0, + 291.0, + 1953.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 379.0, + 213.0, + 394.0, + 213.0, + 394.0, + 222.0, + 379.0, + 222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 213.0, + 863.0, + 213.0, + 863.0, + 222.0, + 832.0, + 222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 931.0, + 730.0, + 995.0, + 730.0, + 995.0, + 794.0, + 931.0, + 794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 661.0, + 851.0, + 711.0, + 851.0, + 711.0, + 911.0, + 661.0, + 911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 625.0, + 251.0, + 647.0, + 251.0, + 647.0, + 269.0, + 625.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1216.75, + 726.0, + 1253.75, + 726.0, + 1253.75, + 796.5, + 1216.75, + 796.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 933.25, + 857.5, + 991.25, + 857.5, + 991.25, + 888.0, + 933.25, + 888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1537.0, + 1040.0, + 1537.0, + 1040.0, + 1593.0, + 291.0, + 1593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1195.0, + 1405.0, + 1195.0, + 1405.0, + 1229.0, + 295.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1225.0, + 933.0, + 1225.0, + 933.0, + 1256.0, + 294.0, + 1256.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 924.0, + 723.0, + 1001.0, + 723.0, + 1001.0, + 805.0, + 924.0, + 805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1205.0, + 722.0, + 1270.0, + 722.0, + 1270.0, + 802.0, + 1205.0, + 802.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 658.0, + 844.0, + 719.0, + 844.0, + 719.0, + 912.0, + 658.0, + 912.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 944.0, + 857.0, + 979.0, + 857.0, + 979.0, + 889.0, + 944.0, + 889.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 631.75, + 253.0, + 643.75, + 253.0, + 643.75, + 267.0, + 631.75, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 904.75, + 368.0, + 917.75, + 368.0, + 917.75, + 380.0, + 904.75, + 380.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 785.0, + 191.0, + 1209.0, + 191.0, + 1209.0, + 226.0, + 785.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 193.0, + 509.0, + 193.0, + 509.0, + 227.0, + 368.0, + 227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 785.0, + 191.0, + 1209.0, + 191.0, + 1209.0, + 226.0, + 785.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1590.0, + 1405.0, + 1590.0, + 1405.0, + 1629.0, + 294.0, + 1629.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1623.0, + 1407.0, + 1623.0, + 1407.0, + 1659.0, + 295.0, + 1659.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1647.0, + 1406.0, + 1647.0, + 1406.0, + 1693.0, + 291.0, + 1693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1684.0, + 1405.0, + 1684.0, + 1405.0, + 1720.0, + 295.0, + 1720.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1713.0, + 1406.0, + 1713.0, + 1406.0, + 1750.0, + 294.0, + 1750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1743.0, + 1406.0, + 1743.0, + 1406.0, + 1780.0, + 294.0, + 1780.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1776.0, + 1403.0, + 1776.0, + 1403.0, + 1808.0, + 295.0, + 1808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1799.0, + 1410.0, + 1799.0, + 1410.0, + 1845.0, + 292.0, + 1845.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1833.0, + 1373.0, + 1833.0, + 1373.0, + 1867.0, + 295.0, + 1867.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1407.0, + 1405.0, + 1407.0, + 1405.0, + 1444.0, + 294.0, + 1444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1439.0, + 1404.0, + 1439.0, + 1404.0, + 1473.0, + 295.0, + 1473.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1471.0, + 1115.0, + 1471.0, + 1115.0, + 1504.0, + 295.0, + 1504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1943.0, + 1408.0, + 1943.0, + 1408.0, + 1980.0, + 294.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1973.0, + 1405.0, + 1973.0, + 1405.0, + 2010.0, + 294.0, + 2010.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 17, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 3, + "poly": [ + 301, + 198, + 1406, + 198, + 1406, + 1849, + 301, + 1849 + ], + "score": 0.961 + }, + { + "category_id": 4, + "poly": [ + 298, + 1864, + 1401, + 1864, + 1401, + 1949, + 298, + 1949 + ], + "score": 0.956 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.848 + }, + { + "category_id": 15, + "poly": [ + 642.0, + 192.0, + 791.0, + 192.0, + 791.0, + 226.0, + 642.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 965.0, + 192.0, + 1020.0, + 192.0, + 1020.0, + 223.0, + 965.0, + 223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1243.0, + 194.0, + 1307.0, + 194.0, + 1307.0, + 225.0, + 1243.0, + 225.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 765.0, + 427.0, + 823.0, + 427.0, + 823.0, + 526.0, + 765.0, + 526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 782.0, + 668.0, + 798.0, + 668.0, + 798.0, + 682.0, + 782.0, + 682.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 619.0, + 992.0, + 637.0, + 992.0, + 637.0, + 1009.0, + 619.0, + 1009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 895.0, + 994.0, + 913.0, + 994.0, + 913.0, + 1009.0, + 895.0, + 1009.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1173.0, + 992.0, + 1191.0, + 992.0, + 1191.0, + 1007.0, + 1173.0, + 1007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 714.0, + 1098.0, + 730.0, + 1098.0, + 730.0, + 1113.0, + 714.0, + 1113.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 992.0, + 1098.0, + 1006.0, + 1098.0, + 1006.0, + 1113.0, + 992.0, + 1113.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1270.0, + 1098.0, + 1284.0, + 1098.0, + 1284.0, + 1111.0, + 1270.0, + 1111.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 819.0, + 1295.0, + 845.0, + 1295.0, + 845.0, + 1317.0, + 819.0, + 1317.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1377.0, + 1299.0, + 1395.0, + 1299.0, + 1395.0, + 1315.0, + 1377.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 1505.0, + 1022.0, + 1505.0, + 1022.0, + 1518.0, + 1006.0, + 1518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1282.0, + 1505.0, + 1302.0, + 1505.0, + 1302.0, + 1520.0, + 1282.0, + 1520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 990.0, + 1552.0, + 1002.0, + 1552.0, + 1002.0, + 1567.0, + 990.0, + 1567.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1266.0, + 1551.0, + 1280.0, + 1551.0, + 1280.0, + 1569.0, + 1266.0, + 1569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1235.0, + 1811.0, + 1252.0, + 1811.0, + 1252.0, + 1824.0, + 1235.0, + 1824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 194.5, + 473.0, + 194.5, + 473.0, + 227.5, + 404.0, + 227.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1863.0, + 1405.0, + 1863.0, + 1405.0, + 1895.0, + 293.0, + 1895.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1891.0, + 1405.0, + 1891.0, + 1405.0, + 1923.0, + 293.0, + 1923.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1920.0, + 910.0, + 1920.0, + 910.0, + 1953.0, + 294.0, + 1953.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 18, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 3, + "poly": [ + 356, + 88, + 1346, + 88, + 1346, + 1762, + 356, + 1762 + ], + "score": 0.93 + }, + { + "category_id": 4, + "poly": [ + 296, + 1795, + 1406, + 1795, + 1406, + 1936, + 296, + 1936 + ], + "score": 0.886 + }, + { + "category_id": 2, + "poly": [ + 835, + 2061, + 865, + 2061, + 865, + 2085, + 835, + 2085 + ], + "score": 0.853 + }, + { + "category_id": 4, + "poly": [ + 721, + 1739, + 977, + 1739, + 977, + 1766, + 721, + 1766 + ], + "score": 0.713 + }, + { + "category_id": 1, + "poly": [ + 296, + 1795, + 1406, + 1795, + 1406, + 1936, + 296, + 1936 + ], + "score": 0.123 + }, + { + "category_id": 15, + "poly": [ + 342.0, + 182.0, + 1133.0, + 182.0, + 1133.0, + 319.0, + 342.0, + 319.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 724.0, + 415.0, + 974.0, + 415.0, + 974.0, + 456.0, + 724.0, + 456.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1043.0, + 510.0, + 1070.0, + 510.0, + 1070.0, + 549.0, + 1043.0, + 549.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 751.0, + 683.0, + 948.0, + 683.0, + 948.0, + 724.0, + 751.0, + 724.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 711.0, + 943.0, + 987.0, + 943.0, + 987.0, + 988.0, + 711.0, + 988.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 747.0, + 1210.0, + 947.0, + 1210.0, + 947.0, + 1244.0, + 747.0, + 1244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1241.0, + 1048.0, + 1241.0, + 1048.0, + 1372.0, + 351.0, + 1372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 732.0, + 1472.0, + 966.0, + 1472.0, + 966.0, + 1507.0, + 732.0, + 1507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 527.0, + 1527.0, + 561.0, + 1527.0, + 561.0, + 1546.0, + 527.0, + 1546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 635.0, + 1526.0, + 679.0, + 1526.0, + 679.0, + 1546.0, + 635.0, + 1546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 772.0, + 1527.0, + 789.0, + 1527.0, + 789.0, + 1540.0, + 772.0, + 1540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1036.0, + 1527.0, + 1057.0, + 1527.0, + 1057.0, + 1540.0, + 1036.0, + 1540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 635.0, + 1592.0, + 662.0, + 1592.0, + 662.0, + 1605.0, + 635.0, + 1605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 719.0, + 1588.0, + 802.0, + 1588.0, + 802.0, + 1620.0, + 719.0, + 1620.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1214.0, + 1664.0, + 1313.0, + 1664.0, + 1313.0, + 1688.0, + 1214.0, + 1688.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 713.0, + 1729.0, + 984.0, + 1729.0, + 984.0, + 1773.0, + 713.0, + 1773.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1795.0, + 1406.0, + 1795.0, + 1406.0, + 1827.0, + 295.0, + 1827.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1823.0, + 1402.0, + 1823.0, + 1402.0, + 1850.0, + 296.0, + 1850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1850.0, + 1404.0, + 1850.0, + 1404.0, + 1882.0, + 294.0, + 1882.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1879.0, + 1404.0, + 1879.0, + 1404.0, + 1911.0, + 295.0, + 1911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1904.0, + 1080.0, + 1904.0, + 1080.0, + 1938.0, + 293.0, + 1938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 829.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2096.0, + 829.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 717.0, + 1734.0, + 982.0, + 1734.0, + 982.0, + 1770.0, + 717.0, + 1770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1795.0, + 1406.0, + 1795.0, + 1406.0, + 1827.0, + 295.0, + 1827.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1823.0, + 1402.0, + 1823.0, + 1402.0, + 1850.0, + 296.0, + 1850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1850.0, + 1404.0, + 1850.0, + 1404.0, + 1882.0, + 294.0, + 1882.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1879.0, + 1404.0, + 1879.0, + 1404.0, + 1911.0, + 295.0, + 1911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1904.0, + 1080.0, + 1904.0, + 1080.0, + 1938.0, + 293.0, + 1938.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 19, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 304, + 747, + 1394, + 747, + 1394, + 1065, + 304, + 1065 + ], + "score": 0.983, + "html": "
MethodBackbone81↑82个8↑REL↓Sq-rel ↓RMS↓RMS log↓
BTS [34]DenseNet-161†0.9560.9930.9980.0590.2452.7560.096
TransDepth [87]ResNet-50+ViT-Bt0.9560.9940.9990.0640.2522.7550.098
DPT [50]ResNet-50+ViT-Bt‡0.9590.9950.9990.06212.5730.092
AdaBins [3]E-B5+mini-ViTt0.9640.9950.9990.0580.1902.3600.088
BinsFormer [35]Swin-Larget0.9740.9970.9990.0520.1512.0980.079
PixelFormer [1]Swin-Larget0.9760.9970.9990.0510.1492.0810.077
MIM [80]SwinV2-LT0.9770.9981.0000.0500.1391.9660.075
samples=1 DDVM samples=2 samples=4Efficient U-NetT#0.9640.9940.9980.0560.3392.7000.091
Efficient U-NetT#0.9650.9940.9980.0550.3252.6600.090
Efficient U-NetT‡0.9650.9940.9980.0550.2922.6130.089
" + }, + { + "category_id": 5, + "poly": [ + 303, + 299, + 1392, + 299, + 1392, + 617, + 303, + 617 + ], + "score": 0.983, + "html": "
MethodArchitecture81↑&↑8↑REL↓RMS↓log10
TransDepth [87]Res-50+ViT-Bt0.9000.9830.9960.1060.3650.045
DPT[50]Res-50+ViT-B†#0.9040.9880.9980.1100.3570.045
AdaBins [3]E-B5+Mini-ViTt0.9030.9840.9970.1030.3640.044
BinsFormer [35]Swin-Larget0.9250.9890.9970.0940.3300.040
PixelFormer [1]Swin-Large†0.9290.9910.9980.0900.3220.039
MIM[80]SwinV2-LT0.9490.9940.9990.0830.2870.035
AiT-P [46]SwinV2-LT0.9530.9930.9990.0760.2790.033
DDVMsamples=1Efficient U-NetT#0.9440.9860.9950.0750.3240.032
samples=2Efficient U-NetT‡0.9440.9870.9960.0740.3190.032
samples=4Efficient U-Net T‡0.9460.9870.9960.0740.3150.032
" + }, + { + "category_id": 1, + "poly": [ + 297, + 1122, + 1405, + 1122, + 1405, + 1306, + 297, + 1306 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 298, + 1400, + 1405, + 1400, + 1405, + 1584, + 298, + 1584 + ], + "score": 0.977 + }, + { + "category_id": 5, + "poly": [ + 418, + 1857, + 1273, + 1857, + 1273, + 1975, + 418, + 1975 + ], + "score": 0.976, + "html": "
81↑82↑83↑REL↓Sq-rel↓RMS↓RMS log↓
L20.9540.9930.9980.0650.3212.7730.099
L10.9640.9940.9980.0560.3392.7000.091
" + }, + { + "category_id": 5, + "poly": [ + 527, + 1652, + 1163, + 1652, + 1163, + 1772, + 527, + 1772 + ], + "score": 0.968, + "html": "
81个8↑8↑REL↓RMS↓log10↓
L20.9320.9810.9940.0850.3490.037
L10.9440.9860.9950.0750.3240.032
" + }, + { + "category_id": 6, + "poly": [ + 297, + 195, + 1406, + 195, + 1406, + 283, + 297, + 283 + ], + "score": 0.95 + }, + { + "category_id": 6, + "poly": [ + 299, + 645, + 1403, + 645, + 1403, + 731, + 299, + 731 + ], + "score": 0.92 + }, + { + "category_id": 6, + "poly": [ + 450, + 1613, + 1247, + 1613, + 1247, + 1644, + 450, + 1644 + ], + "score": 0.914 + }, + { + "category_id": 6, + "poly": [ + 488, + 1817, + 1209, + 1817, + 1209, + 1847, + 488, + 1847 + ], + "score": 0.912 + }, + { + "category_id": 0, + "poly": [ + 298, + 1347, + 493, + 1347, + 493, + 1384, + 298, + 1384 + ], + "score": 0.907 + }, + { + "category_id": 2, + "poly": [ + 834, + 2061, + 863, + 2061, + 863, + 2086, + 834, + 2086 + ], + "score": 0.85 + }, + { + "category_id": 13, + "poly": [ + 503, + 1151, + 625, + 1151, + 625, + 1183, + 503, + 1183 + ], + "score": 0.91, + "latex": "\\cdot \\delta _ { i } < 1 . 2 5 ^ { i }" + }, + { + "category_id": 13, + "poly": [ + 669, + 1153, + 786, + 1153, + 786, + 1184, + 669, + 1184 + ], + "score": 0.88, + "latex": "i \\in { 1 , 2 , 3 }" + }, + { + "category_id": 13, + "poly": [ + 638, + 1403, + 672, + 1403, + 672, + 1432, + 638, + 1432 + ], + "score": 0.88, + "latex": "L _ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 298, + 1433, + 331, + 1433, + 331, + 1462, + 298, + 1462 + ], + "score": 0.87, + "latex": "L _ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 297, + 1184, + 375, + 1184, + 375, + 1216, + 297, + 1216 + ], + "score": 0.84, + "latex": "( l o g _ { 1 0 } )" + }, + { + "category_id": 13, + "poly": [ + 816, + 226, + 834, + 226, + 834, + 254, + 816, + 254 + ], + "score": 0.82, + "latex": "^ \\ddag" + }, + { + "category_id": 13, + "poly": [ + 719, + 675, + 736, + 675, + 736, + 703, + 719, + 703 + ], + "score": 0.81, + "latex": "^ \\ddag" + }, + { + "category_id": 13, + "poly": [ + 340, + 675, + 357, + 675, + 357, + 703, + 340, + 703 + ], + "score": 0.77, + "latex": "^ \\dagger" + }, + { + "category_id": 13, + "poly": [ + 422, + 226, + 439, + 226, + 439, + 254, + 422, + 254 + ], + "score": 0.75, + "latex": "^ \\dagger" + }, + { + "category_id": 13, + "poly": [ + 1075, + 254, + 1093, + 254, + 1093, + 282, + 1075, + 282 + ], + "score": 0.65, + "latex": "\\uparrow" + }, + { + "category_id": 13, + "poly": [ + 1019, + 197, + 1043, + 197, + 1043, + 223, + 1019, + 223 + ], + "score": 0.55, + "latex": "\\top" + }, + { + "category_id": 13, + "poly": [ + 937, + 702, + 955, + 702, + 955, + 730, + 937, + 730 + ], + "score": 0.5, + "latex": "\\uparrow" + }, + { + "category_id": 13, + "poly": [ + 857, + 254, + 879, + 254, + 879, + 282, + 857, + 282 + ], + "score": 0.48, + "latex": "\\downarrow :" + }, + { + "category_id": 13, + "poly": [ + 899, + 197, + 927, + 197, + 927, + 223, + 899, + 223 + ], + "score": 0.28, + "latex": "\\mathbf { v } 2" + }, + { + "category_id": 13, + "poly": [ + 725, + 703, + 749, + 703, + 749, + 729, + 725, + 729 + ], + "score": 0.27, + "latex": "\\downarrow \\colon" + }, + { + "category_id": 13, + "poly": [ + 938, + 646, + 961, + 646, + 961, + 671, + 938, + 671 + ], + "score": 0.26, + "latex": "\\top" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 197.0, + 898.0, + 197.0, + 898.0, + 226.0, + 295.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 928.0, + 197.0, + 1018.0, + 197.0, + 1018.0, + 226.0, + 928.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1044.0, + 197.0, + 1404.0, + 197.0, + 1404.0, + 226.0, + 1044.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 225.0, + 421.0, + 225.0, + 421.0, + 255.0, + 294.0, + 255.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 225.0, + 815.0, + 225.0, + 815.0, + 255.0, + 440.0, + 255.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 835.0, + 225.0, + 1408.0, + 225.0, + 1408.0, + 255.0, + 835.0, + 255.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 254.0, + 856.0, + 254.0, + 856.0, + 284.0, + 296.0, + 284.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 880.0, + 254.0, + 1074.0, + 254.0, + 1074.0, + 284.0, + 880.0, + 284.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1094.0, + 254.0, + 1263.0, + 254.0, + 1263.0, + 284.0, + 1094.0, + 284.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 644.0, + 937.0, + 644.0, + 937.0, + 678.0, + 293.0, + 678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 962.0, + 644.0, + 1408.0, + 644.0, + 1408.0, + 678.0, + 962.0, + 678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 674.0, + 339.0, + 674.0, + 339.0, + 705.0, + 293.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 358.0, + 674.0, + 718.0, + 674.0, + 718.0, + 705.0, + 358.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 737.0, + 674.0, + 1407.0, + 674.0, + 1407.0, + 705.0, + 737.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 700.0, + 724.0, + 700.0, + 724.0, + 732.0, + 293.0, + 732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 750.0, + 700.0, + 936.0, + 700.0, + 936.0, + 732.0, + 750.0, + 732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 956.0, + 700.0, + 1404.0, + 700.0, + 1404.0, + 732.0, + 956.0, + 732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 446.0, + 1608.0, + 1250.0, + 1608.0, + 1250.0, + 1647.0, + 446.0, + 1647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 488.0, + 1815.0, + 1210.0, + 1815.0, + 1210.0, + 1848.0, + 488.0, + 1848.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1343.0, + 498.0, + 1343.0, + 498.0, + 1390.0, + 291.0, + 1390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 830.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2097.0, + 830.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1121.0, + 1406.0, + 1121.0, + 1406.0, + 1157.0, + 295.0, + 1157.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1151.0, + 502.0, + 1151.0, + 502.0, + 1186.0, + 294.0, + 1186.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 626.0, + 1151.0, + 668.0, + 1151.0, + 668.0, + 1186.0, + 626.0, + 1186.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 1151.0, + 1403.0, + 1151.0, + 1403.0, + 1186.0, + 787.0, + 1186.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 376.0, + 1184.0, + 1406.0, + 1184.0, + 1406.0, + 1219.0, + 376.0, + 1219.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1215.0, + 1406.0, + 1215.0, + 1406.0, + 1247.0, + 295.0, + 1247.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1243.0, + 1403.0, + 1243.0, + 1403.0, + 1277.0, + 294.0, + 1277.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1274.0, + 1372.0, + 1274.0, + 1372.0, + 1309.0, + 295.0, + 1309.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1398.0, + 637.0, + 1398.0, + 637.0, + 1436.0, + 293.0, + 1436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 673.0, + 1398.0, + 1406.0, + 1398.0, + 1406.0, + 1436.0, + 673.0, + 1436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 332.0, + 1431.0, + 1406.0, + 1431.0, + 1406.0, + 1467.0, + 332.0, + 1467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1462.0, + 1405.0, + 1462.0, + 1405.0, + 1497.0, + 295.0, + 1497.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1493.0, + 1405.0, + 1493.0, + 1405.0, + 1526.0, + 293.0, + 1526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1521.0, + 1407.0, + 1521.0, + 1407.0, + 1558.0, + 292.0, + 1558.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1553.0, + 518.0, + 1553.0, + 518.0, + 1590.0, + 295.0, + 1590.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 20, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1445, + 1405, + 1445, + 1405, + 1749, + 297, + 1749 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 300, + 1885, + 1403, + 1885, + 1403, + 2007, + 300, + 2007 + ], + "score": 0.979 + }, + { + "category_id": 5, + "poly": [ + 345, + 423, + 1349, + 423, + 1349, + 542, + 345, + 542 + ], + "score": 0.974, + "html": "
81&28↑REL↓Sq-rel↓RMS↓RMS log↓
No self-supervised pre-training0.9520.990(0.9970.0640.389 2.9980.104
With self-supervised pre-training0.9650.9940.9980.0550.3322.6960.091
" + }, + { + "category_id": 5, + "poly": [ + 340, + 234, + 1352, + 234, + 1352, + 355, + 340, + 355 + ], + "score": 0.973, + "html": "
81↑8↑8↑REL↓RMS↓log10↓
No self-supervised pre-training0.9360.9800.9920.0810.3520.035
With self-supervised pre-training0.9440.9860.9950.0750.3240.032
" + }, + { + "category_id": 1, + "poly": [ + 296, + 649, + 1404, + 649, + 1404, + 802, + 296, + 802 + ], + "score": 0.954 + }, + { + "category_id": 4, + "poly": [ + 296, + 1288, + 1401, + 1288, + 1401, + 1347, + 296, + 1347 + ], + "score": 0.941 + }, + { + "category_id": 0, + "poly": [ + 298, + 1390, + 1041, + 1390, + 1041, + 1429, + 298, + 1429 + ], + "score": 0.931 + }, + { + "category_id": 0, + "poly": [ + 301, + 1793, + 784, + 1793, + 784, + 1832, + 301, + 1832 + ], + "score": 0.919 + }, + { + "category_id": 0, + "poly": [ + 296, + 595, + 856, + 595, + 856, + 633, + 296, + 633 + ], + "score": 0.916 + }, + { + "category_id": 6, + "poly": [ + 447, + 196, + 1249, + 196, + 1249, + 227, + 447, + 227 + ], + "score": 0.908 + }, + { + "category_id": 6, + "poly": [ + 455, + 384, + 1241, + 384, + 1241, + 415, + 455, + 415 + ], + "score": 0.901 + }, + { + "category_id": 0, + "poly": [ + 299, + 1848, + 522, + 1848, + 522, + 1879, + 299, + 1879 + ], + "score": 0.9 + }, + { + "category_id": 3, + "poly": [ + 299, + 834, + 1406, + 834, + 1406, + 1272, + 299, + 1272 + ], + "score": 0.899 + }, + { + "category_id": 2, + "poly": [ + 835, + 2061, + 864, + 2061, + 864, + 2085, + 835, + 2085 + ], + "score": 0.844 + }, + { + "category_id": 5, + "poly": [ + 299, + 834, + 1406, + 834, + 1406, + 1272, + 299, + 1272 + ], + "score": 0.111, + "html": "
ImageGround TruthNo refinementWith refinement
+↓+
+
" + }, + { + "category_id": 13, + "poly": [ + 727, + 1507, + 847, + 1507, + 847, + 1536, + 727, + 1536 + ], + "score": 0.9, + "latex": "3 2 0 \\times 4 4 8" + }, + { + "category_id": 13, + "poly": [ + 903, + 1538, + 969, + 1538, + 969, + 1566, + 903, + 1566 + ], + "score": 0.9, + "latex": "2 \\times 5" + }, + { + "category_id": 13, + "poly": [ + 1277, + 1537, + 1397, + 1537, + 1397, + 1566, + 1277, + 1566 + ], + "score": 0.9, + "latex": "3 2 0 \\times 4 4 8" + }, + { + "category_id": 13, + "poly": [ + 1374, + 1916, + 1397, + 1916, + 1397, + 1942, + 1374, + 1942 + ], + "score": 0.43, + "latex": "U _ { ☉ }" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1289.0, + 1403.0, + 1289.0, + 1403.0, + 1320.0, + 296.0, + 1320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1315.0, + 1277.0, + 1315.0, + 1277.0, + 1349.0, + 295.0, + 1349.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1387.0, + 1046.0, + 1387.0, + 1046.0, + 1436.0, + 293.0, + 1436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1791.0, + 788.0, + 1791.0, + 788.0, + 1836.0, + 292.0, + 1836.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 591.0, + 858.0, + 591.0, + 858.0, + 639.0, + 291.0, + 639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 443.0, + 191.0, + 1253.0, + 191.0, + 1253.0, + 232.0, + 443.0, + 232.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 455.0, + 379.0, + 1243.0, + 379.0, + 1243.0, + 420.0, + 455.0, + 420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1843.0, + 525.0, + 1843.0, + 525.0, + 1885.0, + 293.0, + 1885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 643.0, + 836.0, + 788.0, + 836.0, + 788.0, + 865.0, + 643.0, + 865.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 919.0, + 837.0, + 1071.0, + 837.0, + 1071.0, + 867.0, + 919.0, + 867.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1189.0, + 835.0, + 1361.0, + 835.0, + 1361.0, + 870.0, + 1189.0, + 870.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 606.0, + 961.0, + 629.0, + 961.0, + 629.0, + 974.0, + 606.0, + 974.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 883.0, + 961.0, + 905.0, + 961.0, + 905.0, + 975.0, + 883.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1162.0, + 961.0, + 1183.0, + 961.0, + 1183.0, + 975.0, + 1162.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 720.0, + 1147.0, + 743.0, + 1147.0, + 743.0, + 1160.0, + 720.0, + 1160.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1001.0, + 1150.0, + 1020.0, + 1150.0, + 1020.0, + 1159.0, + 1001.0, + 1159.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1276.0, + 1147.0, + 1299.0, + 1147.0, + 1299.0, + 1160.0, + 1276.0, + 1160.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 403.0, + 838.5, + 473.0, + 838.5, + 473.0, + 869.5, + 403.0, + 869.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 829.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2098.0, + 829.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1447.0, + 1405.0, + 1447.0, + 1405.0, + 1478.0, + 296.0, + 1478.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1476.0, + 1405.0, + 1476.0, + 1405.0, + 1511.0, + 294.0, + 1511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1506.0, + 726.0, + 1506.0, + 726.0, + 1541.0, + 295.0, + 1541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 848.0, + 1506.0, + 1406.0, + 1506.0, + 1406.0, + 1541.0, + 848.0, + 1541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1536.0, + 902.0, + 1536.0, + 902.0, + 1572.0, + 294.0, + 1572.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 970.0, + 1536.0, + 1276.0, + 1536.0, + 1276.0, + 1572.0, + 970.0, + 1572.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1398.0, + 1536.0, + 1407.0, + 1536.0, + 1407.0, + 1572.0, + 1398.0, + 1572.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1568.0, + 1407.0, + 1568.0, + 1407.0, + 1603.0, + 295.0, + 1603.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1597.0, + 1405.0, + 1597.0, + 1405.0, + 1634.0, + 291.0, + 1634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1627.0, + 1408.0, + 1627.0, + 1408.0, + 1661.0, + 294.0, + 1661.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1658.0, + 1407.0, + 1658.0, + 1407.0, + 1693.0, + 295.0, + 1693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1690.0, + 1406.0, + 1690.0, + 1406.0, + 1722.0, + 296.0, + 1722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1719.0, + 1176.0, + 1719.0, + 1176.0, + 1751.0, + 296.0, + 1751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1885.0, + 1405.0, + 1885.0, + 1405.0, + 1917.0, + 296.0, + 1917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1914.0, + 1373.0, + 1914.0, + 1373.0, + 1948.0, + 293.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1398.0, + 1914.0, + 1408.0, + 1914.0, + 1408.0, + 1948.0, + 1398.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1945.0, + 1405.0, + 1945.0, + 1405.0, + 1980.0, + 294.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1977.0, + 1405.0, + 1977.0, + 1405.0, + 2010.0, + 296.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 650.0, + 1405.0, + 650.0, + 1405.0, + 683.0, + 296.0, + 683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 678.0, + 1405.0, + 678.0, + 1405.0, + 714.0, + 294.0, + 714.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 710.0, + 1404.0, + 710.0, + 1404.0, + 743.0, + 295.0, + 743.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 736.0, + 1409.0, + 736.0, + 1409.0, + 778.0, + 291.0, + 778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 770.0, + 1122.0, + 770.0, + 1122.0, + 807.0, + 295.0, + 807.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 21, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1159, + 1406, + 1159, + 1406, + 1555, + 298, + 1555 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 299, + 1778, + 1404, + 1778, + 1404, + 1870, + 299, + 1870 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 300, + 1884, + 1404, + 1884, + 1404, + 2007, + 300, + 2007 + ], + "score": 0.972 + }, + { + "category_id": 1, + "poly": [ + 300, + 1570, + 1404, + 1570, + 1404, + 1662, + 300, + 1662 + ], + "score": 0.97 + }, + { + "category_id": 4, + "poly": [ + 297, + 927, + 1402, + 927, + 1402, + 1013, + 297, + 1013 + ], + "score": 0.952 + }, + { + "category_id": 1, + "poly": [ + 297, + 1085, + 1402, + 1085, + 1402, + 1147, + 297, + 1147 + ], + "score": 0.945 + }, + { + "category_id": 0, + "poly": [ + 300, + 1721, + 692, + 1721, + 692, + 1753, + 300, + 1753 + ], + "score": 0.916 + }, + { + "category_id": 6, + "poly": [ + 297, + 196, + 1403, + 196, + 1403, + 255, + 297, + 255 + ], + "score": 0.905 + }, + { + "category_id": 2, + "poly": [ + 835, + 2061, + 864, + 2061, + 864, + 2085, + 835, + 2085 + ], + "score": 0.848 + }, + { + "category_id": 3, + "poly": [ + 305, + 440, + 1401, + 440, + 1401, + 905, + 305, + 905 + ], + "score": 0.601 + }, + { + "category_id": 5, + "poly": [ + 406, + 270, + 1297, + 270, + 1297, + 389, + 406, + 389 + ], + "score": 0.521, + "html": "
Sintel CleanSintel FinalKITTI
RAFT baseline1.272.282.71
RAFT with the coarse-to-fine refinement1.352.262.85
" + }, + { + "category_id": 13, + "poly": [ + 799, + 1161, + 1022, + 1161, + 1022, + 1190, + 799, + 1190 + ], + "score": 0.91, + "latex": "6 4 \\times 6 4 2 5 6 \\times 2 5 6" + }, + { + "category_id": 13, + "poly": [ + 515, + 1601, + 625, + 1601, + 625, + 1630, + 515, + 1630 + ], + "score": 0.89, + "latex": "2 4 0 \\times 3 2 0" + }, + { + "category_id": 13, + "poly": [ + 1073, + 1571, + 1193, + 1571, + 1193, + 1600, + 1073, + 1600 + ], + "score": 0.89, + "latex": "2 5 6 \\times 2 5 6" + }, + { + "category_id": 13, + "poly": [ + 864, + 1601, + 975, + 1601, + 975, + 1630, + 864, + 1630 + ], + "score": 0.89, + "latex": "2 5 6 \\times 3 8 4" + }, + { + "category_id": 13, + "poly": [ + 706, + 1631, + 813, + 1631, + 813, + 1660, + 706, + 1660 + ], + "score": 0.89, + "latex": "3 2 0 \\times 4 4 8" + }, + { + "category_id": 13, + "poly": [ + 1017, + 1631, + 1125, + 1631, + 1125, + 1660, + 1017, + 1660 + ], + "score": 0.88, + "latex": "3 2 0 \\times 7 6 8" + }, + { + "category_id": 13, + "poly": [ + 1218, + 1601, + 1327, + 1601, + 1327, + 1630, + 1218, + 1630 + ], + "score": 0.87, + "latex": "2 5 6 \\times 8 3 2" + }, + { + "category_id": 13, + "poly": [ + 463, + 1915, + 584, + 1915, + 584, + 1946, + 463, + 1946 + ], + "score": 0.87, + "latex": "( 5 4 0 \\times 9 6 0 )" + }, + { + "category_id": 13, + "poly": [ + 695, + 1916, + 816, + 1916, + 816, + 1946, + 695, + 1946 + ], + "score": 0.87, + "latex": "( 5 1 2 \\times 5 1 2 )" + }, + { + "category_id": 13, + "poly": [ + 300, + 1976, + 431, + 1976, + 431, + 2006, + 300, + 2006 + ], + "score": 0.86, + "latex": "( 5 4 0 \\times 1 2 8 0 )" + }, + { + "category_id": 13, + "poly": [ + 526, + 1976, + 655, + 1976, + 655, + 2006, + 526, + 2006 + ], + "score": 0.86, + "latex": "4 3 6 \\times 1 0 2 4 )" + }, + { + "category_id": 13, + "poly": [ + 491, + 1434, + 635, + 1434, + 635, + 1465, + 491, + 1465 + ], + "score": 0.86, + "latex": "C H _ { - } O U T { = } I" + }, + { + "category_id": 13, + "poly": [ + 802, + 1976, + 931, + 1976, + 931, + 2006, + 802, + 2006 + ], + "score": 0.85, + "latex": "( 3 7 5 \\times 1 2 4 2 )" + }, + { + "category_id": 13, + "poly": [ + 1278, + 1885, + 1392, + 1885, + 1392, + 1915, + 1278, + 1915 + ], + "score": 0.85, + "latex": "4 4 8 \\times 5 7 6" + }, + { + "category_id": 13, + "poly": [ + 569, + 1343, + 689, + 1343, + 689, + 1373, + 569, + 1373 + ], + "score": 0.85, + "latex": "C H \\_ I N { = } 6" + }, + { + "category_id": 13, + "poly": [ + 1003, + 1915, + 1122, + 1915, + 1122, + 1945, + 1003, + 1945 + ], + "score": 0.84, + "latex": "( 4 8 0 \\times 6 4 0 )" + }, + { + "category_id": 13, + "poly": [ + 744, + 1343, + 891, + 1343, + 891, + 1374, + 744, + 1374 + ], + "score": 0.83, + "latex": "C H \\_ O U T { = } 3" + }, + { + "category_id": 13, + "poly": [ + 1190, + 1946, + 1306, + 1946, + 1306, + 1976, + 1190, + 1976 + ], + "score": 0.83, + "latex": "( 5 4 0 \\times 9 6 0 )" + }, + { + "category_id": 13, + "poly": [ + 1084, + 1435, + 1203, + 1435, + 1203, + 1464, + 1084, + 1464 + ], + "score": 0.79, + "latex": "C H \\_ I N { = } 8" + }, + { + "category_id": 13, + "poly": [ + 629, + 1464, + 769, + 1464, + 769, + 1495, + 629, + 1495 + ], + "score": 0.76, + "latex": "C H _ { - } O U T { = } 2" + }, + { + "category_id": 13, + "poly": [ + 552, + 1468, + 571, + 1468, + 571, + 1495, + 552, + 1495 + ], + "score": 0.75, + "latex": "y" + }, + { + "category_id": 13, + "poly": [ + 1385, + 1407, + 1406, + 1407, + 1406, + 1432, + 1385, + 1432 + ], + "score": 0.72, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 812, + 987, + 834, + 987, + 834, + 1009, + 812, + 1009 + ], + "score": 0.72, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 484, + 1470, + 503, + 1470, + 503, + 1492, + 484, + 1492 + ], + "score": 0.71, + "latex": "x" + }, + { + "category_id": 13, + "poly": [ + 1384, + 1438, + 1406, + 1438, + 1406, + 1462, + 1384, + 1462 + ], + "score": 0.71, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 1125, + 1404, + 1239, + 1404, + 1239, + 1433, + 1125, + 1433 + ], + "score": 0.63, + "latex": "C H \\_ I N { = } 4" + }, + { + "category_id": 13, + "poly": [ + 540, + 985, + 564, + 985, + 564, + 1009, + 540, + 1009 + ], + "score": 0.5, + "latex": "N" + }, + { + "category_id": 13, + "poly": [ + 1181, + 928, + 1285, + 928, + 1285, + 956, + 1181, + 956 + ], + "score": 0.26, + "latex": "C H \\_ O U T" + }, + { + "category_id": 13, + "poly": [ + 1057, + 927, + 1137, + 927, + 1137, + 956, + 1057, + 956 + ], + "score": 0.25, + "latex": "C H \\_ I N" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 925.0, + 1056.0, + 925.0, + 1056.0, + 960.0, + 293.0, + 960.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1138.0, + 925.0, + 1180.0, + 925.0, + 1180.0, + 960.0, + 1138.0, + 960.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1286.0, + 925.0, + 1407.0, + 925.0, + 1407.0, + 960.0, + 1286.0, + 960.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 956.0, + 1404.0, + 956.0, + 1404.0, + 985.0, + 295.0, + 985.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 984.0, + 539.0, + 984.0, + 539.0, + 1013.0, + 295.0, + 1013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 565.0, + 984.0, + 811.0, + 984.0, + 811.0, + 1013.0, + 565.0, + 1013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 835.0, + 984.0, + 1048.0, + 984.0, + 1048.0, + 1013.0, + 835.0, + 1013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1718.0, + 695.0, + 1718.0, + 695.0, + 1757.0, + 294.0, + 1757.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 197.0, + 1404.0, + 197.0, + 1404.0, + 228.0, + 296.0, + 228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 224.0, + 1404.0, + 224.0, + 1404.0, + 255.0, + 296.0, + 255.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 828.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2096.0, + 828.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 445.0, + 473.0, + 445.0, + 473.0, + 471.0, + 367.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 639.0, + 444.0, + 761.0, + 444.0, + 761.0, + 472.0, + 639.0, + 472.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 915.0, + 444.0, + 1036.0, + 444.0, + 1036.0, + 473.0, + 915.0, + 473.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 696.0, + 476.0, + 707.0, + 476.0, + 707.0, + 487.0, + 696.0, + 487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 378.0, + 507.0, + 462.0, + 507.0, + 462.0, + 533.0, + 378.0, + 533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 654.0, + 507.0, + 747.0, + 507.0, + 747.0, + 531.0, + 654.0, + 531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 910.0, + 497.0, + 1041.0, + 497.0, + 1041.0, + 541.0, + 910.0, + 541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1199.0, + 497.0, + 1329.0, + 497.0, + 1329.0, + 543.0, + 1199.0, + 543.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 545.0, + 417.0, + 545.0, + 417.0, + 570.0, + 329.0, + 570.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 716.0, + 545.0, + 798.0, + 545.0, + 798.0, + 570.0, + 716.0, + 570.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 388.0, + 585.0, + 452.0, + 585.0, + 452.0, + 611.0, + 388.0, + 611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 670.0, + 585.0, + 733.0, + 585.0, + 733.0, + 611.0, + 670.0, + 611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 824.0, + 580.0, + 843.0, + 580.0, + 843.0, + 602.0, + 824.0, + 602.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 951.0, + 578.0, + 998.0, + 578.0, + 998.0, + 606.0, + 951.0, + 606.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1177.0, + 582.0, + 1346.0, + 582.0, + 1346.0, + 608.0, + 1177.0, + 608.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 626.0, + 418.0, + 626.0, + 418.0, + 654.0, + 306.0, + 654.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 627.0, + 827.0, + 627.0, + 827.0, + 652.0, + 721.0, + 652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 874.0, + 625.0, + 973.0, + 625.0, + 973.0, + 641.0, + 874.0, + 641.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1017.0, + 626.0, + 1077.0, + 626.0, + 1077.0, + 642.0, + 1017.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1360.0, + 612.0, + 1401.0, + 612.0, + 1401.0, + 638.0, + 1360.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 925.0, + 649.0, + 1025.0, + 649.0, + 1025.0, + 672.0, + 925.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1072.0, + 647.0, + 1085.0, + 647.0, + 1085.0, + 679.0, + 1072.0, + 679.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1213.0, + 637.0, + 1315.0, + 637.0, + 1315.0, + 666.0, + 1213.0, + 666.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 667.0, + 453.0, + 667.0, + 453.0, + 693.0, + 387.0, + 693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 670.0, + 667.0, + 735.0, + 667.0, + 735.0, + 693.0, + 670.0, + 693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1082.0, + 670.0, + 1120.0, + 670.0, + 1120.0, + 698.0, + 1082.0, + 698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1259.0, + 680.0, + 1269.0, + 680.0, + 1269.0, + 689.0, + 1259.0, + 689.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 708.0, + 421.0, + 708.0, + 421.0, + 736.0, + 306.0, + 736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 720.0, + 708.0, + 832.0, + 708.0, + 832.0, + 736.0, + 720.0, + 736.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 890.0, + 705.0, + 1059.0, + 705.0, + 1059.0, + 733.0, + 890.0, + 733.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1121.0, + 708.0, + 1141.0, + 708.0, + 1141.0, + 731.0, + 1121.0, + 731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1240.0, + 707.0, + 1287.0, + 707.0, + 1287.0, + 733.0, + 1240.0, + 733.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 748.0, + 453.0, + 748.0, + 453.0, + 773.0, + 387.0, + 773.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 669.0, + 748.0, + 735.0, + 748.0, + 735.0, + 773.0, + 669.0, + 773.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 789.0, + 418.0, + 789.0, + 418.0, + 816.0, + 306.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 720.0, + 789.0, + 834.0, + 789.0, + 834.0, + 816.0, + 720.0, + 816.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1259.0, + 789.0, + 1269.0, + 789.0, + 1269.0, + 799.0, + 1259.0, + 799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1055.0, + 795.0, + 1176.0, + 795.0, + 1176.0, + 823.0, + 1055.0, + 823.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 828.0, + 454.0, + 828.0, + 454.0, + 853.0, + 387.0, + 853.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 492.0, + 820.0, + 630.0, + 820.0, + 630.0, + 848.0, + 492.0, + 848.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 666.0, + 828.0, + 736.0, + 828.0, + 736.0, + 857.0, + 666.0, + 857.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1204.0, + 830.0, + 1323.0, + 830.0, + 1323.0, + 858.0, + 1204.0, + 858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 496.0, + 874.0, + 617.0, + 874.0, + 617.0, + 906.0, + 496.0, + 906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 876.0, + 1011.0, + 876.0, + 1011.0, + 906.0, + 941.0, + 906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1228.0, + 874.0, + 1300.0, + 874.0, + 1300.0, + 908.0, + 1228.0, + 908.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1158.0, + 798.0, + 1158.0, + 798.0, + 1197.0, + 295.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1023.0, + 1158.0, + 1404.0, + 1158.0, + 1404.0, + 1197.0, + 1023.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1189.0, + 1408.0, + 1189.0, + 1408.0, + 1227.0, + 293.0, + 1227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1215.0, + 1407.0, + 1215.0, + 1407.0, + 1261.0, + 291.0, + 1261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1252.0, + 1404.0, + 1252.0, + 1404.0, + 1285.0, + 296.0, + 1285.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1282.0, + 1406.0, + 1282.0, + 1406.0, + 1317.0, + 293.0, + 1317.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1313.0, + 1408.0, + 1313.0, + 1408.0, + 1347.0, + 292.0, + 1347.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1343.0, + 568.0, + 1343.0, + 568.0, + 1378.0, + 292.0, + 1378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 690.0, + 1343.0, + 743.0, + 1343.0, + 743.0, + 1378.0, + 690.0, + 1378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 892.0, + 1343.0, + 1408.0, + 1343.0, + 1408.0, + 1378.0, + 892.0, + 1378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1371.0, + 1407.0, + 1371.0, + 1407.0, + 1410.0, + 292.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1403.0, + 1124.0, + 1403.0, + 1124.0, + 1440.0, + 295.0, + 1440.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1240.0, + 1403.0, + 1384.0, + 1403.0, + 1384.0, + 1440.0, + 1240.0, + 1440.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1433.0, + 490.0, + 1433.0, + 490.0, + 1469.0, + 295.0, + 1469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 636.0, + 1433.0, + 1083.0, + 1433.0, + 1083.0, + 1469.0, + 636.0, + 1469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1204.0, + 1433.0, + 1383.0, + 1433.0, + 1383.0, + 1469.0, + 1204.0, + 1469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1463.0, + 483.0, + 1463.0, + 483.0, + 1499.0, + 295.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 504.0, + 1463.0, + 551.0, + 1463.0, + 551.0, + 1499.0, + 504.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 572.0, + 1463.0, + 628.0, + 1463.0, + 628.0, + 1499.0, + 572.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 770.0, + 1463.0, + 1406.0, + 1463.0, + 1406.0, + 1499.0, + 770.0, + 1499.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1493.0, + 1404.0, + 1493.0, + 1404.0, + 1531.0, + 293.0, + 1531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1528.0, + 392.0, + 1528.0, + 392.0, + 1556.0, + 293.0, + 1556.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1777.0, + 1405.0, + 1777.0, + 1405.0, + 1814.0, + 294.0, + 1814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1805.0, + 1405.0, + 1805.0, + 1405.0, + 1847.0, + 293.0, + 1847.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1838.0, + 890.0, + 1838.0, + 890.0, + 1873.0, + 294.0, + 1873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1884.0, + 1277.0, + 1884.0, + 1277.0, + 1920.0, + 293.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1393.0, + 1884.0, + 1409.0, + 1884.0, + 1409.0, + 1920.0, + 1393.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1915.0, + 462.0, + 1915.0, + 462.0, + 1948.0, + 294.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 585.0, + 1915.0, + 694.0, + 1915.0, + 694.0, + 1948.0, + 585.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 817.0, + 1915.0, + 1002.0, + 1915.0, + 1002.0, + 1948.0, + 817.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1123.0, + 1915.0, + 1404.0, + 1915.0, + 1404.0, + 1948.0, + 1123.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1944.0, + 1189.0, + 1944.0, + 1189.0, + 1984.0, + 291.0, + 1984.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1307.0, + 1944.0, + 1406.0, + 1944.0, + 1406.0, + 1984.0, + 1307.0, + 1984.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1975.0, + 299.0, + 1975.0, + 299.0, + 2011.0, + 294.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 432.0, + 1975.0, + 525.0, + 1975.0, + 525.0, + 2011.0, + 432.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 656.0, + 1975.0, + 801.0, + 1975.0, + 801.0, + 2011.0, + 656.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 932.0, + 1975.0, + 946.0, + 1975.0, + 946.0, + 2011.0, + 932.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1568.0, + 1072.0, + 1568.0, + 1072.0, + 1605.0, + 294.0, + 1605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1194.0, + 1568.0, + 1405.0, + 1568.0, + 1405.0, + 1605.0, + 1194.0, + 1605.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1599.0, + 514.0, + 1599.0, + 514.0, + 1633.0, + 295.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 626.0, + 1599.0, + 863.0, + 1599.0, + 863.0, + 1633.0, + 626.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 1599.0, + 1217.0, + 1599.0, + 1217.0, + 1633.0, + 976.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1328.0, + 1599.0, + 1404.0, + 1599.0, + 1404.0, + 1633.0, + 1328.0, + 1633.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1632.0, + 705.0, + 1632.0, + 705.0, + 1664.0, + 295.0, + 1664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 814.0, + 1632.0, + 1016.0, + 1632.0, + 1016.0, + 1664.0, + 814.0, + 1664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1126.0, + 1632.0, + 1137.0, + 1632.0, + 1137.0, + 1664.0, + 1126.0, + 1664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1083.0, + 1404.0, + 1083.0, + 1404.0, + 1120.0, + 293.0, + 1120.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1112.0, + 499.0, + 1112.0, + 499.0, + 1148.0, + 295.0, + 1148.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 22, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 5, + "poly": [ + 433, + 1450, + 1265, + 1450, + 1265, + 1735, + 433, + 1735 + ], + "score": 0.982, + "html": "
NYUKITTI
No infillInfillNo infillInfill
Unroll stepsRELRMSRELRMSRELRMSRELRMS
00.0790.3310.0770.3380.2223.7700.0572.744
10.0760.3240.0750.3240.0852.8440.0562.700
20.0760.3170.0750.3150.0682.7990.0542.629
30.0750.3160.0740.3170.0612.7890.0542.591
40.0750.3160.0740.3150.0592.7390.0532.568
" + }, + { + "category_id": 1, + "poly": [ + 297, + 1096, + 1404, + 1096, + 1404, + 1279, + 297, + 1279 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 299, + 1884, + 1403, + 1884, + 1403, + 2008, + 299, + 2008 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 300, + 1778, + 1403, + 1778, + 1403, + 1870, + 300, + 1870 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 298, + 444, + 1403, + 444, + 1403, + 568, + 298, + 568 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 298, + 278, + 1403, + 278, + 1403, + 430, + 298, + 430 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 299, + 989, + 1404, + 989, + 1404, + 1081, + 299, + 1081 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 299, + 794, + 1405, + 794, + 1405, + 887, + 299, + 887 + ], + "score": 0.972 + }, + { + "category_id": 1, + "poly": [ + 299, + 688, + 1404, + 688, + 1404, + 780, + 299, + 780 + ], + "score": 0.971 + }, + { + "category_id": 1, + "poly": [ + 301, + 582, + 1401, + 582, + 1401, + 674, + 301, + 674 + ], + "score": 0.96 + }, + { + "category_id": 1, + "poly": [ + 297, + 202, + 1401, + 202, + 1401, + 266, + 297, + 266 + ], + "score": 0.952 + }, + { + "category_id": 6, + "poly": [ + 299, + 1319, + 1403, + 1319, + 1403, + 1433, + 299, + 1433 + ], + "score": 0.951 + }, + { + "category_id": 0, + "poly": [ + 298, + 935, + 1061, + 935, + 1061, + 968, + 298, + 968 + ], + "score": 0.932 + }, + { + "category_id": 2, + "poly": [ + 835, + 2061, + 864, + 2061, + 864, + 2085, + 835, + 2085 + ], + "score": 0.862 + }, + { + "category_id": 13, + "poly": [ + 675, + 340, + 799, + 340, + 799, + 369, + 675, + 369 + ], + "score": 0.89, + "latex": "9 6 8 \\times 1 2 9 6" + }, + { + "category_id": 13, + "poly": [ + 926, + 719, + 1050, + 719, + 1050, + 748, + 926, + 748 + ], + "score": 0.89, + "latex": "3 7 0 \\times 1 2 2 6" + }, + { + "category_id": 13, + "poly": [ + 326, + 340, + 437, + 340, + 437, + 370, + 326, + 370 + ], + "score": 0.88, + "latex": "4 8 0 \\times 6 4 0" + }, + { + "category_id": 13, + "poly": [ + 1291, + 613, + 1402, + 613, + 1402, + 642, + 1291, + 642 + ], + "score": 0.88, + "latex": "4 8 0 \\times 6 4 0" + }, + { + "category_id": 13, + "poly": [ + 296, + 400, + 405, + 400, + 405, + 431, + 296, + 431 + ], + "score": 0.86, + "latex": "2 4 0 \\times 3 2 0" + }, + { + "category_id": 13, + "poly": [ + 539, + 1218, + 596, + 1218, + 596, + 1248, + 539, + 1248 + ], + "score": 0.5, + "latex": "8 0 \\mathrm { m } \\mathrm { , }" + }, + { + "category_id": 13, + "poly": [ + 502, + 310, + 567, + 310, + 567, + 339, + 502, + 339 + ], + "score": 0.46, + "latex": "2 . 5 \\mathbf { M }" + }, + { + "category_id": 13, + "poly": [ + 963, + 584, + 995, + 584, + 995, + 611, + 963, + 611 + ], + "score": 0.38, + "latex": "\\nu 2" + }, + { + "category_id": 13, + "poly": [ + 828, + 644, + 875, + 644, + 875, + 672, + 828, + 672 + ], + "score": 0.33, + "latex": "5 0 \\mathrm { k }" + }, + { + "category_id": 13, + "poly": [ + 522, + 477, + 584, + 477, + 584, + 506, + 522, + 506 + ], + "score": 0.27, + "latex": "2 0 0 \\mathrm { k }" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1314.0, + 1405.0, + 1314.0, + 1405.0, + 1356.0, + 292.0, + 1356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1349.0, + 1403.0, + 1349.0, + 1403.0, + 1379.0, + 295.0, + 1379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1376.0, + 1404.0, + 1376.0, + 1404.0, + 1407.0, + 294.0, + 1407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1400.0, + 540.0, + 1400.0, + 540.0, + 1438.0, + 294.0, + 1438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 931.0, + 1066.0, + 931.0, + 1066.0, + 979.0, + 292.0, + 979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 828.0, + 2057.0, + 872.0, + 2057.0, + 872.0, + 2097.0, + 828.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1094.0, + 1405.0, + 1094.0, + 1405.0, + 1131.0, + 294.0, + 1131.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1124.0, + 1405.0, + 1124.0, + 1405.0, + 1161.0, + 292.0, + 1161.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1156.0, + 1405.0, + 1156.0, + 1405.0, + 1191.0, + 295.0, + 1191.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1189.0, + 1404.0, + 1189.0, + 1404.0, + 1220.0, + 296.0, + 1220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1215.0, + 538.0, + 1215.0, + 538.0, + 1253.0, + 294.0, + 1253.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 597.0, + 1215.0, + 1406.0, + 1215.0, + 1406.0, + 1253.0, + 597.0, + 1253.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1249.0, + 1148.0, + 1249.0, + 1148.0, + 1281.0, + 296.0, + 1281.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1884.0, + 1407.0, + 1884.0, + 1407.0, + 1920.0, + 294.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1914.0, + 1407.0, + 1914.0, + 1407.0, + 1951.0, + 294.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1946.0, + 1407.0, + 1946.0, + 1407.0, + 1982.0, + 294.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1976.0, + 1404.0, + 1976.0, + 1404.0, + 2012.0, + 294.0, + 2012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1778.0, + 1405.0, + 1778.0, + 1405.0, + 1812.0, + 295.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1805.0, + 1408.0, + 1805.0, + 1408.0, + 1844.0, + 293.0, + 1844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1840.0, + 620.0, + 1840.0, + 620.0, + 1873.0, + 295.0, + 1873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 443.0, + 1407.0, + 443.0, + 1407.0, + 480.0, + 292.0, + 480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 476.0, + 521.0, + 476.0, + 521.0, + 512.0, + 294.0, + 512.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 585.0, + 476.0, + 1405.0, + 476.0, + 1405.0, + 512.0, + 585.0, + 512.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 508.0, + 1405.0, + 508.0, + 1405.0, + 539.0, + 293.0, + 539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 534.0, + 1058.0, + 534.0, + 1058.0, + 574.0, + 292.0, + 574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 278.0, + 1407.0, + 278.0, + 1407.0, + 315.0, + 294.0, + 315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 307.0, + 501.0, + 307.0, + 501.0, + 346.0, + 293.0, + 346.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 568.0, + 307.0, + 1405.0, + 307.0, + 1405.0, + 346.0, + 568.0, + 346.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 338.0, + 325.0, + 338.0, + 325.0, + 371.0, + 294.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 438.0, + 338.0, + 674.0, + 338.0, + 674.0, + 371.0, + 438.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 800.0, + 338.0, + 1403.0, + 338.0, + 1403.0, + 371.0, + 800.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 368.0, + 1407.0, + 368.0, + 1407.0, + 404.0, + 293.0, + 404.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 406.0, + 399.0, + 415.0, + 399.0, + 415.0, + 433.0, + 406.0, + 433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 987.0, + 1405.0, + 987.0, + 1405.0, + 1027.0, + 293.0, + 1027.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1018.0, + 1404.0, + 1018.0, + 1404.0, + 1054.0, + 294.0, + 1054.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1051.0, + 957.0, + 1051.0, + 957.0, + 1084.0, + 295.0, + 1084.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 791.0, + 1407.0, + 791.0, + 1407.0, + 830.0, + 293.0, + 830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 825.0, + 1405.0, + 825.0, + 1405.0, + 859.0, + 295.0, + 859.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 854.0, + 1047.0, + 854.0, + 1047.0, + 894.0, + 294.0, + 894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 688.0, + 1404.0, + 688.0, + 1404.0, + 722.0, + 295.0, + 722.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 716.0, + 925.0, + 716.0, + 925.0, + 754.0, + 293.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1051.0, + 716.0, + 1405.0, + 716.0, + 1405.0, + 754.0, + 1051.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 748.0, + 1091.0, + 748.0, + 1091.0, + 786.0, + 293.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 581.0, + 962.0, + 581.0, + 962.0, + 618.0, + 295.0, + 618.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 996.0, + 581.0, + 1406.0, + 581.0, + 1406.0, + 618.0, + 996.0, + 618.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 612.0, + 1290.0, + 612.0, + 1290.0, + 646.0, + 296.0, + 646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 644.0, + 827.0, + 644.0, + 827.0, + 677.0, + 295.0, + 677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 876.0, + 644.0, + 1355.0, + 644.0, + 1355.0, + 677.0, + 876.0, + 677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 198.0, + 1403.0, + 198.0, + 1403.0, + 240.0, + 293.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 230.0, + 801.0, + 230.0, + 801.0, + 271.0, + 294.0, + 271.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 23, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 906, + 1404, + 906, + 1404, + 1121, + 298, + 1121 + ], + "score": 0.983 + }, + { + "category_id": 5, + "poly": [ + 396, + 336, + 1301, + 336, + 1301, + 568, + 396, + 568 + ], + "score": 0.982, + "html": "
NYUKITTI
No infillInfillNo infillInfill
RELRMSRELRMSRELRMSRELRMS
Baseline0.0790.3310.0770.3380.2223.7700.0572.744
Self conditioning0.0820.3350.0810.3330.2423.9400.0572.761
Step unrolling0.0760.3240.0750.3240.0852.8440.0562.700
" + }, + { + "category_id": 5, + "poly": [ + 295, + 1691, + 1406, + 1691, + 1406, + 1909, + 295, + 1909 + ], + "score": 0.98, + "html": "
MethodArchitectureResolutionTotal Time [ms]Inference stepsREL↓RMS↓
DPT-HybridNvidia RTX 2080384×384*38*-0.1100.357
DDVMTPU v4240 × 320204240.1040.378
272320.0860.342
544640.0770.324
10891280.0750.324
" + }, + { + "category_id": 1, + "poly": [ + 299, + 1202, + 1404, + 1202, + 1404, + 1296, + 299, + 1296 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 298, + 1310, + 1405, + 1310, + 1405, + 1402, + 298, + 1402 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 298, + 596, + 1405, + 596, + 1405, + 749, + 298, + 749 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 295, + 830, + 1400, + 830, + 1400, + 894, + 295, + 894 + ], + "score": 0.953 + }, + { + "category_id": 1, + "poly": [ + 296, + 1944, + 1401, + 1944, + 1401, + 2008, + 296, + 2008 + ], + "score": 0.944 + }, + { + "category_id": 1, + "poly": [ + 293, + 202, + 1402, + 202, + 1402, + 265, + 293, + 265 + ], + "score": 0.943 + }, + { + "category_id": 6, + "poly": [ + 503, + 291, + 1192, + 291, + 1192, + 323, + 503, + 323 + ], + "score": 0.926 + }, + { + "category_id": 0, + "poly": [ + 299, + 783, + 591, + 783, + 591, + 817, + 299, + 817 + ], + "score": 0.919 + }, + { + "category_id": 0, + "poly": [ + 297, + 1443, + 511, + 1443, + 511, + 1481, + 297, + 1481 + ], + "score": 0.906 + }, + { + "category_id": 0, + "poly": [ + 298, + 1155, + 484, + 1155, + 484, + 1187, + 298, + 1187 + ], + "score": 0.904 + }, + { + "category_id": 2, + "poly": [ + 834, + 2061, + 865, + 2061, + 865, + 2086, + 834, + 2086 + ], + "score": 0.864 + }, + { + "category_id": 6, + "poly": [ + 296, + 1501, + 1405, + 1501, + 1405, + 1672, + 296, + 1672 + ], + "score": 0.71 + }, + { + "category_id": 1, + "poly": [ + 296, + 1501, + 1405, + 1501, + 1405, + 1672, + 296, + 1672 + ], + "score": 0.482 + }, + { + "category_id": 13, + "poly": [ + 472, + 936, + 580, + 936, + 580, + 967, + 472, + 967 + ], + "score": 0.92, + "latex": "1 \\times \\dot { 1 } 0 ^ { - 4 }" + }, + { + "category_id": 13, + "poly": [ + 672, + 967, + 780, + 967, + 780, + 998, + 672, + 998 + ], + "score": 0.92, + "latex": "3 \\times 1 0 ^ { - 5 }" + }, + { + "category_id": 13, + "poly": [ + 298, + 1370, + 411, + 1370, + 411, + 1404, + 298, + 1404 + ], + "score": 0.91, + "latex": "t ^ { \\prime } = 8 / 6 4" + }, + { + "category_id": 13, + "poly": [ + 781, + 1643, + 891, + 1643, + 891, + 1669, + 781, + 1669 + ], + "score": 0.9, + "latex": "4 8 0 \\times 6 4 0" + }, + { + "category_id": 13, + "poly": [ + 864, + 1614, + 974, + 1614, + 974, + 1641, + 864, + 1641 + ], + "score": 0.9, + "latex": "3 8 4 \\times 3 8 4" + }, + { + "category_id": 13, + "poly": [ + 1083, + 909, + 1118, + 909, + 1118, + 938, + 1083, + 938 + ], + "score": 0.88, + "latex": "L _ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 1133, + 833, + 1167, + 833, + 1167, + 862, + 1133, + 862 + ], + "score": 0.88, + "latex": "L _ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 1113, + 1341, + 1241, + 1341, + 1241, + 1373, + 1113, + 1373 + ], + "score": 0.88, + "latex": "t ^ { \\prime } = 3 2 / 6 4" + }, + { + "category_id": 13, + "poly": [ + 711, + 1311, + 766, + 1311, + 766, + 1340, + 711, + 1340 + ], + "score": 0.87, + "latex": "2 \\times 5" + }, + { + "category_id": 13, + "poly": [ + 849, + 1587, + 897, + 1587, + 897, + 1612, + 849, + 1612 + ], + "score": 0.86, + "latex": "{ \\sim } 5 \\mathbf { x }" + }, + { + "category_id": 13, + "poly": [ + 1180, + 1312, + 1208, + 1312, + 1208, + 1339, + 1180, + 1339 + ], + "score": 0.57, + "latex": "\\times" + }, + { + "category_id": 13, + "poly": [ + 443, + 1029, + 491, + 1029, + 491, + 1057, + 443, + 1057 + ], + "score": 0.37, + "latex": "4 0 \\mathrm { k }" + }, + { + "category_id": 13, + "poly": [ + 924, + 832, + 987, + 832, + 987, + 861, + 924, + 861 + ], + "score": 0.36, + "latex": "2 . 8 \\mathbf { M }" + }, + { + "category_id": 13, + "poly": [ + 1065, + 1029, + 1131, + 1029, + 1131, + 1058, + 1065, + 1058 + ], + "score": 0.36, + "latex": "0 . 9 { \\bf M }" + }, + { + "category_id": 13, + "poly": [ + 798, + 1060, + 861, + 1060, + 861, + 1089, + 798, + 1089 + ], + "score": 0.26, + "latex": "3 . 7 \\mathbf { M }" + }, + { + "category_id": 15, + "poly": [ + 504.0, + 286.0, + 1193.0, + 286.0, + 1193.0, + 328.0, + 504.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 778.0, + 595.0, + 778.0, + 595.0, + 825.0, + 292.0, + 825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1438.0, + 516.0, + 1438.0, + 516.0, + 1489.0, + 290.0, + 1489.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1152.0, + 487.0, + 1152.0, + 487.0, + 1193.0, + 294.0, + 1193.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 829.0, + 2056.0, + 871.0, + 2056.0, + 871.0, + 2097.0, + 829.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1503.0, + 1402.0, + 1503.0, + 1402.0, + 1533.0, + 295.0, + 1533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1530.0, + 1405.0, + 1530.0, + 1405.0, + 1562.0, + 294.0, + 1562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1553.0, + 1406.0, + 1553.0, + 1406.0, + 1595.0, + 291.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1585.0, + 848.0, + 1585.0, + 848.0, + 1619.0, + 294.0, + 1619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 898.0, + 1585.0, + 1407.0, + 1585.0, + 1407.0, + 1619.0, + 898.0, + 1619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1614.0, + 863.0, + 1614.0, + 863.0, + 1645.0, + 295.0, + 1645.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 975.0, + 1614.0, + 1406.0, + 1614.0, + 1406.0, + 1645.0, + 975.0, + 1645.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1637.0, + 780.0, + 1637.0, + 780.0, + 1675.0, + 291.0, + 1675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 892.0, + 1637.0, + 1286.0, + 1637.0, + 1286.0, + 1675.0, + 892.0, + 1675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 908.0, + 1082.0, + 908.0, + 1082.0, + 942.0, + 296.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1119.0, + 908.0, + 1405.0, + 908.0, + 1405.0, + 942.0, + 1119.0, + 942.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 934.0, + 471.0, + 934.0, + 471.0, + 974.0, + 292.0, + 974.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 581.0, + 934.0, + 1408.0, + 934.0, + 1408.0, + 974.0, + 581.0, + 974.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 964.0, + 671.0, + 964.0, + 671.0, + 1004.0, + 292.0, + 1004.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 781.0, + 964.0, + 1406.0, + 964.0, + 1406.0, + 1004.0, + 781.0, + 1004.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 998.0, + 1404.0, + 998.0, + 1404.0, + 1032.0, + 294.0, + 1032.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1028.0, + 442.0, + 1028.0, + 442.0, + 1063.0, + 294.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 492.0, + 1028.0, + 1064.0, + 1028.0, + 1064.0, + 1063.0, + 492.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1132.0, + 1028.0, + 1404.0, + 1028.0, + 1404.0, + 1063.0, + 1132.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1058.0, + 797.0, + 1058.0, + 797.0, + 1095.0, + 293.0, + 1095.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 862.0, + 1058.0, + 1406.0, + 1058.0, + 1406.0, + 1095.0, + 862.0, + 1095.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1090.0, + 1052.0, + 1090.0, + 1052.0, + 1124.0, + 294.0, + 1124.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1200.0, + 1404.0, + 1200.0, + 1404.0, + 1239.0, + 293.0, + 1239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1234.0, + 1405.0, + 1234.0, + 1405.0, + 1268.0, + 295.0, + 1268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1266.0, + 634.0, + 1266.0, + 634.0, + 1301.0, + 295.0, + 1301.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1307.0, + 710.0, + 1307.0, + 710.0, + 1345.0, + 295.0, + 1345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 767.0, + 1307.0, + 1179.0, + 1307.0, + 1179.0, + 1345.0, + 767.0, + 1345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1209.0, + 1307.0, + 1407.0, + 1307.0, + 1407.0, + 1345.0, + 1209.0, + 1345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1340.0, + 1112.0, + 1340.0, + 1112.0, + 1373.0, + 292.0, + 1373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1242.0, + 1340.0, + 1405.0, + 1340.0, + 1405.0, + 1373.0, + 1242.0, + 1373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1364.0, + 297.0, + 1364.0, + 297.0, + 1407.0, + 290.0, + 1407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 412.0, + 1364.0, + 424.0, + 1364.0, + 424.0, + 1407.0, + 412.0, + 1407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 593.0, + 1408.0, + 593.0, + 1408.0, + 632.0, + 295.0, + 632.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 625.0, + 1405.0, + 625.0, + 1405.0, + 662.0, + 293.0, + 662.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 654.0, + 1403.0, + 654.0, + 1403.0, + 692.0, + 293.0, + 692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 684.0, + 1406.0, + 684.0, + 1406.0, + 724.0, + 292.0, + 724.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 715.0, + 965.0, + 715.0, + 965.0, + 756.0, + 293.0, + 756.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 830.0, + 923.0, + 830.0, + 923.0, + 866.0, + 295.0, + 866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 988.0, + 830.0, + 1132.0, + 830.0, + 1132.0, + 866.0, + 988.0, + 866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1168.0, + 830.0, + 1405.0, + 830.0, + 1405.0, + 866.0, + 1168.0, + 866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 862.0, + 1277.0, + 862.0, + 1277.0, + 898.0, + 295.0, + 898.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1941.0, + 1405.0, + 1941.0, + 1405.0, + 1982.0, + 294.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1975.0, + 1406.0, + 1975.0, + 1406.0, + 2011.0, + 295.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 199.0, + 1405.0, + 199.0, + 1405.0, + 240.0, + 292.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 235.0, + 613.0, + 235.0, + 613.0, + 267.0, + 296.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1503.0, + 1402.0, + 1503.0, + 1402.0, + 1533.0, + 295.0, + 1533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1530.0, + 1405.0, + 1530.0, + 1405.0, + 1562.0, + 294.0, + 1562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1553.0, + 1406.0, + 1553.0, + 1406.0, + 1595.0, + 291.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1585.0, + 848.0, + 1585.0, + 848.0, + 1619.0, + 294.0, + 1619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 898.0, + 1585.0, + 1407.0, + 1585.0, + 1407.0, + 1619.0, + 898.0, + 1619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1614.0, + 863.0, + 1614.0, + 863.0, + 1645.0, + 295.0, + 1645.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 975.0, + 1614.0, + 1406.0, + 1614.0, + 1406.0, + 1645.0, + 975.0, + 1645.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1637.0, + 780.0, + 1637.0, + 780.0, + 1675.0, + 291.0, + 1675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 892.0, + 1637.0, + 1286.0, + 1637.0, + 1286.0, + 1675.0, + 892.0, + 1675.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 24, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 201, + 1405, + 201, + 1405, + 508, + 297, + 508 + ], + "score": 0.982 + }, + { + "category_id": 3, + "poly": [ + 308, + 642, + 1385, + 642, + 1385, + 1642, + 308, + 1642 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 299, + 521, + 1404, + 521, + 1404, + 615, + 299, + 615 + ], + "score": 0.969 + }, + { + "category_id": 4, + "poly": [ + 297, + 1665, + 1404, + 1665, + 1404, + 1778, + 297, + 1778 + ], + "score": 0.954 + }, + { + "category_id": 2, + "poly": [ + 835, + 2061, + 865, + 2061, + 865, + 2085, + 835, + 2085 + ], + "score": 0.57 + }, + { + "category_id": 2, + "poly": [ + 835, + 2062, + 865, + 2062, + 865, + 2086, + 835, + 2086 + ], + "score": 0.514 + }, + { + "category_id": 13, + "poly": [ + 1096, + 234, + 1154, + 234, + 1154, + 263, + 1096, + 263 + ], + "score": 0.7, + "latex": "{ \\sim } 8 . 5" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 641.0, + 721.0, + 641.0, + 721.0, + 675.0, + 345.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 852.0, + 638.0, + 920.0, + 638.0, + 920.0, + 676.0, + 852.0, + 676.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1124.0, + 637.0, + 1306.0, + 637.0, + 1306.0, + 677.0, + 1124.0, + 677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 309.0, + 793.0, + 345.0, + 793.0, + 345.0, + 903.0, + 309.0, + 903.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 806.0, + 962.0, + 962.0, + 962.0, + 962.0, + 1001.0, + 806.0, + 1001.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1144.0, + 962.0, + 1285.0, + 962.0, + 1285.0, + 1001.0, + 1144.0, + 1001.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 309.0, + 1110.0, + 346.0, + 1110.0, + 346.0, + 1196.0, + 309.0, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 813.0, + 1289.0, + 954.0, + 1289.0, + 954.0, + 1325.0, + 813.0, + 1325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1144.0, + 1288.0, + 1285.0, + 1288.0, + 1285.0, + 1324.0, + 1144.0, + 1324.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 951.0, + 1348.0, + 964.0, + 1348.0, + 964.0, + 1369.0, + 951.0, + 1369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1282.0, + 1349.0, + 1295.0, + 1349.0, + 1295.0, + 1369.0, + 1282.0, + 1369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 311.0, + 1443.0, + 344.0, + 1443.0, + 344.0, + 1541.0, + 311.0, + 1541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 815.0, + 1616.0, + 953.0, + 1616.0, + 953.0, + 1648.0, + 815.0, + 1648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1146.0, + 1616.0, + 1284.0, + 1616.0, + 1284.0, + 1648.0, + 1146.0, + 1648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1667.0, + 1404.0, + 1667.0, + 1404.0, + 1696.0, + 296.0, + 1696.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1695.0, + 1404.0, + 1695.0, + 1404.0, + 1724.0, + 295.0, + 1724.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1720.0, + 1405.0, + 1720.0, + 1405.0, + 1753.0, + 295.0, + 1753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1749.0, + 1361.0, + 1749.0, + 1361.0, + 1779.0, + 296.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 830.0, + 2057.0, + 870.0, + 2057.0, + 870.0, + 2094.0, + 830.0, + 2094.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 830.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2094.0, + 830.0, + 2094.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 204.0, + 1405.0, + 204.0, + 1405.0, + 236.0, + 295.0, + 236.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 229.0, + 1095.0, + 229.0, + 1095.0, + 273.0, + 292.0, + 273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1155.0, + 229.0, + 1406.0, + 229.0, + 1406.0, + 273.0, + 1155.0, + 273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 263.0, + 1406.0, + 263.0, + 1406.0, + 299.0, + 294.0, + 299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 296.0, + 1403.0, + 296.0, + 1403.0, + 328.0, + 296.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 323.0, + 1405.0, + 323.0, + 1405.0, + 360.0, + 294.0, + 360.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 353.0, + 1405.0, + 353.0, + 1405.0, + 389.0, + 294.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 384.0, + 1406.0, + 384.0, + 1406.0, + 419.0, + 294.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 414.0, + 1406.0, + 414.0, + 1406.0, + 451.0, + 291.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 444.0, + 1405.0, + 444.0, + 1405.0, + 482.0, + 292.0, + 482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 471.0, + 578.0, + 471.0, + 578.0, + 512.0, + 294.0, + 512.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 519.0, + 1408.0, + 519.0, + 1408.0, + 557.0, + 294.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 552.0, + 1404.0, + 552.0, + 1404.0, + 586.0, + 295.0, + 586.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 583.0, + 962.0, + 583.0, + 962.0, + 618.0, + 293.0, + 618.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 25, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 299, + 729, + 1405, + 729, + 1405, + 822, + 299, + 822 + ], + "score": 0.97 + }, + { + "category_id": 3, + "poly": [ + 302, + 193, + 1401, + 193, + 1401, + 549, + 302, + 549 + ], + "score": 0.965 + }, + { + "category_id": 4, + "poly": [ + 295, + 564, + 1400, + 564, + 1400, + 651, + 295, + 651 + ], + "score": 0.955 + }, + { + "category_id": 2, + "poly": [ + 835, + 2061, + 864, + 2061, + 864, + 2085, + 835, + 2085 + ], + "score": 0.849 + }, + { + "category_id": 2, + "poly": [ + 916, + 196, + 1231, + 196, + 1231, + 220, + 916, + 220 + ], + "score": 0.372 + }, + { + "category_id": 15, + "poly": [ + 342.0, + 188.0, + 478.0, + 188.0, + 478.0, + 230.0, + 342.0, + 230.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 187.0, + 732.0, + 187.0, + 732.0, + 229.0, + 532.0, + 229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 913.0, + 190.0, + 1236.0, + 190.0, + 1236.0, + 226.0, + 913.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 564.0, + 1404.0, + 564.0, + 1404.0, + 597.0, + 295.0, + 597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 592.0, + 1405.0, + 592.0, + 1405.0, + 624.0, + 293.0, + 624.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 622.0, + 1225.0, + 622.0, + 1225.0, + 653.0, + 294.0, + 653.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 829.0, + 2057.0, + 870.0, + 2057.0, + 870.0, + 2098.0, + 829.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 912.0, + 190.0, + 1233.0, + 190.0, + 1233.0, + 226.0, + 912.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 728.0, + 1403.0, + 728.0, + 1403.0, + 762.0, + 297.0, + 762.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 757.0, + 1409.0, + 757.0, + 1409.0, + 796.0, + 293.0, + 796.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 791.0, + 596.0, + 791.0, + 596.0, + 824.0, + 293.0, + 824.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 26, + "width": 1700, + "height": 2200 + } + } +] \ No newline at end of file diff --git a/parse/dev/k7p_YAO7yE/images/129cef6bd75a15c5201aae4bbb94a7bf0bb3fee667092a26e8678af069f3c2f5.jpg b/parse/dev/k7p_YAO7yE/images/129cef6bd75a15c5201aae4bbb94a7bf0bb3fee667092a26e8678af069f3c2f5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2c0764d11a4dd6e09935ce0380b1475e4a9a3dbd --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/129cef6bd75a15c5201aae4bbb94a7bf0bb3fee667092a26e8678af069f3c2f5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f3fd0512d0b5dc9c5b8aca359d50b0f4a3485de2f6ea366ddd3a443fb7ba7f9 +size 12352 diff --git a/parse/dev/k7p_YAO7yE/images/37c989be0cb836896eeff5c13f0f4620dd8a603625baa4ae2132ba06daa86741.jpg b/parse/dev/k7p_YAO7yE/images/37c989be0cb836896eeff5c13f0f4620dd8a603625baa4ae2132ba06daa86741.jpg new file mode 100644 index 0000000000000000000000000000000000000000..45b8a495da52952d56a82299370d956f67a30051 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/37c989be0cb836896eeff5c13f0f4620dd8a603625baa4ae2132ba06daa86741.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c397b2a99aeed02cad3e1120cc60c2c2bbfd9870e3708792282f5b8c37bd90b3 +size 254700 diff --git a/parse/dev/k7p_YAO7yE/images/442c75196446db9e4d6ce4681b3ea6fdba3f6296d7dac26039c816c843c64ea5.jpg b/parse/dev/k7p_YAO7yE/images/442c75196446db9e4d6ce4681b3ea6fdba3f6296d7dac26039c816c843c64ea5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1fb45733da7246c9a2a21161cf14189ccc22ee70 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/442c75196446db9e4d6ce4681b3ea6fdba3f6296d7dac26039c816c843c64ea5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6466a44a26bc4753f5d9d716e10e7350842342b03a710d92790cbd46ba299701 +size 25930 diff --git a/parse/dev/k7p_YAO7yE/images/44470985fbebc5c3a343371a92291c005aec3f1320774b5f697ec8d4f81d413e.jpg b/parse/dev/k7p_YAO7yE/images/44470985fbebc5c3a343371a92291c005aec3f1320774b5f697ec8d4f81d413e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bd006a6a69c266e518b8ba6ae93324d4e24698b4 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/44470985fbebc5c3a343371a92291c005aec3f1320774b5f697ec8d4f81d413e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1cd2405d43542d6cf7f675de65ebc7ba8018c68213bc7698c9d484ed11ddfef7 +size 7994 diff --git a/parse/dev/k7p_YAO7yE/images/45859304927597db8a40dc2258a0e8ba5abbcc5103ed79c74ea9a277a7a40826.jpg b/parse/dev/k7p_YAO7yE/images/45859304927597db8a40dc2258a0e8ba5abbcc5103ed79c74ea9a277a7a40826.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e87a2d3a38206b9ed36515556b61de15d9c213b4 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/45859304927597db8a40dc2258a0e8ba5abbcc5103ed79c74ea9a277a7a40826.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c1a6abefabca2cc70e2768e647326b654be360a942b757a93d3a8df62d4e2f2 +size 6210 diff --git a/parse/dev/k7p_YAO7yE/images/46ab0585f29d036042366cf8c6c394c662ac1b198d23fb999a1414405369df70.jpg b/parse/dev/k7p_YAO7yE/images/46ab0585f29d036042366cf8c6c394c662ac1b198d23fb999a1414405369df70.jpg new file mode 100644 index 0000000000000000000000000000000000000000..904dd3d04db0307645404cbaeb37986580d484fb --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/46ab0585f29d036042366cf8c6c394c662ac1b198d23fb999a1414405369df70.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ea5aef242cc09ba78adaa4f6ffcd1b9e333b99bd07ff137c16892c7bcbe6d1d +size 43850 diff --git a/parse/dev/k7p_YAO7yE/images/563cced5b8e23d9c10ef88bd7477a2fc359c4540b1d751059ad99cd82c977105.jpg b/parse/dev/k7p_YAO7yE/images/563cced5b8e23d9c10ef88bd7477a2fc359c4540b1d751059ad99cd82c977105.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dab369276d374b3ba79b2e68616990a851be8a58 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/563cced5b8e23d9c10ef88bd7477a2fc359c4540b1d751059ad99cd82c977105.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b3969fb16c6fb85dac8fd7828255d4f5fa70b7b6de4c66c90dd9a10973001698 +size 48712 diff --git a/parse/dev/k7p_YAO7yE/images/586352ca7103a193b42533163558a6f313c4d40e46708360b74a8ac9c17064b4.jpg b/parse/dev/k7p_YAO7yE/images/586352ca7103a193b42533163558a6f313c4d40e46708360b74a8ac9c17064b4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..39dbe579772bed553dbea8ccb9c515b490346ba6 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/586352ca7103a193b42533163558a6f313c4d40e46708360b74a8ac9c17064b4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad1a7f9be82cddbab7a0df8dafc2e0a51ce5c0da924d3b8d6766e57c32651951 +size 17491 diff --git a/parse/dev/k7p_YAO7yE/images/68c3ed4b524c761a7a0f6ce03e76850b566e9b28d521a64607c0eeb09dc08b9e.jpg b/parse/dev/k7p_YAO7yE/images/68c3ed4b524c761a7a0f6ce03e76850b566e9b28d521a64607c0eeb09dc08b9e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e72b39bd5176ac24fc0f103542b3f29ca92936de --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/68c3ed4b524c761a7a0f6ce03e76850b566e9b28d521a64607c0eeb09dc08b9e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ec4b0660364e3757fdf2c32baa9316c4b3e7fe71a85de52970ec3a5b383d5b4 +size 19288 diff --git a/parse/dev/k7p_YAO7yE/images/6bcde0cc44502b91bc98e4a21f189cfb5a64afcd1aa3970ab81de8fd24a98c1d.jpg b/parse/dev/k7p_YAO7yE/images/6bcde0cc44502b91bc98e4a21f189cfb5a64afcd1aa3970ab81de8fd24a98c1d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..be0d60d6a25c81e82459c03f3e9adc30092988a0 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/6bcde0cc44502b91bc98e4a21f189cfb5a64afcd1aa3970ab81de8fd24a98c1d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b568d27c127034a5ea9d1b964825e4757282d22c2f38b30a7b28351cf4976178 +size 6511 diff --git a/parse/dev/k7p_YAO7yE/images/6e91a343b1cb822f74879ec31a00686584ce059e44da28149ba02b412dac0dc6.jpg b/parse/dev/k7p_YAO7yE/images/6e91a343b1cb822f74879ec31a00686584ce059e44da28149ba02b412dac0dc6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ae504eb172ebcc1b3b7d795435cfbf6abfa41fba --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/6e91a343b1cb822f74879ec31a00686584ce059e44da28149ba02b412dac0dc6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f230654de758c3bc36b1f10f3e1011839562796a84370f3840601964eb530b4 +size 15619 diff --git a/parse/dev/k7p_YAO7yE/images/6f61360c6ab7d7d7b1c8c900e57769302de699f87205de417c3fff39b141f380.jpg b/parse/dev/k7p_YAO7yE/images/6f61360c6ab7d7d7b1c8c900e57769302de699f87205de417c3fff39b141f380.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b4e5230449c8d124b41cdefdff66fcb1c9f6f710 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/6f61360c6ab7d7d7b1c8c900e57769302de699f87205de417c3fff39b141f380.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:54975b027f4af6c48ba3eb2bcf14c7a928b4c3227490ab44878abdd58f60ec58 +size 8142 diff --git a/parse/dev/k7p_YAO7yE/images/7ea5a83fd3eb7f61e860733815bc2d3cede683f08fdbd9d04142017e16c5c7fe.jpg b/parse/dev/k7p_YAO7yE/images/7ea5a83fd3eb7f61e860733815bc2d3cede683f08fdbd9d04142017e16c5c7fe.jpg new file mode 100644 index 0000000000000000000000000000000000000000..63f3aa2dd6ad3a02bc4e1807c261b57043569172 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/7ea5a83fd3eb7f61e860733815bc2d3cede683f08fdbd9d04142017e16c5c7fe.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:46eefd925d6f892534155a3602f96197102ea6ac1c8fcef161245b6c03ba0eee +size 66870 diff --git a/parse/dev/k7p_YAO7yE/images/817eb1b088a9d8e03b9f8f2020e74dea75136c5c298aa4a745459ade03f0ada7.jpg b/parse/dev/k7p_YAO7yE/images/817eb1b088a9d8e03b9f8f2020e74dea75136c5c298aa4a745459ade03f0ada7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9494ef01fe5cca86117f00f14e74930fa1816676 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/817eb1b088a9d8e03b9f8f2020e74dea75136c5c298aa4a745459ade03f0ada7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c7073e1601d53535c6da9e0f3bfccadeec4ea2f8a6150b8a0029b1545ce0df3 +size 18959 diff --git a/parse/dev/k7p_YAO7yE/images/857c8dce537cc7014a1d80c0643a123a44a0b6f7e32a7d5e69393cfb898d115b.jpg b/parse/dev/k7p_YAO7yE/images/857c8dce537cc7014a1d80c0643a123a44a0b6f7e32a7d5e69393cfb898d115b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..be6459d453151e62bdd2b7d4d4f8f13ad862fb2f --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/857c8dce537cc7014a1d80c0643a123a44a0b6f7e32a7d5e69393cfb898d115b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce913676b2b6f993b850eb2c21aca352b373254ed57d2aa4074b1ef3f9a5659e +size 165526 diff --git a/parse/dev/k7p_YAO7yE/images/9fd7e0c767542eef4c0e5688261ec38ad88b49157402768f8d0ef5f2904a4248.jpg b/parse/dev/k7p_YAO7yE/images/9fd7e0c767542eef4c0e5688261ec38ad88b49157402768f8d0ef5f2904a4248.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6616b8d52ce6c9275ee976624b1ed49134d37c82 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/9fd7e0c767542eef4c0e5688261ec38ad88b49157402768f8d0ef5f2904a4248.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:84f21e80b1e3311026b71a6766daee2abef1ec68953cf487a1690282bea792dd +size 26301 diff --git a/parse/dev/k7p_YAO7yE/images/a2f5521c677bf10cb505661c95b55a68adc56b0e18766aae1ebfff82c5d6b811.jpg b/parse/dev/k7p_YAO7yE/images/a2f5521c677bf10cb505661c95b55a68adc56b0e18766aae1ebfff82c5d6b811.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b59b9ed8537769001761731162cfaa7e0cbb2c1a --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/a2f5521c677bf10cb505661c95b55a68adc56b0e18766aae1ebfff82c5d6b811.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:805cff198d19323f39e2476aca835591eec34c1f3d8b14cc162c10f6d5b2f90f +size 7052 diff --git a/parse/dev/k7p_YAO7yE/images/ac216522fd7c7ee6ce1d37b51982abac96871f9c6a7ef3d292d5d6464fc56746.jpg b/parse/dev/k7p_YAO7yE/images/ac216522fd7c7ee6ce1d37b51982abac96871f9c6a7ef3d292d5d6464fc56746.jpg new file mode 100644 index 0000000000000000000000000000000000000000..02b17b6e0a4976ddb8e481809683fc1841b78be8 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/ac216522fd7c7ee6ce1d37b51982abac96871f9c6a7ef3d292d5d6464fc56746.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:11343df727406bbfe4cdf7bae0665139e174fc68e708ec1f7be8b2a5ceb10270 +size 22405 diff --git a/parse/dev/k7p_YAO7yE/images/af1caa2c6f8a81947f2073b01aa6b2d24ab89c6980aced5b58ff939053d2983b.jpg b/parse/dev/k7p_YAO7yE/images/af1caa2c6f8a81947f2073b01aa6b2d24ab89c6980aced5b58ff939053d2983b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6fcc7b50d2cf9272e7e6f0d5e8fb1de3e80df341 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/af1caa2c6f8a81947f2073b01aa6b2d24ab89c6980aced5b58ff939053d2983b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1200d7a094dafbf48e6d15f6416949c38b74d765e4ba4a21b80a97ed870f2440 +size 140119 diff --git a/parse/dev/k7p_YAO7yE/images/afda4ddf43be97915bb7d02e7f429e83c03cbc14a83d7682fa07edbe128bba63.jpg b/parse/dev/k7p_YAO7yE/images/afda4ddf43be97915bb7d02e7f429e83c03cbc14a83d7682fa07edbe128bba63.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f5a701ef9f626c7982fe8ef95c86dc9b47390b67 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/afda4ddf43be97915bb7d02e7f429e83c03cbc14a83d7682fa07edbe128bba63.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75f33a76f0f9990921499f8e90b00c341f8598f9dbbd1feb2a62987ee56caa46 +size 4525 diff --git a/parse/dev/k7p_YAO7yE/images/c441aaa6bd3b16d830ee68baf36b34a2b50d6d3aec956daaded1f753890d71f9.jpg b/parse/dev/k7p_YAO7yE/images/c441aaa6bd3b16d830ee68baf36b34a2b50d6d3aec956daaded1f753890d71f9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2d2cbaf5c7092d43b78d5c49b4a682baa462cca5 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/c441aaa6bd3b16d830ee68baf36b34a2b50d6d3aec956daaded1f753890d71f9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e4743914a64a6e37f41ef01705698e7bdd61e4bbaf17ff2cf2c9d3920ff0a8e +size 3516 diff --git a/parse/dev/k7p_YAO7yE/images/cc03d9af623561a51a1cc5c541420049b1469b3c42ce3d88710ecf134923b9d7.jpg b/parse/dev/k7p_YAO7yE/images/cc03d9af623561a51a1cc5c541420049b1469b3c42ce3d88710ecf134923b9d7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..97312a2463db2d844dc262287f884186d0c3623b --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/cc03d9af623561a51a1cc5c541420049b1469b3c42ce3d88710ecf134923b9d7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a55fa04a2f552c97a14896f14a897329d516c533ff8ba0ca33cd449c526fe611 +size 26384 diff --git a/parse/dev/k7p_YAO7yE/images/d50b549d04375dd31dbe89911c772f5401d6716d8bb5d05926baf65971353baf.jpg b/parse/dev/k7p_YAO7yE/images/d50b549d04375dd31dbe89911c772f5401d6716d8bb5d05926baf65971353baf.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5458a1cd9e6c15c428f12994bea94ee29e40c290 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/d50b549d04375dd31dbe89911c772f5401d6716d8bb5d05926baf65971353baf.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87ec8343554f44e6d6bc1bc3a3b348b70d3cbda0559c8c4e549bcf9ad2881b59 +size 168201 diff --git a/parse/dev/k7p_YAO7yE/images/e0cd7e96f42a6e4975c6cefcd7290a0bbd28a809531ed2db6021a04614e8bd94.jpg b/parse/dev/k7p_YAO7yE/images/e0cd7e96f42a6e4975c6cefcd7290a0bbd28a809531ed2db6021a04614e8bd94.jpg new file mode 100644 index 0000000000000000000000000000000000000000..77c2b36d2e4de224f4080098efd7ff508cb20619 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/e0cd7e96f42a6e4975c6cefcd7290a0bbd28a809531ed2db6021a04614e8bd94.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:896e32ee1b8bc6ab40b337739b0fae5978a599769f490cab830484aff47e4ed5 +size 258843 diff --git a/parse/dev/k7p_YAO7yE/images/e6258a2d2ee03801e19e146819b9a2acae5a352c7125878ab519d9fde4613ef9.jpg b/parse/dev/k7p_YAO7yE/images/e6258a2d2ee03801e19e146819b9a2acae5a352c7125878ab519d9fde4613ef9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b2600363ad459040125bd837c6307b3be0516739 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/e6258a2d2ee03801e19e146819b9a2acae5a352c7125878ab519d9fde4613ef9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:123c70a3972650adda3da6edf8fc11bf4e7098258c2727ba4214662e74673895 +size 3335 diff --git a/parse/dev/k7p_YAO7yE/images/e6371de3489ae44f31c91a3f862822dcb247eee3f31af979a218e8a747e726c2.jpg b/parse/dev/k7p_YAO7yE/images/e6371de3489ae44f31c91a3f862822dcb247eee3f31af979a218e8a747e726c2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..21a4b7645dcbad922d13d6356d76d95ed335e408 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/e6371de3489ae44f31c91a3f862822dcb247eee3f31af979a218e8a747e726c2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dde9e8c579c5700597207eafc263b9e9b89b522a1df3e273e634a8c1b96e0b1a +size 64843 diff --git a/parse/dev/k7p_YAO7yE/images/e76048854249de2e942a26492e96088624c2753b3e4c12a77019a1e4f57d41d0.jpg b/parse/dev/k7p_YAO7yE/images/e76048854249de2e942a26492e96088624c2753b3e4c12a77019a1e4f57d41d0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cf72eb5a1a2c99b6397046336a81b60cbb4384b8 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/e76048854249de2e942a26492e96088624c2753b3e4c12a77019a1e4f57d41d0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d8222895696552bac46d98a38d49a0524d45d3ee796f83a1852ac881b39b7184 +size 13895 diff --git a/parse/dev/k7p_YAO7yE/images/ecdea5957c411dc6790c7029760c3d9b5c93f43eab0fd4435cff165f4ffadacb.jpg b/parse/dev/k7p_YAO7yE/images/ecdea5957c411dc6790c7029760c3d9b5c93f43eab0fd4435cff165f4ffadacb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..06c6ff85c88c851ffd57564cf30b69db7c950d91 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/ecdea5957c411dc6790c7029760c3d9b5c93f43eab0fd4435cff165f4ffadacb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9653e109c965990722c3e3246a9ee5ec777aebcc7ab7305f4a9355faaa763dc +size 41566 diff --git a/parse/dev/k7p_YAO7yE/images/f1578167bbba097268e739ea068a619a7b1f2191a0e2e0f3daf046f7fd15e469.jpg b/parse/dev/k7p_YAO7yE/images/f1578167bbba097268e739ea068a619a7b1f2191a0e2e0f3daf046f7fd15e469.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f3a93c983c5a06eb9dec3920afddca56929624ae --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/f1578167bbba097268e739ea068a619a7b1f2191a0e2e0f3daf046f7fd15e469.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9b856eef8bc77ea7bb8d00be597797300336645e658e7ecc3a12204700c3c66 +size 16118 diff --git a/parse/dev/k7p_YAO7yE/images/f895541606768acd1f8539a10b0c33319cc7b76e595c6de6f680cf673fd9fe7b.jpg b/parse/dev/k7p_YAO7yE/images/f895541606768acd1f8539a10b0c33319cc7b76e595c6de6f680cf673fd9fe7b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..886bdba25a3853e0c43a96625e64fbc1d5d03be9 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/f895541606768acd1f8539a10b0c33319cc7b76e595c6de6f680cf673fd9fe7b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d56b27c5684ecb7bc8fa767f8c1d5baf35eaa827a5e3f1134036c7cedd0bad98 +size 59033 diff --git a/parse/dev/k7p_YAO7yE/images/f983c16e84e6f9125bb8b0af0a679c6d077f63efa172dd1cf210ea8ade249fbe.jpg b/parse/dev/k7p_YAO7yE/images/f983c16e84e6f9125bb8b0af0a679c6d077f63efa172dd1cf210ea8ade249fbe.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ea429b67b2d96f8c4227a7fe466eb302ecdb4b3a --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/f983c16e84e6f9125bb8b0af0a679c6d077f63efa172dd1cf210ea8ade249fbe.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:80487c337e9222f843f4efa4e1f66d3f3dcd6ad2c15f4d64ca258a48a9273606 +size 20072 diff --git a/parse/dev/k7p_YAO7yE/images/ff09af545213e2b3c68e0eb6ae545c5ae816d0a04f09104ee2ed6b50ab990bfc.jpg b/parse/dev/k7p_YAO7yE/images/ff09af545213e2b3c68e0eb6ae545c5ae816d0a04f09104ee2ed6b50ab990bfc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..eb0b7e98319992c4967dcefadc136efd0dde11e7 --- /dev/null +++ b/parse/dev/k7p_YAO7yE/images/ff09af545213e2b3c68e0eb6ae545c5ae816d0a04f09104ee2ed6b50ab990bfc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b5fb4ae037c10b5ed9482d63c22a03fa27ee864f2edeeb10cc72097917c41f2e +size 2550 diff --git a/parse/dev/nO5caZwFwYu/images/02db29bd1a00a2d24674766123c6311de2c249ba381d50637f18cb95eaf93a9c.jpg b/parse/dev/nO5caZwFwYu/images/02db29bd1a00a2d24674766123c6311de2c249ba381d50637f18cb95eaf93a9c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..93b8e62d338cc3a88885b3c4e52ab12c4cb7cd99 --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/02db29bd1a00a2d24674766123c6311de2c249ba381d50637f18cb95eaf93a9c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:51365e727d7a0cd9872e5374c4e8f926c5c793c55e7bd5d3704be64aed06f26e +size 8156 diff --git a/parse/dev/nO5caZwFwYu/images/189abc5d6fb79dfe56b65ebb32b7d500676d329066d509c67e35215326a97182.jpg b/parse/dev/nO5caZwFwYu/images/189abc5d6fb79dfe56b65ebb32b7d500676d329066d509c67e35215326a97182.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b8c1943e6ffac4d99e6415ebe54650a9badf69a0 --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/189abc5d6fb79dfe56b65ebb32b7d500676d329066d509c67e35215326a97182.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04a2e956c2d475ecc1ddba4f9728325d552638432b6b3394535d98406829ba5a +size 6338 diff --git a/parse/dev/nO5caZwFwYu/images/2b70c77680f3caf606ae0f0a6c0624de6f9421152730ee2f175b081fef75c62a.jpg b/parse/dev/nO5caZwFwYu/images/2b70c77680f3caf606ae0f0a6c0624de6f9421152730ee2f175b081fef75c62a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..570fc7ba4987d9822d5c12873ed928460fe937ee --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/2b70c77680f3caf606ae0f0a6c0624de6f9421152730ee2f175b081fef75c62a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:934a43054384e044f67c22fd02051cd3738f8c1e58f4414257dc3b78644a86ee +size 99514 diff --git a/parse/dev/nO5caZwFwYu/images/3814f29ab52dc4112005668d02a29044fe9d34e4442175ae8284d12c9a420a11.jpg b/parse/dev/nO5caZwFwYu/images/3814f29ab52dc4112005668d02a29044fe9d34e4442175ae8284d12c9a420a11.jpg new file mode 100644 index 0000000000000000000000000000000000000000..244346cba8aa4a8f7b1dec26165373a9a89477b0 --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/3814f29ab52dc4112005668d02a29044fe9d34e4442175ae8284d12c9a420a11.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8de9d12b8eea06edb1ca3b6befd86da94c12bd2e3940ca28c88b18f48625b5e +size 17964 diff --git a/parse/dev/nO5caZwFwYu/images/3abde8df1e97587e52664ac4987fda3a29e2025873cd251133e478eda9740eb0.jpg b/parse/dev/nO5caZwFwYu/images/3abde8df1e97587e52664ac4987fda3a29e2025873cd251133e478eda9740eb0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9550c503c3b7dfd94ffe42ffa387d18751bfab6c --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/3abde8df1e97587e52664ac4987fda3a29e2025873cd251133e478eda9740eb0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b5e1c8319e7e367b85413b75e7205f08a916e932fbfbff08118a3dea4cad8478 +size 17311 diff --git a/parse/dev/nO5caZwFwYu/images/41ab96920d4ee9221016be5023bf0d3e37f416c536ccee2088f21bf8bca9448d.jpg b/parse/dev/nO5caZwFwYu/images/41ab96920d4ee9221016be5023bf0d3e37f416c536ccee2088f21bf8bca9448d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..056614583cb4a04d9c9efafd5f74031c3764c836 --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/41ab96920d4ee9221016be5023bf0d3e37f416c536ccee2088f21bf8bca9448d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b0d64e7023bd9a2908a2848560f7658d03f50210bcb29bb45e28bcc641b5ef42 +size 47987 diff --git a/parse/dev/nO5caZwFwYu/images/a4efe175befc0bb45d9375e239084557e813c70bf10e86d2d14bc5459fc391f8.jpg b/parse/dev/nO5caZwFwYu/images/a4efe175befc0bb45d9375e239084557e813c70bf10e86d2d14bc5459fc391f8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..773c76648e6df8a28a238b7680afebfb020fc387 --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/a4efe175befc0bb45d9375e239084557e813c70bf10e86d2d14bc5459fc391f8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb4e30392167905a65505f368ddf019522d360741d582246b42e4619729987cc +size 5377 diff --git a/parse/dev/nO5caZwFwYu/images/acaba43ae1df91044310bfda809b60ffddcc8cd8cff8313d303fd2d71b71174d.jpg b/parse/dev/nO5caZwFwYu/images/acaba43ae1df91044310bfda809b60ffddcc8cd8cff8313d303fd2d71b71174d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0319341bb63e69f7d142c43321a1e8f110350e7d --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/acaba43ae1df91044310bfda809b60ffddcc8cd8cff8313d303fd2d71b71174d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:979e86ae22e1a9e019eba1ccabb1715904323537424219202a472ea788a66cdd +size 111447 diff --git a/parse/dev/nO5caZwFwYu/images/aeff2be7dc644f08b6349a6f6cabfb88e1ed829fcc964cb99457f18da9f01b79.jpg b/parse/dev/nO5caZwFwYu/images/aeff2be7dc644f08b6349a6f6cabfb88e1ed829fcc964cb99457f18da9f01b79.jpg new file mode 100644 index 0000000000000000000000000000000000000000..114bc97e9878b277e2675343d3e237a50b273950 --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/aeff2be7dc644f08b6349a6f6cabfb88e1ed829fcc964cb99457f18da9f01b79.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a38d63c1122ed878a05d87866ed0b3c03d032a5e52d507a60caff1b6776affc9 +size 26886 diff --git a/parse/dev/nO5caZwFwYu/images/b850778457b82bd499ebcca9bc2866dca78cef43add4b0a344aaaf470b378364.jpg b/parse/dev/nO5caZwFwYu/images/b850778457b82bd499ebcca9bc2866dca78cef43add4b0a344aaaf470b378364.jpg new file mode 100644 index 0000000000000000000000000000000000000000..20a6602c86f929142b047a03ebd476b3745cd25b --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/b850778457b82bd499ebcca9bc2866dca78cef43add4b0a344aaaf470b378364.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca72f4bdddf41bda32e2a7d7e2a6de73aa7271280810a0f9a42d7389b807843c +size 124399 diff --git a/parse/dev/nO5caZwFwYu/images/c1609797b2d752b07426872754f67e6261c30974ae005d79b301a2100da2fcbd.jpg b/parse/dev/nO5caZwFwYu/images/c1609797b2d752b07426872754f67e6261c30974ae005d79b301a2100da2fcbd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a7d1cd554efc3582155fdceaeecf3949ec435610 --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/c1609797b2d752b07426872754f67e6261c30974ae005d79b301a2100da2fcbd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f5f62331f6c62e1e8fed0d75dc4c2640e6ef2e4532a5a8c0f305a20452f98fb +size 6384 diff --git a/parse/dev/nO5caZwFwYu/images/dafda474a5834db9ed417690205a5bb2e7098e1162c71613804c95c313b3d37d.jpg b/parse/dev/nO5caZwFwYu/images/dafda474a5834db9ed417690205a5bb2e7098e1162c71613804c95c313b3d37d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6f9493953eb54ad8367e9740b25f4ff8862bfbca --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/dafda474a5834db9ed417690205a5bb2e7098e1162c71613804c95c313b3d37d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a7fcc3471a8853e23cd3852ba608150dd52be4acf79523644763699724ecfc7 +size 10858 diff --git a/parse/dev/nO5caZwFwYu/images/f2ce96d91caec68f4b05fe8c69adad675170d3045007f55110ea4d8ab189f66e.jpg b/parse/dev/nO5caZwFwYu/images/f2ce96d91caec68f4b05fe8c69adad675170d3045007f55110ea4d8ab189f66e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..99c19f7c266b195dea898150904660b0aad5de31 --- /dev/null +++ b/parse/dev/nO5caZwFwYu/images/f2ce96d91caec68f4b05fe8c69adad675170d3045007f55110ea4d8ab189f66e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b6c48c94371a31db5835ef3029de727f54af8b7c8ac4f8d60300a4ff298454e +size 66653 diff --git a/parse/dev/pAq8iDy00Oa/images/03d35d80b1e4dd06307f77b1f1c7937f0dd947e6ecc6b7e663a522c84cd26b96.jpg b/parse/dev/pAq8iDy00Oa/images/03d35d80b1e4dd06307f77b1f1c7937f0dd947e6ecc6b7e663a522c84cd26b96.jpg new file mode 100644 index 0000000000000000000000000000000000000000..30299cf397f0971292c6a1a98469b006bee5f1a3 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/03d35d80b1e4dd06307f77b1f1c7937f0dd947e6ecc6b7e663a522c84cd26b96.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:912dd638aede465372daa56b2009dd670ca8d0e89d4a29d619fef72412b40697 +size 65290 diff --git a/parse/dev/pAq8iDy00Oa/images/09025cf267e4c39bfa104828e65ab8aede5afc3f3d03f2c3fd3bf1c41068728b.jpg b/parse/dev/pAq8iDy00Oa/images/09025cf267e4c39bfa104828e65ab8aede5afc3f3d03f2c3fd3bf1c41068728b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fe178188d221751ab2253b6a131960c03101dfff --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/09025cf267e4c39bfa104828e65ab8aede5afc3f3d03f2c3fd3bf1c41068728b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dde3fcca7d8996485c787ecb150016ea0c3c590231985e074f0ca0061656d7f4 +size 8690 diff --git a/parse/dev/pAq8iDy00Oa/images/0fea3f0b1c834c5b4b4ae348ff93b33cfccdb9dbd21d0c0a2e538f6d918805e8.jpg b/parse/dev/pAq8iDy00Oa/images/0fea3f0b1c834c5b4b4ae348ff93b33cfccdb9dbd21d0c0a2e538f6d918805e8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f1b0a582300de2777c48d2f6fc79e1b078e098bd --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/0fea3f0b1c834c5b4b4ae348ff93b33cfccdb9dbd21d0c0a2e538f6d918805e8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6911055604ce1623e0c664ccaec478e624c7c77af4b6c4b2de76d1b9ad6cdd0 +size 12443 diff --git a/parse/dev/pAq8iDy00Oa/images/14275080b9630cab0a79b337ce487bf5c56a8800c25b91e7436a93a35e03365b.jpg b/parse/dev/pAq8iDy00Oa/images/14275080b9630cab0a79b337ce487bf5c56a8800c25b91e7436a93a35e03365b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b13ea10096bdd1d44bff2089768d2f624edbf289 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/14275080b9630cab0a79b337ce487bf5c56a8800c25b91e7436a93a35e03365b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:870d51174560c981ddd46c910fb32222cd55e0a86d9ee9a97dda7c88823a1c6a +size 73756 diff --git a/parse/dev/pAq8iDy00Oa/images/26fff9cf3f8b3cbe2132c6afc7574ce5828125a6d0cf46bacdb4735033dcee23.jpg b/parse/dev/pAq8iDy00Oa/images/26fff9cf3f8b3cbe2132c6afc7574ce5828125a6d0cf46bacdb4735033dcee23.jpg new file mode 100644 index 0000000000000000000000000000000000000000..523d8ba5e609753efbae5a66c26759e6a761bb29 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/26fff9cf3f8b3cbe2132c6afc7574ce5828125a6d0cf46bacdb4735033dcee23.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27a791f5e9d68d303aec9448c4e4fc5132a68cdf847106bba6721b6dd43f6280 +size 55450 diff --git a/parse/dev/pAq8iDy00Oa/images/2886047d301a26689aa12405f936b64e0bdc88d6bde14f6482e0c67836010f48.jpg b/parse/dev/pAq8iDy00Oa/images/2886047d301a26689aa12405f936b64e0bdc88d6bde14f6482e0c67836010f48.jpg new file mode 100644 index 0000000000000000000000000000000000000000..abb9f304d93aac86467a62e89cfab6b6aeaf8faa --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/2886047d301a26689aa12405f936b64e0bdc88d6bde14f6482e0c67836010f48.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62c74368dcb814f0fbf6fd7adf42b46a75911e1acd35b156477750a3cb506de3 +size 12545 diff --git a/parse/dev/pAq8iDy00Oa/images/30d5ac3d6a0bdce8789602f93e565518709a5d0db800bd7284e3d4b473f0ea90.jpg b/parse/dev/pAq8iDy00Oa/images/30d5ac3d6a0bdce8789602f93e565518709a5d0db800bd7284e3d4b473f0ea90.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9fd67a053024065022c15b79beb4162e95c18505 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/30d5ac3d6a0bdce8789602f93e565518709a5d0db800bd7284e3d4b473f0ea90.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0c32c07cabd025e3fbe1acf1c789cb14e96519f86b6b1eacee6f60d50c256fc +size 7089 diff --git a/parse/dev/pAq8iDy00Oa/images/40049cd3bf69b08f2e944550764b41fe8012999330894e16e416fe7a0f6c1118.jpg b/parse/dev/pAq8iDy00Oa/images/40049cd3bf69b08f2e944550764b41fe8012999330894e16e416fe7a0f6c1118.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0c6aa17f1b75f776221a8b585ac8e3f9420987d4 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/40049cd3bf69b08f2e944550764b41fe8012999330894e16e416fe7a0f6c1118.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d437035f2221ad805b4622226ef0937dad77dfd03dad421ba2ec5c122ebf70b3 +size 60083 diff --git a/parse/dev/pAq8iDy00Oa/images/4074edb11d8f9ed6a61449eb6d5ebb240c9c9c1b280d521faa6cb1e5be2d08e4.jpg b/parse/dev/pAq8iDy00Oa/images/4074edb11d8f9ed6a61449eb6d5ebb240c9c9c1b280d521faa6cb1e5be2d08e4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..72c4c724732d90459299cbac9b204a4fc998f7d4 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/4074edb11d8f9ed6a61449eb6d5ebb240c9c9c1b280d521faa6cb1e5be2d08e4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31eb7cb0018265a2640f035860a9b71d62e383ffc49eb34f58324b4b5b25612b +size 5765 diff --git a/parse/dev/pAq8iDy00Oa/images/4bad61c4f761364eacc0505bfa644b9168fd41eb589910085ef29485532fa0ae.jpg b/parse/dev/pAq8iDy00Oa/images/4bad61c4f761364eacc0505bfa644b9168fd41eb589910085ef29485532fa0ae.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5426538cd833602661edd36338742545e9aeb9e3 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/4bad61c4f761364eacc0505bfa644b9168fd41eb589910085ef29485532fa0ae.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3252d9b1436a048dc0e1f3f312857ec867688462b557053fce8370f0a369ce91 +size 20747 diff --git a/parse/dev/pAq8iDy00Oa/images/5928aa6b55e3326ca857ef9256732bcd9a1fbbfc500b4ad44d60cad80a16343f.jpg b/parse/dev/pAq8iDy00Oa/images/5928aa6b55e3326ca857ef9256732bcd9a1fbbfc500b4ad44d60cad80a16343f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..54ac8c3f923922fd28f9508bb7c441d900c13400 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/5928aa6b55e3326ca857ef9256732bcd9a1fbbfc500b4ad44d60cad80a16343f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fef25f8b9d25882122047cb17c7087920977ae35a53cc139c1456e2540b3f8c1 +size 23774 diff --git a/parse/dev/pAq8iDy00Oa/images/618db94709f4a759acbb6494a20ac230fc02327e1cc9ac7db4658c7c84d58582.jpg b/parse/dev/pAq8iDy00Oa/images/618db94709f4a759acbb6494a20ac230fc02327e1cc9ac7db4658c7c84d58582.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5d2e7d05835403b47ab773de179885d1013ad16b --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/618db94709f4a759acbb6494a20ac230fc02327e1cc9ac7db4658c7c84d58582.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc07132cabdf7a397cc7b0b6bdfdd243db27950bab9e24bd9333f129a914b0e3 +size 131821 diff --git a/parse/dev/pAq8iDy00Oa/images/6ad0432691c01929e9a1ae20df9de17337ce8511c0c6b7cc69ac5bbe18e14640.jpg b/parse/dev/pAq8iDy00Oa/images/6ad0432691c01929e9a1ae20df9de17337ce8511c0c6b7cc69ac5bbe18e14640.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2bd339598bef74fb67e6780078e776d568bafab4 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/6ad0432691c01929e9a1ae20df9de17337ce8511c0c6b7cc69ac5bbe18e14640.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8158be2704d0e574d2faf90dac75e333580103cf99c592a322c7a4aed49b3571 +size 7580 diff --git a/parse/dev/pAq8iDy00Oa/images/76984f65fcd53b9ca34a3b0a5ba23faecd581e9d77a5c388a42dbb32e31ace79.jpg b/parse/dev/pAq8iDy00Oa/images/76984f65fcd53b9ca34a3b0a5ba23faecd581e9d77a5c388a42dbb32e31ace79.jpg new file mode 100644 index 0000000000000000000000000000000000000000..de729221c6d0d74b282feadfd420874372b1ad9f --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/76984f65fcd53b9ca34a3b0a5ba23faecd581e9d77a5c388a42dbb32e31ace79.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d6a9ed6129d8f5b28e31037e445ff7cff3a668e104180113c1e858520d454cc +size 3840 diff --git a/parse/dev/pAq8iDy00Oa/images/77c83f5b689fcb1ad9bf4bab06ddf1085b0bda9ce89d9f69f2d7a09d25c5778f.jpg b/parse/dev/pAq8iDy00Oa/images/77c83f5b689fcb1ad9bf4bab06ddf1085b0bda9ce89d9f69f2d7a09d25c5778f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..496e2d4e815865515a1c038e7198bf817562e7b2 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/77c83f5b689fcb1ad9bf4bab06ddf1085b0bda9ce89d9f69f2d7a09d25c5778f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:05c89be3319a1568f35f779d00444b74e8abd1a9a8e9d7a289556c613f29704c +size 47068 diff --git a/parse/dev/pAq8iDy00Oa/images/9b71653cbcbb4e6680dc410b20030ae5566b91cb0daba9f178fed0c0a5d37088.jpg b/parse/dev/pAq8iDy00Oa/images/9b71653cbcbb4e6680dc410b20030ae5566b91cb0daba9f178fed0c0a5d37088.jpg new file mode 100644 index 0000000000000000000000000000000000000000..468d1655f8ff15b795ec6117c4466bb9d69b14e1 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/9b71653cbcbb4e6680dc410b20030ae5566b91cb0daba9f178fed0c0a5d37088.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8d6af1d161d92c1f935988445e06b249d25caca84196415a100689c9cf50422 +size 36083 diff --git a/parse/dev/pAq8iDy00Oa/images/a16dd6bb7c66670a41589485f7987a2d31a7d00542f6a198083d9dc9af6ff61b.jpg b/parse/dev/pAq8iDy00Oa/images/a16dd6bb7c66670a41589485f7987a2d31a7d00542f6a198083d9dc9af6ff61b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a0c28759ed01f7786b6487602fbc172b6b25964c --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/a16dd6bb7c66670a41589485f7987a2d31a7d00542f6a198083d9dc9af6ff61b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91ef44f883690759966735863c085148fc30c7bf0c7cbcc7a4c0a4fffbb0588a +size 9472 diff --git a/parse/dev/pAq8iDy00Oa/images/a3799b77572211dbab5f22daf2443570d2294e5c245cd05d0063e93520e1b294.jpg b/parse/dev/pAq8iDy00Oa/images/a3799b77572211dbab5f22daf2443570d2294e5c245cd05d0063e93520e1b294.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b2abb0433d8c60c5492a3fe49dab5f841c02dd03 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/a3799b77572211dbab5f22daf2443570d2294e5c245cd05d0063e93520e1b294.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a2c3ea46d2b4b2317d05c346078d1e03bc2bd4ce4d622af130464e8c520ce2c9 +size 3623 diff --git a/parse/dev/pAq8iDy00Oa/images/a5291b3e3d5192b17474270feeea3950cd05598e3465afbbdc6268731f9928d5.jpg b/parse/dev/pAq8iDy00Oa/images/a5291b3e3d5192b17474270feeea3950cd05598e3465afbbdc6268731f9928d5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..afeb586714e0ce250f9eb348b701a04bf6d5d430 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/a5291b3e3d5192b17474270feeea3950cd05598e3465afbbdc6268731f9928d5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a16a8cc61d4ea19f609a79c0ef37ec6b235ae7173e49b92096e3fcec3c6b8046 +size 58016 diff --git a/parse/dev/pAq8iDy00Oa/images/b2c5d0300213ab617576b76c33aa8465887dcf521c2754cd198dbe02cbc48169.jpg b/parse/dev/pAq8iDy00Oa/images/b2c5d0300213ab617576b76c33aa8465887dcf521c2754cd198dbe02cbc48169.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4c576dc5a58d4ea817740cda2ff607e865698433 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/b2c5d0300213ab617576b76c33aa8465887dcf521c2754cd198dbe02cbc48169.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19c7ab88d81cfaa0e1dd25562925ca1edb23a95fd9d6e24ce2e47f91e7655668 +size 60505 diff --git a/parse/dev/pAq8iDy00Oa/images/b6c5ea8d6932bc6d55afb9d77fa43f74bfc60d6cdf7ebe6d467c7bd750a15e6f.jpg b/parse/dev/pAq8iDy00Oa/images/b6c5ea8d6932bc6d55afb9d77fa43f74bfc60d6cdf7ebe6d467c7bd750a15e6f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bf094775f0307ef89ddda017aa2fdc8957abd474 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/b6c5ea8d6932bc6d55afb9d77fa43f74bfc60d6cdf7ebe6d467c7bd750a15e6f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6aa0c327b720bde65a53cf8cc36d255bf1af8c9753b0b36f6f06a0731b8e3008 +size 58474 diff --git a/parse/dev/pAq8iDy00Oa/images/c8c7226314e7514453a2ed43835e1b9d04ae9a083f8ea2473e7a83d15d597d38.jpg b/parse/dev/pAq8iDy00Oa/images/c8c7226314e7514453a2ed43835e1b9d04ae9a083f8ea2473e7a83d15d597d38.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0c6981a5842d8eb2392556a1e3fad663764bcf9b --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/c8c7226314e7514453a2ed43835e1b9d04ae9a083f8ea2473e7a83d15d597d38.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df61fa73152f3beddb02e80e872171fad7e72053bc5b6db1fe48d592a3cb1d54 +size 24125 diff --git a/parse/dev/pAq8iDy00Oa/images/d5486479cb00234456f0199c246ff084893deb4f552b369f7ce0347a33f4c24f.jpg b/parse/dev/pAq8iDy00Oa/images/d5486479cb00234456f0199c246ff084893deb4f552b369f7ce0347a33f4c24f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c1385ffc42a240b679c27d234619bd5e83443474 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/d5486479cb00234456f0199c246ff084893deb4f552b369f7ce0347a33f4c24f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e71366b3f6c4c55be332a47ce5876554a31da3da3483b8debac375bb57fbc024 +size 46109 diff --git a/parse/dev/pAq8iDy00Oa/images/e52b810e0ba3bc043c1ba23a4e8f14175b4174059c34862ddcbfb65ddbf0dafc.jpg b/parse/dev/pAq8iDy00Oa/images/e52b810e0ba3bc043c1ba23a4e8f14175b4174059c34862ddcbfb65ddbf0dafc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..861e821825634cb2be5adfb7909734cdc8217961 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/e52b810e0ba3bc043c1ba23a4e8f14175b4174059c34862ddcbfb65ddbf0dafc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:edd9b047a2799919e1a1c1846b62676df6b71fc64f864801b44c7181431c7dad +size 51198 diff --git a/parse/dev/pAq8iDy00Oa/images/ec5809d1a21519078a4f0417fb6ef83f78a0a8d83aef331e9616f00c55b7c47e.jpg b/parse/dev/pAq8iDy00Oa/images/ec5809d1a21519078a4f0417fb6ef83f78a0a8d83aef331e9616f00c55b7c47e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..93727ae46c1b4444936d5d95cdf629e5916e7b8c --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/ec5809d1a21519078a4f0417fb6ef83f78a0a8d83aef331e9616f00c55b7c47e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f64c5367a353cf28e7e2888844a4242807d8c795c78323393c198cd2f174d85e +size 66992 diff --git a/parse/dev/pAq8iDy00Oa/images/f40fc49b87cdcbdfc81b6d609a91eb3109beb60191ec48a266a295c8423c9ba7.jpg b/parse/dev/pAq8iDy00Oa/images/f40fc49b87cdcbdfc81b6d609a91eb3109beb60191ec48a266a295c8423c9ba7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6568b3fb1338b0dd53353e66842ea4a7e41943fc --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/f40fc49b87cdcbdfc81b6d609a91eb3109beb60191ec48a266a295c8423c9ba7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce90cc5ed934439c63c749a131523d86c95da2ac76b6c00e340cb5b450ef66ab +size 37542 diff --git a/parse/dev/pAq8iDy00Oa/images/f53e045990bbf522e285fd3fe66df466f42b095964e4641ac8717badfdf751e3.jpg b/parse/dev/pAq8iDy00Oa/images/f53e045990bbf522e285fd3fe66df466f42b095964e4641ac8717badfdf751e3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cca4aa51ea960b4da6fdeb69508483313f35ec74 --- /dev/null +++ b/parse/dev/pAq8iDy00Oa/images/f53e045990bbf522e285fd3fe66df466f42b095964e4641ac8717badfdf751e3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1538dc70f245afd1b86bd1ac1be4db5c47957c90fea7479b37431d07f4cb3f2f +size 78008 diff --git a/parse/dev/pd1P2eUBVfq/pd1P2eUBVfq_content_list.json b/parse/dev/pd1P2eUBVfq/pd1P2eUBVfq_content_list.json new file mode 100644 index 0000000000000000000000000000000000000000..08a4ac8f7734f33a06adfe543fcd3105804f5045 --- /dev/null +++ b/parse/dev/pd1P2eUBVfq/pd1P2eUBVfq_content_list.json @@ -0,0 +1,3590 @@ +[ + { + "type": "text", + "text": "DIFFUSION MODELS ALREADY HAVE A SEMANTIC LATENT SPACE ", + "text_level": 1, + "bbox": [ + 174, + 130, + 616, + 176 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Mingi Kwon, Jaeseok Jeong, Youngjung Uh∗ ", + "bbox": [ + 186, + 200, + 493, + 215 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Department of Artificial Intelligence \nYonsei University \nSeoul, Republic of Korea \n{kwonmingi,jete jeong,yj.uh}@yonsei.ac.kr ", + "bbox": [ + 184, + 217, + 578, + 270 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "ABSTRACT ", + "text_level": 1, + "bbox": [ + 454, + 308, + 544, + 321 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Diffusion models achieve outstanding generative performance in various domains. Despite their great success, they lack semantic latent space which is essential for controlling the generative process. To address the problem, we propose asymmetric reverse process (Asyrp) which discovers the semantic latent space in frozen pretrained diffusion models. Our semantic latent space, named $h$ -space, has nice properties to accommodate semantic image manipulation: homogeneity, linearity, robustness, and consistency across timesteps. In addition, we introduce a principled design of the generative process for versatile editing and quality boosting by quantifiable measures: editing strength of an interval and quality deficiency at a timestep. Our method is applicable to various architectures $\\mathrm { ( D D P M + + }$ , iDDPM, and ADM) and datasets (CelebA-HQ, AFHQ-dog, LSUN-church, LSUNbedroom, and METFACES). Project page: https://kwonminki.github.io/Asyrp/ ", + "bbox": [ + 233, + 339, + 764, + 506 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "1 INTRODUCTION ", + "text_level": 1, + "bbox": [ + 176, + 534, + 334, + 549 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "In image synthesis, diffusion models have advanced to achieve state-of-the-art performance regarding quality and mode coverage since the introduction of denoising diffusion probabilistic models (Ho et al., 2020). They disrupt images by adding noise through multiple steps of forward process and generate samples by progressive denoising through multiple steps of reverse (i.e., generative) process. Since their deterministic version provides nearly perfect reconstruction of original images (Song et al., 2020a), they are suitable for image editing, which renders target attributes on the real images. However, simply editing the latent variables (i.e., intermediate noisy images) causes degraded results (Kim & Ye, 2021). Instead, they require complicated procedures: providing guidance in the reverse process or finetuning models for an attribute. ", + "bbox": [ + 174, + 565, + 825, + 690 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Figure 1(a-c) briefly illustrates the existing approaches. Image guidance mixes the latent variables of the guiding image with unconditional latent variables (Choi et al., 2021; Lugmayr et al., 2022; Meng et al., 2021). Though it provides some control, it is ambiguous to specify which attribute to reflect among the ones in the guide and the unconditional result, and it lacks intuitive control for the magnitude of change. Classifier guidance manipulates images by imposing gradients of a classifier on the latent variables in the reverse process to match the target class (Dhariwal & Nichol, 2021; Avrahami et al., 2022; Liu et al., 2021). It requires training an extra classifier for the latent variables, i.e., noisy images. Furthermore, computing gradients through the classifier during sampling is costly. Finetuning the whole model can steer the resulting images to the target attribute without the above problems (Kim & Ye, 2021). Still, it requires multiple models to reflect multiple descriptions. ", + "bbox": [ + 174, + 698, + 825, + 837 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "On the other hand, generative adversarial networks (Goodfellow et al., 2020) inherently provide straightforward image editing in their latent space. Given a latent vector for an original image, we can find the direction in the latent space that maximizes the similarity of the resulting image with a target description in CLIP embedding (Patashnik et al., 2021). The latent direction found on one image leads to the same manipulation of other images. However, given a real image, finding its exact latent vector is often challenging and produces unexpected appearance changes. ", + "bbox": [ + 174, + 844, + 823, + 900 + ], + "page_idx": 0 + }, + { + "type": "image", + "img_path": "images/80648f78ecfc7ad5b9897cecadcaf4711f94bdafc585f1c639190f153bb4ad3d.jpg", + "image_caption": [ + "Figure 1: Manipulation approaches for diffusion models. (a) Image guidance suffers ambiguity while controlling the generative process. (b) Classifier guidance requires an extra classifier, is hardly editable, degrades quality, or alters the content. (c) DiffusionCLIP requires fine-tuning the whole model. (d) Our method discovers a semantic latent space of a frozen diffusion model. " + ], + "image_footnote": [], + "bbox": [ + 191, + 102, + 810, + 261 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 345, + 820, + 375 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "It would allow admirable image editing if the diffusion models with the nearly perfect inversion property have such a semantic latent space. Preechakul et al. (2022) introduces an additional input to the reverse diffusion process: a latent vector from an original image embedded by an extra encoder. This latent vector contains the semantics to condition the process. However, it requires training from scratch and does not match with pretrained diffusion models. ", + "bbox": [ + 174, + 381, + 825, + 450 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "In this paper, we propose an asymmetric reverse process (Asyrp) which discovers the semantic latent space of a frozen diffusion model such that modifications in the space edits attributes of the original images. Our semantic latent space, named $h$ -space, has the properties necessary for editing applications as follows. The same shift in this space results in the same attribute change in all images. Linear changes in this space lead to linear changes in attributes. The changes do not degrade the quality of the resulting images. The changes throughout the timesteps are almost identical to each other for a desired attribute change. Figure 1(d) illustrates some of these properties and $\\ S 5 . 3$ provides detailed analyses. To the best of our knowledge, it is the first attempt to discover the semantic latent space in the frozen pretrained diffusion models. Spoiler alert: our semantic latent space is different from the intermediate latent variables in the diffusion process. Moreover, we introduce a principled design of the generative process for versatile editing and quality boosting by quantifiable measures: editing strength of an interval and quality deficiency at a timestep. Extensive experiments demonstrate that our method is generally applicable to various architectures $\\mathrm { ( D D P M + + }$ , iDDPM, and ADM) and datasets (CelebA-HQ, AFHQ-dog, LSUN-church, LSUN-bedroom, and METFACES). ", + "bbox": [ + 173, + 458, + 825, + 651 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "2 BACKGROUND ", + "text_level": 1, + "bbox": [ + 176, + 671, + 326, + 688 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "We briefly describe essential backgrounds. The rest of the related work is deferred to Appendix A. ", + "bbox": [ + 173, + 702, + 816, + 717 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "2.1 DENOISING DIFFUSION PROBABILITY MODEL (DDPM) ", + "text_level": 1, + "bbox": [ + 174, + 732, + 601, + 747 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "DDPM is a latent variable model that learns a data distribution by denoising noisy images (Ho et al., 2020). The forward process diffuses the data samples through Gaussian transitions parameterized with a Markov process: ", + "bbox": [ + 173, + 758, + 826, + 800 + ], + "page_idx": 1 + }, + { + "type": "equation", + "img_path": "images/035cdb05d9c2f6d50e510faab182ddb5c1138b8f9e837a09f7b8bf8817ac6c24.jpg", + "text": "$$\nq \\left( \\pmb { x } _ { t } \\mid \\pmb { \\dot { x } } _ { t - 1 } \\right) = \\mathcal { N } \\left( \\pmb { x } _ { t } ; \\sqrt { 1 - \\beta _ { t } } \\pmb { x } _ { t - 1 } , \\beta _ { t } \\mathbf { I } \\right) = \\mathcal { N } \\left( \\sqrt { \\frac { \\alpha _ { t } } { \\alpha _ { t - 1 } } } \\pmb { x } _ { t - 1 } , \\left( 1 - \\frac { \\alpha _ { t } } { \\alpha _ { t - 1 } } \\right) \\pmb { I } \\right) ,\n$$", + "text_format": "latex", + "bbox": [ + 217, + 799, + 779, + 834 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "where $\\{ \\beta _ { t } \\} _ { t = 1 } ^ { T }$ is the variance schedule and $\\begin{array} { r } { \\alpha _ { t } = \\prod _ { s = 1 } ^ { t } ( 1 - \\beta _ { s } ) } \\end{array}$ . Then the reverse process becomes $\\begin{array} { r } { p _ { \\theta } \\left( \\pmb { x } _ { 0 : T } \\right) : = p \\left( \\pmb { x } _ { T } \\right) \\prod _ { t = 1 } ^ { T } p _ { \\theta } \\left( \\pmb { x } _ { t - 1 } \\mid \\pmb { x } _ { t } \\right) } \\end{array}$ , starting from $\\mathbf { \\boldsymbol { x } } _ { T } \\sim \\mathcal { N } ( 0 , \\mathbf { I } )$ with noise predictor $\\epsilon _ { t } ^ { \\theta }$ : ", + "bbox": [ + 174, + 837, + 830, + 871 + ], + "page_idx": 1 + }, + { + "type": "equation", + "img_path": "images/d7055f0c4357778eb781035893f1f1562f0810b6f8ac46e9dc8ddb106fdd7fcd.jpg", + "text": "$$\n{ \\pmb x } _ { t - 1 } = \\frac { 1 } { \\sqrt { 1 - \\beta _ { t } } } \\left( { \\pmb x } _ { t } - \\frac { \\beta _ { t } } { \\sqrt { 1 - \\alpha _ { t } } } { \\pmb \\epsilon } _ { t } ^ { \\theta } \\left( { \\pmb x } _ { t } \\right) \\right) + \\sigma _ { t } { \\pmb z } _ { t } ,\n$$", + "text_format": "latex", + "bbox": [ + 325, + 872, + 671, + 907 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "where $\\boldsymbol { z } _ { t } \\sim \\mathcal { N } ( 0 , \\mathbf { I } )$ and $\\sigma _ { t } ^ { 2 }$ is a variance of the reverse process which is set to $\\sigma _ { t } ^ { 2 } = \\beta _ { t }$ by DDPM. ", + "bbox": [ + 174, + 909, + 813, + 925 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "2.2 DENOISING DIFFUSION IMPLICIT MODEL (DDIM) ", + "text_level": 1, + "bbox": [ + 173, + 102, + 566, + 118 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "DDIM redefines Eq. (1) as $\\begin{array} { r } { q _ { \\sigma } ( \\pmb { x } _ { t - 1 } | \\pmb { x } _ { t } , \\pmb { x } _ { 0 } ) = \\mathcal { N } ( \\sqrt { \\alpha _ { t - 1 } } \\pmb { x } _ { 0 } + \\sqrt { 1 - \\alpha _ { t - 1 } - \\sigma _ { t } ^ { 2 } } \\cdot \\frac { \\pmb { x } _ { t } - \\sqrt { \\alpha _ { t } } \\pmb { x } _ { 0 } } { \\sqrt { 1 - \\alpha _ { t } } } , \\sigma _ { t } ^ { 2 } I ) } \\end{array}$ which is a non-Markovian process (Song et al., 2020a). Accordingly, the reverse process becomes ", + "bbox": [ + 171, + 131, + 825, + 164 + ], + "page_idx": 2 + }, + { + "type": "equation", + "img_path": "images/ea9fa73c46f30bb8a42ee1c4981dde761fc0b59874a789de6e6aab9c0b57a615.jpg", + "text": "$$\n\\begin{array} { r } { x _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\underbrace { \\left( \\frac { x _ { t } - \\sqrt { 1 - \\alpha _ { t } } \\epsilon _ { t } ^ { \\theta } \\left( x _ { t } \\right) } { \\sqrt { \\alpha _ { t } } } \\right) } _ { \\mathrm { \\normalfont ~ \\ \" p r e d i c t e d ~ \\ b { x } _ 0 ~ \\ \" ~ } } + \\underbrace { \\sqrt { 1 - \\alpha _ { t - 1 } - \\sigma _ { t } ^ { 2 } } \\cdot \\epsilon _ { t } ^ { \\theta } \\left( x _ { t } \\right) } _ { \\mathrm { \\normalfont ~ \\ \" d i r e c t i o n ~ p o i n t i n g ~ t o ~ \\ b { x } _ t \\cdot \\ } } + \\underbrace { \\sigma _ { t } z _ { t } } _ { \\mathrm { \\normalfont ~ \\mathrm { r a n d o m ~ n o i s e } } } \\ , } \\end{array}\n$$", + "text_format": "latex", + "bbox": [ + 220, + 167, + 776, + 220 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "where $\\sigma _ { t } = \\eta \\sqrt { \\left( 1 - \\alpha _ { t - 1 } \\right) / \\left( 1 - \\alpha _ { t } \\right) } \\sqrt { 1 - \\alpha _ { t } / \\alpha _ { t - 1 } }$ . When $\\eta = 1$ for all $t$ , it becomes DDPM. As $\\eta = 0$ , the process becomes deterministic and guarantees nearly perfect inversion. ", + "bbox": [ + 169, + 229, + 825, + 260 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "2.3 IMAGE MANIPULATION WITH CLIP ", + "text_level": 1, + "bbox": [ + 176, + 275, + 459, + 290 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "CLIP learns multimodal embeddings with an image encoder $E _ { I }$ and a text encoder $E _ { T }$ whose similarity indicates semantic similarity between images and texts (Radford et al., 2021). Compared to directly minimizing the cosine distance between the edited image and the target description (Patashnik et al., 2021), directional loss with cosine distance achieves homogeneous editing without mode collapse (Gal et al., 2021): ", + "bbox": [ + 173, + 301, + 825, + 372 + ], + "page_idx": 2 + }, + { + "type": "equation", + "img_path": "images/96a8ada4679fa2db751231f877d8fd3f9c808e36bcb0b4a1c2debc540973182f.jpg", + "text": "$$\n\\mathcal { L } _ { \\mathrm { d i r e c t i o n } } \\left( { \\boldsymbol { x } } ^ { \\mathrm { e d i t } } , { \\boldsymbol { y } } ^ { \\mathrm { t a r g e t } } ; { \\boldsymbol { x } } ^ { \\mathrm { s o u r c e } } , { \\boldsymbol { y } } ^ { \\mathrm { s o u r c e } } \\right) : = 1 - \\frac { \\Delta I \\cdot \\Delta T } { \\| \\Delta I \\| \\| \\Delta T \\| } ,\n$$", + "text_format": "latex", + "bbox": [ + 292, + 375, + 702, + 409 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "where $\\Delta T = E _ { T } \\left( y ^ { \\mathrm { t a r g e t } } \\right) - E _ { T } \\left( y ^ { \\mathrm { s o u r c e } } \\right)$ and $\\Delta I = E _ { I } \\left( { \\pmb x } ^ { \\mathrm { e d i t } } \\right) - E _ { I } \\left( { \\pmb x } ^ { \\mathrm { s o u r c e } } \\right)$ for edited image $\\pmb { x } ^ { \\mathrm { e d i t } }$ , target description $y ^ { \\mathrm { t a r g e t } }$ , original image $\\pmb { x } ^ { \\mathrm { s o u r c e } }$ , and source description $y ^ { \\mathrm { s o u r c e } }$ . We use the prompts ‘smiling face’ and ‘face’ as the target and source descriptions for facial attribute smiling. ", + "bbox": [ + 173, + 412, + 825, + 460 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "3 DISCOVERING SEMANTIC LATENT SPACE IN DIFFUSION MODELS ", + "text_level": 1, + "bbox": [ + 176, + 479, + 736, + 496 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "This section explains why naive approaches do not work and proposes a new controllable reverse process. Then we describe the techniques for controlling the generative process. Throughout this paper, we use an abbreviated version of Eq. (3): ", + "bbox": [ + 173, + 510, + 825, + 553 + ], + "page_idx": 2 + }, + { + "type": "equation", + "img_path": "images/9a23d42adc0c7e01a7729c3269d06d5bf2a09e85d5cf922a14707fa83d940c35.jpg", + "text": "$$\n\\pmb { x } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) ) + \\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) ) + \\sigma _ { t } \\boldsymbol { z } _ { t } ,\n$$", + "text_format": "latex", + "bbox": [ + 330, + 556, + 666, + 577 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "where $\\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )$ denotes the predicted $\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }$ and $\\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )$ denotes the direction pointing to $\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }$ . We omit $\\sigma _ { t } z _ { t }$ for brevity, except when $\\eta \\neq 0$ . We further abbreviate $\\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )$ as $\\mathbf { P } _ { t }$ and $\\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )$ as $\\mathbf { D } _ { t }$ when the context clearly specifies the arguments. ", + "bbox": [ + 174, + 580, + 825, + 625 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "3.1 PROBLEM ", + "text_level": 1, + "bbox": [ + 174, + 640, + 284, + 655 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "We aim to allow semantic latent manipulation of images $\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }$ generated from $\\mathbf { \\nabla } _ { \\mathbf { x } _ { T } }$ given a pretrained and frozen diffusion model. The easiest idea to manipulate $\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }$ is simply updating $\\mathbf { \\nabla } _ { \\mathbf { \\mathcal { X } } \\mathcal { T } }$ to optimize the directional CLIP loss given text prompts with Eq. (4). However, it leads to distorted images or incorrect manipulation (Kim & Ye, 2021). ", + "bbox": [ + 173, + 666, + 825, + 723 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "An alternative approach is to shift the noise $\\epsilon _ { t } ^ { \\theta }$ predicted by the network at each sampling step. However, it does not achieve manipulating $\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }$ because the intermediate changes in $\\mathbf { P } _ { t }$ and $\\mathbf { D } _ { t }$ cancel out each other resulting in the same $p _ { \\theta } ( { \\pmb x } _ { 0 : T } )$ , similarly to destructive interference. ", + "bbox": [ + 173, + 728, + 823, + 772 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Theorem 1. Let $\\epsilon _ { t } ^ { \\theta }$ be a predicted noise during the original reverse process at $t$ and $\\tilde { \\epsilon } _ { t } ^ { \\theta }$ be its shifted counterpart. Then, $\\Delta \\pmb { x } _ { t } = \\tilde { \\pmb { x } } _ { t - 1 } - \\pmb { x } _ { t - 1 }$ is negligible where $\\begin{array} { r } { \\tilde { { \\pmb { x } } } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } { \\bf P } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\pmb { x } } _ { t } ) ) + } \\end{array}$ $\\mathbf { D } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )$ . I.e., the shifted terms of $\\tilde { \\epsilon } _ { t } ^ { \\theta }$ in $\\mathbf { P } _ { t }$ and $\\mathbf { D } _ { t }$ destruct each other in the reverse process. ", + "bbox": [ + 173, + 775, + 825, + 820 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Appendix C proves above theorem. Figure 13(a-b) shows that $\\tilde { \\mathbf { x } } _ { 0 }$ is almost identical to $\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }$ ", + "bbox": [ + 169, + 829, + 766, + 845 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "3.2 ASYMMETRIC REVERSE PROCESS ", + "text_level": 1, + "bbox": [ + 176, + 862, + 446, + 876 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "In order to break the interference, we propose a new controllable reverse process with asymmetry: ", + "bbox": [ + 171, + 887, + 813, + 902 + ], + "page_idx": 2 + }, + { + "type": "equation", + "img_path": "images/bdbf48fe4af9abaa50663aa224fdaed1e2c1a3d853a0ccc0232744adfa33948f.jpg", + "text": "$$\n\\pmb { x } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } \\big ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big ) + \\mathbf { D } _ { t } \\big ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big ) ,\n$$", + "text_format": "latex", + "bbox": [ + 354, + 906, + 642, + 926 + ], + "page_idx": 2 + }, + { + "type": "image", + "img_path": "images/06ae64f291010710a1b60bf1d063bfc7c5ce461052c3aeb0dadeab995109361f.jpg", + "image_caption": [ + "Figure 2: Generative process of Asyrp. The green box on the left illustrates Asyrp which only alters $\\mathbf { P } _ { t }$ while preserving $\\mathbf { D } _ { t }$ shared by DDIM. The right describes that Asyrp modifies the original reverse process toward the target attribute reflecting the change in $h$ -space. " + ], + "image_footnote": [], + "bbox": [ + 189, + 101, + 810, + 218 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "i.e., we modify only $\\mathbf { P } _ { t }$ by shifting $\\epsilon _ { t } ^ { \\theta }$ to $\\tilde { \\epsilon } _ { t } ^ { \\theta }$ while preserving $\\mathbf { D } _ { t }$ . Intuitively, it modifies the original reverse process according to $\\Delta \\epsilon _ { t } = \\widetilde { \\epsilon } _ { t } ^ { \\theta } - \\widetilde { \\epsilon } _ { t } ^ { \\theta }$ while it does not alter the direction toward $\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }$ so that ${ \\mathbf { \\mathcal { x } } } _ { t - 1 }$ follows the original flow $\\mathbf { D } _ { t }$ at each sampling step. Figure 2 illustrates the above intuition. ", + "bbox": [ + 173, + 301, + 826, + 347 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "As in Avrahami et al. (2022), we use the modified $\\mathbf { P } _ { t } ^ { \\mathrm { e d i t } }$ and the original $\\mathbf { P } _ { t } ^ { \\mathrm { s o u r c e } }$ as visual inputs for the directional CLIP loss in Eq. (4), and regularize the difference between the modified $\\mathbf { P } _ { t } ^ { \\mathrm { { e d i t } } }$ and the original $\\mathbf { P } _ { t } ^ { \\mathrm { s o u r c e } }$ . We find $\\begin{array} { r } { \\Delta \\epsilon = \\arg \\operatorname* { m i n } _ { \\Delta \\epsilon } \\mathbb { E } _ { t } \\mathcal { L } ^ { ( t ) } } \\end{array}$ where ", + "bbox": [ + 173, + 351, + 825, + 397 + ], + "page_idx": 3 + }, + { + "type": "equation", + "img_path": "images/bc3060069cd5b867facf9896b4c1a26fdf90a15368d85d880fd3d21e20e88014.jpg", + "text": "$$\n\\begin{array} { r } { \\mathcal { L } ^ { ( t ) } = \\lambda _ { \\mathrm { { C L I P } } } \\mathcal { L } _ { \\mathrm { d i r e c t i o n } } \\left( \\mathbf { P } _ { t } ^ { \\mathrm { { e d i t } } } , y ^ { \\mathrm { { r e f } } } ; \\mathbf { P } _ { t } ^ { \\mathrm { { s o u r c e } } } , y ^ { \\mathrm { { s o u r c e } } } \\right) + \\lambda _ { \\mathrm { { r e c o n } } } \\left| \\mathbf { P } _ { t } ^ { \\mathrm { { e d i t } } } - \\mathbf { P } _ { t } ^ { \\mathrm { { s o u r c e } } } \\right| } \\end{array}\n$$", + "text_format": "latex", + "bbox": [ + 238, + 401, + 758, + 421 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "Although $\\Delta \\epsilon$ indeed renders the attribute in the $\\pmb { x } _ { 0 } ^ { \\mathrm { e d i t } }$ , $\\epsilon$ -space lacks the necessary properties of the semantic latent space in diffusion models that will be described in the following. ", + "bbox": [ + 173, + 433, + 825, + 463 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "3.3 $h$ -space ", + "text_level": 1, + "bbox": [ + 174, + 479, + 266, + 493 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "Note that $\\epsilon _ { t } ^ { \\theta }$ is implemented as U-Net in all state-of-the-art diffusion models. We choose its bottleneck, the deepest feature maps $\\boldsymbol { h } _ { t }$ , to control $\\epsilon _ { t } ^ { \\theta }$ . By design, $h _ { t }$ has smaller spatial resolutions and high-level semantics than $\\epsilon _ { t } ^ { \\theta }$ . Accordingly, the sampling equation becomes ", + "bbox": [ + 174, + 505, + 825, + 549 + ], + "page_idx": 3 + }, + { + "type": "equation", + "img_path": "images/9ea68511b921a6efe10799fe6a19c64192152e029b8e4254229bf856122276cb.jpg", + "text": "$$\n\\pmb { x } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } | \\Delta \\pmb { h } _ { t } ) ) + \\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) ) + \\sigma _ { t } \\boldsymbol { z } _ { t } ,\n$$", + "text_format": "latex", + "bbox": [ + 312, + 565, + 683, + 585 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "where $\\epsilon _ { t } ^ { \\theta } ( \\boldsymbol { x } _ { t } | \\Delta h _ { t } )$ adds $\\Delta { h _ { t } }$ to the original feature maps $h _ { t }$ . The $\\Delta \\boldsymbol { h } _ { t }$ minimizing the same loss in Eq. (7) with $\\dot { \\mathbf { P } } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } | \\Delta \\pmb { h } _ { t } ) )$ instead of $\\mathbf { P } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( x _ { t } ) )$ successfully manipulates the attributes. ", + "bbox": [ + 176, + 588, + 821, + 617 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "We observe that $h$ -space in Asyrp has the following properties that others do not have. ", + "bbox": [ + 174, + 622, + 738, + 638 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "• The same $\\Delta h$ leads to the same effect on different samples. \n• Linearly scaling $\\Delta h$ controls the magnitude of attribute change, even with negative scales. \n• Adding multiple $\\Delta h$ manipulates the corresponding multiple attributes simultaneously. \n• $\\Delta h$ preserves the quality of the resulting images without degradation. \n• $\\Delta { h _ { t } }$ is roughly consistent across different timesteps $t$ . ", + "bbox": [ + 217, + 647, + 823, + 738 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "The above properties are demonstrated thoroughly in $\\ S 5 . 3$ . Appendix D.3 provides details of $h$ -space and suboptimal results from alternative choices. ", + "bbox": [ + 174, + 747, + 823, + 776 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "3.4 IMPLICIT NEURAL DIRECTIONS ", + "text_level": 1, + "bbox": [ + 176, + 792, + 429, + 808 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "Although $\\Delta h$ succeeds in manipulating images, directly optimizing $\\Delta \\boldsymbol { h } _ { t }$ on multiple timesteps requires many iterations of training with a carefully chosen learning rate and its scheduling. Instead, we define an implicit function $f _ { t } ( h _ { t } )$ which produces $\\Delta { h _ { t } }$ for given $h _ { t }$ and $t$ . $\\pmb { f } _ { t }$ is implemented as a small neural network with two $1 \\times 1$ convolutions concatenating timestep $t$ . See Appendix $\\mathrm { E }$ for the details. Accordingly, we optimize the same loss in Eq. (7) with $\\mathbf { P } _ { t } ^ { \\mathrm { e d i t } } = \\mathbf { \\bar { P } } _ { t } \\bigl ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } \\mathbf { \\bar { | } } \\mathbf { \\bar { \\mathbf { f } } } _ { t } ) \\bigr )$ . ", + "bbox": [ + 174, + 818, + 825, + 890 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "Learning $f _ { t }$ is more robust to learning rate settings and converges faster than learning every $\\Delta \\boldsymbol { h } _ { t }$ . In addition, as $\\pmb { f } _ { t }$ learns an implicit function for given timesteps and bottleneck features, it generalizes to unseen timesteps and bottleneck features. The generalization allows us to borrow the accelerated training scheme of DDIM defined on a subsequence $\\{ \\pmb { x } _ { \\tau _ { i } } \\} _ { \\forall i \\in [ 1 , S ] }$ where $\\{ \\tau _ { i } \\}$ is a subsequence of $[ 1 , . . . , T ]$ and $S \\textless T$ . Then, we can use the generative process with a custom subsequence $\\{ \\tilde { \\tau } _ { i } \\}$ with length $\\tilde { S } < T$ through normalization: $\\Delta \\tilde { h } _ { \\tilde { \\tau } } = f _ { \\tilde { \\tau } } ( h _ { \\tilde { \\tau } } ) { S } / { \\tilde { S } }$ . It preserves the amount of $\\sum \\Delta { { h } _ { t } }$ , $\\Delta \\tilde { h } _ { \\tilde { \\tau } } \\tilde { S } = \\Delta h _ { t } S$ . Therefore, we can use $f _ { t }$ trained on any subsequence for any length of the generative process. See Appendix F for details. We use $\\pmb { f } _ { t }$ to get $\\Delta \\boldsymbol { h } _ { t }$ for all experiments except Figure 6. ", + "bbox": [ + 173, + 895, + 825, + 924 + ], + "page_idx": 3 + }, + { + "type": "image", + "img_path": "images/47dbd6b534787f72c5d6825a64bc954aec392ef479c4aed91b47b808b2fa38d3.jpg", + "image_caption": [ + "Figure 3: Intuition for choosing the intervals for editing and quality boosting. We choose the intervals by quantifying two measures (top left inset). The editing strength of an interval $[ T , t ]$ measures its perceptual difference from $T$ until $t$ . We set $[ T , t ]$ to the interval with the smallest editing strength that synthesizes $\\mathbf { P } _ { t }$ close to $_ { \\textbf { \\em x } }$ , i.e., $\\mathrm { L P I P S } ( { \\pmb x } , { \\bf P } _ { t } ) = 0 . 3 3$ . Editing flexibility of an interval $[ t , 0 ]$ measures the potential amount of changes after $t$ . Quality deficiency at $t$ measures the amount of noise in $\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }$ . We set $[ t , 0 ]$ to handle large quality deficiency (i.e., $\\mathrm { L P I P S } ( \\pmb { x } , \\pmb { x } _ { t } ) = 1 . 2 )$ with small editing flexibility. " + ], + "image_footnote": [], + "bbox": [ + 189, + 107, + 802, + 268 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "", + "bbox": [ + 173, + 416, + 825, + 508 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "4 GENERATIVE PROCESS DESIGN ", + "text_level": 1, + "bbox": [ + 176, + 527, + 460, + 542 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "This section describes the entire editing process, which consists of three phases: editing with Asyrp, traditional denoising, and quality boosting. We design formulas to determine the length of each phase with quantifiable measures. ", + "bbox": [ + 174, + 558, + 825, + 601 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "4.1 EDITING PROCESS WITH ASYRP ", + "text_level": 1, + "bbox": [ + 176, + 617, + 434, + 632 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Diffusion models generate the high-level context in the early stage and imperceptible fine details in the later stage (Choi et al., 2022). Likewise, we modify the generative process in the early stage to achieve semantic changes. We refer to the early stage as the editing interval $[ T , t _ { \\mathrm { e d i t } } ]$ . ", + "bbox": [ + 174, + 643, + 825, + 685 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "$\\mathrm { L P I P S } ( { \\pmb x } , { \\pmb P } _ { T } )$ and $\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t } )$ calculate the perceptual distance between the original image and the predicted image at time steps $T$ and $t$ , respectively. Intuitively, the high-level content is already determined by the predicted terms at the respective timesteps and LPIPS measures the remaining component to be edited through the remaining reverse process. Consequently, we define editing strength of an interval $[ T , t ]$ : ", + "bbox": [ + 173, + 693, + 825, + 762 + ], + "page_idx": 4 + }, + { + "type": "equation", + "img_path": "images/1e13f8e22779ba133801ee9ef0228320ea6bf34efecc3a3ff44ec39caedf9159.jpg", + "text": "$$\n\\xi _ { t } = \\mathrm { L P I P S } ( x , \\mathbf { P } _ { T } ) - \\mathrm { L P I P S } ( x , \\mathbf { P } _ { t } )\n$$", + "text_format": "latex", + "bbox": [ + 370, + 768, + 627, + 785 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "indicating the perceptual change from timestep $T$ to $t$ in the original generative process. Figure 3 illustrates $\\mathrm { L P I P S } ( { \\pmb x } , \\cdot )$ for $\\mathbf { P } _ { t }$ and $\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }$ with examples and the inset depicts editing strength. The shorter editing interval has the lower $\\xi _ { t }$ , and the longer editing interval brings more changes to the resulting images. We seek the shortest editing interval which will bring enough distinguishable changes in the images in general. We empirically find that $t _ { \\mathrm { e d i t } }$ with $\\mathrm { L P I P S } ( \\mathbf { x } , \\mathbf { P } _ { t _ { \\mathrm { e d i t } } } ) = 0 . \\dot { 3 } 3$ builds the shortest editing interval with enough editing strength as $\\mathbf { P } _ { t _ { \\mathrm { e d i t } } }$ has nearly all visual attributes in $_ { \\textbf { \\em x } }$ . ", + "bbox": [ + 173, + 790, + 825, + 888 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "However, some attributes require more visual changes than others, e.g., pixar $>$ smile. For such attributes, we increase the editing strength $\\xi _ { t }$ by $\\delta = 0 . 3 3 d ( E _ { T } ( y _ { \\mathrm { s o u r c e } } ) , E _ { T } ( y _ { \\mathrm { t a r g e t } } ) )$ where ", + "bbox": [ + 173, + 895, + 823, + 924 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "$E _ { T } ( \\cdot )$ produces CLIP text embedding, $y _ { ( \\cdot ) }$ denotes the descriptions, and $d ( \\cdot , \\cdot )$ computes the cosine distance between the arguments. Choosing $t _ { \\mathrm { e d i t } }$ with $\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t _ { \\mathrm { e d i t } } } ) = 0 . 3 3 - \\delta$ expands the editing interval to a suitable length. It consistently produces good results in various settings. The supporting experiments are shown in Appendix G. ", + "bbox": [ + 174, + 103, + 825, + 161 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "4.2 QUALITY BOOSTING WITH STOCHASTIC NOISE INJECTION", + "text_level": 1, + "bbox": [ + 173, + 178, + 614, + 191 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Although DDIM achieves nearly perfect inversion by removing stochasticity $( \\eta = 0$ ), Karras et al. (2022) demonstrate that stochasticity improves image quality. Likewise, we inject stochastic noise in the boosting interval $[ t _ { \\mathrm { b o o s t } } , 0 ]$ . ", + "bbox": [ + 176, + 203, + 823, + 247 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Though the longer boosting interval would achieve higher quality, boosting over excessively long intervals would modify the content. Hence, we want to determine the shortest interval that shows enough quality boosting to guarantee minimal change in the content. We consider the noise in the image as the capacity for the quality boosting and define quality deficiency at $t$ $: \\gamma _ { t } = \\mathrm { L P I P S } ( \\pmb { x } , \\pmb { x } _ { t } )$ indicating the amount of noise in $\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }$ compared to the original image. We use $\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }$ instead of $\\mathbf { P } _ { t }$ because we consider the actual image rather than the semantics. Figure 3 inset depicts editing flexibility and quality deficiency. We empirically find that $t _ { \\mathrm { b o o s t } }$ with $\\gamma _ { t _ { \\mathrm { b o o s t } } } ~ = ~ 1 . 2$ achieves quality boosting with minimal content change. We confirmed that the editing strength of the intervals $[ t _ { \\mathrm { b o o s t } } , 0 ]$ is guaranteed to be less than 0.25. In Figure 3, after $t _ { \\mathrm { b o o s t } }$ , $\\mathrm { L P I P S } ( { \\pmb x } , { \\pmb x } _ { t } )$ sharply drops in the original generative process while $\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t } )$ changes little. Note that most of the quality degradation of the resulting images is caused by DDIM reverse process, not by Asyrp. We use this quality boosting for all experiments except ablation in Appendix $_ \\mathrm { H }$ . ", + "bbox": [ + 174, + 252, + 825, + 419 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "4.3 OVERALL PROCESS OF IMAGE EDITING ", + "text_level": 1, + "bbox": [ + 176, + 435, + 483, + 449 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Using $t _ { \\mathrm { e d i t } }$ and $t _ { \\mathrm { b o o s t } }$ determined by the above formulas, we modify the generative process of DDIM with ", + "bbox": [ + 174, + 460, + 826, + 488 + ], + "page_idx": 5 + }, + { + "type": "equation", + "img_path": "images/1fd23e676ae9069b2a174b8dbe503c18b765107d5f7690ab6bab96976aa085a0.jpg", + "text": "$$\np _ { \\theta } ^ { ( t ) } \\left( x _ { t - 1 } \\mid x _ { t } \\right) = \\left\\{ \\begin{array} { l l } { \\mathcal { N } \\left( \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } | f _ { t } ) ) + \\mathbf { D } _ { t } , \\sigma _ { t } ^ { 2 } I \\right) , \\eta = 0 } & { \\mathrm { ~ i f ~ } T \\geq t \\geq t _ { \\mathrm { e d i t } } } \\\\ { \\mathcal { N } \\left( \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) + \\mathbf { D } _ { t } , \\sigma _ { t } ^ { 2 } I \\right) , \\eta = 0 } & { \\mathrm { ~ i f ~ } t _ { \\mathrm { e d i t } } > t \\geq t _ { \\mathrm { b o o s t } } } \\\\ { \\mathcal { N } \\left( \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) + \\mathbf { D } _ { t } , \\sigma _ { t } ^ { 2 } I \\right) , \\eta = 1 } & { \\mathrm { ~ i f ~ } t _ { \\mathrm { b o o s t } } > t } \\end{array} \\right.\n$$", + "text_format": "latex", + "bbox": [ + 189, + 488, + 787, + 542 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "The visual overview and comprehensive algorithms of the entire process are in Appendix I. ", + "bbox": [ + 171, + 550, + 769, + 564 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "5 EXPERIMENTS ", + "text_level": 1, + "bbox": [ + 174, + 583, + 326, + 599 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "In this section, we show the effectiveness of semantic latent editing in $h$ -space with Asyrp on various attributes, datasets and architectures in $\\ S \\ S . 1$ . Moreover, we provide quantitative results including user study in $\\mathrm { ~ \\normalfont ~ \\ S ~ } 5 . 2$ . Lastly, we provide detailed analyses for the properties of the semantic latent space on $h$ -space and alternatives in $\\ S 5 . 3$ . ", + "bbox": [ + 173, + 614, + 825, + 670 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Implementation details. We implement our method on various settings: CelebA-HQ (Karras et al., 2018) and LSUN-bedroom/-church (Yu et al., 2015) on $\\mathrm { D D P M + + }$ (Song et al., 2020b) (Meng et al., 2021); AFHQ-dog (Choi et al., 2020) on iDDPM (Nichol & Dhariwal, 2021); and METFACES (Karras et al., 2020) on ADM with P2-weighting (Dhariwal & Nichol, 2021) (Choi et al., 2022). Please note that all models are official pretrained checkpoints and are kept frozen. Detailed settings including the coefficients for $\\lambda _ { \\mathrm { C L I P } }$ and $\\lambda _ { \\mathrm { { r e c o n } } }$ , and source/target descriptions can be found in Appendix J.1. We train $\\pmb { f } _ { t }$ with $S ~ = ~ 4 0$ for 1 epoch using 1000 samples. The real samples are randomly chosen from each dataset for in-domain-like attributes. For out-of-domainlike attributes, we randomly draw 1,000 latent variables $\\mathbf { \\boldsymbol { x } } _ { T } \\sim \\mathcal { N } ( \\mathbf { \\boldsymbol { 0 } } , I )$ . Details are described in Appendix J.2. Training takes about 20 minutes with three RTX 3090 GPUs. All the images in the figures are not used for training. We set $\\tilde { S } = 1 , 0 0 0$ for inference. The code is available at https://github.com/kwonminki/Asyrp official ", + "bbox": [ + 173, + 684, + 825, + 853 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "5.1 VERSATILITY OF $h$ -space WITH ASYRP ", + "text_level": 1, + "bbox": [ + 176, + 869, + 480, + 883 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Figure 4 shows the effectiveness of our method on various datasets and any existing U-Net based architectures. Our method can synthesize the attributes that are not even included in the training dataset, such as church $ \\{ \\begin{array} { r l } \\end{array} $ department, factory, and temple}. Even for dogs, our method synthesizes smiling Poodle and Yorkshire, the species that barely smile in the dataset. Figure 5 provides results for changing human faces to different identities, painting styles, and ancient primates. More result can be found in Appendix N. Versatility of our method is surprising because we do not alter the models but only shift the bottleneck feature maps in $h$ -space with Asyrp during inference. ", + "bbox": [ + 173, + 895, + 823, + 924 + ], + "page_idx": 5 + }, + { + "type": "image", + "img_path": "images/306d377572a44f6224fe73778efba99d8a0dfab3966d278640038c75118326f5.jpg", + "image_caption": [ + "Figure 4: Editing results of Asyrp on various datasets. We conduct experiments on CelebA-HQ, LSUN-church, METFACES, AFHQ-dog, and LSUN-bedroom. " + ], + "image_footnote": [], + "bbox": [ + 202, + 97, + 794, + 638 + ], + "page_idx": 6 + }, + { + "type": "image", + "img_path": "images/a2ba33a9e5710b55fca14f10f93b7fa652ad4c1a82d2bbb67e10456214757d19.jpg", + "image_caption": [ + "Figure 5: Editing results of Asyrp for unseen domains in CelebA-HQ dataset. " + ], + "image_footnote": [], + "bbox": [ + 272, + 688, + 727, + 761 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "", + "bbox": [ + 173, + 840, + 825, + 922 + ], + "page_idx": 6 + }, + { + "type": "table", + "img_path": "images/16a51330a7547a7d3496a7eba87648c05f50719972f0cfe8bf62f5409f30f898.jpg", + "table_caption": [ + "Table 1: User study with 80 participants. The details are described in $\\ S \\ K . 1$ " + ], + "table_footnote": [], + "table_body": "
CelebA-HQ in-domainCelebA-HQ unseen-domainLSUN-church
qualityattributeoverallqualityattributeoverallqualityattributediversityoverall
Asyrp (ours)98.36%88.13%94.92%71.56%59.84%63.13%73.19%71.81%87.50%76.81%
DiffusionCLIP1.64%11.88%5.08%28.44%40.16%36.88%26.81%28.19%12.50%23.19%
", + "bbox": [ + 174, + 101, + 823, + 147 + ], + "page_idx": 7 + }, + { + "type": "image", + "img_path": "images/66fd5aab2ef8fa4bdecea189af2d5eb233396464cc901dad958ae0cc0424c117.jpg", + "image_caption": [ + "Figure 6: Optimization for smiling on $\\pmb { h }$ -space and ϵ-space. (a) Optimizing $\\Delta { h _ { t } }$ for a sample with smiling results in natural editing while the change due to optimizing $\\Delta \\epsilon _ { t }$ is relatively small. (b) Applying $\\Delta { h _ { t } }$ from (a) to other samples yields the same attribute change while $\\Delta \\epsilon _ { t }$ distorts the images. The result of $\\Delta \\epsilon _ { t }$ is the best sample we could find. " + ], + "image_footnote": [], + "bbox": [ + 187, + 183, + 812, + 265 + ], + "page_idx": 7 + }, + { + "type": "image", + "img_path": "images/7bac52c03fecc44b4f929ac853df2555bd33134ce5593830fb81ee5e8c225924.jpg", + "image_caption": [ + "Figure 7: Linearity of $\\pmb { h }$ -space. Linear interpolation and extrapolation on $h$ -space lead to gradual changes even to the unseen intensity and directions. Right side shows the interpolation results by positive scaling of $\\Delta h _ { t } ^ { \\mathrm { s m i l i n g } }$ . Left side shows the extrapolation results by negative scaling of $\\Delta h _ { t } ^ { s m i l i n g }$ . Note that we did not train $\\pmb { f } _ { t }$ for the negative direction. " + ], + "image_footnote": [], + "bbox": [ + 186, + 344, + 812, + 438 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "5.2 QUANTITATIVE COMPARISON ", + "text_level": 1, + "bbox": [ + 176, + 545, + 416, + 559 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "Considering that our method can be combined with various diffusion models without finetuning, we do not find such a versatile competitor. Nonetheless, we compare Asyrp against DiffusionCLIP using the official code that edits the real images by finetuning the whole model. We asked 80 participants to choose the images with better quality, natural attribute change, and overall preference for given total of 40 sets of original images, ours, and DiffusionCLIP. Table 1 shows that Asyrp outperforms DiffusionCLIP in the all perspectives including the attributes unseen in the training dataset. We list the settings for fair comparison including the questions and example images in Appendix K.1. See $\\textrm { \\AA K } . 2$ for more evaluation metrics: segmentation consistency (SC) and directional CLIP similarity( $S _ { \\mathrm { d i r } } )$ . ", + "bbox": [ + 173, + 570, + 825, + 696 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "5.3 ANALYSIS ON $h$ -space ", + "text_level": 1, + "bbox": [ + 176, + 713, + 366, + 728 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "We provide detailed analyses to validate the properties of semantic latent space for diffusion models: homogeneity, linearity, robustness, and consistency across timesteps. ", + "bbox": [ + 173, + 739, + 821, + 767 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "Homogeneity. Figure 6 illustrates homogeneity of $h$ -space compared to $\\epsilon$ -space. One $\\Delta { h _ { t } }$ optimized for an image results in the same attribute change to other input images. On the other hand, one $\\Delta \\epsilon _ { t }$ optimized for an image distorts other input images. In Figure 10, applying $\\Delta h _ { t } ^ { \\mathrm { m e a n } } =$ $\\begin{array} { r } { \\frac { 1 } { N } \\sum \\Delta h _ { t } ^ { i } } \\end{array}$ produces almost identical results where $i$ indicates indices of $N = 2 0$ random samples. ", + "bbox": [ + 174, + 782, + 823, + 838 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "Linearity. In Figure 7, we observe that linearly scaling a $\\Delta h$ reflects the amount of change in the visual attributes. Surprisingly, it generalizes to negative scales that are not seen during training. Moreover, Figure 8 shows that combinations of different $\\Delta h$ ’s yield their combined semantic changes in the resulting images. Appendix N.2 provides mixed interpolation between multiple attributes. ", + "bbox": [ + 174, + 853, + 823, + 922 + ], + "page_idx": 7 + }, + { + "type": "image", + "img_path": "images/ec4e0c1c15035d0be274617d3f54f6a61905cce32ac51db848c37eaa6738a5c4.jpg", + "image_caption": [ + "Figure 8: Linear combination. Combining multiple ∆hs leads to combined attribute changes. " + ], + "image_footnote": [], + "bbox": [ + 187, + 101, + 810, + 175 + ], + "page_idx": 8 + }, + { + "type": "image", + "img_path": "images/f16b3b6406f8475789e2643d74a9891b20a10d09c81ef00ec5fdfc72bd1f3e33.jpg", + "image_caption": [ + "Figure 9: Ramdom manipulation on $\\pmb { h }$ -space and $\\epsilon$ -space. (a) Adding random noise with magnitude of ∆hsmilingt and random direction in $h$ -space leads to realistic images with small changes. (b) Adding random noise with magnitude of $\\Delta \\epsilon _ { t } ^ { \\mathrm { s m i l i n g } }$ and random direction in $\\epsilon$ -space leads to severely distorted images. (c) Adding random noises with tripled magnitude produce diverse visual changes in realistic images. " + ], + "image_footnote": [], + "bbox": [ + 187, + 213, + 807, + 290 + ], + "page_idx": 8 + }, + { + "type": "image", + "img_path": "images/516dc2690d986f75d0a76feebed8568de3bb5e07882b27922278b8419f1f3e97.jpg", + "image_caption": [ + "Figure 10: Consistency on $\\pmb { h }$ -space. Results of $\\Delta { h _ { t } }$ , $\\Delta h _ { t } ^ { \\mathrm { m e a n } }$ and $\\Delta h ^ { \\mathrm { g l o b a l } }$ are almost identical for in-domain samples. However, we choose $\\Delta { h _ { t } }$ over others to prevent unexpected small difference in the unseen domain. We provide more results in $\\ S \\mathrm { { L . 1 } }$ . " + ], + "image_footnote": [], + "bbox": [ + 187, + 375, + 807, + 464 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "Robustness. Figure 9 compares the effect of adding random noise in $h$ -space and $\\epsilon$ -space. The random noises are chosen to be the vectors with random directions and magnitude of the example $\\Delta { h _ { t } }$ and $\\Delta \\epsilon _ { t }$ in Figure 6 on each space. Perturbation in $h$ -space leads to realistic images with a minimal difference or some semantic changes. On the contrary, perturbation in $\\epsilon$ -space severely distorts the resulting images. See Appendix D.2 for more analyses. ", + "bbox": [ + 174, + 530, + 825, + 599 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "Consistency across timesteps. Recall that $\\Delta \\boldsymbol { h } _ { t }$ for all samples are homogeneous and replacing them by their mea time-invariant $\\Delta h _ { t } ^ { \\mathrm { m e a n } }$ ults. Interesinstead of , in Figure 10, we observe that addalso yields similar results where $\\Delta { h ^ { \\mathrm { g l o b a l } } } ~ \\dot { = } ~ \\frac { 1 } { T _ { e } } \\sum _ { t } \\Delta { h _ { t } ^ { \\mathrm { m e a n } } }$ $\\Delta \\boldsymbol { h } _ { t }$ $T _ { e }$ denotes the length of the editing interval $[ T , t _ { \\mathrm { e d i t } } ]$ . Though we use $\\Delta \\boldsymbol { h } _ { t }$ to deliver the best quality and manipulation, using $\\Delta h _ { t } ^ { \\mathrm { m e a n } }$ or even $\\Delta h ^ { \\mathrm { g l o b a l } }$ with some compromise would be worth trying for simplicity. We report more detail about mean and global direction in Appendix L.1. ", + "bbox": [ + 174, + 616, + 825, + 703 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "6 CONCLUSION ", + "text_level": 1, + "bbox": [ + 174, + 724, + 318, + 739 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "We proposed a new generative process, Asyrp, which facilitates image editing in a semantic latent space $h$ -space for pretrained diffusion models. $h$ -space has nice properties as in the latent space of GANs: homogeneity, linearity, robustness, and consistency across timesteps. The full editing process is designed to achieve versatile editing and high quality by measuring editing strength and quality deficiency at timesteps. We hope that our approach and detailed analyses help cultivate a new paradigm of image editing in the semantic latent space of diffusion models. Combining previous finetuning or guidance techniques would be an interesting research direction. ", + "bbox": [ + 174, + 756, + 825, + 854 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "ACKNOWLEDGMENTS ", + "text_level": 1, + "bbox": [ + 176, + 872, + 326, + 885 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (Ministry of Science and ICT) (No. 2021-0-00155) ", + "bbox": [ + 176, + 895, + 823, + 924 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "REFERENCES ", + "text_level": 1, + "bbox": [ + 174, + 103, + 287, + 117 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Omri Avrahami, Dani Lischinski, and Ohad Fried. Blended diffusion for text-driven editing of natural images. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 18208–18218, 2022. ", + "bbox": [ + 174, + 126, + 823, + 167 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Hila Chefer, Sagie Benaim, Roni Paiss, and Lior Wolf. Image-based clip-guided essence transfer. arXiv preprint arXiv:2110.12427, 2021. ", + "bbox": [ + 171, + 176, + 821, + 205 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Jooyoung Choi, Sungwon Kim, Yonghyun Jeong, Youngjune Gwon, and Sungroh Yoon. Ilvr: Conditioning method for denoising diffusion probabilistic models. arXiv preprint arXiv:2108.02938, 2021. ", + "bbox": [ + 173, + 213, + 823, + 256 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Jooyoung Choi, Jungbeom Lee, Chaehun Shin, Sungwon Kim, Hyunwoo Kim, and Sungroh Yoon. Perception prioritized training of diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11472–11481, 2022. ", + "bbox": [ + 173, + 265, + 825, + 308 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Yunjey Choi, Youngjung Uh, Jaejun Yoo, and Jung-Woo Ha. Stargan v2: Diverse image synthesis for multiple domains. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 8188–8197, 2020. ", + "bbox": [ + 176, + 315, + 823, + 358 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Ning Dai, Jianze Liang, Xipeng Qiu, and Xuanjing Huang. Style transformer: Unpaired text style transfer without disentangled latent representation. arXiv preprint arXiv:1905.05621, 2019. ", + "bbox": [ + 171, + 366, + 823, + 396 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in Neural Information Processing Systems, 34:8780–8794, 2021. ", + "bbox": [ + 171, + 404, + 823, + 433 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Rinon Gal, Or Patashnik, Haggai Maron, Gal Chechik, and Daniel Cohen-Or. Stylegan-nada: Clipguided domain adaptation of image generators. arXiv preprint arXiv:2108.00946, 2021. ", + "bbox": [ + 171, + 440, + 823, + 470 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial networks. Communications of the ACM, 63(11):139–144, 2020. ", + "bbox": [ + 176, + 477, + 823, + 520 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Erik Hark ¨ onen, Aaron Hertzmann, Jaakko Lehtinen, and Sylvain Paris. Ganspace: Discovering ¨ interpretable gan controls. Advances in Neural Information Processing Systems, 33:9841–9850, 2020. ", + "bbox": [ + 176, + 529, + 825, + 571 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in Neural Information Processing Systems, 33:6840–6851, 2020. ", + "bbox": [ + 169, + 579, + 823, + 609 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Tero Karras, Timo Aila, Samuli Laine, and Jaakko Lehtinen. Progressive growing of gans for improved quality, stability, and variation. In International Conference on Learning Representations, 2018. ", + "bbox": [ + 176, + 616, + 825, + 659 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine, Jaakko Lehtinen, and Timo Aila. Training generative adversarial networks with limited data. Advances in Neural Information Processing Systems, 33:12104–12114, 2020. ", + "bbox": [ + 176, + 667, + 823, + 710 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusionbased generative models. arXiv preprint arXiv:2206.00364, 2022. ", + "bbox": [ + 174, + 718, + 823, + 748 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Gwanghyun Kim and Jong Chul Ye. Diffusionclip: Text-guided image manipulation using diffusion models. 2021. ", + "bbox": [ + 173, + 756, + 823, + 785 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Diederik $\\mathrm { \\bf P }$ Kingma and Max Welling. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114, 2013. ", + "bbox": [ + 171, + 792, + 825, + 821 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Cheng-Han Lee, Ziwei Liu, Lingyun Wu, and Ping Luo. Maskgan: Towards diverse and interactive facial image manipulation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 5549–5558, 2020. ", + "bbox": [ + 174, + 830, + 823, + 873 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Huan Ling, Karsten Kreis, Daiqing Li, Seung Wook Kim, Antonio Torralba, and Sanja Fidler. Editgan: High-precision semantic image editing. Advances in Neural Information Processing Systems, 34:16331–16345, 2021. ", + "bbox": [ + 174, + 882, + 825, + 924 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Xihui Liu, Dong Huk Park, Samaneh Azadi, Gong Zhang, Arman Chopikyan, Yuxiao Hu, Humphrey Shi, Anna Rohrbach, and Trevor Darrell. More control for free! image synthesis with semantic diffusion guidance. arXiv preprint arXiv:2112.05744, 2021. ", + "bbox": [ + 176, + 103, + 821, + 146 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Ziwei Liu, Ping Luo, Xiaogang Wang, and Xiaoou Tang. Deep learning face attributes in the wild. In Proceedings of International Conference on Computer Vision (ICCV), December 2015. ", + "bbox": [ + 171, + 154, + 821, + 184 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Andreas Lugmayr, Martin Danelljan, Andres Romero, Fisher Yu, Radu Timofte, and Luc Van Gool. Repaint: Inpainting using denoising diffusion probabilistic models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11461–11471, 2022. ", + "bbox": [ + 176, + 191, + 823, + 234 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Chenlin Meng, Yang Song, Jiaming Song, Jiajun Wu, Jun-Yan Zhu, and Stefano Ermon. Sdedit: Image synthesis and editing with stochastic differential equations. arXiv preprint arXiv:2108.01073, 2021. ", + "bbox": [ + 174, + 242, + 825, + 285 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew, Ilya Sutskever, and Mark Chen. Glide: Towards photorealistic image generation and editing with text-guided diffusion models. arXiv preprint arXiv:2112.10741, 2021. ", + "bbox": [ + 173, + 294, + 823, + 337 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models. In International Conference on Machine Learning, pp. 8162–8171. PMLR, 2021. ", + "bbox": [ + 173, + 344, + 820, + 375 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Or Patashnik, Zongze Wu, Eli Shechtman, Daniel Cohen-Or, and Dani Lischinski. Styleclip: Textdriven manipulation of stylegan imagery. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 2085–2094, 2021. ", + "bbox": [ + 173, + 382, + 825, + 426 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Konpat Preechakul, Nattanat Chatthee, Suttisak Wizadwongsa, and Supasorn Suwajanakorn. Diffusion autoencoders: Toward a meaningful and decodable representation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 10619–10629, 2022. ", + "bbox": [ + 174, + 434, + 825, + 477 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International Conference on Machine Learning, pp. 8748–8763. PMLR, 2021. ", + "bbox": [ + 174, + 484, + 825, + 541 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bjorn Ommer. High- ¨ resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 10684–10695, 2022. ", + "bbox": [ + 173, + 550, + 821, + 594 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "David E Rumelhart, Geoffrey E Hinton, and Ronald J Williams. Learning internal representations by error propagation. Technical report, California Univ San Diego La Jolla Inst for Cognitive Science, 1985. ", + "bbox": [ + 174, + 601, + 821, + 643 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Vikash Sehwag, Caner Hazirbas, Albert Gordo, Firat Ozgenel, and Cristian Canton. Generating high fidelity data from low-density regions using diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11492–11501, 2022. ", + "bbox": [ + 176, + 652, + 823, + 696 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Yujun Shen, Ceyuan Yang, Xiaoou Tang, and Bolei Zhou. Interfacegan: Interpreting the disentangled face representation learned by gans. IEEE transactions on pattern analysis and machine intelligence, 2020. ", + "bbox": [ + 176, + 704, + 823, + 747 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. In International Conference on Machine Learning, pp. 2256–2265. PMLR, 2015. ", + "bbox": [ + 174, + 755, + 823, + 797 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502, 2020a. ", + "bbox": [ + 169, + 806, + 823, + 835 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456, 2020b. ", + "bbox": [ + 173, + 843, + 823, + 886 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Arash Vahdat, Karsten Kreis, and Jan Kautz. Score-based generative modeling in latent space. Advances in Neural Information Processing Systems, 34:11287–11302, 2021. ", + "bbox": [ + 171, + 895, + 820, + 924 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Daniel Watson, William Chan, Jonathan Ho, and Mohammad Norouzi. Learning fast samplers for diffusion models by differentiating through sample quality. In International Conference on Learning Representations, 2022. URL https://openreview.net/forum?id $\\underline { { \\underline { { \\mathbf { \\Pi } } } } }$ VFBjuF8HEp. ", + "bbox": [ + 176, + 103, + 821, + 146 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Zipeng Xu, Tianwei Lin, Hao Tang, Fu Li, Dongliang He, Nicu Sebe, Radu Timofte, Luc Van Gool, and Errui Ding. Predict, prevent, and evaluate: Disentangled text-driven image manipulation empowered by pre-trained vision-language model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 18229–18238, 2022. ", + "bbox": [ + 173, + 155, + 825, + 212 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Changqian Yu, Jingbo Wang, Chao Peng, Changxin Gao, Gang Yu, and Nong Sang. Bisenet: Bilateral segmentation network for real-time semantic segmentation. In Proceedings of the European conference on computer vision (ECCV), pp. 325–341, 2018. ", + "bbox": [ + 174, + 219, + 821, + 263 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Fisher Yu, Ari Seff, Yinda Zhang, Shuran Song, Thomas Funkhouser, and Jianxiong Xiao. Lsun: Construction of a large-scale image dataset using deep learning with humans in the loop. arXiv preprint arXiv:1506.03365, 2015. ", + "bbox": [ + 173, + 271, + 826, + 314 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Oguz Kaan Y ˘ uksel, Enis Simsar, Ezgi G ¨ ulperi Er, and Pinar Yanardag. Latentclr: A contrastive ¨ learning approach for unsupervised discovery of interpretable directions. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 14263–14272, 2021. ", + "bbox": [ + 173, + 323, + 823, + 366 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Bolei Zhou, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Scene parsing through ade20k dataset. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 633–641, 2017. ", + "bbox": [ + 173, + 375, + 823, + 417 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Bolei Zhou, Hang Zhao, Xavier Puig, Tete Xiao, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Semantic understanding of scenes through the ade20k dataset. International Journal of Computer Vision, 127(3):302–321, 2019. ", + "bbox": [ + 174, + 426, + 821, + 469 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Appendix ", + "text_level": 1, + "bbox": [ + 176, + 131, + 316, + 159 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "Table of Contents ", + "text_level": 1, + "bbox": [ + 174, + 183, + 352, + 202 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "A Related work 14 \nB More discussion 1 5 \nC Proof of Theorem 1 1 5 \nD Additional supports for $\\pmb { h }$ -space with Asyrp 17 \nD.1 Random perturbation on $\\epsilon$ -space without Asyrp . 17 \nD.2 Robustness and semantics in $h$ -space and $\\epsilon$ -space with Asyrp 17 \nD.3 Choice of $h$ -space in U-Net 17 \nE Implicit neural directions 17 \n$\\mathbf { F }$ Quality improvements by non-accelerated sampling with scaled $\\Delta h _ { t }$ 18 \nG Editing strength and editing flexibility 18 \nG.1 Editing strength and $t _ { e d i t }$ 18 \nG.2 Editing flexibility and $t _ { \\mathrm { b o o s t } }$ . 18 ", + "bbox": [ + 210, + 208, + 789, + 458 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "H Quality boosting 1 9 ", + "bbox": [ + 209, + 470, + 784, + 484 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "I Algorithm 2 3 ", + "text_level": 1, + "bbox": [ + 205, + 496, + 784, + 511 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "J Training details 2 3 ", + "text_level": 1, + "bbox": [ + 210, + 523, + 787, + 536 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "J.1 Loss coefficients . 23 \nJ.2 Training with random sampling instead of the training datasets . 24 ", + "bbox": [ + 228, + 537, + 785, + 568 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "K Evaluation ", + "text_level": 1, + "bbox": [ + 212, + 579, + 307, + 592 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "2 4 ", + "text_level": 1, + "bbox": [ + 212, + 582, + 789, + 592 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "K.1 User study . 24 \nK.2 Segmentation consistency and directional CLIP similarity 25 ", + "bbox": [ + 233, + 593, + 787, + 625 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "L Directions ", + "text_level": 1, + "bbox": [ + 212, + 636, + 302, + 648 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "26 ", + "text_level": 1, + "bbox": [ + 223, + 637, + 787, + 648 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "L.1 Global direction 26 \nL.2 Compare three methods 29 ", + "bbox": [ + 230, + 651, + 785, + 680 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "M Random sampling ", + "text_level": 1, + "bbox": [ + 212, + 693, + 352, + 707 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "3 0 ", + "bbox": [ + 766, + 693, + 785, + 705 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "N More samples ", + "text_level": 1, + "bbox": [ + 214, + 719, + 326, + 732 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "", + "text_level": 1, + "bbox": [ + 215, + 722, + 789, + 731 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "N.1 ImageNet 32 \nN.2 Multi-interpolation 32 \nN.3 More results on all datasets 32 ", + "bbox": [ + 232, + 734, + 789, + 779 + ], + "page_idx": 12 + }, + { + "type": "image", + "img_path": "images/ccb938a111ed5b23226f87f2c9601955a9f9a495d880770b80c3a6c0651db3e8.jpg", + "image_caption": [ + "Figure 11: Editing dog to smile. We provide high-resolution results of editing real images in the teaser and more. " + ], + "image_footnote": [], + "bbox": [ + 178, + 121, + 825, + 363 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "A RELATED WORK ", + "text_level": 1, + "bbox": [ + 176, + 460, + 344, + 477 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "After Sohl-Dickstein et al. (2015), denoising diffusion probabilistic models (DDPMs) provide a universal approach for generative modeling (Ho et al., 2020). On the other hand, Song et al. (2020b) suggests score-based model and unifies SDEs incorporating diffusion models with score-based models. Subsequent works renovate diffusion models by focusing on architectures, scheduling, weighting, and fast sampling (Nichol & Dhariwal (2021), Karras et al. (2022), Choi et al. (2022), Song et al. (2020a), Watson et al. (2022)). They mainly consider random generation rather than controlled generation. ", + "bbox": [ + 174, + 491, + 825, + 589 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "In the meantime, Dhariwal & Nichol (2021) introduces classifier guidance not only improving the quality of images but also retrieving specific class of images. Since it can apply any guidance, its variants have emerged (Sehwag et al. (2022), Avrahami et al. (2022), Liu et al. (2021), Nichol et al. (2021)). However it requires a noise-dependent classifier (or any off-the-shelf models) and additional cost to compute gradients for the guidance during its sampling process. The other works try to control the generative process using image-space guidance (Choi et al. (2021), Meng et al. (2021), Lugmayr et al. (2022), Avrahami et al. (2022)). They manipulate resulting images by matching noisy images with target images during the reverse process. Still, it is hard to expect delicate control of the reverse process from the image guidance. Furthermore, Preechakul et al. (2022) introduces an extra encoder which encodes the semantic features of a real image in order to condition the generative process. Although the semantics allow one to control diffusion models, it requires additional training from scratch with the encoder and inherently can not use the other pretrained diffusion models. ", + "bbox": [ + 174, + 595, + 825, + 762 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "For controllability, Rombach et al. (2022) and Vahdat et al. (2021) apply another approach which adapts VAE (Kingma & Welling, 2013) and autoencoder (Rumelhart et al., 1985) to diffusion models. In spite of their great success in editing, their diffusion models learn the distribution of the learned embeddings in VAE or autoencoder, not the images. Kim & Ye (2021) proposes another strategy: fine-tuning a whole diffusion model for image editing. It shows valid performance but it requires each fine-tuned model corresponding each attribute. ", + "bbox": [ + 174, + 770, + 825, + 853 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "In comparison, Asyrp enables outstanding manipulation without high computation, specifically designed architectures, or fine-tuning whole models. ", + "bbox": [ + 174, + 859, + 820, + 888 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "Meanwhile, generative adversarial network Goodfellow et al. (2020) address their latent space for image editing (Ling et al. (2021), Hark ¨ onen et al. (2020), Chefer et al. (2021), Shen et al. (2020), ¨ ", + "bbox": [ + 176, + 895, + 820, + 924 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "Yuksel et al. (2021), Patashnik et al. (2021), Gal et al. (2021), Dai et al. (2019), Xu et al. (2022)). ¨ However they have to conduct ‘inversion’ to their latent space for real image editing and ‘GAN inversion’ is often challenging and produces unexpected appearance changes. ", + "bbox": [ + 176, + 103, + 823, + 146 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "On the contrary, Asyrp enables to use latent space of real images by nearly perfect easy inversion of DDIM. ", + "bbox": [ + 173, + 152, + 823, + 181 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "B MORE DISCUSSION ", + "text_level": 1, + "bbox": [ + 176, + 210, + 367, + 227 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "In this section, we discuss the pros and cons of diffusion-model-based and GAN-based methods. \nAnd we provide guidelines for further improvements. ", + "bbox": [ + 176, + 246, + 821, + 275 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "GAN-based latent manipulation methods Patashnik et al. (2021); Gal et al. (2021)require careful inversion from real images to latent codes for real image editing. On the contrary, our proposed method based on diffusion models has a powerful advantage; the sophisticated inversion method is not necessary. This means that we can obtain the latent code of an arbitrary real image even if the image is not in the trained domain. On the other hand, several inversion methods have been proposed for GANs to obtain the latent of the real image, and the corresponding latent manipulation method should be considered for each inversion method. For example, it is difficult to apply the method of editing in $w$ space to the method of inversion using $w ^ { + }$ space. ", + "bbox": [ + 174, + 281, + 825, + 393 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "However, GANs have the advantage of fast sampling. In addition, diffusion models have a relatively slow sampling time. Additionally, we have to be aware of the time steps of diffusion models, which is still less well known. ", + "bbox": [ + 176, + 400, + 825, + 443 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "The advantage of being free from Inversion provides the following milestones. The manipulation in the latent of the diffusion models is the same as the editing in real images. It can be expanded to segmentation, clustering, classification, etc. in $h$ -space for real-world images. ", + "bbox": [ + 174, + 449, + 825, + 491 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "It would be an interesting research direction to employ previous techniques. Our method can be used in conjunction with gradient guidance methods. Although we do not focus on random sampling, ours works effectively for sampling with stochastic. (See $\\ S \\mathbf { M } .$ ) It may bring more diverse methods to steer diffusion models. ", + "bbox": [ + 174, + 498, + 825, + 554 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "$h$ -space in the latent diffusion models such as stable diffusion, is another interesting research direction. The main contribution of our paper is only modifying $\\mathbf { P } _ { t }$ while preserving $\\mathbf { D } _ { t }$ , and can be adapted with latent diffusion models. However, since the latent meaning may be different due to structural differences, research on this is needed. ", + "bbox": [ + 174, + 560, + 825, + 617 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "Furthermore, all of the properties of $h$ -space according to the time step has not been fully discussed so far. Research on them can be expected to expand further. ", + "bbox": [ + 173, + 625, + 823, + 652 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "Limitations. Editing with Asyrp seldom yields changes in overall style or peripheral objects but edits attributes of the main object. Style transfer using frozen diffusion models is our future work. ", + "bbox": [ + 171, + 676, + 823, + 705 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "Societal impact / Ethics statement. Techniques for high-quality image manipulation such as Asyrp should be accompanied by social and/or technical solutions to prevent abuse. We acknowledge the potential ethical implications that may arise from the use of our image manipulation technique, Asyrp. We advocate for the development and implementation of social and technical solutions to prevent potential abuses such as spreading disinformation or propaganda. We are committed to ensuring fairness and non-discrimination, legal compliance, and research integrity in our work. ", + "bbox": [ + 174, + 729, + 825, + 813 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "C PROOF OF THEOREM 1 ", + "text_level": 1, + "bbox": [ + 176, + 842, + 397, + 858 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "Proof of Theorem 1. Let $\\epsilon _ { t } ^ { \\theta }$ be a predicted noise during the original reverse process at $t$ and $\\tilde { \\epsilon } _ { t } ^ { \\theta }$ be its shifted counterpart. Then, $\\Delta \\pmb { x } _ { t } = \\tilde { \\pmb { x } } _ { t - 1 } - \\pmb { x } _ { t - 1 }$ is negligible where $\\tilde { \\pmb { x } } _ { t - 1 } = \\bar { \\sqrt { \\alpha _ { t - 1 } } } \\mathbf { P } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } \\bar { \\Big ) } ) +$ $\\mathbf { D } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )$ . ", + "bbox": [ + 176, + 878, + 823, + 924 + ], + "page_idx": 14 + }, + { + "type": "image", + "img_path": "images/278cf3f351f5df6dbbb8f4ac91469563ba5dc86a95e75f069873149e24faa7f4.jpg", + "image_caption": [ + "Figure 12: Illustration of Theorem 1. Upper blue line describes applying noise $\\tilde { \\epsilon } _ { t } = \\epsilon _ { t } + \\Delta \\epsilon _ { t }$ to produce $\\mathbf { P } _ { t } ( \\tilde { \\epsilon } _ { t } ) =$ the shifted predicted $\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }$ . However, the shift due to $\\Delta \\epsilon _ { t }$ is canceled out by the shift in $\\mathbf { D } _ { t } ( \\tilde { \\epsilon } _ { t } )$ due to $\\Delta \\epsilon _ { t }$ . As a results, applying $\\Delta \\epsilon _ { t }$ both on $\\mathbf { P } _ { t }$ and $\\mathbf { D } _ { t }$ brings identical outputs to the original. " + ], + "image_footnote": [], + "bbox": [ + 310, + 104, + 684, + 311 + ], + "page_idx": 15 + }, + { + "type": "text", + "text": "Define $\\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\bf { x } } _ { t } ) = \\epsilon _ { t } ^ { \\theta } ( { \\bf { x } } _ { t } ) + \\Delta \\epsilon _ { t }$ , $\\{ \\beta _ { t } \\} _ { t = 1 } ^ { T } = \\{ \\beta _ { 1 } = \\beta _ { \\operatorname* { m i n } } , . . . , \\beta _ { T } = \\beta _ { \\operatorname* { m a x } } \\}$ , and $\\begin{array} { r } { \\alpha _ { t } = \\prod _ { s = 1 } ^ { t } ( 1 - } \\end{array}$ $\\beta _ { s , \\ }$ ). Note that $\\beta _ { \\mathrm { m a x } }$ is defined as a small value (e.g., $\\beta _ { \\mathrm { m a x } } = 0 . 0 0 1 )$ and $\\{ \\beta _ { t } \\} _ { t = 1 } ^ { T }$ are defined by a decreasing schedule from $\\beta _ { T } = \\beta _ { \\mathrm { m a x } }$ to $\\beta _ { 1 } = \\beta _ { \\mathrm { m i n } } \\approx 0$ (e.g., $\\beta _ { \\mathrm { m i n } } = 0 . 0 0 0 0 1 \\rangle$ . Then, ", + "bbox": [ + 173, + 416, + 825, + 464 + ], + "page_idx": 15 + }, + { + "type": "equation", + "img_path": "images/01ee3c33e0761b3521ba97b4f84f5ffc229ad08d8505c273d637ac0146bcb0d9.jpg", + "text": "$$\n\\tilde { \\pmb { x } } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } \\big ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big ) + \\mathbf { D } _ { t } \\big ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big )\n$$", + "text_format": "latex", + "bbox": [ + 357, + 484, + 640, + 503 + ], + "page_idx": 15 + }, + { + "type": "equation", + "img_path": "images/2e19aec9451f33520a3a7974bbd5a035ae744237052531f372833c5a99ddff25.jpg", + "text": "$$\n= \\sqrt { \\alpha _ { t - 1 } } \\left( \\frac { x _ { t } - \\sqrt { 1 - \\alpha _ { t } } \\left( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) + \\Delta \\epsilon _ { t } \\right) } { \\sqrt { \\alpha _ { t } } } \\right) + \\sqrt { 1 - \\alpha _ { t - 1 } } \\cdot \\left( \\epsilon _ { t } ^ { ( \\theta ) } \\left( x _ { t } \\right) + \\Delta \\epsilon _ { t } \\right)\n$$", + "text_format": "latex", + "bbox": [ + 232, + 530, + 767, + 573 + ], + "page_idx": 15 + }, + { + "type": "equation", + "img_path": "images/f8f1d07701d3604aa1e847f9afe7750a0c182028698c917afefce6e9a09b95c4.jpg", + "text": "$$\n= \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) + \\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) - \\frac { \\sqrt { \\alpha _ { t - 1 } } \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { \\alpha _ { t } } } \\cdot \\Delta \\epsilon _ { t } + \\sqrt { 1 - \\alpha _ { t - 1 } } \\cdot \\Delta \\epsilon _ { t }\n$$", + "text_format": "latex", + "bbox": [ + 214, + 592, + 759, + 627 + ], + "page_idx": 15 + }, + { + "type": "equation", + "img_path": "images/160589c53833ae678a3010268005cbbbdf94a10e75615dc539618b396960f86e.jpg", + "text": "$$\n= x _ { t - 1 } + \\left( - \\frac { \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } + \\sqrt { 1 - \\alpha _ { t - 1 } } \\right) \\cdot \\Delta \\epsilon _ { t }\n$$", + "text_format": "latex", + "bbox": [ + 346, + 647, + 655, + 681 + ], + "page_idx": 15 + }, + { + "type": "equation", + "img_path": "images/94f6bb57ed66ae0aef5a4609c429c9af874ae238470b1c6b5a2328050c6645f2.jpg", + "text": "$$\n= x _ { t - 1 } + \\left( - \\frac { \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } + \\frac { \\sqrt { 1 - \\prod _ { s = 1 } ^ { t - 1 } \\left( 1 - \\beta _ { s } \\right) } \\sqrt { 1 - \\beta _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } \\right) \\cdot \\Delta \\epsilon _ { t }\n$$", + "text_format": "latex", + "bbox": [ + 281, + 700, + 720, + 752 + ], + "page_idx": 15 + }, + { + "type": "equation", + "img_path": "images/dcc07c1a0353bb620e8ab796ccc6fe1ae0e1a63406cf1e6326a0873ee1a876b7.jpg", + "text": "$$\n= x _ { t - 1 } + \\left( \\frac { \\sqrt { 1 - \\alpha _ { t } - \\beta _ { t } } - \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } \\right) \\cdot \\Delta \\epsilon _ { t } \\quad \\cdot : \\alpha _ { t } = \\prod _ { s = 1 } ^ { t } \\left( 1 - \\beta _ { s } \\right)\n$$", + "text_format": "latex", + "bbox": [ + 266, + 770, + 736, + 813 + ], + "page_idx": 15 + }, + { + "type": "equation", + "img_path": "images/f1b17f28b24912a779196b70a7b90dc23a4b070bc0ac53c342a45f37da7115e9.jpg", + "text": "$$\n\\therefore \\Delta x _ { t } = { \\tilde { x } } _ { t - 1 } - x _ { t - 1 } = \\left( { \\frac { { \\sqrt { 1 - \\alpha _ { t } - \\beta _ { t } } } - { \\sqrt { 1 - \\alpha _ { t } } } } { { \\sqrt { 1 - \\beta _ { t } } } } } \\right) \\cdot \\Delta \\epsilon _ { t } { \\mathrm { ~ i s ~ n e g l i g i b l e ~ } } \\quad \\cdot \\cdot \\beta _ { t } < \\beta _ { m a x }\n$$", + "text_format": "latex", + "bbox": [ + 192, + 839, + 802, + 875 + ], + "page_idx": 15 + }, + { + "type": "text", + "text": "D ADDITIONAL SUPPORTS FOR $h$ -space WITH ASYRP ", + "text_level": 1, + "bbox": [ + 174, + 102, + 625, + 118 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "D.1 RANDOM PERTURBATION ON $\\epsilon$ -space WITHOUT ASYRP ", + "bbox": [ + 174, + 133, + 594, + 147 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "In $\\ S 3 . 1$ , we argue that if both $\\mathbf { P } _ { t }$ and $\\mathbf { D } _ { t }$ are shifted, we can not manipulate $\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }$ . In Figure 13, we do not observe the noticeable difference between (a) and (b) which are the result of the original reverse process of DDIM and the one with shifting both terms, respectively. ", + "bbox": [ + 174, + 159, + 825, + 202 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "D.2 ROBUSTNESS AND SEMANTICS IN $h$ -space AND $\\epsilon$ -space WITH ASYRP ", + "text_level": 1, + "bbox": [ + 176, + 218, + 687, + 233 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "In $\\mathrm { ~ \\ S ~ } 3 . 2$ , we also argue that $h$ -space is more robust than $\\epsilon$ -space with Asyrp. In Figure 13, we observe that small random noise $z \\sim \\mathcal { N } ( 0 , \\mathbf { I } )$ in $\\epsilon$ -space degrades the resulting image without semantic changes (c) and much larger random noise in $h$ -space yields random semantic changes without severe artifacts (d). ", + "bbox": [ + 174, + 244, + 825, + 300 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "Note that diffusion models are designed as latent variable models with learned Gaussian transitions and the reverse process should also be close to Gaussian. Based on the assumption, we manage $\\epsilon _ { t } ^ { \\theta }$ to follow the original Gaussian distribution as follows. Adding $z \\sim \\mathcal { N } ( 0 , \\sigma \\bar { \\bf { I } } )$ to $\\epsilon _ { t } ^ { \\theta }$ expands the distribution of the predicted noise and may produce distorted images. To preserve the distribution,√ we scale $\\tilde { \\epsilon } _ { t } ^ { \\theta } = ( \\epsilon _ { t } ^ { \\theta } + z ) / \\sqrt { 1 ^ { 2 } + { \\sigma } ^ { 2 } }$ . Still, the resulting images are almost identical compared to the original images where $\\tilde { \\epsilon } _ { t } ^ { \\theta } = \\epsilon _ { t } ^ { \\theta } + z$ as shown in Figure 13. It is no wonder that the scaling does not improve the distorted results since the additive random noise disturbs the denoising operation of the predicted random noise. ", + "bbox": [ + 174, + 308, + 825, + 420 + ], + "page_idx": 16 + }, + { + "type": "image", + "img_path": "images/bb209ea4fa736576560335d535c5528804419b23be9234941f9e22111f467649.jpg", + "image_caption": [ + "Figure 13: (a) The reconstructed image by the original DDIM inversion process which is almost indistinguishable to the real image. (b) The result from adding random noise $z$ both on $\\mathbf { P } _ { t }$ and $\\mathbf { D } _ { t }$ . (a) looks identical to (b). The insets of (b, c) depict the full SSIM image from (a). It shows that simply shifting $\\epsilon _ { t } ^ { \\theta }$ without Asyrp does not affect the result. (c) Adding $z \\sim \\mathcal { N }$ to $\\epsilon$ -space with Asyrp easily degrades image with little semantic change. (d) Adding $z \\sim \\mathcal { N }$ to $h$ -space with Asyrp yields random semantic change without image degradation. (e) Correlation between image degradation and noise strength in the two spaces. " + ], + "image_footnote": [], + "bbox": [ + 179, + 435, + 816, + 546 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "D.3 CHOICE OF $h$ -space IN U-NET ", + "text_level": 1, + "bbox": [ + 176, + 689, + 423, + 704 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "As shown in Figure 14, there are many other candidates for $h$ -space in the architecture. Among the layers, we choose the 8th layer, the bridge of the U-Net based architecture. The layer is not influenced by any skip connection, has the smallest spatial dimension with compressed information, and is located just before the upsampling blocks. Thus, we assume that it could possibly be considered as the most suitable latent embedding. To confirm the assumption, we train $\\pmb { f } _ { t }$ on the other layers. The results are shown in Figure 15. We carefully tuned the training hyperparameters $\\lambda _ { C L I P }$ and $\\lambda _ { r e c o n } )$ ) for fair comparison. The 1st to the 6th layers hardly bring visible changes. The 7th and 9th layers bring not only the desired changes but also difficulty in finding optimal hyperparameters. After the 9th layer, the results bear severe artifacts. ", + "bbox": [ + 173, + 715, + 825, + 840 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "E IMPLICIT NEURAL DIRECTIONS ", + "text_level": 1, + "bbox": [ + 176, + 862, + 465, + 876 + ], + "page_idx": 16 + }, + { + "type": "text", + "text": "Figure 16 illustrates the neural implicit function $f _ { t }$ . It has only two 1x1 convolution layers with 512 channels. Note that we haven’t explored the network architecture much. ", + "bbox": [ + 168, + 892, + 823, + 921 + ], + "page_idx": 16 + }, + { + "type": "image", + "img_path": "images/3e8c45121a1bfb589a589081725c2d792d7bced359e7ac19f594c2ade31cda78.jpg", + "image_caption": [ + "Figure 14: Location of $\\pmb { h }$ -space. The U-Net architecture of diffusion models outputs $2 5 6 \\times 2 5 6$ images. Each layer is indexed with a number along the operating sequence of the model. The 8th layer is our $h$ -space which is not directly influenced by a skip connection. " + ], + "image_footnote": [], + "bbox": [ + 186, + 111, + 826, + 380 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "F QUALITY IMPROVEMENTS BY NON-ACCELERATED SAMPLING WITH SCALED $\\Delta h _ { t }$ ", + "text_level": 1, + "bbox": [ + 173, + 474, + 764, + 508 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "Figure 17 shows the quality improvements by non-accelerated sampling with scaled $\\Delta { h _ { t } }$ described in $\\ S \\ S . 4$ . Even with different number of inference steps, we observe similar changes of an attribute if we preserve the sum of $\\Delta \\boldsymbol { h } _ { t }$ . This scaling technique allows non-accelerated sampling with 1000 steps for the models trained by accelerated training with 40 steps. Using non-accelerated sampling with scaled $\\Delta { h _ { t } }$ leads to the same magnitude of manipulation and higher-quality images. In our experiments, it takes about 1.5 seconds to sample for 40 steps and 40 seconds for 1000 steps. ", + "bbox": [ + 173, + 525, + 825, + 609 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "G EDITING STRENGTH AND EDITING FLEXIBILITY ", + "text_level": 1, + "bbox": [ + 174, + 633, + 602, + 648 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "G.1 EDITING STRENGTH AND $t _ { e d i t }$ ", + "text_level": 1, + "bbox": [ + 176, + 666, + 424, + 681 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "Figure 18 shows the results according to $t _ { \\mathrm { e d i t } }$ . If $t _ { \\mathrm { e d i t } }$ is too high, the length of the editing process becomes too short resulting in insufficient changes. On the contrary, too low $t _ { \\mathrm { e d i t } }$ causes excessively unnecessary manipulation from the long editing process. ", + "bbox": [ + 174, + 694, + 825, + 736 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "We observe that $t _ { \\mathrm { e d i t } }$ is one of the important hyperparameters. We argue that the formula for choosing $t _ { \\mathrm { e d i t } }$ using editing strength is reasonable because it applies to all five different datasets despite its sensitivity, even though the choice of $\\mathrm { L P I P S = 0 . 3 3 }$ is empirical. We provide ablation of thresholds in Figure 22. ", + "bbox": [ + 174, + 742, + 825, + 799 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "Additionally, these results imply why we need to use sufficiently low $t _ { \\mathrm { e d i t } }$ in the unseen domains. \nEditing interval with insufficient editing strength struggles to escape from the training domain. ", + "bbox": [ + 169, + 805, + 823, + 834 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "G.2 EDITING FLEXIBILITY AND tboost ", + "text_level": 1, + "bbox": [ + 176, + 854, + 446, + 869 + ], + "page_idx": 17 + }, + { + "type": "text", + "text": "Table 2 shows the prompts, $t _ { \\mathrm { e d i t } }$ , and $t _ { \\mathrm { b o o s t } }$ . Intuitively, the attributes with larger visual changes have smaller cosine similarity and require longer editing interval. Figure 19 shows the average $\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t } )$ and $\\mathrm { L P I P S } ( { \\pmb x } , { \\pmb x } _ { t } )$ of 100 samples on all datasets. Note that $t _ { \\mathrm { e d i t } }$ and $t _ { \\mathrm { b o o s t } }$ dif", + "bbox": [ + 176, + 882, + 823, + 924 + ], + "page_idx": 17 + }, + { + "type": "image", + "img_path": "images/20e18f4b6da57f683105204199ad21e00b13391adc9cd4bf0a95392c4e8d14c8.jpg", + "image_caption": [ + "Figure 15: Exhaustive enumeration over the choices for semantic latent space. We show the result of the training data. We observe that the eighth layer ( $h$ -space) of the U-net suits the best for the semantic latent space. " + ], + "image_footnote": [], + "bbox": [ + 416, + 97, + 689, + 556 + ], + "page_idx": 18 + }, + { + "type": "image", + "img_path": "images/174bbae683903d32c4469c85c1da04852c054dd195ebaf00a66e2a411d7df08d.jpg", + "image_caption": [ + "Figure 16: Illustration of $\\pmb { f } _ { t }$ . We use group norm and swish following DDPM. " + ], + "image_footnote": [], + "bbox": [ + 232, + 631, + 761, + 742 + ], + "page_idx": 18 + }, + { + "type": "text", + "text": "fer across datasets and attributes, and they are chosen by the formulas in $\\ S 4 .$ . We provide ablation of $t _ { \\mathrm { b o o s t } }$ in Figure 23. ", + "bbox": [ + 173, + 808, + 825, + 837 + ], + "page_idx": 18 + }, + { + "type": "text", + "text": "H QUALITY BOOSTING ", + "text_level": 1, + "bbox": [ + 174, + 861, + 380, + 878 + ], + "page_idx": 18 + }, + { + "type": "text", + "text": "We validate the effectiveness of our quality boosting (§ 4.2) in the original DDIM process and in Asyrp. ", + "bbox": [ + 174, + 895, + 825, + 924 + ], + "page_idx": 18 + }, + { + "type": "text", + "text": "Smiling with ", + "bbox": [ + 555, + 106, + 637, + 119 + ], + "page_idx": 19 + }, + { + "type": "image", + "img_path": "images/e119805b4e47eb46571e4d9c938e16083540325efbc089de26542ba2e25cbc3d.jpg", + "image_caption": [ + "Figure 17: Quality improvements by non-accelerated sampling with scaled $\\Delta \\boldsymbol { h } _ { t }$ . We observe quality improvements by non-accelerated (1000-step) sampling with scaled $\\Delta \\boldsymbol { h } _ { t }$ from accelerated (40-step) training described in $\\ S 3 . 4$ . Please zoom in for detailed comparison. " + ], + "image_footnote": [], + "bbox": [ + 196, + 111, + 800, + 609 + ], + "page_idx": 19 + }, + { + "type": "equation", + "img_path": "images/35a76084e6ebb069475f8af856b6fca2df0d3f6c42c0262eb89b774ab89c58e8.jpg", + "text": "$$\nt _ { e d i t } = 9 0 0 ~ t _ { e d i t } = 8 0 0 ~ t _ { e d i t } = 7 0 0 ~ t _ { e d i t } = 6 0 0 ~ t _ { e d i t } = 5 0 0 ~ t _ { e d i t } = 4 0 0 ~ t _ { e d i t } = 3 0 0 ~ t _ { e d i t } = 2 0 0 ~ t _ { e d i t } = 1 0 0\n$$", + "text_format": "latex", + "bbox": [ + 181, + 689, + 823, + 704 + ], + "page_idx": 19 + }, + { + "type": "image", + "img_path": "images/5e9f94c3b8ff254f0b1c246dba27d57fe4749c6b53d4da39b435229817caa682.jpg", + "image_caption": [ + "Figure 18: Importance of choosing proper $t _ { \\mathrm { e d i t } }$ . We explore various $t _ { \\mathrm { e d i t } }$ with smiling. Too short editing interval struggles to manipulate attributes. Excessive editing strength results in degraded images. " + ], + "image_footnote": [], + "bbox": [ + 178, + 707, + 820, + 816 + ], + "page_idx": 19 + }, + { + "type": "table", + "img_path": "images/61cb1f8e818cb9695091612d9e566dc3bfdae8b922b211e762275091594b9337.jpg", + "table_caption": [], + "table_footnote": [], + "table_body": "
Datasetsrc_txttrg_txtCLIP similaritytedittboost
CelebA-HQ LSUN-churchPerson FaceYoung person0.905515167
Smiling face0.899513167
FaceSad face0.894513167
FaceAngry face0.892512167
FaceTanned face0.886512167
FaceDisgusted face0.880511167
Person HumanPerson with makeup0.875509167
PersonZombie0.868506167
PersonPerson with bald head0.861505167
Person with curly hair0.835499167
Human PersonNeanderthal0.802490167
PersonMark Zuckerberg0.797489167
HumanNicolas Cage0.710461167
PhotoPainting in the style of Pixar0.667446167
PhotoPainting in Modigliani style Self-portrait by Frida Kahlo0.565403167
ChurchGothic Church0.443321167
ChurchTemple0.912 0.898371293
ChurchDepartment store0.841367293
ChurchWooden House349293
Church0.793333293
ChurchAncient traditional Asian tower0.784330293
ChurchRed brick wall Church0.774326293
BedroomFactory Princess Bedroom0.702301293
LSUN-bedroomBedroom Hotel Bedroom0.912370221
AFHQDog0.917371221
DogHappy Dog0.883430167
Sleepy Dog0.866422167
DogAngry Dog0.860419167
Dog DogWolf0.850412167
METFACESPainting of a personYorkshire Terrier0.690302167
Painting of a personPainting of a Sad person Painting of a Smiling person0.921 0.908330 320180 180
Painting of a personPainting of a Disgusted person0.879291180
", + "bbox": [ + 173, + 99, + 826, + 525 + ], + "page_idx": 20 + }, + { + "type": "image", + "img_path": "images/d400055c06dce9745d0694a750ea708aaf62c69eec8d8f94c39559d4b4c1588e.jpg", + "image_caption": [ + "Table 2: Prompts, CLIP similarity, $t _ { \\mathrm { e d i t } }$ , and $t _ { \\mathrm { b o o s t } }$ for all attributes in the experiments. ", + "Figure 19: Average LPIPS $( \\boldsymbol { \\mathbf { \\mathit { x } } } , \\boldsymbol { \\mathbf { \\mathit { P } } } _ { t } )$ and LPIPS $( \\pmb { x } , \\pmb { x } _ { t } )$ of 100 samples on all datasets. " + ], + "image_footnote": [], + "bbox": [ + 173, + 574, + 820, + 808 + ], + "page_idx": 20 + }, + { + "type": "text", + "text": "Figure 20 shows the effect of quality boosting with the original DDIM reverse process. Although the reverse process of DDIM has a nearly-perfect inversion property, we observe some noise by zooming in. Our quality boosting improves the quality of a sample and concurrently keeps nearly ", + "bbox": [ + 174, + 881, + 823, + 924 + ], + "page_idx": 20 + }, + { + "type": "image", + "img_path": "images/1434d974f30676869508e856aafe42dbc606c49aa0ad3a871abe5831d5d8d9ad.jpg", + "image_caption": [ + "Figure 20: Ablation study of quality boosting on the original DDIM process without Asyrp. Our quality boosting enhances fine details and prevents images from being noisy in the original DDIM process. Please zoom in for detailed comparison. " + ], + "image_footnote": [], + "bbox": [ + 209, + 104, + 789, + 419 + ], + "page_idx": 21 + }, + { + "type": "text", + "text": "", + "bbox": [ + 571, + 486, + 617, + 498 + ], + "page_idx": 21 + }, + { + "type": "image", + "img_path": "images/c3c2bcb78752a1c696fd47a0788206292074dab31e50f5bcc13a4530d4914cd3.jpg", + "image_caption": [ + "Figure 21: Ablation study of quality boosting with Asyrp. Our quality boosting enhances fine details and prevents images from being noisy. Note that the source of degradation is DDIM process, not Asyrp, confirmed in Figure 20. Please zoom in for detailed comparison. " + ], + "image_footnote": [], + "bbox": [ + 207, + 489, + 790, + 816 + ], + "page_idx": 21 + }, + { + "type": "image", + "img_path": "images/9105ae3ff594358c3dd1ad873a025ed2c74b6b492fe058518f574f1ebe0cc84f.jpg", + "image_caption": [ + "Figure 22: Analyzing the effect of the hyperparameters The figure shows that the calculated parameter works effectively as the maximum boundary of editing while maintaining the quality of the image. " + ], + "image_footnote": [], + "bbox": [ + 210, + 103, + 789, + 287 + ], + "page_idx": 22 + }, + { + "type": "image", + "img_path": "images/067bbe65a525c781959e6f88a9e6ed2322b6ae14dd2301e761c08b57fdce7806.jpg", + "image_caption": [ + "Figure 23: Analyzing the effect of the hyperparameters We observe that the quality of results has robustness to $\\gamma _ { t _ { \\mathrm { b o o s t } } }$ , except for too large $\\gamma _ { t _ { \\mathrm { b o o s t } } }$ . " + ], + "image_footnote": [], + "bbox": [ + 207, + 375, + 789, + 563 + ], + "page_idx": 22 + }, + { + "type": "text", + "text": "perfect inversion property. We observe that $t _ { \\mathrm { b o o s t } }$ is not sensitive, but the larger interval brings the less preservation. ", + "bbox": [ + 176, + 645, + 823, + 674 + ], + "page_idx": 22 + }, + { + "type": "text", + "text": "Figure 21 shows quality improvements by our quality boosting in Asyrp. ", + "bbox": [ + 174, + 680, + 650, + 695 + ], + "page_idx": 22 + }, + { + "type": "text", + "text": "I ALGORITHM ", + "text_level": 1, + "bbox": [ + 174, + 723, + 308, + 739 + ], + "page_idx": 22 + }, + { + "type": "text", + "text": "Figure 24 illustrates generative process. Algorithm 1 and 2 describe training algorithm and inference algorithm of Asyrp, respectively. ", + "bbox": [ + 173, + 760, + 823, + 787 + ], + "page_idx": 22 + }, + { + "type": "text", + "text": "J TRAINING DETAILS ", + "text_level": 1, + "bbox": [ + 176, + 815, + 364, + 832 + ], + "page_idx": 22 + }, + { + "type": "text", + "text": "J.1 LOSS COEFFICIENTS ", + "text_level": 1, + "bbox": [ + 176, + 852, + 352, + 866 + ], + "page_idx": 22 + }, + { + "type": "text", + "text": "Table 3 reports loss coefficients for each attribute. $\\lambda _ { \\mathrm { { C L I P } } } \\mathrm { { s } }$ near 0.8 are suitable for most in-domain attributes. For unseen domains, higher coefficient leads to more noticeable changes. Note that we use $\\lambda _ { \\mathrm { r e c o n } } = \\mathbf { C } \\mathbf { L } \\mathbf { I } \\mathbf { P }$ similarity $^ \\ast 3$ which reduces L1 loss when an attribute needs a lot of change. ", + "bbox": [ + 176, + 882, + 825, + 924 + ], + "page_idx": 22 + }, + { + "type": "image", + "img_path": "images/4a4dfe0ce3ef126a67f1e86d022d8d15a8d17d09cceb878b5663ee16091daf9b.jpg", + "image_caption": [ + "Figure 24: Overview of our generative process. " + ], + "image_footnote": [], + "bbox": [ + 174, + 102, + 823, + 294 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "J.2 TRAINING WITH RANDOM SAMPLING INSTEAD OF THE TRAINING DATASETS ", + "bbox": [ + 174, + 366, + 736, + 380 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "Apparently, for training, inverting real-images can be replaced by random sampling. It refers to using $\\mathbf { \\boldsymbol { x } } _ { T } \\sim \\mathcal { N } ( 0 , \\mathbf { I } )$ instead of $\\begin{array} { r } { \\pmb { x } _ { T } = \\bar { q } ( \\pmb { x } _ { 0 } ) \\cdot \\bar { \\prod } _ { t = 1 } ^ { T } q \\left( \\pmb { x } _ { t } \\mid \\bar { \\pmb { x } } _ { t - 1 } \\right) } \\end{array}$ where $\\pmb { x } _ { 0 } \\sim p _ { d a t a } ( x )$ . It allows us to train Asyrp only with the pretrained network and without extra dataset. Using random samples has tradeoff between preservation of contents and possible amount in editing. It take advantages when a target attribute requires large amount changes. We assume that the inversion of real-image is in the long tail of a Gaussian distribution because of realistic background or detailed clothes. On the other hand, random noise is considered to be closer to the mean of the normal, so it is easier to find directions. It can easily bring larger changes but also easily alter the contents. On the contrary, training with inversion shows the opposite property. ", + "bbox": [ + 173, + 395, + 825, + 525 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "We train $\\pmb { f } _ { t }$ with random sampling for attributes whose identity preservation is not important to take advantage of these properties. The rightmost column in Table 3 shows the choices. ", + "bbox": [ + 173, + 531, + 823, + 559 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "K EVALUATION ", + "text_level": 1, + "bbox": [ + 176, + 590, + 320, + 606 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "K.1 USER STUDY ", + "text_level": 1, + "bbox": [ + 176, + 627, + 308, + 641 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "We conduct user study to compare the performance of Asyrp and DiffusionCLIP (Kim & Ye, 2021) on Celeba-HQ (Liu et al., 2015) and LSUN-church (Yu et al., 2015). We use official checkpoints provided by DiffusionCLIP except for some facial attributes whose checkpoint do not exist. We tried our best to tune their hyperparameters following the manual for fair comparison. Example images are shown in Figure 25-27. ", + "bbox": [ + 174, + 657, + 825, + 728 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "We use smiling and sad for in-domain CelebA-HQ attributes, Pixar and Neanderthal for unseen-domain CelebA-HQ attributes, and department store, ancient, and wooden for LSUN-church. ", + "bbox": [ + 174, + 734, + 825, + 776 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "In unseen domain and Lsun-church, we use official checkpoints provided by DiffusionCLIP. We also randomly select 8 images for each CelebA-HQ attribute and 12 images for each LSUN-church attribute. ", + "bbox": [ + 176, + 784, + 825, + 825 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "We observe that DiffusionCLIP works better in changing the holistic style of images. At the same time, it is short of the ability to bring semantic changes and suffers noisy results and a lack of diversity. The problems would be caused by fine-tuning the whole diffusion model. ", + "bbox": [ + 176, + 832, + 825, + 875 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "We use the following questions for the survey. 1) Quality: Which image quality do you think is better? (clear and less noisy) 2) Attribute: Which image do you think is “Attribute(e.g., Smiling) naturally”? 3) Overall: Which image do you think is better considering the above evaluation criteria? ", + "bbox": [ + 176, + 882, + 825, + 924 + ], + "page_idx": 23 + }, + { + "type": "text", + "text": "Algorithm 1: Editing(Inference) ", + "text_level": 1, + "bbox": [ + 174, + 107, + 392, + 122 + ], + "page_idx": 24 + }, + { + "type": "text", + "text": "Input: $x _ { 0 }$ (Input image), $\\{ f _ { t } ^ { i } \\} _ { i = 1 } ^ { M }$ (M Neural implicit functions for M attributes), $\\{ c _ { i } \\} _ { i = 1 } ^ { M }$ (user defined M scaling coefficients for M attributes), $\\epsilon _ { \\theta }$ (frozen pretrained model), $S _ { f o r }$ (# of inversion steps), $S _ { g e n }$ (# of inference steps), $t _ { \\mathrm { e d i t } }$ (computed from $\\ S 4 . 1$ ), tboost (computed from $\\ S 4 . 2$ ) 1 Function Editor $( x _ { 0 } , \\epsilon _ { \\theta } , \\{ c _ { i } \\} _ { i = 1 } ^ { M } , S _ { f o r } , S _ { g e n } , * )$ : // step 1: Semantic encoding 2 Define $\\{ \\tau _ { s } \\} _ { s = 1 } ^ { S _ { f o r } }$ s.t $\\tau _ { 1 } = 0$ , $\\tau _ { S _ { f o r } } = T$ 3 for $s = 1 , 2 , . . . , S _ { f o r } - 1$ do 4 $\\epsilon \\epsilon _ { \\theta } ( x _ { \\tau _ { s } } , \\tau _ { s } )$ 5 $x _ { \\tau _ { s + 1 } } = \\sqrt { \\alpha _ { \\tau _ { s } } } x _ { \\tau _ { s } } + \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon$ 6 $\\{ \\tilde { \\tau } _ { s } \\} _ { s = 1 } ^ { S _ { g e n } }$ Mas.t $\\tilde { \\tau } _ { 1 } = 0$ a, $\\tilde { \\tau } _ { S _ { e d i t } } = t _ { e d i t }$ , $\\tilde { \\tau } _ { S _ { n o i s e } } = t _ { \\mathrm { b o o s t } }$ and $\\tilde { \\tau } _ { S _ { g e n } } = T$ 7 $\\tilde { x } _ { \\tilde { \\tau } _ { S _ { g e n } } } = x _ { \\tilde { \\tau } _ { S _ { f o r } } }$ 8 for $s = S _ { g e n } , S _ { g e n } - 1 , . . . , 2$ do // phase 1: editing 9 if $s \\geq S _ { e d i t }$ then 10 Extract feature map $h _ { \\tilde { \\tau } _ { s } }$ from $\\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } )$ 11 $\\begin{array} { r l } & { \\Delta h _ { \\tilde { \\tau } _ { s } } = \\frac { S _ { f o r } } { S _ { g e n } } ( \\sum _ { i = 1 } ^ { M } c _ { i } f _ { \\tilde { \\tau } _ { s } } ^ { i } ( h _ { \\tilde { \\tau } _ { s } } ) ) } \\\\ & { \\tilde { \\epsilon } = \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } | \\Delta h _ { \\tilde { \\tau } _ { s } } ) } \\\\ & { \\epsilon = \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } ) } \\\\ & { \\sigma _ { \\tilde { \\tau } _ { s } } = 0 } \\end{array}$ 12 13 14 // phase 2: denoising 15 else if $s \\geq S _ { n o i s e }$ then 16 ϵ˜ = ϵ = ϵθ(˜xτ˜s ) 17 στ˜ = 0 // phase 3: quality boosting 18 else 19 $\\begin{array} { r l } & { \\quad \\Big \\lfloor \\tilde { \\epsilon } = \\epsilon = \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } ) } \\\\ & { \\quad \\Big \\lfloor \\sigma _ { \\tilde { \\tau } _ { s } } = \\sqrt { \\left( 1 - \\alpha _ { \\tilde { \\tau } _ { s } - 1 } \\right) / \\left( 1 - \\alpha _ { \\tilde { \\tau } _ { s } } \\right) } \\sqrt { 1 - \\alpha _ { \\tilde { \\tau } _ { s } } / \\alpha _ { \\tilde { \\tau } _ { s } - 1 } } } \\\\ & { \\quad z \\sim N ( 0 , 1 ) } \\\\ & { \\quad \\tilde { x } _ { \\tilde { \\tau } _ { s - 1 } } = \\sqrt { \\alpha _ { \\tilde { \\tau } _ { s - 1 } } } \\big ( \\frac { \\tilde { x } _ { \\tilde { \\tau } _ { s } } - \\sqrt { 1 - \\alpha _ { \\tilde { \\tau } _ { s } } } \\tilde { \\epsilon } } { \\sqrt { \\alpha _ { \\tilde { \\tau } _ { s } } } } \\big ) + \\sqrt { 1 - \\alpha _ { \\tilde { \\tau } _ { s - 1 } } - \\sigma _ { \\tilde { \\tau } _ { s } } ^ { 2 } } \\epsilon + \\sigma _ { \\tilde { \\tau } _ { s } } z } \\end{array}$ 20 21 22 ", + "bbox": [ + 156, + 123, + 812, + 642 + ], + "page_idx": 24 + }, + { + "type": "text", + "text": "return $\\tilde { x } _ { 0 }$ (manipulated image) ", + "bbox": [ + 176, + 638, + 406, + 654 + ], + "page_idx": 24 + }, + { + "type": "text", + "text": "As for LSUN-church, we provide a set of four images at once and add a question: 3) Diversity: Which group do you think has a more diverse style? 4) Overall: Which image do you think is better considering the above evaluation criteria? ", + "bbox": [ + 176, + 691, + 823, + 734 + ], + "page_idx": 24 + }, + { + "type": "text", + "text": "K.2 SEGMENTATION CONSISTENCY AND DIRECTIONAL CLIP SIMILARITY ", + "text_level": 1, + "bbox": [ + 176, + 751, + 696, + 766 + ], + "page_idx": 24 + }, + { + "type": "text", + "text": "We compare Asyrp and DiffusionCLIP using directional CLIP similarity $( S _ { \\mathrm { d i r } } )$ and segmentationconsistency (SC) following the protocols in DiffusionCLIP in Table 4 and Table 5. A pretrained CLIP (Radford et al., 2021) and segmentation models (Yu et al. (2018); Zhou et al. (2019; 2017); Lee et al. (2020)) are used to compute $S _ { \\mathrm { d i r } }$ and SC, respectively. We choose three attributes (smiling, sad, tanned) for CelebA-HQ-in-domain, two attributes (Pixar,Neaderthal) for CelebA-HQunseen-domain and three attributes (department store, ancient, red brick) for LSUNchurch. ", + "bbox": [ + 173, + 776, + 825, + 875 + ], + "page_idx": 24 + }, + { + "type": "text", + "text": "For a fair comparison, we use official checkpoints of DiffusionCLIP and provide scores of the attributes (tanned, red brick) following Kim & Ye (2021). Regarding attributes without the official checkpoints (smiling,sad), we train DiffusionCLIP by ourselves with the official code. ", + "bbox": [ + 176, + 882, + 823, + 924 + ], + "page_idx": 24 + }, + { + "type": "text", + "text": "Algorithm 2: Training Neural implicit function $f _ { t }$ ", + "text_level": 1, + "bbox": [ + 173, + 107, + 504, + 122 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "Input: $\\epsilon _ { \\theta }$ (pretrained attribute), el), (sou $\\{ x _ { 0 } ^ { ( i ) } \\} _ { i = 1 } ^ { N }$ ages to preco(target text), te), (# $f _ { t }$ (Neural implicitinversion steps), nction of(# of $y _ { s r c }$ $y _ { t a r }$ $S _ { f o r }$ $K$ training epochs), $t _ { \\mathrm { e d i t } }$ (computed from $\\ S 4 . 1 \\}$ , $t _ { \\mathrm { b o o s t } }$ (computed from $\\ S 4 . 2 )$ ) ", + "bbox": [ + 178, + 127, + 797, + 172 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "// step 1: Precompute latents \n1 Define $\\{ \\tau _ { s } \\} _ { s = 1 } ^ { S _ { f o r } }$ s.t $\\tau _ { 1 } = 0 , \\tau _ { S _ { f o r } } = T$ \n2 for $i = 1 , 2 , . . . , N$ do \n3 for $s = 1 , 2 , . . . , S _ { f o r } - 1$ do \n4 $\\epsilon \\epsilon _ { \\theta } ( x _ { \\tau _ { s } } ^ { ( i ) } , \\bar { \\tau _ { s } } )$ \n5 $\\begin{array} { r } { x _ { \\tau _ { s + 1 } } ^ { ( i ) } = \\sqrt { \\alpha _ { \\tau _ { s } } } x _ { \\tau _ { s } } ^ { ( i ) } + \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon } \\end{array}$ 6 Save the latent x(i)τSf or ", + "bbox": [ + 160, + 176, + 460, + 301 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "7 $\\tau _ { S _ { e d i t } } = t _ { e d i t }$ // step 2: Update $\\pmb { f } _ { t }$ 8 for ep ${ \\varkappa } h { = 1 , 2 , . . . , K }$ do ", + "bbox": [ + 158, + 310, + 369, + 351 + ], + "page_idx": 25 + }, + { + "type": "equation", + "img_path": "images/0ba6b4a72d84a3eab40bb6fb5efa2526ffdd9ef453ce460351eeb798a49902e7.jpg", + "text": "$$\n\\begin{array} { r } { P = \\frac { \\tilde { x } _ { \\tau _ { s } } ^ { ( i ) } - \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon _ { \\theta } \\left( \\tilde { x } _ { \\tau _ { s } } ^ { ( i ) } | \\Delta h _ { \\tau _ { s } } \\right) } { \\sqrt { \\alpha _ { \\tau _ { s } } } } ; P _ { s r c } = \\frac { x _ { \\tau _ { s } } ^ { ( i ) } - \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon _ { \\theta } \\left( x _ { \\tau _ { s } } ^ { ( i ) } \\right) } { \\sqrt { \\alpha _ { \\tau _ { s } } } } } \\end{array}\n$$", + "text_format": "latex", + "bbox": [ + 246, + 430, + 642, + 458 + ], + "page_idx": 25 + }, + { + "type": "equation", + "img_path": "images/bc6689a31c48e5761d9ac0c043ce0d1e0ee55fb1266965a19fd3791bbbf44a8e.jpg", + "text": "$$\n\\begin{array} { r l } & { \\tilde { x } _ { \\tau _ { s - 1 } } ^ { ( i ) } = \\sqrt { \\alpha _ { \\tau _ { s - 1 } } } P + \\sqrt { 1 - \\alpha _ { \\tau _ { s - 1 } } } \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tau _ { s } } ^ { ( i ) } ) } \\\\ & { x _ { \\tau _ { s - 1 } } ^ { ( i ) } = \\sqrt { \\alpha _ { \\tau _ { s - 1 } } } P _ { s r c } + \\sqrt { 1 - \\alpha _ { \\tau _ { s - 1 } } } \\epsilon _ { \\theta } ( x _ { \\tau _ { s } } ^ { ( i ) } ) } \\end{array}\n$$", + "text_format": "latex", + "bbox": [ + 246, + 462, + 542, + 503 + ], + "page_idx": 25 + }, + { + "type": "equation", + "img_path": "images/4af7d31768be7cc1ed82bdab507bf120173b9d97531e91f8e08dfa2efa05090c.jpg", + "text": "$$\nE _ { t o t a l } \\lambda _ { C L I P } L _ { d i r e c t i o n } ( P , y _ { t a r } , P _ { s r c } , y _ { s r c } ) + \\lambda _ { r e c o n } | P - P _ { s r c } |\n$$", + "text_format": "latex", + "bbox": [ + 248, + 506, + 692, + 522 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "We use 100 samples per attribute. Asyrp outperforms DiffusionCLIP on $S _ { \\mathrm { d i r } }$ for all attributes. On SC, DiffusionCLIP achieves better or competitive scores. Because DiffusionCLIP manipulates images mostly by focusing on texture or color while preserving structure and shape, it takes advantage of getting higher SC scores. However, in Figure 28, results of smiling show that it is not proper to edit attributes which require structural manipulation. Note that better SC scores do not guarantee better qualitative performance. Results of Pixar also show the similar tendency of each method. We allow more structural changes than DiffusionCLIP while editing. Lower SC of our method comes from desirable structural changes as shown in Figure 28. ", + "bbox": [ + 173, + 589, + 825, + 703 + ], + "page_idx": 25 + }, + { + "type": "equation", + "img_path": "images/53ce9b39242e61d513e443f453e9d9b59f2f8d6c59eb55f50e784775b71557e9.jpg", + "text": "$$\nS _ { \\mathrm { d i r } } \\left( x _ { \\mathrm { g e n } } , y _ { \\mathrm { t a r } } ; x _ { \\mathrm { r e f } } , y _ { \\mathrm { r e f } } \\right) : = \\frac { \\left. \\Delta I , \\Delta T \\right. } { \\left\\| \\Delta I \\right\\| \\left\\| \\Delta T \\right\\| } ,\n$$", + "text_format": "latex", + "bbox": [ + 346, + 728, + 619, + 762 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "L DIRECTIONS ", + "text_level": 1, + "bbox": [ + 174, + 787, + 313, + 804 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "L.1 GLOBAL DIRECTION ", + "text_level": 1, + "bbox": [ + 174, + 819, + 357, + 834 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "Figure 29 and Figure 30 show that the effects of mean direction and global direction are quite similar with $\\Delta h _ { t }$ by $\\mathbf { \\Delta } f _ { t }$ in various attributes. We compute mean direction and global direction from 20 different images. ", + "bbox": [ + 174, + 845, + 825, + 888 + ], + "page_idx": 25 + }, + { + "type": "text", + "text": "We argue that $h$ -space is roughly homogeneous across samples and timesteps. However, we observe that $h$ -space is not completely independent to the conditions especially on unseen-domain (See Figure 29). Note that unseen-domains require a longer editing interval with small $t _ { \\mathrm { e d i t } }$ . Therefore, we conjecture that the consistency of $h$ -space decreases at the end of the generative process. It is supported by additional experiments that L2 distance between $\\Delta h _ { t }$ and global direction gradually increases along with timesteps. We leave a more detailed analysis on $h$ -space at different timesteps as a future work. ", + "bbox": [ + 174, + 895, + 821, + 924 + ], + "page_idx": 25 + }, + { + "type": "table", + "img_path": "images/93e57daef13c9d000d22d16cc5b1feb897cc08af451e893d40410c75fce7b88a.jpg", + "table_caption": [ + "Table 3: The coefficients range from 0.5 to 0.8 for in-domain attributes. Unseen domains need slightly stronger $\\lambda _ { \\mathrm { { C L I P } } } \\mathrm { { s } }$ . We also report which attributes we train with random noise sampling. The criterion is which attributes require relatively less maintenance of identity. We use $\\lambda _ { \\mathrm { { r e c o n } } } =$ CLIP similarity $^ \\ast 3$ which reduces L1 loss when an attribute needs a lot of change. " + ], + "table_footnote": [], + "table_body": "
Datasetsrc_txttrg_txtXCLIPXreconfrom random noise
CelebA-HQPersonYoung person0.80.905*3X
FaceSmiling face0.80.899*3X
FaceSad face0.80.894*3X
FaceAngry face0.80.892*3X
FaceTanned face0.80.886*3X
FaceDisgusted face0.80.880*3X
PersonPerson with makeup0.80.875*3X
Person PhotoPerson with curly hair0.80.835*3X
Self-portrait by Frida Kahlo0.50.44330
HumanZombie0.60.868*30
PersonNicolas Cage0.80.710*30
Human PhotoPainting in the style of Pixar0.80.667*30
HumanPainting in Modigliani style0.80.565*30
LSUN-churchChurchNeanderthal Gothic church1.20.802*30
ChurchTemple0.80.912*30
ChurchDepartment store0.80.898*30
ChurchWooden house0.80.841*30
ChurchAncient traditional Asian tower0.8 0.80.793*30
ChurchRed brick wall Church0.80.784*3 0.774*30
ChurchFactory0.80.702*30
LSUN-bedroomBedroomPrincess bedroom1.50.912*30 0
BedroomHotel bedroom1.50.917*3
AFHQ-dogDogHappy dog0
DogSleepy dog1.50.883*30
Dog1.50.866*30
DogWolf Yorkshire terrier1.70.850*30
METFACESPainting of a personPainting of a Sad person2 1.50.690*3 0.921*30 X
Painting of a personPainting of a Smiling person1.50.908*3X
Painting of a personPainting of a Disgusted person1.50.879*3X
", + "bbox": [ + 173, + 175, + 825, + 536 + ], + "page_idx": 26 + }, + { + "type": "table", + "img_path": "images/39128b22784c462f3d60b7d120d611eaf69db40d8f68d9dcfcebcfddaf3e23f5.jpg", + "table_caption": [ + "Table 4: Quantitative evaluation on CelebA-HQ. " + ], + "table_footnote": [], + "table_body": "
CelebA-HQ-in-domainCelebA-HQ-unseen-domain
SmilingSadTannedPixarNeanderthal
SdirSCSdirSCSdirSCSdirSCSdirSC
Asyrp (ours)0.92189.02%0.96488.90%0.99185.71%0.95676.51%0.80579.03
DiffusionCLIP0.81391.41%0.76089.93%0.88892.85%0.81189.91%0.66181.23%
", + "bbox": [ + 173, + 760, + 821, + 823 + ], + "page_idx": 26 + }, + { + "type": "image", + "img_path": "images/c2cf22d51b44b51faa69e1a51e8e027ba84a40344d50621d1a284856a9eb758c.jpg", + "image_caption": [ + "Figure 25: Asyrp vs. DiffusionCLIP on CelebA-HQ in-domain attributes. We observe that DiffusionCLIP struggles to change semantic facial attributes. Their official checkpoints do not exist and we ran careful hyperparameter tuning for training. " + ], + "image_footnote": [], + "bbox": [ + 176, + 102, + 820, + 347 + ], + "page_idx": 27 + }, + { + "type": "image", + "img_path": "images/06f561db0257bb962e8c406afb37da80494f09913966d224a12c97e0991f5899.jpg", + "image_caption": [ + "Figure 26: Asyrp vs. DiffusionCLIP on CelebA-HQ unseen-domain attributes. We use the official checkpoint provided by DiffusionCLIP. Asyrp works well even for the unseen-domains. " + ], + "image_footnote": [], + "bbox": [ + 179, + 414, + 820, + 657 + ], + "page_idx": 27 + }, + { + "type": "table", + "img_path": "images/63d4f8f633262547023dff557121d24f1c8c7ecb03ea93fc468a9540dcddfdf3.jpg", + "table_caption": [ + "Table 5: Quantitative evaluation on LSUN-church. " + ], + "table_footnote": [], + "table_body": "
LSUN_church
Department storeancientred brick
SdirSCSdirSCSdirSC
Asyrp (ours)0.77857.62%0.94362.65%0.98965.83%
DIffusionCLIP0.66154.50%0.90764.82%0.96465.02%
", + "bbox": [ + 243, + 727, + 754, + 800 + ], + "page_idx": 27 + }, + { + "type": "text", + "text": "", + "bbox": [ + 173, + 853, + 825, + 924 + ], + "page_idx": 27 + }, + { + "type": "image", + "img_path": "images/e6959cd9188841bfcc00b46aeed46d7800c43970a4f909834f9efccdda44607a.jpg", + "image_caption": [ + "Figure 27: Asyrp vs. DiffusionCLIP on LSUN-church. Note that we use the official checkpoint provided by DiffusionCLIP. We observe DiffusionCLIP produces narrow range of styles while Asyrp produces diverse styles. " + ], + "image_footnote": [], + "bbox": [ + 171, + 93, + 820, + 601 + ], + "page_idx": 28 + }, + { + "type": "text", + "text": "L.2 COMPARE THREE METHODS ", + "text_level": 1, + "bbox": [ + 176, + 690, + 410, + 705 + ], + "page_idx": 28 + }, + { + "type": "text", + "text": "In this section, we compare three methods: implicit neural direction $f _ { t }$ , optimized $\\Delta h _ { t }$ , and optimized ∆hglobal. ", + "bbox": [ + 171, + 717, + 823, + 746 + ], + "page_idx": 28 + }, + { + "type": "text", + "text": "Training time $f _ { t } \\approx \\Delta h ^ { g l o b a l } < \\Delta h _ { t }$ ", + "bbox": [ + 176, + 761, + 431, + 779 + ], + "page_idx": 28 + }, + { + "type": "text", + "text": "We have to optimize each $\\Delta h _ { t }$ for each time step $t$ . Additionally, it needs specific hyperparameters for each $\\Delta h _ { t }$ , e.g., higher learning rates for larger $t$ . On the contrary, time-consuming for $f _ { t }$ is similar to optimizing ∆hglobal. ", + "bbox": [ + 174, + 784, + 825, + 827 + ], + "page_idx": 28 + }, + { + "type": "text", + "text": "Quality ", + "text_level": 1, + "bbox": [ + 174, + 840, + 390, + 858 + ], + "page_idx": 28 + }, + { + "type": "text", + "text": "$f _ { t }$ and $\\Delta h _ { t }$ , where directions can be obtained for each timestep, have the best quality. As can be shown in Figure 10, $\\Delta h ^ { g l o b a l }$ is sometimes accompanied by slight differences in hair, etc. ", + "bbox": [ + 174, + 864, + 825, + 893 + ], + "page_idx": 28 + }, + { + "type": "text", + "text": "Extensibility $f _ { t } > \\Delta h _ { t } > \\Delta h ^ { g l o b a l }$ ", + "bbox": [ + 174, + 907, + 423, + 924 + ], + "page_idx": 28 + }, + { + "type": "image", + "img_path": "images/4bcaa6683a66fe5dc47e975b0ea521a3cef4e214da2c1f2d293b268e362cc228.jpg", + "image_caption": [ + "Figure 28: Example segmentations for computing segmentation-consistency (SC). " + ], + "image_footnote": [], + "bbox": [ + 259, + 104, + 692, + 630 + ], + "page_idx": 29 + }, + { + "type": "text", + "text": "$\\Delta h _ { t }$ can be obtained from $f _ { t }$ , and $\\Delta h ^ { g l o b a l }$ can be obtained by aggregating $\\Delta h _ { t }$ . \nWe opt to use $f _ { t }$ for above three advantages. ", + "bbox": [ + 174, + 718, + 702, + 756 + ], + "page_idx": 29 + }, + { + "type": "text", + "text": "M RANDOM SAMPLING ", + "text_level": 1, + "bbox": [ + 174, + 813, + 385, + 829 + ], + "page_idx": 29 + }, + { + "type": "text", + "text": "We conduct extra experiments: generating images with target attributes using Asyrp not from inversion but from random Gaussian noises. As a consequence, the generative process can be used for conditional random sampling. We provide the results in Figure 31. However, it is beyond the scope of this paper. ", + "bbox": [ + 174, + 867, + 825, + 924 + ], + "page_idx": 29 + }, + { + "type": "image", + "img_path": "images/03e1b521e960c51738ecc1ae3c6697bcf45be2371c9a8ab319c4542be8d594bf.jpg", + "image_caption": [ + "Figure 29: We compute mean direction and global direction from 20 other images on CelebAHQ. The effect of mean direction and global direction are quite similar with $\\Delta h _ { t }$ by $f _ { t }$ at diverse attributes. " + ], + "image_footnote": [], + "bbox": [ + 187, + 103, + 818, + 406 + ], + "page_idx": 30 + }, + { + "type": "image", + "img_path": "images/94795dd3dafb5d37e50068f7124646c932ba91d53b723f68b82fcc14beab1bf8.jpg", + "image_caption": [ + "Figure 30: We obtain mean direction and global direction from 20 other images on LSUN-church. The effect of mean direction and global direction are quite similar with $\\Delta h _ { t }$ by $f _ { t }$ at diverse attributes. " + ], + "image_footnote": [], + "bbox": [ + 176, + 479, + 820, + 669 + ], + "page_idx": 30 + }, + { + "type": "image", + "img_path": "images/6eb1c92e476116c392feb38047e039fb60ec3d76570c709ea5ec47a32b19d996.jpg", + "image_caption": [ + "Figure 31: Uncurated random sampling. We generate images from random noise with Asyrp. Although we do not focus on these results, Asyrp can be used for conditional sampling. " + ], + "image_footnote": [], + "bbox": [ + 176, + 746, + 821, + 840 + ], + "page_idx": 30 + }, + { + "type": "text", + "text": "N MORE SAMPLES ", + "text_level": 1, + "bbox": [ + 176, + 102, + 344, + 118 + ], + "page_idx": 31 + }, + { + "type": "text", + "text": "N.1 IMAGENET", + "text_level": 1, + "bbox": [ + 174, + 133, + 297, + 148 + ], + "page_idx": 31 + }, + { + "type": "text", + "text": "We conduct extra experiments: editing images with target class using Asyrp with ImageNet pretrained model. We verified that models trained on large datasets, such as ImageNet, can be edited using Asyrp. However, we also observed that in this case the latent space is not partitioned by classes. For an orange, we have different latents for a single orange, for many oranges, for a cross-section of cut orange, and for a single piece of orange. Therefore, we learned the implicit function by collecting similar images to find the direction. ", + "bbox": [ + 174, + 159, + 826, + 242 + ], + "page_idx": 31 + }, + { + "type": "image", + "img_path": "images/774272204ff4ad42b215ddad0f822037c3a70d6b9b02276bd89a0e47fea88d21.jpg", + "image_caption": [ + "Figure 32: Result of Asyrp in ImageNet. The result shows that Asyrp works even in ImageNet dataset. " + ], + "image_footnote": [], + "bbox": [ + 259, + 258, + 738, + 332 + ], + "page_idx": 31 + }, + { + "type": "text", + "text": "N.2 MULTI-INTERPOLATION", + "text_level": 1, + "bbox": [ + 174, + 406, + 385, + 421 + ], + "page_idx": 31 + }, + { + "type": "text", + "text": "Figure 33 provides mixed interpolation between multiple attributes. We observe that any interpolation with any attribute is possible. ", + "bbox": [ + 174, + 433, + 821, + 462 + ], + "page_idx": 31 + }, + { + "type": "text", + "text": "N.3 MORE RESULTS ON ALL DATASETS ", + "text_level": 1, + "bbox": [ + 176, + 478, + 457, + 492 + ], + "page_idx": 31 + }, + { + "type": "text", + "text": "We provide more results on CelebA-HQ (Figure 34), LSUN-church (Figure 35), AFHQ, LSUNbedroom, METFACES (Figure 36). ", + "bbox": [ + 173, + 503, + 823, + 534 + ], + "page_idx": 31 + }, + { + "type": "image", + "img_path": "images/b9bf01755a1cba0600b8b0a808d391a7b78c8c875350410552a6a08c44d09b8a.jpg", + "image_caption": [ + "$\\Delta h$ of smiling and young. " + ], + "image_footnote": [], + "bbox": [ + 264, + 116, + 728, + 897 + ], + "page_idx": 32 + }, + { + "type": "image", + "img_path": "images/e936bdacea2d549661e4189967aeb6d92e07c0baab9c7e9b2fce8aea9a2adbf3.jpg", + "image_caption": [], + "image_footnote": [], + "bbox": [ + 173, + 213, + 823, + 486 + ], + "page_idx": 33 + }, + { + "type": "image", + "img_path": "images/377b9520a7229fe39d63b39397e4df0928ea29df4c7531e407e9ec506cc84359.jpg", + "image_caption": [ + "Figure 34: We provide more results on CelebA-HQ dataset. " + ], + "image_footnote": [], + "bbox": [ + 174, + 497, + 823, + 768 + ], + "page_idx": 33 + }, + { + "type": "image", + "img_path": "images/23c8ad00b208a9a83c86351a6572870e2502164c093437e52becf10d8cf7b5ae.jpg", + "image_caption": [ + "Figure 35: We provide more results on LSUN-church dataset. " + ], + "image_footnote": [], + "bbox": [ + 174, + 133, + 820, + 462 + ], + "page_idx": 34 + }, + { + "type": "image", + "img_path": "images/01bb6a5ce5257428dc08aae084bd10e960340eda2d2141d0c31b80453e60a4df.jpg", + "image_caption": [ + "Figure 36: We provide more results on AFHQ, LSUN-bedroom and METFACES datasets respectively. " + ], + "image_footnote": [], + "bbox": [ + 189, + 571, + 812, + 840 + ], + "page_idx": 34 + } +] \ No newline at end of file diff --git a/parse/dev/pd1P2eUBVfq/pd1P2eUBVfq_middle.json b/parse/dev/pd1P2eUBVfq/pd1P2eUBVfq_middle.json new file mode 100644 index 0000000000000000000000000000000000000000..9720907e0c1c9bd55823940e19ea21123eb57dda --- /dev/null +++ b/parse/dev/pd1P2eUBVfq/pd1P2eUBVfq_middle.json @@ -0,0 +1,76491 @@ +{ + "pdf_info": [ + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 103, + 377, + 140 + ], + "lines": [ + { + "bbox": [ + 105, + 102, + 379, + 121 + ], + "spans": [ + { + "bbox": [ + 105, + 102, + 379, + 121 + ], + "score": 1.0, + "content": "DIFFUSION MODELS ALREADY HAVE", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 123, + 318, + 141 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 318, + 141 + ], + "score": 1.0, + "content": "A SEMANTIC LATENT SPACE", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 114, + 159, + 302, + 171 + ], + "lines": [ + { + "bbox": [ + 112, + 158, + 304, + 174 + ], + "spans": [ + { + "bbox": [ + 112, + 158, + 304, + 174 + ], + "score": 1.0, + "content": "Mingi Kwon, Jaeseok Jeong, Youngjung Uh∗", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 2 + }, + { + "type": "text", + "bbox": [ + 113, + 172, + 354, + 214 + ], + "lines": [ + { + "bbox": [ + 111, + 169, + 261, + 182 + ], + "spans": [ + { + "bbox": [ + 111, + 169, + 261, + 182 + ], + "score": 1.0, + "content": "Department of Artificial Intelligence", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 112, + 180, + 187, + 194 + ], + "spans": [ + { + "bbox": [ + 112, + 180, + 187, + 194 + ], + "score": 1.0, + "content": "Yonsei University", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 111, + 191, + 217, + 204 + ], + "spans": [ + { + "bbox": [ + 111, + 191, + 217, + 204 + ], + "score": 1.0, + "content": "Seoul, Republic of Korea", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 112, + 203, + 356, + 216 + ], + "spans": [ + { + "bbox": [ + 112, + 203, + 356, + 216 + ], + "score": 1.0, + "content": "{kwonmingi,jete jeong,yj.uh}@yonsei.ac.kr", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + }, + { + "type": "title", + "bbox": [ + 278, + 244, + 333, + 255 + ], + "lines": [ + { + "bbox": [ + 276, + 242, + 336, + 258 + ], + "spans": [ + { + "bbox": [ + 276, + 242, + 336, + 258 + ], + "score": 1.0, + "content": "ABSTRACT", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 143, + 269, + 468, + 401 + ], + "lines": [ + { + "bbox": [ + 141, + 268, + 469, + 281 + ], + "spans": [ + { + "bbox": [ + 141, + 268, + 469, + 281 + ], + "score": 1.0, + "content": "Diffusion models achieve outstanding generative performance in various domains.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 141, + 280, + 469, + 293 + ], + "spans": [ + { + "bbox": [ + 141, + 280, + 469, + 293 + ], + "score": 1.0, + "content": "Despite their great success, they lack semantic latent space which is essen-", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 141, + 290, + 469, + 304 + ], + "spans": [ + { + "bbox": [ + 141, + 290, + 469, + 304 + ], + "score": 1.0, + "content": "tial for controlling the generative process. To address the problem, we propose", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 302, + 470, + 315 + ], + "spans": [ + { + "bbox": [ + 141, + 302, + 470, + 315 + ], + "score": 1.0, + "content": "asymmetric reverse process (Asyrp) which discovers the semantic latent space", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 312, + 470, + 326 + ], + "spans": [ + { + "bbox": [ + 141, + 312, + 433, + 326 + ], + "score": 1.0, + "content": "in frozen pretrained diffusion models. Our semantic latent space, named", + "type": "text" + }, + { + "bbox": [ + 433, + 313, + 439, + 323 + ], + "score": 0.64, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 440, + 312, + 470, + 326 + ], + "score": 1.0, + "content": "-space,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 324, + 470, + 337 + ], + "spans": [ + { + "bbox": [ + 141, + 324, + 470, + 337 + ], + "score": 1.0, + "content": "has nice properties to accommodate semantic image manipulation: homogeneity,", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 335, + 470, + 347 + ], + "spans": [ + { + "bbox": [ + 141, + 335, + 470, + 347 + ], + "score": 1.0, + "content": "linearity, robustness, and consistency across timesteps. In addition, we introduce a", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 347, + 469, + 358 + ], + "spans": [ + { + "bbox": [ + 141, + 347, + 469, + 358 + ], + "score": 1.0, + "content": "principled design of the generative process for versatile editing and quality boost-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 357, + 470, + 370 + ], + "spans": [ + { + "bbox": [ + 141, + 357, + 470, + 370 + ], + "score": 1.0, + "content": "ing by quantifiable measures: editing strength of an interval and quality deficiency", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 368, + 469, + 380 + ], + "spans": [ + { + "bbox": [ + 141, + 368, + 406, + 380 + ], + "score": 1.0, + "content": "at a timestep. Our method is applicable to various architectures", + "type": "text" + }, + { + "bbox": [ + 406, + 368, + 449, + 379 + ], + "score": 0.65, + "content": "\\mathrm { ( D D P M + + }", + "type": "inline_equation" + }, + { + "bbox": [ + 450, + 368, + 469, + 380 + ], + "score": 1.0, + "content": ", iD-", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 378, + 470, + 392 + ], + "spans": [ + { + "bbox": [ + 141, + 378, + 470, + 392 + ], + "score": 1.0, + "content": "DPM, and ADM) and datasets (CelebA-HQ, AFHQ-dog, LSUN-church, LSUN-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 388, + 453, + 403 + ], + "spans": [ + { + "bbox": [ + 141, + 388, + 453, + 403 + ], + "score": 1.0, + "content": "bedroom, and METFACES). Project page: https://kwonminki.github.io/Asyrp/", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 13.5 + }, + { + "type": "title", + "bbox": [ + 108, + 423, + 205, + 435 + ], + "lines": [ + { + "bbox": [ + 105, + 422, + 208, + 439 + ], + "spans": [ + { + "bbox": [ + 105, + 422, + 208, + 439 + ], + "score": 1.0, + "content": "1 INTRODUCTION", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 448, + 505, + 547 + ], + "lines": [ + { + "bbox": [ + 106, + 449, + 504, + 461 + ], + "spans": [ + { + "bbox": [ + 106, + 449, + 504, + 461 + ], + "score": 1.0, + "content": "In image synthesis, diffusion models have advanced to achieve state-of-the-art performance regard-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 459, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 505, + 472 + ], + "score": 1.0, + "content": "ing quality and mode coverage since the introduction of denoising diffusion probabilistic models", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 470, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 505, + 484 + ], + "score": 1.0, + "content": "(Ho et al., 2020). They disrupt images by adding noise through multiple steps of forward process", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 481, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 481, + 505, + 495 + ], + "score": 1.0, + "content": "and generate samples by progressive denoising through multiple steps of reverse (i.e., generative)", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 493, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 505, + 506 + ], + "score": 1.0, + "content": "process. Since their deterministic version provides nearly perfect reconstruction of original images", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 504, + 505, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 504, + 505, + 516 + ], + "score": 1.0, + "content": "(Song et al., 2020a), they are suitable for image editing, which renders target attributes on the real", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 514, + 504, + 527 + ], + "spans": [ + { + "bbox": [ + 105, + 514, + 504, + 527 + ], + "score": 1.0, + "content": "images. However, simply editing the latent variables (i.e., intermediate noisy images) causes de-", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 104, + 524, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 104, + 524, + 506, + 540 + ], + "score": 1.0, + "content": "graded results (Kim & Ye, 2021). Instead, they require complicated procedures: providing guidance", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 537, + 343, + 549 + ], + "spans": [ + { + "bbox": [ + 106, + 537, + 343, + 549 + ], + "score": 1.0, + "content": "in the reverse process or finetuning models for an attribute.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 107, + 553, + 505, + 663 + ], + "lines": [ + { + "bbox": [ + 105, + 553, + 505, + 566 + ], + "spans": [ + { + "bbox": [ + 105, + 553, + 505, + 566 + ], + "score": 1.0, + "content": "Figure 1(a-c) briefly illustrates the existing approaches. Image guidance mixes the latent variables", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "score": 1.0, + "content": "of the guiding image with unconditional latent variables (Choi et al., 2021; Lugmayr et al., 2022;", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 575, + 506, + 587 + ], + "spans": [ + { + "bbox": [ + 105, + 575, + 506, + 587 + ], + "score": 1.0, + "content": "Meng et al., 2021). Though it provides some control, it is ambiguous to specify which attribute to", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 586, + 505, + 599 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 505, + 599 + ], + "score": 1.0, + "content": "reflect among the ones in the guide and the unconditional result, and it lacks intuitive control for the", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 104, + 596, + 506, + 610 + ], + "spans": [ + { + "bbox": [ + 104, + 596, + 506, + 610 + ], + "score": 1.0, + "content": "magnitude of change. Classifier guidance manipulates images by imposing gradients of a classifier", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 608, + 505, + 621 + ], + "spans": [ + { + "bbox": [ + 105, + 608, + 505, + 621 + ], + "score": 1.0, + "content": "on the latent variables in the reverse process to match the target class (Dhariwal & Nichol, 2021;", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 618, + 506, + 632 + ], + "spans": [ + { + "bbox": [ + 105, + 618, + 506, + 632 + ], + "score": 1.0, + "content": "Avrahami et al., 2022; Liu et al., 2021). It requires training an extra classifier for the latent variables,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 630, + 505, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 505, + 643 + ], + "score": 1.0, + "content": "i.e., noisy images. Furthermore, computing gradients through the classifier during sampling is costly.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 640, + 505, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 640, + 505, + 654 + ], + "score": 1.0, + "content": "Finetuning the whole model can steer the resulting images to the target attribute without the above", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 652, + 481, + 665 + ], + "spans": [ + { + "bbox": [ + 105, + 652, + 481, + 665 + ], + "score": 1.0, + "content": "problems (Kim & Ye, 2021). Still, it requires multiple models to reflect multiple descriptions.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 34.5 + }, + { + "type": "text", + "bbox": [ + 107, + 669, + 504, + 713 + ], + "lines": [ + { + "bbox": [ + 106, + 669, + 505, + 681 + ], + "spans": [ + { + "bbox": [ + 106, + 669, + 505, + 681 + ], + "score": 1.0, + "content": "On the other hand, generative adversarial networks (Goodfellow et al., 2020) inherently provide", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 680, + 505, + 692 + ], + "spans": [ + { + "bbox": [ + 105, + 680, + 505, + 692 + ], + "score": 1.0, + "content": "straightforward image editing in their latent space. Given a latent vector for an original image, we", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 691, + 505, + 703 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 505, + 703 + ], + "score": 1.0, + "content": "can find the direction in the latent space that maximizes the similarity of the resulting image with", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 702, + 506, + 714 + ], + "spans": [ + { + "bbox": [ + 105, + 702, + 506, + 714 + ], + "score": 1.0, + "content": "a target description in CLIP embedding (Patashnik et al., 2021). The latent direction found on one", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 41.5 + } + ], + "page_idx": 0, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 119, + 722, + 200, + 731 + ], + "lines": [ + { + "bbox": [ + 119, + 720, + 202, + 734 + ], + "spans": [ + { + "bbox": [ + 119, + 720, + 202, + 734 + ], + "score": 1.0, + "content": "∗corresponding author", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "score": 1.0, + "content": "1", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 103, + 377, + 140 + ], + "lines": [ + { + "bbox": [ + 105, + 102, + 379, + 121 + ], + "spans": [ + { + "bbox": [ + 105, + 102, + 379, + 121 + ], + "score": 1.0, + "content": "DIFFUSION MODELS ALREADY HAVE", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 123, + 318, + 141 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 318, + 141 + ], + "score": 1.0, + "content": "A SEMANTIC LATENT SPACE", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 114, + 159, + 302, + 171 + ], + "lines": [ + { + "bbox": [ + 112, + 158, + 304, + 174 + ], + "spans": [ + { + "bbox": [ + 112, + 158, + 304, + 174 + ], + "score": 1.0, + "content": "Mingi Kwon, Jaeseok Jeong, Youngjung Uh∗", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 2, + "bbox_fs": [ + 112, + 158, + 304, + 174 + ] + }, + { + "type": "list", + "bbox": [ + 113, + 172, + 354, + 214 + ], + "lines": [ + { + "bbox": [ + 111, + 169, + 261, + 182 + ], + "spans": [ + { + "bbox": [ + 111, + 169, + 261, + 182 + ], + "score": 1.0, + "content": "Department of Artificial Intelligence", + "type": "text" + } + ], + "index": 3, + "is_list_start_line": true + }, + { + "bbox": [ + 112, + 180, + 187, + 194 + ], + "spans": [ + { + "bbox": [ + 112, + 180, + 187, + 194 + ], + "score": 1.0, + "content": "Yonsei University", + "type": "text" + } + ], + "index": 4, + "is_list_start_line": true + }, + { + "bbox": [ + 111, + 191, + 217, + 204 + ], + "spans": [ + { + "bbox": [ + 111, + 191, + 217, + 204 + ], + "score": 1.0, + "content": "Seoul, Republic of Korea", + "type": "text" + } + ], + "index": 5, + "is_list_start_line": true + }, + { + "bbox": [ + 112, + 203, + 356, + 216 + ], + "spans": [ + { + "bbox": [ + 112, + 203, + 356, + 216 + ], + "score": 1.0, + "content": "{kwonmingi,jete jeong,yj.uh}@yonsei.ac.kr", + "type": "text" + } + ], + "index": 6, + "is_list_start_line": true + } + ], + "index": 4.5, + "bbox_fs": [ + 111, + 169, + 356, + 216 + ] + }, + { + "type": "title", + "bbox": [ + 278, + 244, + 333, + 255 + ], + "lines": [ + { + "bbox": [ + 276, + 242, + 336, + 258 + ], + "spans": [ + { + "bbox": [ + 276, + 242, + 336, + 258 + ], + "score": 1.0, + "content": "ABSTRACT", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 143, + 269, + 468, + 401 + ], + "lines": [ + { + "bbox": [ + 141, + 268, + 469, + 281 + ], + "spans": [ + { + "bbox": [ + 141, + 268, + 469, + 281 + ], + "score": 1.0, + "content": "Diffusion models achieve outstanding generative performance in various domains.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 141, + 280, + 469, + 293 + ], + "spans": [ + { + "bbox": [ + 141, + 280, + 469, + 293 + ], + "score": 1.0, + "content": "Despite their great success, they lack semantic latent space which is essen-", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 141, + 290, + 469, + 304 + ], + "spans": [ + { + "bbox": [ + 141, + 290, + 469, + 304 + ], + "score": 1.0, + "content": "tial for controlling the generative process. To address the problem, we propose", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 302, + 470, + 315 + ], + "spans": [ + { + "bbox": [ + 141, + 302, + 470, + 315 + ], + "score": 1.0, + "content": "asymmetric reverse process (Asyrp) which discovers the semantic latent space", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 312, + 470, + 326 + ], + "spans": [ + { + "bbox": [ + 141, + 312, + 433, + 326 + ], + "score": 1.0, + "content": "in frozen pretrained diffusion models. Our semantic latent space, named", + "type": "text" + }, + { + "bbox": [ + 433, + 313, + 439, + 323 + ], + "score": 0.64, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 440, + 312, + 470, + 326 + ], + "score": 1.0, + "content": "-space,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 324, + 470, + 337 + ], + "spans": [ + { + "bbox": [ + 141, + 324, + 470, + 337 + ], + "score": 1.0, + "content": "has nice properties to accommodate semantic image manipulation: homogeneity,", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 335, + 470, + 347 + ], + "spans": [ + { + "bbox": [ + 141, + 335, + 470, + 347 + ], + "score": 1.0, + "content": "linearity, robustness, and consistency across timesteps. In addition, we introduce a", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 347, + 469, + 358 + ], + "spans": [ + { + "bbox": [ + 141, + 347, + 469, + 358 + ], + "score": 1.0, + "content": "principled design of the generative process for versatile editing and quality boost-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 357, + 470, + 370 + ], + "spans": [ + { + "bbox": [ + 141, + 357, + 470, + 370 + ], + "score": 1.0, + "content": "ing by quantifiable measures: editing strength of an interval and quality deficiency", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 368, + 469, + 380 + ], + "spans": [ + { + "bbox": [ + 141, + 368, + 406, + 380 + ], + "score": 1.0, + "content": "at a timestep. Our method is applicable to various architectures", + "type": "text" + }, + { + "bbox": [ + 406, + 368, + 449, + 379 + ], + "score": 0.65, + "content": "\\mathrm { ( D D P M + + }", + "type": "inline_equation" + }, + { + "bbox": [ + 450, + 368, + 469, + 380 + ], + "score": 1.0, + "content": ", iD-", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 378, + 470, + 392 + ], + "spans": [ + { + "bbox": [ + 141, + 378, + 470, + 392 + ], + "score": 1.0, + "content": "DPM, and ADM) and datasets (CelebA-HQ, AFHQ-dog, LSUN-church, LSUN-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 388, + 453, + 403 + ], + "spans": [ + { + "bbox": [ + 141, + 388, + 453, + 403 + ], + "score": 1.0, + "content": "bedroom, and METFACES). Project page: https://kwonminki.github.io/Asyrp/", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 13.5, + "bbox_fs": [ + 141, + 268, + 470, + 403 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 423, + 205, + 435 + ], + "lines": [ + { + "bbox": [ + 105, + 422, + 208, + 439 + ], + "spans": [ + { + "bbox": [ + 105, + 422, + 208, + 439 + ], + "score": 1.0, + "content": "1 INTRODUCTION", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 448, + 505, + 547 + ], + "lines": [ + { + "bbox": [ + 106, + 449, + 504, + 461 + ], + "spans": [ + { + "bbox": [ + 106, + 449, + 504, + 461 + ], + "score": 1.0, + "content": "In image synthesis, diffusion models have advanced to achieve state-of-the-art performance regard-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 459, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 505, + 472 + ], + "score": 1.0, + "content": "ing quality and mode coverage since the introduction of denoising diffusion probabilistic models", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 470, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 505, + 484 + ], + "score": 1.0, + "content": "(Ho et al., 2020). They disrupt images by adding noise through multiple steps of forward process", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 481, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 481, + 505, + 495 + ], + "score": 1.0, + "content": "and generate samples by progressive denoising through multiple steps of reverse (i.e., generative)", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 493, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 505, + 506 + ], + "score": 1.0, + "content": "process. Since their deterministic version provides nearly perfect reconstruction of original images", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 504, + 505, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 504, + 505, + 516 + ], + "score": 1.0, + "content": "(Song et al., 2020a), they are suitable for image editing, which renders target attributes on the real", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 514, + 504, + 527 + ], + "spans": [ + { + "bbox": [ + 105, + 514, + 504, + 527 + ], + "score": 1.0, + "content": "images. However, simply editing the latent variables (i.e., intermediate noisy images) causes de-", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 104, + 524, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 104, + 524, + 506, + 540 + ], + "score": 1.0, + "content": "graded results (Kim & Ye, 2021). Instead, they require complicated procedures: providing guidance", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 537, + 343, + 549 + ], + "spans": [ + { + "bbox": [ + 106, + 537, + 343, + 549 + ], + "score": 1.0, + "content": "in the reverse process or finetuning models for an attribute.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 25, + "bbox_fs": [ + 104, + 449, + 506, + 549 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 553, + 505, + 663 + ], + "lines": [ + { + "bbox": [ + 105, + 553, + 505, + 566 + ], + "spans": [ + { + "bbox": [ + 105, + 553, + 505, + 566 + ], + "score": 1.0, + "content": "Figure 1(a-c) briefly illustrates the existing approaches. Image guidance mixes the latent variables", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 505, + 577 + ], + "score": 1.0, + "content": "of the guiding image with unconditional latent variables (Choi et al., 2021; Lugmayr et al., 2022;", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 575, + 506, + 587 + ], + "spans": [ + { + "bbox": [ + 105, + 575, + 506, + 587 + ], + "score": 1.0, + "content": "Meng et al., 2021). Though it provides some control, it is ambiguous to specify which attribute to", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 586, + 505, + 599 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 505, + 599 + ], + "score": 1.0, + "content": "reflect among the ones in the guide and the unconditional result, and it lacks intuitive control for the", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 104, + 596, + 506, + 610 + ], + "spans": [ + { + "bbox": [ + 104, + 596, + 506, + 610 + ], + "score": 1.0, + "content": "magnitude of change. Classifier guidance manipulates images by imposing gradients of a classifier", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 608, + 505, + 621 + ], + "spans": [ + { + "bbox": [ + 105, + 608, + 505, + 621 + ], + "score": 1.0, + "content": "on the latent variables in the reverse process to match the target class (Dhariwal & Nichol, 2021;", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 618, + 506, + 632 + ], + "spans": [ + { + "bbox": [ + 105, + 618, + 506, + 632 + ], + "score": 1.0, + "content": "Avrahami et al., 2022; Liu et al., 2021). It requires training an extra classifier for the latent variables,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 630, + 505, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 505, + 643 + ], + "score": 1.0, + "content": "i.e., noisy images. Furthermore, computing gradients through the classifier during sampling is costly.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 640, + 505, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 640, + 505, + 654 + ], + "score": 1.0, + "content": "Finetuning the whole model can steer the resulting images to the target attribute without the above", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 652, + 481, + 665 + ], + "spans": [ + { + "bbox": [ + 105, + 652, + 481, + 665 + ], + "score": 1.0, + "content": "problems (Kim & Ye, 2021). Still, it requires multiple models to reflect multiple descriptions.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 34.5, + "bbox_fs": [ + 104, + 553, + 506, + 665 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 669, + 504, + 713 + ], + "lines": [ + { + "bbox": [ + 106, + 669, + 505, + 681 + ], + "spans": [ + { + "bbox": [ + 106, + 669, + 505, + 681 + ], + "score": 1.0, + "content": "On the other hand, generative adversarial networks (Goodfellow et al., 2020) inherently provide", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 680, + 505, + 692 + ], + "spans": [ + { + "bbox": [ + 105, + 680, + 505, + 692 + ], + "score": 1.0, + "content": "straightforward image editing in their latent space. Given a latent vector for an original image, we", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 691, + 505, + 703 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 505, + 703 + ], + "score": 1.0, + "content": "can find the direction in the latent space that maximizes the similarity of the resulting image with", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 702, + 506, + 714 + ], + "spans": [ + { + "bbox": [ + 105, + 702, + 506, + 714 + ], + "score": 1.0, + "content": "a target description in CLIP embedding (Patashnik et al., 2021). The latent direction found on one", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 274, + 504, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 274, + 504, + 287 + ], + "score": 1.0, + "content": "image leads to the same manipulation of other images. However, given a real image, finding its exact", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 284, + 426, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 284, + 426, + 298 + ], + "score": 1.0, + "content": "latent vector is often challenging and produces unexpected appearance changes.", + "type": "text", + "cross_page": true + } + ], + "index": 8 + } + ], + "index": 41.5, + "bbox_fs": [ + 105, + 669, + 506, + 714 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 117, + 81, + 496, + 207 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 117, + 81, + 496, + 207 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 117, + 81, + 496, + 207 + ], + "spans": [ + { + "bbox": [ + 117, + 81, + 496, + 207 + ], + "score": 0.971, + "type": "image", + "image_path": "80648f78ecfc7ad5b9897cecadcaf4711f94bdafc585f1c639190f153bb4ad3d.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 117, + 81, + 496, + 123.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 117, + 123.0, + 496, + 165.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 117, + 165.0, + 496, + 207.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 217, + 505, + 262 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 217, + 504, + 230 + ], + "spans": [ + { + "bbox": [ + 105, + 217, + 504, + 230 + ], + "score": 1.0, + "content": "Figure 1: Manipulation approaches for diffusion models. (a) Image guidance suffers ambiguity", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 228, + 505, + 241 + ], + "spans": [ + { + "bbox": [ + 105, + 228, + 505, + 241 + ], + "score": 1.0, + "content": "while controlling the generative process. (b) Classifier guidance requires an extra classifier, is hardly", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 240, + 505, + 252 + ], + "spans": [ + { + "bbox": [ + 106, + 240, + 505, + 252 + ], + "score": 1.0, + "content": "editable, degrades quality, or alters the content. (c) DiffusionCLIP requires fine-tuning the whole", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 251, + 447, + 262 + ], + "spans": [ + { + "bbox": [ + 106, + 251, + 447, + 262 + ], + "score": 1.0, + "content": "model. (d) Our method discovers a semantic latent space of a frozen diffusion model.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "text", + "bbox": [ + 107, + 274, + 502, + 297 + ], + "lines": [ + { + "bbox": [ + 106, + 274, + 504, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 274, + 504, + 287 + ], + "score": 1.0, + "content": "image leads to the same manipulation of other images. However, given a real image, finding its exact", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 284, + 426, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 284, + 426, + 298 + ], + "score": 1.0, + "content": "latent vector is often challenging and produces unexpected appearance changes.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 107, + 302, + 505, + 357 + ], + "lines": [ + { + "bbox": [ + 105, + 302, + 505, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 302, + 505, + 314 + ], + "score": 1.0, + "content": "It would allow admirable image editing if the diffusion models with the nearly perfect inversion", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 313, + 505, + 325 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 505, + 325 + ], + "score": 1.0, + "content": "property have such a semantic latent space. Preechakul et al. (2022) introduces an additional input to", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 325, + 505, + 336 + ], + "spans": [ + { + "bbox": [ + 106, + 325, + 505, + 336 + ], + "score": 1.0, + "content": "the reverse diffusion process: a latent vector from an original image embedded by an extra encoder.", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 335, + 505, + 347 + ], + "spans": [ + { + "bbox": [ + 105, + 335, + 505, + 347 + ], + "score": 1.0, + "content": "This latent vector contains the semantics to condition the process. However, it requires training from", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 347, + 352, + 358 + ], + "spans": [ + { + "bbox": [ + 106, + 347, + 352, + 358 + ], + "score": 1.0, + "content": "scratch and does not match with pretrained diffusion models.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 106, + 363, + 505, + 516 + ], + "lines": [ + { + "bbox": [ + 105, + 363, + 506, + 376 + ], + "spans": [ + { + "bbox": [ + 105, + 363, + 506, + 376 + ], + "score": 1.0, + "content": "In this paper, we propose an asymmetric reverse process (Asyrp) which discovers the semantic latent", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 373, + 505, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 373, + 505, + 387 + ], + "score": 1.0, + "content": "space of a frozen diffusion model such that modifications in the space edits attributes of the original", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 385, + 504, + 397 + ], + "spans": [ + { + "bbox": [ + 106, + 386, + 276, + 397 + ], + "score": 1.0, + "content": "images. Our semantic latent space, named", + "type": "text" + }, + { + "bbox": [ + 276, + 385, + 282, + 395 + ], + "score": 0.66, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 283, + 386, + 504, + 397 + ], + "score": 1.0, + "content": "-space, has the properties necessary for editing applica-", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 396, + 506, + 408 + ], + "spans": [ + { + "bbox": [ + 105, + 396, + 506, + 408 + ], + "score": 1.0, + "content": "tions as follows. The same shift in this space results in the same attribute change in all images. Linear", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 406, + 506, + 419 + ], + "spans": [ + { + "bbox": [ + 105, + 406, + 506, + 419 + ], + "score": 1.0, + "content": "changes in this space lead to linear changes in attributes. The changes do not degrade the quality of", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 417, + 506, + 430 + ], + "spans": [ + { + "bbox": [ + 105, + 417, + 506, + 430 + ], + "score": 1.0, + "content": "the resulting images. The changes throughout the timesteps are almost identical to each other for a", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 428, + 505, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 428, + 412, + 441 + ], + "score": 1.0, + "content": "desired attribute change. Figure 1(d) illustrates some of these properties and", + "type": "text" + }, + { + "bbox": [ + 412, + 429, + 433, + 440 + ], + "score": 0.85, + "content": "\\ S 5 . 3", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 428, + 505, + 441 + ], + "score": 1.0, + "content": "provides detailed", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 440, + 504, + 452 + ], + "spans": [ + { + "bbox": [ + 106, + 440, + 504, + 452 + ], + "score": 1.0, + "content": "analyses. To the best of our knowledge, it is the first attempt to discover the semantic latent space in", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 450, + 506, + 463 + ], + "spans": [ + { + "bbox": [ + 105, + 450, + 506, + 463 + ], + "score": 1.0, + "content": "the frozen pretrained diffusion models. Spoiler alert: our semantic latent space is different from the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 460, + 505, + 475 + ], + "spans": [ + { + "bbox": [ + 105, + 460, + 505, + 475 + ], + "score": 1.0, + "content": "intermediate latent variables in the diffusion process. Moreover, we introduce a principled design", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 471, + 505, + 485 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 505, + 485 + ], + "score": 1.0, + "content": "of the generative process for versatile editing and quality boosting by quantifiable measures: edit-", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 483, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 506, + 496 + ], + "score": 1.0, + "content": "ing strength of an interval and quality deficiency at a timestep. Extensive experiments demonstrate", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 493, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 360, + 507 + ], + "score": 1.0, + "content": "that our method is generally applicable to various architectures", + "type": "text" + }, + { + "bbox": [ + 360, + 495, + 402, + 505 + ], + "score": 0.62, + "content": "\\mathrm { ( D D P M + + }", + "type": "inline_equation" + }, + { + "bbox": [ + 403, + 493, + 506, + 507 + ], + "score": 1.0, + "content": ", iDDPM, and ADM) and", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 505, + 452, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 452, + 518 + ], + "score": 1.0, + "content": "datasets (CelebA-HQ, AFHQ-dog, LSUN-church, LSUN-bedroom, and METFACES).", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 20.5 + }, + { + "type": "title", + "bbox": [ + 108, + 532, + 200, + 545 + ], + "lines": [ + { + "bbox": [ + 104, + 530, + 201, + 548 + ], + "spans": [ + { + "bbox": [ + 104, + 530, + 201, + 548 + ], + "score": 1.0, + "content": "2 BACKGROUND", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 106, + 556, + 500, + 568 + ], + "lines": [ + { + "bbox": [ + 106, + 555, + 501, + 569 + ], + "spans": [ + { + "bbox": [ + 106, + 555, + 501, + 569 + ], + "score": 1.0, + "content": "We briefly describe essential backgrounds. The rest of the related work is deferred to Appendix A.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "title", + "bbox": [ + 107, + 580, + 368, + 592 + ], + "lines": [ + { + "bbox": [ + 105, + 579, + 369, + 595 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 369, + 595 + ], + "score": 1.0, + "content": "2.1 DENOISING DIFFUSION PROBABILITY MODEL (DDPM)", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 106, + 601, + 506, + 634 + ], + "lines": [ + { + "bbox": [ + 105, + 600, + 505, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 600, + 505, + 613 + ], + "score": 1.0, + "content": "DDPM is a latent variable model that learns a data distribution by denoising noisy images (Ho et al.,", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 611, + 506, + 626 + ], + "spans": [ + { + "bbox": [ + 105, + 611, + 506, + 626 + ], + "score": 1.0, + "content": "2020). The forward process diffuses the data samples through Gaussian transitions parameterized", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 622, + 203, + 636 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 203, + 636 + ], + "score": 1.0, + "content": "with a Markov process:", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32 + }, + { + "type": "interline_equation", + "bbox": [ + 133, + 633, + 477, + 661 + ], + "lines": [ + { + "bbox": [ + 133, + 633, + 477, + 661 + ], + "spans": [ + { + "bbox": [ + 133, + 633, + 477, + 661 + ], + "score": 0.92, + "content": "q \\left( \\pmb { x } _ { t } \\mid \\pmb { \\dot { x } } _ { t - 1 } \\right) = \\mathcal { N } \\left( \\pmb { x } _ { t } ; \\sqrt { 1 - \\beta _ { t } } \\pmb { x } _ { t - 1 } , \\beta _ { t } \\mathbf { I } \\right) = \\mathcal { N } \\left( \\sqrt { \\frac { \\alpha _ { t } } { \\alpha _ { t - 1 } } } \\pmb { x } _ { t - 1 } , \\left( 1 - \\frac { \\alpha _ { t } } { \\alpha _ { t - 1 } } \\right) \\pmb { I } \\right) ,", + "type": "interline_equation", + "image_path": "035cdb05d9c2f6d50e510faab182ddb5c1138b8f9e837a09f7b8bf8817ac6c24.jpg" + } + ] + } + ], + "index": 35, + "virtual_lines": [ + { + "bbox": [ + 133, + 633, + 477, + 642.3333333333334 + ], + "spans": [], + "index": 34 + }, + { + "bbox": [ + 133, + 642.3333333333334, + 477, + 651.6666666666667 + ], + "spans": [], + "index": 35 + }, + { + "bbox": [ + 133, + 651.6666666666667, + 477, + 661.0000000000001 + ], + "spans": [], + "index": 36 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 663, + 508, + 690 + ], + "lines": [ + { + "bbox": [ + 105, + 662, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 662, + 133, + 678 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 133, + 663, + 167, + 677 + ], + "score": 0.93, + "content": "\\{ \\beta _ { t } \\} _ { t = 1 } ^ { T }", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 662, + 281, + 678 + ], + "score": 1.0, + "content": "is the variance schedule and", + "type": "text" + }, + { + "bbox": [ + 282, + 663, + 364, + 677 + ], + "score": 0.93, + "content": "\\begin{array} { r } { \\alpha _ { t } = \\prod _ { s = 1 } ^ { t } ( 1 - \\beta _ { s } ) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 364, + 662, + 506, + 678 + ], + "score": 1.0, + "content": ". Then the reverse process becomes", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 675, + 488, + 693 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 273, + 690 + ], + "score": 0.9, + "content": "\\begin{array} { r } { p _ { \\theta } \\left( \\pmb { x } _ { 0 : T } \\right) : = p \\left( \\pmb { x } _ { T } \\right) \\prod _ { t = 1 } ^ { T } p _ { \\theta } \\left( \\pmb { x } _ { t - 1 } \\mid \\pmb { x } _ { t } \\right) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 273, + 675, + 331, + 693 + ], + "score": 1.0, + "content": ", starting from", + "type": "text" + }, + { + "bbox": [ + 331, + 678, + 389, + 690 + ], + "score": 0.92, + "content": "\\mathbf { \\boldsymbol { x } } _ { T } \\sim \\mathcal { N } ( 0 , \\mathbf { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 389, + 675, + 473, + 693 + ], + "score": 1.0, + "content": "with noise predictor", + "type": "text" + }, + { + "bbox": [ + 473, + 677, + 483, + 690 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 483, + 675, + 488, + 693 + ], + "score": 1.0, + "content": ":", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 37.5 + }, + { + "type": "interline_equation", + "bbox": [ + 199, + 691, + 411, + 719 + ], + "lines": [ + { + "bbox": [ + 199, + 691, + 411, + 719 + ], + "spans": [ + { + "bbox": [ + 199, + 691, + 411, + 719 + ], + "score": 0.93, + "content": "{ \\pmb x } _ { t - 1 } = \\frac { 1 } { \\sqrt { 1 - \\beta _ { t } } } \\left( { \\pmb x } _ { t } - \\frac { \\beta _ { t } } { \\sqrt { 1 - \\alpha _ { t } } } { \\pmb \\epsilon } _ { t } ^ { \\theta } \\left( { \\pmb x } _ { t } \\right) \\right) + \\sigma _ { t } { \\pmb z } _ { t } ,", + "type": "interline_equation", + "image_path": "d7055f0c4357778eb781035893f1f1562f0810b6f8ac46e9dc8ddb106fdd7fcd.jpg" + } + ] + } + ], + "index": 39.5, + "virtual_lines": [ + { + "bbox": [ + 199, + 691, + 411, + 705.0 + ], + "spans": [], + "index": 39 + }, + { + "bbox": [ + 199, + 705.0, + 411, + 719.0 + ], + "spans": [], + "index": 40 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 720, + 498, + 733 + ], + "lines": [ + { + "bbox": [ + 106, + 718, + 500, + 734 + ], + "spans": [ + { + "bbox": [ + 106, + 718, + 133, + 734 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 133, + 720, + 188, + 732 + ], + "score": 0.93, + "content": "\\boldsymbol { z } _ { t } \\sim \\mathcal { N } ( 0 , \\mathbf { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 188, + 718, + 206, + 734 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 206, + 720, + 218, + 732 + ], + "score": 0.9, + "content": "\\sigma _ { t } ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 218, + 718, + 419, + 734 + ], + "score": 1.0, + "content": "is a variance of the reverse process which is set to", + "type": "text" + }, + { + "bbox": [ + 420, + 720, + 453, + 732 + ], + "score": 0.93, + "content": "\\sigma _ { t } ^ { 2 } = \\beta _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 454, + 718, + 500, + 734 + ], + "score": 1.0, + "content": "by DDPM.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 41 + } + ], + "page_idx": 1, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 763 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 763 + ], + "score": 1.0, + "content": "2", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 117, + 81, + 496, + 207 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 117, + 81, + 496, + 207 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 117, + 81, + 496, + 207 + ], + "spans": [ + { + "bbox": [ + 117, + 81, + 496, + 207 + ], + "score": 0.971, + "type": "image", + "image_path": "80648f78ecfc7ad5b9897cecadcaf4711f94bdafc585f1c639190f153bb4ad3d.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 117, + 81, + 496, + 123.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 117, + 123.0, + 496, + 165.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 117, + 165.0, + 496, + 207.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 217, + 505, + 262 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 217, + 504, + 230 + ], + "spans": [ + { + "bbox": [ + 105, + 217, + 504, + 230 + ], + "score": 1.0, + "content": "Figure 1: Manipulation approaches for diffusion models. (a) Image guidance suffers ambiguity", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 228, + 505, + 241 + ], + "spans": [ + { + "bbox": [ + 105, + 228, + 505, + 241 + ], + "score": 1.0, + "content": "while controlling the generative process. (b) Classifier guidance requires an extra classifier, is hardly", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 240, + 505, + 252 + ], + "spans": [ + { + "bbox": [ + 106, + 240, + 505, + 252 + ], + "score": 1.0, + "content": "editable, degrades quality, or alters the content. (c) DiffusionCLIP requires fine-tuning the whole", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 251, + 447, + 262 + ], + "spans": [ + { + "bbox": [ + 106, + 251, + 447, + 262 + ], + "score": 1.0, + "content": "model. (d) Our method discovers a semantic latent space of a frozen diffusion model.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "text", + "bbox": [ + 107, + 274, + 502, + 297 + ], + "lines": [], + "index": 7.5, + "bbox_fs": [ + 105, + 274, + 504, + 298 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 302, + 505, + 357 + ], + "lines": [ + { + "bbox": [ + 105, + 302, + 505, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 302, + 505, + 314 + ], + "score": 1.0, + "content": "It would allow admirable image editing if the diffusion models with the nearly perfect inversion", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 313, + 505, + 325 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 505, + 325 + ], + "score": 1.0, + "content": "property have such a semantic latent space. Preechakul et al. (2022) introduces an additional input to", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 325, + 505, + 336 + ], + "spans": [ + { + "bbox": [ + 106, + 325, + 505, + 336 + ], + "score": 1.0, + "content": "the reverse diffusion process: a latent vector from an original image embedded by an extra encoder.", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 335, + 505, + 347 + ], + "spans": [ + { + "bbox": [ + 105, + 335, + 505, + 347 + ], + "score": 1.0, + "content": "This latent vector contains the semantics to condition the process. However, it requires training from", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 347, + 352, + 358 + ], + "spans": [ + { + "bbox": [ + 106, + 347, + 352, + 358 + ], + "score": 1.0, + "content": "scratch and does not match with pretrained diffusion models.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 302, + 505, + 358 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 363, + 505, + 516 + ], + "lines": [ + { + "bbox": [ + 105, + 363, + 506, + 376 + ], + "spans": [ + { + "bbox": [ + 105, + 363, + 506, + 376 + ], + "score": 1.0, + "content": "In this paper, we propose an asymmetric reverse process (Asyrp) which discovers the semantic latent", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 373, + 505, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 373, + 505, + 387 + ], + "score": 1.0, + "content": "space of a frozen diffusion model such that modifications in the space edits attributes of the original", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 385, + 504, + 397 + ], + "spans": [ + { + "bbox": [ + 106, + 386, + 276, + 397 + ], + "score": 1.0, + "content": "images. Our semantic latent space, named", + "type": "text" + }, + { + "bbox": [ + 276, + 385, + 282, + 395 + ], + "score": 0.66, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 283, + 386, + 504, + 397 + ], + "score": 1.0, + "content": "-space, has the properties necessary for editing applica-", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 396, + 506, + 408 + ], + "spans": [ + { + "bbox": [ + 105, + 396, + 506, + 408 + ], + "score": 1.0, + "content": "tions as follows. The same shift in this space results in the same attribute change in all images. Linear", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 406, + 506, + 419 + ], + "spans": [ + { + "bbox": [ + 105, + 406, + 506, + 419 + ], + "score": 1.0, + "content": "changes in this space lead to linear changes in attributes. The changes do not degrade the quality of", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 417, + 506, + 430 + ], + "spans": [ + { + "bbox": [ + 105, + 417, + 506, + 430 + ], + "score": 1.0, + "content": "the resulting images. The changes throughout the timesteps are almost identical to each other for a", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 428, + 505, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 428, + 412, + 441 + ], + "score": 1.0, + "content": "desired attribute change. Figure 1(d) illustrates some of these properties and", + "type": "text" + }, + { + "bbox": [ + 412, + 429, + 433, + 440 + ], + "score": 0.85, + "content": "\\ S 5 . 3", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 428, + 505, + 441 + ], + "score": 1.0, + "content": "provides detailed", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 440, + 504, + 452 + ], + "spans": [ + { + "bbox": [ + 106, + 440, + 504, + 452 + ], + "score": 1.0, + "content": "analyses. To the best of our knowledge, it is the first attempt to discover the semantic latent space in", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 450, + 506, + 463 + ], + "spans": [ + { + "bbox": [ + 105, + 450, + 506, + 463 + ], + "score": 1.0, + "content": "the frozen pretrained diffusion models. Spoiler alert: our semantic latent space is different from the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 460, + 505, + 475 + ], + "spans": [ + { + "bbox": [ + 105, + 460, + 505, + 475 + ], + "score": 1.0, + "content": "intermediate latent variables in the diffusion process. Moreover, we introduce a principled design", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 471, + 505, + 485 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 505, + 485 + ], + "score": 1.0, + "content": "of the generative process for versatile editing and quality boosting by quantifiable measures: edit-", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 483, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 506, + 496 + ], + "score": 1.0, + "content": "ing strength of an interval and quality deficiency at a timestep. Extensive experiments demonstrate", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 493, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 493, + 360, + 507 + ], + "score": 1.0, + "content": "that our method is generally applicable to various architectures", + "type": "text" + }, + { + "bbox": [ + 360, + 495, + 402, + 505 + ], + "score": 0.62, + "content": "\\mathrm { ( D D P M + + }", + "type": "inline_equation" + }, + { + "bbox": [ + 403, + 493, + 506, + 507 + ], + "score": 1.0, + "content": ", iDDPM, and ADM) and", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 505, + 452, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 452, + 518 + ], + "score": 1.0, + "content": "datasets (CelebA-HQ, AFHQ-dog, LSUN-church, LSUN-bedroom, and METFACES).", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 20.5, + "bbox_fs": [ + 105, + 363, + 506, + 518 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 532, + 200, + 545 + ], + "lines": [ + { + "bbox": [ + 104, + 530, + 201, + 548 + ], + "spans": [ + { + "bbox": [ + 104, + 530, + 201, + 548 + ], + "score": 1.0, + "content": "2 BACKGROUND", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 106, + 556, + 500, + 568 + ], + "lines": [ + { + "bbox": [ + 106, + 555, + 501, + 569 + ], + "spans": [ + { + "bbox": [ + 106, + 555, + 501, + 569 + ], + "score": 1.0, + "content": "We briefly describe essential backgrounds. The rest of the related work is deferred to Appendix A.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29, + "bbox_fs": [ + 106, + 555, + 501, + 569 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 580, + 368, + 592 + ], + "lines": [ + { + "bbox": [ + 105, + 579, + 369, + 595 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 369, + 595 + ], + "score": 1.0, + "content": "2.1 DENOISING DIFFUSION PROBABILITY MODEL (DDPM)", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 106, + 601, + 506, + 634 + ], + "lines": [ + { + "bbox": [ + 105, + 600, + 505, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 600, + 505, + 613 + ], + "score": 1.0, + "content": "DDPM is a latent variable model that learns a data distribution by denoising noisy images (Ho et al.,", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 611, + 506, + 626 + ], + "spans": [ + { + "bbox": [ + 105, + 611, + 506, + 626 + ], + "score": 1.0, + "content": "2020). The forward process diffuses the data samples through Gaussian transitions parameterized", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 622, + 203, + 636 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 203, + 636 + ], + "score": 1.0, + "content": "with a Markov process:", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32, + "bbox_fs": [ + 105, + 600, + 506, + 636 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 133, + 633, + 477, + 661 + ], + "lines": [ + { + "bbox": [ + 133, + 633, + 477, + 661 + ], + "spans": [ + { + "bbox": [ + 133, + 633, + 477, + 661 + ], + "score": 0.92, + "content": "q \\left( \\pmb { x } _ { t } \\mid \\pmb { \\dot { x } } _ { t - 1 } \\right) = \\mathcal { N } \\left( \\pmb { x } _ { t } ; \\sqrt { 1 - \\beta _ { t } } \\pmb { x } _ { t - 1 } , \\beta _ { t } \\mathbf { I } \\right) = \\mathcal { N } \\left( \\sqrt { \\frac { \\alpha _ { t } } { \\alpha _ { t - 1 } } } \\pmb { x } _ { t - 1 } , \\left( 1 - \\frac { \\alpha _ { t } } { \\alpha _ { t - 1 } } \\right) \\pmb { I } \\right) ,", + "type": "interline_equation", + "image_path": "035cdb05d9c2f6d50e510faab182ddb5c1138b8f9e837a09f7b8bf8817ac6c24.jpg" + } + ] + } + ], + "index": 35, + "virtual_lines": [ + { + "bbox": [ + 133, + 633, + 477, + 642.3333333333334 + ], + "spans": [], + "index": 34 + }, + { + "bbox": [ + 133, + 642.3333333333334, + 477, + 651.6666666666667 + ], + "spans": [], + "index": 35 + }, + { + "bbox": [ + 133, + 651.6666666666667, + 477, + 661.0000000000001 + ], + "spans": [], + "index": 36 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 663, + 508, + 690 + ], + "lines": [ + { + "bbox": [ + 105, + 662, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 662, + 133, + 678 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 133, + 663, + 167, + 677 + ], + "score": 0.93, + "content": "\\{ \\beta _ { t } \\} _ { t = 1 } ^ { T }", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 662, + 281, + 678 + ], + "score": 1.0, + "content": "is the variance schedule and", + "type": "text" + }, + { + "bbox": [ + 282, + 663, + 364, + 677 + ], + "score": 0.93, + "content": "\\begin{array} { r } { \\alpha _ { t } = \\prod _ { s = 1 } ^ { t } ( 1 - \\beta _ { s } ) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 364, + 662, + 506, + 678 + ], + "score": 1.0, + "content": ". Then the reverse process becomes", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 675, + 488, + 693 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 273, + 690 + ], + "score": 0.9, + "content": "\\begin{array} { r } { p _ { \\theta } \\left( \\pmb { x } _ { 0 : T } \\right) : = p \\left( \\pmb { x } _ { T } \\right) \\prod _ { t = 1 } ^ { T } p _ { \\theta } \\left( \\pmb { x } _ { t - 1 } \\mid \\pmb { x } _ { t } \\right) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 273, + 675, + 331, + 693 + ], + "score": 1.0, + "content": ", starting from", + "type": "text" + }, + { + "bbox": [ + 331, + 678, + 389, + 690 + ], + "score": 0.92, + "content": "\\mathbf { \\boldsymbol { x } } _ { T } \\sim \\mathcal { N } ( 0 , \\mathbf { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 389, + 675, + 473, + 693 + ], + "score": 1.0, + "content": "with noise predictor", + "type": "text" + }, + { + "bbox": [ + 473, + 677, + 483, + 690 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 483, + 675, + 488, + 693 + ], + "score": 1.0, + "content": ":", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 37.5, + "bbox_fs": [ + 105, + 662, + 506, + 693 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 199, + 691, + 411, + 719 + ], + "lines": [ + { + "bbox": [ + 199, + 691, + 411, + 719 + ], + "spans": [ + { + "bbox": [ + 199, + 691, + 411, + 719 + ], + "score": 0.93, + "content": "{ \\pmb x } _ { t - 1 } = \\frac { 1 } { \\sqrt { 1 - \\beta _ { t } } } \\left( { \\pmb x } _ { t } - \\frac { \\beta _ { t } } { \\sqrt { 1 - \\alpha _ { t } } } { \\pmb \\epsilon } _ { t } ^ { \\theta } \\left( { \\pmb x } _ { t } \\right) \\right) + \\sigma _ { t } { \\pmb z } _ { t } ,", + "type": "interline_equation", + "image_path": "d7055f0c4357778eb781035893f1f1562f0810b6f8ac46e9dc8ddb106fdd7fcd.jpg" + } + ] + } + ], + "index": 39.5, + "virtual_lines": [ + { + "bbox": [ + 199, + 691, + 411, + 705.0 + ], + "spans": [], + "index": 39 + }, + { + "bbox": [ + 199, + 705.0, + 411, + 719.0 + ], + "spans": [], + "index": 40 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 720, + 498, + 733 + ], + "lines": [ + { + "bbox": [ + 106, + 718, + 500, + 734 + ], + "spans": [ + { + "bbox": [ + 106, + 718, + 133, + 734 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 133, + 720, + 188, + 732 + ], + "score": 0.93, + "content": "\\boldsymbol { z } _ { t } \\sim \\mathcal { N } ( 0 , \\mathbf { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 188, + 718, + 206, + 734 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 206, + 720, + 218, + 732 + ], + "score": 0.9, + "content": "\\sigma _ { t } ^ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 218, + 718, + 419, + 734 + ], + "score": 1.0, + "content": "is a variance of the reverse process which is set to", + "type": "text" + }, + { + "bbox": [ + 420, + 720, + 453, + 732 + ], + "score": 0.93, + "content": "\\sigma _ { t } ^ { 2 } = \\beta _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 454, + 718, + 500, + 734 + ], + "score": 1.0, + "content": "by DDPM.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 41, + "bbox_fs": [ + 106, + 718, + 500, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 106, + 81, + 347, + 94 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 347, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 347, + 96 + ], + "score": 1.0, + "content": "2.2 DENOISING DIFFUSION IMPLICIT MODEL (DDIM)", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 105, + 104, + 505, + 130 + ], + "lines": [ + { + "bbox": [ + 104, + 100, + 505, + 122 + ], + "spans": [ + { + "bbox": [ + 104, + 100, + 214, + 122 + ], + "score": 1.0, + "content": "DDIM redefines Eq. (1) as", + "type": "text" + }, + { + "bbox": [ + 214, + 103, + 505, + 119 + ], + "score": 0.87, + "content": "\\begin{array} { r } { q _ { \\sigma } ( \\pmb { x } _ { t - 1 } | \\pmb { x } _ { t } , \\pmb { x } _ { 0 } ) = \\mathcal { N } ( \\sqrt { \\alpha _ { t - 1 } } \\pmb { x } _ { 0 } + \\sqrt { 1 - \\alpha _ { t - 1 } - \\sigma _ { t } ^ { 2 } } \\cdot \\frac { \\pmb { x } _ { t } - \\sqrt { \\alpha _ { t } } \\pmb { x } _ { 0 } } { \\sqrt { 1 - \\alpha _ { t } } } , \\sigma _ { t } ^ { 2 } I ) } \\end{array}", + "type": "inline_equation" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 117, + 500, + 131 + ], + "spans": [ + { + "bbox": [ + 106, + 117, + 500, + 131 + ], + "score": 1.0, + "content": "which is a non-Markovian process (Song et al., 2020a). Accordingly, the reverse process becomes", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1.5 + }, + { + "type": "interline_equation", + "bbox": [ + 135, + 133, + 475, + 175 + ], + "lines": [ + { + "bbox": [ + 135, + 133, + 475, + 175 + ], + "spans": [ + { + "bbox": [ + 135, + 133, + 475, + 175 + ], + "score": 0.82, + "content": "\\begin{array} { r } { x _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\underbrace { \\left( \\frac { x _ { t } - \\sqrt { 1 - \\alpha _ { t } } \\epsilon _ { t } ^ { \\theta } \\left( x _ { t } \\right) } { \\sqrt { \\alpha _ { t } } } \\right) } _ { \\mathrm { \\normalfont ~ \\ \" p r e d i c t e d ~ \\ b { x } _ 0 ~ \\ \" ~ } } + \\underbrace { \\sqrt { 1 - \\alpha _ { t - 1 } - \\sigma _ { t } ^ { 2 } } \\cdot \\epsilon _ { t } ^ { \\theta } \\left( x _ { t } \\right) } _ { \\mathrm { \\normalfont ~ \\ \" d i r e c t i o n ~ p o i n t i n g ~ t o ~ \\ b { x } _ t \\cdot \\ } } + \\underbrace { \\sigma _ { t } z _ { t } } _ { \\mathrm { \\normalfont ~ \\mathrm { r a n d o m ~ n o i s e } } } \\ , } \\end{array}", + "type": "interline_equation", + "image_path": "ea9fa73c46f30bb8a42ee1c4981dde761fc0b59874a789de6e6aab9c0b57a615.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 135, + 133, + 475, + 147.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 135, + 147.0, + 475, + 161.0 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 135, + 161.0, + 475, + 175.0 + ], + "spans": [], + "index": 5 + } + ] + }, + { + "type": "text", + "bbox": [ + 104, + 182, + 505, + 206 + ], + "lines": [ + { + "bbox": [ + 106, + 182, + 505, + 196 + ], + "spans": [ + { + "bbox": [ + 106, + 182, + 132, + 195 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 133, + 182, + 323, + 196 + ], + "score": 0.91, + "content": "\\sigma _ { t } = \\eta \\sqrt { \\left( 1 - \\alpha _ { t - 1 } \\right) / \\left( 1 - \\alpha _ { t } \\right) } \\sqrt { 1 - \\alpha _ { t } / \\alpha _ { t - 1 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 323, + 182, + 352, + 195 + ], + "score": 1.0, + "content": ". When", + "type": "text" + }, + { + "bbox": [ + 352, + 184, + 377, + 195 + ], + "score": 0.91, + "content": "\\eta = 1", + "type": "inline_equation" + }, + { + "bbox": [ + 377, + 182, + 404, + 195 + ], + "score": 1.0, + "content": "for all", + "type": "text" + }, + { + "bbox": [ + 404, + 185, + 409, + 194 + ], + "score": 0.72, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 410, + 182, + 505, + 195 + ], + "score": 1.0, + "content": ", it becomes DDPM. As", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 107, + 194, + 435, + 206 + ], + "spans": [ + { + "bbox": [ + 107, + 195, + 131, + 206 + ], + "score": 0.89, + "content": "\\eta = 0", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 194, + 435, + 206 + ], + "score": 1.0, + "content": ", the process becomes deterministic and guarantees nearly perfect inversion.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5 + }, + { + "type": "title", + "bbox": [ + 108, + 218, + 281, + 230 + ], + "lines": [ + { + "bbox": [ + 106, + 218, + 282, + 231 + ], + "spans": [ + { + "bbox": [ + 106, + 218, + 282, + 231 + ], + "score": 1.0, + "content": "2.3 IMAGE MANIPULATION WITH CLIP", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 106, + 239, + 505, + 295 + ], + "lines": [ + { + "bbox": [ + 106, + 239, + 504, + 251 + ], + "spans": [ + { + "bbox": [ + 106, + 239, + 350, + 251 + ], + "score": 1.0, + "content": "CLIP learns multimodal embeddings with an image encoder", + "type": "text" + }, + { + "bbox": [ + 350, + 240, + 363, + 250 + ], + "score": 0.88, + "content": "E _ { I }", + "type": "inline_equation" + }, + { + "bbox": [ + 363, + 239, + 440, + 251 + ], + "score": 1.0, + "content": "and a text encoder", + "type": "text" + }, + { + "bbox": [ + 441, + 240, + 455, + 250 + ], + "score": 0.89, + "content": "E _ { T }", + "type": "inline_equation" + }, + { + "bbox": [ + 455, + 239, + 504, + 251 + ], + "score": 1.0, + "content": "whose sim-", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 250, + 505, + 262 + ], + "spans": [ + { + "bbox": [ + 106, + 250, + 505, + 262 + ], + "score": 1.0, + "content": "ilarity indicates semantic similarity between images and texts (Radford et al., 2021). Compared to", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 261, + 504, + 273 + ], + "spans": [ + { + "bbox": [ + 106, + 261, + 504, + 273 + ], + "score": 1.0, + "content": "directly minimizing the cosine distance between the edited image and the target description (Patash-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 271, + 505, + 285 + ], + "spans": [ + { + "bbox": [ + 105, + 271, + 505, + 285 + ], + "score": 1.0, + "content": "nik et al., 2021), directional loss with cosine distance achieves homogeneous editing without mode", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 283, + 215, + 297 + ], + "spans": [ + { + "bbox": [ + 106, + 283, + 215, + 297 + ], + "score": 1.0, + "content": "collapse (Gal et al., 2021):", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 11 + }, + { + "type": "interline_equation", + "bbox": [ + 179, + 297, + 430, + 324 + ], + "lines": [ + { + "bbox": [ + 179, + 297, + 430, + 324 + ], + "spans": [ + { + "bbox": [ + 179, + 297, + 430, + 324 + ], + "score": 0.91, + "content": "\\mathcal { L } _ { \\mathrm { d i r e c t i o n } } \\left( { \\boldsymbol { x } } ^ { \\mathrm { e d i t } } , { \\boldsymbol { y } } ^ { \\mathrm { t a r g e t } } ; { \\boldsymbol { x } } ^ { \\mathrm { s o u r c e } } , { \\boldsymbol { y } } ^ { \\mathrm { s o u r c e } } \\right) : = 1 - \\frac { \\Delta I \\cdot \\Delta T } { \\| \\Delta I \\| \\| \\Delta T \\| } ,", + "type": "interline_equation", + "image_path": "96a8ada4679fa2db751231f877d8fd3f9c808e36bcb0b4a1c2debc540973182f.jpg" + } + ] + } + ], + "index": 14, + "virtual_lines": [ + { + "bbox": [ + 179, + 297, + 430, + 324 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 327, + 505, + 365 + ], + "lines": [ + { + "bbox": [ + 104, + 324, + 507, + 344 + ], + "spans": [ + { + "bbox": [ + 104, + 324, + 133, + 344 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 134, + 328, + 279, + 341 + ], + "score": 0.9, + "content": "\\Delta T = E _ { T } \\left( y ^ { \\mathrm { t a r g e t } } \\right) - E _ { T } \\left( y ^ { \\mathrm { s o u r c e } } \\right)", + "type": "inline_equation" + }, + { + "bbox": [ + 280, + 324, + 298, + 344 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 299, + 327, + 434, + 342 + ], + "score": 0.91, + "content": "\\Delta I = E _ { I } \\left( { \\pmb x } ^ { \\mathrm { e d i t } } \\right) - E _ { I } \\left( { \\pmb x } ^ { \\mathrm { s o u r c e } } \\right)", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 324, + 507, + 344 + ], + "score": 1.0, + "content": "for edited image", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 339, + 506, + 355 + ], + "spans": [ + { + "bbox": [ + 106, + 341, + 128, + 352 + ], + "score": 0.86, + "content": "\\pmb { x } ^ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 128, + 339, + 205, + 355 + ], + "score": 1.0, + "content": ", target description", + "type": "text" + }, + { + "bbox": [ + 205, + 342, + 232, + 353 + ], + "score": 0.89, + "content": "y ^ { \\mathrm { t a r g e t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 232, + 339, + 298, + 355 + ], + "score": 1.0, + "content": ", original image", + "type": "text" + }, + { + "bbox": [ + 298, + 342, + 327, + 352 + ], + "score": 0.84, + "content": "\\pmb { x } ^ { \\mathrm { s o u r c e } }", + "type": "inline_equation" + }, + { + "bbox": [ + 328, + 339, + 425, + 355 + ], + "score": 1.0, + "content": ", and source description", + "type": "text" + }, + { + "bbox": [ + 426, + 342, + 454, + 353 + ], + "score": 0.81, + "content": "y ^ { \\mathrm { s o u r c e } }", + "type": "inline_equation" + }, + { + "bbox": [ + 454, + 339, + 506, + 355 + ], + "score": 1.0, + "content": ". We use the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 352, + 505, + 366 + ], + "spans": [ + { + "bbox": [ + 105, + 352, + 505, + 366 + ], + "score": 1.0, + "content": "prompts ‘smiling face’ and ‘face’ as the target and source descriptions for facial attribute smiling.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16 + }, + { + "type": "title", + "bbox": [ + 108, + 380, + 451, + 393 + ], + "lines": [ + { + "bbox": [ + 104, + 379, + 452, + 395 + ], + "spans": [ + { + "bbox": [ + 104, + 379, + 452, + 395 + ], + "score": 1.0, + "content": "3 DISCOVERING SEMANTIC LATENT SPACE IN DIFFUSION MODELS", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 106, + 404, + 505, + 438 + ], + "lines": [ + { + "bbox": [ + 106, + 405, + 505, + 417 + ], + "spans": [ + { + "bbox": [ + 106, + 405, + 505, + 417 + ], + "score": 1.0, + "content": "This section explains why naive approaches do not work and proposes a new controllable reverse", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "score": 1.0, + "content": "process. Then we describe the techniques for controlling the generative process. Throughout this", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 426, + 300, + 439 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 300, + 439 + ], + "score": 1.0, + "content": "paper, we use an abbreviated version of Eq. (3):", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20 + }, + { + "type": "interline_equation", + "bbox": [ + 202, + 441, + 408, + 457 + ], + "lines": [ + { + "bbox": [ + 202, + 441, + 408, + 457 + ], + "spans": [ + { + "bbox": [ + 202, + 441, + 408, + 457 + ], + "score": 0.89, + "content": "\\pmb { x } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) ) + \\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) ) + \\sigma _ { t } \\boldsymbol { z } _ { t } ,", + "type": "interline_equation", + "image_path": "9a23d42adc0c7e01a7729c3269d06d5bf2a09e85d5cf922a14707fa83d940c35.jpg" + } + ] + } + ], + "index": 22, + "virtual_lines": [ + { + "bbox": [ + 202, + 441, + 408, + 457 + ], + "spans": [], + "index": 22 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 460, + 505, + 495 + ], + "lines": [ + { + "bbox": [ + 105, + 460, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 460, + 133, + 474 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 133, + 460, + 180, + 473 + ], + "score": 0.92, + "content": "\\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 180, + 460, + 267, + 474 + ], + "score": 1.0, + "content": "denotes the predicted", + "type": "text" + }, + { + "bbox": [ + 267, + 462, + 279, + 472 + ], + "score": 0.86, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 280, + 460, + 296, + 474 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 297, + 460, + 344, + 473 + ], + "score": 0.92, + "content": "\\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 344, + 460, + 475, + 474 + ], + "score": 1.0, + "content": "denotes the direction pointing to", + "type": "text" + }, + { + "bbox": [ + 475, + 463, + 486, + 472 + ], + "score": 0.82, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 486, + 460, + 505, + 474 + ], + "score": 1.0, + "content": ". We", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 470, + 503, + 486 + ], + "spans": [ + { + "bbox": [ + 105, + 470, + 127, + 486 + ], + "score": 1.0, + "content": "omit", + "type": "text" + }, + { + "bbox": [ + 128, + 474, + 147, + 483 + ], + "score": 0.88, + "content": "\\sigma _ { t } z _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 147, + 470, + 247, + 486 + ], + "score": 1.0, + "content": "for brevity, except when", + "type": "text" + }, + { + "bbox": [ + 247, + 473, + 272, + 484 + ], + "score": 0.91, + "content": "\\eta \\neq 0", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 470, + 365, + 486 + ], + "score": 1.0, + "content": ". We further abbreviate", + "type": "text" + }, + { + "bbox": [ + 365, + 471, + 412, + 484 + ], + "score": 0.94, + "content": "\\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 412, + 470, + 424, + 486 + ], + "score": 1.0, + "content": "as", + "type": "text" + }, + { + "bbox": [ + 424, + 472, + 437, + 483 + ], + "score": 0.89, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 437, + 470, + 455, + 486 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 456, + 472, + 503, + 484 + ], + "score": 0.91, + "content": "\\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 483, + 330, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 117, + 496 + ], + "score": 1.0, + "content": "as", + "type": "text" + }, + { + "bbox": [ + 117, + 484, + 131, + 494 + ], + "score": 0.88, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 483, + 330, + 496 + ], + "score": 1.0, + "content": "when the context clearly specifies the arguments.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24 + }, + { + "type": "title", + "bbox": [ + 107, + 507, + 174, + 519 + ], + "lines": [ + { + "bbox": [ + 105, + 506, + 175, + 520 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 175, + 520 + ], + "score": 1.0, + "content": "3.1 PROBLEM", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 26 + }, + { + "type": "text", + "bbox": [ + 106, + 528, + 505, + 573 + ], + "lines": [ + { + "bbox": [ + 106, + 528, + 506, + 541 + ], + "spans": [ + { + "bbox": [ + 106, + 528, + 336, + 541 + ], + "score": 1.0, + "content": "We aim to allow semantic latent manipulation of images", + "type": "text" + }, + { + "bbox": [ + 336, + 531, + 348, + 540 + ], + "score": 0.85, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 528, + 413, + 541 + ], + "score": 1.0, + "content": "generated from", + "type": "text" + }, + { + "bbox": [ + 414, + 530, + 428, + 540 + ], + "score": 0.85, + "content": "\\mathbf { \\nabla } _ { \\mathbf { x } _ { T } }", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 528, + 506, + 541 + ], + "score": 1.0, + "content": "given a pretrained", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 539, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 539, + 348, + 552 + ], + "score": 1.0, + "content": "and frozen diffusion model. The easiest idea to manipulate", + "type": "text" + }, + { + "bbox": [ + 348, + 541, + 361, + 550 + ], + "score": 0.87, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 361, + 539, + 440, + 552 + ], + "score": 1.0, + "content": "is simply updating", + "type": "text" + }, + { + "bbox": [ + 441, + 541, + 455, + 550 + ], + "score": 0.87, + "content": "\\mathbf { \\nabla } _ { \\mathbf { \\mathcal { X } } \\mathcal { T } }", + "type": "inline_equation" + }, + { + "bbox": [ + 455, + 539, + 505, + 552 + ], + "score": 1.0, + "content": "to optimize", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 550, + 505, + 562 + ], + "spans": [ + { + "bbox": [ + 105, + 550, + 505, + 562 + ], + "score": 1.0, + "content": "the directional CLIP loss given text prompts with Eq. (4). However, it leads to distorted images or", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 561, + 277, + 573 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 277, + 573 + ], + "score": 1.0, + "content": "incorrect manipulation (Kim & Ye, 2021).", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 28.5 + }, + { + "type": "text", + "bbox": [ + 106, + 577, + 504, + 612 + ], + "lines": [ + { + "bbox": [ + 104, + 576, + 506, + 592 + ], + "spans": [ + { + "bbox": [ + 104, + 576, + 291, + 592 + ], + "score": 1.0, + "content": "An alternative approach is to shift the noise", + "type": "text" + }, + { + "bbox": [ + 291, + 577, + 302, + 590 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 303, + 576, + 506, + 592 + ], + "score": 1.0, + "content": "predicted by the network at each sampling step.", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 589, + 505, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 589, + 276, + 601 + ], + "score": 1.0, + "content": "However, it does not achieve manipulating", + "type": "text" + }, + { + "bbox": [ + 276, + 591, + 288, + 600 + ], + "score": 0.86, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 289, + 589, + 433, + 601 + ], + "score": 1.0, + "content": "because the intermediate changes in", + "type": "text" + }, + { + "bbox": [ + 433, + 589, + 445, + 600 + ], + "score": 0.89, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 446, + 589, + 462, + 601 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 463, + 590, + 476, + 600 + ], + "score": 0.89, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 477, + 589, + 505, + 601 + ], + "score": 1.0, + "content": "cancel", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 600, + 440, + 612 + ], + "spans": [ + { + "bbox": [ + 106, + 600, + 250, + 612 + ], + "score": 1.0, + "content": "out each other resulting in the same", + "type": "text" + }, + { + "bbox": [ + 251, + 600, + 288, + 612 + ], + "score": 0.93, + "content": "p _ { \\theta } ( { \\pmb x } _ { 0 : T } )", + "type": "inline_equation" + }, + { + "bbox": [ + 288, + 600, + 440, + 612 + ], + "score": 1.0, + "content": ", similarly to destructive interference.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32 + }, + { + "type": "text", + "bbox": [ + 106, + 614, + 505, + 650 + ], + "lines": [ + { + "bbox": [ + 105, + 612, + 507, + 628 + ], + "spans": [ + { + "bbox": [ + 105, + 612, + 178, + 628 + ], + "score": 1.0, + "content": "Theorem 1. Let", + "type": "text" + }, + { + "bbox": [ + 179, + 613, + 189, + 626 + ], + "score": 0.88, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 612, + 440, + 628 + ], + "score": 1.0, + "content": "be a predicted noise during the original reverse process at", + "type": "text" + }, + { + "bbox": [ + 440, + 615, + 446, + 624 + ], + "score": 0.59, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 446, + 612, + 466, + 628 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 467, + 613, + 477, + 625 + ], + "score": 0.84, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 612, + 507, + 628 + ], + "score": 1.0, + "content": "be its", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 624, + 505, + 638 + ], + "spans": [ + { + "bbox": [ + 106, + 624, + 213, + 638 + ], + "score": 1.0, + "content": "shifted counterpart. Then,", + "type": "text" + }, + { + "bbox": [ + 214, + 626, + 300, + 637 + ], + "score": 0.92, + "content": "\\Delta \\pmb { x } _ { t } = \\tilde { \\pmb { x } } _ { t - 1 } - \\pmb { x } _ { t - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 301, + 624, + 380, + 638 + ], + "score": 1.0, + "content": "is negligible where", + "type": "text" + }, + { + "bbox": [ + 380, + 625, + 505, + 638 + ], + "score": 0.88, + "content": "\\begin{array} { r } { \\tilde { { \\pmb { x } } } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } { \\bf P } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\pmb { x } } _ { t } ) ) + } \\end{array}", + "type": "inline_equation" + } + ], + "index": 35 + }, + { + "bbox": [ + 107, + 636, + 493, + 651 + ], + "spans": [ + { + "bbox": [ + 107, + 637, + 154, + 650 + ], + "score": 0.92, + "content": "\\mathbf { D } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 154, + 636, + 255, + 651 + ], + "score": 1.0, + "content": ". I.e., the shifted terms of", + "type": "text" + }, + { + "bbox": [ + 255, + 637, + 265, + 650 + ], + "score": 0.87, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 266, + 636, + 276, + 651 + ], + "score": 1.0, + "content": "in", + "type": "text" + }, + { + "bbox": [ + 277, + 638, + 289, + 649 + ], + "score": 0.82, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 636, + 308, + 651 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 308, + 638, + 322, + 649 + ], + "score": 0.88, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 322, + 636, + 493, + 651 + ], + "score": 1.0, + "content": "destruct each other in the reverse process.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35 + }, + { + "type": "text", + "bbox": [ + 104, + 657, + 469, + 670 + ], + "lines": [ + { + "bbox": [ + 105, + 656, + 465, + 673 + ], + "spans": [ + { + "bbox": [ + 105, + 656, + 354, + 673 + ], + "score": 1.0, + "content": "Appendix C proves above theorem. Figure 13(a-b) shows that", + "type": "text" + }, + { + "bbox": [ + 355, + 659, + 367, + 669 + ], + "score": 0.89, + "content": "\\tilde { \\mathbf { x } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 367, + 656, + 453, + 673 + ], + "score": 1.0, + "content": "is almost identical to", + "type": "text" + }, + { + "bbox": [ + 453, + 660, + 465, + 669 + ], + "score": 0.85, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "title", + "bbox": [ + 108, + 683, + 273, + 694 + ], + "lines": [ + { + "bbox": [ + 106, + 683, + 274, + 695 + ], + "spans": [ + { + "bbox": [ + 106, + 683, + 274, + 695 + ], + "score": 1.0, + "content": "3.2 ASYMMETRIC REVERSE PROCESS", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 105, + 703, + 498, + 715 + ], + "lines": [ + { + "bbox": [ + 105, + 701, + 500, + 717 + ], + "spans": [ + { + "bbox": [ + 105, + 701, + 500, + 717 + ], + "score": 1.0, + "content": "In order to break the interference, we propose a new controllable reverse process with asymmetry:", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 39 + }, + { + "type": "interline_equation", + "bbox": [ + 217, + 718, + 393, + 734 + ], + "lines": [ + { + "bbox": [ + 217, + 718, + 393, + 734 + ], + "spans": [ + { + "bbox": [ + 217, + 718, + 393, + 734 + ], + "score": 0.91, + "content": "\\pmb { x } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } \\big ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big ) + \\mathbf { D } _ { t } \\big ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big ) ,", + "type": "interline_equation", + "image_path": "bdbf48fe4af9abaa50663aa224fdaed1e2c1a3d853a0ccc0232744adfa33948f.jpg" + } + ] + } + ], + "index": 40, + "virtual_lines": [ + { + "bbox": [ + 217, + 718, + 393, + 734 + ], + "spans": [], + "index": 40 + } + ] + } + ], + "page_idx": 2, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 26, + 293, + 38 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 294, + 39 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 294, + 39 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "score": 1.0, + "content": "3", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 106, + 81, + 347, + 94 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 347, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 347, + 96 + ], + "score": 1.0, + "content": "2.2 DENOISING DIFFUSION IMPLICIT MODEL (DDIM)", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 105, + 104, + 505, + 130 + ], + "lines": [ + { + "bbox": [ + 104, + 100, + 505, + 122 + ], + "spans": [ + { + "bbox": [ + 104, + 100, + 214, + 122 + ], + "score": 1.0, + "content": "DDIM redefines Eq. (1) as", + "type": "text" + }, + { + "bbox": [ + 214, + 103, + 505, + 119 + ], + "score": 0.87, + "content": "\\begin{array} { r } { q _ { \\sigma } ( \\pmb { x } _ { t - 1 } | \\pmb { x } _ { t } , \\pmb { x } _ { 0 } ) = \\mathcal { N } ( \\sqrt { \\alpha _ { t - 1 } } \\pmb { x } _ { 0 } + \\sqrt { 1 - \\alpha _ { t - 1 } - \\sigma _ { t } ^ { 2 } } \\cdot \\frac { \\pmb { x } _ { t } - \\sqrt { \\alpha _ { t } } \\pmb { x } _ { 0 } } { \\sqrt { 1 - \\alpha _ { t } } } , \\sigma _ { t } ^ { 2 } I ) } \\end{array}", + "type": "inline_equation" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 117, + 500, + 131 + ], + "spans": [ + { + "bbox": [ + 106, + 117, + 500, + 131 + ], + "score": 1.0, + "content": "which is a non-Markovian process (Song et al., 2020a). Accordingly, the reverse process becomes", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1.5, + "bbox_fs": [ + 104, + 100, + 505, + 131 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 135, + 133, + 475, + 175 + ], + "lines": [ + { + "bbox": [ + 135, + 133, + 475, + 175 + ], + "spans": [ + { + "bbox": [ + 135, + 133, + 475, + 175 + ], + "score": 0.82, + "content": "\\begin{array} { r } { x _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\underbrace { \\left( \\frac { x _ { t } - \\sqrt { 1 - \\alpha _ { t } } \\epsilon _ { t } ^ { \\theta } \\left( x _ { t } \\right) } { \\sqrt { \\alpha _ { t } } } \\right) } _ { \\mathrm { \\normalfont ~ \\ \" p r e d i c t e d ~ \\ b { x } _ 0 ~ \\ \" ~ } } + \\underbrace { \\sqrt { 1 - \\alpha _ { t - 1 } - \\sigma _ { t } ^ { 2 } } \\cdot \\epsilon _ { t } ^ { \\theta } \\left( x _ { t } \\right) } _ { \\mathrm { \\normalfont ~ \\ \" d i r e c t i o n ~ p o i n t i n g ~ t o ~ \\ b { x } _ t \\cdot \\ } } + \\underbrace { \\sigma _ { t } z _ { t } } _ { \\mathrm { \\normalfont ~ \\mathrm { r a n d o m ~ n o i s e } } } \\ , } \\end{array}", + "type": "interline_equation", + "image_path": "ea9fa73c46f30bb8a42ee1c4981dde761fc0b59874a789de6e6aab9c0b57a615.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 135, + 133, + 475, + 147.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 135, + 147.0, + 475, + 161.0 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 135, + 161.0, + 475, + 175.0 + ], + "spans": [], + "index": 5 + } + ] + }, + { + "type": "text", + "bbox": [ + 104, + 182, + 505, + 206 + ], + "lines": [ + { + "bbox": [ + 106, + 182, + 505, + 196 + ], + "spans": [ + { + "bbox": [ + 106, + 182, + 132, + 195 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 133, + 182, + 323, + 196 + ], + "score": 0.91, + "content": "\\sigma _ { t } = \\eta \\sqrt { \\left( 1 - \\alpha _ { t - 1 } \\right) / \\left( 1 - \\alpha _ { t } \\right) } \\sqrt { 1 - \\alpha _ { t } / \\alpha _ { t - 1 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 323, + 182, + 352, + 195 + ], + "score": 1.0, + "content": ". When", + "type": "text" + }, + { + "bbox": [ + 352, + 184, + 377, + 195 + ], + "score": 0.91, + "content": "\\eta = 1", + "type": "inline_equation" + }, + { + "bbox": [ + 377, + 182, + 404, + 195 + ], + "score": 1.0, + "content": "for all", + "type": "text" + }, + { + "bbox": [ + 404, + 185, + 409, + 194 + ], + "score": 0.72, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 410, + 182, + 505, + 195 + ], + "score": 1.0, + "content": ", it becomes DDPM. As", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 107, + 194, + 435, + 206 + ], + "spans": [ + { + "bbox": [ + 107, + 195, + 131, + 206 + ], + "score": 0.89, + "content": "\\eta = 0", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 194, + 435, + 206 + ], + "score": 1.0, + "content": ", the process becomes deterministic and guarantees nearly perfect inversion.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5, + "bbox_fs": [ + 106, + 182, + 505, + 206 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 218, + 281, + 230 + ], + "lines": [ + { + "bbox": [ + 106, + 218, + 282, + 231 + ], + "spans": [ + { + "bbox": [ + 106, + 218, + 282, + 231 + ], + "score": 1.0, + "content": "2.3 IMAGE MANIPULATION WITH CLIP", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 106, + 239, + 505, + 295 + ], + "lines": [ + { + "bbox": [ + 106, + 239, + 504, + 251 + ], + "spans": [ + { + "bbox": [ + 106, + 239, + 350, + 251 + ], + "score": 1.0, + "content": "CLIP learns multimodal embeddings with an image encoder", + "type": "text" + }, + { + "bbox": [ + 350, + 240, + 363, + 250 + ], + "score": 0.88, + "content": "E _ { I }", + "type": "inline_equation" + }, + { + "bbox": [ + 363, + 239, + 440, + 251 + ], + "score": 1.0, + "content": "and a text encoder", + "type": "text" + }, + { + "bbox": [ + 441, + 240, + 455, + 250 + ], + "score": 0.89, + "content": "E _ { T }", + "type": "inline_equation" + }, + { + "bbox": [ + 455, + 239, + 504, + 251 + ], + "score": 1.0, + "content": "whose sim-", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 250, + 505, + 262 + ], + "spans": [ + { + "bbox": [ + 106, + 250, + 505, + 262 + ], + "score": 1.0, + "content": "ilarity indicates semantic similarity between images and texts (Radford et al., 2021). Compared to", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 261, + 504, + 273 + ], + "spans": [ + { + "bbox": [ + 106, + 261, + 504, + 273 + ], + "score": 1.0, + "content": "directly minimizing the cosine distance between the edited image and the target description (Patash-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 271, + 505, + 285 + ], + "spans": [ + { + "bbox": [ + 105, + 271, + 505, + 285 + ], + "score": 1.0, + "content": "nik et al., 2021), directional loss with cosine distance achieves homogeneous editing without mode", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 283, + 215, + 297 + ], + "spans": [ + { + "bbox": [ + 106, + 283, + 215, + 297 + ], + "score": 1.0, + "content": "collapse (Gal et al., 2021):", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 239, + 505, + 297 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 179, + 297, + 430, + 324 + ], + "lines": [ + { + "bbox": [ + 179, + 297, + 430, + 324 + ], + "spans": [ + { + "bbox": [ + 179, + 297, + 430, + 324 + ], + "score": 0.91, + "content": "\\mathcal { L } _ { \\mathrm { d i r e c t i o n } } \\left( { \\boldsymbol { x } } ^ { \\mathrm { e d i t } } , { \\boldsymbol { y } } ^ { \\mathrm { t a r g e t } } ; { \\boldsymbol { x } } ^ { \\mathrm { s o u r c e } } , { \\boldsymbol { y } } ^ { \\mathrm { s o u r c e } } \\right) : = 1 - \\frac { \\Delta I \\cdot \\Delta T } { \\| \\Delta I \\| \\| \\Delta T \\| } ,", + "type": "interline_equation", + "image_path": "96a8ada4679fa2db751231f877d8fd3f9c808e36bcb0b4a1c2debc540973182f.jpg" + } + ] + } + ], + "index": 14, + "virtual_lines": [ + { + "bbox": [ + 179, + 297, + 430, + 324 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 327, + 505, + 365 + ], + "lines": [ + { + "bbox": [ + 104, + 324, + 507, + 344 + ], + "spans": [ + { + "bbox": [ + 104, + 324, + 133, + 344 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 134, + 328, + 279, + 341 + ], + "score": 0.9, + "content": "\\Delta T = E _ { T } \\left( y ^ { \\mathrm { t a r g e t } } \\right) - E _ { T } \\left( y ^ { \\mathrm { s o u r c e } } \\right)", + "type": "inline_equation" + }, + { + "bbox": [ + 280, + 324, + 298, + 344 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 299, + 327, + 434, + 342 + ], + "score": 0.91, + "content": "\\Delta I = E _ { I } \\left( { \\pmb x } ^ { \\mathrm { e d i t } } \\right) - E _ { I } \\left( { \\pmb x } ^ { \\mathrm { s o u r c e } } \\right)", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 324, + 507, + 344 + ], + "score": 1.0, + "content": "for edited image", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 339, + 506, + 355 + ], + "spans": [ + { + "bbox": [ + 106, + 341, + 128, + 352 + ], + "score": 0.86, + "content": "\\pmb { x } ^ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 128, + 339, + 205, + 355 + ], + "score": 1.0, + "content": ", target description", + "type": "text" + }, + { + "bbox": [ + 205, + 342, + 232, + 353 + ], + "score": 0.89, + "content": "y ^ { \\mathrm { t a r g e t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 232, + 339, + 298, + 355 + ], + "score": 1.0, + "content": ", original image", + "type": "text" + }, + { + "bbox": [ + 298, + 342, + 327, + 352 + ], + "score": 0.84, + "content": "\\pmb { x } ^ { \\mathrm { s o u r c e } }", + "type": "inline_equation" + }, + { + "bbox": [ + 328, + 339, + 425, + 355 + ], + "score": 1.0, + "content": ", and source description", + "type": "text" + }, + { + "bbox": [ + 426, + 342, + 454, + 353 + ], + "score": 0.81, + "content": "y ^ { \\mathrm { s o u r c e } }", + "type": "inline_equation" + }, + { + "bbox": [ + 454, + 339, + 506, + 355 + ], + "score": 1.0, + "content": ". We use the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 352, + 505, + 366 + ], + "spans": [ + { + "bbox": [ + 105, + 352, + 505, + 366 + ], + "score": 1.0, + "content": "prompts ‘smiling face’ and ‘face’ as the target and source descriptions for facial attribute smiling.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16, + "bbox_fs": [ + 104, + 324, + 507, + 366 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 380, + 451, + 393 + ], + "lines": [ + { + "bbox": [ + 104, + 379, + 452, + 395 + ], + "spans": [ + { + "bbox": [ + 104, + 379, + 452, + 395 + ], + "score": 1.0, + "content": "3 DISCOVERING SEMANTIC LATENT SPACE IN DIFFUSION MODELS", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 106, + 404, + 505, + 438 + ], + "lines": [ + { + "bbox": [ + 106, + 405, + 505, + 417 + ], + "spans": [ + { + "bbox": [ + 106, + 405, + 505, + 417 + ], + "score": 1.0, + "content": "This section explains why naive approaches do not work and proposes a new controllable reverse", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 505, + 428 + ], + "score": 1.0, + "content": "process. Then we describe the techniques for controlling the generative process. Throughout this", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 426, + 300, + 439 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 300, + 439 + ], + "score": 1.0, + "content": "paper, we use an abbreviated version of Eq. (3):", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20, + "bbox_fs": [ + 105, + 405, + 505, + 439 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 202, + 441, + 408, + 457 + ], + "lines": [ + { + "bbox": [ + 202, + 441, + 408, + 457 + ], + "spans": [ + { + "bbox": [ + 202, + 441, + 408, + 457 + ], + "score": 0.89, + "content": "\\pmb { x } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) ) + \\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) ) + \\sigma _ { t } \\boldsymbol { z } _ { t } ,", + "type": "interline_equation", + "image_path": "9a23d42adc0c7e01a7729c3269d06d5bf2a09e85d5cf922a14707fa83d940c35.jpg" + } + ] + } + ], + "index": 22, + "virtual_lines": [ + { + "bbox": [ + 202, + 441, + 408, + 457 + ], + "spans": [], + "index": 22 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 460, + 505, + 495 + ], + "lines": [ + { + "bbox": [ + 105, + 460, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 460, + 133, + 474 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 133, + 460, + 180, + 473 + ], + "score": 0.92, + "content": "\\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 180, + 460, + 267, + 474 + ], + "score": 1.0, + "content": "denotes the predicted", + "type": "text" + }, + { + "bbox": [ + 267, + 462, + 279, + 472 + ], + "score": 0.86, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 280, + 460, + 296, + 474 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 297, + 460, + 344, + 473 + ], + "score": 0.92, + "content": "\\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 344, + 460, + 475, + 474 + ], + "score": 1.0, + "content": "denotes the direction pointing to", + "type": "text" + }, + { + "bbox": [ + 475, + 463, + 486, + 472 + ], + "score": 0.82, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 486, + 460, + 505, + 474 + ], + "score": 1.0, + "content": ". We", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 470, + 503, + 486 + ], + "spans": [ + { + "bbox": [ + 105, + 470, + 127, + 486 + ], + "score": 1.0, + "content": "omit", + "type": "text" + }, + { + "bbox": [ + 128, + 474, + 147, + 483 + ], + "score": 0.88, + "content": "\\sigma _ { t } z _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 147, + 470, + 247, + 486 + ], + "score": 1.0, + "content": "for brevity, except when", + "type": "text" + }, + { + "bbox": [ + 247, + 473, + 272, + 484 + ], + "score": 0.91, + "content": "\\eta \\neq 0", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 470, + 365, + 486 + ], + "score": 1.0, + "content": ". We further abbreviate", + "type": "text" + }, + { + "bbox": [ + 365, + 471, + 412, + 484 + ], + "score": 0.94, + "content": "\\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 412, + 470, + 424, + 486 + ], + "score": 1.0, + "content": "as", + "type": "text" + }, + { + "bbox": [ + 424, + 472, + 437, + 483 + ], + "score": 0.89, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 437, + 470, + 455, + 486 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 456, + 472, + 503, + 484 + ], + "score": 0.91, + "content": "\\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 483, + 330, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 117, + 496 + ], + "score": 1.0, + "content": "as", + "type": "text" + }, + { + "bbox": [ + 117, + 484, + 131, + 494 + ], + "score": 0.88, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 483, + 330, + 496 + ], + "score": 1.0, + "content": "when the context clearly specifies the arguments.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24, + "bbox_fs": [ + 105, + 460, + 505, + 496 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 507, + 174, + 519 + ], + "lines": [ + { + "bbox": [ + 105, + 506, + 175, + 520 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 175, + 520 + ], + "score": 1.0, + "content": "3.1 PROBLEM", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 26 + }, + { + "type": "text", + "bbox": [ + 106, + 528, + 505, + 573 + ], + "lines": [ + { + "bbox": [ + 106, + 528, + 506, + 541 + ], + "spans": [ + { + "bbox": [ + 106, + 528, + 336, + 541 + ], + "score": 1.0, + "content": "We aim to allow semantic latent manipulation of images", + "type": "text" + }, + { + "bbox": [ + 336, + 531, + 348, + 540 + ], + "score": 0.85, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 528, + 413, + 541 + ], + "score": 1.0, + "content": "generated from", + "type": "text" + }, + { + "bbox": [ + 414, + 530, + 428, + 540 + ], + "score": 0.85, + "content": "\\mathbf { \\nabla } _ { \\mathbf { x } _ { T } }", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 528, + 506, + 541 + ], + "score": 1.0, + "content": "given a pretrained", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 539, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 539, + 348, + 552 + ], + "score": 1.0, + "content": "and frozen diffusion model. The easiest idea to manipulate", + "type": "text" + }, + { + "bbox": [ + 348, + 541, + 361, + 550 + ], + "score": 0.87, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 361, + 539, + 440, + 552 + ], + "score": 1.0, + "content": "is simply updating", + "type": "text" + }, + { + "bbox": [ + 441, + 541, + 455, + 550 + ], + "score": 0.87, + "content": "\\mathbf { \\nabla } _ { \\mathbf { \\mathcal { X } } \\mathcal { T } }", + "type": "inline_equation" + }, + { + "bbox": [ + 455, + 539, + 505, + 552 + ], + "score": 1.0, + "content": "to optimize", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 550, + 505, + 562 + ], + "spans": [ + { + "bbox": [ + 105, + 550, + 505, + 562 + ], + "score": 1.0, + "content": "the directional CLIP loss given text prompts with Eq. (4). However, it leads to distorted images or", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 561, + 277, + 573 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 277, + 573 + ], + "score": 1.0, + "content": "incorrect manipulation (Kim & Ye, 2021).", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 28.5, + "bbox_fs": [ + 105, + 528, + 506, + 573 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 577, + 504, + 612 + ], + "lines": [ + { + "bbox": [ + 104, + 576, + 506, + 592 + ], + "spans": [ + { + "bbox": [ + 104, + 576, + 291, + 592 + ], + "score": 1.0, + "content": "An alternative approach is to shift the noise", + "type": "text" + }, + { + "bbox": [ + 291, + 577, + 302, + 590 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 303, + 576, + 506, + 592 + ], + "score": 1.0, + "content": "predicted by the network at each sampling step.", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 589, + 505, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 589, + 276, + 601 + ], + "score": 1.0, + "content": "However, it does not achieve manipulating", + "type": "text" + }, + { + "bbox": [ + 276, + 591, + 288, + 600 + ], + "score": 0.86, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 289, + 589, + 433, + 601 + ], + "score": 1.0, + "content": "because the intermediate changes in", + "type": "text" + }, + { + "bbox": [ + 433, + 589, + 445, + 600 + ], + "score": 0.89, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 446, + 589, + 462, + 601 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 463, + 590, + 476, + 600 + ], + "score": 0.89, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 477, + 589, + 505, + 601 + ], + "score": 1.0, + "content": "cancel", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 600, + 440, + 612 + ], + "spans": [ + { + "bbox": [ + 106, + 600, + 250, + 612 + ], + "score": 1.0, + "content": "out each other resulting in the same", + "type": "text" + }, + { + "bbox": [ + 251, + 600, + 288, + 612 + ], + "score": 0.93, + "content": "p _ { \\theta } ( { \\pmb x } _ { 0 : T } )", + "type": "inline_equation" + }, + { + "bbox": [ + 288, + 600, + 440, + 612 + ], + "score": 1.0, + "content": ", similarly to destructive interference.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32, + "bbox_fs": [ + 104, + 576, + 506, + 612 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 614, + 505, + 650 + ], + "lines": [ + { + "bbox": [ + 105, + 612, + 507, + 628 + ], + "spans": [ + { + "bbox": [ + 105, + 612, + 178, + 628 + ], + "score": 1.0, + "content": "Theorem 1. Let", + "type": "text" + }, + { + "bbox": [ + 179, + 613, + 189, + 626 + ], + "score": 0.88, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 612, + 440, + 628 + ], + "score": 1.0, + "content": "be a predicted noise during the original reverse process at", + "type": "text" + }, + { + "bbox": [ + 440, + 615, + 446, + 624 + ], + "score": 0.59, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 446, + 612, + 466, + 628 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 467, + 613, + 477, + 625 + ], + "score": 0.84, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 612, + 507, + 628 + ], + "score": 1.0, + "content": "be its", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 624, + 505, + 638 + ], + "spans": [ + { + "bbox": [ + 106, + 624, + 213, + 638 + ], + "score": 1.0, + "content": "shifted counterpart. Then,", + "type": "text" + }, + { + "bbox": [ + 214, + 626, + 300, + 637 + ], + "score": 0.92, + "content": "\\Delta \\pmb { x } _ { t } = \\tilde { \\pmb { x } } _ { t - 1 } - \\pmb { x } _ { t - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 301, + 624, + 380, + 638 + ], + "score": 1.0, + "content": "is negligible where", + "type": "text" + }, + { + "bbox": [ + 380, + 625, + 505, + 638 + ], + "score": 0.88, + "content": "\\begin{array} { r } { \\tilde { { \\pmb { x } } } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } { \\bf P } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\pmb { x } } _ { t } ) ) + } \\end{array}", + "type": "inline_equation" + } + ], + "index": 35 + }, + { + "bbox": [ + 107, + 636, + 493, + 651 + ], + "spans": [ + { + "bbox": [ + 107, + 637, + 154, + 650 + ], + "score": 0.92, + "content": "\\mathbf { D } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 154, + 636, + 255, + 651 + ], + "score": 1.0, + "content": ". I.e., the shifted terms of", + "type": "text" + }, + { + "bbox": [ + 255, + 637, + 265, + 650 + ], + "score": 0.87, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 266, + 636, + 276, + 651 + ], + "score": 1.0, + "content": "in", + "type": "text" + }, + { + "bbox": [ + 277, + 638, + 289, + 649 + ], + "score": 0.82, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 636, + 308, + 651 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 308, + 638, + 322, + 649 + ], + "score": 0.88, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 322, + 636, + 493, + 651 + ], + "score": 1.0, + "content": "destruct each other in the reverse process.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35, + "bbox_fs": [ + 105, + 612, + 507, + 651 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 657, + 469, + 670 + ], + "lines": [ + { + "bbox": [ + 105, + 656, + 465, + 673 + ], + "spans": [ + { + "bbox": [ + 105, + 656, + 354, + 673 + ], + "score": 1.0, + "content": "Appendix C proves above theorem. Figure 13(a-b) shows that", + "type": "text" + }, + { + "bbox": [ + 355, + 659, + 367, + 669 + ], + "score": 0.89, + "content": "\\tilde { \\mathbf { x } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 367, + 656, + 453, + 673 + ], + "score": 1.0, + "content": "is almost identical to", + "type": "text" + }, + { + "bbox": [ + 453, + 660, + 465, + 669 + ], + "score": 0.85, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + } + ], + "index": 37 + } + ], + "index": 37, + "bbox_fs": [ + 105, + 656, + 465, + 673 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 683, + 273, + 694 + ], + "lines": [ + { + "bbox": [ + 106, + 683, + 274, + 695 + ], + "spans": [ + { + "bbox": [ + 106, + 683, + 274, + 695 + ], + "score": 1.0, + "content": "3.2 ASYMMETRIC REVERSE PROCESS", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 105, + 703, + 498, + 715 + ], + "lines": [ + { + "bbox": [ + 105, + 701, + 500, + 717 + ], + "spans": [ + { + "bbox": [ + 105, + 701, + 500, + 717 + ], + "score": 1.0, + "content": "In order to break the interference, we propose a new controllable reverse process with asymmetry:", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 39, + "bbox_fs": [ + 105, + 701, + 500, + 717 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 217, + 718, + 393, + 734 + ], + "lines": [ + { + "bbox": [ + 217, + 718, + 393, + 734 + ], + "spans": [ + { + "bbox": [ + 217, + 718, + 393, + 734 + ], + "score": 0.91, + "content": "\\pmb { x } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } \\big ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big ) + \\mathbf { D } _ { t } \\big ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big ) ,", + "type": "interline_equation", + "image_path": "bdbf48fe4af9abaa50663aa224fdaed1e2c1a3d853a0ccc0232744adfa33948f.jpg" + } + ] + } + ], + "index": 40, + "virtual_lines": [ + { + "bbox": [ + 217, + 718, + 393, + 734 + ], + "spans": [], + "index": 40 + } + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 116, + 80, + 496, + 173 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 116, + 80, + 496, + 173 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 116, + 80, + 496, + 173 + ], + "spans": [ + { + "bbox": [ + 116, + 80, + 496, + 173 + ], + "score": 0.962, + "type": "image", + "image_path": "06ae64f291010710a1b60bf1d063bfc7c5ce461052c3aeb0dadeab995109361f.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 116, + 80, + 496, + 111.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 116, + 111.0, + 496, + 142.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 116, + 142.0, + 496, + 173.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 183, + 505, + 217 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 183, + 505, + 196 + ], + "spans": [ + { + "bbox": [ + 106, + 183, + 505, + 196 + ], + "score": 1.0, + "content": "Figure 2: Generative process of Asyrp. The green box on the left illustrates Asyrp which only", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 193, + 506, + 207 + ], + "spans": [ + { + "bbox": [ + 105, + 193, + 130, + 207 + ], + "score": 1.0, + "content": "alters", + "type": "text" + }, + { + "bbox": [ + 130, + 194, + 143, + 205 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 143, + 193, + 212, + 207 + ], + "score": 1.0, + "content": "while preserving", + "type": "text" + }, + { + "bbox": [ + 213, + 194, + 226, + 205 + ], + "score": 0.89, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 226, + 193, + 506, + 207 + ], + "score": 1.0, + "content": "shared by DDIM. The right describes that Asyrp modifies the original", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 205, + 407, + 219 + ], + "spans": [ + { + "bbox": [ + 105, + 205, + 370, + 219 + ], + "score": 1.0, + "content": "reverse process toward the target attribute reflecting the change in", + "type": "text" + }, + { + "bbox": [ + 370, + 206, + 377, + 215 + ], + "score": 0.53, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 377, + 205, + 407, + 219 + ], + "score": 1.0, + "content": "-space.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 106, + 239, + 506, + 275 + ], + "lines": [ + { + "bbox": [ + 105, + 239, + 506, + 253 + ], + "spans": [ + { + "bbox": [ + 105, + 239, + 188, + 253 + ], + "score": 1.0, + "content": "i.e., we modify only", + "type": "text" + }, + { + "bbox": [ + 189, + 241, + 201, + 252 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 201, + 239, + 247, + 253 + ], + "score": 1.0, + "content": "by shifting", + "type": "text" + }, + { + "bbox": [ + 248, + 240, + 258, + 252 + ], + "score": 0.88, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 258, + 239, + 269, + 253 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 269, + 240, + 280, + 252 + ], + "score": 0.86, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 280, + 239, + 349, + 253 + ], + "score": 1.0, + "content": "while preserving", + "type": "text" + }, + { + "bbox": [ + 350, + 241, + 363, + 252 + ], + "score": 0.87, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 363, + 239, + 506, + 253 + ], + "score": 1.0, + "content": ". Intuitively, it modifies the original", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 250, + 506, + 266 + ], + "spans": [ + { + "bbox": [ + 105, + 250, + 223, + 266 + ], + "score": 1.0, + "content": "reverse process according to", + "type": "text" + }, + { + "bbox": [ + 224, + 252, + 287, + 264 + ], + "score": 0.92, + "content": "\\Delta \\epsilon _ { t } = \\widetilde { \\epsilon } _ { t } ^ { \\theta } - \\widetilde { \\epsilon } _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 287, + 250, + 463, + 266 + ], + "score": 1.0, + "content": "while it does not alter the direction toward", + "type": "text" + }, + { + "bbox": [ + 463, + 253, + 474, + 263 + ], + "score": 0.86, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 475, + 250, + 506, + 266 + ], + "score": 1.0, + "content": "so that", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 262, + 491, + 275 + ], + "spans": [ + { + "bbox": [ + 106, + 264, + 128, + 274 + ], + "score": 0.86, + "content": "{ \\mathbf { \\mathcal { x } } } _ { t - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 128, + 262, + 229, + 275 + ], + "score": 1.0, + "content": "follows the original flow", + "type": "text" + }, + { + "bbox": [ + 230, + 263, + 243, + 273 + ], + "score": 0.87, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 243, + 262, + 491, + 275 + ], + "score": 1.0, + "content": "at each sampling step. Figure 2 illustrates the above intuition.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 106, + 278, + 505, + 315 + ], + "lines": [ + { + "bbox": [ + 105, + 277, + 507, + 292 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 306, + 292 + ], + "score": 1.0, + "content": "As in Avrahami et al. (2022), we use the modified", + "type": "text" + }, + { + "bbox": [ + 307, + 279, + 329, + 291 + ], + "score": 0.89, + "content": "\\mathbf { P } _ { t } ^ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 329, + 277, + 395, + 292 + ], + "score": 1.0, + "content": "and the original", + "type": "text" + }, + { + "bbox": [ + 395, + 280, + 426, + 291 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t } ^ { \\mathrm { s o u r c e } }", + "type": "inline_equation" + }, + { + "bbox": [ + 426, + 277, + 507, + 292 + ], + "score": 1.0, + "content": "as visual inputs for", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 104, + 289, + 506, + 303 + ], + "spans": [ + { + "bbox": [ + 104, + 289, + 463, + 303 + ], + "score": 1.0, + "content": "the directional CLIP loss in Eq. (4), and regularize the difference between the modified", + "type": "text" + }, + { + "bbox": [ + 464, + 290, + 486, + 303 + ], + "score": 0.91, + "content": "\\mathbf { P } _ { t } ^ { \\mathrm { { e d i t } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 486, + 289, + 506, + 303 + ], + "score": 1.0, + "content": "and", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 104, + 300, + 354, + 316 + ], + "spans": [ + { + "bbox": [ + 104, + 300, + 154, + 316 + ], + "score": 1.0, + "content": "the original", + "type": "text" + }, + { + "bbox": [ + 155, + 303, + 185, + 315 + ], + "score": 0.87, + "content": "\\mathbf { P } _ { t } ^ { \\mathrm { s o u r c e } }", + "type": "inline_equation" + }, + { + "bbox": [ + 186, + 300, + 223, + 316 + ], + "score": 1.0, + "content": ". We find", + "type": "text" + }, + { + "bbox": [ + 224, + 302, + 324, + 315 + ], + "score": 0.92, + "content": "\\begin{array} { r } { \\Delta \\epsilon = \\arg \\operatorname* { m i n } _ { \\Delta \\epsilon } \\mathbb { E } _ { t } \\mathcal { L } ^ { ( t ) } } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 300, + 354, + 316 + ], + "score": 1.0, + "content": "where", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + }, + { + "type": "interline_equation", + "bbox": [ + 146, + 318, + 464, + 334 + ], + "lines": [ + { + "bbox": [ + 146, + 318, + 464, + 334 + ], + "spans": [ + { + "bbox": [ + 146, + 318, + 464, + 334 + ], + "score": 0.88, + "content": "\\begin{array} { r } { \\mathcal { L } ^ { ( t ) } = \\lambda _ { \\mathrm { { C L I P } } } \\mathcal { L } _ { \\mathrm { d i r e c t i o n } } \\left( \\mathbf { P } _ { t } ^ { \\mathrm { { e d i t } } } , y ^ { \\mathrm { { r e f } } } ; \\mathbf { P } _ { t } ^ { \\mathrm { { s o u r c e } } } , y ^ { \\mathrm { { s o u r c e } } } \\right) + \\lambda _ { \\mathrm { { r e c o n } } } \\left| \\mathbf { P } _ { t } ^ { \\mathrm { { e d i t } } } - \\mathbf { P } _ { t } ^ { \\mathrm { { s o u r c e } } } \\right| } \\end{array}", + "type": "interline_equation", + "image_path": "bc3060069cd5b867facf9896b4c1a26fdf90a15368d85d880fd3d21e20e88014.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 146, + 318, + 464, + 334 + ], + "spans": [], + "index": 12 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 343, + 505, + 367 + ], + "lines": [ + { + "bbox": [ + 102, + 341, + 508, + 361 + ], + "spans": [ + { + "bbox": [ + 102, + 341, + 147, + 361 + ], + "score": 1.0, + "content": "Although", + "type": "text" + }, + { + "bbox": [ + 147, + 345, + 162, + 355 + ], + "score": 0.83, + "content": "\\Delta \\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 162, + 341, + 299, + 361 + ], + "score": 1.0, + "content": "indeed renders the attribute in the", + "type": "text" + }, + { + "bbox": [ + 299, + 344, + 321, + 357 + ], + "score": 0.9, + "content": "\\pmb { x } _ { 0 } ^ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 321, + 341, + 324, + 361 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 325, + 347, + 331, + 355 + ], + "score": 0.69, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 341, + 508, + 361 + ], + "score": 1.0, + "content": "-space lacks the necessary properties of the", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 104, + 354, + 430, + 370 + ], + "spans": [ + { + "bbox": [ + 104, + 354, + 430, + 370 + ], + "score": 1.0, + "content": "semantic latent space in diffusion models that will be described in the following.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5 + }, + { + "type": "title", + "bbox": [ + 107, + 380, + 163, + 391 + ], + "lines": [ + { + "bbox": [ + 105, + 378, + 165, + 393 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 130, + 393 + ], + "score": 1.0, + "content": "3.3", + "type": "text" + }, + { + "bbox": [ + 130, + 381, + 137, + 390 + ], + "score": 0.3, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 378, + 165, + 393 + ], + "score": 1.0, + "content": "-space", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 107, + 400, + 505, + 435 + ], + "lines": [ + { + "bbox": [ + 105, + 399, + 505, + 413 + ], + "spans": [ + { + "bbox": [ + 105, + 399, + 146, + 413 + ], + "score": 1.0, + "content": "Note that", + "type": "text" + }, + { + "bbox": [ + 146, + 400, + 157, + 413 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 157, + 399, + 505, + 413 + ], + "score": 1.0, + "content": "is implemented as U-Net in all state-of-the-art diffusion models. We choose its bottle-", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 411, + 505, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 232, + 424 + ], + "score": 1.0, + "content": "neck, the deepest feature maps", + "type": "text" + }, + { + "bbox": [ + 233, + 412, + 244, + 423 + ], + "score": 0.87, + "content": "\\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 245, + 411, + 290, + 424 + ], + "score": 1.0, + "content": ", to control", + "type": "text" + }, + { + "bbox": [ + 290, + 411, + 301, + 424 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 301, + 411, + 351, + 424 + ], + "score": 1.0, + "content": ". By design,", + "type": "text" + }, + { + "bbox": [ + 351, + 412, + 362, + 423 + ], + "score": 0.87, + "content": "h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 363, + 411, + 505, + 424 + ], + "score": 1.0, + "content": "has smaller spatial resolutions and", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 104, + 421, + 408, + 437 + ], + "spans": [ + { + "bbox": [ + 104, + 421, + 210, + 437 + ], + "score": 1.0, + "content": "high-level semantics than", + "type": "text" + }, + { + "bbox": [ + 210, + 422, + 221, + 435 + ], + "score": 0.88, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 221, + 421, + 408, + 437 + ], + "score": 1.0, + "content": ". Accordingly, the sampling equation becomes", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17 + }, + { + "type": "interline_equation", + "bbox": [ + 191, + 448, + 418, + 464 + ], + "lines": [ + { + "bbox": [ + 191, + 448, + 418, + 464 + ], + "spans": [ + { + "bbox": [ + 191, + 448, + 418, + 464 + ], + "score": 0.9, + "content": "\\pmb { x } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } | \\Delta \\pmb { h } _ { t } ) ) + \\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) ) + \\sigma _ { t } \\boldsymbol { z } _ { t } ,", + "type": "interline_equation", + "image_path": "9ea68511b921a6efe10799fe6a19c64192152e029b8e4254229bf856122276cb.jpg" + } + ] + } + ], + "index": 19, + "virtual_lines": [ + { + "bbox": [ + 191, + 448, + 418, + 464 + ], + "spans": [], + "index": 19 + } + ] + }, + { + "type": "text", + "bbox": [ + 108, + 466, + 503, + 489 + ], + "lines": [ + { + "bbox": [ + 106, + 465, + 505, + 479 + ], + "spans": [ + { + "bbox": [ + 106, + 465, + 133, + 479 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 133, + 465, + 182, + 478 + ], + "score": 0.91, + "content": "\\epsilon _ { t } ^ { \\theta } ( \\boldsymbol { x } _ { t } | \\Delta h _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 183, + 465, + 205, + 479 + ], + "score": 1.0, + "content": "adds", + "type": "text" + }, + { + "bbox": [ + 205, + 466, + 224, + 477 + ], + "score": 0.89, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 465, + 337, + 479 + ], + "score": 1.0, + "content": "to the original feature maps", + "type": "text" + }, + { + "bbox": [ + 337, + 466, + 348, + 477 + ], + "score": 0.88, + "content": "h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 465, + 370, + 479 + ], + "score": 1.0, + "content": ". The", + "type": "text" + }, + { + "bbox": [ + 370, + 466, + 390, + 477 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 390, + 465, + 505, + 479 + ], + "score": 1.0, + "content": "minimizing the same loss in", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 476, + 476, + 490 + ], + "spans": [ + { + "bbox": [ + 105, + 476, + 157, + 490 + ], + "score": 1.0, + "content": "Eq. (7) with", + "type": "text" + }, + { + "bbox": [ + 158, + 477, + 224, + 489 + ], + "score": 0.9, + "content": "\\dot { \\mathbf { P } } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } | \\Delta \\pmb { h } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 476, + 268, + 490 + ], + "score": 1.0, + "content": "instead of", + "type": "text" + }, + { + "bbox": [ + 268, + 477, + 315, + 489 + ], + "score": 0.93, + "content": "\\mathbf { P } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( x _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 316, + 476, + 476, + 490 + ], + "score": 1.0, + "content": "successfully manipulates the attributes.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20.5 + }, + { + "type": "text", + "bbox": [ + 107, + 493, + 452, + 506 + ], + "lines": [ + { + "bbox": [ + 106, + 493, + 453, + 507 + ], + "spans": [ + { + "bbox": [ + 106, + 493, + 172, + 507 + ], + "score": 1.0, + "content": "We observe that", + "type": "text" + }, + { + "bbox": [ + 173, + 495, + 179, + 504 + ], + "score": 0.76, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 180, + 493, + 453, + 507 + ], + "score": 1.0, + "content": "-space in Asyrp has the following properties that others do not have.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 133, + 513, + 504, + 585 + ], + "lines": [ + { + "bbox": [ + 133, + 513, + 382, + 526 + ], + "spans": [ + { + "bbox": [ + 133, + 513, + 183, + 526 + ], + "score": 1.0, + "content": "• The same", + "type": "text" + }, + { + "bbox": [ + 184, + 514, + 200, + 524 + ], + "score": 0.82, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 513, + 382, + 526 + ], + "score": 1.0, + "content": "leads to the same effect on different samples.", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 133, + 529, + 505, + 542 + ], + "spans": [ + { + "bbox": [ + 133, + 529, + 209, + 542 + ], + "score": 1.0, + "content": "• Linearly scaling", + "type": "text" + }, + { + "bbox": [ + 209, + 529, + 226, + 539 + ], + "score": 0.79, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 226, + 529, + 505, + 542 + ], + "score": 1.0, + "content": "controls the magnitude of attribute change, even with negative scales.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 134, + 543, + 492, + 556 + ], + "spans": [ + { + "bbox": [ + 134, + 543, + 210, + 556 + ], + "score": 1.0, + "content": "• Adding multiple", + "type": "text" + }, + { + "bbox": [ + 211, + 543, + 227, + 554 + ], + "score": 0.78, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 227, + 543, + 492, + 556 + ], + "score": 1.0, + "content": "manipulates the corresponding multiple attributes simultaneously.", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 133, + 557, + 423, + 571 + ], + "spans": [ + { + "bbox": [ + 133, + 557, + 142, + 571 + ], + "score": 1.0, + "content": "•", + "type": "text" + }, + { + "bbox": [ + 142, + 558, + 159, + 568 + ], + "score": 0.75, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 557, + 423, + 571 + ], + "score": 1.0, + "content": "preserves the quality of the resulting images without degradation.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 132, + 572, + 362, + 585 + ], + "spans": [ + { + "bbox": [ + 132, + 572, + 142, + 585 + ], + "score": 1.0, + "content": "•", + "type": "text" + }, + { + "bbox": [ + 142, + 572, + 162, + 583 + ], + "score": 0.89, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 162, + 572, + 352, + 585 + ], + "score": 1.0, + "content": "is roughly consistent across different timesteps", + "type": "text" + }, + { + "bbox": [ + 352, + 574, + 357, + 582 + ], + "score": 0.72, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 357, + 572, + 362, + 585 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 107, + 592, + 504, + 615 + ], + "lines": [ + { + "bbox": [ + 105, + 591, + 506, + 606 + ], + "spans": [ + { + "bbox": [ + 105, + 591, + 316, + 606 + ], + "score": 1.0, + "content": "The above properties are demonstrated thoroughly in", + "type": "text" + }, + { + "bbox": [ + 316, + 593, + 336, + 604 + ], + "score": 0.69, + "content": "\\ S 5 . 3", + "type": "inline_equation" + }, + { + "bbox": [ + 337, + 591, + 471, + 606 + ], + "score": 1.0, + "content": ". Appendix D.3 provides details of", + "type": "text" + }, + { + "bbox": [ + 472, + 594, + 478, + 603 + ], + "score": 0.77, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 591, + 506, + 606 + ], + "score": 1.0, + "content": "-space", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 604, + 300, + 617 + ], + "spans": [ + { + "bbox": [ + 105, + 604, + 300, + 617 + ], + "score": 1.0, + "content": "and suboptimal results from alternative choices.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 28.5 + }, + { + "type": "title", + "bbox": [ + 108, + 628, + 263, + 640 + ], + "lines": [ + { + "bbox": [ + 106, + 628, + 264, + 641 + ], + "spans": [ + { + "bbox": [ + 106, + 628, + 264, + 641 + ], + "score": 1.0, + "content": "3.4 IMPLICIT NEURAL DIRECTIONS", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 107, + 648, + 505, + 705 + ], + "lines": [ + { + "bbox": [ + 105, + 648, + 505, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 147, + 662 + ], + "score": 1.0, + "content": "Although", + "type": "text" + }, + { + "bbox": [ + 147, + 649, + 163, + 659 + ], + "score": 0.75, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 164, + 648, + 381, + 662 + ], + "score": 1.0, + "content": "succeeds in manipulating images, directly optimizing", + "type": "text" + }, + { + "bbox": [ + 381, + 649, + 401, + 660 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 401, + 648, + 505, + 662 + ], + "score": 1.0, + "content": "on multiple timesteps re-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 660, + 505, + 673 + ], + "spans": [ + { + "bbox": [ + 105, + 660, + 505, + 673 + ], + "score": 1.0, + "content": "quires many iterations of training with a carefully chosen learning rate and its scheduling. Instead,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 670, + 506, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 228, + 684 + ], + "score": 1.0, + "content": "we define an implicit function", + "type": "text" + }, + { + "bbox": [ + 228, + 671, + 256, + 682 + ], + "score": 0.92, + "content": "f _ { t } ( h _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 256, + 670, + 322, + 684 + ], + "score": 1.0, + "content": "which produces", + "type": "text" + }, + { + "bbox": [ + 323, + 671, + 342, + 682 + ], + "score": 0.9, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 342, + 670, + 381, + 684 + ], + "score": 1.0, + "content": "for given", + "type": "text" + }, + { + "bbox": [ + 381, + 671, + 393, + 682 + ], + "score": 0.87, + "content": "h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 393, + 670, + 411, + 684 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 411, + 672, + 416, + 681 + ], + "score": 0.56, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 416, + 670, + 419, + 684 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 420, + 671, + 430, + 682 + ], + "score": 0.81, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 430, + 670, + 506, + 684 + ], + "score": 1.0, + "content": "is implemented as", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 681, + 506, + 694 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 240, + 694 + ], + "score": 1.0, + "content": "a small neural network with two", + "type": "text" + }, + { + "bbox": [ + 240, + 682, + 263, + 692 + ], + "score": 0.89, + "content": "1 \\times 1", + "type": "inline_equation" + }, + { + "bbox": [ + 264, + 681, + 413, + 694 + ], + "score": 1.0, + "content": "convolutions concatenating timestep", + "type": "text" + }, + { + "bbox": [ + 414, + 683, + 419, + 691 + ], + "score": 0.53, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 681, + 482, + 694 + ], + "score": 1.0, + "content": ". See Appendix", + "type": "text" + }, + { + "bbox": [ + 482, + 682, + 489, + 691 + ], + "score": 0.43, + "content": "\\mathrm { E }", + "type": "inline_equation" + }, + { + "bbox": [ + 490, + 681, + 506, + 694 + ], + "score": 1.0, + "content": "for", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 104, + 688, + 473, + 707 + ], + "spans": [ + { + "bbox": [ + 104, + 688, + 374, + 707 + ], + "score": 1.0, + "content": "the details. Accordingly, we optimize the same loss in Eq. (7) with", + "type": "text" + }, + { + "bbox": [ + 374, + 692, + 468, + 705 + ], + "score": 0.91, + "content": "\\mathbf { P } _ { t } ^ { \\mathrm { e d i t } } = \\mathbf { \\bar { P } } _ { t } \\bigl ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } \\mathbf { \\bar { | } } \\mathbf { \\bar { \\mathbf { f } } } _ { t } ) \\bigr )", + "type": "inline_equation" + }, + { + "bbox": [ + 468, + 688, + 473, + 707 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 33 + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 145, + 722 + ], + "score": 1.0, + "content": "Learning", + "type": "text" + }, + { + "bbox": [ + 145, + 710, + 155, + 721 + ], + "score": 0.86, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 709, + 471, + 722 + ], + "score": 1.0, + "content": "is more robust to learning rate settings and converges faster than learning every", + "type": "text" + }, + { + "bbox": [ + 471, + 710, + 490, + 721 + ], + "score": 0.89, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 491, + 709, + 505, + 722 + ], + "score": 1.0, + "content": ". In", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 720, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 155, + 733 + ], + "score": 1.0, + "content": "addition, as", + "type": "text" + }, + { + "bbox": [ + 155, + 721, + 165, + 732 + ], + "score": 0.87, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 166, + 720, + 505, + 733 + ], + "score": 1.0, + "content": "learns an implicit function for given timesteps and bottleneck features, it generalizes", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5 + } + ], + "page_idx": 3, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 310, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 310, + 762 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 12, + "width": 8 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 116, + 80, + 496, + 173 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 116, + 80, + 496, + 173 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 116, + 80, + 496, + 173 + ], + "spans": [ + { + "bbox": [ + 116, + 80, + 496, + 173 + ], + "score": 0.962, + "type": "image", + "image_path": "06ae64f291010710a1b60bf1d063bfc7c5ce461052c3aeb0dadeab995109361f.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 116, + 80, + 496, + 111.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 116, + 111.0, + 496, + 142.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 116, + 142.0, + 496, + 173.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 183, + 505, + 217 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 183, + 505, + 196 + ], + "spans": [ + { + "bbox": [ + 106, + 183, + 505, + 196 + ], + "score": 1.0, + "content": "Figure 2: Generative process of Asyrp. The green box on the left illustrates Asyrp which only", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 193, + 506, + 207 + ], + "spans": [ + { + "bbox": [ + 105, + 193, + 130, + 207 + ], + "score": 1.0, + "content": "alters", + "type": "text" + }, + { + "bbox": [ + 130, + 194, + 143, + 205 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 143, + 193, + 212, + 207 + ], + "score": 1.0, + "content": "while preserving", + "type": "text" + }, + { + "bbox": [ + 213, + 194, + 226, + 205 + ], + "score": 0.89, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 226, + 193, + 506, + 207 + ], + "score": 1.0, + "content": "shared by DDIM. The right describes that Asyrp modifies the original", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 205, + 407, + 219 + ], + "spans": [ + { + "bbox": [ + 105, + 205, + 370, + 219 + ], + "score": 1.0, + "content": "reverse process toward the target attribute reflecting the change in", + "type": "text" + }, + { + "bbox": [ + 370, + 206, + 377, + 215 + ], + "score": 0.53, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 377, + 205, + 407, + 219 + ], + "score": 1.0, + "content": "-space.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 106, + 239, + 506, + 275 + ], + "lines": [ + { + "bbox": [ + 105, + 239, + 506, + 253 + ], + "spans": [ + { + "bbox": [ + 105, + 239, + 188, + 253 + ], + "score": 1.0, + "content": "i.e., we modify only", + "type": "text" + }, + { + "bbox": [ + 189, + 241, + 201, + 252 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 201, + 239, + 247, + 253 + ], + "score": 1.0, + "content": "by shifting", + "type": "text" + }, + { + "bbox": [ + 248, + 240, + 258, + 252 + ], + "score": 0.88, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 258, + 239, + 269, + 253 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 269, + 240, + 280, + 252 + ], + "score": 0.86, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 280, + 239, + 349, + 253 + ], + "score": 1.0, + "content": "while preserving", + "type": "text" + }, + { + "bbox": [ + 350, + 241, + 363, + 252 + ], + "score": 0.87, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 363, + 239, + 506, + 253 + ], + "score": 1.0, + "content": ". Intuitively, it modifies the original", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 250, + 506, + 266 + ], + "spans": [ + { + "bbox": [ + 105, + 250, + 223, + 266 + ], + "score": 1.0, + "content": "reverse process according to", + "type": "text" + }, + { + "bbox": [ + 224, + 252, + 287, + 264 + ], + "score": 0.92, + "content": "\\Delta \\epsilon _ { t } = \\widetilde { \\epsilon } _ { t } ^ { \\theta } - \\widetilde { \\epsilon } _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 287, + 250, + 463, + 266 + ], + "score": 1.0, + "content": "while it does not alter the direction toward", + "type": "text" + }, + { + "bbox": [ + 463, + 253, + 474, + 263 + ], + "score": 0.86, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 475, + 250, + 506, + 266 + ], + "score": 1.0, + "content": "so that", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 262, + 491, + 275 + ], + "spans": [ + { + "bbox": [ + 106, + 264, + 128, + 274 + ], + "score": 0.86, + "content": "{ \\mathbf { \\mathcal { x } } } _ { t - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 128, + 262, + 229, + 275 + ], + "score": 1.0, + "content": "follows the original flow", + "type": "text" + }, + { + "bbox": [ + 230, + 263, + 243, + 273 + ], + "score": 0.87, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 243, + 262, + 491, + 275 + ], + "score": 1.0, + "content": "at each sampling step. Figure 2 illustrates the above intuition.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7, + "bbox_fs": [ + 105, + 239, + 506, + 275 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 278, + 505, + 315 + ], + "lines": [ + { + "bbox": [ + 105, + 277, + 507, + 292 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 306, + 292 + ], + "score": 1.0, + "content": "As in Avrahami et al. (2022), we use the modified", + "type": "text" + }, + { + "bbox": [ + 307, + 279, + 329, + 291 + ], + "score": 0.89, + "content": "\\mathbf { P } _ { t } ^ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 329, + 277, + 395, + 292 + ], + "score": 1.0, + "content": "and the original", + "type": "text" + }, + { + "bbox": [ + 395, + 280, + 426, + 291 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t } ^ { \\mathrm { s o u r c e } }", + "type": "inline_equation" + }, + { + "bbox": [ + 426, + 277, + 507, + 292 + ], + "score": 1.0, + "content": "as visual inputs for", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 104, + 289, + 506, + 303 + ], + "spans": [ + { + "bbox": [ + 104, + 289, + 463, + 303 + ], + "score": 1.0, + "content": "the directional CLIP loss in Eq. (4), and regularize the difference between the modified", + "type": "text" + }, + { + "bbox": [ + 464, + 290, + 486, + 303 + ], + "score": 0.91, + "content": "\\mathbf { P } _ { t } ^ { \\mathrm { { e d i t } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 486, + 289, + 506, + 303 + ], + "score": 1.0, + "content": "and", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 104, + 300, + 354, + 316 + ], + "spans": [ + { + "bbox": [ + 104, + 300, + 154, + 316 + ], + "score": 1.0, + "content": "the original", + "type": "text" + }, + { + "bbox": [ + 155, + 303, + 185, + 315 + ], + "score": 0.87, + "content": "\\mathbf { P } _ { t } ^ { \\mathrm { s o u r c e } }", + "type": "inline_equation" + }, + { + "bbox": [ + 186, + 300, + 223, + 316 + ], + "score": 1.0, + "content": ". We find", + "type": "text" + }, + { + "bbox": [ + 224, + 302, + 324, + 315 + ], + "score": 0.92, + "content": "\\begin{array} { r } { \\Delta \\epsilon = \\arg \\operatorname* { m i n } _ { \\Delta \\epsilon } \\mathbb { E } _ { t } \\mathcal { L } ^ { ( t ) } } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 300, + 354, + 316 + ], + "score": 1.0, + "content": "where", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10, + "bbox_fs": [ + 104, + 277, + 507, + 316 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 146, + 318, + 464, + 334 + ], + "lines": [ + { + "bbox": [ + 146, + 318, + 464, + 334 + ], + "spans": [ + { + "bbox": [ + 146, + 318, + 464, + 334 + ], + "score": 0.88, + "content": "\\begin{array} { r } { \\mathcal { L } ^ { ( t ) } = \\lambda _ { \\mathrm { { C L I P } } } \\mathcal { L } _ { \\mathrm { d i r e c t i o n } } \\left( \\mathbf { P } _ { t } ^ { \\mathrm { { e d i t } } } , y ^ { \\mathrm { { r e f } } } ; \\mathbf { P } _ { t } ^ { \\mathrm { { s o u r c e } } } , y ^ { \\mathrm { { s o u r c e } } } \\right) + \\lambda _ { \\mathrm { { r e c o n } } } \\left| \\mathbf { P } _ { t } ^ { \\mathrm { { e d i t } } } - \\mathbf { P } _ { t } ^ { \\mathrm { { s o u r c e } } } \\right| } \\end{array}", + "type": "interline_equation", + "image_path": "bc3060069cd5b867facf9896b4c1a26fdf90a15368d85d880fd3d21e20e88014.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 146, + 318, + 464, + 334 + ], + "spans": [], + "index": 12 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 343, + 505, + 367 + ], + "lines": [ + { + "bbox": [ + 102, + 341, + 508, + 361 + ], + "spans": [ + { + "bbox": [ + 102, + 341, + 147, + 361 + ], + "score": 1.0, + "content": "Although", + "type": "text" + }, + { + "bbox": [ + 147, + 345, + 162, + 355 + ], + "score": 0.83, + "content": "\\Delta \\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 162, + 341, + 299, + 361 + ], + "score": 1.0, + "content": "indeed renders the attribute in the", + "type": "text" + }, + { + "bbox": [ + 299, + 344, + 321, + 357 + ], + "score": 0.9, + "content": "\\pmb { x } _ { 0 } ^ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 321, + 341, + 324, + 361 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 325, + 347, + 331, + 355 + ], + "score": 0.69, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 341, + 508, + 361 + ], + "score": 1.0, + "content": "-space lacks the necessary properties of the", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 104, + 354, + 430, + 370 + ], + "spans": [ + { + "bbox": [ + 104, + 354, + 430, + 370 + ], + "score": 1.0, + "content": "semantic latent space in diffusion models that will be described in the following.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5, + "bbox_fs": [ + 102, + 341, + 508, + 370 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 380, + 163, + 391 + ], + "lines": [ + { + "bbox": [ + 105, + 378, + 165, + 393 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 130, + 393 + ], + "score": 1.0, + "content": "3.3", + "type": "text" + }, + { + "bbox": [ + 130, + 381, + 137, + 390 + ], + "score": 0.3, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 378, + 165, + 393 + ], + "score": 1.0, + "content": "-space", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 107, + 400, + 505, + 435 + ], + "lines": [ + { + "bbox": [ + 105, + 399, + 505, + 413 + ], + "spans": [ + { + "bbox": [ + 105, + 399, + 146, + 413 + ], + "score": 1.0, + "content": "Note that", + "type": "text" + }, + { + "bbox": [ + 146, + 400, + 157, + 413 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 157, + 399, + 505, + 413 + ], + "score": 1.0, + "content": "is implemented as U-Net in all state-of-the-art diffusion models. We choose its bottle-", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 411, + 505, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 232, + 424 + ], + "score": 1.0, + "content": "neck, the deepest feature maps", + "type": "text" + }, + { + "bbox": [ + 233, + 412, + 244, + 423 + ], + "score": 0.87, + "content": "\\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 245, + 411, + 290, + 424 + ], + "score": 1.0, + "content": ", to control", + "type": "text" + }, + { + "bbox": [ + 290, + 411, + 301, + 424 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 301, + 411, + 351, + 424 + ], + "score": 1.0, + "content": ". By design,", + "type": "text" + }, + { + "bbox": [ + 351, + 412, + 362, + 423 + ], + "score": 0.87, + "content": "h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 363, + 411, + 505, + 424 + ], + "score": 1.0, + "content": "has smaller spatial resolutions and", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 104, + 421, + 408, + 437 + ], + "spans": [ + { + "bbox": [ + 104, + 421, + 210, + 437 + ], + "score": 1.0, + "content": "high-level semantics than", + "type": "text" + }, + { + "bbox": [ + 210, + 422, + 221, + 435 + ], + "score": 0.88, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 221, + 421, + 408, + 437 + ], + "score": 1.0, + "content": ". Accordingly, the sampling equation becomes", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17, + "bbox_fs": [ + 104, + 399, + 505, + 437 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 191, + 448, + 418, + 464 + ], + "lines": [ + { + "bbox": [ + 191, + 448, + 418, + 464 + ], + "spans": [ + { + "bbox": [ + 191, + 448, + 418, + 464 + ], + "score": 0.9, + "content": "\\pmb { x } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } | \\Delta \\pmb { h } _ { t } ) ) + \\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) ) + \\sigma _ { t } \\boldsymbol { z } _ { t } ,", + "type": "interline_equation", + "image_path": "9ea68511b921a6efe10799fe6a19c64192152e029b8e4254229bf856122276cb.jpg" + } + ] + } + ], + "index": 19, + "virtual_lines": [ + { + "bbox": [ + 191, + 448, + 418, + 464 + ], + "spans": [], + "index": 19 + } + ] + }, + { + "type": "text", + "bbox": [ + 108, + 466, + 503, + 489 + ], + "lines": [ + { + "bbox": [ + 106, + 465, + 505, + 479 + ], + "spans": [ + { + "bbox": [ + 106, + 465, + 133, + 479 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 133, + 465, + 182, + 478 + ], + "score": 0.91, + "content": "\\epsilon _ { t } ^ { \\theta } ( \\boldsymbol { x } _ { t } | \\Delta h _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 183, + 465, + 205, + 479 + ], + "score": 1.0, + "content": "adds", + "type": "text" + }, + { + "bbox": [ + 205, + 466, + 224, + 477 + ], + "score": 0.89, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 465, + 337, + 479 + ], + "score": 1.0, + "content": "to the original feature maps", + "type": "text" + }, + { + "bbox": [ + 337, + 466, + 348, + 477 + ], + "score": 0.88, + "content": "h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 465, + 370, + 479 + ], + "score": 1.0, + "content": ". The", + "type": "text" + }, + { + "bbox": [ + 370, + 466, + 390, + 477 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 390, + 465, + 505, + 479 + ], + "score": 1.0, + "content": "minimizing the same loss in", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 476, + 476, + 490 + ], + "spans": [ + { + "bbox": [ + 105, + 476, + 157, + 490 + ], + "score": 1.0, + "content": "Eq. (7) with", + "type": "text" + }, + { + "bbox": [ + 158, + 477, + 224, + 489 + ], + "score": 0.9, + "content": "\\dot { \\mathbf { P } } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( \\pmb { x } _ { t } | \\Delta \\pmb { h } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 476, + 268, + 490 + ], + "score": 1.0, + "content": "instead of", + "type": "text" + }, + { + "bbox": [ + 268, + 477, + 315, + 489 + ], + "score": 0.93, + "content": "\\mathbf { P } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( x _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 316, + 476, + 476, + 490 + ], + "score": 1.0, + "content": "successfully manipulates the attributes.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20.5, + "bbox_fs": [ + 105, + 465, + 505, + 490 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 493, + 452, + 506 + ], + "lines": [ + { + "bbox": [ + 106, + 493, + 453, + 507 + ], + "spans": [ + { + "bbox": [ + 106, + 493, + 172, + 507 + ], + "score": 1.0, + "content": "We observe that", + "type": "text" + }, + { + "bbox": [ + 173, + 495, + 179, + 504 + ], + "score": 0.76, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 180, + 493, + 453, + 507 + ], + "score": 1.0, + "content": "-space in Asyrp has the following properties that others do not have.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 22, + "bbox_fs": [ + 106, + 493, + 453, + 507 + ] + }, + { + "type": "list", + "bbox": [ + 133, + 513, + 504, + 585 + ], + "lines": [ + { + "bbox": [ + 133, + 513, + 382, + 526 + ], + "spans": [ + { + "bbox": [ + 133, + 513, + 183, + 526 + ], + "score": 1.0, + "content": "• The same", + "type": "text" + }, + { + "bbox": [ + 184, + 514, + 200, + 524 + ], + "score": 0.82, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 513, + 382, + 526 + ], + "score": 1.0, + "content": "leads to the same effect on different samples.", + "type": "text" + } + ], + "index": 23, + "is_list_end_line": true + }, + { + "bbox": [ + 133, + 529, + 505, + 542 + ], + "spans": [ + { + "bbox": [ + 133, + 529, + 209, + 542 + ], + "score": 1.0, + "content": "• Linearly scaling", + "type": "text" + }, + { + "bbox": [ + 209, + 529, + 226, + 539 + ], + "score": 0.79, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 226, + 529, + 505, + 542 + ], + "score": 1.0, + "content": "controls the magnitude of attribute change, even with negative scales.", + "type": "text" + } + ], + "index": 24, + "is_list_start_line": true, + "is_list_end_line": true + }, + { + "bbox": [ + 134, + 543, + 492, + 556 + ], + "spans": [ + { + "bbox": [ + 134, + 543, + 210, + 556 + ], + "score": 1.0, + "content": "• Adding multiple", + "type": "text" + }, + { + "bbox": [ + 211, + 543, + 227, + 554 + ], + "score": 0.78, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 227, + 543, + 492, + 556 + ], + "score": 1.0, + "content": "manipulates the corresponding multiple attributes simultaneously.", + "type": "text" + } + ], + "index": 25, + "is_list_start_line": true, + "is_list_end_line": true + }, + { + "bbox": [ + 133, + 557, + 423, + 571 + ], + "spans": [ + { + "bbox": [ + 133, + 557, + 142, + 571 + ], + "score": 1.0, + "content": "•", + "type": "text" + }, + { + "bbox": [ + 142, + 558, + 159, + 568 + ], + "score": 0.75, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 557, + 423, + 571 + ], + "score": 1.0, + "content": "preserves the quality of the resulting images without degradation.", + "type": "text" + } + ], + "index": 26, + "is_list_start_line": true, + "is_list_end_line": true + }, + { + "bbox": [ + 132, + 572, + 362, + 585 + ], + "spans": [ + { + "bbox": [ + 132, + 572, + 142, + 585 + ], + "score": 1.0, + "content": "•", + "type": "text" + }, + { + "bbox": [ + 142, + 572, + 162, + 583 + ], + "score": 0.89, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 162, + 572, + 352, + 585 + ], + "score": 1.0, + "content": "is roughly consistent across different timesteps", + "type": "text" + }, + { + "bbox": [ + 352, + 574, + 357, + 582 + ], + "score": 0.72, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 357, + 572, + 362, + 585 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 27, + "is_list_start_line": true, + "is_list_end_line": true + } + ], + "index": 25, + "bbox_fs": [ + 132, + 513, + 505, + 585 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 592, + 504, + 615 + ], + "lines": [ + { + "bbox": [ + 105, + 591, + 506, + 606 + ], + "spans": [ + { + "bbox": [ + 105, + 591, + 316, + 606 + ], + "score": 1.0, + "content": "The above properties are demonstrated thoroughly in", + "type": "text" + }, + { + "bbox": [ + 316, + 593, + 336, + 604 + ], + "score": 0.69, + "content": "\\ S 5 . 3", + "type": "inline_equation" + }, + { + "bbox": [ + 337, + 591, + 471, + 606 + ], + "score": 1.0, + "content": ". Appendix D.3 provides details of", + "type": "text" + }, + { + "bbox": [ + 472, + 594, + 478, + 603 + ], + "score": 0.77, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 591, + 506, + 606 + ], + "score": 1.0, + "content": "-space", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 604, + 300, + 617 + ], + "spans": [ + { + "bbox": [ + 105, + 604, + 300, + 617 + ], + "score": 1.0, + "content": "and suboptimal results from alternative choices.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 28.5, + "bbox_fs": [ + 105, + 591, + 506, + 617 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 628, + 263, + 640 + ], + "lines": [ + { + "bbox": [ + 106, + 628, + 264, + 641 + ], + "spans": [ + { + "bbox": [ + 106, + 628, + 264, + 641 + ], + "score": 1.0, + "content": "3.4 IMPLICIT NEURAL DIRECTIONS", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 107, + 648, + 505, + 705 + ], + "lines": [ + { + "bbox": [ + 105, + 648, + 505, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 147, + 662 + ], + "score": 1.0, + "content": "Although", + "type": "text" + }, + { + "bbox": [ + 147, + 649, + 163, + 659 + ], + "score": 0.75, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 164, + 648, + 381, + 662 + ], + "score": 1.0, + "content": "succeeds in manipulating images, directly optimizing", + "type": "text" + }, + { + "bbox": [ + 381, + 649, + 401, + 660 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 401, + 648, + 505, + 662 + ], + "score": 1.0, + "content": "on multiple timesteps re-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 660, + 505, + 673 + ], + "spans": [ + { + "bbox": [ + 105, + 660, + 505, + 673 + ], + "score": 1.0, + "content": "quires many iterations of training with a carefully chosen learning rate and its scheduling. Instead,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 670, + 506, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 228, + 684 + ], + "score": 1.0, + "content": "we define an implicit function", + "type": "text" + }, + { + "bbox": [ + 228, + 671, + 256, + 682 + ], + "score": 0.92, + "content": "f _ { t } ( h _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 256, + 670, + 322, + 684 + ], + "score": 1.0, + "content": "which produces", + "type": "text" + }, + { + "bbox": [ + 323, + 671, + 342, + 682 + ], + "score": 0.9, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 342, + 670, + 381, + 684 + ], + "score": 1.0, + "content": "for given", + "type": "text" + }, + { + "bbox": [ + 381, + 671, + 393, + 682 + ], + "score": 0.87, + "content": "h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 393, + 670, + 411, + 684 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 411, + 672, + 416, + 681 + ], + "score": 0.56, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 416, + 670, + 419, + 684 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 420, + 671, + 430, + 682 + ], + "score": 0.81, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 430, + 670, + 506, + 684 + ], + "score": 1.0, + "content": "is implemented as", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 681, + 506, + 694 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 240, + 694 + ], + "score": 1.0, + "content": "a small neural network with two", + "type": "text" + }, + { + "bbox": [ + 240, + 682, + 263, + 692 + ], + "score": 0.89, + "content": "1 \\times 1", + "type": "inline_equation" + }, + { + "bbox": [ + 264, + 681, + 413, + 694 + ], + "score": 1.0, + "content": "convolutions concatenating timestep", + "type": "text" + }, + { + "bbox": [ + 414, + 683, + 419, + 691 + ], + "score": 0.53, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 681, + 482, + 694 + ], + "score": 1.0, + "content": ". See Appendix", + "type": "text" + }, + { + "bbox": [ + 482, + 682, + 489, + 691 + ], + "score": 0.43, + "content": "\\mathrm { E }", + "type": "inline_equation" + }, + { + "bbox": [ + 490, + 681, + 506, + 694 + ], + "score": 1.0, + "content": "for", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 104, + 688, + 473, + 707 + ], + "spans": [ + { + "bbox": [ + 104, + 688, + 374, + 707 + ], + "score": 1.0, + "content": "the details. Accordingly, we optimize the same loss in Eq. (7) with", + "type": "text" + }, + { + "bbox": [ + 374, + 692, + 468, + 705 + ], + "score": 0.91, + "content": "\\mathbf { P } _ { t } ^ { \\mathrm { e d i t } } = \\mathbf { \\bar { P } } _ { t } \\bigl ( \\epsilon _ { t } ^ { \\theta } ( { \\pmb x } _ { t } \\mathbf { \\bar { | } } \\mathbf { \\bar { \\mathbf { f } } } _ { t } ) \\bigr )", + "type": "inline_equation" + }, + { + "bbox": [ + 468, + 688, + 473, + 707 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 33, + "bbox_fs": [ + 104, + 648, + 506, + 707 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 145, + 722 + ], + "score": 1.0, + "content": "Learning", + "type": "text" + }, + { + "bbox": [ + 145, + 710, + 155, + 721 + ], + "score": 0.86, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 709, + 471, + 722 + ], + "score": 1.0, + "content": "is more robust to learning rate settings and converges faster than learning every", + "type": "text" + }, + { + "bbox": [ + 471, + 710, + 490, + 721 + ], + "score": 0.89, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 491, + 709, + 505, + 722 + ], + "score": 1.0, + "content": ". In", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 720, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 155, + 733 + ], + "score": 1.0, + "content": "addition, as", + "type": "text" + }, + { + "bbox": [ + 155, + 721, + 165, + 732 + ], + "score": 0.87, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 166, + 720, + 505, + 733 + ], + "score": 1.0, + "content": "learns an implicit function for given timesteps and bottleneck features, it generalizes", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 331, + 505, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 331, + 505, + 343 + ], + "score": 1.0, + "content": "to unseen timesteps and bottleneck features. The generalization allows us to borrow the accelerated", + "type": "text", + "cross_page": true + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 341, + 506, + 356 + ], + "spans": [ + { + "bbox": [ + 105, + 341, + 320, + 356 + ], + "score": 1.0, + "content": "training scheme of DDIM defined on a subsequence", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 320, + 342, + 374, + 355 + ], + "score": 0.93, + "content": "\\{ \\pmb { x } _ { \\tau _ { i } } \\} _ { \\forall i \\in [ 1 , S ] }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 375, + 341, + 403, + 356 + ], + "score": 1.0, + "content": "where", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 404, + 342, + 423, + 354 + ], + "score": 0.92, + "content": "\\{ \\tau _ { i } \\}", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 423, + 341, + 506, + 356 + ], + "score": 1.0, + "content": "is a subsequence of", + "type": "text", + "cross_page": true + } + ], + "index": 11 + }, + { + "bbox": [ + 107, + 352, + 504, + 368 + ], + "spans": [ + { + "bbox": [ + 107, + 354, + 142, + 366 + ], + "score": 0.89, + "content": "[ 1 , . . . , T ]", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 143, + 352, + 163, + 368 + ], + "score": 1.0, + "content": "and", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 163, + 354, + 195, + 364 + ], + "score": 0.92, + "content": "S \\textless T", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 195, + 352, + 485, + 368 + ], + "score": 1.0, + "content": ". Then, we can use the generative process with a custom subsequence", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 485, + 354, + 504, + 366 + ], + "score": 0.9, + "content": "\\{ \\tilde { \\tau } _ { i } \\}", + "type": "inline_equation", + "cross_page": true + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 365, + 505, + 380 + ], + "spans": [ + { + "bbox": [ + 105, + 365, + 154, + 380 + ], + "score": 1.0, + "content": "with length", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 154, + 365, + 182, + 377 + ], + "score": 0.91, + "content": "\\tilde { S } < T", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 183, + 365, + 277, + 380 + ], + "score": 1.0, + "content": "through normalization:", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 278, + 365, + 360, + 379 + ], + "score": 0.94, + "content": "\\Delta \\tilde { h } _ { \\tilde { \\tau } } = f _ { \\tilde { \\tau } } ( h _ { \\tilde { \\tau } } ) { S } / { \\tilde { S } }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 361, + 365, + 469, + 380 + ], + "score": 1.0, + "content": ". It preserves the amount of", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 470, + 366, + 501, + 379 + ], + "score": 0.9, + "content": "\\sum \\Delta { { h } _ { t } }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 501, + 365, + 505, + 380 + ], + "score": 1.0, + "content": ",", + "type": "text", + "cross_page": true + } + ], + "index": 13 + }, + { + "bbox": [ + 107, + 377, + 505, + 394 + ], + "spans": [ + { + "bbox": [ + 107, + 378, + 173, + 390 + ], + "score": 0.92, + "content": "\\Delta \\tilde { h } _ { \\tilde { \\tau } } \\tilde { S } = \\Delta h _ { t } S", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 174, + 377, + 266, + 394 + ], + "score": 1.0, + "content": ". Therefore, we can use", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 267, + 380, + 277, + 391 + ], + "score": 0.87, + "content": "f _ { t }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 277, + 377, + 505, + 394 + ], + "score": 1.0, + "content": "trained on any subsequence for any length of the genera-", + "type": "text", + "cross_page": true + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 390, + 504, + 404 + ], + "spans": [ + { + "bbox": [ + 105, + 390, + 301, + 404 + ], + "score": 1.0, + "content": "tive process. See Appendix F for details. We use", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 302, + 391, + 312, + 402 + ], + "score": 0.88, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 312, + 390, + 338, + 404 + ], + "score": 1.0, + "content": "to get", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 338, + 391, + 358, + 402 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 358, + 390, + 504, + 404 + ], + "score": 1.0, + "content": "for all experiments except Figure 6.", + "type": "text", + "cross_page": true + } + ], + "index": 15 + } + ], + "index": 36.5, + "bbox_fs": [ + 105, + 709, + 505, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 116, + 85, + 491, + 213 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 116, + 85, + 491, + 213 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 116, + 85, + 491, + 213 + ], + "spans": [ + { + "bbox": [ + 116, + 85, + 491, + 213 + ], + "score": 0.973, + "type": "image", + "image_path": "47dbd6b534787f72c5d6825a64bc954aec392ef479c4aed91b47b808b2fa38d3.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 116, + 85, + 491, + 127.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 116, + 127.66666666666666, + 491, + 170.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 116, + 170.33333333333331, + 491, + 212.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 228, + 505, + 306 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 228, + 505, + 241 + ], + "spans": [ + { + "bbox": [ + 106, + 228, + 505, + 241 + ], + "score": 1.0, + "content": "Figure 3: Intuition for choosing the intervals for editing and quality boosting. We choose the", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 239, + 505, + 253 + ], + "spans": [ + { + "bbox": [ + 105, + 239, + 460, + 253 + ], + "score": 1.0, + "content": "intervals by quantifying two measures (top left inset). The editing strength of an interval", + "type": "text" + }, + { + "bbox": [ + 460, + 240, + 481, + 252 + ], + "score": 0.91, + "content": "[ T , t ]", + "type": "inline_equation" + }, + { + "bbox": [ + 482, + 239, + 505, + 253 + ], + "score": 1.0, + "content": "mea-", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 104, + 250, + 506, + 265 + ], + "spans": [ + { + "bbox": [ + 104, + 250, + 251, + 265 + ], + "score": 1.0, + "content": "sures its perceptual difference from", + "type": "text" + }, + { + "bbox": [ + 252, + 251, + 261, + 261 + ], + "score": 0.8, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 250, + 282, + 265 + ], + "score": 1.0, + "content": "until", + "type": "text" + }, + { + "bbox": [ + 282, + 252, + 288, + 261 + ], + "score": 0.59, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 289, + 250, + 322, + 265 + ], + "score": 1.0, + "content": ". We set", + "type": "text" + }, + { + "bbox": [ + 322, + 251, + 343, + 263 + ], + "score": 0.89, + "content": "[ T , t ]", + "type": "inline_equation" + }, + { + "bbox": [ + 344, + 250, + 506, + 265 + ], + "score": 1.0, + "content": "to the interval with the smallest editing", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 262, + 504, + 274 + ], + "spans": [ + { + "bbox": [ + 106, + 262, + 206, + 273 + ], + "score": 1.0, + "content": "strength that synthesizes", + "type": "text" + }, + { + "bbox": [ + 206, + 262, + 219, + 273 + ], + "score": 0.89, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 219, + 262, + 253, + 273 + ], + "score": 1.0, + "content": "close to", + "type": "text" + }, + { + "bbox": [ + 253, + 264, + 261, + 272 + ], + "score": 0.76, + "content": "_ { \\textbf { \\em x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 262, + 282, + 273 + ], + "score": 1.0, + "content": ", i.e.,", + "type": "text" + }, + { + "bbox": [ + 282, + 262, + 373, + 274 + ], + "score": 0.73, + "content": "\\mathrm { L P I P S } ( { \\pmb x } , { \\bf P } _ { t } ) = 0 . 3 3", + "type": "inline_equation" + }, + { + "bbox": [ + 373, + 262, + 504, + 273 + ], + "score": 1.0, + "content": ". Editing flexibility of an interval", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 107, + 272, + 506, + 286 + ], + "spans": [ + { + "bbox": [ + 107, + 273, + 126, + 285 + ], + "score": 0.88, + "content": "[ t , 0 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 127, + 272, + 318, + 286 + ], + "score": 1.0, + "content": "measures the potential amount of changes after", + "type": "text" + }, + { + "bbox": [ + 318, + 275, + 323, + 283 + ], + "score": 0.71, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 323, + 272, + 412, + 286 + ], + "score": 1.0, + "content": ". Quality deficiency at", + "type": "text" + }, + { + "bbox": [ + 412, + 274, + 417, + 283 + ], + "score": 0.75, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 417, + 272, + 506, + 286 + ], + "score": 1.0, + "content": "measures the amount", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 283, + 506, + 297 + ], + "spans": [ + { + "bbox": [ + 105, + 283, + 150, + 297 + ], + "score": 1.0, + "content": "of noise in", + "type": "text" + }, + { + "bbox": [ + 151, + 285, + 162, + 295 + ], + "score": 0.86, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 162, + 283, + 195, + 297 + ], + "score": 1.0, + "content": ". We set", + "type": "text" + }, + { + "bbox": [ + 195, + 284, + 214, + 296 + ], + "score": 0.88, + "content": "[ t , 0 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 215, + 283, + 371, + 297 + ], + "score": 1.0, + "content": "to handle large quality deficiency (i.e.,", + "type": "text" + }, + { + "bbox": [ + 371, + 284, + 458, + 296 + ], + "score": 0.84, + "content": "\\mathrm { L P I P S } ( \\pmb { x } , \\pmb { x } _ { t } ) = 1 . 2 )", + "type": "inline_equation" + }, + { + "bbox": [ + 459, + 283, + 506, + 297 + ], + "score": 1.0, + "content": "with small", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 294, + 180, + 309 + ], + "spans": [ + { + "bbox": [ + 105, + 294, + 180, + 309 + ], + "score": 1.0, + "content": "editing flexibility.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 6 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 106, + 330, + 505, + 403 + ], + "lines": [ + { + "bbox": [ + 106, + 331, + 505, + 343 + ], + "spans": [ + { + "bbox": [ + 106, + 331, + 505, + 343 + ], + "score": 1.0, + "content": "to unseen timesteps and bottleneck features. The generalization allows us to borrow the accelerated", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 341, + 506, + 356 + ], + "spans": [ + { + "bbox": [ + 105, + 341, + 320, + 356 + ], + "score": 1.0, + "content": "training scheme of DDIM defined on a subsequence", + "type": "text" + }, + { + "bbox": [ + 320, + 342, + 374, + 355 + ], + "score": 0.93, + "content": "\\{ \\pmb { x } _ { \\tau _ { i } } \\} _ { \\forall i \\in [ 1 , S ] }", + "type": "inline_equation" + }, + { + "bbox": [ + 375, + 341, + 403, + 356 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 404, + 342, + 423, + 354 + ], + "score": 0.92, + "content": "\\{ \\tau _ { i } \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 423, + 341, + 506, + 356 + ], + "score": 1.0, + "content": "is a subsequence of", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 107, + 352, + 504, + 368 + ], + "spans": [ + { + "bbox": [ + 107, + 354, + 142, + 366 + ], + "score": 0.89, + "content": "[ 1 , . . . , T ]", + "type": "inline_equation" + }, + { + "bbox": [ + 143, + 352, + 163, + 368 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 163, + 354, + 195, + 364 + ], + "score": 0.92, + "content": "S \\textless T", + "type": "inline_equation" + }, + { + "bbox": [ + 195, + 352, + 485, + 368 + ], + "score": 1.0, + "content": ". Then, we can use the generative process with a custom subsequence", + "type": "text" + }, + { + "bbox": [ + 485, + 354, + 504, + 366 + ], + "score": 0.9, + "content": "\\{ \\tilde { \\tau } _ { i } \\}", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 365, + 505, + 380 + ], + "spans": [ + { + "bbox": [ + 105, + 365, + 154, + 380 + ], + "score": 1.0, + "content": "with length", + "type": "text" + }, + { + "bbox": [ + 154, + 365, + 182, + 377 + ], + "score": 0.91, + "content": "\\tilde { S } < T", + "type": "inline_equation" + }, + { + "bbox": [ + 183, + 365, + 277, + 380 + ], + "score": 1.0, + "content": "through normalization:", + "type": "text" + }, + { + "bbox": [ + 278, + 365, + 360, + 379 + ], + "score": 0.94, + "content": "\\Delta \\tilde { h } _ { \\tilde { \\tau } } = f _ { \\tilde { \\tau } } ( h _ { \\tilde { \\tau } } ) { S } / { \\tilde { S } }", + "type": "inline_equation" + }, + { + "bbox": [ + 361, + 365, + 469, + 380 + ], + "score": 1.0, + "content": ". It preserves the amount of", + "type": "text" + }, + { + "bbox": [ + 470, + 366, + 501, + 379 + ], + "score": 0.9, + "content": "\\sum \\Delta { { h } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 501, + 365, + 505, + 380 + ], + "score": 1.0, + "content": ",", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 107, + 377, + 505, + 394 + ], + "spans": [ + { + "bbox": [ + 107, + 378, + 173, + 390 + ], + "score": 0.92, + "content": "\\Delta \\tilde { h } _ { \\tilde { \\tau } } \\tilde { S } = \\Delta h _ { t } S", + "type": "inline_equation" + }, + { + "bbox": [ + 174, + 377, + 266, + 394 + ], + "score": 1.0, + "content": ". Therefore, we can use", + "type": "text" + }, + { + "bbox": [ + 267, + 380, + 277, + 391 + ], + "score": 0.87, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 277, + 377, + 505, + 394 + ], + "score": 1.0, + "content": "trained on any subsequence for any length of the genera-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 390, + 504, + 404 + ], + "spans": [ + { + "bbox": [ + 105, + 390, + 301, + 404 + ], + "score": 1.0, + "content": "tive process. See Appendix F for details. We use", + "type": "text" + }, + { + "bbox": [ + 302, + 391, + 312, + 402 + ], + "score": 0.88, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 312, + 390, + 338, + 404 + ], + "score": 1.0, + "content": "to get", + "type": "text" + }, + { + "bbox": [ + 338, + 391, + 358, + 402 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 358, + 390, + 504, + 404 + ], + "score": 1.0, + "content": "for all experiments except Figure 6.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 12.5 + }, + { + "type": "title", + "bbox": [ + 108, + 418, + 282, + 430 + ], + "lines": [ + { + "bbox": [ + 104, + 416, + 285, + 434 + ], + "spans": [ + { + "bbox": [ + 104, + 416, + 285, + 434 + ], + "score": 1.0, + "content": "4 GENERATIVE PROCESS DESIGN", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 107, + 442, + 505, + 476 + ], + "lines": [ + { + "bbox": [ + 105, + 441, + 505, + 457 + ], + "spans": [ + { + "bbox": [ + 105, + 441, + 505, + 457 + ], + "score": 1.0, + "content": "This section describes the entire editing process, which consists of three phases: editing with Asyrp,", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 454, + 505, + 466 + ], + "spans": [ + { + "bbox": [ + 105, + 454, + 505, + 466 + ], + "score": 1.0, + "content": "traditional denoising, and quality boosting. We design formulas to determine the length of each", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 465, + 243, + 477 + ], + "spans": [ + { + "bbox": [ + 106, + 465, + 243, + 477 + ], + "score": 1.0, + "content": "phase with quantifiable measures.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18 + }, + { + "type": "title", + "bbox": [ + 108, + 489, + 266, + 501 + ], + "lines": [ + { + "bbox": [ + 106, + 489, + 267, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 267, + 502 + ], + "score": 1.0, + "content": "4.1 EDITING PROCESS WITH ASYRP", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 510, + 505, + 543 + ], + "lines": [ + { + "bbox": [ + 106, + 510, + 505, + 522 + ], + "spans": [ + { + "bbox": [ + 106, + 510, + 505, + 522 + ], + "score": 1.0, + "content": "Diffusion models generate the high-level context in the early stage and imperceptible fine details in", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 520, + 506, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 520, + 506, + 534 + ], + "score": 1.0, + "content": "the later stage (Choi et al., 2022). Likewise, we modify the generative process in the early stage to", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 531, + 448, + 546 + ], + "spans": [ + { + "bbox": [ + 105, + 531, + 410, + 546 + ], + "score": 1.0, + "content": "achieve semantic changes. We refer to the early stage as the editing interval", + "type": "text" + }, + { + "bbox": [ + 410, + 532, + 444, + 544 + ], + "score": 0.92, + "content": "[ T , t _ { \\mathrm { e d i t } } ]", + "type": "inline_equation" + }, + { + "bbox": [ + 445, + 531, + 448, + 546 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 106, + 549, + 505, + 604 + ], + "lines": [ + { + "bbox": [ + 106, + 549, + 505, + 561 + ], + "spans": [ + { + "bbox": [ + 106, + 549, + 169, + 561 + ], + "score": 0.77, + "content": "\\mathrm { L P I P S } ( { \\pmb x } , { \\pmb P } _ { T } )", + "type": "inline_equation" + }, + { + "bbox": [ + 169, + 549, + 187, + 561 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 187, + 549, + 247, + 561 + ], + "score": 0.87, + "content": "\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 247, + 549, + 505, + 561 + ], + "score": 1.0, + "content": "calculate the perceptual distance between the original image and", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 560, + 504, + 572 + ], + "spans": [ + { + "bbox": [ + 106, + 560, + 240, + 572 + ], + "score": 1.0, + "content": "the predicted image at time steps", + "type": "text" + }, + { + "bbox": [ + 241, + 561, + 249, + 570 + ], + "score": 0.82, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 250, + 560, + 267, + 572 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 268, + 561, + 273, + 570 + ], + "score": 0.74, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 273, + 560, + 504, + 572 + ], + "score": 1.0, + "content": ", respectively. Intuitively, the high-level content is already", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 570, + 505, + 585 + ], + "spans": [ + { + "bbox": [ + 105, + 570, + 505, + 585 + ], + "score": 1.0, + "content": "determined by the predicted terms at the respective timesteps and LPIPS measures the remaining", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 104, + 580, + 505, + 596 + ], + "spans": [ + { + "bbox": [ + 104, + 580, + 505, + 596 + ], + "score": 1.0, + "content": "component to be edited through the remaining reverse process. Consequently, we define editing", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 592, + 222, + 606 + ], + "spans": [ + { + "bbox": [ + 105, + 592, + 197, + 606 + ], + "score": 1.0, + "content": "strength of an interval", + "type": "text" + }, + { + "bbox": [ + 197, + 593, + 218, + 605 + ], + "score": 0.91, + "content": "[ T , t ]", + "type": "inline_equation" + }, + { + "bbox": [ + 218, + 592, + 222, + 606 + ], + "score": 1.0, + "content": ":", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 26 + }, + { + "type": "interline_equation", + "bbox": [ + 227, + 609, + 384, + 622 + ], + "lines": [ + { + "bbox": [ + 227, + 609, + 384, + 622 + ], + "spans": [ + { + "bbox": [ + 227, + 609, + 384, + 622 + ], + "score": 0.9, + "content": "\\xi _ { t } = \\mathrm { L P I P S } ( x , \\mathbf { P } _ { T } ) - \\mathrm { L P I P S } ( x , \\mathbf { P } _ { t } )", + "type": "interline_equation", + "image_path": "1e13f8e22779ba133801ee9ef0228320ea6bf34efecc3a3ff44ec39caedf9159.jpg" + } + ] + } + ], + "index": 29, + "virtual_lines": [ + { + "bbox": [ + 227, + 609, + 384, + 622 + ], + "spans": [], + "index": 29 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 626, + 505, + 704 + ], + "lines": [ + { + "bbox": [ + 105, + 626, + 505, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 300, + 640 + ], + "score": 1.0, + "content": "indicating the perceptual change from timestep", + "type": "text" + }, + { + "bbox": [ + 300, + 627, + 309, + 637 + ], + "score": 0.82, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 626, + 321, + 640 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 321, + 628, + 327, + 637 + ], + "score": 0.77, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 327, + 626, + 505, + 640 + ], + "score": 1.0, + "content": "in the original generative process. Figure 3", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 637, + 506, + 651 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 149, + 651 + ], + "score": 1.0, + "content": "illustrates", + "type": "text" + }, + { + "bbox": [ + 149, + 638, + 200, + 650 + ], + "score": 0.76, + "content": "\\mathrm { L P I P S } ( { \\pmb x } , \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 201, + 637, + 218, + 651 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 218, + 638, + 231, + 649 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 232, + 637, + 252, + 651 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 252, + 639, + 263, + 649 + ], + "score": 0.87, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 263, + 637, + 506, + 651 + ], + "score": 1.0, + "content": "with examples and the inset depicts editing strength. The", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 649, + 505, + 661 + ], + "spans": [ + { + "bbox": [ + 106, + 649, + 262, + 661 + ], + "score": 1.0, + "content": "shorter editing interval has the lower", + "type": "text" + }, + { + "bbox": [ + 263, + 650, + 272, + 660 + ], + "score": 0.88, + "content": "\\xi _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 649, + 505, + 661 + ], + "score": 1.0, + "content": ", and the longer editing interval brings more changes to", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "score": 1.0, + "content": "the resulting images. We seek the shortest editing interval which will bring enough distinguishable", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 104, + 669, + 506, + 686 + ], + "spans": [ + { + "bbox": [ + 104, + 669, + 335, + 686 + ], + "score": 1.0, + "content": "changes in the images in general. We empirically find that", + "type": "text" + }, + { + "bbox": [ + 335, + 671, + 354, + 682 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 354, + 669, + 375, + 686 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 375, + 671, + 478, + 683 + ], + "score": 0.8, + "content": "\\mathrm { L P I P S } ( \\mathbf { x } , \\mathbf { P } _ { t _ { \\mathrm { e d i t } } } ) = 0 . \\dot { 3 } 3", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 669, + 506, + 686 + ], + "score": 1.0, + "content": "builds", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 680, + 505, + 696 + ], + "spans": [ + { + "bbox": [ + 105, + 680, + 346, + 696 + ], + "score": 1.0, + "content": "the shortest editing interval with enough editing strength as", + "type": "text" + }, + { + "bbox": [ + 347, + 682, + 370, + 694 + ], + "score": 0.92, + "content": "\\mathbf { P } _ { t _ { \\mathrm { e d i t } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 371, + 680, + 505, + 696 + ], + "score": 1.0, + "content": "has nearly all visual attributes in", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 107, + 694, + 119, + 704 + ], + "spans": [ + { + "bbox": [ + 107, + 695, + 114, + 703 + ], + "score": 0.68, + "content": "_ { \\textbf { \\em x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 115, + 694, + 119, + 704 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 33 + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 708, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 439, + 722 + ], + "score": 1.0, + "content": "However, some attributes require more visual changes than others, e.g., pixar", + "type": "text" + }, + { + "bbox": [ + 440, + 711, + 451, + 720 + ], + "score": 0.65, + "content": ">", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 708, + 506, + 722 + ], + "score": 1.0, + "content": "smile. For", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 300, + 734 + ], + "score": 1.0, + "content": "such attributes, we increase the editing strength", + "type": "text" + }, + { + "bbox": [ + 300, + 721, + 310, + 732 + ], + "score": 0.88, + "content": "\\xi _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 310, + 720, + 325, + 734 + ], + "score": 1.0, + "content": "by", + "type": "text" + }, + { + "bbox": [ + 325, + 721, + 475, + 733 + ], + "score": 0.91, + "content": "\\delta = 0 . 3 3 d ( E _ { T } ( y _ { \\mathrm { s o u r c e } } ) , E _ { T } ( y _ { \\mathrm { t a r g e t } } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 476, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "where", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 37.5 + } + ], + "page_idx": 4, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "score": 1.0, + "content": "5", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 116, + 85, + 491, + 213 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 116, + 85, + 491, + 213 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 116, + 85, + 491, + 213 + ], + "spans": [ + { + "bbox": [ + 116, + 85, + 491, + 213 + ], + "score": 0.973, + "type": "image", + "image_path": "47dbd6b534787f72c5d6825a64bc954aec392ef479c4aed91b47b808b2fa38d3.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 116, + 85, + 491, + 127.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 116, + 127.66666666666666, + 491, + 170.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 116, + 170.33333333333331, + 491, + 212.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 228, + 505, + 306 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 228, + 505, + 241 + ], + "spans": [ + { + "bbox": [ + 106, + 228, + 505, + 241 + ], + "score": 1.0, + "content": "Figure 3: Intuition for choosing the intervals for editing and quality boosting. We choose the", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 239, + 505, + 253 + ], + "spans": [ + { + "bbox": [ + 105, + 239, + 460, + 253 + ], + "score": 1.0, + "content": "intervals by quantifying two measures (top left inset). The editing strength of an interval", + "type": "text" + }, + { + "bbox": [ + 460, + 240, + 481, + 252 + ], + "score": 0.91, + "content": "[ T , t ]", + "type": "inline_equation" + }, + { + "bbox": [ + 482, + 239, + 505, + 253 + ], + "score": 1.0, + "content": "mea-", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 104, + 250, + 506, + 265 + ], + "spans": [ + { + "bbox": [ + 104, + 250, + 251, + 265 + ], + "score": 1.0, + "content": "sures its perceptual difference from", + "type": "text" + }, + { + "bbox": [ + 252, + 251, + 261, + 261 + ], + "score": 0.8, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 250, + 282, + 265 + ], + "score": 1.0, + "content": "until", + "type": "text" + }, + { + "bbox": [ + 282, + 252, + 288, + 261 + ], + "score": 0.59, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 289, + 250, + 322, + 265 + ], + "score": 1.0, + "content": ". We set", + "type": "text" + }, + { + "bbox": [ + 322, + 251, + 343, + 263 + ], + "score": 0.89, + "content": "[ T , t ]", + "type": "inline_equation" + }, + { + "bbox": [ + 344, + 250, + 506, + 265 + ], + "score": 1.0, + "content": "to the interval with the smallest editing", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 262, + 504, + 274 + ], + "spans": [ + { + "bbox": [ + 106, + 262, + 206, + 273 + ], + "score": 1.0, + "content": "strength that synthesizes", + "type": "text" + }, + { + "bbox": [ + 206, + 262, + 219, + 273 + ], + "score": 0.89, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 219, + 262, + 253, + 273 + ], + "score": 1.0, + "content": "close to", + "type": "text" + }, + { + "bbox": [ + 253, + 264, + 261, + 272 + ], + "score": 0.76, + "content": "_ { \\textbf { \\em x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 262, + 282, + 273 + ], + "score": 1.0, + "content": ", i.e.,", + "type": "text" + }, + { + "bbox": [ + 282, + 262, + 373, + 274 + ], + "score": 0.73, + "content": "\\mathrm { L P I P S } ( { \\pmb x } , { \\bf P } _ { t } ) = 0 . 3 3", + "type": "inline_equation" + }, + { + "bbox": [ + 373, + 262, + 504, + 273 + ], + "score": 1.0, + "content": ". Editing flexibility of an interval", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 107, + 272, + 506, + 286 + ], + "spans": [ + { + "bbox": [ + 107, + 273, + 126, + 285 + ], + "score": 0.88, + "content": "[ t , 0 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 127, + 272, + 318, + 286 + ], + "score": 1.0, + "content": "measures the potential amount of changes after", + "type": "text" + }, + { + "bbox": [ + 318, + 275, + 323, + 283 + ], + "score": 0.71, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 323, + 272, + 412, + 286 + ], + "score": 1.0, + "content": ". Quality deficiency at", + "type": "text" + }, + { + "bbox": [ + 412, + 274, + 417, + 283 + ], + "score": 0.75, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 417, + 272, + 506, + 286 + ], + "score": 1.0, + "content": "measures the amount", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 283, + 506, + 297 + ], + "spans": [ + { + "bbox": [ + 105, + 283, + 150, + 297 + ], + "score": 1.0, + "content": "of noise in", + "type": "text" + }, + { + "bbox": [ + 151, + 285, + 162, + 295 + ], + "score": 0.86, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 162, + 283, + 195, + 297 + ], + "score": 1.0, + "content": ". We set", + "type": "text" + }, + { + "bbox": [ + 195, + 284, + 214, + 296 + ], + "score": 0.88, + "content": "[ t , 0 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 215, + 283, + 371, + 297 + ], + "score": 1.0, + "content": "to handle large quality deficiency (i.e.,", + "type": "text" + }, + { + "bbox": [ + 371, + 284, + 458, + 296 + ], + "score": 0.84, + "content": "\\mathrm { L P I P S } ( \\pmb { x } , \\pmb { x } _ { t } ) = 1 . 2 )", + "type": "inline_equation" + }, + { + "bbox": [ + 459, + 283, + 506, + 297 + ], + "score": 1.0, + "content": "with small", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 294, + 180, + 309 + ], + "spans": [ + { + "bbox": [ + 105, + 294, + 180, + 309 + ], + "score": 1.0, + "content": "editing flexibility.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 6 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 106, + 330, + 505, + 403 + ], + "lines": [], + "index": 12.5, + "bbox_fs": [ + 105, + 331, + 506, + 404 + ], + "lines_deleted": true + }, + { + "type": "title", + "bbox": [ + 108, + 418, + 282, + 430 + ], + "lines": [ + { + "bbox": [ + 104, + 416, + 285, + 434 + ], + "spans": [ + { + "bbox": [ + 104, + 416, + 285, + 434 + ], + "score": 1.0, + "content": "4 GENERATIVE PROCESS DESIGN", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 107, + 442, + 505, + 476 + ], + "lines": [ + { + "bbox": [ + 105, + 441, + 505, + 457 + ], + "spans": [ + { + "bbox": [ + 105, + 441, + 505, + 457 + ], + "score": 1.0, + "content": "This section describes the entire editing process, which consists of three phases: editing with Asyrp,", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 454, + 505, + 466 + ], + "spans": [ + { + "bbox": [ + 105, + 454, + 505, + 466 + ], + "score": 1.0, + "content": "traditional denoising, and quality boosting. We design formulas to determine the length of each", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 465, + 243, + 477 + ], + "spans": [ + { + "bbox": [ + 106, + 465, + 243, + 477 + ], + "score": 1.0, + "content": "phase with quantifiable measures.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18, + "bbox_fs": [ + 105, + 441, + 505, + 477 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 489, + 266, + 501 + ], + "lines": [ + { + "bbox": [ + 106, + 489, + 267, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 267, + 502 + ], + "score": 1.0, + "content": "4.1 EDITING PROCESS WITH ASYRP", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 510, + 505, + 543 + ], + "lines": [ + { + "bbox": [ + 106, + 510, + 505, + 522 + ], + "spans": [ + { + "bbox": [ + 106, + 510, + 505, + 522 + ], + "score": 1.0, + "content": "Diffusion models generate the high-level context in the early stage and imperceptible fine details in", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 520, + 506, + 534 + ], + "spans": [ + { + "bbox": [ + 105, + 520, + 506, + 534 + ], + "score": 1.0, + "content": "the later stage (Choi et al., 2022). Likewise, we modify the generative process in the early stage to", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 531, + 448, + 546 + ], + "spans": [ + { + "bbox": [ + 105, + 531, + 410, + 546 + ], + "score": 1.0, + "content": "achieve semantic changes. We refer to the early stage as the editing interval", + "type": "text" + }, + { + "bbox": [ + 410, + 532, + 444, + 544 + ], + "score": 0.92, + "content": "[ T , t _ { \\mathrm { e d i t } } ]", + "type": "inline_equation" + }, + { + "bbox": [ + 445, + 531, + 448, + 546 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22, + "bbox_fs": [ + 105, + 510, + 506, + 546 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 549, + 505, + 604 + ], + "lines": [ + { + "bbox": [ + 106, + 549, + 505, + 561 + ], + "spans": [ + { + "bbox": [ + 106, + 549, + 169, + 561 + ], + "score": 0.77, + "content": "\\mathrm { L P I P S } ( { \\pmb x } , { \\pmb P } _ { T } )", + "type": "inline_equation" + }, + { + "bbox": [ + 169, + 549, + 187, + 561 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 187, + 549, + 247, + 561 + ], + "score": 0.87, + "content": "\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 247, + 549, + 505, + 561 + ], + "score": 1.0, + "content": "calculate the perceptual distance between the original image and", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 560, + 504, + 572 + ], + "spans": [ + { + "bbox": [ + 106, + 560, + 240, + 572 + ], + "score": 1.0, + "content": "the predicted image at time steps", + "type": "text" + }, + { + "bbox": [ + 241, + 561, + 249, + 570 + ], + "score": 0.82, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 250, + 560, + 267, + 572 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 268, + 561, + 273, + 570 + ], + "score": 0.74, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 273, + 560, + 504, + 572 + ], + "score": 1.0, + "content": ", respectively. Intuitively, the high-level content is already", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 570, + 505, + 585 + ], + "spans": [ + { + "bbox": [ + 105, + 570, + 505, + 585 + ], + "score": 1.0, + "content": "determined by the predicted terms at the respective timesteps and LPIPS measures the remaining", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 104, + 580, + 505, + 596 + ], + "spans": [ + { + "bbox": [ + 104, + 580, + 505, + 596 + ], + "score": 1.0, + "content": "component to be edited through the remaining reverse process. Consequently, we define editing", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 592, + 222, + 606 + ], + "spans": [ + { + "bbox": [ + 105, + 592, + 197, + 606 + ], + "score": 1.0, + "content": "strength of an interval", + "type": "text" + }, + { + "bbox": [ + 197, + 593, + 218, + 605 + ], + "score": 0.91, + "content": "[ T , t ]", + "type": "inline_equation" + }, + { + "bbox": [ + 218, + 592, + 222, + 606 + ], + "score": 1.0, + "content": ":", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 26, + "bbox_fs": [ + 104, + 549, + 505, + 606 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 227, + 609, + 384, + 622 + ], + "lines": [ + { + "bbox": [ + 227, + 609, + 384, + 622 + ], + "spans": [ + { + "bbox": [ + 227, + 609, + 384, + 622 + ], + "score": 0.9, + "content": "\\xi _ { t } = \\mathrm { L P I P S } ( x , \\mathbf { P } _ { T } ) - \\mathrm { L P I P S } ( x , \\mathbf { P } _ { t } )", + "type": "interline_equation", + "image_path": "1e13f8e22779ba133801ee9ef0228320ea6bf34efecc3a3ff44ec39caedf9159.jpg" + } + ] + } + ], + "index": 29, + "virtual_lines": [ + { + "bbox": [ + 227, + 609, + 384, + 622 + ], + "spans": [], + "index": 29 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 626, + 505, + 704 + ], + "lines": [ + { + "bbox": [ + 105, + 626, + 505, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 300, + 640 + ], + "score": 1.0, + "content": "indicating the perceptual change from timestep", + "type": "text" + }, + { + "bbox": [ + 300, + 627, + 309, + 637 + ], + "score": 0.82, + "content": "T", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 626, + 321, + 640 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 321, + 628, + 327, + 637 + ], + "score": 0.77, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 327, + 626, + 505, + 640 + ], + "score": 1.0, + "content": "in the original generative process. Figure 3", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 637, + 506, + 651 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 149, + 651 + ], + "score": 1.0, + "content": "illustrates", + "type": "text" + }, + { + "bbox": [ + 149, + 638, + 200, + 650 + ], + "score": 0.76, + "content": "\\mathrm { L P I P S } ( { \\pmb x } , \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 201, + 637, + 218, + 651 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 218, + 638, + 231, + 649 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 232, + 637, + 252, + 651 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 252, + 639, + 263, + 649 + ], + "score": 0.87, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 263, + 637, + 506, + 651 + ], + "score": 1.0, + "content": "with examples and the inset depicts editing strength. The", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 649, + 505, + 661 + ], + "spans": [ + { + "bbox": [ + 106, + 649, + 262, + 661 + ], + "score": 1.0, + "content": "shorter editing interval has the lower", + "type": "text" + }, + { + "bbox": [ + 263, + 650, + 272, + 660 + ], + "score": 0.88, + "content": "\\xi _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 649, + 505, + 661 + ], + "score": 1.0, + "content": ", and the longer editing interval brings more changes to", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "score": 1.0, + "content": "the resulting images. We seek the shortest editing interval which will bring enough distinguishable", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 104, + 669, + 506, + 686 + ], + "spans": [ + { + "bbox": [ + 104, + 669, + 335, + 686 + ], + "score": 1.0, + "content": "changes in the images in general. We empirically find that", + "type": "text" + }, + { + "bbox": [ + 335, + 671, + 354, + 682 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 354, + 669, + 375, + 686 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 375, + 671, + 478, + 683 + ], + "score": 0.8, + "content": "\\mathrm { L P I P S } ( \\mathbf { x } , \\mathbf { P } _ { t _ { \\mathrm { e d i t } } } ) = 0 . \\dot { 3 } 3", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 669, + 506, + 686 + ], + "score": 1.0, + "content": "builds", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 680, + 505, + 696 + ], + "spans": [ + { + "bbox": [ + 105, + 680, + 346, + 696 + ], + "score": 1.0, + "content": "the shortest editing interval with enough editing strength as", + "type": "text" + }, + { + "bbox": [ + 347, + 682, + 370, + 694 + ], + "score": 0.92, + "content": "\\mathbf { P } _ { t _ { \\mathrm { e d i t } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 371, + 680, + 505, + 696 + ], + "score": 1.0, + "content": "has nearly all visual attributes in", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 107, + 694, + 119, + 704 + ], + "spans": [ + { + "bbox": [ + 107, + 695, + 114, + 703 + ], + "score": 0.68, + "content": "_ { \\textbf { \\em x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 115, + 694, + 119, + 704 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 33, + "bbox_fs": [ + 104, + 626, + 506, + 704 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 708, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 439, + 722 + ], + "score": 1.0, + "content": "However, some attributes require more visual changes than others, e.g., pixar", + "type": "text" + }, + { + "bbox": [ + 440, + 711, + 451, + 720 + ], + "score": 0.65, + "content": ">", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 708, + 506, + 722 + ], + "score": 1.0, + "content": "smile. For", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 300, + 734 + ], + "score": 1.0, + "content": "such attributes, we increase the editing strength", + "type": "text" + }, + { + "bbox": [ + 300, + 721, + 310, + 732 + ], + "score": 0.88, + "content": "\\xi _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 310, + 720, + 325, + 734 + ], + "score": 1.0, + "content": "by", + "type": "text" + }, + { + "bbox": [ + 325, + 721, + 475, + 733 + ], + "score": 0.91, + "content": "\\delta = 0 . 3 3 d ( E _ { T } ( y _ { \\mathrm { s o u r c e } } ) , E _ { T } ( y _ { \\mathrm { t a r g e t } } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 476, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "where", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 37.5, + "bbox_fs": [ + 105, + 708, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 128 + ], + "lines": [ + { + "bbox": [ + 107, + 81, + 505, + 96 + ], + "spans": [ + { + "bbox": [ + 107, + 82, + 132, + 95 + ], + "score": 0.91, + "content": "E _ { T } ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 132, + 81, + 262, + 96 + ], + "score": 1.0, + "content": "produces CLIP text embedding,", + "type": "text" + }, + { + "bbox": [ + 262, + 84, + 277, + 95 + ], + "score": 0.84, + "content": "y _ { ( \\cdot ) }", + "type": "inline_equation" + }, + { + "bbox": [ + 277, + 81, + 396, + 96 + ], + "score": 1.0, + "content": "denotes the descriptions, and", + "type": "text" + }, + { + "bbox": [ + 397, + 82, + 420, + 94 + ], + "score": 0.91, + "content": "d ( \\cdot , \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 421, + 81, + 505, + 96 + ], + "score": 1.0, + "content": "computes the cosine", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 104, + 92, + 506, + 109 + ], + "spans": [ + { + "bbox": [ + 104, + 92, + 284, + 109 + ], + "score": 1.0, + "content": "distance between the arguments. Choosing", + "type": "text" + }, + { + "bbox": [ + 284, + 95, + 303, + 106 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 303, + 92, + 326, + 109 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 326, + 94, + 452, + 107 + ], + "score": 0.89, + "content": "\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t _ { \\mathrm { e d i t } } } ) = 0 . 3 3 - \\delta", + "type": "inline_equation" + }, + { + "bbox": [ + 453, + 92, + 506, + 109 + ], + "score": 1.0, + "content": "expands the", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 105, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 105, + 105, + 505, + 118 + ], + "score": 1.0, + "content": "editing interval to a suitable length. It consistently produces good results in various settings. The", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 117, + 309, + 128 + ], + "spans": [ + { + "bbox": [ + 105, + 117, + 309, + 128 + ], + "score": 1.0, + "content": "supporting experiments are shown in Appendix G.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 1.5 + }, + { + "type": "title", + "bbox": [ + 106, + 141, + 376, + 152 + ], + "lines": [ + { + "bbox": [ + 106, + 140, + 378, + 153 + ], + "spans": [ + { + "bbox": [ + 106, + 140, + 378, + 153 + ], + "score": 1.0, + "content": "4.2 QUALITY BOOSTING WITH STOCHASTIC NOISE INJECTION", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 108, + 161, + 504, + 196 + ], + "lines": [ + { + "bbox": [ + 105, + 160, + 505, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 160, + 418, + 174 + ], + "score": 1.0, + "content": "Although DDIM achieves nearly perfect inversion by removing stochasticity", + "type": "text" + }, + { + "bbox": [ + 419, + 162, + 447, + 173 + ], + "score": 0.87, + "content": "( \\eta = 0", + "type": "inline_equation" + }, + { + "bbox": [ + 447, + 160, + 505, + 174 + ], + "score": 1.0, + "content": "), Karras et al.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 172, + 505, + 185 + ], + "spans": [ + { + "bbox": [ + 106, + 172, + 505, + 185 + ], + "score": 1.0, + "content": "(2022) demonstrate that stochasticity improves image quality. Likewise, we inject stochastic noise", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 183, + 244, + 197 + ], + "spans": [ + { + "bbox": [ + 105, + 183, + 201, + 197 + ], + "score": 1.0, + "content": "in the boosting interval", + "type": "text" + }, + { + "bbox": [ + 201, + 183, + 240, + 195 + ], + "score": 0.93, + "content": "[ t _ { \\mathrm { b o o s t } } , 0 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 240, + 183, + 244, + 197 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 107, + 200, + 505, + 332 + ], + "lines": [ + { + "bbox": [ + 105, + 199, + 505, + 214 + ], + "spans": [ + { + "bbox": [ + 105, + 199, + 505, + 214 + ], + "score": 1.0, + "content": "Though the longer boosting interval would achieve higher quality, boosting over excessively long", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 211, + 506, + 223 + ], + "spans": [ + { + "bbox": [ + 105, + 211, + 506, + 223 + ], + "score": 1.0, + "content": "intervals would modify the content. Hence, we want to determine the shortest interval that shows", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 222, + 505, + 234 + ], + "spans": [ + { + "bbox": [ + 105, + 222, + 505, + 234 + ], + "score": 1.0, + "content": "enough quality boosting to guarantee minimal change in the content. We consider the noise in the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 232, + 504, + 246 + ], + "spans": [ + { + "bbox": [ + 105, + 232, + 414, + 246 + ], + "score": 1.0, + "content": "image as the capacity for the quality boosting and define quality deficiency at", + "type": "text" + }, + { + "bbox": [ + 414, + 234, + 419, + 243 + ], + "score": 0.36, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 233, + 504, + 244 + ], + "score": 0.89, + "content": ": \\gamma _ { t } = \\mathrm { L P I P S } ( \\pmb { x } , \\pmb { x } _ { t } )", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 243, + 506, + 256 + ], + "spans": [ + { + "bbox": [ + 105, + 243, + 236, + 256 + ], + "score": 1.0, + "content": "indicating the amount of noise in", + "type": "text" + }, + { + "bbox": [ + 237, + 245, + 248, + 255 + ], + "score": 0.86, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 243, + 405, + 256 + ], + "score": 1.0, + "content": "compared to the original image. We use", + "type": "text" + }, + { + "bbox": [ + 406, + 245, + 417, + 255 + ], + "score": 0.87, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 417, + 243, + 457, + 256 + ], + "score": 1.0, + "content": "instead of", + "type": "text" + }, + { + "bbox": [ + 458, + 245, + 470, + 255 + ], + "score": 0.87, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 471, + 243, + 506, + 256 + ], + "score": 1.0, + "content": "because", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 255, + 505, + 267 + ], + "spans": [ + { + "bbox": [ + 106, + 255, + 505, + 267 + ], + "score": 1.0, + "content": "we consider the actual image rather than the semantics. Figure 3 inset depicts editing flexibility and", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 264, + 505, + 281 + ], + "spans": [ + { + "bbox": [ + 105, + 264, + 289, + 281 + ], + "score": 1.0, + "content": "quality deficiency. We empirically find that", + "type": "text" + }, + { + "bbox": [ + 289, + 267, + 314, + 277 + ], + "score": 0.88, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 264, + 338, + 281 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 338, + 266, + 397, + 278 + ], + "score": 0.91, + "content": "\\gamma _ { t _ { \\mathrm { b o o s t } } } ~ = ~ 1 . 2", + "type": "inline_equation" + }, + { + "bbox": [ + 397, + 264, + 505, + 281 + ], + "score": 1.0, + "content": "achieves quality boosting", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 275, + 506, + 290 + ], + "spans": [ + { + "bbox": [ + 105, + 275, + 455, + 290 + ], + "score": 1.0, + "content": "with minimal content change. We confirmed that the editing strength of the intervals", + "type": "text" + }, + { + "bbox": [ + 455, + 277, + 494, + 289 + ], + "score": 0.93, + "content": "[ t _ { \\mathrm { b o o s t } } , 0 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 494, + 275, + 506, + 290 + ], + "score": 1.0, + "content": "is", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 104, + 287, + 507, + 301 + ], + "spans": [ + { + "bbox": [ + 104, + 287, + 302, + 301 + ], + "score": 1.0, + "content": "guaranteed to be less than 0.25. In Figure 3, after", + "type": "text" + }, + { + "bbox": [ + 303, + 288, + 327, + 299 + ], + "score": 0.89, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 327, + 287, + 330, + 301 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 331, + 288, + 390, + 299 + ], + "score": 0.67, + "content": "\\mathrm { L P I P S } ( { \\pmb x } , { \\pmb x } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 390, + 287, + 507, + 301 + ], + "score": 1.0, + "content": "sharply drops in the original", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 104, + 298, + 507, + 312 + ], + "spans": [ + { + "bbox": [ + 104, + 298, + 207, + 312 + ], + "score": 1.0, + "content": "generative process while", + "type": "text" + }, + { + "bbox": [ + 207, + 299, + 268, + 311 + ], + "score": 0.71, + "content": "\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 268, + 298, + 507, + 312 + ], + "score": 1.0, + "content": "changes little. Note that most of the quality degradation of", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 308, + 505, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 308, + 505, + 324 + ], + "score": 1.0, + "content": "the resulting images is caused by DDIM reverse process, not by Asyrp. We use this quality boosting", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 321, + 311, + 333 + ], + "spans": [ + { + "bbox": [ + 105, + 321, + 299, + 333 + ], + "score": 1.0, + "content": "for all experiments except ablation in Appendix", + "type": "text" + }, + { + "bbox": [ + 299, + 321, + 307, + 330 + ], + "score": 0.38, + "content": "_ \\mathrm { H }", + "type": "inline_equation" + }, + { + "bbox": [ + 308, + 321, + 311, + 333 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 13.5 + }, + { + "type": "title", + "bbox": [ + 108, + 345, + 296, + 356 + ], + "lines": [ + { + "bbox": [ + 105, + 344, + 298, + 357 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 298, + 357 + ], + "score": 1.0, + "content": "4.3 OVERALL PROCESS OF IMAGE EDITING", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 365, + 506, + 387 + ], + "lines": [ + { + "bbox": [ + 105, + 364, + 506, + 378 + ], + "spans": [ + { + "bbox": [ + 105, + 364, + 132, + 378 + ], + "score": 1.0, + "content": "Using", + "type": "text" + }, + { + "bbox": [ + 132, + 367, + 151, + 377 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 151, + 364, + 168, + 378 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 168, + 366, + 192, + 377 + ], + "score": 0.88, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 192, + 364, + 506, + 378 + ], + "score": 1.0, + "content": "determined by the above formulas, we modify the generative process of DDIM", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 375, + 128, + 389 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 128, + 389 + ], + "score": 1.0, + "content": "with", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5 + }, + { + "type": "interline_equation", + "bbox": [ + 116, + 387, + 482, + 430 + ], + "lines": [ + { + "bbox": [ + 116, + 387, + 482, + 430 + ], + "spans": [ + { + "bbox": [ + 116, + 387, + 482, + 430 + ], + "score": 0.93, + "content": "p _ { \\theta } ^ { ( t ) } \\left( x _ { t - 1 } \\mid x _ { t } \\right) = \\left\\{ \\begin{array} { l l } { \\mathcal { N } \\left( \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } | f _ { t } ) ) + \\mathbf { D } _ { t } , \\sigma _ { t } ^ { 2 } I \\right) , \\eta = 0 } & { \\mathrm { ~ i f ~ } T \\geq t \\geq t _ { \\mathrm { e d i t } } } \\\\ { \\mathcal { N } \\left( \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) + \\mathbf { D } _ { t } , \\sigma _ { t } ^ { 2 } I \\right) , \\eta = 0 } & { \\mathrm { ~ i f ~ } t _ { \\mathrm { e d i t } } > t \\geq t _ { \\mathrm { b o o s t } } } \\\\ { \\mathcal { N } \\left( \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) + \\mathbf { D } _ { t } , \\sigma _ { t } ^ { 2 } I \\right) , \\eta = 1 } & { \\mathrm { ~ i f ~ } t _ { \\mathrm { b o o s t } } > t } \\end{array} \\right.", + "type": "interline_equation", + "image_path": "1fd23e676ae9069b2a174b8dbe503c18b765107d5f7690ab6bab96976aa085a0.jpg" + } + ] + } + ], + "index": 24, + "virtual_lines": [ + { + "bbox": [ + 116, + 387, + 482, + 401.3333333333333 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 116, + 401.3333333333333, + 482, + 415.66666666666663 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 116, + 415.66666666666663, + 482, + 429.99999999999994 + ], + "spans": [], + "index": 25 + } + ] + }, + { + "type": "text", + "bbox": [ + 105, + 436, + 471, + 447 + ], + "lines": [ + { + "bbox": [ + 105, + 435, + 473, + 449 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 473, + 449 + ], + "score": 1.0, + "content": "The visual overview and comprehensive algorithms of the entire process are in Appendix I.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 26 + }, + { + "type": "title", + "bbox": [ + 107, + 462, + 200, + 475 + ], + "lines": [ + { + "bbox": [ + 105, + 462, + 201, + 477 + ], + "spans": [ + { + "bbox": [ + 105, + 462, + 201, + 477 + ], + "score": 1.0, + "content": "5 EXPERIMENTS", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 106, + 487, + 505, + 531 + ], + "lines": [ + { + "bbox": [ + 105, + 487, + 505, + 499 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 381, + 499 + ], + "score": 1.0, + "content": "In this section, we show the effectiveness of semantic latent editing in", + "type": "text" + }, + { + "bbox": [ + 382, + 488, + 388, + 497 + ], + "score": 0.61, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 389, + 487, + 505, + 499 + ], + "score": 1.0, + "content": "-space with Asyrp on various", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 497, + 506, + 511 + ], + "spans": [ + { + "bbox": [ + 105, + 497, + 267, + 511 + ], + "score": 1.0, + "content": "attributes, datasets and architectures in", + "type": "text" + }, + { + "bbox": [ + 267, + 498, + 288, + 510 + ], + "score": 0.54, + "content": "\\ S \\ S . 1", + "type": "inline_equation" + }, + { + "bbox": [ + 288, + 497, + 506, + 511 + ], + "score": 1.0, + "content": ". Moreover, we provide quantitative results including", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 509, + 505, + 521 + ], + "spans": [ + { + "bbox": [ + 106, + 510, + 162, + 521 + ], + "score": 1.0, + "content": "user study in", + "type": "text" + }, + { + "bbox": [ + 163, + 509, + 185, + 521 + ], + "score": 0.63, + "content": "\\mathrm { ~ \\normalfont ~ \\ S ~ } 5 . 2", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 510, + 505, + 521 + ], + "score": 1.0, + "content": ". Lastly, we provide detailed analyses for the properties of the semantic latent", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 520, + 276, + 532 + ], + "spans": [ + { + "bbox": [ + 105, + 521, + 143, + 532 + ], + "score": 1.0, + "content": "space on", + "type": "text" + }, + { + "bbox": [ + 144, + 521, + 150, + 530 + ], + "score": 0.64, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 150, + 521, + 252, + 532 + ], + "score": 1.0, + "content": "-space and alternatives in", + "type": "text" + }, + { + "bbox": [ + 252, + 520, + 273, + 532 + ], + "score": 0.85, + "content": "\\ S 5 . 3", + "type": "inline_equation" + }, + { + "bbox": [ + 273, + 521, + 276, + 532 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 29.5 + }, + { + "type": "text", + "bbox": [ + 106, + 542, + 505, + 676 + ], + "lines": [ + { + "bbox": [ + 105, + 542, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 505, + 555 + ], + "score": 1.0, + "content": "Implementation details. We implement our method on various settings: CelebA-HQ (Karras", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 104, + 552, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 104, + 552, + 352, + 567 + ], + "score": 1.0, + "content": "et al., 2018) and LSUN-bedroom/-church (Yu et al., 2015) on", + "type": "text" + }, + { + "bbox": [ + 353, + 554, + 394, + 564 + ], + "score": 0.57, + "content": "\\mathrm { D D P M + + }", + "type": "inline_equation" + }, + { + "bbox": [ + 395, + 552, + 505, + 567 + ], + "score": 1.0, + "content": "(Song et al., 2020b) (Meng", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 565, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 505, + 577 + ], + "score": 1.0, + "content": "et al., 2021); AFHQ-dog (Choi et al., 2020) on iDDPM (Nichol & Dhariwal, 2021); and MET-", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 104, + 574, + 505, + 589 + ], + "spans": [ + { + "bbox": [ + 104, + 574, + 505, + 589 + ], + "score": 1.0, + "content": "FACES (Karras et al., 2020) on ADM with P2-weighting (Dhariwal & Nichol, 2021) (Choi et al.,", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 586, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 505, + 600 + ], + "score": 1.0, + "content": "2022). Please note that all models are official pretrained checkpoints and are kept frozen. De-", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 104, + 597, + 506, + 612 + ], + "spans": [ + { + "bbox": [ + 104, + 597, + 287, + 612 + ], + "score": 1.0, + "content": "tailed settings including the coefficients for", + "type": "text" + }, + { + "bbox": [ + 287, + 598, + 313, + 609 + ], + "score": 0.86, + "content": "\\lambda _ { \\mathrm { C L I P } }", + "type": "inline_equation" + }, + { + "bbox": [ + 314, + 597, + 333, + 612 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 334, + 598, + 360, + 609 + ], + "score": 0.92, + "content": "\\lambda _ { \\mathrm { { r e c o n } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 360, + 597, + 506, + 612 + ], + "score": 1.0, + "content": ", and source/target descriptions can", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 609, + 505, + 622 + ], + "spans": [ + { + "bbox": [ + 106, + 609, + 257, + 622 + ], + "score": 1.0, + "content": "be found in Appendix J.1. We train", + "type": "text" + }, + { + "bbox": [ + 257, + 609, + 268, + 620 + ], + "score": 0.87, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 268, + 609, + 291, + 622 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 291, + 609, + 327, + 619 + ], + "score": 0.89, + "content": "S ~ = ~ 4 0", + "type": "inline_equation" + }, + { + "bbox": [ + 328, + 609, + 505, + 622 + ], + "score": 1.0, + "content": "for 1 epoch using 1000 samples. The real", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 620, + 505, + 632 + ], + "spans": [ + { + "bbox": [ + 105, + 620, + 505, + 632 + ], + "score": 1.0, + "content": "samples are randomly chosen from each dataset for in-domain-like attributes. For out-of-domain-", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 630, + 505, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 338, + 643 + ], + "score": 1.0, + "content": "like attributes, we randomly draw 1,000 latent variables", + "type": "text" + }, + { + "bbox": [ + 338, + 630, + 401, + 642 + ], + "score": 0.92, + "content": "\\mathbf { \\boldsymbol { x } } _ { T } \\sim \\mathcal { N } ( \\mathbf { \\boldsymbol { 0 } } , I )", + "type": "inline_equation" + }, + { + "bbox": [ + 401, + 630, + 505, + 643 + ], + "score": 1.0, + "content": ". Details are described in", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 641, + 506, + 655 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 506, + 655 + ], + "score": 1.0, + "content": "Appendix J.2. Training takes about 20 minutes with three RTX 3090 GPUs. All the images in", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 652, + 505, + 665 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 292, + 665 + ], + "score": 1.0, + "content": "the figures are not used for training. We set", + "type": "text" + }, + { + "bbox": [ + 292, + 652, + 343, + 665 + ], + "score": 0.92, + "content": "\\tilde { S } = 1 , 0 0 0", + "type": "inline_equation" + }, + { + "bbox": [ + 343, + 653, + 505, + 665 + ], + "score": 1.0, + "content": "for inference. The code is available at", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 665, + 289, + 677 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 289, + 677 + ], + "score": 1.0, + "content": "https://github.com/kwonminki/Asyrp official", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 37.5 + }, + { + "type": "title", + "bbox": [ + 108, + 689, + 294, + 700 + ], + "lines": [ + { + "bbox": [ + 105, + 687, + 295, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 204, + 702 + ], + "score": 1.0, + "content": "5.1 VERSATILITY OF", + "type": "text" + }, + { + "bbox": [ + 204, + 690, + 210, + 699 + ], + "score": 0.63, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 211, + 687, + 295, + 702 + ], + "score": 1.0, + "content": "-space WITH ASYRP", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 44 + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "Figure 4 shows the effectiveness of our method on various datasets and any existing U-Net based", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 720, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 505, + 733 + ], + "score": 1.0, + "content": "architectures. Our method can synthesize the attributes that are not even included in the training", + "type": "text" + } + ], + "index": 46 + } + ], + "index": 45.5 + } + ], + "page_idx": 5, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "score": 1.0, + "content": "6", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 128 + ], + "lines": [ + { + "bbox": [ + 107, + 81, + 505, + 96 + ], + "spans": [ + { + "bbox": [ + 107, + 82, + 132, + 95 + ], + "score": 0.91, + "content": "E _ { T } ( \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 132, + 81, + 262, + 96 + ], + "score": 1.0, + "content": "produces CLIP text embedding,", + "type": "text" + }, + { + "bbox": [ + 262, + 84, + 277, + 95 + ], + "score": 0.84, + "content": "y _ { ( \\cdot ) }", + "type": "inline_equation" + }, + { + "bbox": [ + 277, + 81, + 396, + 96 + ], + "score": 1.0, + "content": "denotes the descriptions, and", + "type": "text" + }, + { + "bbox": [ + 397, + 82, + 420, + 94 + ], + "score": 0.91, + "content": "d ( \\cdot , \\cdot )", + "type": "inline_equation" + }, + { + "bbox": [ + 421, + 81, + 505, + 96 + ], + "score": 1.0, + "content": "computes the cosine", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 104, + 92, + 506, + 109 + ], + "spans": [ + { + "bbox": [ + 104, + 92, + 284, + 109 + ], + "score": 1.0, + "content": "distance between the arguments. Choosing", + "type": "text" + }, + { + "bbox": [ + 284, + 95, + 303, + 106 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 303, + 92, + 326, + 109 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 326, + 94, + 452, + 107 + ], + "score": 0.89, + "content": "\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t _ { \\mathrm { e d i t } } } ) = 0 . 3 3 - \\delta", + "type": "inline_equation" + }, + { + "bbox": [ + 453, + 92, + 506, + 109 + ], + "score": 1.0, + "content": "expands the", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 105, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 105, + 105, + 505, + 118 + ], + "score": 1.0, + "content": "editing interval to a suitable length. It consistently produces good results in various settings. The", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 117, + 309, + 128 + ], + "spans": [ + { + "bbox": [ + 105, + 117, + 309, + 128 + ], + "score": 1.0, + "content": "supporting experiments are shown in Appendix G.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 1.5, + "bbox_fs": [ + 104, + 81, + 506, + 128 + ] + }, + { + "type": "title", + "bbox": [ + 106, + 141, + 376, + 152 + ], + "lines": [ + { + "bbox": [ + 106, + 140, + 378, + 153 + ], + "spans": [ + { + "bbox": [ + 106, + 140, + 378, + 153 + ], + "score": 1.0, + "content": "4.2 QUALITY BOOSTING WITH STOCHASTIC NOISE INJECTION", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 108, + 161, + 504, + 196 + ], + "lines": [ + { + "bbox": [ + 105, + 160, + 505, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 160, + 418, + 174 + ], + "score": 1.0, + "content": "Although DDIM achieves nearly perfect inversion by removing stochasticity", + "type": "text" + }, + { + "bbox": [ + 419, + 162, + 447, + 173 + ], + "score": 0.87, + "content": "( \\eta = 0", + "type": "inline_equation" + }, + { + "bbox": [ + 447, + 160, + 505, + 174 + ], + "score": 1.0, + "content": "), Karras et al.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 172, + 505, + 185 + ], + "spans": [ + { + "bbox": [ + 106, + 172, + 505, + 185 + ], + "score": 1.0, + "content": "(2022) demonstrate that stochasticity improves image quality. Likewise, we inject stochastic noise", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 183, + 244, + 197 + ], + "spans": [ + { + "bbox": [ + 105, + 183, + 201, + 197 + ], + "score": 1.0, + "content": "in the boosting interval", + "type": "text" + }, + { + "bbox": [ + 201, + 183, + 240, + 195 + ], + "score": 0.93, + "content": "[ t _ { \\mathrm { b o o s t } } , 0 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 240, + 183, + 244, + 197 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6, + "bbox_fs": [ + 105, + 160, + 505, + 197 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 200, + 505, + 332 + ], + "lines": [ + { + "bbox": [ + 105, + 199, + 505, + 214 + ], + "spans": [ + { + "bbox": [ + 105, + 199, + 505, + 214 + ], + "score": 1.0, + "content": "Though the longer boosting interval would achieve higher quality, boosting over excessively long", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 211, + 506, + 223 + ], + "spans": [ + { + "bbox": [ + 105, + 211, + 506, + 223 + ], + "score": 1.0, + "content": "intervals would modify the content. Hence, we want to determine the shortest interval that shows", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 222, + 505, + 234 + ], + "spans": [ + { + "bbox": [ + 105, + 222, + 505, + 234 + ], + "score": 1.0, + "content": "enough quality boosting to guarantee minimal change in the content. We consider the noise in the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 232, + 504, + 246 + ], + "spans": [ + { + "bbox": [ + 105, + 232, + 414, + 246 + ], + "score": 1.0, + "content": "image as the capacity for the quality boosting and define quality deficiency at", + "type": "text" + }, + { + "bbox": [ + 414, + 234, + 419, + 243 + ], + "score": 0.36, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 233, + 504, + 244 + ], + "score": 0.89, + "content": ": \\gamma _ { t } = \\mathrm { L P I P S } ( \\pmb { x } , \\pmb { x } _ { t } )", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 243, + 506, + 256 + ], + "spans": [ + { + "bbox": [ + 105, + 243, + 236, + 256 + ], + "score": 1.0, + "content": "indicating the amount of noise in", + "type": "text" + }, + { + "bbox": [ + 237, + 245, + 248, + 255 + ], + "score": 0.86, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 243, + 405, + 256 + ], + "score": 1.0, + "content": "compared to the original image. We use", + "type": "text" + }, + { + "bbox": [ + 406, + 245, + 417, + 255 + ], + "score": 0.87, + "content": "\\mathbf { \\Delta } _ { \\mathbf { \\mathcal { X } } _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 417, + 243, + 457, + 256 + ], + "score": 1.0, + "content": "instead of", + "type": "text" + }, + { + "bbox": [ + 458, + 245, + 470, + 255 + ], + "score": 0.87, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 471, + 243, + 506, + 256 + ], + "score": 1.0, + "content": "because", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 255, + 505, + 267 + ], + "spans": [ + { + "bbox": [ + 106, + 255, + 505, + 267 + ], + "score": 1.0, + "content": "we consider the actual image rather than the semantics. Figure 3 inset depicts editing flexibility and", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 264, + 505, + 281 + ], + "spans": [ + { + "bbox": [ + 105, + 264, + 289, + 281 + ], + "score": 1.0, + "content": "quality deficiency. We empirically find that", + "type": "text" + }, + { + "bbox": [ + 289, + 267, + 314, + 277 + ], + "score": 0.88, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 264, + 338, + 281 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 338, + 266, + 397, + 278 + ], + "score": 0.91, + "content": "\\gamma _ { t _ { \\mathrm { b o o s t } } } ~ = ~ 1 . 2", + "type": "inline_equation" + }, + { + "bbox": [ + 397, + 264, + 505, + 281 + ], + "score": 1.0, + "content": "achieves quality boosting", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 275, + 506, + 290 + ], + "spans": [ + { + "bbox": [ + 105, + 275, + 455, + 290 + ], + "score": 1.0, + "content": "with minimal content change. We confirmed that the editing strength of the intervals", + "type": "text" + }, + { + "bbox": [ + 455, + 277, + 494, + 289 + ], + "score": 0.93, + "content": "[ t _ { \\mathrm { b o o s t } } , 0 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 494, + 275, + 506, + 290 + ], + "score": 1.0, + "content": "is", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 104, + 287, + 507, + 301 + ], + "spans": [ + { + "bbox": [ + 104, + 287, + 302, + 301 + ], + "score": 1.0, + "content": "guaranteed to be less than 0.25. In Figure 3, after", + "type": "text" + }, + { + "bbox": [ + 303, + 288, + 327, + 299 + ], + "score": 0.89, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 327, + 287, + 330, + 301 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 331, + 288, + 390, + 299 + ], + "score": 0.67, + "content": "\\mathrm { L P I P S } ( { \\pmb x } , { \\pmb x } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 390, + 287, + 507, + 301 + ], + "score": 1.0, + "content": "sharply drops in the original", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 104, + 298, + 507, + 312 + ], + "spans": [ + { + "bbox": [ + 104, + 298, + 207, + 312 + ], + "score": 1.0, + "content": "generative process while", + "type": "text" + }, + { + "bbox": [ + 207, + 299, + 268, + 311 + ], + "score": 0.71, + "content": "\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 268, + 298, + 507, + 312 + ], + "score": 1.0, + "content": "changes little. Note that most of the quality degradation of", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 308, + 505, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 308, + 505, + 324 + ], + "score": 1.0, + "content": "the resulting images is caused by DDIM reverse process, not by Asyrp. We use this quality boosting", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 321, + 311, + 333 + ], + "spans": [ + { + "bbox": [ + 105, + 321, + 299, + 333 + ], + "score": 1.0, + "content": "for all experiments except ablation in Appendix", + "type": "text" + }, + { + "bbox": [ + 299, + 321, + 307, + 330 + ], + "score": 0.38, + "content": "_ \\mathrm { H }", + "type": "inline_equation" + }, + { + "bbox": [ + 308, + 321, + 311, + 333 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 13.5, + "bbox_fs": [ + 104, + 199, + 507, + 333 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 345, + 296, + 356 + ], + "lines": [ + { + "bbox": [ + 105, + 344, + 298, + 357 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 298, + 357 + ], + "score": 1.0, + "content": "4.3 OVERALL PROCESS OF IMAGE EDITING", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 365, + 506, + 387 + ], + "lines": [ + { + "bbox": [ + 105, + 364, + 506, + 378 + ], + "spans": [ + { + "bbox": [ + 105, + 364, + 132, + 378 + ], + "score": 1.0, + "content": "Using", + "type": "text" + }, + { + "bbox": [ + 132, + 367, + 151, + 377 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 151, + 364, + 168, + 378 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 168, + 366, + 192, + 377 + ], + "score": 0.88, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 192, + 364, + 506, + 378 + ], + "score": 1.0, + "content": "determined by the above formulas, we modify the generative process of DDIM", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 375, + 128, + 389 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 128, + 389 + ], + "score": 1.0, + "content": "with", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5, + "bbox_fs": [ + 105, + 364, + 506, + 389 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 116, + 387, + 482, + 430 + ], + "lines": [ + { + "bbox": [ + 116, + 387, + 482, + 430 + ], + "spans": [ + { + "bbox": [ + 116, + 387, + 482, + 430 + ], + "score": 0.93, + "content": "p _ { \\theta } ^ { ( t ) } \\left( x _ { t - 1 } \\mid x _ { t } \\right) = \\left\\{ \\begin{array} { l l } { \\mathcal { N } \\left( \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } | f _ { t } ) ) + \\mathbf { D } _ { t } , \\sigma _ { t } ^ { 2 } I \\right) , \\eta = 0 } & { \\mathrm { ~ i f ~ } T \\geq t \\geq t _ { \\mathrm { e d i t } } } \\\\ { \\mathcal { N } \\left( \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) + \\mathbf { D } _ { t } , \\sigma _ { t } ^ { 2 } I \\right) , \\eta = 0 } & { \\mathrm { ~ i f ~ } t _ { \\mathrm { e d i t } } > t \\geq t _ { \\mathrm { b o o s t } } } \\\\ { \\mathcal { N } \\left( \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) + \\mathbf { D } _ { t } , \\sigma _ { t } ^ { 2 } I \\right) , \\eta = 1 } & { \\mathrm { ~ i f ~ } t _ { \\mathrm { b o o s t } } > t } \\end{array} \\right.", + "type": "interline_equation", + "image_path": "1fd23e676ae9069b2a174b8dbe503c18b765107d5f7690ab6bab96976aa085a0.jpg" + } + ] + } + ], + "index": 24, + "virtual_lines": [ + { + "bbox": [ + 116, + 387, + 482, + 401.3333333333333 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 116, + 401.3333333333333, + 482, + 415.66666666666663 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 116, + 415.66666666666663, + 482, + 429.99999999999994 + ], + "spans": [], + "index": 25 + } + ] + }, + { + "type": "text", + "bbox": [ + 105, + 436, + 471, + 447 + ], + "lines": [ + { + "bbox": [ + 105, + 435, + 473, + 449 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 473, + 449 + ], + "score": 1.0, + "content": "The visual overview and comprehensive algorithms of the entire process are in Appendix I.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 26, + "bbox_fs": [ + 105, + 435, + 473, + 449 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 462, + 200, + 475 + ], + "lines": [ + { + "bbox": [ + 105, + 462, + 201, + 477 + ], + "spans": [ + { + "bbox": [ + 105, + 462, + 201, + 477 + ], + "score": 1.0, + "content": "5 EXPERIMENTS", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 106, + 487, + 505, + 531 + ], + "lines": [ + { + "bbox": [ + 105, + 487, + 505, + 499 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 381, + 499 + ], + "score": 1.0, + "content": "In this section, we show the effectiveness of semantic latent editing in", + "type": "text" + }, + { + "bbox": [ + 382, + 488, + 388, + 497 + ], + "score": 0.61, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 389, + 487, + 505, + 499 + ], + "score": 1.0, + "content": "-space with Asyrp on various", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 497, + 506, + 511 + ], + "spans": [ + { + "bbox": [ + 105, + 497, + 267, + 511 + ], + "score": 1.0, + "content": "attributes, datasets and architectures in", + "type": "text" + }, + { + "bbox": [ + 267, + 498, + 288, + 510 + ], + "score": 0.54, + "content": "\\ S \\ S . 1", + "type": "inline_equation" + }, + { + "bbox": [ + 288, + 497, + 506, + 511 + ], + "score": 1.0, + "content": ". Moreover, we provide quantitative results including", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 509, + 505, + 521 + ], + "spans": [ + { + "bbox": [ + 106, + 510, + 162, + 521 + ], + "score": 1.0, + "content": "user study in", + "type": "text" + }, + { + "bbox": [ + 163, + 509, + 185, + 521 + ], + "score": 0.63, + "content": "\\mathrm { ~ \\normalfont ~ \\ S ~ } 5 . 2", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 510, + 505, + 521 + ], + "score": 1.0, + "content": ". Lastly, we provide detailed analyses for the properties of the semantic latent", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 520, + 276, + 532 + ], + "spans": [ + { + "bbox": [ + 105, + 521, + 143, + 532 + ], + "score": 1.0, + "content": "space on", + "type": "text" + }, + { + "bbox": [ + 144, + 521, + 150, + 530 + ], + "score": 0.64, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 150, + 521, + 252, + 532 + ], + "score": 1.0, + "content": "-space and alternatives in", + "type": "text" + }, + { + "bbox": [ + 252, + 520, + 273, + 532 + ], + "score": 0.85, + "content": "\\ S 5 . 3", + "type": "inline_equation" + }, + { + "bbox": [ + 273, + 521, + 276, + 532 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 29.5, + "bbox_fs": [ + 105, + 487, + 506, + 532 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 542, + 505, + 676 + ], + "lines": [ + { + "bbox": [ + 105, + 542, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 505, + 555 + ], + "score": 1.0, + "content": "Implementation details. We implement our method on various settings: CelebA-HQ (Karras", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 104, + 552, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 104, + 552, + 352, + 567 + ], + "score": 1.0, + "content": "et al., 2018) and LSUN-bedroom/-church (Yu et al., 2015) on", + "type": "text" + }, + { + "bbox": [ + 353, + 554, + 394, + 564 + ], + "score": 0.57, + "content": "\\mathrm { D D P M + + }", + "type": "inline_equation" + }, + { + "bbox": [ + 395, + 552, + 505, + 567 + ], + "score": 1.0, + "content": "(Song et al., 2020b) (Meng", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 565, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 505, + 577 + ], + "score": 1.0, + "content": "et al., 2021); AFHQ-dog (Choi et al., 2020) on iDDPM (Nichol & Dhariwal, 2021); and MET-", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 104, + 574, + 505, + 589 + ], + "spans": [ + { + "bbox": [ + 104, + 574, + 505, + 589 + ], + "score": 1.0, + "content": "FACES (Karras et al., 2020) on ADM with P2-weighting (Dhariwal & Nichol, 2021) (Choi et al.,", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 586, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 505, + 600 + ], + "score": 1.0, + "content": "2022). Please note that all models are official pretrained checkpoints and are kept frozen. De-", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 104, + 597, + 506, + 612 + ], + "spans": [ + { + "bbox": [ + 104, + 597, + 287, + 612 + ], + "score": 1.0, + "content": "tailed settings including the coefficients for", + "type": "text" + }, + { + "bbox": [ + 287, + 598, + 313, + 609 + ], + "score": 0.86, + "content": "\\lambda _ { \\mathrm { C L I P } }", + "type": "inline_equation" + }, + { + "bbox": [ + 314, + 597, + 333, + 612 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 334, + 598, + 360, + 609 + ], + "score": 0.92, + "content": "\\lambda _ { \\mathrm { { r e c o n } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 360, + 597, + 506, + 612 + ], + "score": 1.0, + "content": ", and source/target descriptions can", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 609, + 505, + 622 + ], + "spans": [ + { + "bbox": [ + 106, + 609, + 257, + 622 + ], + "score": 1.0, + "content": "be found in Appendix J.1. We train", + "type": "text" + }, + { + "bbox": [ + 257, + 609, + 268, + 620 + ], + "score": 0.87, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 268, + 609, + 291, + 622 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 291, + 609, + 327, + 619 + ], + "score": 0.89, + "content": "S ~ = ~ 4 0", + "type": "inline_equation" + }, + { + "bbox": [ + 328, + 609, + 505, + 622 + ], + "score": 1.0, + "content": "for 1 epoch using 1000 samples. The real", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 620, + 505, + 632 + ], + "spans": [ + { + "bbox": [ + 105, + 620, + 505, + 632 + ], + "score": 1.0, + "content": "samples are randomly chosen from each dataset for in-domain-like attributes. For out-of-domain-", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 630, + 505, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 338, + 643 + ], + "score": 1.0, + "content": "like attributes, we randomly draw 1,000 latent variables", + "type": "text" + }, + { + "bbox": [ + 338, + 630, + 401, + 642 + ], + "score": 0.92, + "content": "\\mathbf { \\boldsymbol { x } } _ { T } \\sim \\mathcal { N } ( \\mathbf { \\boldsymbol { 0 } } , I )", + "type": "inline_equation" + }, + { + "bbox": [ + 401, + 630, + 505, + 643 + ], + "score": 1.0, + "content": ". Details are described in", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 641, + 506, + 655 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 506, + 655 + ], + "score": 1.0, + "content": "Appendix J.2. Training takes about 20 minutes with three RTX 3090 GPUs. All the images in", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 652, + 505, + 665 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 292, + 665 + ], + "score": 1.0, + "content": "the figures are not used for training. We set", + "type": "text" + }, + { + "bbox": [ + 292, + 652, + 343, + 665 + ], + "score": 0.92, + "content": "\\tilde { S } = 1 , 0 0 0", + "type": "inline_equation" + }, + { + "bbox": [ + 343, + 653, + 505, + 665 + ], + "score": 1.0, + "content": "for inference. The code is available at", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 665, + 289, + 677 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 289, + 677 + ], + "score": 1.0, + "content": "https://github.com/kwonminki/Asyrp official", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 37.5, + "bbox_fs": [ + 104, + 542, + 506, + 677 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 689, + 294, + 700 + ], + "lines": [ + { + "bbox": [ + 105, + 687, + 295, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 204, + 702 + ], + "score": 1.0, + "content": "5.1 VERSATILITY OF", + "type": "text" + }, + { + "bbox": [ + 204, + 690, + 210, + 699 + ], + "score": 0.63, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 211, + 687, + 295, + 702 + ], + "score": 1.0, + "content": "-space WITH ASYRP", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 44 + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "Figure 4 shows the effectiveness of our method on various datasets and any existing U-Net based", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 720, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 505, + 733 + ], + "score": 1.0, + "content": "architectures. Our method can synthesize the attributes that are not even included in the training", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 106, + 665, + 505, + 677 + ], + "spans": [ + { + "bbox": [ + 106, + 665, + 201, + 677 + ], + "score": 1.0, + "content": "dataset, such as church", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 201, + 666, + 221, + 677 + ], + "score": 0.52, + "content": " \\{ \\begin{array} { r l } \\end{array} ", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 221, + 665, + 505, + 677 + ], + "score": 1.0, + "content": "department, factory, and temple}. Even for dogs, our method", + "type": "text", + "cross_page": true + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 677, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 506, + 689 + ], + "score": 1.0, + "content": "synthesizes smiling Poodle and Yorkshire, the species that barely smile in the dataset. Figure 5", + "type": "text", + "cross_page": true + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "score": 1.0, + "content": "provides results for changing human faces to different identities, painting styles, and ancient pri-", + "type": "text", + "cross_page": true + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "mates. More result can be found in Appendix N. Versatility of our method is surprising because", + "type": "text", + "cross_page": true + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 708, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 395, + 723 + ], + "score": 1.0, + "content": "we do not alter the models but only shift the bottleneck feature maps in", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 396, + 710, + 402, + 720 + ], + "score": 0.67, + "content": "h", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 402, + 708, + 505, + 723 + ], + "score": 1.0, + "content": "-space with Asyrp during", + "type": "text", + "cross_page": true + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 720, + 149, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 149, + 733 + ], + "score": 1.0, + "content": "inference.", + "type": "text", + "cross_page": true + } + ], + "index": 14 + } + ], + "index": 45.5, + "bbox_fs": [ + 105, + 709, + 505, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 124, + 77, + 486, + 506 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 124, + 77, + 486, + 506 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 124, + 77, + 486, + 506 + ], + "spans": [ + { + "bbox": [ + 124, + 77, + 486, + 506 + ], + "score": 0.967, + "type": "image", + "image_path": "306d377572a44f6224fe73778efba99d8a0dfab3966d278640038c75118326f5.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 124, + 77, + 486, + 220.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 124, + 220.0, + 486, + 363.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 124, + 363.0, + 486, + 506.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 105, + 517, + 504, + 540 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 515, + 505, + 531 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 505, + 531 + ], + "score": 1.0, + "content": "Figure 4: Editing results of Asyrp on various datasets. We conduct experiments on CelebA-HQ,", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 527, + 357, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 527, + 357, + 541 + ], + "score": 1.0, + "content": "LSUN-church, METFACES, AFHQ-dog, and LSUN-bedroom.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "image", + "bbox": [ + 167, + 545, + 445, + 603 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 167, + 545, + 445, + 603 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 167, + 545, + 445, + 603 + ], + "spans": [ + { + "bbox": [ + 167, + 545, + 445, + 603 + ], + "score": 0.967, + "type": "image", + "image_path": "a2ba33a9e5710b55fca14f10f93b7fa652ad4c1a82d2bbb67e10456214757d19.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 167, + 545, + 445, + 564.3333333333334 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 167, + 564.3333333333334, + 445, + 583.6666666666667 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 167, + 583.6666666666667, + 445, + 603.0000000000001 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 140, + 614, + 468, + 626 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 138, + 613, + 470, + 628 + ], + "spans": [ + { + "bbox": [ + 138, + 613, + 470, + 628 + ], + "score": 1.0, + "content": "Figure 5: Editing results of Asyrp for unseen domains in CelebA-HQ dataset.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + } + ], + "index": 7.0 + }, + { + "type": "text", + "bbox": [ + 106, + 666, + 505, + 731 + ], + "lines": [ + { + "bbox": [ + 106, + 665, + 505, + 677 + ], + "spans": [ + { + "bbox": [ + 106, + 665, + 201, + 677 + ], + "score": 1.0, + "content": "dataset, such as church", + "type": "text" + }, + { + "bbox": [ + 201, + 666, + 221, + 677 + ], + "score": 0.52, + "content": " \\{ \\begin{array} { r l } \\end{array} ", + "type": "inline_equation" + }, + { + "bbox": [ + 221, + 665, + 505, + 677 + ], + "score": 1.0, + "content": "department, factory, and temple}. Even for dogs, our method", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 677, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 506, + 689 + ], + "score": 1.0, + "content": "synthesizes smiling Poodle and Yorkshire, the species that barely smile in the dataset. Figure 5", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "score": 1.0, + "content": "provides results for changing human faces to different identities, painting styles, and ancient pri-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "mates. More result can be found in Appendix N. Versatility of our method is surprising because", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 708, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 395, + 723 + ], + "score": 1.0, + "content": "we do not alter the models but only shift the bottleneck feature maps in", + "type": "text" + }, + { + "bbox": [ + 396, + 710, + 402, + 720 + ], + "score": 0.67, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 402, + 708, + 505, + 723 + ], + "score": 1.0, + "content": "-space with Asyrp during", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 720, + 149, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 149, + 733 + ], + "score": 1.0, + "content": "inference.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 11.5 + } + ], + "page_idx": 6, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 309, + 762 + ], + "score": 1.0, + "content": "7", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 124, + 77, + 486, + 506 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 124, + 77, + 486, + 506 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 124, + 77, + 486, + 506 + ], + "spans": [ + { + "bbox": [ + 124, + 77, + 486, + 506 + ], + "score": 0.967, + "type": "image", + "image_path": "306d377572a44f6224fe73778efba99d8a0dfab3966d278640038c75118326f5.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 124, + 77, + 486, + 220.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 124, + 220.0, + 486, + 363.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 124, + 363.0, + 486, + 506.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 105, + 517, + 504, + 540 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 515, + 505, + 531 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 505, + 531 + ], + "score": 1.0, + "content": "Figure 4: Editing results of Asyrp on various datasets. We conduct experiments on CelebA-HQ,", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 527, + 357, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 527, + 357, + 541 + ], + "score": 1.0, + "content": "LSUN-church, METFACES, AFHQ-dog, and LSUN-bedroom.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "image", + "bbox": [ + 167, + 545, + 445, + 603 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 167, + 545, + 445, + 603 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 167, + 545, + 445, + 603 + ], + "spans": [ + { + "bbox": [ + 167, + 545, + 445, + 603 + ], + "score": 0.967, + "type": "image", + "image_path": "a2ba33a9e5710b55fca14f10f93b7fa652ad4c1a82d2bbb67e10456214757d19.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 167, + 545, + 445, + 564.3333333333334 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 167, + 564.3333333333334, + 445, + 583.6666666666667 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 167, + 583.6666666666667, + 445, + 603.0000000000001 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 140, + 614, + 468, + 626 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 138, + 613, + 470, + 628 + ], + "spans": [ + { + "bbox": [ + 138, + 613, + 470, + 628 + ], + "score": 1.0, + "content": "Figure 5: Editing results of Asyrp for unseen domains in CelebA-HQ dataset.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + } + ], + "index": 7.0 + }, + { + "type": "text", + "bbox": [ + 106, + 666, + 505, + 731 + ], + "lines": [], + "index": 11.5, + "bbox_fs": [ + 105, + 665, + 506, + 733 + ], + "lines_deleted": true + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 107, + 80, + 504, + 117 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 107, + 80, + 504, + 117 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 80, + 504, + 117 + ], + "spans": [ + { + "bbox": [ + 107, + 80, + 504, + 117 + ], + "score": 0.961, + "html": "
CelebA-HQ in-domainCelebA-HQ unseen-domainLSUN-church
qualityattributeoverallqualityattributeoverallqualityattributediversityoverall
Asyrp (ours)98.36%88.13%94.92%71.56%59.84%63.13%73.19%71.81%87.50%76.81%
DiffusionCLIP1.64%11.88%5.08%28.44%40.16%36.88%26.81%28.19%12.50%23.19%
", + "type": "table", + "image_path": "16a51330a7547a7d3496a7eba87648c05f50719972f0cfe8bf62f5409f30f898.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 80, + 504, + 92.33333333333333 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 92.33333333333333, + 504, + 104.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 104.66666666666666, + 504, + 116.99999999999999 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 154, + 125, + 456, + 137 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 153, + 124, + 456, + 138 + ], + "spans": [ + { + "bbox": [ + 153, + 124, + 433, + 138 + ], + "score": 1.0, + "content": "Table 1: User study with 80 participants. The details are described in", + "type": "text" + }, + { + "bbox": [ + 433, + 125, + 456, + 136 + ], + "score": 0.47, + "content": "\\ S \\ K . 1", + "type": "inline_equation" + } + ], + "index": 3 + } + ], + "index": 3 + } + ], + "index": 2.0 + }, + { + "type": "image", + "bbox": [ + 115, + 145, + 497, + 210 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 145, + 497, + 210 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 115, + 145, + 497, + 210 + ], + "spans": [ + { + "bbox": [ + 115, + 145, + 497, + 210 + ], + "score": 0.955, + "type": "image", + "image_path": "66fd5aab2ef8fa4bdecea189af2d5eb233396464cc901dad958ae0cc0424c117.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 115, + 145, + 497, + 166.66666666666666 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 115, + 166.66666666666666, + 497, + 188.33333333333331 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 115, + 188.33333333333331, + 497, + 209.99999999999997 + ], + "spans": [], + "index": 6 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 222, + 505, + 267 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 223, + 504, + 235 + ], + "spans": [ + { + "bbox": [ + 105, + 223, + 280, + 235 + ], + "score": 1.0, + "content": "Figure 6: Optimization for smiling on", + "type": "text" + }, + { + "bbox": [ + 280, + 223, + 287, + 233 + ], + "score": 0.39, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 288, + 223, + 432, + 235 + ], + "score": 1.0, + "content": "-space and ϵ-space. (a) Optimizing", + "type": "text" + }, + { + "bbox": [ + 432, + 223, + 451, + 234 + ], + "score": 0.9, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 223, + 504, + 235 + ], + "score": 1.0, + "content": "for a sample", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 234, + 504, + 245 + ], + "spans": [ + { + "bbox": [ + 105, + 234, + 411, + 245 + ], + "score": 1.0, + "content": "with smiling results in natural editing while the change due to optimizing", + "type": "text" + }, + { + "bbox": [ + 411, + 234, + 429, + 244 + ], + "score": 0.91, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 429, + 234, + 504, + 245 + ], + "score": 1.0, + "content": "is relatively small.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 245, + 505, + 257 + ], + "spans": [ + { + "bbox": [ + 106, + 245, + 160, + 257 + ], + "score": 1.0, + "content": "(b) Applying", + "type": "text" + }, + { + "bbox": [ + 161, + 245, + 180, + 255 + ], + "score": 0.91, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 181, + 245, + 439, + 257 + ], + "score": 1.0, + "content": "from (a) to other samples yields the same attribute change while", + "type": "text" + }, + { + "bbox": [ + 439, + 245, + 457, + 255 + ], + "score": 0.9, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 458, + 245, + 505, + 257 + ], + "score": 1.0, + "content": "distorts the", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 256, + 344, + 267 + ], + "spans": [ + { + "bbox": [ + 105, + 256, + 193, + 267 + ], + "score": 1.0, + "content": "images. The result of", + "type": "text" + }, + { + "bbox": [ + 194, + 256, + 211, + 266 + ], + "score": 0.91, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 211, + 256, + 344, + 267 + ], + "score": 1.0, + "content": "is the best sample we could find.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 8.5 + } + ], + "index": 6.75 + }, + { + "type": "image", + "bbox": [ + 114, + 273, + 497, + 347 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 114, + 273, + 497, + 347 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 114, + 273, + 497, + 347 + ], + "spans": [ + { + "bbox": [ + 114, + 273, + 497, + 347 + ], + "score": 0.964, + "type": "image", + "image_path": "7bac52c03fecc44b4f929ac853df2555bd33134ce5593830fb81ee5e8c225924.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 114, + 273, + 497, + 297.6666666666667 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 114, + 297.6666666666667, + 497, + 322.33333333333337 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 114, + 322.33333333333337, + 497, + 347.00000000000006 + ], + "spans": [], + "index": 13 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 358, + 505, + 406 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 358, + 505, + 371 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 205, + 371 + ], + "score": 1.0, + "content": "Figure 7: Linearity of", + "type": "text" + }, + { + "bbox": [ + 206, + 359, + 213, + 369 + ], + "score": 0.58, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 213, + 358, + 415, + 371 + ], + "score": 1.0, + "content": "-space. Linear interpolation and extrapolation on", + "type": "text" + }, + { + "bbox": [ + 415, + 359, + 422, + 369 + ], + "score": 0.67, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 422, + 358, + 505, + 371 + ], + "score": 1.0, + "content": "-space lead to grad-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 370, + 505, + 382 + ], + "spans": [ + { + "bbox": [ + 106, + 370, + 505, + 382 + ], + "score": 1.0, + "content": "ual changes even to the unseen intensity and directions. Right side shows the interpolation results", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 104, + 378, + 507, + 396 + ], + "spans": [ + { + "bbox": [ + 104, + 378, + 197, + 396 + ], + "score": 1.0, + "content": "by positive scaling of", + "type": "text" + }, + { + "bbox": [ + 197, + 380, + 242, + 394 + ], + "score": 0.93, + "content": "\\Delta h _ { t } ^ { \\mathrm { s m i l i n g } }", + "type": "inline_equation" + }, + { + "bbox": [ + 242, + 378, + 507, + 396 + ], + "score": 1.0, + "content": ". Left side shows the extrapolation results by negative scaling of", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 107, + 391, + 376, + 408 + ], + "spans": [ + { + "bbox": [ + 107, + 393, + 150, + 407 + ], + "score": 0.92, + "content": "\\Delta h _ { t } ^ { s m i l i n g }", + "type": "inline_equation" + }, + { + "bbox": [ + 150, + 391, + 259, + 408 + ], + "score": 1.0, + "content": ". Note that we did not train", + "type": "text" + }, + { + "bbox": [ + 259, + 395, + 269, + 406 + ], + "score": 0.86, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 391, + 376, + 408 + ], + "score": 1.0, + "content": "for the negative direction.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 15.5 + } + ], + "index": 13.75 + }, + { + "type": "title", + "bbox": [ + 108, + 432, + 255, + 443 + ], + "lines": [ + { + "bbox": [ + 105, + 432, + 257, + 445 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 257, + 445 + ], + "score": 1.0, + "content": "5.2 QUANTITATIVE COMPARISON", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 106, + 452, + 505, + 552 + ], + "lines": [ + { + "bbox": [ + 105, + 451, + 506, + 466 + ], + "spans": [ + { + "bbox": [ + 105, + 451, + 506, + 466 + ], + "score": 1.0, + "content": "Considering that our method can be combined with various diffusion models without finetuning,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 464, + 504, + 475 + ], + "spans": [ + { + "bbox": [ + 106, + 464, + 504, + 475 + ], + "score": 1.0, + "content": "we do not find such a versatile competitor. Nonetheless, we compare Asyrp against DiffusionCLIP", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 474, + 505, + 487 + ], + "spans": [ + { + "bbox": [ + 105, + 474, + 505, + 487 + ], + "score": 1.0, + "content": "using the official code that edits the real images by finetuning the whole model. We asked 80 partic-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 485, + 505, + 498 + ], + "spans": [ + { + "bbox": [ + 106, + 485, + 505, + 498 + ], + "score": 1.0, + "content": "ipants to choose the images with better quality, natural attribute change, and overall preference for", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 496, + 505, + 509 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 505, + 509 + ], + "score": 1.0, + "content": "given total of 40 sets of original images, ours, and DiffusionCLIP. Table 1 shows that Asyrp outper-", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 506, + 504, + 520 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 504, + 520 + ], + "score": 1.0, + "content": "forms DiffusionCLIP in the all perspectives including the attributes unseen in the training dataset.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 516, + 505, + 532 + ], + "spans": [ + { + "bbox": [ + 105, + 516, + 505, + 532 + ], + "score": 1.0, + "content": "We list the settings for fair comparison including the questions and example images in Appendix", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 528, + 505, + 543 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 144, + 543 + ], + "score": 1.0, + "content": "K.1. See", + "type": "text" + }, + { + "bbox": [ + 144, + 529, + 169, + 541 + ], + "score": 0.62, + "content": "\\textrm { \\AA K } . 2", + "type": "inline_equation" + }, + { + "bbox": [ + 169, + 528, + 505, + 543 + ], + "score": 1.0, + "content": "for more evaluation metrics: segmentation consistency (SC) and directional CLIP", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 538, + 173, + 553 + ], + "spans": [ + { + "bbox": [ + 105, + 538, + 148, + 553 + ], + "score": 1.0, + "content": "similarity(", + "type": "text" + }, + { + "bbox": [ + 148, + 541, + 168, + 551 + ], + "score": 0.84, + "content": "S _ { \\mathrm { d i r } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 168, + 538, + 173, + 553 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 23 + }, + { + "type": "title", + "bbox": [ + 108, + 565, + 224, + 577 + ], + "lines": [ + { + "bbox": [ + 105, + 563, + 226, + 579 + ], + "spans": [ + { + "bbox": [ + 105, + 563, + 191, + 579 + ], + "score": 1.0, + "content": "5.3 ANALYSIS ON", + "type": "text" + }, + { + "bbox": [ + 191, + 566, + 198, + 575 + ], + "score": 0.38, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 198, + 563, + 226, + 579 + ], + "score": 1.0, + "content": "-space", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 106, + 586, + 503, + 608 + ], + "lines": [ + { + "bbox": [ + 106, + 585, + 505, + 598 + ], + "spans": [ + { + "bbox": [ + 106, + 585, + 505, + 598 + ], + "score": 1.0, + "content": "We provide detailed analyses to validate the properties of semantic latent space for diffusion models:", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 596, + 383, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 596, + 383, + 609 + ], + "score": 1.0, + "content": "homogeneity, linearity, robustness, and consistency across timesteps.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5 + }, + { + "type": "text", + "bbox": [ + 107, + 620, + 504, + 664 + ], + "lines": [ + { + "bbox": [ + 106, + 621, + 504, + 632 + ], + "spans": [ + { + "bbox": [ + 106, + 621, + 319, + 632 + ], + "score": 1.0, + "content": "Homogeneity. Figure 6 illustrates homogeneity of", + "type": "text" + }, + { + "bbox": [ + 320, + 621, + 326, + 630 + ], + "score": 0.75, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 326, + 621, + 406, + 632 + ], + "score": 1.0, + "content": "-space compared to", + "type": "text" + }, + { + "bbox": [ + 407, + 623, + 413, + 630 + ], + "score": 0.61, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 414, + 621, + 462, + 632 + ], + "score": 1.0, + "content": "-space. One", + "type": "text" + }, + { + "bbox": [ + 462, + 621, + 482, + 631 + ], + "score": 0.9, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 482, + 621, + 504, + 632 + ], + "score": 1.0, + "content": "opti-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 631, + 505, + 643 + ], + "spans": [ + { + "bbox": [ + 106, + 631, + 505, + 643 + ], + "score": 1.0, + "content": "mized for an image results in the same attribute change to other input images. On the other hand,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 104, + 641, + 504, + 655 + ], + "spans": [ + { + "bbox": [ + 104, + 641, + 124, + 655 + ], + "score": 1.0, + "content": "one", + "type": "text" + }, + { + "bbox": [ + 125, + 642, + 142, + 653 + ], + "score": 0.87, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 143, + 641, + 456, + 655 + ], + "score": 1.0, + "content": "optimized for an image distorts other input images. In Figure 10, applying", + "type": "text" + }, + { + "bbox": [ + 456, + 642, + 504, + 654 + ], + "score": 0.91, + "content": "\\Delta h _ { t } ^ { \\mathrm { m e a n } } =", + "type": "inline_equation" + } + ], + "index": 33 + }, + { + "bbox": [ + 107, + 652, + 502, + 668 + ], + "spans": [ + { + "bbox": [ + 107, + 653, + 149, + 666 + ], + "score": 0.91, + "content": "\\begin{array} { r } { \\frac { 1 } { N } \\sum \\Delta h _ { t } ^ { i } } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 150, + 652, + 310, + 668 + ], + "score": 1.0, + "content": "produces almost identical results where", + "type": "text" + }, + { + "bbox": [ + 311, + 654, + 315, + 663 + ], + "score": 0.72, + "content": "i", + "type": "inline_equation" + }, + { + "bbox": [ + 316, + 652, + 396, + 668 + ], + "score": 1.0, + "content": "indicates indices of", + "type": "text" + }, + { + "bbox": [ + 396, + 654, + 430, + 663 + ], + "score": 0.9, + "content": "N = 2 0", + "type": "inline_equation" + }, + { + "bbox": [ + 430, + 652, + 502, + 668 + ], + "score": 1.0, + "content": "random samples.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 32.5 + }, + { + "type": "text", + "bbox": [ + 107, + 676, + 504, + 731 + ], + "lines": [ + { + "bbox": [ + 106, + 677, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 352, + 689 + ], + "score": 1.0, + "content": "Linearity. In Figure 7, we observe that linearly scaling a", + "type": "text" + }, + { + "bbox": [ + 352, + 677, + 368, + 687 + ], + "score": 0.76, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 368, + 677, + 505, + 689 + ], + "score": 1.0, + "content": "reflects the amount of change in", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "score": 1.0, + "content": "the visual attributes. Surprisingly, it generalizes to negative scales that are not seen during train-", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 356, + 711 + ], + "score": 1.0, + "content": "ing. Moreover, Figure 8 shows that combinations of different", + "type": "text" + }, + { + "bbox": [ + 356, + 699, + 372, + 709 + ], + "score": 0.77, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 372, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "’s yield their combined semantic", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "changes in the resulting images. Appendix N.2 provides mixed interpolation between multiple at-", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 720, + 142, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 142, + 734 + ], + "score": 1.0, + "content": "tributes.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 37 + } + ], + "page_idx": 7, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 300, + 750, + 309, + 761 + ], + "spans": [ + { + "bbox": [ + 300, + 750, + 309, + 761 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 11, + "width": 9 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 107, + 80, + 504, + 117 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 107, + 80, + 504, + 117 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 80, + 504, + 117 + ], + "spans": [ + { + "bbox": [ + 107, + 80, + 504, + 117 + ], + "score": 0.961, + "html": "
CelebA-HQ in-domainCelebA-HQ unseen-domainLSUN-church
qualityattributeoverallqualityattributeoverallqualityattributediversityoverall
Asyrp (ours)98.36%88.13%94.92%71.56%59.84%63.13%73.19%71.81%87.50%76.81%
DiffusionCLIP1.64%11.88%5.08%28.44%40.16%36.88%26.81%28.19%12.50%23.19%
", + "type": "table", + "image_path": "16a51330a7547a7d3496a7eba87648c05f50719972f0cfe8bf62f5409f30f898.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 80, + 504, + 92.33333333333333 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 92.33333333333333, + 504, + 104.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 104.66666666666666, + 504, + 116.99999999999999 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 154, + 125, + 456, + 137 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 153, + 124, + 456, + 138 + ], + "spans": [ + { + "bbox": [ + 153, + 124, + 433, + 138 + ], + "score": 1.0, + "content": "Table 1: User study with 80 participants. The details are described in", + "type": "text" + }, + { + "bbox": [ + 433, + 125, + 456, + 136 + ], + "score": 0.47, + "content": "\\ S \\ K . 1", + "type": "inline_equation" + } + ], + "index": 3 + } + ], + "index": 3 + } + ], + "index": 2.0 + }, + { + "type": "image", + "bbox": [ + 115, + 145, + 497, + 210 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 145, + 497, + 210 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 115, + 145, + 497, + 210 + ], + "spans": [ + { + "bbox": [ + 115, + 145, + 497, + 210 + ], + "score": 0.955, + "type": "image", + "image_path": "66fd5aab2ef8fa4bdecea189af2d5eb233396464cc901dad958ae0cc0424c117.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 115, + 145, + 497, + 166.66666666666666 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 115, + 166.66666666666666, + 497, + 188.33333333333331 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 115, + 188.33333333333331, + 497, + 209.99999999999997 + ], + "spans": [], + "index": 6 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 222, + 505, + 267 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 223, + 504, + 235 + ], + "spans": [ + { + "bbox": [ + 105, + 223, + 280, + 235 + ], + "score": 1.0, + "content": "Figure 6: Optimization for smiling on", + "type": "text" + }, + { + "bbox": [ + 280, + 223, + 287, + 233 + ], + "score": 0.39, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 288, + 223, + 432, + 235 + ], + "score": 1.0, + "content": "-space and ϵ-space. (a) Optimizing", + "type": "text" + }, + { + "bbox": [ + 432, + 223, + 451, + 234 + ], + "score": 0.9, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 223, + 504, + 235 + ], + "score": 1.0, + "content": "for a sample", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 234, + 504, + 245 + ], + "spans": [ + { + "bbox": [ + 105, + 234, + 411, + 245 + ], + "score": 1.0, + "content": "with smiling results in natural editing while the change due to optimizing", + "type": "text" + }, + { + "bbox": [ + 411, + 234, + 429, + 244 + ], + "score": 0.91, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 429, + 234, + 504, + 245 + ], + "score": 1.0, + "content": "is relatively small.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 245, + 505, + 257 + ], + "spans": [ + { + "bbox": [ + 106, + 245, + 160, + 257 + ], + "score": 1.0, + "content": "(b) Applying", + "type": "text" + }, + { + "bbox": [ + 161, + 245, + 180, + 255 + ], + "score": 0.91, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 181, + 245, + 439, + 257 + ], + "score": 1.0, + "content": "from (a) to other samples yields the same attribute change while", + "type": "text" + }, + { + "bbox": [ + 439, + 245, + 457, + 255 + ], + "score": 0.9, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 458, + 245, + 505, + 257 + ], + "score": 1.0, + "content": "distorts the", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 256, + 344, + 267 + ], + "spans": [ + { + "bbox": [ + 105, + 256, + 193, + 267 + ], + "score": 1.0, + "content": "images. The result of", + "type": "text" + }, + { + "bbox": [ + 194, + 256, + 211, + 266 + ], + "score": 0.91, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 211, + 256, + 344, + 267 + ], + "score": 1.0, + "content": "is the best sample we could find.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 8.5 + } + ], + "index": 6.75 + }, + { + "type": "image", + "bbox": [ + 114, + 273, + 497, + 347 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 114, + 273, + 497, + 347 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 114, + 273, + 497, + 347 + ], + "spans": [ + { + "bbox": [ + 114, + 273, + 497, + 347 + ], + "score": 0.964, + "type": "image", + "image_path": "7bac52c03fecc44b4f929ac853df2555bd33134ce5593830fb81ee5e8c225924.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 114, + 273, + 497, + 297.6666666666667 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 114, + 297.6666666666667, + 497, + 322.33333333333337 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 114, + 322.33333333333337, + 497, + 347.00000000000006 + ], + "spans": [], + "index": 13 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 358, + 505, + 406 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 358, + 505, + 371 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 205, + 371 + ], + "score": 1.0, + "content": "Figure 7: Linearity of", + "type": "text" + }, + { + "bbox": [ + 206, + 359, + 213, + 369 + ], + "score": 0.58, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 213, + 358, + 415, + 371 + ], + "score": 1.0, + "content": "-space. Linear interpolation and extrapolation on", + "type": "text" + }, + { + "bbox": [ + 415, + 359, + 422, + 369 + ], + "score": 0.67, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 422, + 358, + 505, + 371 + ], + "score": 1.0, + "content": "-space lead to grad-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 370, + 505, + 382 + ], + "spans": [ + { + "bbox": [ + 106, + 370, + 505, + 382 + ], + "score": 1.0, + "content": "ual changes even to the unseen intensity and directions. Right side shows the interpolation results", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 104, + 378, + 507, + 396 + ], + "spans": [ + { + "bbox": [ + 104, + 378, + 197, + 396 + ], + "score": 1.0, + "content": "by positive scaling of", + "type": "text" + }, + { + "bbox": [ + 197, + 380, + 242, + 394 + ], + "score": 0.93, + "content": "\\Delta h _ { t } ^ { \\mathrm { s m i l i n g } }", + "type": "inline_equation" + }, + { + "bbox": [ + 242, + 378, + 507, + 396 + ], + "score": 1.0, + "content": ". Left side shows the extrapolation results by negative scaling of", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 107, + 391, + 376, + 408 + ], + "spans": [ + { + "bbox": [ + 107, + 393, + 150, + 407 + ], + "score": 0.92, + "content": "\\Delta h _ { t } ^ { s m i l i n g }", + "type": "inline_equation" + }, + { + "bbox": [ + 150, + 391, + 259, + 408 + ], + "score": 1.0, + "content": ". Note that we did not train", + "type": "text" + }, + { + "bbox": [ + 259, + 395, + 269, + 406 + ], + "score": 0.86, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 391, + 376, + 408 + ], + "score": 1.0, + "content": "for the negative direction.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 15.5 + } + ], + "index": 13.75 + }, + { + "type": "title", + "bbox": [ + 108, + 432, + 255, + 443 + ], + "lines": [ + { + "bbox": [ + 105, + 432, + 257, + 445 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 257, + 445 + ], + "score": 1.0, + "content": "5.2 QUANTITATIVE COMPARISON", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 106, + 452, + 505, + 552 + ], + "lines": [ + { + "bbox": [ + 105, + 451, + 506, + 466 + ], + "spans": [ + { + "bbox": [ + 105, + 451, + 506, + 466 + ], + "score": 1.0, + "content": "Considering that our method can be combined with various diffusion models without finetuning,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 464, + 504, + 475 + ], + "spans": [ + { + "bbox": [ + 106, + 464, + 504, + 475 + ], + "score": 1.0, + "content": "we do not find such a versatile competitor. Nonetheless, we compare Asyrp against DiffusionCLIP", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 474, + 505, + 487 + ], + "spans": [ + { + "bbox": [ + 105, + 474, + 505, + 487 + ], + "score": 1.0, + "content": "using the official code that edits the real images by finetuning the whole model. We asked 80 partic-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 485, + 505, + 498 + ], + "spans": [ + { + "bbox": [ + 106, + 485, + 505, + 498 + ], + "score": 1.0, + "content": "ipants to choose the images with better quality, natural attribute change, and overall preference for", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 496, + 505, + 509 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 505, + 509 + ], + "score": 1.0, + "content": "given total of 40 sets of original images, ours, and DiffusionCLIP. Table 1 shows that Asyrp outper-", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 506, + 504, + 520 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 504, + 520 + ], + "score": 1.0, + "content": "forms DiffusionCLIP in the all perspectives including the attributes unseen in the training dataset.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 516, + 505, + 532 + ], + "spans": [ + { + "bbox": [ + 105, + 516, + 505, + 532 + ], + "score": 1.0, + "content": "We list the settings for fair comparison including the questions and example images in Appendix", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 528, + 505, + 543 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 144, + 543 + ], + "score": 1.0, + "content": "K.1. See", + "type": "text" + }, + { + "bbox": [ + 144, + 529, + 169, + 541 + ], + "score": 0.62, + "content": "\\textrm { \\AA K } . 2", + "type": "inline_equation" + }, + { + "bbox": [ + 169, + 528, + 505, + 543 + ], + "score": 1.0, + "content": "for more evaluation metrics: segmentation consistency (SC) and directional CLIP", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 538, + 173, + 553 + ], + "spans": [ + { + "bbox": [ + 105, + 538, + 148, + 553 + ], + "score": 1.0, + "content": "similarity(", + "type": "text" + }, + { + "bbox": [ + 148, + 541, + 168, + 551 + ], + "score": 0.84, + "content": "S _ { \\mathrm { d i r } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 168, + 538, + 173, + 553 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 23, + "bbox_fs": [ + 105, + 451, + 506, + 553 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 565, + 224, + 577 + ], + "lines": [ + { + "bbox": [ + 105, + 563, + 226, + 579 + ], + "spans": [ + { + "bbox": [ + 105, + 563, + 191, + 579 + ], + "score": 1.0, + "content": "5.3 ANALYSIS ON", + "type": "text" + }, + { + "bbox": [ + 191, + 566, + 198, + 575 + ], + "score": 0.38, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 198, + 563, + 226, + 579 + ], + "score": 1.0, + "content": "-space", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 106, + 586, + 503, + 608 + ], + "lines": [ + { + "bbox": [ + 106, + 585, + 505, + 598 + ], + "spans": [ + { + "bbox": [ + 106, + 585, + 505, + 598 + ], + "score": 1.0, + "content": "We provide detailed analyses to validate the properties of semantic latent space for diffusion models:", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 596, + 383, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 596, + 383, + 609 + ], + "score": 1.0, + "content": "homogeneity, linearity, robustness, and consistency across timesteps.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5, + "bbox_fs": [ + 105, + 585, + 505, + 609 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 620, + 504, + 664 + ], + "lines": [ + { + "bbox": [ + 106, + 621, + 504, + 632 + ], + "spans": [ + { + "bbox": [ + 106, + 621, + 319, + 632 + ], + "score": 1.0, + "content": "Homogeneity. Figure 6 illustrates homogeneity of", + "type": "text" + }, + { + "bbox": [ + 320, + 621, + 326, + 630 + ], + "score": 0.75, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 326, + 621, + 406, + 632 + ], + "score": 1.0, + "content": "-space compared to", + "type": "text" + }, + { + "bbox": [ + 407, + 623, + 413, + 630 + ], + "score": 0.61, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 414, + 621, + 462, + 632 + ], + "score": 1.0, + "content": "-space. One", + "type": "text" + }, + { + "bbox": [ + 462, + 621, + 482, + 631 + ], + "score": 0.9, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 482, + 621, + 504, + 632 + ], + "score": 1.0, + "content": "opti-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 631, + 505, + 643 + ], + "spans": [ + { + "bbox": [ + 106, + 631, + 505, + 643 + ], + "score": 1.0, + "content": "mized for an image results in the same attribute change to other input images. On the other hand,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 104, + 641, + 504, + 655 + ], + "spans": [ + { + "bbox": [ + 104, + 641, + 124, + 655 + ], + "score": 1.0, + "content": "one", + "type": "text" + }, + { + "bbox": [ + 125, + 642, + 142, + 653 + ], + "score": 0.87, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 143, + 641, + 456, + 655 + ], + "score": 1.0, + "content": "optimized for an image distorts other input images. In Figure 10, applying", + "type": "text" + }, + { + "bbox": [ + 456, + 642, + 504, + 654 + ], + "score": 0.91, + "content": "\\Delta h _ { t } ^ { \\mathrm { m e a n } } =", + "type": "inline_equation" + } + ], + "index": 33 + }, + { + "bbox": [ + 107, + 652, + 502, + 668 + ], + "spans": [ + { + "bbox": [ + 107, + 653, + 149, + 666 + ], + "score": 0.91, + "content": "\\begin{array} { r } { \\frac { 1 } { N } \\sum \\Delta h _ { t } ^ { i } } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 150, + 652, + 310, + 668 + ], + "score": 1.0, + "content": "produces almost identical results where", + "type": "text" + }, + { + "bbox": [ + 311, + 654, + 315, + 663 + ], + "score": 0.72, + "content": "i", + "type": "inline_equation" + }, + { + "bbox": [ + 316, + 652, + 396, + 668 + ], + "score": 1.0, + "content": "indicates indices of", + "type": "text" + }, + { + "bbox": [ + 396, + 654, + 430, + 663 + ], + "score": 0.9, + "content": "N = 2 0", + "type": "inline_equation" + }, + { + "bbox": [ + 430, + 652, + 502, + 668 + ], + "score": 1.0, + "content": "random samples.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 32.5, + "bbox_fs": [ + 104, + 621, + 505, + 668 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 676, + 504, + 731 + ], + "lines": [ + { + "bbox": [ + 106, + 677, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 352, + 689 + ], + "score": 1.0, + "content": "Linearity. In Figure 7, we observe that linearly scaling a", + "type": "text" + }, + { + "bbox": [ + 352, + 677, + 368, + 687 + ], + "score": 0.76, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 368, + 677, + 505, + 689 + ], + "score": 1.0, + "content": "reflects the amount of change in", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "score": 1.0, + "content": "the visual attributes. Surprisingly, it generalizes to negative scales that are not seen during train-", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 356, + 711 + ], + "score": 1.0, + "content": "ing. Moreover, Figure 8 shows that combinations of different", + "type": "text" + }, + { + "bbox": [ + 356, + 699, + 372, + 709 + ], + "score": 0.77, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 372, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "’s yield their combined semantic", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "changes in the resulting images. Appendix N.2 provides mixed interpolation between multiple at-", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 720, + 142, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 142, + 734 + ], + "score": 1.0, + "content": "tributes.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 37, + "bbox_fs": [ + 105, + 677, + 505, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 115, + 80, + 496, + 139 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 80, + 496, + 139 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 115, + 80, + 496, + 139 + ], + "spans": [ + { + "bbox": [ + 115, + 80, + 496, + 139 + ], + "score": 0.955, + "type": "image", + "image_path": "ec4e0c1c15035d0be274617d3f54f6a61905cce32ac51db848c37eaa6738a5c4.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 115, + 80, + 496, + 99.66666666666667 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 115, + 99.66666666666667, + 496, + 119.33333333333334 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 115, + 119.33333333333334, + 496, + 139.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 112, + 150, + 494, + 163 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 112, + 149, + 496, + 164 + ], + "spans": [ + { + "bbox": [ + 112, + 149, + 496, + 164 + ], + "score": 1.0, + "content": "Figure 8: Linear combination. Combining multiple ∆hs leads to combined attribute changes.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 3 + } + ], + "index": 2.0 + }, + { + "type": "image", + "bbox": [ + 115, + 169, + 494, + 230 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 169, + 494, + 230 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 115, + 169, + 494, + 230 + ], + "spans": [ + { + "bbox": [ + 115, + 169, + 494, + 230 + ], + "score": 0.963, + "type": "image", + "image_path": "f16b3b6406f8475789e2643d74a9891b20a10d09c81ef00ec5fdfc72bd1f3e33.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 115, + 169, + 494, + 189.33333333333334 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 115, + 189.33333333333334, + 494, + 209.66666666666669 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 115, + 209.66666666666669, + 494, + 230.00000000000003 + ], + "spans": [], + "index": 6 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 233, + 505, + 291 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 232, + 504, + 247 + ], + "spans": [ + { + "bbox": [ + 105, + 232, + 265, + 247 + ], + "score": 1.0, + "content": "Figure 9: Ramdom manipulation on", + "type": "text" + }, + { + "bbox": [ + 265, + 233, + 272, + 243 + ], + "score": 0.37, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 273, + 232, + 319, + 247 + ], + "score": 1.0, + "content": "-space and", + "type": "text" + }, + { + "bbox": [ + 320, + 235, + 325, + 243 + ], + "score": 0.61, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 326, + 232, + 504, + 247 + ], + "score": 1.0, + "content": "-space. (a) Adding random noise with mag-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 104, + 242, + 504, + 259 + ], + "spans": [ + { + "bbox": [ + 104, + 242, + 195, + 259 + ], + "score": 1.0, + "content": "nitude of ∆hsmilingt", + "type": "text" + }, + { + "bbox": [ + 191, + 244, + 290, + 258 + ], + "score": 1.0, + "content": "and random direction in", + "type": "text" + }, + { + "bbox": [ + 291, + 246, + 297, + 255 + ], + "score": 0.54, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 298, + 244, + 504, + 258 + ], + "score": 1.0, + "content": "-space leads to realistic images with small changes.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 103, + 254, + 507, + 273 + ], + "spans": [ + { + "bbox": [ + 103, + 254, + 290, + 273 + ], + "score": 1.0, + "content": "(b) Adding random noise with magnitude of", + "type": "text" + }, + { + "bbox": [ + 290, + 256, + 335, + 270 + ], + "score": 0.92, + "content": "\\Delta \\epsilon _ { t } ^ { \\mathrm { s m i l i n g } }", + "type": "inline_equation" + }, + { + "bbox": [ + 335, + 254, + 437, + 273 + ], + "score": 1.0, + "content": "and random direction in", + "type": "text" + }, + { + "bbox": [ + 437, + 260, + 444, + 268 + ], + "score": 0.5, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 444, + 254, + 507, + 273 + ], + "score": 1.0, + "content": "-space leads to", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 268, + 505, + 281 + ], + "spans": [ + { + "bbox": [ + 105, + 268, + 505, + 281 + ], + "score": 1.0, + "content": "severely distorted images. (c) Adding random noises with tripled magnitude produce diverse visual", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 280, + 218, + 293 + ], + "spans": [ + { + "bbox": [ + 106, + 280, + 218, + 293 + ], + "score": 1.0, + "content": "changes in realistic images.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 9 + } + ], + "index": 7.0 + }, + { + "type": "image", + "bbox": [ + 115, + 297, + 494, + 368 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 297, + 494, + 368 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 115, + 297, + 494, + 368 + ], + "spans": [ + { + "bbox": [ + 115, + 297, + 494, + 368 + ], + "score": 0.954, + "type": "image", + "image_path": "516dc2690d986f75d0a76feebed8568de3bb5e07882b27922278b8419f1f3e97.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 115, + 297, + 494, + 320.6666666666667 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 115, + 320.6666666666667, + 494, + 344.33333333333337 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 115, + 344.33333333333337, + 494, + 368.00000000000006 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 370, + 504, + 404 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 105, + 367, + 507, + 384 + ], + "spans": [ + { + "bbox": [ + 105, + 367, + 217, + 384 + ], + "score": 1.0, + "content": "Figure 10: Consistency on", + "type": "text" + }, + { + "bbox": [ + 218, + 370, + 225, + 380 + ], + "score": 0.39, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 367, + 297, + 384 + ], + "score": 1.0, + "content": "-space. Results of", + "type": "text" + }, + { + "bbox": [ + 297, + 370, + 316, + 381 + ], + "score": 0.85, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 316, + 367, + 320, + 384 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 320, + 370, + 355, + 382 + ], + "score": 0.88, + "content": "\\Delta h _ { t } ^ { \\mathrm { m e a n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 355, + 367, + 372, + 384 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 373, + 369, + 410, + 380 + ], + "score": 0.86, + "content": "\\Delta h ^ { \\mathrm { g l o b a l } }", + "type": "inline_equation" + }, + { + "bbox": [ + 410, + 367, + 507, + 384 + ], + "score": 1.0, + "content": "are almost identical for", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 381, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 106, + 381, + 271, + 394 + ], + "score": 1.0, + "content": "in-domain samples. However, we choose", + "type": "text" + }, + { + "bbox": [ + 271, + 381, + 291, + 392 + ], + "score": 0.9, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 291, + 381, + 506, + 394 + ], + "score": 1.0, + "content": "over others to prevent unexpected small difference in", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 392, + 322, + 405 + ], + "spans": [ + { + "bbox": [ + 106, + 392, + 296, + 405 + ], + "score": 1.0, + "content": "the unseen domain. We provide more results in", + "type": "text" + }, + { + "bbox": [ + 296, + 392, + 317, + 404 + ], + "score": 0.74, + "content": "\\ S \\mathrm { { L . 1 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 318, + 392, + 322, + 405 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16 + } + ], + "index": 14.5 + }, + { + "type": "text", + "bbox": [ + 107, + 420, + 505, + 475 + ], + "lines": [ + { + "bbox": [ + 106, + 420, + 505, + 432 + ], + "spans": [ + { + "bbox": [ + 106, + 420, + 398, + 432 + ], + "score": 1.0, + "content": "Robustness. Figure 9 compares the effect of adding random noise in", + "type": "text" + }, + { + "bbox": [ + 399, + 421, + 405, + 430 + ], + "score": 0.68, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 406, + 420, + 451, + 432 + ], + "score": 1.0, + "content": "-space and", + "type": "text" + }, + { + "bbox": [ + 451, + 422, + 457, + 430 + ], + "score": 0.46, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 457, + 420, + 505, + 432 + ], + "score": 1.0, + "content": "-space. The", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 431, + 505, + 443 + ], + "spans": [ + { + "bbox": [ + 106, + 431, + 505, + 443 + ], + "score": 1.0, + "content": "random noises are chosen to be the vectors with random directions and magnitude of the example", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 107, + 442, + 506, + 455 + ], + "spans": [ + { + "bbox": [ + 107, + 442, + 126, + 453 + ], + "score": 0.89, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 127, + 442, + 146, + 455 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 146, + 442, + 164, + 453 + ], + "score": 0.9, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 164, + 442, + 341, + 455 + ], + "score": 1.0, + "content": "in Figure 6 on each space. Perturbation in", + "type": "text" + }, + { + "bbox": [ + 341, + 443, + 348, + 452 + ], + "score": 0.74, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 348, + 442, + 506, + 455 + ], + "score": 1.0, + "content": "-space leads to realistic images with a", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 452, + 505, + 466 + ], + "spans": [ + { + "bbox": [ + 105, + 452, + 435, + 466 + ], + "score": 1.0, + "content": "minimal difference or some semantic changes. On the contrary, perturbation in", + "type": "text" + }, + { + "bbox": [ + 436, + 455, + 442, + 463 + ], + "score": 0.66, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 442, + 452, + 505, + 466 + ], + "score": 1.0, + "content": "-space severely", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 464, + 376, + 477 + ], + "spans": [ + { + "bbox": [ + 106, + 464, + 376, + 477 + ], + "score": 1.0, + "content": "distorts the resulting images. See Appendix D.2 for more analyses.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 488, + 505, + 557 + ], + "lines": [ + { + "bbox": [ + 106, + 487, + 505, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 487, + 289, + 502 + ], + "score": 1.0, + "content": "Consistency across timesteps. Recall that", + "type": "text" + }, + { + "bbox": [ + 289, + 489, + 309, + 500 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 487, + 505, + 502 + ], + "score": 1.0, + "content": "for all samples are homogeneous and replacing", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 102, + 496, + 492, + 529 + ], + "spans": [ + { + "bbox": [ + 102, + 496, + 173, + 529 + ], + "score": 1.0, + "content": "them by their mea time-invariant", + "type": "text" + }, + { + "bbox": [ + 184, + 500, + 219, + 510 + ], + "score": 0.86, + "content": "\\Delta h _ { t } ^ { \\mathrm { m e a n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 496, + 335, + 529 + ], + "score": 1.0, + "content": "ults. Interesinstead of", + "type": "text" + }, + { + "bbox": [ + 356, + 496, + 492, + 529 + ], + "score": 1.0, + "content": ", in Figure 10, we observe that addalso yields similar results where", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 173, + 510, + 504, + 524 + ], + "spans": [ + { + "bbox": [ + 173, + 510, + 290, + 524 + ], + "score": 0.91, + "content": "\\Delta { h ^ { \\mathrm { g l o b a l } } } ~ \\dot { = } ~ \\frac { 1 } { T _ { e } } \\sum _ { t } \\Delta { h _ { t } ^ { \\mathrm { m e a n } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 336, + 511, + 355, + 522 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 492, + 511, + 504, + 522 + ], + "score": 0.86, + "content": "T _ { e }", + "type": "inline_equation" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 523, + 505, + 537 + ], + "spans": [ + { + "bbox": [ + 105, + 523, + 273, + 537 + ], + "score": 1.0, + "content": "denotes the length of the editing interval", + "type": "text" + }, + { + "bbox": [ + 274, + 523, + 309, + 535 + ], + "score": 0.88, + "content": "[ T , t _ { \\mathrm { e d i t } } ]", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 523, + 378, + 537 + ], + "score": 1.0, + "content": ". Though we use", + "type": "text" + }, + { + "bbox": [ + 378, + 523, + 398, + 534 + ], + "score": 0.91, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 398, + 523, + 505, + 537 + ], + "score": 1.0, + "content": "to deliver the best quality", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 103, + 529, + 507, + 550 + ], + "spans": [ + { + "bbox": [ + 103, + 529, + 207, + 550 + ], + "score": 1.0, + "content": "and manipulation, using", + "type": "text" + }, + { + "bbox": [ + 207, + 534, + 241, + 546 + ], + "score": 0.91, + "content": "\\Delta h _ { t } ^ { \\mathrm { m e a n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 242, + 529, + 276, + 550 + ], + "score": 1.0, + "content": "or even", + "type": "text" + }, + { + "bbox": [ + 276, + 535, + 314, + 545 + ], + "score": 0.85, + "content": "\\Delta h ^ { \\mathrm { g l o b a l } }", + "type": "inline_equation" + }, + { + "bbox": [ + 314, + 529, + 507, + 550 + ], + "score": 1.0, + "content": "with some compromise would be worth trying", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 546, + 456, + 558 + ], + "spans": [ + { + "bbox": [ + 106, + 546, + 456, + 558 + ], + "score": 1.0, + "content": "for simplicity. We report more detail about mean and global direction in Appendix L.1.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 25.5 + }, + { + "type": "title", + "bbox": [ + 107, + 574, + 195, + 586 + ], + "lines": [ + { + "bbox": [ + 105, + 572, + 197, + 590 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 197, + 590 + ], + "score": 1.0, + "content": "6 CONCLUSION", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 599, + 505, + 677 + ], + "lines": [ + { + "bbox": [ + 105, + 599, + 506, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 599, + 506, + 613 + ], + "score": 1.0, + "content": "We proposed a new generative process, Asyrp, which facilitates image editing in a semantic latent", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 132, + 624 + ], + "score": 1.0, + "content": "space", + "type": "text" + }, + { + "bbox": [ + 132, + 611, + 138, + 621 + ], + "score": 0.68, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 139, + 610, + 300, + 624 + ], + "score": 1.0, + "content": "-space for pretrained diffusion models.", + "type": "text" + }, + { + "bbox": [ + 300, + 611, + 307, + 621 + ], + "score": 0.62, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 307, + 610, + 506, + 624 + ], + "score": 1.0, + "content": "-space has nice properties as in the latent space", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 621, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 505, + 635 + ], + "score": 1.0, + "content": "of GANs: homogeneity, linearity, robustness, and consistency across timesteps. The full editing", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 632, + 505, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 505, + 646 + ], + "score": 1.0, + "content": "process is designed to achieve versatile editing and high quality by measuring editing strength and", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "score": 1.0, + "content": "quality deficiency at timesteps. We hope that our approach and detailed analyses help cultivate a new", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 654, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 505, + 667 + ], + "score": 1.0, + "content": "paradigm of image editing in the semantic latent space of diffusion models. Combining previous", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 666, + 415, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 415, + 678 + ], + "score": 1.0, + "content": "finetuning or guidance techniques would be an interesting research direction.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 33 + }, + { + "type": "title", + "bbox": [ + 108, + 691, + 200, + 701 + ], + "lines": [ + { + "bbox": [ + 107, + 691, + 200, + 702 + ], + "spans": [ + { + "bbox": [ + 107, + 691, + 200, + 702 + ], + "score": 1.0, + "content": "ACKNOWLEDGMENTS", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 108, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "This work was supported by the National Research Foundation of Korea (NRF) grant funded by the", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 720, + 388, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 388, + 733 + ], + "score": 1.0, + "content": "Korea government (Ministry of Science and ICT) (No. 2021-0-00155)", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5 + } + ], + "page_idx": 8, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "score": 1.0, + "content": "9", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 115, + 80, + 496, + 139 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 80, + 496, + 139 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 115, + 80, + 496, + 139 + ], + "spans": [ + { + "bbox": [ + 115, + 80, + 496, + 139 + ], + "score": 0.955, + "type": "image", + "image_path": "ec4e0c1c15035d0be274617d3f54f6a61905cce32ac51db848c37eaa6738a5c4.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 115, + 80, + 496, + 99.66666666666667 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 115, + 99.66666666666667, + 496, + 119.33333333333334 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 115, + 119.33333333333334, + 496, + 139.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 112, + 150, + 494, + 163 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 112, + 149, + 496, + 164 + ], + "spans": [ + { + "bbox": [ + 112, + 149, + 496, + 164 + ], + "score": 1.0, + "content": "Figure 8: Linear combination. Combining multiple ∆hs leads to combined attribute changes.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 3 + } + ], + "index": 2.0 + }, + { + "type": "image", + "bbox": [ + 115, + 169, + 494, + 230 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 169, + 494, + 230 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 115, + 169, + 494, + 230 + ], + "spans": [ + { + "bbox": [ + 115, + 169, + 494, + 230 + ], + "score": 0.963, + "type": "image", + "image_path": "f16b3b6406f8475789e2643d74a9891b20a10d09c81ef00ec5fdfc72bd1f3e33.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 115, + 169, + 494, + 189.33333333333334 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 115, + 189.33333333333334, + 494, + 209.66666666666669 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 115, + 209.66666666666669, + 494, + 230.00000000000003 + ], + "spans": [], + "index": 6 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 233, + 505, + 291 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 232, + 504, + 247 + ], + "spans": [ + { + "bbox": [ + 105, + 232, + 265, + 247 + ], + "score": 1.0, + "content": "Figure 9: Ramdom manipulation on", + "type": "text" + }, + { + "bbox": [ + 265, + 233, + 272, + 243 + ], + "score": 0.37, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 273, + 232, + 319, + 247 + ], + "score": 1.0, + "content": "-space and", + "type": "text" + }, + { + "bbox": [ + 320, + 235, + 325, + 243 + ], + "score": 0.61, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 326, + 232, + 504, + 247 + ], + "score": 1.0, + "content": "-space. (a) Adding random noise with mag-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 104, + 242, + 504, + 259 + ], + "spans": [ + { + "bbox": [ + 104, + 242, + 195, + 259 + ], + "score": 1.0, + "content": "nitude of ∆hsmilingt", + "type": "text" + }, + { + "bbox": [ + 191, + 244, + 290, + 258 + ], + "score": 1.0, + "content": "and random direction in", + "type": "text" + }, + { + "bbox": [ + 291, + 246, + 297, + 255 + ], + "score": 0.54, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 298, + 244, + 504, + 258 + ], + "score": 1.0, + "content": "-space leads to realistic images with small changes.", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 103, + 254, + 507, + 273 + ], + "spans": [ + { + "bbox": [ + 103, + 254, + 290, + 273 + ], + "score": 1.0, + "content": "(b) Adding random noise with magnitude of", + "type": "text" + }, + { + "bbox": [ + 290, + 256, + 335, + 270 + ], + "score": 0.92, + "content": "\\Delta \\epsilon _ { t } ^ { \\mathrm { s m i l i n g } }", + "type": "inline_equation" + }, + { + "bbox": [ + 335, + 254, + 437, + 273 + ], + "score": 1.0, + "content": "and random direction in", + "type": "text" + }, + { + "bbox": [ + 437, + 260, + 444, + 268 + ], + "score": 0.5, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 444, + 254, + 507, + 273 + ], + "score": 1.0, + "content": "-space leads to", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 268, + 505, + 281 + ], + "spans": [ + { + "bbox": [ + 105, + 268, + 505, + 281 + ], + "score": 1.0, + "content": "severely distorted images. (c) Adding random noises with tripled magnitude produce diverse visual", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 280, + 218, + 293 + ], + "spans": [ + { + "bbox": [ + 106, + 280, + 218, + 293 + ], + "score": 1.0, + "content": "changes in realistic images.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 9 + } + ], + "index": 7.0 + }, + { + "type": "image", + "bbox": [ + 115, + 297, + 494, + 368 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 297, + 494, + 368 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 115, + 297, + 494, + 368 + ], + "spans": [ + { + "bbox": [ + 115, + 297, + 494, + 368 + ], + "score": 0.954, + "type": "image", + "image_path": "516dc2690d986f75d0a76feebed8568de3bb5e07882b27922278b8419f1f3e97.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 115, + 297, + 494, + 320.6666666666667 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 115, + 320.6666666666667, + 494, + 344.33333333333337 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 115, + 344.33333333333337, + 494, + 368.00000000000006 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 370, + 504, + 404 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 105, + 367, + 507, + 384 + ], + "spans": [ + { + "bbox": [ + 105, + 367, + 217, + 384 + ], + "score": 1.0, + "content": "Figure 10: Consistency on", + "type": "text" + }, + { + "bbox": [ + 218, + 370, + 225, + 380 + ], + "score": 0.39, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 367, + 297, + 384 + ], + "score": 1.0, + "content": "-space. Results of", + "type": "text" + }, + { + "bbox": [ + 297, + 370, + 316, + 381 + ], + "score": 0.85, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 316, + 367, + 320, + 384 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 320, + 370, + 355, + 382 + ], + "score": 0.88, + "content": "\\Delta h _ { t } ^ { \\mathrm { m e a n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 355, + 367, + 372, + 384 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 373, + 369, + 410, + 380 + ], + "score": 0.86, + "content": "\\Delta h ^ { \\mathrm { g l o b a l } }", + "type": "inline_equation" + }, + { + "bbox": [ + 410, + 367, + 507, + 384 + ], + "score": 1.0, + "content": "are almost identical for", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 381, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 106, + 381, + 271, + 394 + ], + "score": 1.0, + "content": "in-domain samples. However, we choose", + "type": "text" + }, + { + "bbox": [ + 271, + 381, + 291, + 392 + ], + "score": 0.9, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 291, + 381, + 506, + 394 + ], + "score": 1.0, + "content": "over others to prevent unexpected small difference in", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 392, + 322, + 405 + ], + "spans": [ + { + "bbox": [ + 106, + 392, + 296, + 405 + ], + "score": 1.0, + "content": "the unseen domain. We provide more results in", + "type": "text" + }, + { + "bbox": [ + 296, + 392, + 317, + 404 + ], + "score": 0.74, + "content": "\\ S \\mathrm { { L . 1 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 318, + 392, + 322, + 405 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16 + } + ], + "index": 14.5 + }, + { + "type": "text", + "bbox": [ + 107, + 420, + 505, + 475 + ], + "lines": [ + { + "bbox": [ + 106, + 420, + 505, + 432 + ], + "spans": [ + { + "bbox": [ + 106, + 420, + 398, + 432 + ], + "score": 1.0, + "content": "Robustness. Figure 9 compares the effect of adding random noise in", + "type": "text" + }, + { + "bbox": [ + 399, + 421, + 405, + 430 + ], + "score": 0.68, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 406, + 420, + 451, + 432 + ], + "score": 1.0, + "content": "-space and", + "type": "text" + }, + { + "bbox": [ + 451, + 422, + 457, + 430 + ], + "score": 0.46, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 457, + 420, + 505, + 432 + ], + "score": 1.0, + "content": "-space. The", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 431, + 505, + 443 + ], + "spans": [ + { + "bbox": [ + 106, + 431, + 505, + 443 + ], + "score": 1.0, + "content": "random noises are chosen to be the vectors with random directions and magnitude of the example", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 107, + 442, + 506, + 455 + ], + "spans": [ + { + "bbox": [ + 107, + 442, + 126, + 453 + ], + "score": 0.89, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 127, + 442, + 146, + 455 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 146, + 442, + 164, + 453 + ], + "score": 0.9, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 164, + 442, + 341, + 455 + ], + "score": 1.0, + "content": "in Figure 6 on each space. Perturbation in", + "type": "text" + }, + { + "bbox": [ + 341, + 443, + 348, + 452 + ], + "score": 0.74, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 348, + 442, + 506, + 455 + ], + "score": 1.0, + "content": "-space leads to realistic images with a", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 452, + 505, + 466 + ], + "spans": [ + { + "bbox": [ + 105, + 452, + 435, + 466 + ], + "score": 1.0, + "content": "minimal difference or some semantic changes. On the contrary, perturbation in", + "type": "text" + }, + { + "bbox": [ + 436, + 455, + 442, + 463 + ], + "score": 0.66, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 442, + 452, + 505, + 466 + ], + "score": 1.0, + "content": "-space severely", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 464, + 376, + 477 + ], + "spans": [ + { + "bbox": [ + 106, + 464, + 376, + 477 + ], + "score": 1.0, + "content": "distorts the resulting images. See Appendix D.2 for more analyses.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 20, + "bbox_fs": [ + 105, + 420, + 506, + 477 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 488, + 505, + 557 + ], + "lines": [ + { + "bbox": [ + 106, + 487, + 505, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 487, + 289, + 502 + ], + "score": 1.0, + "content": "Consistency across timesteps. Recall that", + "type": "text" + }, + { + "bbox": [ + 289, + 489, + 309, + 500 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 487, + 505, + 502 + ], + "score": 1.0, + "content": "for all samples are homogeneous and replacing", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 102, + 496, + 492, + 529 + ], + "spans": [ + { + "bbox": [ + 102, + 496, + 173, + 529 + ], + "score": 1.0, + "content": "them by their mea time-invariant", + "type": "text" + }, + { + "bbox": [ + 184, + 500, + 219, + 510 + ], + "score": 0.86, + "content": "\\Delta h _ { t } ^ { \\mathrm { m e a n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 496, + 335, + 529 + ], + "score": 1.0, + "content": "ults. Interesinstead of", + "type": "text" + }, + { + "bbox": [ + 356, + 496, + 492, + 529 + ], + "score": 1.0, + "content": ", in Figure 10, we observe that addalso yields similar results where", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 173, + 510, + 504, + 524 + ], + "spans": [ + { + "bbox": [ + 173, + 510, + 290, + 524 + ], + "score": 0.91, + "content": "\\Delta { h ^ { \\mathrm { g l o b a l } } } ~ \\dot { = } ~ \\frac { 1 } { T _ { e } } \\sum _ { t } \\Delta { h _ { t } ^ { \\mathrm { m e a n } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 336, + 511, + 355, + 522 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 492, + 511, + 504, + 522 + ], + "score": 0.86, + "content": "T _ { e }", + "type": "inline_equation" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 523, + 505, + 537 + ], + "spans": [ + { + "bbox": [ + 105, + 523, + 273, + 537 + ], + "score": 1.0, + "content": "denotes the length of the editing interval", + "type": "text" + }, + { + "bbox": [ + 274, + 523, + 309, + 535 + ], + "score": 0.88, + "content": "[ T , t _ { \\mathrm { e d i t } } ]", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 523, + 378, + 537 + ], + "score": 1.0, + "content": ". Though we use", + "type": "text" + }, + { + "bbox": [ + 378, + 523, + 398, + 534 + ], + "score": 0.91, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 398, + 523, + 505, + 537 + ], + "score": 1.0, + "content": "to deliver the best quality", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 103, + 529, + 507, + 550 + ], + "spans": [ + { + "bbox": [ + 103, + 529, + 207, + 550 + ], + "score": 1.0, + "content": "and manipulation, using", + "type": "text" + }, + { + "bbox": [ + 207, + 534, + 241, + 546 + ], + "score": 0.91, + "content": "\\Delta h _ { t } ^ { \\mathrm { m e a n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 242, + 529, + 276, + 550 + ], + "score": 1.0, + "content": "or even", + "type": "text" + }, + { + "bbox": [ + 276, + 535, + 314, + 545 + ], + "score": 0.85, + "content": "\\Delta h ^ { \\mathrm { g l o b a l } }", + "type": "inline_equation" + }, + { + "bbox": [ + 314, + 529, + 507, + 550 + ], + "score": 1.0, + "content": "with some compromise would be worth trying", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 546, + 456, + 558 + ], + "spans": [ + { + "bbox": [ + 106, + 546, + 456, + 558 + ], + "score": 1.0, + "content": "for simplicity. We report more detail about mean and global direction in Appendix L.1.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 25.5, + "bbox_fs": [ + 102, + 487, + 507, + 558 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 574, + 195, + 586 + ], + "lines": [ + { + "bbox": [ + 105, + 572, + 197, + 590 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 197, + 590 + ], + "score": 1.0, + "content": "6 CONCLUSION", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 599, + 505, + 677 + ], + "lines": [ + { + "bbox": [ + 105, + 599, + 506, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 599, + 506, + 613 + ], + "score": 1.0, + "content": "We proposed a new generative process, Asyrp, which facilitates image editing in a semantic latent", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 132, + 624 + ], + "score": 1.0, + "content": "space", + "type": "text" + }, + { + "bbox": [ + 132, + 611, + 138, + 621 + ], + "score": 0.68, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 139, + 610, + 300, + 624 + ], + "score": 1.0, + "content": "-space for pretrained diffusion models.", + "type": "text" + }, + { + "bbox": [ + 300, + 611, + 307, + 621 + ], + "score": 0.62, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 307, + 610, + 506, + 624 + ], + "score": 1.0, + "content": "-space has nice properties as in the latent space", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 621, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 505, + 635 + ], + "score": 1.0, + "content": "of GANs: homogeneity, linearity, robustness, and consistency across timesteps. The full editing", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 632, + 505, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 505, + 646 + ], + "score": 1.0, + "content": "process is designed to achieve versatile editing and high quality by measuring editing strength and", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "score": 1.0, + "content": "quality deficiency at timesteps. We hope that our approach and detailed analyses help cultivate a new", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 654, + 505, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 505, + 667 + ], + "score": 1.0, + "content": "paradigm of image editing in the semantic latent space of diffusion models. Combining previous", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 666, + 415, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 415, + 678 + ], + "score": 1.0, + "content": "finetuning or guidance techniques would be an interesting research direction.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 33, + "bbox_fs": [ + 105, + 599, + 506, + 678 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 691, + 200, + 701 + ], + "lines": [ + { + "bbox": [ + 107, + 691, + 200, + 702 + ], + "spans": [ + { + "bbox": [ + 107, + 691, + 200, + 702 + ], + "score": 1.0, + "content": "ACKNOWLEDGMENTS", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 108, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "This work was supported by the National Research Foundation of Korea (NRF) grant funded by the", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 720, + 388, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 388, + 733 + ], + "score": 1.0, + "content": "Korea government (Ministry of Science and ICT) (No. 2021-0-00155)", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5, + "bbox_fs": [ + 106, + 709, + 505, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 82, + 176, + 93 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 176, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 176, + 95 + ], + "score": 1.0, + "content": "REFERENCES", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 100, + 504, + 133 + ], + "lines": [ + { + "bbox": [ + 106, + 99, + 506, + 113 + ], + "spans": [ + { + "bbox": [ + 106, + 99, + 506, + 113 + ], + "score": 1.0, + "content": "Omri Avrahami, Dani Lischinski, and Ohad Fried. Blended diffusion for text-driven editing of", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 116, + 111, + 506, + 123 + ], + "spans": [ + { + "bbox": [ + 116, + 111, + 506, + 123 + ], + "score": 1.0, + "content": "natural images. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 116, + 122, + 269, + 134 + ], + "spans": [ + { + "bbox": [ + 116, + 122, + 269, + 134 + ], + "score": 1.0, + "content": "Recognition, pp. 18208–18218, 2022.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2 + }, + { + "type": "text", + "bbox": [ + 105, + 140, + 503, + 163 + ], + "lines": [ + { + "bbox": [ + 105, + 139, + 504, + 153 + ], + "spans": [ + { + "bbox": [ + 105, + 139, + 504, + 153 + ], + "score": 1.0, + "content": "Hila Chefer, Sagie Benaim, Roni Paiss, and Lior Wolf. Image-based clip-guided essence transfer.", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 116, + 151, + 278, + 163 + ], + "spans": [ + { + "bbox": [ + 116, + 151, + 278, + 163 + ], + "score": 1.0, + "content": "arXiv preprint arXiv:2110.12427, 2021.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4.5 + }, + { + "type": "text", + "bbox": [ + 106, + 169, + 504, + 203 + ], + "lines": [ + { + "bbox": [ + 105, + 168, + 504, + 183 + ], + "spans": [ + { + "bbox": [ + 105, + 168, + 504, + 183 + ], + "score": 1.0, + "content": "Jooyoung Choi, Sungwon Kim, Yonghyun Jeong, Youngjune Gwon, and Sungroh Yoon. Ilvr: Con-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 115, + 180, + 505, + 194 + ], + "spans": [ + { + "bbox": [ + 115, + 180, + 505, + 194 + ], + "score": 1.0, + "content": "ditioning method for denoising diffusion probabilistic models. arXiv preprint arXiv:2108.02938,", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 115, + 191, + 144, + 204 + ], + "spans": [ + { + "bbox": [ + 115, + 191, + 144, + 204 + ], + "score": 1.0, + "content": "2021.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 106, + 210, + 505, + 244 + ], + "lines": [ + { + "bbox": [ + 106, + 210, + 504, + 223 + ], + "spans": [ + { + "bbox": [ + 106, + 210, + 504, + 223 + ], + "score": 1.0, + "content": "Jooyoung Choi, Jungbeom Lee, Chaehun Shin, Sungwon Kim, Hyunwoo Kim, and Sungroh Yoon.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 115, + 220, + 504, + 234 + ], + "spans": [ + { + "bbox": [ + 115, + 220, + 504, + 234 + ], + "score": 1.0, + "content": "Perception prioritized training of diffusion models. In Proceedings of the IEEE/CVF Conference", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 115, + 231, + 400, + 245 + ], + "spans": [ + { + "bbox": [ + 115, + 231, + 400, + 245 + ], + "score": 1.0, + "content": "on Computer Vision and Pattern Recognition, pp. 11472–11481, 2022.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + }, + { + "type": "text", + "bbox": [ + 108, + 250, + 504, + 284 + ], + "lines": [ + { + "bbox": [ + 106, + 250, + 505, + 263 + ], + "spans": [ + { + "bbox": [ + 106, + 250, + 505, + 263 + ], + "score": 1.0, + "content": "Yunjey Choi, Youngjung Uh, Jaejun Yoo, and Jung-Woo Ha. Stargan v2: Diverse image synthesis for", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 115, + 261, + 505, + 275 + ], + "spans": [ + { + "bbox": [ + 115, + 261, + 505, + 275 + ], + "score": 1.0, + "content": "multiple domains. In Proceedings of the IEEE/CVF conference on computer vision and pattern", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 273, + 257, + 284 + ], + "spans": [ + { + "bbox": [ + 115, + 273, + 257, + 284 + ], + "score": 1.0, + "content": "recognition, pp. 8188–8197, 2020.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 105, + 290, + 504, + 314 + ], + "lines": [ + { + "bbox": [ + 106, + 290, + 505, + 303 + ], + "spans": [ + { + "bbox": [ + 106, + 290, + 505, + 303 + ], + "score": 1.0, + "content": "Ning Dai, Jianze Liang, Xipeng Qiu, and Xuanjing Huang. Style transformer: Unpaired text style", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 301, + 485, + 314 + ], + "spans": [ + { + "bbox": [ + 115, + 301, + 485, + 314 + ], + "score": 1.0, + "content": "transfer without disentangled latent representation. arXiv preprint arXiv:1905.05621, 2019.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5 + }, + { + "type": "text", + "bbox": [ + 105, + 320, + 504, + 343 + ], + "lines": [ + { + "bbox": [ + 105, + 320, + 505, + 334 + ], + "spans": [ + { + "bbox": [ + 105, + 320, + 505, + 334 + ], + "score": 1.0, + "content": "Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 115, + 331, + 378, + 343 + ], + "spans": [ + { + "bbox": [ + 115, + 331, + 378, + 343 + ], + "score": 1.0, + "content": "in Neural Information Processing Systems, 34:8780–8794, 2021.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5 + }, + { + "type": "text", + "bbox": [ + 105, + 349, + 504, + 373 + ], + "lines": [ + { + "bbox": [ + 105, + 348, + 505, + 363 + ], + "spans": [ + { + "bbox": [ + 105, + 348, + 505, + 363 + ], + "score": 1.0, + "content": "Rinon Gal, Or Patashnik, Haggai Maron, Gal Chechik, and Daniel Cohen-Or. Stylegan-nada: Clip-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 360, + 470, + 374 + ], + "spans": [ + { + "bbox": [ + 115, + 360, + 470, + 374 + ], + "score": 1.0, + "content": "guided domain adaptation of image generators. arXiv preprint arXiv:2108.00946, 2021.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5 + }, + { + "type": "text", + "bbox": [ + 108, + 378, + 504, + 412 + ], + "lines": [ + { + "bbox": [ + 105, + 378, + 506, + 392 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 506, + 392 + ], + "score": 1.0, + "content": "Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 390, + 505, + 402 + ], + "spans": [ + { + "bbox": [ + 115, + 390, + 505, + 402 + ], + "score": 1.0, + "content": "Aaron Courville, and Yoshua Bengio. Generative adversarial networks. Communications of the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 401, + 236, + 412 + ], + "spans": [ + { + "bbox": [ + 115, + 401, + 236, + 412 + ], + "score": 1.0, + "content": "ACM, 63(11):139–144, 2020.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 108, + 419, + 505, + 453 + ], + "lines": [ + { + "bbox": [ + 105, + 417, + 505, + 433 + ], + "spans": [ + { + "bbox": [ + 105, + 417, + 505, + 433 + ], + "score": 1.0, + "content": "Erik Hark ¨ onen, Aaron Hertzmann, Jaakko Lehtinen, and Sylvain Paris. Ganspace: Discovering ¨", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 430, + 505, + 443 + ], + "spans": [ + { + "bbox": [ + 115, + 430, + 505, + 443 + ], + "score": 1.0, + "content": "interpretable gan controls. Advances in Neural Information Processing Systems, 33:9841–9850,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 115, + 441, + 142, + 453 + ], + "spans": [ + { + "bbox": [ + 115, + 441, + 142, + 453 + ], + "score": 1.0, + "content": "2020.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 104, + 459, + 504, + 483 + ], + "lines": [ + { + "bbox": [ + 105, + 459, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 505, + 472 + ], + "score": 1.0, + "content": "Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 470, + 367, + 483 + ], + "spans": [ + { + "bbox": [ + 115, + 470, + 367, + 483 + ], + "score": 1.0, + "content": "Neural Information Processing Systems, 33:6840–6851, 2020.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 108, + 488, + 505, + 522 + ], + "lines": [ + { + "bbox": [ + 106, + 488, + 504, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 488, + 504, + 502 + ], + "score": 1.0, + "content": "Tero Karras, Timo Aila, Samuli Laine, and Jaakko Lehtinen. Progressive growing of gans for im-", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 115, + 499, + 506, + 514 + ], + "spans": [ + { + "bbox": [ + 115, + 499, + 506, + 514 + ], + "score": 1.0, + "content": "proved quality, stability, and variation. In International Conference on Learning Representations,", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 511, + 142, + 523 + ], + "spans": [ + { + "bbox": [ + 115, + 511, + 142, + 523 + ], + "score": 1.0, + "content": "2018.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 108, + 529, + 504, + 563 + ], + "lines": [ + { + "bbox": [ + 105, + 528, + 506, + 543 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 506, + 543 + ], + "score": 1.0, + "content": "Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine, Jaakko Lehtinen, and Timo Aila. Training", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 115, + 540, + 506, + 554 + ], + "spans": [ + { + "bbox": [ + 115, + 540, + 506, + 554 + ], + "score": 1.0, + "content": "generative adversarial networks with limited data. Advances in Neural Information Processing", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 552, + 250, + 563 + ], + "spans": [ + { + "bbox": [ + 115, + 552, + 250, + 563 + ], + "score": 1.0, + "content": "Systems, 33:12104–12114, 2020.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33 + }, + { + "type": "text", + "bbox": [ + 107, + 569, + 504, + 593 + ], + "lines": [ + { + "bbox": [ + 105, + 569, + 504, + 583 + ], + "spans": [ + { + "bbox": [ + 105, + 569, + 504, + 583 + ], + "score": 1.0, + "content": "Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusion-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 116, + 581, + 382, + 593 + ], + "spans": [ + { + "bbox": [ + 116, + 581, + 382, + 593 + ], + "score": 1.0, + "content": "based generative models. arXiv preprint arXiv:2206.00364, 2022.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5 + }, + { + "type": "text", + "bbox": [ + 106, + 599, + 504, + 622 + ], + "lines": [ + { + "bbox": [ + 105, + 597, + 505, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 505, + 613 + ], + "score": 1.0, + "content": "Gwanghyun Kim and Jong Chul Ye. Diffusionclip: Text-guided image manipulation using diffusion", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 609, + 176, + 622 + ], + "spans": [ + { + "bbox": [ + 115, + 609, + 176, + 622 + ], + "score": 1.0, + "content": "models. 2021.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 37.5 + }, + { + "type": "text", + "bbox": [ + 105, + 628, + 505, + 651 + ], + "lines": [ + { + "bbox": [ + 105, + 627, + 506, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 146, + 643 + ], + "score": 1.0, + "content": "Diederik", + "type": "text" + }, + { + "bbox": [ + 146, + 629, + 155, + 639 + ], + "score": 0.26, + "content": "\\mathrm { \\bf P }", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 627, + 506, + 643 + ], + "score": 1.0, + "content": "Kingma and Max Welling. Auto-encoding variational bayes. arXiv preprint", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 115, + 639, + 214, + 651 + ], + "spans": [ + { + "bbox": [ + 115, + 639, + 214, + 651 + ], + "score": 1.0, + "content": "arXiv:1312.6114, 2013.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5 + }, + { + "type": "text", + "bbox": [ + 107, + 658, + 504, + 692 + ], + "lines": [ + { + "bbox": [ + 106, + 658, + 505, + 670 + ], + "spans": [ + { + "bbox": [ + 106, + 658, + 505, + 670 + ], + "score": 1.0, + "content": "Cheng-Han Lee, Ziwei Liu, Lingyun Wu, and Ping Luo. Maskgan: Towards diverse and interactive", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 116, + 669, + 505, + 682 + ], + "spans": [ + { + "bbox": [ + 116, + 669, + 505, + 682 + ], + "score": 1.0, + "content": "facial image manipulation. In Proceedings of the IEEE/CVF Conference on Computer Vision and", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 116, + 680, + 290, + 692 + ], + "spans": [ + { + "bbox": [ + 116, + 680, + 290, + 692 + ], + "score": 1.0, + "content": "Pattern Recognition, pp. 5549–5558, 2020.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42 + }, + { + "type": "text", + "bbox": [ + 107, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "Huan Ling, Karsten Kreis, Daiqing Li, Seung Wook Kim, Antonio Torralba, and Sanja Fidler. Edit-", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 709, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 115, + 709, + 505, + 723 + ], + "score": 1.0, + "content": "gan: High-precision semantic image editing. Advances in Neural Information Processing Systems,", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 116, + 720, + 214, + 732 + ], + "spans": [ + { + "bbox": [ + 116, + 720, + 214, + 732 + ], + "score": 1.0, + "content": "34:16331–16345, 2021.", + "type": "text" + } + ], + "index": 46 + } + ], + "index": 45 + } + ], + "page_idx": 9, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "10", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 82, + 176, + 93 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 176, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 176, + 95 + ], + "score": 1.0, + "content": "REFERENCES", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 100, + 504, + 133 + ], + "lines": [ + { + "bbox": [ + 106, + 99, + 506, + 113 + ], + "spans": [ + { + "bbox": [ + 106, + 99, + 506, + 113 + ], + "score": 1.0, + "content": "Omri Avrahami, Dani Lischinski, and Ohad Fried. Blended diffusion for text-driven editing of", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 116, + 111, + 506, + 123 + ], + "spans": [ + { + "bbox": [ + 116, + 111, + 506, + 123 + ], + "score": 1.0, + "content": "natural images. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 116, + 122, + 269, + 134 + ], + "spans": [ + { + "bbox": [ + 116, + 122, + 269, + 134 + ], + "score": 1.0, + "content": "Recognition, pp. 18208–18218, 2022.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2, + "bbox_fs": [ + 106, + 99, + 506, + 134 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 140, + 503, + 163 + ], + "lines": [ + { + "bbox": [ + 105, + 139, + 504, + 153 + ], + "spans": [ + { + "bbox": [ + 105, + 139, + 504, + 153 + ], + "score": 1.0, + "content": "Hila Chefer, Sagie Benaim, Roni Paiss, and Lior Wolf. Image-based clip-guided essence transfer.", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 116, + 151, + 278, + 163 + ], + "spans": [ + { + "bbox": [ + 116, + 151, + 278, + 163 + ], + "score": 1.0, + "content": "arXiv preprint arXiv:2110.12427, 2021.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4.5, + "bbox_fs": [ + 105, + 139, + 504, + 163 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 169, + 504, + 203 + ], + "lines": [ + { + "bbox": [ + 105, + 168, + 504, + 183 + ], + "spans": [ + { + "bbox": [ + 105, + 168, + 504, + 183 + ], + "score": 1.0, + "content": "Jooyoung Choi, Sungwon Kim, Yonghyun Jeong, Youngjune Gwon, and Sungroh Yoon. Ilvr: Con-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 115, + 180, + 505, + 194 + ], + "spans": [ + { + "bbox": [ + 115, + 180, + 505, + 194 + ], + "score": 1.0, + "content": "ditioning method for denoising diffusion probabilistic models. arXiv preprint arXiv:2108.02938,", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 115, + 191, + 144, + 204 + ], + "spans": [ + { + "bbox": [ + 115, + 191, + 144, + 204 + ], + "score": 1.0, + "content": "2021.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7, + "bbox_fs": [ + 105, + 168, + 505, + 204 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 210, + 505, + 244 + ], + "lines": [ + { + "bbox": [ + 106, + 210, + 504, + 223 + ], + "spans": [ + { + "bbox": [ + 106, + 210, + 504, + 223 + ], + "score": 1.0, + "content": "Jooyoung Choi, Jungbeom Lee, Chaehun Shin, Sungwon Kim, Hyunwoo Kim, and Sungroh Yoon.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 115, + 220, + 504, + 234 + ], + "spans": [ + { + "bbox": [ + 115, + 220, + 504, + 234 + ], + "score": 1.0, + "content": "Perception prioritized training of diffusion models. In Proceedings of the IEEE/CVF Conference", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 115, + 231, + 400, + 245 + ], + "spans": [ + { + "bbox": [ + 115, + 231, + 400, + 245 + ], + "score": 1.0, + "content": "on Computer Vision and Pattern Recognition, pp. 11472–11481, 2022.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10, + "bbox_fs": [ + 106, + 210, + 504, + 245 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 250, + 504, + 284 + ], + "lines": [ + { + "bbox": [ + 106, + 250, + 505, + 263 + ], + "spans": [ + { + "bbox": [ + 106, + 250, + 505, + 263 + ], + "score": 1.0, + "content": "Yunjey Choi, Youngjung Uh, Jaejun Yoo, and Jung-Woo Ha. Stargan v2: Diverse image synthesis for", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 115, + 261, + 505, + 275 + ], + "spans": [ + { + "bbox": [ + 115, + 261, + 505, + 275 + ], + "score": 1.0, + "content": "multiple domains. In Proceedings of the IEEE/CVF conference on computer vision and pattern", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 273, + 257, + 284 + ], + "spans": [ + { + "bbox": [ + 115, + 273, + 257, + 284 + ], + "score": 1.0, + "content": "recognition, pp. 8188–8197, 2020.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13, + "bbox_fs": [ + 106, + 250, + 505, + 284 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 290, + 504, + 314 + ], + "lines": [ + { + "bbox": [ + 106, + 290, + 505, + 303 + ], + "spans": [ + { + "bbox": [ + 106, + 290, + 505, + 303 + ], + "score": 1.0, + "content": "Ning Dai, Jianze Liang, Xipeng Qiu, and Xuanjing Huang. Style transformer: Unpaired text style", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 301, + 485, + 314 + ], + "spans": [ + { + "bbox": [ + 115, + 301, + 485, + 314 + ], + "score": 1.0, + "content": "transfer without disentangled latent representation. arXiv preprint arXiv:1905.05621, 2019.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5, + "bbox_fs": [ + 106, + 290, + 505, + 314 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 320, + 504, + 343 + ], + "lines": [ + { + "bbox": [ + 105, + 320, + 505, + 334 + ], + "spans": [ + { + "bbox": [ + 105, + 320, + 505, + 334 + ], + "score": 1.0, + "content": "Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 115, + 331, + 378, + 343 + ], + "spans": [ + { + "bbox": [ + 115, + 331, + 378, + 343 + ], + "score": 1.0, + "content": "in Neural Information Processing Systems, 34:8780–8794, 2021.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5, + "bbox_fs": [ + 105, + 320, + 505, + 343 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 349, + 504, + 373 + ], + "lines": [ + { + "bbox": [ + 105, + 348, + 505, + 363 + ], + "spans": [ + { + "bbox": [ + 105, + 348, + 505, + 363 + ], + "score": 1.0, + "content": "Rinon Gal, Or Patashnik, Haggai Maron, Gal Chechik, and Daniel Cohen-Or. Stylegan-nada: Clip-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 360, + 470, + 374 + ], + "spans": [ + { + "bbox": [ + 115, + 360, + 470, + 374 + ], + "score": 1.0, + "content": "guided domain adaptation of image generators. arXiv preprint arXiv:2108.00946, 2021.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5, + "bbox_fs": [ + 105, + 348, + 505, + 374 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 378, + 504, + 412 + ], + "lines": [ + { + "bbox": [ + 105, + 378, + 506, + 392 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 506, + 392 + ], + "score": 1.0, + "content": "Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 390, + 505, + 402 + ], + "spans": [ + { + "bbox": [ + 115, + 390, + 505, + 402 + ], + "score": 1.0, + "content": "Aaron Courville, and Yoshua Bengio. Generative adversarial networks. Communications of the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 401, + 236, + 412 + ], + "spans": [ + { + "bbox": [ + 115, + 401, + 236, + 412 + ], + "score": 1.0, + "content": "ACM, 63(11):139–144, 2020.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22, + "bbox_fs": [ + 105, + 378, + 506, + 412 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 419, + 505, + 453 + ], + "lines": [ + { + "bbox": [ + 105, + 417, + 505, + 433 + ], + "spans": [ + { + "bbox": [ + 105, + 417, + 505, + 433 + ], + "score": 1.0, + "content": "Erik Hark ¨ onen, Aaron Hertzmann, Jaakko Lehtinen, and Sylvain Paris. Ganspace: Discovering ¨", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 430, + 505, + 443 + ], + "spans": [ + { + "bbox": [ + 115, + 430, + 505, + 443 + ], + "score": 1.0, + "content": "interpretable gan controls. Advances in Neural Information Processing Systems, 33:9841–9850,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 115, + 441, + 142, + 453 + ], + "spans": [ + { + "bbox": [ + 115, + 441, + 142, + 453 + ], + "score": 1.0, + "content": "2020.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25, + "bbox_fs": [ + 105, + 417, + 505, + 453 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 459, + 504, + 483 + ], + "lines": [ + { + "bbox": [ + 105, + 459, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 105, + 459, + 505, + 472 + ], + "score": 1.0, + "content": "Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 470, + 367, + 483 + ], + "spans": [ + { + "bbox": [ + 115, + 470, + 367, + 483 + ], + "score": 1.0, + "content": "Neural Information Processing Systems, 33:6840–6851, 2020.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27.5, + "bbox_fs": [ + 105, + 459, + 505, + 483 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 488, + 505, + 522 + ], + "lines": [ + { + "bbox": [ + 106, + 488, + 504, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 488, + 504, + 502 + ], + "score": 1.0, + "content": "Tero Karras, Timo Aila, Samuli Laine, and Jaakko Lehtinen. Progressive growing of gans for im-", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 115, + 499, + 506, + 514 + ], + "spans": [ + { + "bbox": [ + 115, + 499, + 506, + 514 + ], + "score": 1.0, + "content": "proved quality, stability, and variation. In International Conference on Learning Representations,", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 511, + 142, + 523 + ], + "spans": [ + { + "bbox": [ + 115, + 511, + 142, + 523 + ], + "score": 1.0, + "content": "2018.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30, + "bbox_fs": [ + 106, + 488, + 506, + 523 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 529, + 504, + 563 + ], + "lines": [ + { + "bbox": [ + 105, + 528, + 506, + 543 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 506, + 543 + ], + "score": 1.0, + "content": "Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine, Jaakko Lehtinen, and Timo Aila. Training", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 115, + 540, + 506, + 554 + ], + "spans": [ + { + "bbox": [ + 115, + 540, + 506, + 554 + ], + "score": 1.0, + "content": "generative adversarial networks with limited data. Advances in Neural Information Processing", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 552, + 250, + 563 + ], + "spans": [ + { + "bbox": [ + 115, + 552, + 250, + 563 + ], + "score": 1.0, + "content": "Systems, 33:12104–12114, 2020.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33, + "bbox_fs": [ + 105, + 528, + 506, + 563 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 569, + 504, + 593 + ], + "lines": [ + { + "bbox": [ + 105, + 569, + 504, + 583 + ], + "spans": [ + { + "bbox": [ + 105, + 569, + 504, + 583 + ], + "score": 1.0, + "content": "Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusion-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 116, + 581, + 382, + 593 + ], + "spans": [ + { + "bbox": [ + 116, + 581, + 382, + 593 + ], + "score": 1.0, + "content": "based generative models. arXiv preprint arXiv:2206.00364, 2022.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5, + "bbox_fs": [ + 105, + 569, + 504, + 593 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 599, + 504, + 622 + ], + "lines": [ + { + "bbox": [ + 105, + 597, + 505, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 505, + 613 + ], + "score": 1.0, + "content": "Gwanghyun Kim and Jong Chul Ye. Diffusionclip: Text-guided image manipulation using diffusion", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 609, + 176, + 622 + ], + "spans": [ + { + "bbox": [ + 115, + 609, + 176, + 622 + ], + "score": 1.0, + "content": "models. 2021.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 37.5, + "bbox_fs": [ + 105, + 597, + 505, + 622 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 628, + 505, + 651 + ], + "lines": [ + { + "bbox": [ + 105, + 627, + 506, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 146, + 643 + ], + "score": 1.0, + "content": "Diederik", + "type": "text" + }, + { + "bbox": [ + 146, + 629, + 155, + 639 + ], + "score": 0.26, + "content": "\\mathrm { \\bf P }", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 627, + 506, + 643 + ], + "score": 1.0, + "content": "Kingma and Max Welling. Auto-encoding variational bayes. arXiv preprint", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 115, + 639, + 214, + 651 + ], + "spans": [ + { + "bbox": [ + 115, + 639, + 214, + 651 + ], + "score": 1.0, + "content": "arXiv:1312.6114, 2013.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5, + "bbox_fs": [ + 105, + 627, + 506, + 651 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 658, + 504, + 692 + ], + "lines": [ + { + "bbox": [ + 106, + 658, + 505, + 670 + ], + "spans": [ + { + "bbox": [ + 106, + 658, + 505, + 670 + ], + "score": 1.0, + "content": "Cheng-Han Lee, Ziwei Liu, Lingyun Wu, and Ping Luo. Maskgan: Towards diverse and interactive", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 116, + 669, + 505, + 682 + ], + "spans": [ + { + "bbox": [ + 116, + 669, + 505, + 682 + ], + "score": 1.0, + "content": "facial image manipulation. In Proceedings of the IEEE/CVF Conference on Computer Vision and", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 116, + 680, + 290, + 692 + ], + "spans": [ + { + "bbox": [ + 116, + 680, + 290, + 692 + ], + "score": 1.0, + "content": "Pattern Recognition, pp. 5549–5558, 2020.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42, + "bbox_fs": [ + 106, + 658, + 505, + 692 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "Huan Ling, Karsten Kreis, Daiqing Li, Seung Wook Kim, Antonio Torralba, and Sanja Fidler. Edit-", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 709, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 115, + 709, + 505, + 723 + ], + "score": 1.0, + "content": "gan: High-precision semantic image editing. Advances in Neural Information Processing Systems,", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 116, + 720, + 214, + 732 + ], + "spans": [ + { + "bbox": [ + 116, + 720, + 214, + 732 + ], + "score": 1.0, + "content": "34:16331–16345, 2021.", + "type": "text" + } + ], + "index": 46 + } + ], + "index": 45, + "bbox_fs": [ + 105, + 698, + 505, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 108, + 82, + 503, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 505, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 505, + 96 + ], + "score": 1.0, + "content": "Xihui Liu, Dong Huk Park, Samaneh Azadi, Gong Zhang, Arman Chopikyan, Yuxiao Hu, Humphrey", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 505, + 106 + ], + "score": 1.0, + "content": "Shi, Anna Rohrbach, and Trevor Darrell. More control for free! image synthesis with semantic", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 104, + 359, + 118 + ], + "spans": [ + { + "bbox": [ + 115, + 104, + 359, + 118 + ], + "score": 1.0, + "content": "diffusion guidance. arXiv preprint arXiv:2112.05744, 2021.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 105, + 122, + 503, + 146 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 504, + 136 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 504, + 136 + ], + "score": 1.0, + "content": "Ziwei Liu, Ping Luo, Xiaogang Wang, and Xiaoou Tang. Deep learning face attributes in the wild.", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 133, + 478, + 146 + ], + "spans": [ + { + "bbox": [ + 115, + 133, + 478, + 146 + ], + "score": 1.0, + "content": "In Proceedings of International Conference on Computer Vision (ICCV), December 2015.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 108, + 152, + 504, + 186 + ], + "lines": [ + { + "bbox": [ + 106, + 152, + 505, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 152, + 505, + 165 + ], + "score": 1.0, + "content": "Andreas Lugmayr, Martin Danelljan, Andres Romero, Fisher Yu, Radu Timofte, and Luc Van Gool.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 163, + 505, + 176 + ], + "spans": [ + { + "bbox": [ + 115, + 163, + 505, + 176 + ], + "score": 1.0, + "content": "Repaint: Inpainting using denoising diffusion probabilistic models. In Proceedings of the", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 115, + 174, + 495, + 188 + ], + "spans": [ + { + "bbox": [ + 115, + 174, + 495, + 188 + ], + "score": 1.0, + "content": "IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11461–11471, 2022.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 107, + 192, + 505, + 226 + ], + "lines": [ + { + "bbox": [ + 105, + 192, + 505, + 206 + ], + "spans": [ + { + "bbox": [ + 105, + 192, + 505, + 206 + ], + "score": 1.0, + "content": "Chenlin Meng, Yang Song, Jiaming Song, Jiajun Wu, Jun-Yan Zhu, and Stefano Ermon. Sdedit: Im-", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 115, + 204, + 505, + 217 + ], + "spans": [ + { + "bbox": [ + 115, + 204, + 505, + 217 + ], + "score": 1.0, + "content": "age synthesis and editing with stochastic differential equations. arXiv preprint arXiv:2108.01073,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 115, + 214, + 143, + 227 + ], + "spans": [ + { + "bbox": [ + 115, + 214, + 143, + 227 + ], + "score": 1.0, + "content": "2021.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 106, + 233, + 504, + 267 + ], + "lines": [ + { + "bbox": [ + 106, + 232, + 505, + 246 + ], + "spans": [ + { + "bbox": [ + 106, + 232, + 505, + 246 + ], + "score": 1.0, + "content": "Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 116, + 244, + 505, + 258 + ], + "spans": [ + { + "bbox": [ + 116, + 244, + 505, + 258 + ], + "score": 1.0, + "content": "Ilya Sutskever, and Mark Chen. Glide: Towards photorealistic image generation and editing with", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 116, + 256, + 400, + 268 + ], + "spans": [ + { + "bbox": [ + 116, + 256, + 400, + 268 + ], + "score": 1.0, + "content": "text-guided diffusion models. arXiv preprint arXiv:2112.10741, 2021.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12 + }, + { + "type": "text", + "bbox": [ + 106, + 273, + 502, + 297 + ], + "lines": [ + { + "bbox": [ + 106, + 274, + 504, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 274, + 504, + 287 + ], + "score": 1.0, + "content": "Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models.", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 115, + 285, + 444, + 298 + ], + "spans": [ + { + "bbox": [ + 115, + 285, + 444, + 298 + ], + "score": 1.0, + "content": "In International Conference on Machine Learning, pp. 8162–8171. PMLR, 2021.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5 + }, + { + "type": "text", + "bbox": [ + 106, + 303, + 505, + 338 + ], + "lines": [ + { + "bbox": [ + 105, + 303, + 505, + 317 + ], + "spans": [ + { + "bbox": [ + 105, + 303, + 505, + 317 + ], + "score": 1.0, + "content": "Or Patashnik, Zongze Wu, Eli Shechtman, Daniel Cohen-Or, and Dani Lischinski. Styleclip: Text-", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 315, + 505, + 328 + ], + "spans": [ + { + "bbox": [ + 115, + 315, + 505, + 328 + ], + "score": 1.0, + "content": "driven manipulation of stylegan imagery. In Proceedings of the IEEE/CVF International Confer-", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 115, + 326, + 312, + 338 + ], + "spans": [ + { + "bbox": [ + 115, + 326, + 312, + 338 + ], + "score": 1.0, + "content": "ence on Computer Vision, pp. 2085–2094, 2021.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17 + }, + { + "type": "text", + "bbox": [ + 107, + 344, + 505, + 378 + ], + "lines": [ + { + "bbox": [ + 105, + 344, + 505, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 505, + 358 + ], + "score": 1.0, + "content": "Konpat Preechakul, Nattanat Chatthee, Suttisak Wizadwongsa, and Supasorn Suwajanakorn. Dif-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 354, + 505, + 369 + ], + "spans": [ + { + "bbox": [ + 115, + 354, + 505, + 369 + ], + "score": 1.0, + "content": "fusion autoencoders: Toward a meaningful and decodable representation. In Proceedings of the", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 116, + 367, + 493, + 379 + ], + "spans": [ + { + "bbox": [ + 116, + 367, + 493, + 379 + ], + "score": 1.0, + "content": "IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 10619–10629, 2022.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 384, + 505, + 429 + ], + "lines": [ + { + "bbox": [ + 105, + 384, + 505, + 398 + ], + "spans": [ + { + "bbox": [ + 105, + 384, + 505, + 398 + ], + "score": 1.0, + "content": "Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 395, + 506, + 408 + ], + "spans": [ + { + "bbox": [ + 115, + 395, + 506, + 408 + ], + "score": 1.0, + "content": "Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 115, + 407, + 505, + 420 + ], + "spans": [ + { + "bbox": [ + 115, + 407, + 505, + 420 + ], + "score": 1.0, + "content": "models from natural language supervision. In International Conference on Machine Learning,", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 418, + 239, + 430 + ], + "spans": [ + { + "bbox": [ + 115, + 418, + 239, + 430 + ], + "score": 1.0, + "content": "pp. 8748–8763. PMLR, 2021.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 23.5 + }, + { + "type": "text", + "bbox": [ + 106, + 436, + 503, + 471 + ], + "lines": [ + { + "bbox": [ + 105, + 436, + 504, + 449 + ], + "spans": [ + { + "bbox": [ + 105, + 436, + 504, + 449 + ], + "score": 1.0, + "content": "Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bjorn Ommer. High- ¨", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 115, + 447, + 505, + 460 + ], + "spans": [ + { + "bbox": [ + 115, + 447, + 505, + 460 + ], + "score": 1.0, + "content": "resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Con-", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 459, + 431, + 471 + ], + "spans": [ + { + "bbox": [ + 115, + 459, + 431, + 471 + ], + "score": 1.0, + "content": "ference on Computer Vision and Pattern Recognition, pp. 10684–10695, 2022.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 107, + 476, + 503, + 510 + ], + "lines": [ + { + "bbox": [ + 105, + 476, + 504, + 489 + ], + "spans": [ + { + "bbox": [ + 105, + 476, + 504, + 489 + ], + "score": 1.0, + "content": "David E Rumelhart, Geoffrey E Hinton, and Ronald J Williams. Learning internal representations", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 488, + 505, + 501 + ], + "spans": [ + { + "bbox": [ + 116, + 488, + 505, + 501 + ], + "score": 1.0, + "content": "by error propagation. Technical report, California Univ San Diego La Jolla Inst for Cognitive", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 116, + 499, + 177, + 510 + ], + "spans": [ + { + "bbox": [ + 116, + 499, + 177, + 510 + ], + "score": 1.0, + "content": "Science, 1985.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 108, + 517, + 504, + 552 + ], + "lines": [ + { + "bbox": [ + 106, + 516, + 506, + 531 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 506, + 531 + ], + "score": 1.0, + "content": "Vikash Sehwag, Caner Hazirbas, Albert Gordo, Firat Ozgenel, and Cristian Canton. Generating high", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 115, + 528, + 506, + 541 + ], + "spans": [ + { + "bbox": [ + 115, + 528, + 506, + 541 + ], + "score": 1.0, + "content": "fidelity data from low-density regions using diffusion models. In Proceedings of the IEEE/CVF", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 117, + 539, + 448, + 553 + ], + "spans": [ + { + "bbox": [ + 117, + 539, + 448, + 553 + ], + "score": 1.0, + "content": "Conference on Computer Vision and Pattern Recognition, pp. 11492–11501, 2022.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33 + }, + { + "type": "text", + "bbox": [ + 108, + 558, + 504, + 592 + ], + "lines": [ + { + "bbox": [ + 105, + 556, + 505, + 572 + ], + "spans": [ + { + "bbox": [ + 105, + 556, + 505, + 572 + ], + "score": 1.0, + "content": "Yujun Shen, Ceyuan Yang, Xiaoou Tang, and Bolei Zhou. Interfacegan: Interpreting the disentan-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 569, + 505, + 582 + ], + "spans": [ + { + "bbox": [ + 115, + 569, + 505, + 582 + ], + "score": 1.0, + "content": "gled face representation learned by gans. IEEE transactions on pattern analysis and machine", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 580, + 192, + 592 + ], + "spans": [ + { + "bbox": [ + 115, + 580, + 192, + 592 + ], + "score": 1.0, + "content": "intelligence, 2020.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36 + }, + { + "type": "text", + "bbox": [ + 107, + 598, + 504, + 632 + ], + "lines": [ + { + "bbox": [ + 105, + 597, + 505, + 612 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 505, + 612 + ], + "score": 1.0, + "content": "Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 116, + 610, + 505, + 622 + ], + "spans": [ + { + "bbox": [ + 116, + 610, + 505, + 622 + ], + "score": 1.0, + "content": "learning using nonequilibrium thermodynamics. In International Conference on Machine Learn-", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 116, + 621, + 256, + 632 + ], + "spans": [ + { + "bbox": [ + 116, + 621, + 256, + 632 + ], + "score": 1.0, + "content": "ing, pp. 2256–2265. PMLR, 2015.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39 + }, + { + "type": "text", + "bbox": [ + 104, + 639, + 504, + 662 + ], + "lines": [ + { + "bbox": [ + 105, + 638, + 506, + 652 + ], + "spans": [ + { + "bbox": [ + 105, + 638, + 506, + 652 + ], + "score": 1.0, + "content": "Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 650, + 258, + 663 + ], + "spans": [ + { + "bbox": [ + 115, + 650, + 258, + 663 + ], + "score": 1.0, + "content": "preprint arXiv:2010.02502, 2020a.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41.5 + }, + { + "type": "text", + "bbox": [ + 106, + 668, + 504, + 702 + ], + "lines": [ + { + "bbox": [ + 106, + 668, + 505, + 681 + ], + "spans": [ + { + "bbox": [ + 106, + 668, + 505, + 681 + ], + "score": 1.0, + "content": "Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 115, + 680, + 505, + 693 + ], + "spans": [ + { + "bbox": [ + 115, + 680, + 505, + 693 + ], + "score": 1.0, + "content": "Poole. Score-based generative modeling through stochastic differential equations. arXiv preprint", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 691, + 224, + 702 + ], + "spans": [ + { + "bbox": [ + 115, + 691, + 224, + 702 + ], + "score": 1.0, + "content": "arXiv:2011.13456, 2020b.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44 + }, + { + "type": "text", + "bbox": [ + 105, + 709, + 502, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 708, + 504, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 504, + 722 + ], + "score": 1.0, + "content": "Arash Vahdat, Karsten Kreis, and Jan Kautz. Score-based generative modeling in latent space.", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 115, + 720, + 428, + 733 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 428, + 733 + ], + "score": 1.0, + "content": "Advances in Neural Information Processing Systems, 34:11287–11302, 2021.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 46.5 + } + ], + "page_idx": 10, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 765 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 765 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 108, + 82, + 503, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 505, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 505, + 96 + ], + "score": 1.0, + "content": "Xihui Liu, Dong Huk Park, Samaneh Azadi, Gong Zhang, Arman Chopikyan, Yuxiao Hu, Humphrey", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 505, + 106 + ], + "score": 1.0, + "content": "Shi, Anna Rohrbach, and Trevor Darrell. More control for free! image synthesis with semantic", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 104, + 359, + 118 + ], + "spans": [ + { + "bbox": [ + 115, + 104, + 359, + 118 + ], + "score": 1.0, + "content": "diffusion guidance. arXiv preprint arXiv:2112.05744, 2021.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1, + "bbox_fs": [ + 105, + 81, + 505, + 118 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 122, + 503, + 146 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 504, + 136 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 504, + 136 + ], + "score": 1.0, + "content": "Ziwei Liu, Ping Luo, Xiaogang Wang, and Xiaoou Tang. Deep learning face attributes in the wild.", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 133, + 478, + 146 + ], + "spans": [ + { + "bbox": [ + 115, + 133, + 478, + 146 + ], + "score": 1.0, + "content": "In Proceedings of International Conference on Computer Vision (ICCV), December 2015.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5, + "bbox_fs": [ + 106, + 123, + 504, + 146 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 152, + 504, + 186 + ], + "lines": [ + { + "bbox": [ + 106, + 152, + 505, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 152, + 505, + 165 + ], + "score": 1.0, + "content": "Andreas Lugmayr, Martin Danelljan, Andres Romero, Fisher Yu, Radu Timofte, and Luc Van Gool.", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 163, + 505, + 176 + ], + "spans": [ + { + "bbox": [ + 115, + 163, + 505, + 176 + ], + "score": 1.0, + "content": "Repaint: Inpainting using denoising diffusion probabilistic models. In Proceedings of the", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 115, + 174, + 495, + 188 + ], + "spans": [ + { + "bbox": [ + 115, + 174, + 495, + 188 + ], + "score": 1.0, + "content": "IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11461–11471, 2022.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6, + "bbox_fs": [ + 106, + 152, + 505, + 188 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 192, + 505, + 226 + ], + "lines": [ + { + "bbox": [ + 105, + 192, + 505, + 206 + ], + "spans": [ + { + "bbox": [ + 105, + 192, + 505, + 206 + ], + "score": 1.0, + "content": "Chenlin Meng, Yang Song, Jiaming Song, Jiajun Wu, Jun-Yan Zhu, and Stefano Ermon. Sdedit: Im-", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 115, + 204, + 505, + 217 + ], + "spans": [ + { + "bbox": [ + 115, + 204, + 505, + 217 + ], + "score": 1.0, + "content": "age synthesis and editing with stochastic differential equations. arXiv preprint arXiv:2108.01073,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 115, + 214, + 143, + 227 + ], + "spans": [ + { + "bbox": [ + 115, + 214, + 143, + 227 + ], + "score": 1.0, + "content": "2021.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9, + "bbox_fs": [ + 105, + 192, + 505, + 227 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 233, + 504, + 267 + ], + "lines": [ + { + "bbox": [ + 106, + 232, + 505, + 246 + ], + "spans": [ + { + "bbox": [ + 106, + 232, + 505, + 246 + ], + "score": 1.0, + "content": "Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 116, + 244, + 505, + 258 + ], + "spans": [ + { + "bbox": [ + 116, + 244, + 505, + 258 + ], + "score": 1.0, + "content": "Ilya Sutskever, and Mark Chen. Glide: Towards photorealistic image generation and editing with", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 116, + 256, + 400, + 268 + ], + "spans": [ + { + "bbox": [ + 116, + 256, + 400, + 268 + ], + "score": 1.0, + "content": "text-guided diffusion models. arXiv preprint arXiv:2112.10741, 2021.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12, + "bbox_fs": [ + 106, + 232, + 505, + 268 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 273, + 502, + 297 + ], + "lines": [ + { + "bbox": [ + 106, + 274, + 504, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 274, + 504, + 287 + ], + "score": 1.0, + "content": "Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models.", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 115, + 285, + 444, + 298 + ], + "spans": [ + { + "bbox": [ + 115, + 285, + 444, + 298 + ], + "score": 1.0, + "content": "In International Conference on Machine Learning, pp. 8162–8171. PMLR, 2021.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5, + "bbox_fs": [ + 106, + 274, + 504, + 298 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 303, + 505, + 338 + ], + "lines": [ + { + "bbox": [ + 105, + 303, + 505, + 317 + ], + "spans": [ + { + "bbox": [ + 105, + 303, + 505, + 317 + ], + "score": 1.0, + "content": "Or Patashnik, Zongze Wu, Eli Shechtman, Daniel Cohen-Or, and Dani Lischinski. Styleclip: Text-", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 315, + 505, + 328 + ], + "spans": [ + { + "bbox": [ + 115, + 315, + 505, + 328 + ], + "score": 1.0, + "content": "driven manipulation of stylegan imagery. In Proceedings of the IEEE/CVF International Confer-", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 115, + 326, + 312, + 338 + ], + "spans": [ + { + "bbox": [ + 115, + 326, + 312, + 338 + ], + "score": 1.0, + "content": "ence on Computer Vision, pp. 2085–2094, 2021.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17, + "bbox_fs": [ + 105, + 303, + 505, + 338 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 344, + 505, + 378 + ], + "lines": [ + { + "bbox": [ + 105, + 344, + 505, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 505, + 358 + ], + "score": 1.0, + "content": "Konpat Preechakul, Nattanat Chatthee, Suttisak Wizadwongsa, and Supasorn Suwajanakorn. Dif-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 354, + 505, + 369 + ], + "spans": [ + { + "bbox": [ + 115, + 354, + 505, + 369 + ], + "score": 1.0, + "content": "fusion autoencoders: Toward a meaningful and decodable representation. In Proceedings of the", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 116, + 367, + 493, + 379 + ], + "spans": [ + { + "bbox": [ + 116, + 367, + 493, + 379 + ], + "score": 1.0, + "content": "IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 10619–10629, 2022.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20, + "bbox_fs": [ + 105, + 344, + 505, + 379 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 384, + 505, + 429 + ], + "lines": [ + { + "bbox": [ + 105, + 384, + 505, + 398 + ], + "spans": [ + { + "bbox": [ + 105, + 384, + 505, + 398 + ], + "score": 1.0, + "content": "Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 395, + 506, + 408 + ], + "spans": [ + { + "bbox": [ + 115, + 395, + 506, + 408 + ], + "score": 1.0, + "content": "Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 115, + 407, + 505, + 420 + ], + "spans": [ + { + "bbox": [ + 115, + 407, + 505, + 420 + ], + "score": 1.0, + "content": "models from natural language supervision. In International Conference on Machine Learning,", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 418, + 239, + 430 + ], + "spans": [ + { + "bbox": [ + 115, + 418, + 239, + 430 + ], + "score": 1.0, + "content": "pp. 8748–8763. PMLR, 2021.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 23.5, + "bbox_fs": [ + 105, + 384, + 506, + 430 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 436, + 503, + 471 + ], + "lines": [ + { + "bbox": [ + 105, + 436, + 504, + 449 + ], + "spans": [ + { + "bbox": [ + 105, + 436, + 504, + 449 + ], + "score": 1.0, + "content": "Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bjorn Ommer. High- ¨", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 115, + 447, + 505, + 460 + ], + "spans": [ + { + "bbox": [ + 115, + 447, + 505, + 460 + ], + "score": 1.0, + "content": "resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Con-", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 459, + 431, + 471 + ], + "spans": [ + { + "bbox": [ + 115, + 459, + 431, + 471 + ], + "score": 1.0, + "content": "ference on Computer Vision and Pattern Recognition, pp. 10684–10695, 2022.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27, + "bbox_fs": [ + 105, + 436, + 505, + 471 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 476, + 503, + 510 + ], + "lines": [ + { + "bbox": [ + 105, + 476, + 504, + 489 + ], + "spans": [ + { + "bbox": [ + 105, + 476, + 504, + 489 + ], + "score": 1.0, + "content": "David E Rumelhart, Geoffrey E Hinton, and Ronald J Williams. Learning internal representations", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 488, + 505, + 501 + ], + "spans": [ + { + "bbox": [ + 116, + 488, + 505, + 501 + ], + "score": 1.0, + "content": "by error propagation. Technical report, California Univ San Diego La Jolla Inst for Cognitive", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 116, + 499, + 177, + 510 + ], + "spans": [ + { + "bbox": [ + 116, + 499, + 177, + 510 + ], + "score": 1.0, + "content": "Science, 1985.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30, + "bbox_fs": [ + 105, + 476, + 505, + 510 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 517, + 504, + 552 + ], + "lines": [ + { + "bbox": [ + 106, + 516, + 506, + 531 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 506, + 531 + ], + "score": 1.0, + "content": "Vikash Sehwag, Caner Hazirbas, Albert Gordo, Firat Ozgenel, and Cristian Canton. Generating high", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 115, + 528, + 506, + 541 + ], + "spans": [ + { + "bbox": [ + 115, + 528, + 506, + 541 + ], + "score": 1.0, + "content": "fidelity data from low-density regions using diffusion models. In Proceedings of the IEEE/CVF", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 117, + 539, + 448, + 553 + ], + "spans": [ + { + "bbox": [ + 117, + 539, + 448, + 553 + ], + "score": 1.0, + "content": "Conference on Computer Vision and Pattern Recognition, pp. 11492–11501, 2022.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33, + "bbox_fs": [ + 106, + 516, + 506, + 553 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 558, + 504, + 592 + ], + "lines": [ + { + "bbox": [ + 105, + 556, + 505, + 572 + ], + "spans": [ + { + "bbox": [ + 105, + 556, + 505, + 572 + ], + "score": 1.0, + "content": "Yujun Shen, Ceyuan Yang, Xiaoou Tang, and Bolei Zhou. Interfacegan: Interpreting the disentan-", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 569, + 505, + 582 + ], + "spans": [ + { + "bbox": [ + 115, + 569, + 505, + 582 + ], + "score": 1.0, + "content": "gled face representation learned by gans. IEEE transactions on pattern analysis and machine", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 580, + 192, + 592 + ], + "spans": [ + { + "bbox": [ + 115, + 580, + 192, + 592 + ], + "score": 1.0, + "content": "intelligence, 2020.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36, + "bbox_fs": [ + 105, + 556, + 505, + 592 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 598, + 504, + 632 + ], + "lines": [ + { + "bbox": [ + 105, + 597, + 505, + 612 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 505, + 612 + ], + "score": 1.0, + "content": "Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 116, + 610, + 505, + 622 + ], + "spans": [ + { + "bbox": [ + 116, + 610, + 505, + 622 + ], + "score": 1.0, + "content": "learning using nonequilibrium thermodynamics. In International Conference on Machine Learn-", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 116, + 621, + 256, + 632 + ], + "spans": [ + { + "bbox": [ + 116, + 621, + 256, + 632 + ], + "score": 1.0, + "content": "ing, pp. 2256–2265. PMLR, 2015.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39, + "bbox_fs": [ + 105, + 597, + 505, + 632 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 639, + 504, + 662 + ], + "lines": [ + { + "bbox": [ + 105, + 638, + 506, + 652 + ], + "spans": [ + { + "bbox": [ + 105, + 638, + 506, + 652 + ], + "score": 1.0, + "content": "Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 650, + 258, + 663 + ], + "spans": [ + { + "bbox": [ + 115, + 650, + 258, + 663 + ], + "score": 1.0, + "content": "preprint arXiv:2010.02502, 2020a.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41.5, + "bbox_fs": [ + 105, + 638, + 506, + 663 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 668, + 504, + 702 + ], + "lines": [ + { + "bbox": [ + 106, + 668, + 505, + 681 + ], + "spans": [ + { + "bbox": [ + 106, + 668, + 505, + 681 + ], + "score": 1.0, + "content": "Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 115, + 680, + 505, + 693 + ], + "spans": [ + { + "bbox": [ + 115, + 680, + 505, + 693 + ], + "score": 1.0, + "content": "Poole. Score-based generative modeling through stochastic differential equations. arXiv preprint", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 691, + 224, + 702 + ], + "spans": [ + { + "bbox": [ + 115, + 691, + 224, + 702 + ], + "score": 1.0, + "content": "arXiv:2011.13456, 2020b.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44, + "bbox_fs": [ + 106, + 668, + 505, + 702 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 709, + 502, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 708, + 504, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 504, + 722 + ], + "score": 1.0, + "content": "Arash Vahdat, Karsten Kreis, and Jan Kautz. Score-based generative modeling in latent space.", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 115, + 720, + 428, + 733 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 428, + 733 + ], + "score": 1.0, + "content": "Advances in Neural Information Processing Systems, 34:11287–11302, 2021.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 46.5, + "bbox_fs": [ + 105, + 708, + 504, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 108, + 82, + 503, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 82, + 505, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 505, + 96 + ], + "score": 1.0, + "content": "Daniel Watson, William Chan, Jonathan Ho, and Mohammad Norouzi. Learning fast samplers for", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 505, + 106 + ], + "score": 1.0, + "content": "diffusion models by differentiating through sample quality. In International Conference on Learn-", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 104, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 115, + 104, + 434, + 118 + ], + "score": 1.0, + "content": "ing Representations, 2022. URL https://openreview.net/forum?id", + "type": "text" + }, + { + "bbox": [ + 435, + 106, + 441, + 114 + ], + "score": 0.45, + "content": "\\underline { { \\underline { { \\mathbf { \\Pi } } } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 442, + 104, + 505, + 118 + ], + "score": 1.0, + "content": "VFBjuF8HEp.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 106, + 123, + 505, + 168 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "score": 1.0, + "content": "Zipeng Xu, Tianwei Lin, Hao Tang, Fu Li, Dongliang He, Nicu Sebe, Radu Timofte, Luc Van Gool,", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 134, + 505, + 147 + ], + "spans": [ + { + "bbox": [ + 115, + 134, + 505, + 147 + ], + "score": 1.0, + "content": "and Errui Ding. Predict, prevent, and evaluate: Disentangled text-driven image manipulation", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 145, + 506, + 158 + ], + "spans": [ + { + "bbox": [ + 115, + 145, + 506, + 158 + ], + "score": 1.0, + "content": "empowered by pre-trained vision-language model. In Proceedings of the IEEE/CVF Conference", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 156, + 399, + 168 + ], + "spans": [ + { + "bbox": [ + 115, + 156, + 399, + 168 + ], + "score": 1.0, + "content": "on Computer Vision and Pattern Recognition, pp. 18229–18238, 2022.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + }, + { + "type": "text", + "bbox": [ + 107, + 174, + 503, + 209 + ], + "lines": [ + { + "bbox": [ + 106, + 174, + 505, + 188 + ], + "spans": [ + { + "bbox": [ + 106, + 174, + 505, + 188 + ], + "score": 1.0, + "content": "Changqian Yu, Jingbo Wang, Chao Peng, Changxin Gao, Gang Yu, and Nong Sang. Bisenet: Bilat-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 116, + 187, + 505, + 199 + ], + "spans": [ + { + "bbox": [ + 116, + 187, + 505, + 199 + ], + "score": 1.0, + "content": "eral segmentation network for real-time semantic segmentation. In Proceedings of the European", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 117, + 197, + 358, + 209 + ], + "spans": [ + { + "bbox": [ + 117, + 197, + 358, + 209 + ], + "score": 1.0, + "content": "conference on computer vision (ECCV), pp. 325–341, 2018.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 106, + 215, + 506, + 249 + ], + "lines": [ + { + "bbox": [ + 105, + 215, + 505, + 229 + ], + "spans": [ + { + "bbox": [ + 105, + 215, + 505, + 229 + ], + "score": 1.0, + "content": "Fisher Yu, Ari Seff, Yinda Zhang, Shuran Song, Thomas Funkhouser, and Jianxiong Xiao. Lsun:", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 116, + 227, + 505, + 239 + ], + "spans": [ + { + "bbox": [ + 116, + 227, + 505, + 239 + ], + "score": 1.0, + "content": "Construction of a large-scale image dataset using deep learning with humans in the loop. arXiv", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 115, + 239, + 254, + 250 + ], + "spans": [ + { + "bbox": [ + 115, + 239, + 254, + 250 + ], + "score": 1.0, + "content": "preprint arXiv:1506.03365, 2015.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 106, + 256, + 504, + 290 + ], + "lines": [ + { + "bbox": [ + 106, + 256, + 506, + 270 + ], + "spans": [ + { + "bbox": [ + 106, + 256, + 506, + 270 + ], + "score": 1.0, + "content": "Oguz Kaan Y ˘ uksel, Enis Simsar, Ezgi G ¨ ulperi Er, and Pinar Yanardag. Latentclr: A contrastive ¨", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 267, + 505, + 281 + ], + "spans": [ + { + "bbox": [ + 115, + 267, + 505, + 281 + ], + "score": 1.0, + "content": "learning approach for unsupervised discovery of interpretable directions. In Proceedings of the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 116, + 279, + 450, + 291 + ], + "spans": [ + { + "bbox": [ + 116, + 279, + 450, + 291 + ], + "score": 1.0, + "content": "IEEE/CVF International Conference on Computer Vision, pp. 14263–14272, 2021.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14 + }, + { + "type": "text", + "bbox": [ + 106, + 297, + 504, + 331 + ], + "lines": [ + { + "bbox": [ + 106, + 297, + 505, + 310 + ], + "spans": [ + { + "bbox": [ + 106, + 297, + 505, + 310 + ], + "score": 1.0, + "content": "Bolei Zhou, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Scene", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 308, + 506, + 321 + ], + "spans": [ + { + "bbox": [ + 115, + 308, + 506, + 321 + ], + "score": 1.0, + "content": "parsing through ade20k dataset. In Proceedings of the IEEE conference on computer vision and", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 115, + 320, + 278, + 332 + ], + "spans": [ + { + "bbox": [ + 115, + 320, + 278, + 332 + ], + "score": 1.0, + "content": "pattern recognition, pp. 633–641, 2017.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17 + }, + { + "type": "text", + "bbox": [ + 107, + 338, + 503, + 372 + ], + "lines": [ + { + "bbox": [ + 106, + 338, + 504, + 351 + ], + "spans": [ + { + "bbox": [ + 106, + 338, + 504, + 351 + ], + "score": 1.0, + "content": "Bolei Zhou, Hang Zhao, Xavier Puig, Tete Xiao, Sanja Fidler, Adela Barriuso, and Antonio Torralba.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 116, + 349, + 505, + 362 + ], + "spans": [ + { + "bbox": [ + 116, + 349, + 505, + 362 + ], + "score": 1.0, + "content": "Semantic understanding of scenes through the ade20k dataset. International Journal of Computer", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 118, + 361, + 240, + 372 + ], + "spans": [ + { + "bbox": [ + 118, + 361, + 240, + 372 + ], + "score": 1.0, + "content": "Vision, 127(3):302–321, 2019.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20 + } + ], + "page_idx": 11, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "12", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 108, + 82, + 503, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 82, + 505, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 505, + 96 + ], + "score": 1.0, + "content": "Daniel Watson, William Chan, Jonathan Ho, and Mohammad Norouzi. Learning fast samplers for", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 505, + 106 + ], + "score": 1.0, + "content": "diffusion models by differentiating through sample quality. In International Conference on Learn-", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 104, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 115, + 104, + 434, + 118 + ], + "score": 1.0, + "content": "ing Representations, 2022. URL https://openreview.net/forum?id", + "type": "text" + }, + { + "bbox": [ + 435, + 106, + 441, + 114 + ], + "score": 0.45, + "content": "\\underline { { \\underline { { \\mathbf { \\Pi } } } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 442, + 104, + 505, + 118 + ], + "score": 1.0, + "content": "VFBjuF8HEp.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1, + "bbox_fs": [ + 105, + 82, + 505, + 118 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 123, + 505, + 168 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "score": 1.0, + "content": "Zipeng Xu, Tianwei Lin, Hao Tang, Fu Li, Dongliang He, Nicu Sebe, Radu Timofte, Luc Van Gool,", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 134, + 505, + 147 + ], + "spans": [ + { + "bbox": [ + 115, + 134, + 505, + 147 + ], + "score": 1.0, + "content": "and Errui Ding. Predict, prevent, and evaluate: Disentangled text-driven image manipulation", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 145, + 506, + 158 + ], + "spans": [ + { + "bbox": [ + 115, + 145, + 506, + 158 + ], + "score": 1.0, + "content": "empowered by pre-trained vision-language model. In Proceedings of the IEEE/CVF Conference", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 156, + 399, + 168 + ], + "spans": [ + { + "bbox": [ + 115, + 156, + 399, + 168 + ], + "score": 1.0, + "content": "on Computer Vision and Pattern Recognition, pp. 18229–18238, 2022.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5, + "bbox_fs": [ + 106, + 123, + 506, + 168 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 174, + 503, + 209 + ], + "lines": [ + { + "bbox": [ + 106, + 174, + 505, + 188 + ], + "spans": [ + { + "bbox": [ + 106, + 174, + 505, + 188 + ], + "score": 1.0, + "content": "Changqian Yu, Jingbo Wang, Chao Peng, Changxin Gao, Gang Yu, and Nong Sang. Bisenet: Bilat-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 116, + 187, + 505, + 199 + ], + "spans": [ + { + "bbox": [ + 116, + 187, + 505, + 199 + ], + "score": 1.0, + "content": "eral segmentation network for real-time semantic segmentation. In Proceedings of the European", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 117, + 197, + 358, + 209 + ], + "spans": [ + { + "bbox": [ + 117, + 197, + 358, + 209 + ], + "score": 1.0, + "content": "conference on computer vision (ECCV), pp. 325–341, 2018.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8, + "bbox_fs": [ + 106, + 174, + 505, + 209 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 215, + 506, + 249 + ], + "lines": [ + { + "bbox": [ + 105, + 215, + 505, + 229 + ], + "spans": [ + { + "bbox": [ + 105, + 215, + 505, + 229 + ], + "score": 1.0, + "content": "Fisher Yu, Ari Seff, Yinda Zhang, Shuran Song, Thomas Funkhouser, and Jianxiong Xiao. Lsun:", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 116, + 227, + 505, + 239 + ], + "spans": [ + { + "bbox": [ + 116, + 227, + 505, + 239 + ], + "score": 1.0, + "content": "Construction of a large-scale image dataset using deep learning with humans in the loop. arXiv", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 115, + 239, + 254, + 250 + ], + "spans": [ + { + "bbox": [ + 115, + 239, + 254, + 250 + ], + "score": 1.0, + "content": "preprint arXiv:1506.03365, 2015.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 215, + 505, + 250 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 256, + 504, + 290 + ], + "lines": [ + { + "bbox": [ + 106, + 256, + 506, + 270 + ], + "spans": [ + { + "bbox": [ + 106, + 256, + 506, + 270 + ], + "score": 1.0, + "content": "Oguz Kaan Y ˘ uksel, Enis Simsar, Ezgi G ¨ ulperi Er, and Pinar Yanardag. Latentclr: A contrastive ¨", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 267, + 505, + 281 + ], + "spans": [ + { + "bbox": [ + 115, + 267, + 505, + 281 + ], + "score": 1.0, + "content": "learning approach for unsupervised discovery of interpretable directions. In Proceedings of the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 116, + 279, + 450, + 291 + ], + "spans": [ + { + "bbox": [ + 116, + 279, + 450, + 291 + ], + "score": 1.0, + "content": "IEEE/CVF International Conference on Computer Vision, pp. 14263–14272, 2021.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14, + "bbox_fs": [ + 106, + 256, + 506, + 291 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 297, + 504, + 331 + ], + "lines": [ + { + "bbox": [ + 106, + 297, + 505, + 310 + ], + "spans": [ + { + "bbox": [ + 106, + 297, + 505, + 310 + ], + "score": 1.0, + "content": "Bolei Zhou, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Scene", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 308, + 506, + 321 + ], + "spans": [ + { + "bbox": [ + 115, + 308, + 506, + 321 + ], + "score": 1.0, + "content": "parsing through ade20k dataset. In Proceedings of the IEEE conference on computer vision and", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 115, + 320, + 278, + 332 + ], + "spans": [ + { + "bbox": [ + 115, + 320, + 278, + 332 + ], + "score": 1.0, + "content": "pattern recognition, pp. 633–641, 2017.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17, + "bbox_fs": [ + 106, + 297, + 506, + 332 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 338, + 503, + 372 + ], + "lines": [ + { + "bbox": [ + 106, + 338, + 504, + 351 + ], + "spans": [ + { + "bbox": [ + 106, + 338, + 504, + 351 + ], + "score": 1.0, + "content": "Bolei Zhou, Hang Zhao, Xavier Puig, Tete Xiao, Sanja Fidler, Adela Barriuso, and Antonio Torralba.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 116, + 349, + 505, + 362 + ], + "spans": [ + { + "bbox": [ + 116, + 349, + 505, + 362 + ], + "score": 1.0, + "content": "Semantic understanding of scenes through the ade20k dataset. International Journal of Computer", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 118, + 361, + 240, + 372 + ], + "spans": [ + { + "bbox": [ + 118, + 361, + 240, + 372 + ], + "score": 1.0, + "content": "Vision, 127(3):302–321, 2019.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20, + "bbox_fs": [ + 106, + 338, + 505, + 372 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 104, + 194, + 126 + ], + "lines": [ + { + "bbox": [ + 104, + 101, + 196, + 128 + ], + "spans": [ + { + "bbox": [ + 104, + 101, + 196, + 128 + ], + "score": 1.0, + "content": "Appendix", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "title", + "bbox": [ + 107, + 145, + 216, + 160 + ], + "lines": [ + { + "bbox": [ + 106, + 144, + 217, + 163 + ], + "spans": [ + { + "bbox": [ + 106, + 144, + 217, + 163 + ], + "score": 1.0, + "content": "Table of Contents", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 129, + 165, + 483, + 363 + ], + "lines": [ + { + "bbox": [ + 129, + 165, + 483, + 178 + ], + "spans": [ + { + "bbox": [ + 129, + 166, + 198, + 178 + ], + "score": 1.0, + "content": "A Related work", + "type": "text" + }, + { + "bbox": [ + 469, + 165, + 483, + 178 + ], + "score": 1.0, + "content": "14", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 129, + 186, + 483, + 201 + ], + "spans": [ + { + "bbox": [ + 129, + 187, + 208, + 199 + ], + "score": 1.0, + "content": "B More discussion", + "type": "text" + }, + { + "bbox": [ + 468, + 186, + 483, + 201 + ], + "score": 1.0, + "content": "1 5", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 130, + 207, + 482, + 221 + ], + "spans": [ + { + "bbox": [ + 130, + 208, + 221, + 220 + ], + "score": 1.0, + "content": "C Proof of Theorem 1", + "type": "text" + }, + { + "bbox": [ + 469, + 207, + 482, + 221 + ], + "score": 1.0, + "content": "1 5", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 128, + 228, + 482, + 242 + ], + "spans": [ + { + "bbox": [ + 128, + 228, + 235, + 242 + ], + "score": 1.0, + "content": "D Additional supports for", + "type": "text" + }, + { + "bbox": [ + 236, + 230, + 242, + 239 + ], + "score": 0.46, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 242, + 228, + 312, + 242 + ], + "score": 1.0, + "content": "-space with Asyrp", + "type": "text" + }, + { + "bbox": [ + 468, + 228, + 482, + 240 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 141, + 240, + 482, + 254 + ], + "spans": [ + { + "bbox": [ + 141, + 240, + 254, + 254 + ], + "score": 1.0, + "content": "D.1 Random perturbation on", + "type": "text" + }, + { + "bbox": [ + 254, + 244, + 259, + 250 + ], + "score": 0.48, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 260, + 240, + 344, + 254 + ], + "score": 1.0, + "content": "-space without Asyrp .", + "type": "text" + }, + { + "bbox": [ + 468, + 240, + 482, + 253 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 142, + 252, + 482, + 265 + ], + "spans": [ + { + "bbox": [ + 142, + 252, + 269, + 265 + ], + "score": 1.0, + "content": "D.2 Robustness and semantics in", + "type": "text" + }, + { + "bbox": [ + 269, + 254, + 275, + 262 + ], + "score": 0.65, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 276, + 252, + 315, + 265 + ], + "score": 1.0, + "content": "-space and", + "type": "text" + }, + { + "bbox": [ + 315, + 255, + 320, + 262 + ], + "score": 0.39, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 320, + 252, + 388, + 265 + ], + "score": 1.0, + "content": "-space with Asyrp", + "type": "text" + }, + { + "bbox": [ + 468, + 252, + 482, + 264 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 142, + 264, + 483, + 279 + ], + "spans": [ + { + "bbox": [ + 142, + 264, + 201, + 277 + ], + "score": 1.0, + "content": "D.3 Choice of", + "type": "text" + }, + { + "bbox": [ + 202, + 266, + 208, + 275 + ], + "score": 0.62, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 208, + 264, + 269, + 277 + ], + "score": 1.0, + "content": "-space in U-Net", + "type": "text" + }, + { + "bbox": [ + 468, + 264, + 483, + 279 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 129, + 285, + 483, + 300 + ], + "spans": [ + { + "bbox": [ + 129, + 285, + 245, + 298 + ], + "score": 1.0, + "content": "E Implicit neural directions", + "type": "text" + }, + { + "bbox": [ + 468, + 285, + 483, + 300 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 129, + 306, + 483, + 320 + ], + "spans": [ + { + "bbox": [ + 129, + 307, + 138, + 317 + ], + "score": 0.28, + "content": "\\mathbf { F }", + "type": "inline_equation" + }, + { + "bbox": [ + 139, + 306, + 388, + 320 + ], + "score": 1.0, + "content": "Quality improvements by non-accelerated sampling with scaled", + "type": "text" + }, + { + "bbox": [ + 389, + 307, + 406, + 317 + ], + "score": 0.71, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 468, + 306, + 483, + 319 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 129, + 326, + 483, + 340 + ], + "spans": [ + { + "bbox": [ + 129, + 327, + 291, + 340 + ], + "score": 1.0, + "content": "G Editing strength and editing flexibility", + "type": "text" + }, + { + "bbox": [ + 468, + 326, + 483, + 339 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 142, + 338, + 483, + 353 + ], + "spans": [ + { + "bbox": [ + 142, + 338, + 239, + 353 + ], + "score": 1.0, + "content": "G.1 Editing strength and", + "type": "text" + }, + { + "bbox": [ + 240, + 340, + 257, + 350 + ], + "score": 0.83, + "content": "t _ { e d i t }", + "type": "inline_equation" + }, + { + "bbox": [ + 468, + 339, + 483, + 351 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 142, + 350, + 483, + 365 + ], + "spans": [ + { + "bbox": [ + 142, + 350, + 245, + 365 + ], + "score": 1.0, + "content": "G.2 Editing flexibility and", + "type": "text" + }, + { + "bbox": [ + 245, + 352, + 268, + 362 + ], + "score": 0.79, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 268, + 350, + 278, + 365 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 469, + 351, + 483, + 363 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 128, + 373, + 480, + 384 + ], + "lines": [ + { + "bbox": [ + 129, + 370, + 482, + 387 + ], + "spans": [ + { + "bbox": [ + 129, + 370, + 210, + 387 + ], + "score": 1.0, + "content": "H Quality boosting", + "type": "text" + }, + { + "bbox": [ + 469, + 372, + 482, + 384 + ], + "score": 1.0, + "content": "1 9", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14 + }, + { + "type": "title", + "bbox": [ + 126, + 393, + 480, + 405 + ], + "lines": [ + { + "bbox": [ + 127, + 392, + 482, + 407 + ], + "spans": [ + { + "bbox": [ + 127, + 392, + 187, + 407 + ], + "score": 1.0, + "content": "I Algorithm", + "type": "text" + }, + { + "bbox": [ + 468, + 392, + 482, + 405 + ], + "score": 1.0, + "content": "2 3", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "title", + "bbox": [ + 129, + 415, + 482, + 425 + ], + "lines": [ + { + "bbox": [ + 128, + 413, + 482, + 428 + ], + "spans": [ + { + "bbox": [ + 128, + 414, + 207, + 426 + ], + "score": 1.0, + "content": "J Training details", + "type": "text" + }, + { + "bbox": [ + 468, + 413, + 482, + 428 + ], + "score": 1.0, + "content": "2 3", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 140, + 426, + 481, + 450 + ], + "lines": [ + { + "bbox": [ + 142, + 425, + 482, + 438 + ], + "spans": [ + { + "bbox": [ + 142, + 425, + 236, + 437 + ], + "score": 1.0, + "content": "J.1 Loss coefficients .", + "type": "text" + }, + { + "bbox": [ + 468, + 425, + 482, + 438 + ], + "score": 1.0, + "content": "23", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 142, + 437, + 482, + 451 + ], + "spans": [ + { + "bbox": [ + 142, + 437, + 399, + 451 + ], + "score": 1.0, + "content": "J.2 Training with random sampling instead of the training datasets .", + "type": "text" + }, + { + "bbox": [ + 468, + 438, + 482, + 449 + ], + "score": 1.0, + "content": "24", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5 + }, + { + "type": "title", + "bbox": [ + 130, + 459, + 188, + 469 + ], + "lines": [ + { + "bbox": [ + 129, + 458, + 188, + 470 + ], + "spans": [ + { + "bbox": [ + 129, + 458, + 188, + 470 + ], + "score": 1.0, + "content": "K Evaluation", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "title", + "bbox": [ + 130, + 461, + 483, + 469 + ], + "lines": [ + { + "bbox": [ + 469, + 458, + 482, + 471 + ], + "spans": [ + { + "bbox": [ + 469, + 458, + 482, + 471 + ], + "score": 1.0, + "content": "2 4", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 143, + 470, + 482, + 495 + ], + "lines": [ + { + "bbox": [ + 143, + 470, + 482, + 483 + ], + "spans": [ + { + "bbox": [ + 143, + 471, + 213, + 482 + ], + "score": 1.0, + "content": "K.1 User study .", + "type": "text" + }, + { + "bbox": [ + 468, + 470, + 482, + 483 + ], + "score": 1.0, + "content": "24", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 142, + 482, + 482, + 496 + ], + "spans": [ + { + "bbox": [ + 142, + 482, + 374, + 496 + ], + "score": 1.0, + "content": "K.2 Segmentation consistency and directional CLIP similarity", + "type": "text" + }, + { + "bbox": [ + 468, + 482, + 482, + 495 + ], + "score": 1.0, + "content": "25", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5 + }, + { + "type": "title", + "bbox": [ + 130, + 504, + 185, + 514 + ], + "lines": [ + { + "bbox": [ + 129, + 503, + 186, + 516 + ], + "spans": [ + { + "bbox": [ + 129, + 503, + 186, + 516 + ], + "score": 1.0, + "content": "L Directions", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 23 + }, + { + "type": "title", + "bbox": [ + 137, + 505, + 482, + 514 + ], + "lines": [ + { + "bbox": [ + 468, + 503, + 483, + 516 + ], + "spans": [ + { + "bbox": [ + 468, + 503, + 483, + 516 + ], + "score": 1.0, + "content": "26", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 141, + 516, + 481, + 539 + ], + "lines": [ + { + "bbox": [ + 142, + 515, + 482, + 527 + ], + "spans": [ + { + "bbox": [ + 142, + 515, + 228, + 527 + ], + "score": 1.0, + "content": "L.1 Global direction", + "type": "text" + }, + { + "bbox": [ + 468, + 515, + 482, + 527 + ], + "score": 1.0, + "content": "26", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 142, + 527, + 482, + 540 + ], + "spans": [ + { + "bbox": [ + 142, + 527, + 255, + 540 + ], + "score": 1.0, + "content": "L.2 Compare three methods", + "type": "text" + }, + { + "bbox": [ + 468, + 527, + 482, + 539 + ], + "score": 1.0, + "content": "29", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25.5 + }, + { + "type": "title", + "bbox": [ + 130, + 549, + 216, + 560 + ], + "lines": [ + { + "bbox": [ + 129, + 546, + 217, + 563 + ], + "spans": [ + { + "bbox": [ + 129, + 546, + 217, + 563 + ], + "score": 1.0, + "content": "M Random sampling", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 469, + 549, + 481, + 559 + ], + "lines": [ + { + "bbox": [ + 468, + 549, + 482, + 561 + ], + "spans": [ + { + "bbox": [ + 468, + 549, + 482, + 561 + ], + "score": 1.0, + "content": "3 0", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "title", + "bbox": [ + 131, + 570, + 200, + 580 + ], + "lines": [ + { + "bbox": [ + 129, + 568, + 200, + 582 + ], + "spans": [ + { + "bbox": [ + 129, + 568, + 200, + 582 + ], + "score": 1.0, + "content": "N More samples", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "title", + "bbox": [ + 132, + 572, + 483, + 579 + ], + "lines": [ + { + "bbox": [ + 132, + 572, + 483, + 579 + ], + "spans": [], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 142, + 582, + 483, + 617 + ], + "lines": [ + { + "bbox": [ + 142, + 580, + 482, + 594 + ], + "spans": [ + { + "bbox": [ + 142, + 580, + 206, + 594 + ], + "score": 1.0, + "content": "N.1 ImageNet", + "type": "text" + }, + { + "bbox": [ + 469, + 581, + 482, + 593 + ], + "score": 1.0, + "content": "32", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 142, + 593, + 482, + 605 + ], + "spans": [ + { + "bbox": [ + 142, + 593, + 237, + 605 + ], + "score": 1.0, + "content": "N.2 Multi-interpolation", + "type": "text" + }, + { + "bbox": [ + 469, + 593, + 482, + 605 + ], + "score": 1.0, + "content": "32", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 142, + 605, + 482, + 617 + ], + "spans": [ + { + "bbox": [ + 142, + 605, + 267, + 617 + ], + "score": 1.0, + "content": "N.3 More results on all datasets", + "type": "text" + }, + { + "bbox": [ + 468, + 605, + 482, + 617 + ], + "score": 1.0, + "content": "32", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32 + } + ], + "page_idx": 12, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 26, + 293, + 38 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 293, + 39 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 293, + 39 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "13", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 104, + 194, + 126 + ], + "lines": [ + { + "bbox": [ + 104, + 101, + 196, + 128 + ], + "spans": [ + { + "bbox": [ + 104, + 101, + 196, + 128 + ], + "score": 1.0, + "content": "Appendix", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "title", + "bbox": [ + 107, + 145, + 216, + 160 + ], + "lines": [ + { + "bbox": [ + 106, + 144, + 217, + 163 + ], + "spans": [ + { + "bbox": [ + 106, + 144, + 217, + 163 + ], + "score": 1.0, + "content": "Table of Contents", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1 + }, + { + "type": "index", + "bbox": [ + 129, + 165, + 483, + 363 + ], + "lines": [ + { + "bbox": [ + 129, + 165, + 483, + 178 + ], + "spans": [ + { + "bbox": [ + 129, + 166, + 198, + 178 + ], + "score": 1.0, + "content": "A Related work", + "type": "text" + }, + { + "bbox": [ + 469, + 165, + 483, + 178 + ], + "score": 1.0, + "content": "14", + "type": "text" + } + ], + "index": 2, + "is_list_start_line": true + }, + { + "bbox": [ + 129, + 186, + 483, + 201 + ], + "spans": [ + { + "bbox": [ + 129, + 187, + 208, + 199 + ], + "score": 1.0, + "content": "B More discussion", + "type": "text" + }, + { + "bbox": [ + 468, + 186, + 483, + 201 + ], + "score": 1.0, + "content": "1 5", + "type": "text" + } + ], + "index": 3, + "is_list_start_line": true + }, + { + "bbox": [ + 130, + 207, + 482, + 221 + ], + "spans": [ + { + "bbox": [ + 130, + 208, + 221, + 220 + ], + "score": 1.0, + "content": "C Proof of Theorem 1", + "type": "text" + }, + { + "bbox": [ + 469, + 207, + 482, + 221 + ], + "score": 1.0, + "content": "1 5", + "type": "text" + } + ], + "index": 4, + "is_list_start_line": true + }, + { + "bbox": [ + 128, + 228, + 482, + 242 + ], + "spans": [ + { + "bbox": [ + 128, + 228, + 235, + 242 + ], + "score": 1.0, + "content": "D Additional supports for", + "type": "text" + }, + { + "bbox": [ + 236, + 230, + 242, + 239 + ], + "score": 0.46, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 242, + 228, + 312, + 242 + ], + "score": 1.0, + "content": "-space with Asyrp", + "type": "text" + }, + { + "bbox": [ + 468, + 228, + 482, + 240 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ], + "index": 5, + "is_list_start_line": true + }, + { + "bbox": [ + 141, + 240, + 482, + 254 + ], + "spans": [ + { + "bbox": [ + 141, + 240, + 254, + 254 + ], + "score": 1.0, + "content": "D.1 Random perturbation on", + "type": "text" + }, + { + "bbox": [ + 254, + 244, + 259, + 250 + ], + "score": 0.48, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 260, + 240, + 344, + 254 + ], + "score": 1.0, + "content": "-space without Asyrp .", + "type": "text" + }, + { + "bbox": [ + 468, + 240, + 482, + 253 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ], + "index": 6, + "is_list_start_line": true + }, + { + "bbox": [ + 142, + 252, + 482, + 265 + ], + "spans": [ + { + "bbox": [ + 142, + 252, + 269, + 265 + ], + "score": 1.0, + "content": "D.2 Robustness and semantics in", + "type": "text" + }, + { + "bbox": [ + 269, + 254, + 275, + 262 + ], + "score": 0.65, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 276, + 252, + 315, + 265 + ], + "score": 1.0, + "content": "-space and", + "type": "text" + }, + { + "bbox": [ + 315, + 255, + 320, + 262 + ], + "score": 0.39, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 320, + 252, + 388, + 265 + ], + "score": 1.0, + "content": "-space with Asyrp", + "type": "text" + }, + { + "bbox": [ + 468, + 252, + 482, + 264 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ], + "index": 7, + "is_list_start_line": true + }, + { + "bbox": [ + 142, + 264, + 483, + 279 + ], + "spans": [ + { + "bbox": [ + 142, + 264, + 201, + 277 + ], + "score": 1.0, + "content": "D.3 Choice of", + "type": "text" + }, + { + "bbox": [ + 202, + 266, + 208, + 275 + ], + "score": 0.62, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 208, + 264, + 269, + 277 + ], + "score": 1.0, + "content": "-space in U-Net", + "type": "text" + }, + { + "bbox": [ + 468, + 264, + 483, + 279 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ], + "index": 8, + "is_list_start_line": true + }, + { + "bbox": [ + 129, + 285, + 483, + 300 + ], + "spans": [ + { + "bbox": [ + 129, + 285, + 245, + 298 + ], + "score": 1.0, + "content": "E Implicit neural directions", + "type": "text" + }, + { + "bbox": [ + 468, + 285, + 483, + 300 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ], + "index": 9, + "is_list_start_line": true + }, + { + "bbox": [ + 129, + 306, + 483, + 320 + ], + "spans": [ + { + "bbox": [ + 129, + 307, + 138, + 317 + ], + "score": 0.28, + "content": "\\mathbf { F }", + "type": "inline_equation" + }, + { + "bbox": [ + 139, + 306, + 388, + 320 + ], + "score": 1.0, + "content": "Quality improvements by non-accelerated sampling with scaled", + "type": "text" + }, + { + "bbox": [ + 389, + 307, + 406, + 317 + ], + "score": 0.71, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 468, + 306, + 483, + 319 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ], + "index": 10, + "is_list_start_line": true + }, + { + "bbox": [ + 129, + 326, + 483, + 340 + ], + "spans": [ + { + "bbox": [ + 129, + 327, + 291, + 340 + ], + "score": 1.0, + "content": "G Editing strength and editing flexibility", + "type": "text" + }, + { + "bbox": [ + 468, + 326, + 483, + 339 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ], + "index": 11, + "is_list_start_line": true + }, + { + "bbox": [ + 142, + 338, + 483, + 353 + ], + "spans": [ + { + "bbox": [ + 142, + 338, + 239, + 353 + ], + "score": 1.0, + "content": "G.1 Editing strength and", + "type": "text" + }, + { + "bbox": [ + 240, + 340, + 257, + 350 + ], + "score": 0.83, + "content": "t _ { e d i t }", + "type": "inline_equation" + }, + { + "bbox": [ + 468, + 339, + 483, + 351 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ], + "index": 12, + "is_list_start_line": true + }, + { + "bbox": [ + 142, + 350, + 483, + 365 + ], + "spans": [ + { + "bbox": [ + 142, + 350, + 245, + 365 + ], + "score": 1.0, + "content": "G.2 Editing flexibility and", + "type": "text" + }, + { + "bbox": [ + 245, + 352, + 268, + 362 + ], + "score": 0.79, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 268, + 350, + 278, + 365 + ], + "score": 1.0, + "content": ".", + "type": "text" + }, + { + "bbox": [ + 469, + 351, + 483, + 363 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ], + "index": 13, + "is_list_start_line": true + } + ], + "index": 7.5, + "bbox_fs": [ + 128, + 165, + 483, + 365 + ] + }, + { + "type": "text", + "bbox": [ + 128, + 373, + 480, + 384 + ], + "lines": [ + { + "bbox": [ + 129, + 370, + 482, + 387 + ], + "spans": [ + { + "bbox": [ + 129, + 370, + 210, + 387 + ], + "score": 1.0, + "content": "H Quality boosting", + "type": "text" + }, + { + "bbox": [ + 469, + 372, + 482, + 384 + ], + "score": 1.0, + "content": "1 9", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14, + "bbox_fs": [ + 129, + 370, + 482, + 387 + ] + }, + { + "type": "title", + "bbox": [ + 126, + 393, + 480, + 405 + ], + "lines": [ + { + "bbox": [ + 127, + 392, + 482, + 407 + ], + "spans": [ + { + "bbox": [ + 127, + 392, + 187, + 407 + ], + "score": 1.0, + "content": "I Algorithm", + "type": "text" + }, + { + "bbox": [ + 468, + 392, + 482, + 405 + ], + "score": 1.0, + "content": "2 3", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "title", + "bbox": [ + 129, + 415, + 482, + 425 + ], + "lines": [ + { + "bbox": [ + 128, + 413, + 482, + 428 + ], + "spans": [ + { + "bbox": [ + 128, + 414, + 207, + 426 + ], + "score": 1.0, + "content": "J Training details", + "type": "text" + }, + { + "bbox": [ + 468, + 413, + 482, + 428 + ], + "score": 1.0, + "content": "2 3", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "index", + "bbox": [ + 140, + 426, + 481, + 450 + ], + "lines": [ + { + "bbox": [ + 142, + 425, + 482, + 438 + ], + "spans": [ + { + "bbox": [ + 142, + 425, + 236, + 437 + ], + "score": 1.0, + "content": "J.1 Loss coefficients .", + "type": "text" + }, + { + "bbox": [ + 468, + 425, + 482, + 438 + ], + "score": 1.0, + "content": "23", + "type": "text" + } + ], + "index": 17, + "is_list_start_line": true + }, + { + "bbox": [ + 142, + 437, + 482, + 451 + ], + "spans": [ + { + "bbox": [ + 142, + 437, + 399, + 451 + ], + "score": 1.0, + "content": "J.2 Training with random sampling instead of the training datasets .", + "type": "text" + }, + { + "bbox": [ + 468, + 438, + 482, + 449 + ], + "score": 1.0, + "content": "24", + "type": "text" + } + ], + "index": 18, + "is_list_start_line": true + } + ], + "index": 17.5, + "bbox_fs": [ + 142, + 425, + 482, + 451 + ] + }, + { + "type": "title", + "bbox": [ + 130, + 459, + 188, + 469 + ], + "lines": [ + { + "bbox": [ + 129, + 458, + 188, + 470 + ], + "spans": [ + { + "bbox": [ + 129, + 458, + 188, + 470 + ], + "score": 1.0, + "content": "K Evaluation", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "title", + "bbox": [ + 130, + 461, + 483, + 469 + ], + "lines": [ + { + "bbox": [ + 469, + 458, + 482, + 471 + ], + "spans": [ + { + "bbox": [ + 469, + 458, + 482, + 471 + ], + "score": 1.0, + "content": "2 4", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "index", + "bbox": [ + 143, + 470, + 482, + 495 + ], + "lines": [ + { + "bbox": [ + 143, + 470, + 482, + 483 + ], + "spans": [ + { + "bbox": [ + 143, + 471, + 213, + 482 + ], + "score": 1.0, + "content": "K.1 User study .", + "type": "text" + }, + { + "bbox": [ + 468, + 470, + 482, + 483 + ], + "score": 1.0, + "content": "24", + "type": "text" + } + ], + "index": 21, + "is_list_start_line": true + }, + { + "bbox": [ + 142, + 482, + 482, + 496 + ], + "spans": [ + { + "bbox": [ + 142, + 482, + 374, + 496 + ], + "score": 1.0, + "content": "K.2 Segmentation consistency and directional CLIP similarity", + "type": "text" + }, + { + "bbox": [ + 468, + 482, + 482, + 495 + ], + "score": 1.0, + "content": "25", + "type": "text" + } + ], + "index": 22, + "is_list_start_line": true + } + ], + "index": 21.5, + "bbox_fs": [ + 142, + 470, + 482, + 496 + ] + }, + { + "type": "title", + "bbox": [ + 130, + 504, + 185, + 514 + ], + "lines": [ + { + "bbox": [ + 129, + 503, + 186, + 516 + ], + "spans": [ + { + "bbox": [ + 129, + 503, + 186, + 516 + ], + "score": 1.0, + "content": "L Directions", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 23 + }, + { + "type": "title", + "bbox": [ + 137, + 505, + 482, + 514 + ], + "lines": [ + { + "bbox": [ + 468, + 503, + 483, + 516 + ], + "spans": [ + { + "bbox": [ + 468, + 503, + 483, + 516 + ], + "score": 1.0, + "content": "26", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "index", + "bbox": [ + 141, + 516, + 481, + 539 + ], + "lines": [ + { + "bbox": [ + 142, + 515, + 482, + 527 + ], + "spans": [ + { + "bbox": [ + 142, + 515, + 228, + 527 + ], + "score": 1.0, + "content": "L.1 Global direction", + "type": "text" + }, + { + "bbox": [ + 468, + 515, + 482, + 527 + ], + "score": 1.0, + "content": "26", + "type": "text" + } + ], + "index": 25, + "is_list_start_line": true + }, + { + "bbox": [ + 142, + 527, + 482, + 540 + ], + "spans": [ + { + "bbox": [ + 142, + 527, + 255, + 540 + ], + "score": 1.0, + "content": "L.2 Compare three methods", + "type": "text" + }, + { + "bbox": [ + 468, + 527, + 482, + 539 + ], + "score": 1.0, + "content": "29", + "type": "text" + } + ], + "index": 26, + "is_list_start_line": true + } + ], + "index": 25.5, + "bbox_fs": [ + 142, + 515, + 482, + 540 + ] + }, + { + "type": "title", + "bbox": [ + 130, + 549, + 216, + 560 + ], + "lines": [ + { + "bbox": [ + 129, + 546, + 217, + 563 + ], + "spans": [ + { + "bbox": [ + 129, + 546, + 217, + 563 + ], + "score": 1.0, + "content": "M Random sampling", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 469, + 549, + 481, + 559 + ], + "lines": [ + { + "bbox": [ + 468, + 549, + 482, + 561 + ], + "spans": [ + { + "bbox": [ + 468, + 549, + 482, + 561 + ], + "score": 1.0, + "content": "3 0", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28, + "bbox_fs": [ + 468, + 549, + 482, + 561 + ] + }, + { + "type": "title", + "bbox": [ + 131, + 570, + 200, + 580 + ], + "lines": [ + { + "bbox": [ + 129, + 568, + 200, + 582 + ], + "spans": [ + { + "bbox": [ + 129, + 568, + 200, + 582 + ], + "score": 1.0, + "content": "N More samples", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "title", + "bbox": [ + 132, + 572, + 483, + 579 + ], + "lines": [ + { + "bbox": [ + 132, + 572, + 483, + 579 + ], + "spans": [], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "index", + "bbox": [ + 142, + 582, + 483, + 617 + ], + "lines": [ + { + "bbox": [ + 142, + 580, + 482, + 594 + ], + "spans": [ + { + "bbox": [ + 142, + 580, + 206, + 594 + ], + "score": 1.0, + "content": "N.1 ImageNet", + "type": "text" + }, + { + "bbox": [ + 469, + 581, + 482, + 593 + ], + "score": 1.0, + "content": "32", + "type": "text" + } + ], + "index": 31, + "is_list_start_line": true + }, + { + "bbox": [ + 142, + 593, + 482, + 605 + ], + "spans": [ + { + "bbox": [ + 142, + 593, + 237, + 605 + ], + "score": 1.0, + "content": "N.2 Multi-interpolation", + "type": "text" + }, + { + "bbox": [ + 469, + 593, + 482, + 605 + ], + "score": 1.0, + "content": "32", + "type": "text" + } + ], + "index": 32, + "is_list_start_line": true + }, + { + "bbox": [ + 142, + 605, + 482, + 617 + ], + "spans": [ + { + "bbox": [ + 142, + 605, + 267, + 617 + ], + "score": 1.0, + "content": "N.3 More results on all datasets", + "type": "text" + }, + { + "bbox": [ + 468, + 605, + 482, + 617 + ], + "score": 1.0, + "content": "32", + "type": "text" + } + ], + "index": 33, + "is_list_start_line": true + } + ], + "index": 32, + "bbox_fs": [ + 142, + 580, + 482, + 617 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 109, + 96, + 505, + 288 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 109, + 96, + 505, + 288 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 109, + 96, + 505, + 288 + ], + "spans": [ + { + "bbox": [ + 109, + 96, + 505, + 288 + ], + "score": 0.974, + "type": "image", + "image_path": "ccb938a111ed5b23226f87f2c9601955a9f9a495d880770b80c3a6c0651db3e8.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 109, + 96, + 505, + 160.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 109, + 160.0, + 505, + 224.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 109, + 224.0, + 505, + 288.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 316, + 504, + 339 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 315, + 505, + 329 + ], + "spans": [ + { + "bbox": [ + 105, + 315, + 505, + 329 + ], + "score": 1.0, + "content": "Figure 11: Editing dog to smile. We provide high-resolution results of editing real images in the", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 328, + 175, + 339 + ], + "spans": [ + { + "bbox": [ + 106, + 328, + 175, + 339 + ], + "score": 1.0, + "content": "teaser and more.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "title", + "bbox": [ + 108, + 365, + 211, + 378 + ], + "lines": [ + { + "bbox": [ + 105, + 363, + 213, + 380 + ], + "spans": [ + { + "bbox": [ + 105, + 363, + 213, + 380 + ], + "score": 1.0, + "content": "A RELATED WORK", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 107, + 389, + 505, + 467 + ], + "lines": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "score": 1.0, + "content": "After Sohl-Dickstein et al. (2015), denoising diffusion probabilistic models (DDPMs) provide a uni-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 400, + 505, + 414 + ], + "spans": [ + { + "bbox": [ + 105, + 400, + 505, + 414 + ], + "score": 1.0, + "content": "versal approach for generative modeling (Ho et al., 2020). On the other hand, Song et al. (2020b)", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 411, + 505, + 425 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 505, + 425 + ], + "score": 1.0, + "content": "suggests score-based model and unifies SDEs incorporating diffusion models with score-based mod-", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 423, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 105, + 423, + 505, + 436 + ], + "score": 1.0, + "content": "els. Subsequent works renovate diffusion models by focusing on architectures, scheduling, weight-", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 433, + 505, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 433, + 505, + 447 + ], + "score": 1.0, + "content": "ing, and fast sampling (Nichol & Dhariwal (2021), Karras et al. (2022), Choi et al. (2022), Song", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 444, + 505, + 457 + ], + "spans": [ + { + "bbox": [ + 105, + 444, + 505, + 457 + ], + "score": 1.0, + "content": "et al. (2020a), Watson et al. (2022)). They mainly consider random generation rather than controlled", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 456, + 154, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 154, + 468 + ], + "score": 1.0, + "content": "generation.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 107, + 472, + 505, + 604 + ], + "lines": [ + { + "bbox": [ + 105, + 472, + 505, + 485 + ], + "spans": [ + { + "bbox": [ + 105, + 472, + 505, + 485 + ], + "score": 1.0, + "content": "In the meantime, Dhariwal & Nichol (2021) introduces classifier guidance not only improving the", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 484, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 484, + 506, + 496 + ], + "score": 1.0, + "content": "quality of images but also retrieving specific class of images. Since it can apply any guidance, its", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 495, + 505, + 507 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 505, + 507 + ], + "score": 1.0, + "content": "variants have emerged (Sehwag et al. (2022), Avrahami et al. (2022), Liu et al. (2021), Nichol et al.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 506, + 505, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 505, + 518 + ], + "score": 1.0, + "content": "(2021)). However it requires a noise-dependent classifier (or any off-the-shelf models) and additional", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 517, + 506, + 529 + ], + "spans": [ + { + "bbox": [ + 105, + 517, + 506, + 529 + ], + "score": 1.0, + "content": "cost to compute gradients for the guidance during its sampling process. The other works try to", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 527, + 506, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 527, + 506, + 541 + ], + "score": 1.0, + "content": "control the generative process using image-space guidance (Choi et al. (2021), Meng et al. (2021),", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 537, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 552 + ], + "score": 1.0, + "content": "Lugmayr et al. (2022), Avrahami et al. (2022)). They manipulate resulting images by matching noisy", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 549, + 505, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 549, + 505, + 563 + ], + "score": 1.0, + "content": "images with target images during the reverse process. Still, it is hard to expect delicate control of the", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 560, + 506, + 573 + ], + "spans": [ + { + "bbox": [ + 105, + 560, + 506, + 573 + ], + "score": 1.0, + "content": "reverse process from the image guidance. Furthermore, Preechakul et al. (2022) introduces an extra", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 571, + 505, + 583 + ], + "spans": [ + { + "bbox": [ + 105, + 571, + 505, + 583 + ], + "score": 1.0, + "content": "encoder which encodes the semantic features of a real image in order to condition the generative", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 582, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 582, + 506, + 596 + ], + "score": 1.0, + "content": "process. Although the semantics allow one to control diffusion models, it requires additional training", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 593, + 487, + 605 + ], + "spans": [ + { + "bbox": [ + 105, + 593, + 487, + 605 + ], + "score": 1.0, + "content": "from scratch with the encoder and inherently can not use the other pretrained diffusion models.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 18.5 + }, + { + "type": "text", + "bbox": [ + 107, + 610, + 505, + 676 + ], + "lines": [ + { + "bbox": [ + 105, + 609, + 505, + 622 + ], + "spans": [ + { + "bbox": [ + 105, + 609, + 505, + 622 + ], + "score": 1.0, + "content": "For controllability, Rombach et al. (2022) and Vahdat et al. (2021) apply another approach which", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 621, + 505, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 505, + 633 + ], + "score": 1.0, + "content": "adapts VAE (Kingma & Welling, 2013) and autoencoder (Rumelhart et al., 1985) to diffusion mod-", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 631, + 505, + 644 + ], + "spans": [ + { + "bbox": [ + 105, + 631, + 505, + 644 + ], + "score": 1.0, + "content": "els. In spite of their great success in editing, their diffusion models learn the distribution of the", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 642, + 505, + 655 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 505, + 655 + ], + "score": 1.0, + "content": "learned embeddings in VAE or autoencoder, not the images. Kim & Ye (2021) proposes another", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "score": 1.0, + "content": "strategy: fine-tuning a whole diffusion model for image editing. It shows valid performance but it", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 665, + 350, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 665, + 350, + 678 + ], + "score": 1.0, + "content": "requires each fine-tuned model corresponding each attribute.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 107, + 681, + 502, + 704 + ], + "lines": [ + { + "bbox": [ + 105, + 681, + 504, + 695 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 504, + 695 + ], + "score": 1.0, + "content": "In comparison, Asyrp enables outstanding manipulation without high computation, specifically de-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 693, + 308, + 705 + ], + "spans": [ + { + "bbox": [ + 105, + 693, + 308, + 705 + ], + "score": 1.0, + "content": "signed architectures, or fine-tuning whole models.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 108, + 709, + 502, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 504, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 504, + 722 + ], + "score": 1.0, + "content": "Meanwhile, generative adversarial network Goodfellow et al. (2020) address their latent space for", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "score": 1.0, + "content": "image editing (Ling et al. (2021), Hark ¨ onen et al. (2020), Chefer et al. (2021), Shen et al. (2020), ¨", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5 + } + ], + "page_idx": 13, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 752, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 14 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 109, + 96, + 505, + 288 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 109, + 96, + 505, + 288 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 109, + 96, + 505, + 288 + ], + "spans": [ + { + "bbox": [ + 109, + 96, + 505, + 288 + ], + "score": 0.974, + "type": "image", + "image_path": "ccb938a111ed5b23226f87f2c9601955a9f9a495d880770b80c3a6c0651db3e8.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 109, + 96, + 505, + 160.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 109, + 160.0, + 505, + 224.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 109, + 224.0, + 505, + 288.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 316, + 504, + 339 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 315, + 505, + 329 + ], + "spans": [ + { + "bbox": [ + 105, + 315, + 505, + 329 + ], + "score": 1.0, + "content": "Figure 11: Editing dog to smile. We provide high-resolution results of editing real images in the", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 328, + 175, + 339 + ], + "spans": [ + { + "bbox": [ + 106, + 328, + 175, + 339 + ], + "score": 1.0, + "content": "teaser and more.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + } + ], + "index": 2.25 + }, + { + "type": "title", + "bbox": [ + 108, + 365, + 211, + 378 + ], + "lines": [ + { + "bbox": [ + 105, + 363, + 213, + 380 + ], + "spans": [ + { + "bbox": [ + 105, + 363, + 213, + 380 + ], + "score": 1.0, + "content": "A RELATED WORK", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 107, + 389, + 505, + 467 + ], + "lines": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "score": 1.0, + "content": "After Sohl-Dickstein et al. (2015), denoising diffusion probabilistic models (DDPMs) provide a uni-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 400, + 505, + 414 + ], + "spans": [ + { + "bbox": [ + 105, + 400, + 505, + 414 + ], + "score": 1.0, + "content": "versal approach for generative modeling (Ho et al., 2020). On the other hand, Song et al. (2020b)", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 411, + 505, + 425 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 505, + 425 + ], + "score": 1.0, + "content": "suggests score-based model and unifies SDEs incorporating diffusion models with score-based mod-", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 423, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 105, + 423, + 505, + 436 + ], + "score": 1.0, + "content": "els. Subsequent works renovate diffusion models by focusing on architectures, scheduling, weight-", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 433, + 505, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 433, + 505, + 447 + ], + "score": 1.0, + "content": "ing, and fast sampling (Nichol & Dhariwal (2021), Karras et al. (2022), Choi et al. (2022), Song", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 444, + 505, + 457 + ], + "spans": [ + { + "bbox": [ + 105, + 444, + 505, + 457 + ], + "score": 1.0, + "content": "et al. (2020a), Watson et al. (2022)). They mainly consider random generation rather than controlled", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 456, + 154, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 154, + 468 + ], + "score": 1.0, + "content": "generation.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 9, + "bbox_fs": [ + 105, + 389, + 505, + 468 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 472, + 505, + 604 + ], + "lines": [ + { + "bbox": [ + 105, + 472, + 505, + 485 + ], + "spans": [ + { + "bbox": [ + 105, + 472, + 505, + 485 + ], + "score": 1.0, + "content": "In the meantime, Dhariwal & Nichol (2021) introduces classifier guidance not only improving the", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 484, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 484, + 506, + 496 + ], + "score": 1.0, + "content": "quality of images but also retrieving specific class of images. Since it can apply any guidance, its", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 495, + 505, + 507 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 505, + 507 + ], + "score": 1.0, + "content": "variants have emerged (Sehwag et al. (2022), Avrahami et al. (2022), Liu et al. (2021), Nichol et al.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 506, + 505, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 505, + 518 + ], + "score": 1.0, + "content": "(2021)). However it requires a noise-dependent classifier (or any off-the-shelf models) and additional", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 517, + 506, + 529 + ], + "spans": [ + { + "bbox": [ + 105, + 517, + 506, + 529 + ], + "score": 1.0, + "content": "cost to compute gradients for the guidance during its sampling process. The other works try to", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 527, + 506, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 527, + 506, + 541 + ], + "score": 1.0, + "content": "control the generative process using image-space guidance (Choi et al. (2021), Meng et al. (2021),", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 537, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 552 + ], + "score": 1.0, + "content": "Lugmayr et al. (2022), Avrahami et al. (2022)). They manipulate resulting images by matching noisy", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 549, + 505, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 549, + 505, + 563 + ], + "score": 1.0, + "content": "images with target images during the reverse process. Still, it is hard to expect delicate control of the", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 560, + 506, + 573 + ], + "spans": [ + { + "bbox": [ + 105, + 560, + 506, + 573 + ], + "score": 1.0, + "content": "reverse process from the image guidance. Furthermore, Preechakul et al. (2022) introduces an extra", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 571, + 505, + 583 + ], + "spans": [ + { + "bbox": [ + 105, + 571, + 505, + 583 + ], + "score": 1.0, + "content": "encoder which encodes the semantic features of a real image in order to condition the generative", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 582, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 582, + 506, + 596 + ], + "score": 1.0, + "content": "process. Although the semantics allow one to control diffusion models, it requires additional training", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 593, + 487, + 605 + ], + "spans": [ + { + "bbox": [ + 105, + 593, + 487, + 605 + ], + "score": 1.0, + "content": "from scratch with the encoder and inherently can not use the other pretrained diffusion models.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 18.5, + "bbox_fs": [ + 105, + 472, + 506, + 605 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 610, + 505, + 676 + ], + "lines": [ + { + "bbox": [ + 105, + 609, + 505, + 622 + ], + "spans": [ + { + "bbox": [ + 105, + 609, + 505, + 622 + ], + "score": 1.0, + "content": "For controllability, Rombach et al. (2022) and Vahdat et al. (2021) apply another approach which", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 621, + 505, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 505, + 633 + ], + "score": 1.0, + "content": "adapts VAE (Kingma & Welling, 2013) and autoencoder (Rumelhart et al., 1985) to diffusion mod-", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 631, + 505, + 644 + ], + "spans": [ + { + "bbox": [ + 105, + 631, + 505, + 644 + ], + "score": 1.0, + "content": "els. In spite of their great success in editing, their diffusion models learn the distribution of the", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 642, + 505, + 655 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 505, + 655 + ], + "score": 1.0, + "content": "learned embeddings in VAE or autoencoder, not the images. Kim & Ye (2021) proposes another", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "score": 1.0, + "content": "strategy: fine-tuning a whole diffusion model for image editing. It shows valid performance but it", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 665, + 350, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 665, + 350, + 678 + ], + "score": 1.0, + "content": "requires each fine-tuned model corresponding each attribute.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 27.5, + "bbox_fs": [ + 105, + 609, + 505, + 678 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 681, + 502, + 704 + ], + "lines": [ + { + "bbox": [ + 105, + 681, + 504, + 695 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 504, + 695 + ], + "score": 1.0, + "content": "In comparison, Asyrp enables outstanding manipulation without high computation, specifically de-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 693, + 308, + 705 + ], + "spans": [ + { + "bbox": [ + 105, + 693, + 308, + 705 + ], + "score": 1.0, + "content": "signed architectures, or fine-tuning whole models.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5, + "bbox_fs": [ + 105, + 681, + 504, + 705 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 709, + 502, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 504, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 504, + 722 + ], + "score": 1.0, + "content": "Meanwhile, generative adversarial network Goodfellow et al. (2020) address their latent space for", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "score": 1.0, + "content": "image editing (Ling et al. (2021), Hark ¨ onen et al. (2020), Chefer et al. (2021), Shen et al. (2020), ¨", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5, + "bbox_fs": [ + 106, + 709, + 505, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 108, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 505, + 95 + ], + "score": 1.0, + "content": "Yuksel et al. (2021), Patashnik et al. (2021), Gal et al. (2021), Dai et al. (2019), Xu et al. (2022)). ¨", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 93, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 106, + 93, + 505, + 106 + ], + "score": 1.0, + "content": "However they have to conduct ‘inversion’ to their latent space for real image editing and ‘GAN", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 103, + 417, + 118 + ], + "spans": [ + { + "bbox": [ + 105, + 103, + 417, + 118 + ], + "score": 1.0, + "content": "inversion’ is often challenging and produces unexpected appearance changes.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 106, + 121, + 504, + 144 + ], + "lines": [ + { + "bbox": [ + 105, + 120, + 506, + 134 + ], + "spans": [ + { + "bbox": [ + 105, + 120, + 506, + 134 + ], + "score": 1.0, + "content": "On the contrary, Asyrp enables to use latent space of real images by nearly perfect easy inversion of", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 131, + 138, + 143 + ], + "spans": [ + { + "bbox": [ + 105, + 131, + 138, + 143 + ], + "score": 1.0, + "content": "DDIM.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + }, + { + "type": "title", + "bbox": [ + 108, + 167, + 225, + 180 + ], + "lines": [ + { + "bbox": [ + 104, + 165, + 228, + 181 + ], + "spans": [ + { + "bbox": [ + 104, + 165, + 228, + 181 + ], + "score": 1.0, + "content": "B MORE DISCUSSION", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 108, + 195, + 503, + 218 + ], + "lines": [ + { + "bbox": [ + 106, + 195, + 505, + 208 + ], + "spans": [ + { + "bbox": [ + 106, + 195, + 505, + 208 + ], + "score": 1.0, + "content": "In this section, we discuss the pros and cons of diffusion-model-based and GAN-based methods.", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 208, + 321, + 219 + ], + "spans": [ + { + "bbox": [ + 106, + 208, + 321, + 219 + ], + "score": 1.0, + "content": "And we provide guidelines for further improvements.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 107, + 223, + 505, + 312 + ], + "lines": [ + { + "bbox": [ + 105, + 223, + 505, + 236 + ], + "spans": [ + { + "bbox": [ + 105, + 223, + 505, + 236 + ], + "score": 1.0, + "content": "GAN-based latent manipulation methods Patashnik et al. (2021); Gal et al. (2021)require careful", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 234, + 506, + 248 + ], + "spans": [ + { + "bbox": [ + 105, + 234, + 506, + 248 + ], + "score": 1.0, + "content": "inversion from real images to latent codes for real image editing. On the contrary, our proposed", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 246, + 505, + 258 + ], + "spans": [ + { + "bbox": [ + 106, + 246, + 505, + 258 + ], + "score": 1.0, + "content": "method based on diffusion models has a powerful advantage; the sophisticated inversion method is", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 257, + 505, + 268 + ], + "spans": [ + { + "bbox": [ + 106, + 257, + 505, + 268 + ], + "score": 1.0, + "content": "not necessary. This means that we can obtain the latent code of an arbitrary real image even if the", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 268, + 505, + 280 + ], + "spans": [ + { + "bbox": [ + 106, + 268, + 505, + 280 + ], + "score": 1.0, + "content": "image is not in the trained domain. On the other hand, several inversion methods have been proposed", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 278, + 506, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 278, + 506, + 291 + ], + "score": 1.0, + "content": "for GANs to obtain the latent of the real image, and the corresponding latent manipulation method", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 289, + 506, + 303 + ], + "spans": [ + { + "bbox": [ + 105, + 289, + 506, + 303 + ], + "score": 1.0, + "content": "should be considered for each inversion method. For example, it is difficult to apply the method of", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 300, + 357, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 300, + 146, + 314 + ], + "score": 1.0, + "content": "editing in", + "type": "text" + }, + { + "bbox": [ + 147, + 302, + 156, + 311 + ], + "score": 0.69, + "content": "w", + "type": "inline_equation" + }, + { + "bbox": [ + 156, + 300, + 312, + 314 + ], + "score": 1.0, + "content": "space to the method of inversion using", + "type": "text" + }, + { + "bbox": [ + 313, + 300, + 328, + 311 + ], + "score": 0.88, + "content": "w ^ { + }", + "type": "inline_equation" + }, + { + "bbox": [ + 328, + 300, + 357, + 314 + ], + "score": 1.0, + "content": "space.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 11.5 + }, + { + "type": "text", + "bbox": [ + 108, + 317, + 505, + 351 + ], + "lines": [ + { + "bbox": [ + 106, + 317, + 505, + 330 + ], + "spans": [ + { + "bbox": [ + 106, + 317, + 505, + 330 + ], + "score": 1.0, + "content": "However, GANs have the advantage of fast sampling. In addition, diffusion models have a relatively", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 328, + 505, + 340 + ], + "spans": [ + { + "bbox": [ + 106, + 328, + 505, + 340 + ], + "score": 1.0, + "content": "slow sampling time. Additionally, we have to be aware of the time steps of diffusion models, which", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 338, + 202, + 352 + ], + "spans": [ + { + "bbox": [ + 105, + 338, + 202, + 352 + ], + "score": 1.0, + "content": "is still less well known.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17 + }, + { + "type": "text", + "bbox": [ + 107, + 356, + 505, + 389 + ], + "lines": [ + { + "bbox": [ + 106, + 356, + 505, + 369 + ], + "spans": [ + { + "bbox": [ + 106, + 356, + 505, + 369 + ], + "score": 1.0, + "content": "The advantage of being free from Inversion provides the following milestones. The manipulation in", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 367, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 106, + 367, + 505, + 379 + ], + "score": 1.0, + "content": "the latent of the diffusion models is the same as the editing in real images. It can be expanded to", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 378, + 418, + 392 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 292, + 392 + ], + "score": 1.0, + "content": "segmentation, clustering, classification, etc. in", + "type": "text" + }, + { + "bbox": [ + 293, + 379, + 299, + 388 + ], + "score": 0.71, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 299, + 378, + 418, + 392 + ], + "score": 1.0, + "content": "-space for real-world images.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 395, + 505, + 439 + ], + "lines": [ + { + "bbox": [ + 105, + 395, + 505, + 408 + ], + "spans": [ + { + "bbox": [ + 105, + 395, + 505, + 408 + ], + "score": 1.0, + "content": "It would be an interesting research direction to employ previous techniques. Our method can be used", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 406, + 506, + 419 + ], + "spans": [ + { + "bbox": [ + 105, + 406, + 506, + 419 + ], + "score": 1.0, + "content": "in conjunction with gradient guidance methods. Although we do not focus on random sampling, ours", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 417, + 506, + 430 + ], + "spans": [ + { + "bbox": [ + 105, + 417, + 312, + 430 + ], + "score": 1.0, + "content": "works effectively for sampling with stochastic. (See", + "type": "text" + }, + { + "bbox": [ + 312, + 417, + 330, + 429 + ], + "score": 0.5, + "content": "\\ S \\mathbf { M } .", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 417, + 506, + 430 + ], + "score": 1.0, + "content": ") It may bring more diverse methods to steer", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 428, + 178, + 439 + ], + "spans": [ + { + "bbox": [ + 106, + 428, + 178, + 439 + ], + "score": 1.0, + "content": "diffusion models.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 23.5 + }, + { + "type": "text", + "bbox": [ + 107, + 444, + 505, + 489 + ], + "lines": [ + { + "bbox": [ + 106, + 445, + 505, + 457 + ], + "spans": [ + { + "bbox": [ + 106, + 446, + 113, + 455 + ], + "score": 0.5, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 113, + 445, + 505, + 457 + ], + "score": 1.0, + "content": "-space in the latent diffusion models such as stable diffusion, is another interesting research di-", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 456, + 505, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 357, + 468 + ], + "score": 1.0, + "content": "rection. The main contribution of our paper is only modifying", + "type": "text" + }, + { + "bbox": [ + 358, + 456, + 371, + 468 + ], + "score": 0.87, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 371, + 456, + 441, + 468 + ], + "score": 1.0, + "content": "while preserving", + "type": "text" + }, + { + "bbox": [ + 442, + 456, + 455, + 467 + ], + "score": 0.87, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 455, + 456, + 505, + 468 + ], + "score": 1.0, + "content": ", and can be", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 467, + 505, + 479 + ], + "spans": [ + { + "bbox": [ + 105, + 467, + 505, + 479 + ], + "score": 1.0, + "content": "adapted with latent diffusion models. However, since the latent meaning may be different due to", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 479, + 302, + 489 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 302, + 489 + ], + "score": 1.0, + "content": "structural differences, research on this is needed.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 106, + 495, + 504, + 517 + ], + "lines": [ + { + "bbox": [ + 106, + 494, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 106, + 494, + 252, + 508 + ], + "score": 1.0, + "content": "Furthermore, all of the properties of", + "type": "text" + }, + { + "bbox": [ + 252, + 496, + 258, + 505 + ], + "score": 0.71, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 259, + 494, + 505, + 508 + ], + "score": 1.0, + "content": "-space according to the time step has not been fully discussed", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 506, + 346, + 519 + ], + "spans": [ + { + "bbox": [ + 106, + 506, + 346, + 519 + ], + "score": 1.0, + "content": "so far. Research on them can be expected to expand further.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30.5 + }, + { + "type": "text", + "bbox": [ + 105, + 536, + 504, + 559 + ], + "lines": [ + { + "bbox": [ + 105, + 535, + 505, + 549 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 505, + 549 + ], + "score": 1.0, + "content": "Limitations. Editing with Asyrp seldom yields changes in overall style or peripheral objects but", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 547, + 498, + 560 + ], + "spans": [ + { + "bbox": [ + 106, + 547, + 498, + 560 + ], + "score": 1.0, + "content": "edits attributes of the main object. Style transfer using frozen diffusion models is our future work.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32.5 + }, + { + "type": "text", + "bbox": [ + 107, + 578, + 505, + 644 + ], + "lines": [ + { + "bbox": [ + 105, + 577, + 505, + 591 + ], + "spans": [ + { + "bbox": [ + 105, + 577, + 505, + 591 + ], + "score": 1.0, + "content": "Societal impact / Ethics statement. Techniques for high-quality image manipulation such as", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 588, + 505, + 602 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 505, + 602 + ], + "score": 1.0, + "content": "Asyrp should be accompanied by social and/or technical solutions to prevent abuse. We acknowledge", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 600, + 506, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 600, + 506, + 613 + ], + "score": 1.0, + "content": "the potential ethical implications that may arise from the use of our image manipulation technique,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 611, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 611, + 506, + 624 + ], + "score": 1.0, + "content": "Asyrp. We advocate for the development and implementation of social and technical solutions to", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "score": 1.0, + "content": "prevent potential abuses such as spreading disinformation or propaganda. We are committed to en-", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 633, + 477, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 633, + 477, + 646 + ], + "score": 1.0, + "content": "suring fairness and non-discrimination, legal compliance, and research integrity in our work.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 36.5 + }, + { + "type": "title", + "bbox": [ + 108, + 667, + 243, + 680 + ], + "lines": [ + { + "bbox": [ + 106, + 666, + 245, + 682 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 245, + 682 + ], + "score": 1.0, + "content": "C PROOF OF THEOREM 1", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 40 + }, + { + "type": "text", + "bbox": [ + 108, + 696, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 695, + 506, + 710 + ], + "spans": [ + { + "bbox": [ + 105, + 695, + 211, + 710 + ], + "score": 1.0, + "content": "Proof of Theorem 1. Let", + "type": "text" + }, + { + "bbox": [ + 212, + 696, + 222, + 708 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 222, + 695, + 457, + 710 + ], + "score": 1.0, + "content": "be a predicted noise during the original reverse process at", + "type": "text" + }, + { + "bbox": [ + 457, + 698, + 463, + 707 + ], + "score": 0.76, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 463, + 695, + 481, + 710 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 481, + 696, + 491, + 708 + ], + "score": 0.88, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 492, + 695, + 506, + 710 + ], + "score": 1.0, + "content": "be", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 706, + 505, + 721 + ], + "spans": [ + { + "bbox": [ + 105, + 706, + 222, + 721 + ], + "score": 1.0, + "content": "its shifted counterpart. Then,", + "type": "text" + }, + { + "bbox": [ + 223, + 708, + 305, + 719 + ], + "score": 0.92, + "content": "\\Delta \\pmb { x } _ { t } = \\tilde { \\pmb { x } } _ { t - 1 } - \\pmb { x } _ { t - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 306, + 706, + 383, + 721 + ], + "score": 1.0, + "content": "is negligible where", + "type": "text" + }, + { + "bbox": [ + 384, + 707, + 505, + 721 + ], + "score": 0.93, + "content": "\\tilde { \\pmb { x } } _ { t - 1 } = \\bar { \\sqrt { \\alpha _ { t - 1 } } } \\mathbf { P } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } \\bar { \\Big ) } ) +", + "type": "inline_equation" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 717, + 159, + 735 + ], + "spans": [ + { + "bbox": [ + 106, + 719, + 154, + 733 + ], + "score": 0.91, + "content": "\\mathbf { D } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 717, + 159, + 735 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42 + } + ], + "page_idx": 14, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "15", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 108, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 505, + 95 + ], + "score": 1.0, + "content": "Yuksel et al. (2021), Patashnik et al. (2021), Gal et al. (2021), Dai et al. (2019), Xu et al. (2022)). ¨", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 93, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 106, + 93, + 505, + 106 + ], + "score": 1.0, + "content": "However they have to conduct ‘inversion’ to their latent space for real image editing and ‘GAN", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 103, + 417, + 118 + ], + "spans": [ + { + "bbox": [ + 105, + 103, + 417, + 118 + ], + "score": 1.0, + "content": "inversion’ is often challenging and produces unexpected appearance changes.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1, + "bbox_fs": [ + 105, + 81, + 505, + 118 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 121, + 504, + 144 + ], + "lines": [ + { + "bbox": [ + 105, + 120, + 506, + 134 + ], + "spans": [ + { + "bbox": [ + 105, + 120, + 506, + 134 + ], + "score": 1.0, + "content": "On the contrary, Asyrp enables to use latent space of real images by nearly perfect easy inversion of", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 131, + 138, + 143 + ], + "spans": [ + { + "bbox": [ + 105, + 131, + 138, + 143 + ], + "score": 1.0, + "content": "DDIM.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5, + "bbox_fs": [ + 105, + 120, + 506, + 143 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 167, + 225, + 180 + ], + "lines": [ + { + "bbox": [ + 104, + 165, + 228, + 181 + ], + "spans": [ + { + "bbox": [ + 104, + 165, + 228, + 181 + ], + "score": 1.0, + "content": "B MORE DISCUSSION", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "list", + "bbox": [ + 108, + 195, + 503, + 218 + ], + "lines": [ + { + "bbox": [ + 106, + 195, + 505, + 208 + ], + "spans": [ + { + "bbox": [ + 106, + 195, + 505, + 208 + ], + "score": 1.0, + "content": "In this section, we discuss the pros and cons of diffusion-model-based and GAN-based methods.", + "type": "text" + } + ], + "index": 6, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 208, + 321, + 219 + ], + "spans": [ + { + "bbox": [ + 106, + 208, + 321, + 219 + ], + "score": 1.0, + "content": "And we provide guidelines for further improvements.", + "type": "text" + } + ], + "index": 7, + "is_list_start_line": true, + "is_list_end_line": true + } + ], + "index": 6.5, + "bbox_fs": [ + 106, + 195, + 505, + 219 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 223, + 505, + 312 + ], + "lines": [ + { + "bbox": [ + 105, + 223, + 505, + 236 + ], + "spans": [ + { + "bbox": [ + 105, + 223, + 505, + 236 + ], + "score": 1.0, + "content": "GAN-based latent manipulation methods Patashnik et al. (2021); Gal et al. (2021)require careful", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 234, + 506, + 248 + ], + "spans": [ + { + "bbox": [ + 105, + 234, + 506, + 248 + ], + "score": 1.0, + "content": "inversion from real images to latent codes for real image editing. On the contrary, our proposed", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 246, + 505, + 258 + ], + "spans": [ + { + "bbox": [ + 106, + 246, + 505, + 258 + ], + "score": 1.0, + "content": "method based on diffusion models has a powerful advantage; the sophisticated inversion method is", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 257, + 505, + 268 + ], + "spans": [ + { + "bbox": [ + 106, + 257, + 505, + 268 + ], + "score": 1.0, + "content": "not necessary. This means that we can obtain the latent code of an arbitrary real image even if the", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 268, + 505, + 280 + ], + "spans": [ + { + "bbox": [ + 106, + 268, + 505, + 280 + ], + "score": 1.0, + "content": "image is not in the trained domain. On the other hand, several inversion methods have been proposed", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 278, + 506, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 278, + 506, + 291 + ], + "score": 1.0, + "content": "for GANs to obtain the latent of the real image, and the corresponding latent manipulation method", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 289, + 506, + 303 + ], + "spans": [ + { + "bbox": [ + 105, + 289, + 506, + 303 + ], + "score": 1.0, + "content": "should be considered for each inversion method. For example, it is difficult to apply the method of", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 300, + 357, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 300, + 146, + 314 + ], + "score": 1.0, + "content": "editing in", + "type": "text" + }, + { + "bbox": [ + 147, + 302, + 156, + 311 + ], + "score": 0.69, + "content": "w", + "type": "inline_equation" + }, + { + "bbox": [ + 156, + 300, + 312, + 314 + ], + "score": 1.0, + "content": "space to the method of inversion using", + "type": "text" + }, + { + "bbox": [ + 313, + 300, + 328, + 311 + ], + "score": 0.88, + "content": "w ^ { + }", + "type": "inline_equation" + }, + { + "bbox": [ + 328, + 300, + 357, + 314 + ], + "score": 1.0, + "content": "space.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 11.5, + "bbox_fs": [ + 105, + 223, + 506, + 314 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 317, + 505, + 351 + ], + "lines": [ + { + "bbox": [ + 106, + 317, + 505, + 330 + ], + "spans": [ + { + "bbox": [ + 106, + 317, + 505, + 330 + ], + "score": 1.0, + "content": "However, GANs have the advantage of fast sampling. In addition, diffusion models have a relatively", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 328, + 505, + 340 + ], + "spans": [ + { + "bbox": [ + 106, + 328, + 505, + 340 + ], + "score": 1.0, + "content": "slow sampling time. Additionally, we have to be aware of the time steps of diffusion models, which", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 338, + 202, + 352 + ], + "spans": [ + { + "bbox": [ + 105, + 338, + 202, + 352 + ], + "score": 1.0, + "content": "is still less well known.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17, + "bbox_fs": [ + 105, + 317, + 505, + 352 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 356, + 505, + 389 + ], + "lines": [ + { + "bbox": [ + 106, + 356, + 505, + 369 + ], + "spans": [ + { + "bbox": [ + 106, + 356, + 505, + 369 + ], + "score": 1.0, + "content": "The advantage of being free from Inversion provides the following milestones. The manipulation in", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 367, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 106, + 367, + 505, + 379 + ], + "score": 1.0, + "content": "the latent of the diffusion models is the same as the editing in real images. It can be expanded to", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 378, + 418, + 392 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 292, + 392 + ], + "score": 1.0, + "content": "segmentation, clustering, classification, etc. in", + "type": "text" + }, + { + "bbox": [ + 293, + 379, + 299, + 388 + ], + "score": 0.71, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 299, + 378, + 418, + 392 + ], + "score": 1.0, + "content": "-space for real-world images.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20, + "bbox_fs": [ + 105, + 356, + 505, + 392 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 395, + 505, + 439 + ], + "lines": [ + { + "bbox": [ + 105, + 395, + 505, + 408 + ], + "spans": [ + { + "bbox": [ + 105, + 395, + 505, + 408 + ], + "score": 1.0, + "content": "It would be an interesting research direction to employ previous techniques. Our method can be used", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 406, + 506, + 419 + ], + "spans": [ + { + "bbox": [ + 105, + 406, + 506, + 419 + ], + "score": 1.0, + "content": "in conjunction with gradient guidance methods. Although we do not focus on random sampling, ours", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 417, + 506, + 430 + ], + "spans": [ + { + "bbox": [ + 105, + 417, + 312, + 430 + ], + "score": 1.0, + "content": "works effectively for sampling with stochastic. (See", + "type": "text" + }, + { + "bbox": [ + 312, + 417, + 330, + 429 + ], + "score": 0.5, + "content": "\\ S \\mathbf { M } .", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 417, + 506, + 430 + ], + "score": 1.0, + "content": ") It may bring more diverse methods to steer", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 428, + 178, + 439 + ], + "spans": [ + { + "bbox": [ + 106, + 428, + 178, + 439 + ], + "score": 1.0, + "content": "diffusion models.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 23.5, + "bbox_fs": [ + 105, + 395, + 506, + 439 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 444, + 505, + 489 + ], + "lines": [ + { + "bbox": [ + 106, + 445, + 505, + 457 + ], + "spans": [ + { + "bbox": [ + 106, + 446, + 113, + 455 + ], + "score": 0.5, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 113, + 445, + 505, + 457 + ], + "score": 1.0, + "content": "-space in the latent diffusion models such as stable diffusion, is another interesting research di-", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 456, + 505, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 357, + 468 + ], + "score": 1.0, + "content": "rection. The main contribution of our paper is only modifying", + "type": "text" + }, + { + "bbox": [ + 358, + 456, + 371, + 468 + ], + "score": 0.87, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 371, + 456, + 441, + 468 + ], + "score": 1.0, + "content": "while preserving", + "type": "text" + }, + { + "bbox": [ + 442, + 456, + 455, + 467 + ], + "score": 0.87, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 455, + 456, + 505, + 468 + ], + "score": 1.0, + "content": ", and can be", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 467, + 505, + 479 + ], + "spans": [ + { + "bbox": [ + 105, + 467, + 505, + 479 + ], + "score": 1.0, + "content": "adapted with latent diffusion models. However, since the latent meaning may be different due to", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 479, + 302, + 489 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 302, + 489 + ], + "score": 1.0, + "content": "structural differences, research on this is needed.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 27.5, + "bbox_fs": [ + 105, + 445, + 505, + 489 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 495, + 504, + 517 + ], + "lines": [ + { + "bbox": [ + 106, + 494, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 106, + 494, + 252, + 508 + ], + "score": 1.0, + "content": "Furthermore, all of the properties of", + "type": "text" + }, + { + "bbox": [ + 252, + 496, + 258, + 505 + ], + "score": 0.71, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 259, + 494, + 505, + 508 + ], + "score": 1.0, + "content": "-space according to the time step has not been fully discussed", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 506, + 346, + 519 + ], + "spans": [ + { + "bbox": [ + 106, + 506, + 346, + 519 + ], + "score": 1.0, + "content": "so far. Research on them can be expected to expand further.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30.5, + "bbox_fs": [ + 106, + 494, + 505, + 519 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 536, + 504, + 559 + ], + "lines": [ + { + "bbox": [ + 105, + 535, + 505, + 549 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 505, + 549 + ], + "score": 1.0, + "content": "Limitations. Editing with Asyrp seldom yields changes in overall style or peripheral objects but", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 547, + 498, + 560 + ], + "spans": [ + { + "bbox": [ + 106, + 547, + 498, + 560 + ], + "score": 1.0, + "content": "edits attributes of the main object. Style transfer using frozen diffusion models is our future work.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32.5, + "bbox_fs": [ + 105, + 535, + 505, + 560 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 578, + 505, + 644 + ], + "lines": [ + { + "bbox": [ + 105, + 577, + 505, + 591 + ], + "spans": [ + { + "bbox": [ + 105, + 577, + 505, + 591 + ], + "score": 1.0, + "content": "Societal impact / Ethics statement. Techniques for high-quality image manipulation such as", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 588, + 505, + 602 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 505, + 602 + ], + "score": 1.0, + "content": "Asyrp should be accompanied by social and/or technical solutions to prevent abuse. We acknowledge", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 600, + 506, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 600, + 506, + 613 + ], + "score": 1.0, + "content": "the potential ethical implications that may arise from the use of our image manipulation technique,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 611, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 611, + 506, + 624 + ], + "score": 1.0, + "content": "Asyrp. We advocate for the development and implementation of social and technical solutions to", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "score": 1.0, + "content": "prevent potential abuses such as spreading disinformation or propaganda. We are committed to en-", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 633, + 477, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 633, + 477, + 646 + ], + "score": 1.0, + "content": "suring fairness and non-discrimination, legal compliance, and research integrity in our work.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 36.5, + "bbox_fs": [ + 105, + 577, + 506, + 646 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 667, + 243, + 680 + ], + "lines": [ + { + "bbox": [ + 106, + 666, + 245, + 682 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 245, + 682 + ], + "score": 1.0, + "content": "C PROOF OF THEOREM 1", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 40 + }, + { + "type": "text", + "bbox": [ + 108, + 696, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 695, + 506, + 710 + ], + "spans": [ + { + "bbox": [ + 105, + 695, + 211, + 710 + ], + "score": 1.0, + "content": "Proof of Theorem 1. Let", + "type": "text" + }, + { + "bbox": [ + 212, + 696, + 222, + 708 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 222, + 695, + 457, + 710 + ], + "score": 1.0, + "content": "be a predicted noise during the original reverse process at", + "type": "text" + }, + { + "bbox": [ + 457, + 698, + 463, + 707 + ], + "score": 0.76, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 463, + 695, + 481, + 710 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 481, + 696, + 491, + 708 + ], + "score": 0.88, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 492, + 695, + 506, + 710 + ], + "score": 1.0, + "content": "be", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 706, + 505, + 721 + ], + "spans": [ + { + "bbox": [ + 105, + 706, + 222, + 721 + ], + "score": 1.0, + "content": "its shifted counterpart. Then,", + "type": "text" + }, + { + "bbox": [ + 223, + 708, + 305, + 719 + ], + "score": 0.92, + "content": "\\Delta \\pmb { x } _ { t } = \\tilde { \\pmb { x } } _ { t - 1 } - \\pmb { x } _ { t - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 306, + 706, + 383, + 721 + ], + "score": 1.0, + "content": "is negligible where", + "type": "text" + }, + { + "bbox": [ + 384, + 707, + 505, + 721 + ], + "score": 0.93, + "content": "\\tilde { \\pmb { x } } _ { t - 1 } = \\bar { \\sqrt { \\alpha _ { t - 1 } } } \\mathbf { P } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } \\bar { \\Big ) } ) +", + "type": "inline_equation" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 717, + 159, + 735 + ], + "spans": [ + { + "bbox": [ + 106, + 719, + 154, + 733 + ], + "score": 0.91, + "content": "\\mathbf { D } _ { t } ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\pmb x } _ { t } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 717, + 159, + 735 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42, + "bbox_fs": [ + 105, + 695, + 506, + 735 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 190, + 83, + 419, + 247 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 190, + 83, + 419, + 247 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 190, + 83, + 419, + 247 + ], + "spans": [ + { + "bbox": [ + 190, + 83, + 419, + 247 + ], + "score": 0.966, + "type": "image", + "image_path": "278cf3f351f5df6dbbb8f4ac91469563ba5dc86a95e75f069873149e24faa7f4.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 190, + 83, + 419, + 97.9090909090909 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 190, + 97.9090909090909, + 419, + 112.81818181818181 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 190, + 112.81818181818181, + 419, + 127.72727272727272 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 190, + 127.72727272727272, + 419, + 142.63636363636363 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 190, + 142.63636363636363, + 419, + 157.54545454545453 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 190, + 157.54545454545453, + 419, + 172.45454545454544 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 190, + 172.45454545454544, + 419, + 187.36363636363635 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 190, + 187.36363636363635, + 419, + 202.27272727272725 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 190, + 202.27272727272725, + 419, + 217.18181818181816 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 190, + 217.18181818181816, + 419, + 232.09090909090907 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 190, + 232.09090909090907, + 419, + 246.99999999999997 + ], + "spans": [], + "index": 10 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 260, + 505, + 305 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 259, + 504, + 273 + ], + "spans": [ + { + "bbox": [ + 105, + 259, + 440, + 273 + ], + "score": 1.0, + "content": "Figure 12: Illustration of Theorem 1. Upper blue line describes applying noise", + "type": "text" + }, + { + "bbox": [ + 440, + 261, + 504, + 271 + ], + "score": 0.91, + "content": "\\tilde { \\epsilon } _ { t } = \\epsilon _ { t } + \\Delta \\epsilon _ { t }", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 271, + 505, + 283 + ], + "spans": [ + { + "bbox": [ + 105, + 271, + 151, + 283 + ], + "score": 1.0, + "content": "to produce", + "type": "text" + }, + { + "bbox": [ + 152, + 271, + 191, + 283 + ], + "score": 0.9, + "content": "\\mathbf { P } _ { t } ( \\tilde { \\epsilon } _ { t } ) =", + "type": "inline_equation" + }, + { + "bbox": [ + 192, + 271, + 276, + 283 + ], + "score": 1.0, + "content": "the shifted predicted", + "type": "text" + }, + { + "bbox": [ + 276, + 272, + 288, + 282 + ], + "score": 0.86, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 289, + 271, + 396, + 283 + ], + "score": 1.0, + "content": ". However, the shift due to", + "type": "text" + }, + { + "bbox": [ + 396, + 272, + 414, + 282 + ], + "score": 0.9, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 414, + 271, + 505, + 283 + ], + "score": 1.0, + "content": "is canceled out by the", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 281, + 506, + 295 + ], + "spans": [ + { + "bbox": [ + 105, + 281, + 136, + 295 + ], + "score": 1.0, + "content": "shift in", + "type": "text" + }, + { + "bbox": [ + 137, + 282, + 166, + 294 + ], + "score": 0.92, + "content": "\\mathbf { D } _ { t } ( \\tilde { \\epsilon } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 166, + 281, + 194, + 295 + ], + "score": 1.0, + "content": "due to", + "type": "text" + }, + { + "bbox": [ + 194, + 282, + 211, + 293 + ], + "score": 0.89, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 212, + 281, + 303, + 295 + ], + "score": 1.0, + "content": ". As a results, applying", + "type": "text" + }, + { + "bbox": [ + 303, + 282, + 320, + 293 + ], + "score": 0.9, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 321, + 281, + 353, + 295 + ], + "score": 1.0, + "content": "both on", + "type": "text" + }, + { + "bbox": [ + 354, + 282, + 366, + 293 + ], + "score": 0.89, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 367, + 281, + 384, + 295 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 385, + 282, + 398, + 293 + ], + "score": 0.89, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 398, + 281, + 506, + 295 + ], + "score": 1.0, + "content": "brings identical outputs to", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 293, + 158, + 306 + ], + "spans": [ + { + "bbox": [ + 105, + 293, + 158, + 306 + ], + "score": 1.0, + "content": "the original.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 12.5 + } + ], + "index": 8.75 + }, + { + "type": "text", + "bbox": [ + 106, + 330, + 505, + 368 + ], + "lines": [ + { + "bbox": [ + 104, + 329, + 506, + 348 + ], + "spans": [ + { + "bbox": [ + 104, + 329, + 136, + 348 + ], + "score": 1.0, + "content": "Define", + "type": "text" + }, + { + "bbox": [ + 137, + 331, + 238, + 344 + ], + "score": 0.89, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\bf { x } } _ { t } ) = \\epsilon _ { t } ^ { \\theta } ( { \\bf { x } } _ { t } ) + \\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 329, + 243, + 348 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 243, + 331, + 411, + 344 + ], + "score": 0.91, + "content": "\\{ \\beta _ { t } \\} _ { t = 1 } ^ { T } = \\{ \\beta _ { 1 } = \\beta _ { \\operatorname* { m i n } } , . . . , \\beta _ { T } = \\beta _ { \\operatorname* { m a x } } \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 411, + 329, + 433, + 348 + ], + "score": 1.0, + "content": ", and", + "type": "text" + }, + { + "bbox": [ + 433, + 330, + 506, + 344 + ], + "score": 0.92, + "content": "\\begin{array} { r } { \\alpha _ { t } = \\prod _ { s = 1 } ^ { t } ( 1 - } \\end{array}", + "type": "inline_equation" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 340, + 507, + 360 + ], + "spans": [ + { + "bbox": [ + 106, + 344, + 119, + 356 + ], + "score": 0.79, + "content": "\\beta _ { s , \\ }", + "type": "inline_equation" + }, + { + "bbox": [ + 119, + 340, + 166, + 360 + ], + "score": 1.0, + "content": "). Note that", + "type": "text" + }, + { + "bbox": [ + 167, + 345, + 186, + 356 + ], + "score": 0.93, + "content": "\\beta _ { \\mathrm { m a x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 186, + 340, + 321, + 360 + ], + "score": 1.0, + "content": "is defined as a small value (e.g.,", + "type": "text" + }, + { + "bbox": [ + 321, + 345, + 381, + 356 + ], + "score": 0.88, + "content": "\\beta _ { \\mathrm { m a x } } = 0 . 0 0 1 )", + "type": "inline_equation" + }, + { + "bbox": [ + 382, + 340, + 402, + 360 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 402, + 344, + 435, + 356 + ], + "score": 0.93, + "content": "\\{ \\beta _ { t } \\} _ { t = 1 } ^ { T }", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 340, + 507, + 360 + ], + "score": 1.0, + "content": "are defined by a", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 354, + 450, + 369 + ], + "spans": [ + { + "bbox": [ + 106, + 354, + 211, + 369 + ], + "score": 1.0, + "content": "decreasing schedule from", + "type": "text" + }, + { + "bbox": [ + 211, + 356, + 255, + 367 + ], + "score": 0.93, + "content": "\\beta _ { T } = \\beta _ { \\mathrm { m a x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 256, + 354, + 267, + 369 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 267, + 356, + 327, + 367 + ], + "score": 0.91, + "content": "\\beta _ { 1 } = \\beta _ { \\mathrm { m i n } } \\approx 0", + "type": "inline_equation" + }, + { + "bbox": [ + 327, + 354, + 350, + 369 + ], + "score": 1.0, + "content": "(e.g.,", + "type": "text" + }, + { + "bbox": [ + 351, + 356, + 416, + 367 + ], + "score": 0.88, + "content": "\\beta _ { \\mathrm { m i n } } = 0 . 0 0 0 0 1 \\rangle", + "type": "inline_equation" + }, + { + "bbox": [ + 417, + 354, + 450, + 369 + ], + "score": 1.0, + "content": ". Then,", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16 + }, + { + "type": "interline_equation", + "bbox": [ + 219, + 384, + 392, + 399 + ], + "lines": [ + { + "bbox": [ + 219, + 384, + 392, + 399 + ], + "spans": [ + { + "bbox": [ + 219, + 384, + 392, + 399 + ], + "score": 0.91, + "content": "\\tilde { \\pmb { x } } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } \\big ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big ) + \\mathbf { D } _ { t } \\big ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big )", + "type": "interline_equation", + "image_path": "01ee3c33e0761b3521ba97b4f84f5ffc229ad08d8505c273d637ac0146bcb0d9.jpg" + } + ] + } + ], + "index": 18, + "virtual_lines": [ + { + "bbox": [ + 219, + 384, + 392, + 399 + ], + "spans": [], + "index": 18 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 142, + 420, + 470, + 454 + ], + "lines": [ + { + "bbox": [ + 142, + 420, + 470, + 454 + ], + "spans": [ + { + "bbox": [ + 142, + 420, + 470, + 454 + ], + "score": 0.93, + "content": "= \\sqrt { \\alpha _ { t - 1 } } \\left( \\frac { x _ { t } - \\sqrt { 1 - \\alpha _ { t } } \\left( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) + \\Delta \\epsilon _ { t } \\right) } { \\sqrt { \\alpha _ { t } } } \\right) + \\sqrt { 1 - \\alpha _ { t - 1 } } \\cdot \\left( \\epsilon _ { t } ^ { ( \\theta ) } \\left( x _ { t } \\right) + \\Delta \\epsilon _ { t } \\right)", + "type": "interline_equation", + "image_path": "2e19aec9451f33520a3a7974bbd5a035ae744237052531f372833c5a99ddff25.jpg" + } + ] + } + ], + "index": 20, + "virtual_lines": [ + { + "bbox": [ + 142, + 420, + 470, + 431.3333333333333 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 142, + 431.3333333333333, + 470, + 442.66666666666663 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 142, + 442.66666666666663, + 470, + 453.99999999999994 + ], + "spans": [], + "index": 21 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 131, + 469, + 465, + 497 + ], + "lines": [ + { + "bbox": [ + 131, + 469, + 465, + 497 + ], + "spans": [ + { + "bbox": [ + 131, + 469, + 465, + 497 + ], + "score": 0.9, + "content": "= \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) + \\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) - \\frac { \\sqrt { \\alpha _ { t - 1 } } \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { \\alpha _ { t } } } \\cdot \\Delta \\epsilon _ { t } + \\sqrt { 1 - \\alpha _ { t - 1 } } \\cdot \\Delta \\epsilon _ { t }", + "type": "interline_equation", + "image_path": "f8f1d07701d3604aa1e847f9afe7750a0c182028698c917afefce6e9a09b95c4.jpg" + } + ] + } + ], + "index": 22, + "virtual_lines": [ + { + "bbox": [ + 131, + 469, + 465, + 497 + ], + "spans": [], + "index": 22 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 212, + 513, + 401, + 540 + ], + "lines": [ + { + "bbox": [ + 212, + 513, + 401, + 540 + ], + "spans": [ + { + "bbox": [ + 212, + 513, + 401, + 540 + ], + "score": 0.93, + "content": "= x _ { t - 1 } + \\left( - \\frac { \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } + \\sqrt { 1 - \\alpha _ { t - 1 } } \\right) \\cdot \\Delta \\epsilon _ { t }", + "type": "interline_equation", + "image_path": "160589c53833ae678a3010268005cbbbdf94a10e75615dc539618b396960f86e.jpg" + } + ] + } + ], + "index": 23, + "virtual_lines": [ + { + "bbox": [ + 212, + 513, + 401, + 540 + ], + "spans": [], + "index": 23 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 172, + 555, + 441, + 596 + ], + "lines": [ + { + "bbox": [ + 172, + 555, + 441, + 596 + ], + "spans": [ + { + "bbox": [ + 172, + 555, + 441, + 596 + ], + "score": 0.94, + "content": "= x _ { t - 1 } + \\left( - \\frac { \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } + \\frac { \\sqrt { 1 - \\prod _ { s = 1 } ^ { t - 1 } \\left( 1 - \\beta _ { s } \\right) } \\sqrt { 1 - \\beta _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } \\right) \\cdot \\Delta \\epsilon _ { t }", + "type": "interline_equation", + "image_path": "94f6bb57ed66ae0aef5a4609c429c9af874ae238470b1c6b5a2328050c6645f2.jpg" + } + ] + } + ], + "index": 25, + "virtual_lines": [ + { + "bbox": [ + 172, + 555, + 441, + 568.6666666666666 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 172, + 568.6666666666666, + 441, + 582.3333333333333 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 172, + 582.3333333333333, + 441, + 595.9999999999999 + ], + "spans": [], + "index": 26 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 163, + 610, + 451, + 644 + ], + "lines": [ + { + "bbox": [ + 163, + 610, + 451, + 644 + ], + "spans": [ + { + "bbox": [ + 163, + 610, + 451, + 644 + ], + "score": 0.92, + "content": "= x _ { t - 1 } + \\left( \\frac { \\sqrt { 1 - \\alpha _ { t } - \\beta _ { t } } - \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } \\right) \\cdot \\Delta \\epsilon _ { t } \\quad \\cdot : \\alpha _ { t } = \\prod _ { s = 1 } ^ { t } \\left( 1 - \\beta _ { s } \\right)", + "type": "interline_equation", + "image_path": "dcc07c1a0353bb620e8ab796ccc6fe1ae0e1a63406cf1e6326a0873ee1a876b7.jpg" + } + ] + } + ], + "index": 28, + "virtual_lines": [ + { + "bbox": [ + 163, + 610, + 451, + 621.3333333333334 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 163, + 621.3333333333334, + 451, + 632.6666666666667 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 163, + 632.6666666666667, + 451, + 644.0000000000001 + ], + "spans": [], + "index": 29 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 118, + 665, + 491, + 693 + ], + "lines": [ + { + "bbox": [ + 118, + 665, + 491, + 693 + ], + "spans": [ + { + "bbox": [ + 118, + 665, + 491, + 693 + ], + "score": 0.93, + "content": "\\therefore \\Delta x _ { t } = { \\tilde { x } } _ { t - 1 } - x _ { t - 1 } = \\left( { \\frac { { \\sqrt { 1 - \\alpha _ { t } - \\beta _ { t } } } - { \\sqrt { 1 - \\alpha _ { t } } } } { { \\sqrt { 1 - \\beta _ { t } } } } } \\right) \\cdot \\Delta \\epsilon _ { t } { \\mathrm { ~ i s ~ n e g l i g i b l e ~ } } \\quad \\cdot \\cdot \\beta _ { t } < \\beta _ { m a x }", + "type": "interline_equation", + "image_path": "f1b17f28b24912a779196b70a7b90dc23a4b070bc0ac53c342a45f37da7115e9.jpg" + } + ] + } + ], + "index": 30, + "virtual_lines": [ + { + "bbox": [ + 118, + 665, + 491, + 693 + ], + "spans": [], + "index": 30 + } + ] + } + ], + "page_idx": 15, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 761 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "16", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 494, + 708, + 504, + 718 + ], + "lines": [ + { + "bbox": [ + 497, + 709, + 504, + 717 + ], + "spans": [ + { + "bbox": [ + 497, + 709, + 504, + 717 + ], + "score": 0.997, + "content": "□", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 190, + 83, + 419, + 247 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 190, + 83, + 419, + 247 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 190, + 83, + 419, + 247 + ], + "spans": [ + { + "bbox": [ + 190, + 83, + 419, + 247 + ], + "score": 0.966, + "type": "image", + "image_path": "278cf3f351f5df6dbbb8f4ac91469563ba5dc86a95e75f069873149e24faa7f4.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 190, + 83, + 419, + 97.9090909090909 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 190, + 97.9090909090909, + 419, + 112.81818181818181 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 190, + 112.81818181818181, + 419, + 127.72727272727272 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 190, + 127.72727272727272, + 419, + 142.63636363636363 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 190, + 142.63636363636363, + 419, + 157.54545454545453 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 190, + 157.54545454545453, + 419, + 172.45454545454544 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 190, + 172.45454545454544, + 419, + 187.36363636363635 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 190, + 187.36363636363635, + 419, + 202.27272727272725 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 190, + 202.27272727272725, + 419, + 217.18181818181816 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 190, + 217.18181818181816, + 419, + 232.09090909090907 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 190, + 232.09090909090907, + 419, + 246.99999999999997 + ], + "spans": [], + "index": 10 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 260, + 505, + 305 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 259, + 504, + 273 + ], + "spans": [ + { + "bbox": [ + 105, + 259, + 440, + 273 + ], + "score": 1.0, + "content": "Figure 12: Illustration of Theorem 1. Upper blue line describes applying noise", + "type": "text" + }, + { + "bbox": [ + 440, + 261, + 504, + 271 + ], + "score": 0.91, + "content": "\\tilde { \\epsilon } _ { t } = \\epsilon _ { t } + \\Delta \\epsilon _ { t }", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 271, + 505, + 283 + ], + "spans": [ + { + "bbox": [ + 105, + 271, + 151, + 283 + ], + "score": 1.0, + "content": "to produce", + "type": "text" + }, + { + "bbox": [ + 152, + 271, + 191, + 283 + ], + "score": 0.9, + "content": "\\mathbf { P } _ { t } ( \\tilde { \\epsilon } _ { t } ) =", + "type": "inline_equation" + }, + { + "bbox": [ + 192, + 271, + 276, + 283 + ], + "score": 1.0, + "content": "the shifted predicted", + "type": "text" + }, + { + "bbox": [ + 276, + 272, + 288, + 282 + ], + "score": 0.86, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 289, + 271, + 396, + 283 + ], + "score": 1.0, + "content": ". However, the shift due to", + "type": "text" + }, + { + "bbox": [ + 396, + 272, + 414, + 282 + ], + "score": 0.9, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 414, + 271, + 505, + 283 + ], + "score": 1.0, + "content": "is canceled out by the", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 281, + 506, + 295 + ], + "spans": [ + { + "bbox": [ + 105, + 281, + 136, + 295 + ], + "score": 1.0, + "content": "shift in", + "type": "text" + }, + { + "bbox": [ + 137, + 282, + 166, + 294 + ], + "score": 0.92, + "content": "\\mathbf { D } _ { t } ( \\tilde { \\epsilon } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 166, + 281, + 194, + 295 + ], + "score": 1.0, + "content": "due to", + "type": "text" + }, + { + "bbox": [ + 194, + 282, + 211, + 293 + ], + "score": 0.89, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 212, + 281, + 303, + 295 + ], + "score": 1.0, + "content": ". As a results, applying", + "type": "text" + }, + { + "bbox": [ + 303, + 282, + 320, + 293 + ], + "score": 0.9, + "content": "\\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 321, + 281, + 353, + 295 + ], + "score": 1.0, + "content": "both on", + "type": "text" + }, + { + "bbox": [ + 354, + 282, + 366, + 293 + ], + "score": 0.89, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 367, + 281, + 384, + 295 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 385, + 282, + 398, + 293 + ], + "score": 0.89, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 398, + 281, + 506, + 295 + ], + "score": 1.0, + "content": "brings identical outputs to", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 293, + 158, + 306 + ], + "spans": [ + { + "bbox": [ + 105, + 293, + 158, + 306 + ], + "score": 1.0, + "content": "the original.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 12.5 + } + ], + "index": 8.75 + }, + { + "type": "text", + "bbox": [ + 106, + 330, + 505, + 368 + ], + "lines": [ + { + "bbox": [ + 104, + 329, + 506, + 348 + ], + "spans": [ + { + "bbox": [ + 104, + 329, + 136, + 348 + ], + "score": 1.0, + "content": "Define", + "type": "text" + }, + { + "bbox": [ + 137, + 331, + 238, + 344 + ], + "score": 0.89, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta } ( { \\bf { x } } _ { t } ) = \\epsilon _ { t } ^ { \\theta } ( { \\bf { x } } _ { t } ) + \\Delta \\epsilon _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 329, + 243, + 348 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 243, + 331, + 411, + 344 + ], + "score": 0.91, + "content": "\\{ \\beta _ { t } \\} _ { t = 1 } ^ { T } = \\{ \\beta _ { 1 } = \\beta _ { \\operatorname* { m i n } } , . . . , \\beta _ { T } = \\beta _ { \\operatorname* { m a x } } \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 411, + 329, + 433, + 348 + ], + "score": 1.0, + "content": ", and", + "type": "text" + }, + { + "bbox": [ + 433, + 330, + 506, + 344 + ], + "score": 0.92, + "content": "\\begin{array} { r } { \\alpha _ { t } = \\prod _ { s = 1 } ^ { t } ( 1 - } \\end{array}", + "type": "inline_equation" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 340, + 507, + 360 + ], + "spans": [ + { + "bbox": [ + 106, + 344, + 119, + 356 + ], + "score": 0.79, + "content": "\\beta _ { s , \\ }", + "type": "inline_equation" + }, + { + "bbox": [ + 119, + 340, + 166, + 360 + ], + "score": 1.0, + "content": "). Note that", + "type": "text" + }, + { + "bbox": [ + 167, + 345, + 186, + 356 + ], + "score": 0.93, + "content": "\\beta _ { \\mathrm { m a x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 186, + 340, + 321, + 360 + ], + "score": 1.0, + "content": "is defined as a small value (e.g.,", + "type": "text" + }, + { + "bbox": [ + 321, + 345, + 381, + 356 + ], + "score": 0.88, + "content": "\\beta _ { \\mathrm { m a x } } = 0 . 0 0 1 )", + "type": "inline_equation" + }, + { + "bbox": [ + 382, + 340, + 402, + 360 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 402, + 344, + 435, + 356 + ], + "score": 0.93, + "content": "\\{ \\beta _ { t } \\} _ { t = 1 } ^ { T }", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 340, + 507, + 360 + ], + "score": 1.0, + "content": "are defined by a", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 354, + 450, + 369 + ], + "spans": [ + { + "bbox": [ + 106, + 354, + 211, + 369 + ], + "score": 1.0, + "content": "decreasing schedule from", + "type": "text" + }, + { + "bbox": [ + 211, + 356, + 255, + 367 + ], + "score": 0.93, + "content": "\\beta _ { T } = \\beta _ { \\mathrm { m a x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 256, + 354, + 267, + 369 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 267, + 356, + 327, + 367 + ], + "score": 0.91, + "content": "\\beta _ { 1 } = \\beta _ { \\mathrm { m i n } } \\approx 0", + "type": "inline_equation" + }, + { + "bbox": [ + 327, + 354, + 350, + 369 + ], + "score": 1.0, + "content": "(e.g.,", + "type": "text" + }, + { + "bbox": [ + 351, + 356, + 416, + 367 + ], + "score": 0.88, + "content": "\\beta _ { \\mathrm { m i n } } = 0 . 0 0 0 0 1 \\rangle", + "type": "inline_equation" + }, + { + "bbox": [ + 417, + 354, + 450, + 369 + ], + "score": 1.0, + "content": ". Then,", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16, + "bbox_fs": [ + 104, + 329, + 507, + 369 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 219, + 384, + 392, + 399 + ], + "lines": [ + { + "bbox": [ + 219, + 384, + 392, + 399 + ], + "spans": [ + { + "bbox": [ + 219, + 384, + 392, + 399 + ], + "score": 0.91, + "content": "\\tilde { \\pmb { x } } _ { t - 1 } = \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } \\big ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big ) + \\mathbf { D } _ { t } \\big ( \\tilde { \\epsilon } _ { t } ^ { \\theta } ( \\pmb { x } _ { t } ) \\big )", + "type": "interline_equation", + "image_path": "01ee3c33e0761b3521ba97b4f84f5ffc229ad08d8505c273d637ac0146bcb0d9.jpg" + } + ] + } + ], + "index": 18, + "virtual_lines": [ + { + "bbox": [ + 219, + 384, + 392, + 399 + ], + "spans": [], + "index": 18 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 142, + 420, + 470, + 454 + ], + "lines": [ + { + "bbox": [ + 142, + 420, + 470, + 454 + ], + "spans": [ + { + "bbox": [ + 142, + 420, + 470, + 454 + ], + "score": 0.93, + "content": "= \\sqrt { \\alpha _ { t - 1 } } \\left( \\frac { x _ { t } - \\sqrt { 1 - \\alpha _ { t } } \\left( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) + \\Delta \\epsilon _ { t } \\right) } { \\sqrt { \\alpha _ { t } } } \\right) + \\sqrt { 1 - \\alpha _ { t - 1 } } \\cdot \\left( \\epsilon _ { t } ^ { ( \\theta ) } \\left( x _ { t } \\right) + \\Delta \\epsilon _ { t } \\right)", + "type": "interline_equation", + "image_path": "2e19aec9451f33520a3a7974bbd5a035ae744237052531f372833c5a99ddff25.jpg" + } + ] + } + ], + "index": 20, + "virtual_lines": [ + { + "bbox": [ + 142, + 420, + 470, + 431.3333333333333 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 142, + 431.3333333333333, + 470, + 442.66666666666663 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 142, + 442.66666666666663, + 470, + 453.99999999999994 + ], + "spans": [], + "index": 21 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 131, + 469, + 465, + 497 + ], + "lines": [ + { + "bbox": [ + 131, + 469, + 465, + 497 + ], + "spans": [ + { + "bbox": [ + 131, + 469, + 465, + 497 + ], + "score": 0.9, + "content": "= \\sqrt { \\alpha _ { t - 1 } } \\mathbf { P } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) + \\mathbf { D } _ { t } ( \\epsilon _ { t } ^ { \\theta } ( x _ { t } ) ) - \\frac { \\sqrt { \\alpha _ { t - 1 } } \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { \\alpha _ { t } } } \\cdot \\Delta \\epsilon _ { t } + \\sqrt { 1 - \\alpha _ { t - 1 } } \\cdot \\Delta \\epsilon _ { t }", + "type": "interline_equation", + "image_path": "f8f1d07701d3604aa1e847f9afe7750a0c182028698c917afefce6e9a09b95c4.jpg" + } + ] + } + ], + "index": 22, + "virtual_lines": [ + { + "bbox": [ + 131, + 469, + 465, + 497 + ], + "spans": [], + "index": 22 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 212, + 513, + 401, + 540 + ], + "lines": [ + { + "bbox": [ + 212, + 513, + 401, + 540 + ], + "spans": [ + { + "bbox": [ + 212, + 513, + 401, + 540 + ], + "score": 0.93, + "content": "= x _ { t - 1 } + \\left( - \\frac { \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } + \\sqrt { 1 - \\alpha _ { t - 1 } } \\right) \\cdot \\Delta \\epsilon _ { t }", + "type": "interline_equation", + "image_path": "160589c53833ae678a3010268005cbbbdf94a10e75615dc539618b396960f86e.jpg" + } + ] + } + ], + "index": 23, + "virtual_lines": [ + { + "bbox": [ + 212, + 513, + 401, + 540 + ], + "spans": [], + "index": 23 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 172, + 555, + 441, + 596 + ], + "lines": [ + { + "bbox": [ + 172, + 555, + 441, + 596 + ], + "spans": [ + { + "bbox": [ + 172, + 555, + 441, + 596 + ], + "score": 0.94, + "content": "= x _ { t - 1 } + \\left( - \\frac { \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } + \\frac { \\sqrt { 1 - \\prod _ { s = 1 } ^ { t - 1 } \\left( 1 - \\beta _ { s } \\right) } \\sqrt { 1 - \\beta _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } \\right) \\cdot \\Delta \\epsilon _ { t }", + "type": "interline_equation", + "image_path": "94f6bb57ed66ae0aef5a4609c429c9af874ae238470b1c6b5a2328050c6645f2.jpg" + } + ] + } + ], + "index": 25, + "virtual_lines": [ + { + "bbox": [ + 172, + 555, + 441, + 568.6666666666666 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 172, + 568.6666666666666, + 441, + 582.3333333333333 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 172, + 582.3333333333333, + 441, + 595.9999999999999 + ], + "spans": [], + "index": 26 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 163, + 610, + 451, + 644 + ], + "lines": [ + { + "bbox": [ + 163, + 610, + 451, + 644 + ], + "spans": [ + { + "bbox": [ + 163, + 610, + 451, + 644 + ], + "score": 0.92, + "content": "= x _ { t - 1 } + \\left( \\frac { \\sqrt { 1 - \\alpha _ { t } - \\beta _ { t } } - \\sqrt { 1 - \\alpha _ { t } } } { \\sqrt { 1 - \\beta _ { t } } } \\right) \\cdot \\Delta \\epsilon _ { t } \\quad \\cdot : \\alpha _ { t } = \\prod _ { s = 1 } ^ { t } \\left( 1 - \\beta _ { s } \\right)", + "type": "interline_equation", + "image_path": "dcc07c1a0353bb620e8ab796ccc6fe1ae0e1a63406cf1e6326a0873ee1a876b7.jpg" + } + ] + } + ], + "index": 28, + "virtual_lines": [ + { + "bbox": [ + 163, + 610, + 451, + 621.3333333333334 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 163, + 621.3333333333334, + 451, + 632.6666666666667 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 163, + 632.6666666666667, + 451, + 644.0000000000001 + ], + "spans": [], + "index": 29 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 118, + 665, + 491, + 693 + ], + "lines": [ + { + "bbox": [ + 118, + 665, + 491, + 693 + ], + "spans": [ + { + "bbox": [ + 118, + 665, + 491, + 693 + ], + "score": 0.93, + "content": "\\therefore \\Delta x _ { t } = { \\tilde { x } } _ { t - 1 } - x _ { t - 1 } = \\left( { \\frac { { \\sqrt { 1 - \\alpha _ { t } - \\beta _ { t } } } - { \\sqrt { 1 - \\alpha _ { t } } } } { { \\sqrt { 1 - \\beta _ { t } } } } } \\right) \\cdot \\Delta \\epsilon _ { t } { \\mathrm { ~ i s ~ n e g l i g i b l e ~ } } \\quad \\cdot \\cdot \\beta _ { t } < \\beta _ { m a x }", + "type": "interline_equation", + "image_path": "f1b17f28b24912a779196b70a7b90dc23a4b070bc0ac53c342a45f37da7115e9.jpg" + } + ] + } + ], + "index": 30, + "virtual_lines": [ + { + "bbox": [ + 118, + 665, + 491, + 693 + ], + "spans": [], + "index": 30 + } + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 81, + 383, + 94 + ], + "lines": [ + { + "bbox": [ + 105, + 79, + 384, + 97 + ], + "spans": [ + { + "bbox": [ + 105, + 79, + 275, + 97 + ], + "score": 1.0, + "content": "D ADDITIONAL SUPPORTS FOR", + "type": "text" + }, + { + "bbox": [ + 275, + 82, + 282, + 93 + ], + "score": 0.59, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 283, + 79, + 384, + 97 + ], + "score": 1.0, + "content": "-space WITH ASYRP", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 106, + 364, + 117 + ], + "lines": [ + { + "bbox": [ + 105, + 106, + 365, + 119 + ], + "spans": [ + { + "bbox": [ + 105, + 106, + 257, + 119 + ], + "score": 1.0, + "content": "D.1 RANDOM PERTURBATION ON", + "type": "text" + }, + { + "bbox": [ + 258, + 109, + 263, + 116 + ], + "score": 0.29, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 263, + 106, + 365, + 119 + ], + "score": 1.0, + "content": "-space WITHOUT ASYRP", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 107, + 126, + 505, + 160 + ], + "lines": [ + { + "bbox": [ + 105, + 127, + 504, + 139 + ], + "spans": [ + { + "bbox": [ + 105, + 127, + 117, + 139 + ], + "score": 1.0, + "content": "In", + "type": "text" + }, + { + "bbox": [ + 117, + 127, + 137, + 138 + ], + "score": 0.6, + "content": "\\ S 3 . 1", + "type": "inline_equation" + }, + { + "bbox": [ + 138, + 127, + 225, + 139 + ], + "score": 1.0, + "content": ", we argue that if both", + "type": "text" + }, + { + "bbox": [ + 225, + 127, + 238, + 138 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 238, + 127, + 255, + 139 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 256, + 127, + 269, + 138 + ], + "score": 0.89, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 127, + 408, + 139 + ], + "score": 1.0, + "content": "are shifted, we can not manipulate", + "type": "text" + }, + { + "bbox": [ + 409, + 128, + 421, + 138 + ], + "score": 0.85, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 421, + 127, + 504, + 139 + ], + "score": 1.0, + "content": ". In Figure 13, we do", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 137, + 506, + 151 + ], + "spans": [ + { + "bbox": [ + 105, + 137, + 506, + 151 + ], + "score": 1.0, + "content": "not observe the noticeable difference between (a) and (b) which are the result of the original reverse", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 149, + 380, + 162 + ], + "spans": [ + { + "bbox": [ + 105, + 149, + 380, + 162 + ], + "score": 1.0, + "content": "process of DDIM and the one with shifting both terms, respectively.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3 + }, + { + "type": "title", + "bbox": [ + 108, + 173, + 421, + 185 + ], + "lines": [ + { + "bbox": [ + 105, + 172, + 422, + 187 + ], + "spans": [ + { + "bbox": [ + 105, + 172, + 277, + 187 + ], + "score": 1.0, + "content": "D.2 ROBUSTNESS AND SEMANTICS IN", + "type": "text" + }, + { + "bbox": [ + 277, + 174, + 284, + 183 + ], + "score": 0.7, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 284, + 172, + 331, + 187 + ], + "score": 1.0, + "content": "-space AND", + "type": "text" + }, + { + "bbox": [ + 332, + 176, + 337, + 183 + ], + "score": 0.5, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 338, + 172, + 422, + 187 + ], + "score": 1.0, + "content": "-space WITH ASYRP", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 107, + 194, + 505, + 238 + ], + "lines": [ + { + "bbox": [ + 105, + 193, + 505, + 207 + ], + "spans": [ + { + "bbox": [ + 105, + 193, + 118, + 207 + ], + "score": 1.0, + "content": "In", + "type": "text" + }, + { + "bbox": [ + 118, + 194, + 141, + 206 + ], + "score": 0.8, + "content": "\\mathrm { ~ \\ S ~ } 3 . 2", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 193, + 226, + 207 + ], + "score": 1.0, + "content": ", we also argue that", + "type": "text" + }, + { + "bbox": [ + 226, + 195, + 232, + 204 + ], + "score": 0.73, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 233, + 193, + 345, + 207 + ], + "score": 1.0, + "content": "-space is more robust than", + "type": "text" + }, + { + "bbox": [ + 346, + 196, + 352, + 204 + ], + "score": 0.6, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 352, + 193, + 505, + 207 + ], + "score": 1.0, + "content": "-space with Asyrp. In Figure 13, we", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 204, + 506, + 218 + ], + "spans": [ + { + "bbox": [ + 105, + 204, + 245, + 218 + ], + "score": 1.0, + "content": "observe that small random noise", + "type": "text" + }, + { + "bbox": [ + 245, + 205, + 301, + 217 + ], + "score": 0.93, + "content": "z \\sim \\mathcal { N } ( 0 , \\mathbf { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 301, + 204, + 315, + 218 + ], + "score": 1.0, + "content": "in", + "type": "text" + }, + { + "bbox": [ + 316, + 207, + 322, + 216 + ], + "score": 0.64, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 322, + 204, + 506, + 218 + ], + "score": 1.0, + "content": "-space degrades the resulting image without", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 216, + 505, + 228 + ], + "spans": [ + { + "bbox": [ + 105, + 216, + 335, + 228 + ], + "score": 1.0, + "content": "semantic changes (c) and much larger random noise in", + "type": "text" + }, + { + "bbox": [ + 335, + 217, + 342, + 226 + ], + "score": 0.75, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 342, + 216, + 505, + 228 + ], + "score": 1.0, + "content": "-space yields random semantic changes", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 227, + 218, + 239 + ], + "spans": [ + { + "bbox": [ + 106, + 227, + 218, + 239 + ], + "score": 1.0, + "content": "without severe artifacts (d).", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 107, + 244, + 505, + 333 + ], + "lines": [ + { + "bbox": [ + 106, + 244, + 505, + 255 + ], + "spans": [ + { + "bbox": [ + 106, + 244, + 505, + 255 + ], + "score": 1.0, + "content": "Note that diffusion models are designed as latent variable models with learned Gaussian transitions", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 104, + 252, + 504, + 270 + ], + "spans": [ + { + "bbox": [ + 104, + 252, + 493, + 270 + ], + "score": 1.0, + "content": "and the reverse process should also be close to Gaussian. Based on the assumption, we manage", + "type": "text" + }, + { + "bbox": [ + 493, + 254, + 504, + 267 + ], + "score": 0.87, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 263, + 506, + 280 + ], + "spans": [ + { + "bbox": [ + 104, + 263, + 367, + 280 + ], + "score": 1.0, + "content": "to follow the original Gaussian distribution as follows. Adding", + "type": "text" + }, + { + "bbox": [ + 367, + 266, + 428, + 278 + ], + "score": 0.93, + "content": "z \\sim \\mathcal { N } ( 0 , \\sigma \\bar { \\bf { I } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 263, + 441, + 280 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 441, + 265, + 452, + 278 + ], + "score": 0.88, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 452, + 263, + 506, + 280 + ], + "score": 1.0, + "content": "expands the", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 276, + 506, + 290 + ], + "spans": [ + { + "bbox": [ + 105, + 276, + 506, + 290 + ], + "score": 1.0, + "content": "distribution of the predicted noise and may produce distorted images. To preserve the distribution,√", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 288, + 506, + 302 + ], + "spans": [ + { + "bbox": [ + 105, + 288, + 143, + 302 + ], + "score": 1.0, + "content": "we scale", + "type": "text" + }, + { + "bbox": [ + 144, + 288, + 248, + 301 + ], + "score": 0.91, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta } = ( \\epsilon _ { t } ^ { \\theta } + z ) / \\sqrt { 1 ^ { 2 } + { \\sigma } ^ { 2 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 288, + 506, + 302 + ], + "score": 1.0, + "content": ". Still, the resulting images are almost identical compared to the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 104, + 298, + 506, + 314 + ], + "spans": [ + { + "bbox": [ + 104, + 298, + 198, + 314 + ], + "score": 1.0, + "content": "original images where", + "type": "text" + }, + { + "bbox": [ + 198, + 300, + 249, + 312 + ], + "score": 0.9, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta } = \\epsilon _ { t } ^ { \\theta } + z", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 298, + 506, + 314 + ], + "score": 1.0, + "content": "as shown in Figure 13. It is no wonder that the scaling does not", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 311, + 505, + 323 + ], + "spans": [ + { + "bbox": [ + 105, + 311, + 505, + 323 + ], + "score": 1.0, + "content": "improve the distorted results since the additive random noise disturbs the denoising operation of the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 322, + 206, + 334 + ], + "spans": [ + { + "bbox": [ + 105, + 322, + 206, + 334 + ], + "score": 1.0, + "content": "predicted random noise.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 13.5 + }, + { + "type": "image", + "bbox": [ + 110, + 345, + 500, + 433 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 110, + 345, + 500, + 433 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 110, + 345, + 500, + 433 + ], + "spans": [ + { + "bbox": [ + 110, + 345, + 500, + 433 + ], + "score": 0.961, + "type": "image", + "image_path": "bb209ea4fa736576560335d535c5528804419b23be9234941f9e22111f467649.jpg" + } + ] + } + ], + "index": 19, + "virtual_lines": [ + { + "bbox": [ + 110, + 345, + 500, + 374.3333333333333 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 110, + 374.3333333333333, + 500, + 403.66666666666663 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 110, + 403.66666666666663, + 500, + 432.99999999999994 + ], + "spans": [], + "index": 20 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 444, + 505, + 522 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 444, + 506, + 458 + ], + "spans": [ + { + "bbox": [ + 105, + 444, + 506, + 458 + ], + "score": 1.0, + "content": "Figure 13: (a) The reconstructed image by the original DDIM inversion process which is almost", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 455, + 504, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 415, + 468 + ], + "score": 1.0, + "content": "indistinguishable to the real image. (b) The result from adding random noise", + "type": "text" + }, + { + "bbox": [ + 416, + 457, + 423, + 466 + ], + "score": 0.76, + "content": "z", + "type": "inline_equation" + }, + { + "bbox": [ + 423, + 455, + 456, + 468 + ], + "score": 1.0, + "content": "both on", + "type": "text" + }, + { + "bbox": [ + 456, + 456, + 469, + 467 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 470, + 455, + 487, + 468 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 488, + 456, + 501, + 467 + ], + "score": 0.88, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 501, + 455, + 504, + 468 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 466, + 505, + 479 + ], + "spans": [ + { + "bbox": [ + 106, + 466, + 505, + 479 + ], + "score": 1.0, + "content": "(a) looks identical to (b). The insets of (b, c) depict the full SSIM image from (a). It shows that", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 104, + 476, + 506, + 492 + ], + "spans": [ + { + "bbox": [ + 104, + 476, + 168, + 492 + ], + "score": 1.0, + "content": "simply shifting", + "type": "text" + }, + { + "bbox": [ + 168, + 477, + 179, + 489 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 179, + 476, + 385, + 492 + ], + "score": 1.0, + "content": "without Asyrp does not affect the result. (c) Adding", + "type": "text" + }, + { + "bbox": [ + 386, + 478, + 415, + 487 + ], + "score": 0.9, + "content": "z \\sim \\mathcal { N }", + "type": "inline_equation" + }, + { + "bbox": [ + 415, + 476, + 426, + 492 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 426, + 479, + 432, + 487 + ], + "score": 0.63, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 432, + 476, + 506, + 492 + ], + "score": 1.0, + "content": "-space with Asyrp", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 488, + 505, + 502 + ], + "spans": [ + { + "bbox": [ + 105, + 488, + 356, + 502 + ], + "score": 1.0, + "content": "easily degrades image with little semantic change. (d) Adding", + "type": "text" + }, + { + "bbox": [ + 356, + 489, + 386, + 499 + ], + "score": 0.9, + "content": "z \\sim \\mathcal { N }", + "type": "inline_equation" + }, + { + "bbox": [ + 387, + 488, + 397, + 502 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 398, + 489, + 405, + 499 + ], + "score": 0.82, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 405, + 488, + 505, + 502 + ], + "score": 1.0, + "content": "-space with Asyrp yields", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 104, + 499, + 506, + 512 + ], + "spans": [ + { + "bbox": [ + 104, + 499, + 506, + 512 + ], + "score": 1.0, + "content": "random semantic change without image degradation. (e) Correlation between image degradation and", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 511, + 237, + 523 + ], + "spans": [ + { + "bbox": [ + 106, + 511, + 237, + 523 + ], + "score": 1.0, + "content": "noise strength in the two spaces.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 24 + } + ], + "index": 21.5 + }, + { + "type": "title", + "bbox": [ + 108, + 546, + 259, + 558 + ], + "lines": [ + { + "bbox": [ + 105, + 544, + 261, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 182, + 560 + ], + "score": 1.0, + "content": "D.3 CHOICE OF", + "type": "text" + }, + { + "bbox": [ + 182, + 547, + 189, + 556 + ], + "score": 0.53, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 544, + 261, + 560 + ], + "score": 1.0, + "content": "-space IN U-NET", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 106, + 567, + 505, + 666 + ], + "lines": [ + { + "bbox": [ + 106, + 567, + 505, + 579 + ], + "spans": [ + { + "bbox": [ + 106, + 567, + 346, + 579 + ], + "score": 1.0, + "content": "As shown in Figure 14, there are many other candidates for", + "type": "text" + }, + { + "bbox": [ + 347, + 568, + 353, + 577 + ], + "score": 0.69, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 353, + 567, + 505, + 579 + ], + "score": 1.0, + "content": "-space in the architecture. Among the", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 578, + 504, + 590 + ], + "spans": [ + { + "bbox": [ + 106, + 578, + 504, + 590 + ], + "score": 1.0, + "content": "layers, we choose the 8th layer, the bridge of the U-Net based architecture. The layer is not influ-", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 590, + 505, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 590, + 505, + 601 + ], + "score": 1.0, + "content": "enced by any skip connection, has the smallest spatial dimension with compressed information, and", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 599, + 505, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 599, + 505, + 613 + ], + "score": 1.0, + "content": "is located just before the upsampling blocks. Thus, we assume that it could possibly be considered as", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 610, + 505, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 396, + 624 + ], + "score": 1.0, + "content": "the most suitable latent embedding. To confirm the assumption, we train", + "type": "text" + }, + { + "bbox": [ + 397, + 611, + 407, + 622 + ], + "score": 0.88, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 407, + 610, + 505, + 624 + ], + "score": 1.0, + "content": "on the other layers. The", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 104, + 620, + 506, + 636 + ], + "spans": [ + { + "bbox": [ + 104, + 620, + 425, + 636 + ], + "score": 1.0, + "content": "results are shown in Figure 15. We carefully tuned the training hyperparameters", + "type": "text" + }, + { + "bbox": [ + 426, + 622, + 455, + 633 + ], + "score": 0.89, + "content": "\\lambda _ { C L I P }", + "type": "inline_equation" + }, + { + "bbox": [ + 456, + 620, + 473, + 636 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 473, + 622, + 502, + 633 + ], + "score": 0.86, + "content": "\\lambda _ { r e c o n } )", + "type": "inline_equation" + }, + { + "bbox": [ + 502, + 620, + 506, + 636 + ], + "score": 1.0, + "content": ")", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 104, + 631, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 104, + 631, + 506, + 646 + ], + "score": 1.0, + "content": "for fair comparison. The 1st to the 6th layers hardly bring visible changes. The 7th and 9th layers", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "score": 1.0, + "content": "bring not only the desired changes but also difficulty in finding optimal hyperparameters. After the", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 655, + 273, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 273, + 667 + ], + "score": 1.0, + "content": "9th layer, the results bear severe artifacts.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 33 + }, + { + "type": "title", + "bbox": [ + 108, + 683, + 285, + 694 + ], + "lines": [ + { + "bbox": [ + 105, + 681, + 286, + 697 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 286, + 697 + ], + "score": 1.0, + "content": "E IMPLICIT NEURAL DIRECTIONS", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 103, + 707, + 504, + 730 + ], + "lines": [ + { + "bbox": [ + 106, + 706, + 505, + 719 + ], + "spans": [ + { + "bbox": [ + 106, + 706, + 299, + 719 + ], + "score": 1.0, + "content": "Figure 16 illustrates the neural implicit function", + "type": "text" + }, + { + "bbox": [ + 300, + 708, + 309, + 719 + ], + "score": 0.83, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 706, + 505, + 719 + ], + "score": 1.0, + "content": ". It has only two 1x1 convolution layers with 512", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 718, + 394, + 730 + ], + "spans": [ + { + "bbox": [ + 106, + 718, + 394, + 730 + ], + "score": 1.0, + "content": "channels. Note that we haven’t explored the network architecture much.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5 + } + ], + "page_idx": 16, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "17", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 81, + 383, + 94 + ], + "lines": [ + { + "bbox": [ + 105, + 79, + 384, + 97 + ], + "spans": [ + { + "bbox": [ + 105, + 79, + 275, + 97 + ], + "score": 1.0, + "content": "D ADDITIONAL SUPPORTS FOR", + "type": "text" + }, + { + "bbox": [ + 275, + 82, + 282, + 93 + ], + "score": 0.59, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 283, + 79, + 384, + 97 + ], + "score": 1.0, + "content": "-space WITH ASYRP", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 106, + 364, + 117 + ], + "lines": [ + { + "bbox": [ + 105, + 106, + 365, + 119 + ], + "spans": [ + { + "bbox": [ + 105, + 106, + 257, + 119 + ], + "score": 1.0, + "content": "D.1 RANDOM PERTURBATION ON", + "type": "text" + }, + { + "bbox": [ + 258, + 109, + 263, + 116 + ], + "score": 0.29, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 263, + 106, + 365, + 119 + ], + "score": 1.0, + "content": "-space WITHOUT ASYRP", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1, + "bbox_fs": [ + 105, + 106, + 365, + 119 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 126, + 505, + 160 + ], + "lines": [ + { + "bbox": [ + 105, + 127, + 504, + 139 + ], + "spans": [ + { + "bbox": [ + 105, + 127, + 117, + 139 + ], + "score": 1.0, + "content": "In", + "type": "text" + }, + { + "bbox": [ + 117, + 127, + 137, + 138 + ], + "score": 0.6, + "content": "\\ S 3 . 1", + "type": "inline_equation" + }, + { + "bbox": [ + 138, + 127, + 225, + 139 + ], + "score": 1.0, + "content": ", we argue that if both", + "type": "text" + }, + { + "bbox": [ + 225, + 127, + 238, + 138 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 238, + 127, + 255, + 139 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 256, + 127, + 269, + 138 + ], + "score": 0.89, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 127, + 408, + 139 + ], + "score": 1.0, + "content": "are shifted, we can not manipulate", + "type": "text" + }, + { + "bbox": [ + 409, + 128, + 421, + 138 + ], + "score": 0.85, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 421, + 127, + 504, + 139 + ], + "score": 1.0, + "content": ". In Figure 13, we do", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 137, + 506, + 151 + ], + "spans": [ + { + "bbox": [ + 105, + 137, + 506, + 151 + ], + "score": 1.0, + "content": "not observe the noticeable difference between (a) and (b) which are the result of the original reverse", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 149, + 380, + 162 + ], + "spans": [ + { + "bbox": [ + 105, + 149, + 380, + 162 + ], + "score": 1.0, + "content": "process of DDIM and the one with shifting both terms, respectively.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3, + "bbox_fs": [ + 105, + 127, + 506, + 162 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 173, + 421, + 185 + ], + "lines": [ + { + "bbox": [ + 105, + 172, + 422, + 187 + ], + "spans": [ + { + "bbox": [ + 105, + 172, + 277, + 187 + ], + "score": 1.0, + "content": "D.2 ROBUSTNESS AND SEMANTICS IN", + "type": "text" + }, + { + "bbox": [ + 277, + 174, + 284, + 183 + ], + "score": 0.7, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 284, + 172, + 331, + 187 + ], + "score": 1.0, + "content": "-space AND", + "type": "text" + }, + { + "bbox": [ + 332, + 176, + 337, + 183 + ], + "score": 0.5, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 338, + 172, + 422, + 187 + ], + "score": 1.0, + "content": "-space WITH ASYRP", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 107, + 194, + 505, + 238 + ], + "lines": [ + { + "bbox": [ + 105, + 193, + 505, + 207 + ], + "spans": [ + { + "bbox": [ + 105, + 193, + 118, + 207 + ], + "score": 1.0, + "content": "In", + "type": "text" + }, + { + "bbox": [ + 118, + 194, + 141, + 206 + ], + "score": 0.8, + "content": "\\mathrm { ~ \\ S ~ } 3 . 2", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 193, + 226, + 207 + ], + "score": 1.0, + "content": ", we also argue that", + "type": "text" + }, + { + "bbox": [ + 226, + 195, + 232, + 204 + ], + "score": 0.73, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 233, + 193, + 345, + 207 + ], + "score": 1.0, + "content": "-space is more robust than", + "type": "text" + }, + { + "bbox": [ + 346, + 196, + 352, + 204 + ], + "score": 0.6, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 352, + 193, + 505, + 207 + ], + "score": 1.0, + "content": "-space with Asyrp. In Figure 13, we", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 204, + 506, + 218 + ], + "spans": [ + { + "bbox": [ + 105, + 204, + 245, + 218 + ], + "score": 1.0, + "content": "observe that small random noise", + "type": "text" + }, + { + "bbox": [ + 245, + 205, + 301, + 217 + ], + "score": 0.93, + "content": "z \\sim \\mathcal { N } ( 0 , \\mathbf { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 301, + 204, + 315, + 218 + ], + "score": 1.0, + "content": "in", + "type": "text" + }, + { + "bbox": [ + 316, + 207, + 322, + 216 + ], + "score": 0.64, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 322, + 204, + 506, + 218 + ], + "score": 1.0, + "content": "-space degrades the resulting image without", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 216, + 505, + 228 + ], + "spans": [ + { + "bbox": [ + 105, + 216, + 335, + 228 + ], + "score": 1.0, + "content": "semantic changes (c) and much larger random noise in", + "type": "text" + }, + { + "bbox": [ + 335, + 217, + 342, + 226 + ], + "score": 0.75, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 342, + 216, + 505, + 228 + ], + "score": 1.0, + "content": "-space yields random semantic changes", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 227, + 218, + 239 + ], + "spans": [ + { + "bbox": [ + 106, + 227, + 218, + 239 + ], + "score": 1.0, + "content": "without severe artifacts (d).", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 7.5, + "bbox_fs": [ + 105, + 193, + 506, + 239 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 244, + 505, + 333 + ], + "lines": [ + { + "bbox": [ + 106, + 244, + 505, + 255 + ], + "spans": [ + { + "bbox": [ + 106, + 244, + 505, + 255 + ], + "score": 1.0, + "content": "Note that diffusion models are designed as latent variable models with learned Gaussian transitions", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 104, + 252, + 504, + 270 + ], + "spans": [ + { + "bbox": [ + 104, + 252, + 493, + 270 + ], + "score": 1.0, + "content": "and the reverse process should also be close to Gaussian. Based on the assumption, we manage", + "type": "text" + }, + { + "bbox": [ + 493, + 254, + 504, + 267 + ], + "score": 0.87, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 263, + 506, + 280 + ], + "spans": [ + { + "bbox": [ + 104, + 263, + 367, + 280 + ], + "score": 1.0, + "content": "to follow the original Gaussian distribution as follows. Adding", + "type": "text" + }, + { + "bbox": [ + 367, + 266, + 428, + 278 + ], + "score": 0.93, + "content": "z \\sim \\mathcal { N } ( 0 , \\sigma \\bar { \\bf { I } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 263, + 441, + 280 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 441, + 265, + 452, + 278 + ], + "score": 0.88, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 452, + 263, + 506, + 280 + ], + "score": 1.0, + "content": "expands the", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 276, + 506, + 290 + ], + "spans": [ + { + "bbox": [ + 105, + 276, + 506, + 290 + ], + "score": 1.0, + "content": "distribution of the predicted noise and may produce distorted images. To preserve the distribution,√", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 288, + 506, + 302 + ], + "spans": [ + { + "bbox": [ + 105, + 288, + 143, + 302 + ], + "score": 1.0, + "content": "we scale", + "type": "text" + }, + { + "bbox": [ + 144, + 288, + 248, + 301 + ], + "score": 0.91, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta } = ( \\epsilon _ { t } ^ { \\theta } + z ) / \\sqrt { 1 ^ { 2 } + { \\sigma } ^ { 2 } }", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 288, + 506, + 302 + ], + "score": 1.0, + "content": ". Still, the resulting images are almost identical compared to the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 104, + 298, + 506, + 314 + ], + "spans": [ + { + "bbox": [ + 104, + 298, + 198, + 314 + ], + "score": 1.0, + "content": "original images where", + "type": "text" + }, + { + "bbox": [ + 198, + 300, + 249, + 312 + ], + "score": 0.9, + "content": "\\tilde { \\epsilon } _ { t } ^ { \\theta } = \\epsilon _ { t } ^ { \\theta } + z", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 298, + 506, + 314 + ], + "score": 1.0, + "content": "as shown in Figure 13. It is no wonder that the scaling does not", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 311, + 505, + 323 + ], + "spans": [ + { + "bbox": [ + 105, + 311, + 505, + 323 + ], + "score": 1.0, + "content": "improve the distorted results since the additive random noise disturbs the denoising operation of the", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 322, + 206, + 334 + ], + "spans": [ + { + "bbox": [ + 105, + 322, + 206, + 334 + ], + "score": 1.0, + "content": "predicted random noise.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 13.5, + "bbox_fs": [ + 104, + 244, + 506, + 334 + ] + }, + { + "type": "image", + "bbox": [ + 110, + 345, + 500, + 433 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 110, + 345, + 500, + 433 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 110, + 345, + 500, + 433 + ], + "spans": [ + { + "bbox": [ + 110, + 345, + 500, + 433 + ], + "score": 0.961, + "type": "image", + "image_path": "bb209ea4fa736576560335d535c5528804419b23be9234941f9e22111f467649.jpg" + } + ] + } + ], + "index": 19, + "virtual_lines": [ + { + "bbox": [ + 110, + 345, + 500, + 374.3333333333333 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 110, + 374.3333333333333, + 500, + 403.66666666666663 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 110, + 403.66666666666663, + 500, + 432.99999999999994 + ], + "spans": [], + "index": 20 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 444, + 505, + 522 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 444, + 506, + 458 + ], + "spans": [ + { + "bbox": [ + 105, + 444, + 506, + 458 + ], + "score": 1.0, + "content": "Figure 13: (a) The reconstructed image by the original DDIM inversion process which is almost", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 455, + 504, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 415, + 468 + ], + "score": 1.0, + "content": "indistinguishable to the real image. (b) The result from adding random noise", + "type": "text" + }, + { + "bbox": [ + 416, + 457, + 423, + 466 + ], + "score": 0.76, + "content": "z", + "type": "inline_equation" + }, + { + "bbox": [ + 423, + 455, + 456, + 468 + ], + "score": 1.0, + "content": "both on", + "type": "text" + }, + { + "bbox": [ + 456, + 456, + 469, + 467 + ], + "score": 0.88, + "content": "\\mathbf { P } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 470, + 455, + 487, + 468 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 488, + 456, + 501, + 467 + ], + "score": 0.88, + "content": "\\mathbf { D } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 501, + 455, + 504, + 468 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 466, + 505, + 479 + ], + "spans": [ + { + "bbox": [ + 106, + 466, + 505, + 479 + ], + "score": 1.0, + "content": "(a) looks identical to (b). The insets of (b, c) depict the full SSIM image from (a). It shows that", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 104, + 476, + 506, + 492 + ], + "spans": [ + { + "bbox": [ + 104, + 476, + 168, + 492 + ], + "score": 1.0, + "content": "simply shifting", + "type": "text" + }, + { + "bbox": [ + 168, + 477, + 179, + 489 + ], + "score": 0.89, + "content": "\\epsilon _ { t } ^ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 179, + 476, + 385, + 492 + ], + "score": 1.0, + "content": "without Asyrp does not affect the result. (c) Adding", + "type": "text" + }, + { + "bbox": [ + 386, + 478, + 415, + 487 + ], + "score": 0.9, + "content": "z \\sim \\mathcal { N }", + "type": "inline_equation" + }, + { + "bbox": [ + 415, + 476, + 426, + 492 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 426, + 479, + 432, + 487 + ], + "score": 0.63, + "content": "\\epsilon", + "type": "inline_equation" + }, + { + "bbox": [ + 432, + 476, + 506, + 492 + ], + "score": 1.0, + "content": "-space with Asyrp", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 488, + 505, + 502 + ], + "spans": [ + { + "bbox": [ + 105, + 488, + 356, + 502 + ], + "score": 1.0, + "content": "easily degrades image with little semantic change. (d) Adding", + "type": "text" + }, + { + "bbox": [ + 356, + 489, + 386, + 499 + ], + "score": 0.9, + "content": "z \\sim \\mathcal { N }", + "type": "inline_equation" + }, + { + "bbox": [ + 387, + 488, + 397, + 502 + ], + "score": 1.0, + "content": "to", + "type": "text" + }, + { + "bbox": [ + 398, + 489, + 405, + 499 + ], + "score": 0.82, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 405, + 488, + 505, + 502 + ], + "score": 1.0, + "content": "-space with Asyrp yields", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 104, + 499, + 506, + 512 + ], + "spans": [ + { + "bbox": [ + 104, + 499, + 506, + 512 + ], + "score": 1.0, + "content": "random semantic change without image degradation. (e) Correlation between image degradation and", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 511, + 237, + 523 + ], + "spans": [ + { + "bbox": [ + 106, + 511, + 237, + 523 + ], + "score": 1.0, + "content": "noise strength in the two spaces.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 24 + } + ], + "index": 21.5 + }, + { + "type": "title", + "bbox": [ + 108, + 546, + 259, + 558 + ], + "lines": [ + { + "bbox": [ + 105, + 544, + 261, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 182, + 560 + ], + "score": 1.0, + "content": "D.3 CHOICE OF", + "type": "text" + }, + { + "bbox": [ + 182, + 547, + 189, + 556 + ], + "score": 0.53, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 544, + 261, + 560 + ], + "score": 1.0, + "content": "-space IN U-NET", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 106, + 567, + 505, + 666 + ], + "lines": [ + { + "bbox": [ + 106, + 567, + 505, + 579 + ], + "spans": [ + { + "bbox": [ + 106, + 567, + 346, + 579 + ], + "score": 1.0, + "content": "As shown in Figure 14, there are many other candidates for", + "type": "text" + }, + { + "bbox": [ + 347, + 568, + 353, + 577 + ], + "score": 0.69, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 353, + 567, + 505, + 579 + ], + "score": 1.0, + "content": "-space in the architecture. Among the", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 578, + 504, + 590 + ], + "spans": [ + { + "bbox": [ + 106, + 578, + 504, + 590 + ], + "score": 1.0, + "content": "layers, we choose the 8th layer, the bridge of the U-Net based architecture. The layer is not influ-", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 590, + 505, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 590, + 505, + 601 + ], + "score": 1.0, + "content": "enced by any skip connection, has the smallest spatial dimension with compressed information, and", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 599, + 505, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 599, + 505, + 613 + ], + "score": 1.0, + "content": "is located just before the upsampling blocks. Thus, we assume that it could possibly be considered as", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 610, + 505, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 396, + 624 + ], + "score": 1.0, + "content": "the most suitable latent embedding. To confirm the assumption, we train", + "type": "text" + }, + { + "bbox": [ + 397, + 611, + 407, + 622 + ], + "score": 0.88, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 407, + 610, + 505, + 624 + ], + "score": 1.0, + "content": "on the other layers. The", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 104, + 620, + 506, + 636 + ], + "spans": [ + { + "bbox": [ + 104, + 620, + 425, + 636 + ], + "score": 1.0, + "content": "results are shown in Figure 15. We carefully tuned the training hyperparameters", + "type": "text" + }, + { + "bbox": [ + 426, + 622, + 455, + 633 + ], + "score": 0.89, + "content": "\\lambda _ { C L I P }", + "type": "inline_equation" + }, + { + "bbox": [ + 456, + 620, + 473, + 636 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 473, + 622, + 502, + 633 + ], + "score": 0.86, + "content": "\\lambda _ { r e c o n } )", + "type": "inline_equation" + }, + { + "bbox": [ + 502, + 620, + 506, + 636 + ], + "score": 1.0, + "content": ")", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 104, + 631, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 104, + 631, + 506, + 646 + ], + "score": 1.0, + "content": "for fair comparison. The 1st to the 6th layers hardly bring visible changes. The 7th and 9th layers", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 506, + 657 + ], + "score": 1.0, + "content": "bring not only the desired changes but also difficulty in finding optimal hyperparameters. After the", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 655, + 273, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 273, + 667 + ], + "score": 1.0, + "content": "9th layer, the results bear severe artifacts.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 33, + "bbox_fs": [ + 104, + 567, + 506, + 667 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 683, + 285, + 694 + ], + "lines": [ + { + "bbox": [ + 105, + 681, + 286, + 697 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 286, + 697 + ], + "score": 1.0, + "content": "E IMPLICIT NEURAL DIRECTIONS", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 103, + 707, + 504, + 730 + ], + "lines": [ + { + "bbox": [ + 106, + 706, + 505, + 719 + ], + "spans": [ + { + "bbox": [ + 106, + 706, + 299, + 719 + ], + "score": 1.0, + "content": "Figure 16 illustrates the neural implicit function", + "type": "text" + }, + { + "bbox": [ + 300, + 708, + 309, + 719 + ], + "score": 0.83, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 706, + 505, + 719 + ], + "score": 1.0, + "content": ". It has only two 1x1 convolution layers with 512", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 718, + 394, + 730 + ], + "spans": [ + { + "bbox": [ + 106, + 718, + 394, + 730 + ], + "score": 1.0, + "content": "channels. Note that we haven’t explored the network architecture much.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5, + "bbox_fs": [ + 106, + 706, + 505, + 730 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 114, + 88, + 506, + 301 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 114, + 88, + 506, + 301 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 114, + 88, + 506, + 301 + ], + "spans": [ + { + "bbox": [ + 114, + 88, + 506, + 301 + ], + "score": 0.972, + "type": "image", + "image_path": "3e8c45121a1bfb589a589081725c2d792d7bced359e7ac19f594c2ade31cda78.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 114, + 88, + 506, + 159.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 114, + 159.0, + 506, + 230.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 114, + 230.0, + 506, + 301.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 316, + 506, + 350 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 315, + 504, + 329 + ], + "spans": [ + { + "bbox": [ + 105, + 315, + 205, + 329 + ], + "score": 1.0, + "content": "Figure 14: Location of", + "type": "text" + }, + { + "bbox": [ + 205, + 317, + 212, + 326 + ], + "score": 0.41, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 213, + 315, + 459, + 329 + ], + "score": 1.0, + "content": "-space. The U-Net architecture of diffusion models outputs", + "type": "text" + }, + { + "bbox": [ + 459, + 316, + 504, + 327 + ], + "score": 0.87, + "content": "2 5 6 \\times 2 5 6", + "type": "inline_equation" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 327, + 505, + 340 + ], + "spans": [ + { + "bbox": [ + 105, + 327, + 505, + 340 + ], + "score": 1.0, + "content": "images. Each layer is indexed with a number along the operating sequence of the model. The 8th", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 338, + 402, + 351 + ], + "spans": [ + { + "bbox": [ + 106, + 338, + 153, + 351 + ], + "score": 1.0, + "content": "layer is our", + "type": "text" + }, + { + "bbox": [ + 154, + 339, + 160, + 348 + ], + "score": 0.65, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 160, + 338, + 402, + 351 + ], + "score": 1.0, + "content": "-space which is not directly influenced by a skip connection.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "title", + "bbox": [ + 106, + 376, + 468, + 403 + ], + "lines": [ + { + "bbox": [ + 104, + 374, + 470, + 390 + ], + "spans": [ + { + "bbox": [ + 104, + 374, + 470, + 390 + ], + "score": 1.0, + "content": "F QUALITY IMPROVEMENTS BY NON-ACCELERATED SAMPLING WITH", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 125, + 390, + 191, + 404 + ], + "spans": [ + { + "bbox": [ + 125, + 390, + 169, + 404 + ], + "score": 1.0, + "content": "SCALED", + "type": "text" + }, + { + "bbox": [ + 169, + 390, + 191, + 403 + ], + "score": 0.89, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + } + ], + "index": 7 + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 106, + 416, + 505, + 483 + ], + "lines": [ + { + "bbox": [ + 105, + 416, + 506, + 429 + ], + "spans": [ + { + "bbox": [ + 105, + 416, + 443, + 429 + ], + "score": 1.0, + "content": "Figure 17 shows the quality improvements by non-accelerated sampling with scaled", + "type": "text" + }, + { + "bbox": [ + 444, + 417, + 463, + 428 + ], + "score": 0.89, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 463, + 416, + 506, + 429 + ], + "score": 1.0, + "content": "described", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 428, + 505, + 439 + ], + "spans": [ + { + "bbox": [ + 106, + 428, + 117, + 439 + ], + "score": 1.0, + "content": "in", + "type": "text" + }, + { + "bbox": [ + 117, + 428, + 138, + 439 + ], + "score": 0.53, + "content": "\\ S \\ S . 4", + "type": "inline_equation" + }, + { + "bbox": [ + 138, + 428, + 505, + 439 + ], + "score": 1.0, + "content": ". Even with different number of inference steps, we observe similar changes of an attribute", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 438, + 505, + 451 + ], + "spans": [ + { + "bbox": [ + 105, + 438, + 212, + 451 + ], + "score": 1.0, + "content": "if we preserve the sum of", + "type": "text" + }, + { + "bbox": [ + 212, + 439, + 231, + 450 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 231, + 438, + 505, + 451 + ], + "score": 1.0, + "content": ". This scaling technique allows non-accelerated sampling with 1000", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 104, + 448, + 506, + 464 + ], + "spans": [ + { + "bbox": [ + 104, + 448, + 506, + 464 + ], + "score": 1.0, + "content": "steps for the models trained by accelerated training with 40 steps. Using non-accelerated sampling", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 460, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 460, + 156, + 473 + ], + "score": 1.0, + "content": "with scaled", + "type": "text" + }, + { + "bbox": [ + 156, + 461, + 176, + 471 + ], + "score": 0.9, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 176, + 460, + 505, + 473 + ], + "score": 1.0, + "content": "leads to the same magnitude of manipulation and higher-quality images. In our", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 472, + 479, + 485 + ], + "spans": [ + { + "bbox": [ + 106, + 472, + 479, + 485 + ], + "score": 1.0, + "content": "experiments, it takes about 1.5 seconds to sample for 40 steps and 40 seconds for 1000 steps.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 10.5 + }, + { + "type": "title", + "bbox": [ + 107, + 502, + 369, + 514 + ], + "lines": [ + { + "bbox": [ + 106, + 502, + 370, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 502, + 370, + 516 + ], + "score": 1.0, + "content": "G EDITING STRENGTH AND EDITING FLEXIBILITY", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14 + }, + { + "type": "title", + "bbox": [ + 108, + 528, + 260, + 540 + ], + "lines": [ + { + "bbox": [ + 105, + 526, + 261, + 542 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 241, + 542 + ], + "score": 1.0, + "content": "G.1 EDITING STRENGTH AND", + "type": "text" + }, + { + "bbox": [ + 241, + 529, + 261, + 540 + ], + "score": 0.65, + "content": "t _ { e d i t }", + "type": "inline_equation" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 107, + 550, + 505, + 583 + ], + "lines": [ + { + "bbox": [ + 105, + 549, + 505, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 549, + 272, + 563 + ], + "score": 1.0, + "content": "Figure 18 shows the results according to", + "type": "text" + }, + { + "bbox": [ + 272, + 551, + 291, + 561 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 291, + 549, + 304, + 563 + ], + "score": 1.0, + "content": ". If", + "type": "text" + }, + { + "bbox": [ + 305, + 551, + 323, + 561 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 549, + 505, + 563 + ], + "score": 1.0, + "content": "is too high, the length of the editing process", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 561, + 504, + 574 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 409, + 574 + ], + "score": 1.0, + "content": "becomes too short resulting in insufficient changes. On the contrary, too low", + "type": "text" + }, + { + "bbox": [ + 410, + 562, + 428, + 572 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 561, + 504, + 574 + ], + "score": 1.0, + "content": "causes excessively", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 572, + 334, + 585 + ], + "spans": [ + { + "bbox": [ + 106, + 572, + 334, + 585 + ], + "score": 1.0, + "content": "unnecessary manipulation from the long editing process.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17 + }, + { + "type": "text", + "bbox": [ + 107, + 588, + 505, + 633 + ], + "lines": [ + { + "bbox": [ + 106, + 588, + 505, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 172, + 601 + ], + "score": 1.0, + "content": "We observe that", + "type": "text" + }, + { + "bbox": [ + 172, + 590, + 191, + 600 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 191, + 588, + 505, + 601 + ], + "score": 1.0, + "content": "is one of the important hyperparameters. We argue that the formula for choos-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 600, + 505, + 612 + ], + "spans": [ + { + "bbox": [ + 105, + 600, + 121, + 612 + ], + "score": 1.0, + "content": "ing", + "type": "text" + }, + { + "bbox": [ + 121, + 600, + 140, + 611 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 140, + 600, + 505, + 612 + ], + "score": 1.0, + "content": "using editing strength is reasonable because it applies to all five different datasets despite its", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 610, + 505, + 622 + ], + "spans": [ + { + "bbox": [ + 105, + 611, + 255, + 622 + ], + "score": 1.0, + "content": "sensitivity, even though the choice of", + "type": "text" + }, + { + "bbox": [ + 255, + 610, + 317, + 622 + ], + "score": 0.44, + "content": "\\mathrm { L P I P S = 0 . 3 3 }", + "type": "inline_equation" + }, + { + "bbox": [ + 317, + 611, + 505, + 622 + ], + "score": 1.0, + "content": "is empirical. We provide ablation of thresholds", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 621, + 161, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 161, + 634 + ], + "score": 1.0, + "content": "in Figure 22.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 20.5 + }, + { + "type": "text", + "bbox": [ + 104, + 638, + 504, + 661 + ], + "lines": [ + { + "bbox": [ + 106, + 638, + 505, + 651 + ], + "spans": [ + { + "bbox": [ + 106, + 638, + 389, + 651 + ], + "score": 1.0, + "content": "Additionally, these results imply why we need to use sufficiently low", + "type": "text" + }, + { + "bbox": [ + 389, + 639, + 408, + 650 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 408, + 638, + 505, + 651 + ], + "score": 1.0, + "content": "in the unseen domains.", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 649, + 486, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 486, + 663 + ], + "score": 1.0, + "content": "Editing interval with insufficient editing strength struggles to escape from the training domain.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 23.5 + }, + { + "type": "title", + "bbox": [ + 108, + 677, + 273, + 689 + ], + "lines": [ + { + "bbox": [ + 105, + 675, + 275, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 675, + 275, + 691 + ], + "score": 1.0, + "content": "G.2 EDITING FLEXIBILITY AND tboost", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 222, + 712 + ], + "score": 1.0, + "content": "Table 2 shows the prompts,", + "type": "text" + }, + { + "bbox": [ + 223, + 700, + 241, + 710 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 241, + 698, + 263, + 712 + ], + "score": 1.0, + "content": ", and", + "type": "text" + }, + { + "bbox": [ + 264, + 700, + 288, + 710 + ], + "score": 0.89, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 288, + 698, + 506, + 712 + ], + "score": 1.0, + "content": ". Intuitively, the attributes with larger visual changes", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 708, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 506, + 723 + ], + "score": 1.0, + "content": "have smaller cosine similarity and require longer editing interval. Figure 19 shows the average", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 107, + 719, + 506, + 734 + ], + "spans": [ + { + "bbox": [ + 107, + 720, + 167, + 732 + ], + "score": 0.68, + "content": "\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 719, + 187, + 734 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 187, + 721, + 246, + 732 + ], + "score": 0.79, + "content": "\\mathrm { L P I P S } ( { \\pmb x } , { \\pmb x } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 246, + 719, + 421, + 734 + ], + "score": 1.0, + "content": "of 100 samples on all datasets. Note that", + "type": "text" + }, + { + "bbox": [ + 422, + 721, + 441, + 732 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 441, + 719, + 461, + 734 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 461, + 721, + 486, + 732 + ], + "score": 0.88, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 486, + 719, + 506, + 734 + ], + "score": 1.0, + "content": "dif-", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27 + } + ], + "page_idx": 17, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 301, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 763 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 763 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 114, + 88, + 506, + 301 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 114, + 88, + 506, + 301 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 114, + 88, + 506, + 301 + ], + "spans": [ + { + "bbox": [ + 114, + 88, + 506, + 301 + ], + "score": 0.972, + "type": "image", + "image_path": "3e8c45121a1bfb589a589081725c2d792d7bced359e7ac19f594c2ade31cda78.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 114, + 88, + 506, + 159.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 114, + 159.0, + 506, + 230.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 114, + 230.0, + 506, + 301.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 316, + 506, + 350 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 315, + 504, + 329 + ], + "spans": [ + { + "bbox": [ + 105, + 315, + 205, + 329 + ], + "score": 1.0, + "content": "Figure 14: Location of", + "type": "text" + }, + { + "bbox": [ + 205, + 317, + 212, + 326 + ], + "score": 0.41, + "content": "\\pmb { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 213, + 315, + 459, + 329 + ], + "score": 1.0, + "content": "-space. The U-Net architecture of diffusion models outputs", + "type": "text" + }, + { + "bbox": [ + 459, + 316, + 504, + 327 + ], + "score": 0.87, + "content": "2 5 6 \\times 2 5 6", + "type": "inline_equation" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 327, + 505, + 340 + ], + "spans": [ + { + "bbox": [ + 105, + 327, + 505, + 340 + ], + "score": 1.0, + "content": "images. Each layer is indexed with a number along the operating sequence of the model. The 8th", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 338, + 402, + 351 + ], + "spans": [ + { + "bbox": [ + 106, + 338, + 153, + 351 + ], + "score": 1.0, + "content": "layer is our", + "type": "text" + }, + { + "bbox": [ + 154, + 339, + 160, + 348 + ], + "score": 0.65, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 160, + 338, + 402, + 351 + ], + "score": 1.0, + "content": "-space which is not directly influenced by a skip connection.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "title", + "bbox": [ + 106, + 376, + 468, + 403 + ], + "lines": [ + { + "bbox": [ + 104, + 374, + 470, + 390 + ], + "spans": [ + { + "bbox": [ + 104, + 374, + 470, + 390 + ], + "score": 1.0, + "content": "F QUALITY IMPROVEMENTS BY NON-ACCELERATED SAMPLING WITH", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 125, + 390, + 191, + 404 + ], + "spans": [ + { + "bbox": [ + 125, + 390, + 169, + 404 + ], + "score": 1.0, + "content": "SCALED", + "type": "text" + }, + { + "bbox": [ + 169, + 390, + 191, + 403 + ], + "score": 0.89, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + } + ], + "index": 7 + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 106, + 416, + 505, + 483 + ], + "lines": [ + { + "bbox": [ + 105, + 416, + 506, + 429 + ], + "spans": [ + { + "bbox": [ + 105, + 416, + 443, + 429 + ], + "score": 1.0, + "content": "Figure 17 shows the quality improvements by non-accelerated sampling with scaled", + "type": "text" + }, + { + "bbox": [ + 444, + 417, + 463, + 428 + ], + "score": 0.89, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 463, + 416, + 506, + 429 + ], + "score": 1.0, + "content": "described", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 428, + 505, + 439 + ], + "spans": [ + { + "bbox": [ + 106, + 428, + 117, + 439 + ], + "score": 1.0, + "content": "in", + "type": "text" + }, + { + "bbox": [ + 117, + 428, + 138, + 439 + ], + "score": 0.53, + "content": "\\ S \\ S . 4", + "type": "inline_equation" + }, + { + "bbox": [ + 138, + 428, + 505, + 439 + ], + "score": 1.0, + "content": ". Even with different number of inference steps, we observe similar changes of an attribute", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 438, + 505, + 451 + ], + "spans": [ + { + "bbox": [ + 105, + 438, + 212, + 451 + ], + "score": 1.0, + "content": "if we preserve the sum of", + "type": "text" + }, + { + "bbox": [ + 212, + 439, + 231, + 450 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 231, + 438, + 505, + 451 + ], + "score": 1.0, + "content": ". This scaling technique allows non-accelerated sampling with 1000", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 104, + 448, + 506, + 464 + ], + "spans": [ + { + "bbox": [ + 104, + 448, + 506, + 464 + ], + "score": 1.0, + "content": "steps for the models trained by accelerated training with 40 steps. Using non-accelerated sampling", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 460, + 505, + 473 + ], + "spans": [ + { + "bbox": [ + 105, + 460, + 156, + 473 + ], + "score": 1.0, + "content": "with scaled", + "type": "text" + }, + { + "bbox": [ + 156, + 461, + 176, + 471 + ], + "score": 0.9, + "content": "\\Delta { h _ { t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 176, + 460, + 505, + 473 + ], + "score": 1.0, + "content": "leads to the same magnitude of manipulation and higher-quality images. In our", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 472, + 479, + 485 + ], + "spans": [ + { + "bbox": [ + 106, + 472, + 479, + 485 + ], + "score": 1.0, + "content": "experiments, it takes about 1.5 seconds to sample for 40 steps and 40 seconds for 1000 steps.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 10.5, + "bbox_fs": [ + 104, + 416, + 506, + 485 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 502, + 369, + 514 + ], + "lines": [ + { + "bbox": [ + 106, + 502, + 370, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 502, + 370, + 516 + ], + "score": 1.0, + "content": "G EDITING STRENGTH AND EDITING FLEXIBILITY", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14 + }, + { + "type": "title", + "bbox": [ + 108, + 528, + 260, + 540 + ], + "lines": [ + { + "bbox": [ + 105, + 526, + 261, + 542 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 241, + 542 + ], + "score": 1.0, + "content": "G.1 EDITING STRENGTH AND", + "type": "text" + }, + { + "bbox": [ + 241, + 529, + 261, + 540 + ], + "score": 0.65, + "content": "t _ { e d i t }", + "type": "inline_equation" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 107, + 550, + 505, + 583 + ], + "lines": [ + { + "bbox": [ + 105, + 549, + 505, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 549, + 272, + 563 + ], + "score": 1.0, + "content": "Figure 18 shows the results according to", + "type": "text" + }, + { + "bbox": [ + 272, + 551, + 291, + 561 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 291, + 549, + 304, + 563 + ], + "score": 1.0, + "content": ". If", + "type": "text" + }, + { + "bbox": [ + 305, + 551, + 323, + 561 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 549, + 505, + 563 + ], + "score": 1.0, + "content": "is too high, the length of the editing process", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 561, + 504, + 574 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 409, + 574 + ], + "score": 1.0, + "content": "becomes too short resulting in insufficient changes. On the contrary, too low", + "type": "text" + }, + { + "bbox": [ + 410, + 562, + 428, + 572 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 561, + 504, + 574 + ], + "score": 1.0, + "content": "causes excessively", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 572, + 334, + 585 + ], + "spans": [ + { + "bbox": [ + 106, + 572, + 334, + 585 + ], + "score": 1.0, + "content": "unnecessary manipulation from the long editing process.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17, + "bbox_fs": [ + 105, + 549, + 505, + 585 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 588, + 505, + 633 + ], + "lines": [ + { + "bbox": [ + 106, + 588, + 505, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 172, + 601 + ], + "score": 1.0, + "content": "We observe that", + "type": "text" + }, + { + "bbox": [ + 172, + 590, + 191, + 600 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 191, + 588, + 505, + 601 + ], + "score": 1.0, + "content": "is one of the important hyperparameters. We argue that the formula for choos-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 600, + 505, + 612 + ], + "spans": [ + { + "bbox": [ + 105, + 600, + 121, + 612 + ], + "score": 1.0, + "content": "ing", + "type": "text" + }, + { + "bbox": [ + 121, + 600, + 140, + 611 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 140, + 600, + 505, + 612 + ], + "score": 1.0, + "content": "using editing strength is reasonable because it applies to all five different datasets despite its", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 610, + 505, + 622 + ], + "spans": [ + { + "bbox": [ + 105, + 611, + 255, + 622 + ], + "score": 1.0, + "content": "sensitivity, even though the choice of", + "type": "text" + }, + { + "bbox": [ + 255, + 610, + 317, + 622 + ], + "score": 0.44, + "content": "\\mathrm { L P I P S = 0 . 3 3 }", + "type": "inline_equation" + }, + { + "bbox": [ + 317, + 611, + 505, + 622 + ], + "score": 1.0, + "content": "is empirical. We provide ablation of thresholds", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 621, + 161, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 621, + 161, + 634 + ], + "score": 1.0, + "content": "in Figure 22.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 20.5, + "bbox_fs": [ + 105, + 588, + 505, + 634 + ] + }, + { + "type": "list", + "bbox": [ + 104, + 638, + 504, + 661 + ], + "lines": [ + { + "bbox": [ + 106, + 638, + 505, + 651 + ], + "spans": [ + { + "bbox": [ + 106, + 638, + 389, + 651 + ], + "score": 1.0, + "content": "Additionally, these results imply why we need to use sufficiently low", + "type": "text" + }, + { + "bbox": [ + 389, + 639, + 408, + 650 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 408, + 638, + 505, + 651 + ], + "score": 1.0, + "content": "in the unseen domains.", + "type": "text" + } + ], + "index": 23, + "is_list_end_line": true + }, + { + "bbox": [ + 105, + 649, + 486, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 486, + 663 + ], + "score": 1.0, + "content": "Editing interval with insufficient editing strength struggles to escape from the training domain.", + "type": "text" + } + ], + "index": 24, + "is_list_start_line": true, + "is_list_end_line": true + } + ], + "index": 23.5, + "bbox_fs": [ + 105, + 638, + 505, + 663 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 677, + 273, + 689 + ], + "lines": [ + { + "bbox": [ + 105, + 675, + 275, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 675, + 275, + 691 + ], + "score": 1.0, + "content": "G.2 EDITING FLEXIBILITY AND tboost", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 222, + 712 + ], + "score": 1.0, + "content": "Table 2 shows the prompts,", + "type": "text" + }, + { + "bbox": [ + 223, + 700, + 241, + 710 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 241, + 698, + 263, + 712 + ], + "score": 1.0, + "content": ", and", + "type": "text" + }, + { + "bbox": [ + 264, + 700, + 288, + 710 + ], + "score": 0.89, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 288, + 698, + 506, + 712 + ], + "score": 1.0, + "content": ". Intuitively, the attributes with larger visual changes", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 708, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 506, + 723 + ], + "score": 1.0, + "content": "have smaller cosine similarity and require longer editing interval. Figure 19 shows the average", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 107, + 719, + 506, + 734 + ], + "spans": [ + { + "bbox": [ + 107, + 720, + 167, + 732 + ], + "score": 0.68, + "content": "\\mathrm { L P I P S } ( \\pmb { x } , \\mathbf { P } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 719, + 187, + 734 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 187, + 721, + 246, + 732 + ], + "score": 0.79, + "content": "\\mathrm { L P I P S } ( { \\pmb x } , { \\pmb x } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 246, + 719, + 421, + 734 + ], + "score": 1.0, + "content": "of 100 samples on all datasets. Note that", + "type": "text" + }, + { + "bbox": [ + 422, + 721, + 441, + 732 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 441, + 719, + 461, + 734 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 461, + 721, + 486, + 732 + ], + "score": 0.88, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 486, + 719, + 506, + 734 + ], + "score": 1.0, + "content": "dif-", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27, + "bbox_fs": [ + 105, + 698, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 255, + 77, + 422, + 441 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 255, + 77, + 422, + 441 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 255, + 77, + 422, + 441 + ], + "spans": [ + { + "bbox": [ + 255, + 77, + 422, + 441 + ], + "score": 0.738, + "type": "image", + "image_path": "20e18f4b6da57f683105204199ad21e00b13391adc9cd4bf0a95392c4e8d14c8.jpg" + } + ] + } + ], + "index": 13.5, + "virtual_lines": [ + { + "bbox": [ + 255, + 77, + 422, + 90.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 255, + 90.0, + 422, + 103.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 255, + 103.0, + 422, + 116.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 255, + 116.0, + 422, + 129.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 255, + 129.0, + 422, + 142.0 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 255, + 142.0, + 422, + 155.0 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 255, + 155.0, + 422, + 168.0 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 255, + 168.0, + 422, + 181.0 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 255, + 181.0, + 422, + 194.0 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 255, + 194.0, + 422, + 207.0 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 255, + 207.0, + 422, + 220.0 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 255, + 220.0, + 422, + 233.0 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 255, + 233.0, + 422, + 246.0 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 255, + 246.0, + 422, + 259.0 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 255, + 259.0, + 422, + 272.0 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 255, + 272.0, + 422, + 285.0 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 255, + 285.0, + 422, + 298.0 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 255, + 298.0, + 422, + 311.0 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 255, + 311.0, + 422, + 324.0 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 255, + 324.0, + 422, + 337.0 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 255, + 337.0, + 422, + 350.0 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 255, + 350.0, + 422, + 363.0 + ], + "spans": [], + "index": 21 + }, + { + "bbox": [ + 255, + 363.0, + 422, + 376.0 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 255, + 376.0, + 422, + 389.0 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 255, + 389.0, + 422, + 402.0 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 255, + 402.0, + 422, + 415.0 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 255, + 415.0, + 422, + 428.0 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 255, + 428.0, + 422, + 441.0 + ], + "spans": [], + "index": 27 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 455, + 506, + 489 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 455, + 505, + 467 + ], + "spans": [ + { + "bbox": [ + 106, + 455, + 505, + 467 + ], + "score": 1.0, + "content": "Figure 15: Exhaustive enumeration over the choices for semantic latent space. We show the", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 466, + 505, + 479 + ], + "spans": [ + { + "bbox": [ + 105, + 466, + 348, + 479 + ], + "score": 1.0, + "content": "result of the training data. We observe that the eighth layer (", + "type": "text" + }, + { + "bbox": [ + 348, + 467, + 354, + 477 + ], + "score": 0.4, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 354, + 466, + 505, + 479 + ], + "score": 1.0, + "content": "-space) of the U-net suits the best for", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 477, + 211, + 490 + ], + "spans": [ + { + "bbox": [ + 105, + 477, + 211, + 490 + ], + "score": 1.0, + "content": "the semantic latent space.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29 + } + ], + "index": 21.25 + }, + { + "type": "image", + "bbox": [ + 142, + 500, + 466, + 588 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 142, + 500, + 466, + 588 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 142, + 500, + 466, + 588 + ], + "spans": [ + { + "bbox": [ + 142, + 500, + 466, + 588 + ], + "score": 0.966, + "type": "image", + "image_path": "174bbae683903d32c4469c85c1da04852c054dd195ebaf00a66e2a411d7df08d.jpg" + } + ] + } + ], + "index": 32, + "virtual_lines": [ + { + "bbox": [ + 142, + 500, + 466, + 529.3333333333334 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 142, + 529.3333333333334, + 466, + 558.6666666666667 + ], + "spans": [], + "index": 32 + }, + { + "bbox": [ + 142, + 558.6666666666667, + 466, + 588.0000000000001 + ], + "spans": [], + "index": 33 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 144, + 600, + 464, + 613 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 144, + 599, + 464, + 614 + ], + "spans": [ + { + "bbox": [ + 144, + 599, + 253, + 614 + ], + "score": 1.0, + "content": "Figure 16: Illustration of", + "type": "text" + }, + { + "bbox": [ + 253, + 601, + 262, + 612 + ], + "score": 0.79, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 263, + 599, + 464, + 614 + ], + "score": 1.0, + "content": ". We use group norm and swish following DDPM.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 34 + } + ], + "index": 33.0 + }, + { + "type": "text", + "bbox": [ + 106, + 640, + 505, + 663 + ], + "lines": [ + { + "bbox": [ + 106, + 640, + 506, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 640, + 394, + 653 + ], + "score": 1.0, + "content": "fer across datasets and attributes, and they are chosen by the formulas in", + "type": "text" + }, + { + "bbox": [ + 395, + 641, + 409, + 652 + ], + "score": 0.37, + "content": "\\ S 4 .", + "type": "inline_equation" + }, + { + "bbox": [ + 409, + 640, + 506, + 653 + ], + "score": 1.0, + "content": ". We provide ablation of", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 651, + 185, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 652, + 131, + 663 + ], + "score": 0.81, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 651, + 185, + 664 + ], + "score": 1.0, + "content": "in Figure 23.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5 + }, + { + "type": "title", + "bbox": [ + 107, + 682, + 233, + 696 + ], + "lines": [ + { + "bbox": [ + 104, + 681, + 235, + 699 + ], + "spans": [ + { + "bbox": [ + 104, + 681, + 235, + 699 + ], + "score": 1.0, + "content": "H QUALITY BOOSTING", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 107, + 709, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 707, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 707, + 506, + 723 + ], + "score": 1.0, + "content": "We validate the effectiveness of our quality boosting (§ 4.2) in the original DDIM process and in", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 104, + 717, + 138, + 736 + ], + "spans": [ + { + "bbox": [ + 104, + 717, + 138, + 736 + ], + "score": 1.0, + "content": "Asyrp.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5 + } + ], + "page_idx": 18, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "19", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 255, + 77, + 422, + 441 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 255, + 77, + 422, + 441 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 255, + 77, + 422, + 441 + ], + "spans": [ + { + "bbox": [ + 255, + 77, + 422, + 441 + ], + "score": 0.738, + "type": "image", + "image_path": "20e18f4b6da57f683105204199ad21e00b13391adc9cd4bf0a95392c4e8d14c8.jpg" + } + ] + } + ], + "index": 13.5, + "virtual_lines": [ + { + "bbox": [ + 255, + 77, + 422, + 90.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 255, + 90.0, + 422, + 103.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 255, + 103.0, + 422, + 116.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 255, + 116.0, + 422, + 129.0 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 255, + 129.0, + 422, + 142.0 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 255, + 142.0, + 422, + 155.0 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 255, + 155.0, + 422, + 168.0 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 255, + 168.0, + 422, + 181.0 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 255, + 181.0, + 422, + 194.0 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 255, + 194.0, + 422, + 207.0 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 255, + 207.0, + 422, + 220.0 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 255, + 220.0, + 422, + 233.0 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 255, + 233.0, + 422, + 246.0 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 255, + 246.0, + 422, + 259.0 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 255, + 259.0, + 422, + 272.0 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 255, + 272.0, + 422, + 285.0 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 255, + 285.0, + 422, + 298.0 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 255, + 298.0, + 422, + 311.0 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 255, + 311.0, + 422, + 324.0 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 255, + 324.0, + 422, + 337.0 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 255, + 337.0, + 422, + 350.0 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 255, + 350.0, + 422, + 363.0 + ], + "spans": [], + "index": 21 + }, + { + "bbox": [ + 255, + 363.0, + 422, + 376.0 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 255, + 376.0, + 422, + 389.0 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 255, + 389.0, + 422, + 402.0 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 255, + 402.0, + 422, + 415.0 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 255, + 415.0, + 422, + 428.0 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 255, + 428.0, + 422, + 441.0 + ], + "spans": [], + "index": 27 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 455, + 506, + 489 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 455, + 505, + 467 + ], + "spans": [ + { + "bbox": [ + 106, + 455, + 505, + 467 + ], + "score": 1.0, + "content": "Figure 15: Exhaustive enumeration over the choices for semantic latent space. We show the", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 466, + 505, + 479 + ], + "spans": [ + { + "bbox": [ + 105, + 466, + 348, + 479 + ], + "score": 1.0, + "content": "result of the training data. We observe that the eighth layer (", + "type": "text" + }, + { + "bbox": [ + 348, + 467, + 354, + 477 + ], + "score": 0.4, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 354, + 466, + 505, + 479 + ], + "score": 1.0, + "content": "-space) of the U-net suits the best for", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 477, + 211, + 490 + ], + "spans": [ + { + "bbox": [ + 105, + 477, + 211, + 490 + ], + "score": 1.0, + "content": "the semantic latent space.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29 + } + ], + "index": 21.25 + }, + { + "type": "image", + "bbox": [ + 142, + 500, + 466, + 588 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 142, + 500, + 466, + 588 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 142, + 500, + 466, + 588 + ], + "spans": [ + { + "bbox": [ + 142, + 500, + 466, + 588 + ], + "score": 0.966, + "type": "image", + "image_path": "174bbae683903d32c4469c85c1da04852c054dd195ebaf00a66e2a411d7df08d.jpg" + } + ] + } + ], + "index": 32, + "virtual_lines": [ + { + "bbox": [ + 142, + 500, + 466, + 529.3333333333334 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 142, + 529.3333333333334, + 466, + 558.6666666666667 + ], + "spans": [], + "index": 32 + }, + { + "bbox": [ + 142, + 558.6666666666667, + 466, + 588.0000000000001 + ], + "spans": [], + "index": 33 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 144, + 600, + 464, + 613 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 144, + 599, + 464, + 614 + ], + "spans": [ + { + "bbox": [ + 144, + 599, + 253, + 614 + ], + "score": 1.0, + "content": "Figure 16: Illustration of", + "type": "text" + }, + { + "bbox": [ + 253, + 601, + 262, + 612 + ], + "score": 0.79, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 263, + 599, + 464, + 614 + ], + "score": 1.0, + "content": ". We use group norm and swish following DDPM.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 34 + } + ], + "index": 33.0 + }, + { + "type": "text", + "bbox": [ + 106, + 640, + 505, + 663 + ], + "lines": [ + { + "bbox": [ + 106, + 640, + 506, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 640, + 394, + 653 + ], + "score": 1.0, + "content": "fer across datasets and attributes, and they are chosen by the formulas in", + "type": "text" + }, + { + "bbox": [ + 395, + 641, + 409, + 652 + ], + "score": 0.37, + "content": "\\ S 4 .", + "type": "inline_equation" + }, + { + "bbox": [ + 409, + 640, + 506, + 653 + ], + "score": 1.0, + "content": ". We provide ablation of", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 651, + 185, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 652, + 131, + 663 + ], + "score": 0.81, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 651, + 185, + 664 + ], + "score": 1.0, + "content": "in Figure 23.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5, + "bbox_fs": [ + 106, + 640, + 506, + 664 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 682, + 233, + 696 + ], + "lines": [ + { + "bbox": [ + 104, + 681, + 235, + 699 + ], + "spans": [ + { + "bbox": [ + 104, + 681, + 235, + 699 + ], + "score": 1.0, + "content": "H QUALITY BOOSTING", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 107, + 709, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 707, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 707, + 506, + 723 + ], + "score": 1.0, + "content": "We validate the effectiveness of our quality boosting (§ 4.2) in the original DDIM process and in", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 104, + 717, + 138, + 736 + ], + "spans": [ + { + "bbox": [ + 104, + 717, + 138, + 736 + ], + "score": 1.0, + "content": "Asyrp.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5, + "bbox_fs": [ + 104, + 707, + 506, + 736 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 340, + 84, + 390, + 95 + ], + "lines": [ + { + "bbox": [ + 340, + 83, + 391, + 96 + ], + "spans": [ + { + "bbox": [ + 340, + 83, + 391, + 96 + ], + "score": 1.0, + "content": "Smiling with", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "image", + "bbox": [ + 120, + 88, + 490, + 483 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 120, + 88, + 490, + 483 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 120, + 88, + 490, + 483 + ], + "spans": [ + { + "bbox": [ + 120, + 88, + 490, + 483 + ], + "score": 0.956, + "type": "image", + "image_path": "e119805b4e47eb46571e4d9c938e16083540325efbc089de26542ba2e25cbc3d.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 120, + 88, + 490, + 219.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 120, + 219.66666666666666, + 490, + 351.3333333333333 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 120, + 351.3333333333333, + 490, + 483.0 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 497, + 503, + 531 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 497, + 505, + 510 + ], + "spans": [ + { + "bbox": [ + 105, + 497, + 431, + 510 + ], + "score": 1.0, + "content": "Figure 17: Quality improvements by non-accelerated sampling with scaled", + "type": "text" + }, + { + "bbox": [ + 432, + 498, + 451, + 509 + ], + "score": 0.89, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 497, + 505, + 510 + ], + "score": 1.0, + "content": ". We observe", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 509, + 505, + 521 + ], + "spans": [ + { + "bbox": [ + 105, + 509, + 413, + 521 + ], + "score": 1.0, + "content": "quality improvements by non-accelerated (1000-step) sampling with scaled", + "type": "text" + }, + { + "bbox": [ + 414, + 509, + 433, + 520 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 509, + 505, + 521 + ], + "score": 1.0, + "content": "from accelerated", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 520, + 416, + 532 + ], + "spans": [ + { + "bbox": [ + 106, + 520, + 229, + 532 + ], + "score": 1.0, + "content": "(40-step) training described in", + "type": "text" + }, + { + "bbox": [ + 229, + 520, + 250, + 531 + ], + "score": 0.6, + "content": "\\ S 3 . 4", + "type": "inline_equation" + }, + { + "bbox": [ + 250, + 520, + 416, + 532 + ], + "score": 1.0, + "content": ". Please zoom in for detailed comparison.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5 + } + ], + "index": 3.5 + }, + { + "type": "interline_equation", + "bbox": [ + 111, + 546, + 504, + 558 + ], + "lines": [ + { + "bbox": [ + 111, + 546, + 504, + 558 + ], + "spans": [ + { + "bbox": [ + 111, + 546, + 504, + 558 + ], + "score": 0.75, + "content": "t _ { e d i t } = 9 0 0 ~ t _ { e d i t } = 8 0 0 ~ t _ { e d i t } = 7 0 0 ~ t _ { e d i t } = 6 0 0 ~ t _ { e d i t } = 5 0 0 ~ t _ { e d i t } = 4 0 0 ~ t _ { e d i t } = 3 0 0 ~ t _ { e d i t } = 2 0 0 ~ t _ { e d i t } = 1 0 0", + "type": "interline_equation", + "image_path": "35a76084e6ebb069475f8af856b6fca2df0d3f6c42c0262eb89b774ab89c58e8.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 111, + 546, + 504, + 558 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "image", + "bbox": [ + 109, + 560, + 502, + 647 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 109, + 560, + 502, + 647 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 109, + 560, + 502, + 647 + ], + "spans": [ + { + "bbox": [ + 109, + 560, + 502, + 647 + ], + "score": 0.932, + "type": "image", + "image_path": "5e9f94c3b8ff254f0b1c246dba27d57fe4749c6b53d4da39b435229817caa682.jpg" + } + ] + } + ], + "index": 9, + "virtual_lines": [ + { + "bbox": [ + 109, + 560, + 502, + 589.0 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 109, + 589.0, + 502, + 618.0 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 109, + 618.0, + 502, + 647.0 + ], + "spans": [], + "index": 10 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 660, + 504, + 694 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 659, + 506, + 673 + ], + "spans": [ + { + "bbox": [ + 106, + 659, + 292, + 673 + ], + "score": 1.0, + "content": "Figure 18: Importance of choosing proper", + "type": "text" + }, + { + "bbox": [ + 293, + 661, + 311, + 672 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 312, + 659, + 397, + 673 + ], + "score": 1.0, + "content": ". We explore various", + "type": "text" + }, + { + "bbox": [ + 397, + 661, + 417, + 672 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 417, + 659, + 506, + 673 + ], + "score": 1.0, + "content": "with smiling. Too", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 670, + 506, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 506, + 684 + ], + "score": 1.0, + "content": "short editing interval struggles to manipulate attributes. Excessive editing strength results in de-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 682, + 170, + 696 + ], + "spans": [ + { + "bbox": [ + 105, + 682, + 170, + 696 + ], + "score": 1.0, + "content": "graded images.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12 + } + ], + "index": 10.5 + } + ], + "page_idx": 19, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 340, + 84, + 390, + 95 + ], + "lines": [ + { + "bbox": [ + 340, + 83, + 391, + 96 + ], + "spans": [ + { + "bbox": [ + 340, + 83, + 391, + 96 + ], + "score": 1.0, + "content": "Smiling with", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0, + "bbox_fs": [ + 340, + 83, + 391, + 96 + ] + }, + { + "type": "image", + "bbox": [ + 120, + 88, + 490, + 483 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 120, + 88, + 490, + 483 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 120, + 88, + 490, + 483 + ], + "spans": [ + { + "bbox": [ + 120, + 88, + 490, + 483 + ], + "score": 0.956, + "type": "image", + "image_path": "e119805b4e47eb46571e4d9c938e16083540325efbc089de26542ba2e25cbc3d.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 120, + 88, + 490, + 219.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 120, + 219.66666666666666, + 490, + 351.3333333333333 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 120, + 351.3333333333333, + 490, + 483.0 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 497, + 503, + 531 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 497, + 505, + 510 + ], + "spans": [ + { + "bbox": [ + 105, + 497, + 431, + 510 + ], + "score": 1.0, + "content": "Figure 17: Quality improvements by non-accelerated sampling with scaled", + "type": "text" + }, + { + "bbox": [ + 432, + 498, + 451, + 509 + ], + "score": 0.89, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 497, + 505, + 510 + ], + "score": 1.0, + "content": ". We observe", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 509, + 505, + 521 + ], + "spans": [ + { + "bbox": [ + 105, + 509, + 413, + 521 + ], + "score": 1.0, + "content": "quality improvements by non-accelerated (1000-step) sampling with scaled", + "type": "text" + }, + { + "bbox": [ + 414, + 509, + 433, + 520 + ], + "score": 0.9, + "content": "\\Delta \\boldsymbol { h } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 509, + 505, + 521 + ], + "score": 1.0, + "content": "from accelerated", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 520, + 416, + 532 + ], + "spans": [ + { + "bbox": [ + 106, + 520, + 229, + 532 + ], + "score": 1.0, + "content": "(40-step) training described in", + "type": "text" + }, + { + "bbox": [ + 229, + 520, + 250, + 531 + ], + "score": 0.6, + "content": "\\ S 3 . 4", + "type": "inline_equation" + }, + { + "bbox": [ + 250, + 520, + 416, + 532 + ], + "score": 1.0, + "content": ". Please zoom in for detailed comparison.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5 + } + ], + "index": 3.5 + }, + { + "type": "interline_equation", + "bbox": [ + 111, + 546, + 504, + 558 + ], + "lines": [ + { + "bbox": [ + 111, + 546, + 504, + 558 + ], + "spans": [ + { + "bbox": [ + 111, + 546, + 504, + 558 + ], + "score": 0.75, + "content": "t _ { e d i t } = 9 0 0 ~ t _ { e d i t } = 8 0 0 ~ t _ { e d i t } = 7 0 0 ~ t _ { e d i t } = 6 0 0 ~ t _ { e d i t } = 5 0 0 ~ t _ { e d i t } = 4 0 0 ~ t _ { e d i t } = 3 0 0 ~ t _ { e d i t } = 2 0 0 ~ t _ { e d i t } = 1 0 0", + "type": "interline_equation", + "image_path": "35a76084e6ebb069475f8af856b6fca2df0d3f6c42c0262eb89b774ab89c58e8.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 111, + 546, + 504, + 558 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "image", + "bbox": [ + 109, + 560, + 502, + 647 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 109, + 560, + 502, + 647 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 109, + 560, + 502, + 647 + ], + "spans": [ + { + "bbox": [ + 109, + 560, + 502, + 647 + ], + "score": 0.932, + "type": "image", + "image_path": "5e9f94c3b8ff254f0b1c246dba27d57fe4749c6b53d4da39b435229817caa682.jpg" + } + ] + } + ], + "index": 9, + "virtual_lines": [ + { + "bbox": [ + 109, + 560, + 502, + 589.0 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 109, + 589.0, + 502, + 618.0 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 109, + 618.0, + 502, + 647.0 + ], + "spans": [], + "index": 10 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 660, + 504, + 694 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 659, + 506, + 673 + ], + "spans": [ + { + "bbox": [ + 106, + 659, + 292, + 673 + ], + "score": 1.0, + "content": "Figure 18: Importance of choosing proper", + "type": "text" + }, + { + "bbox": [ + 293, + 661, + 311, + 672 + ], + "score": 0.9, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 312, + 659, + 397, + 673 + ], + "score": 1.0, + "content": ". We explore various", + "type": "text" + }, + { + "bbox": [ + 397, + 661, + 417, + 672 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 417, + 659, + 506, + 673 + ], + "score": 1.0, + "content": "with smiling. Too", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 670, + 506, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 506, + 684 + ], + "score": 1.0, + "content": "short editing interval struggles to manipulate attributes. Excessive editing strength results in de-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 682, + 170, + 696 + ], + "spans": [ + { + "bbox": [ + 105, + 682, + 170, + 696 + ], + "score": 1.0, + "content": "graded images.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12 + } + ], + "index": 10.5 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 79, + 506, + 416 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 79, + 506, + 416 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 79, + 506, + 416 + ], + "spans": [ + { + "bbox": [ + 106, + 79, + 506, + 416 + ], + "score": 0.981, + "html": "
Datasetsrc_txttrg_txtCLIP similaritytedittboost
CelebA-HQ LSUN-churchPerson FaceYoung person0.905515167
Smiling face0.899513167
FaceSad face0.894513167
FaceAngry face0.892512167
FaceTanned face0.886512167
FaceDisgusted face0.880511167
Person HumanPerson with makeup0.875509167
PersonZombie0.868506167
PersonPerson with bald head0.861505167
Person with curly hair0.835499167
Human PersonNeanderthal0.802490167
PersonMark Zuckerberg0.797489167
HumanNicolas Cage0.710461167
PhotoPainting in the style of Pixar0.667446167
PhotoPainting in Modigliani style Self-portrait by Frida Kahlo0.565403167
ChurchGothic Church0.443321167
ChurchTemple0.912 0.898371293
ChurchDepartment store0.841367293
ChurchWooden House349293
Church0.793333293
ChurchAncient traditional Asian tower0.784330293
ChurchRed brick wall Church0.774326293
BedroomFactory Princess Bedroom0.702301293
LSUN-bedroomBedroom Hotel Bedroom0.912370221
AFHQDog0.917371221
DogHappy Dog0.883430167
Sleepy Dog0.866422167
DogAngry Dog0.860419167
Dog DogWolf0.850412167
METFACESPainting of a personYorkshire Terrier0.690302167
Painting of a personPainting of a Sad person Painting of a Smiling person0.921 0.908330 320180 180
Painting of a personPainting of a Disgusted person0.879291180
", + "type": "table", + "image_path": "61cb1f8e818cb9695091612d9e566dc3bfdae8b922b211e762275091594b9337.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 79, + 506, + 191.33333333333331 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 191.33333333333331, + 506, + 303.66666666666663 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 303.66666666666663, + 506, + 415.99999999999994 + ], + "spans": [], + "index": 2 + } + ] + } + ], + "index": 1 + }, + { + "type": "image", + "bbox": [ + 106, + 455, + 502, + 640 + ], + "blocks": [ + { + "type": "image_caption", + "bbox": [ + 129, + 424, + 479, + 436 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 129, + 423, + 481, + 437 + ], + "spans": [ + { + "bbox": [ + 129, + 423, + 272, + 437 + ], + "score": 1.0, + "content": "Table 2: Prompts, CLIP similarity,", + "type": "text" + }, + { + "bbox": [ + 272, + 425, + 290, + 435 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 291, + 423, + 311, + 437 + ], + "score": 1.0, + "content": ", and", + "type": "text" + }, + { + "bbox": [ + 311, + 425, + 335, + 435 + ], + "score": 0.89, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 336, + 423, + 481, + 437 + ], + "score": 1.0, + "content": "for all attributes in the experiments.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 3 + }, + { + "type": "image_body", + "bbox": [ + 106, + 455, + 502, + 640 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 455, + 502, + 640 + ], + "spans": [ + { + "bbox": [ + 106, + 455, + 502, + 640 + ], + "score": 0.968, + "type": "image", + "image_path": "d400055c06dce9745d0694a750ea708aaf62c69eec8d8f94c39559d4b4c1588e.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 106, + 455, + 502, + 516.6666666666666 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 106, + 516.6666666666666, + 502, + 578.3333333333333 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 106, + 578.3333333333333, + 502, + 639.9999999999999 + ], + "spans": [], + "index": 6 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 128, + 652, + 480, + 666 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 127, + 652, + 481, + 667 + ], + "spans": [ + { + "bbox": [ + 127, + 652, + 240, + 667 + ], + "score": 1.0, + "content": "Figure 19: Average LPIPS", + "type": "text" + }, + { + "bbox": [ + 241, + 653, + 271, + 665 + ], + "score": 0.8, + "content": "( \\boldsymbol { \\mathbf { \\mathit { x } } } , \\boldsymbol { \\mathbf { \\mathit { P } } } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 271, + 652, + 320, + 667 + ], + "score": 1.0, + "content": "and LPIPS", + "type": "text" + }, + { + "bbox": [ + 321, + 653, + 350, + 665 + ], + "score": 0.59, + "content": "( \\pmb { x } , \\pmb { x } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 350, + 652, + 481, + 667 + ], + "score": 1.0, + "content": "of 100 samples on all datasets.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 7 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 107, + 698, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "Figure 20 shows the effect of quality boosting with the original DDIM reverse process. Although", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 710, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 505, + 722 + ], + "score": 1.0, + "content": "the reverse process of DDIM has a nearly-perfect inversion property, we observe some noise by", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 721, + 504, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 504, + 733 + ], + "score": 1.0, + "content": "zooming in. Our quality boosting improves the quality of a sample and concurrently keeps nearly", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9 + } + ], + "page_idx": 20, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 312, + 765 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 312, + 765 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 14 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 79, + 506, + 416 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 79, + 506, + 416 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 79, + 506, + 416 + ], + "spans": [ + { + "bbox": [ + 106, + 79, + 506, + 416 + ], + "score": 0.981, + "html": "
Datasetsrc_txttrg_txtCLIP similaritytedittboost
CelebA-HQ LSUN-churchPerson FaceYoung person0.905515167
Smiling face0.899513167
FaceSad face0.894513167
FaceAngry face0.892512167
FaceTanned face0.886512167
FaceDisgusted face0.880511167
Person HumanPerson with makeup0.875509167
PersonZombie0.868506167
PersonPerson with bald head0.861505167
Person with curly hair0.835499167
Human PersonNeanderthal0.802490167
PersonMark Zuckerberg0.797489167
HumanNicolas Cage0.710461167
PhotoPainting in the style of Pixar0.667446167
PhotoPainting in Modigliani style Self-portrait by Frida Kahlo0.565403167
ChurchGothic Church0.443321167
ChurchTemple0.912 0.898371293
ChurchDepartment store0.841367293
ChurchWooden House349293
Church0.793333293
ChurchAncient traditional Asian tower0.784330293
ChurchRed brick wall Church0.774326293
BedroomFactory Princess Bedroom0.702301293
LSUN-bedroomBedroom Hotel Bedroom0.912370221
AFHQDog0.917371221
DogHappy Dog0.883430167
Sleepy Dog0.866422167
DogAngry Dog0.860419167
Dog DogWolf0.850412167
METFACESPainting of a personYorkshire Terrier0.690302167
Painting of a personPainting of a Sad person Painting of a Smiling person0.921 0.908330 320180 180
Painting of a personPainting of a Disgusted person0.879291180
", + "type": "table", + "image_path": "61cb1f8e818cb9695091612d9e566dc3bfdae8b922b211e762275091594b9337.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 79, + 506, + 191.33333333333331 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 191.33333333333331, + 506, + 303.66666666666663 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 303.66666666666663, + 506, + 415.99999999999994 + ], + "spans": [], + "index": 2 + } + ] + } + ], + "index": 1 + }, + { + "type": "image", + "bbox": [ + 106, + 455, + 502, + 640 + ], + "blocks": [ + { + "type": "image_caption", + "bbox": [ + 129, + 424, + 479, + 436 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 129, + 423, + 481, + 437 + ], + "spans": [ + { + "bbox": [ + 129, + 423, + 272, + 437 + ], + "score": 1.0, + "content": "Table 2: Prompts, CLIP similarity,", + "type": "text" + }, + { + "bbox": [ + 272, + 425, + 290, + 435 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 291, + 423, + 311, + 437 + ], + "score": 1.0, + "content": ", and", + "type": "text" + }, + { + "bbox": [ + 311, + 425, + 335, + 435 + ], + "score": 0.89, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 336, + 423, + 481, + 437 + ], + "score": 1.0, + "content": "for all attributes in the experiments.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 3 + }, + { + "type": "image_body", + "bbox": [ + 106, + 455, + 502, + 640 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 455, + 502, + 640 + ], + "spans": [ + { + "bbox": [ + 106, + 455, + 502, + 640 + ], + "score": 0.968, + "type": "image", + "image_path": "d400055c06dce9745d0694a750ea708aaf62c69eec8d8f94c39559d4b4c1588e.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 106, + 455, + 502, + 516.6666666666666 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 106, + 516.6666666666666, + 502, + 578.3333333333333 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 106, + 578.3333333333333, + 502, + 639.9999999999999 + ], + "spans": [], + "index": 6 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 128, + 652, + 480, + 666 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 127, + 652, + 481, + 667 + ], + "spans": [ + { + "bbox": [ + 127, + 652, + 240, + 667 + ], + "score": 1.0, + "content": "Figure 19: Average LPIPS", + "type": "text" + }, + { + "bbox": [ + 241, + 653, + 271, + 665 + ], + "score": 0.8, + "content": "( \\boldsymbol { \\mathbf { \\mathit { x } } } , \\boldsymbol { \\mathbf { \\mathit { P } } } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 271, + 652, + 320, + 667 + ], + "score": 1.0, + "content": "and LPIPS", + "type": "text" + }, + { + "bbox": [ + 321, + 653, + 350, + 665 + ], + "score": 0.59, + "content": "( \\pmb { x } , \\pmb { x } _ { t } )", + "type": "inline_equation" + }, + { + "bbox": [ + 350, + 652, + 481, + 667 + ], + "score": 1.0, + "content": "of 100 samples on all datasets.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 7 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 107, + 698, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "Figure 20 shows the effect of quality boosting with the original DDIM reverse process. Although", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 710, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 505, + 722 + ], + "score": 1.0, + "content": "the reverse process of DDIM has a nearly-perfect inversion property, we observe some noise by", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 721, + 504, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 504, + 733 + ], + "score": 1.0, + "content": "zooming in. Our quality boosting improves the quality of a sample and concurrently keeps nearly", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9, + "bbox_fs": [ + 105, + 698, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 128, + 83, + 483, + 332 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 128, + 83, + 483, + 332 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 128, + 83, + 483, + 332 + ], + "spans": [ + { + "bbox": [ + 128, + 83, + 483, + 332 + ], + "score": 0.97, + "type": "image", + "image_path": "1434d974f30676869508e856aafe42dbc606c49aa0ad3a871abe5831d5d8d9ad.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 128, + 83, + 483, + 166.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 128, + 166.0, + 483, + 249.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 128, + 249.0, + 483, + 332.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 345, + 505, + 380 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 345, + 505, + 359 + ], + "spans": [ + { + "bbox": [ + 106, + 345, + 505, + 359 + ], + "score": 1.0, + "content": "Figure 20: Ablation study of quality boosting on the original DDIM process without Asyrp. Our", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 357, + 505, + 370 + ], + "spans": [ + { + "bbox": [ + 105, + 357, + 505, + 370 + ], + "score": 1.0, + "content": "quality boosting enhances fine details and prevents images from being noisy in the original DDIM", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 368, + 304, + 381 + ], + "spans": [ + { + "bbox": [ + 105, + 368, + 304, + 381 + ], + "score": 1.0, + "content": "process. Please zoom in for detailed comparison.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 350, + 385, + 378, + 395 + ], + "lines": [ + { + "bbox": [ + 350, + 385, + 378, + 395 + ], + "spans": [], + "index": 6 + } + ], + "index": 6 + }, + { + "type": "image", + "bbox": [ + 127, + 388, + 484, + 647 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 127, + 388, + 484, + 647 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 127, + 388, + 484, + 647 + ], + "spans": [ + { + "bbox": [ + 127, + 388, + 484, + 647 + ], + "score": 0.967, + "type": "image", + "image_path": "c3c2bcb78752a1c696fd47a0788206292074dab31e50f5bcc13a4530d4914cd3.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 127, + 388, + 484, + 474.3333333333333 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 127, + 474.3333333333333, + 484, + 560.6666666666666 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 127, + 560.6666666666666, + 484, + 647.0 + ], + "spans": [], + "index": 9 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 659, + 505, + 693 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 658, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 658, + 505, + 672 + ], + "score": 1.0, + "content": "Figure 21: Ablation study of quality boosting with Asyrp. Our quality boosting enhances fine", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 669, + 505, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 669, + 505, + 684 + ], + "score": 1.0, + "content": "details and prevents images from being noisy. Note that the source of degradation is DDIM process,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 681, + 410, + 694 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 410, + 694 + ], + "score": 1.0, + "content": "not Asyrp, confirmed in Figure 20. Please zoom in for detailed comparison.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11 + } + ], + "index": 9.5 + } + ], + "page_idx": 21, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 765 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 765 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 128, + 83, + 483, + 332 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 128, + 83, + 483, + 332 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 128, + 83, + 483, + 332 + ], + "spans": [ + { + "bbox": [ + 128, + 83, + 483, + 332 + ], + "score": 0.97, + "type": "image", + "image_path": "1434d974f30676869508e856aafe42dbc606c49aa0ad3a871abe5831d5d8d9ad.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 128, + 83, + 483, + 166.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 128, + 166.0, + 483, + 249.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 128, + 249.0, + 483, + 332.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 345, + 505, + 380 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 345, + 505, + 359 + ], + "spans": [ + { + "bbox": [ + 106, + 345, + 505, + 359 + ], + "score": 1.0, + "content": "Figure 20: Ablation study of quality boosting on the original DDIM process without Asyrp. Our", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 357, + 505, + 370 + ], + "spans": [ + { + "bbox": [ + 105, + 357, + 505, + 370 + ], + "score": 1.0, + "content": "quality boosting enhances fine details and prevents images from being noisy in the original DDIM", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 368, + 304, + 381 + ], + "spans": [ + { + "bbox": [ + 105, + 368, + 304, + 381 + ], + "score": 1.0, + "content": "process. Please zoom in for detailed comparison.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 350, + 385, + 378, + 395 + ], + "lines": [ + { + "bbox": [ + 350, + 385, + 378, + 395 + ], + "spans": [], + "index": 6 + } + ], + "index": 6, + "bbox_fs": [ + 350, + 385, + 378, + 395 + ] + }, + { + "type": "image", + "bbox": [ + 127, + 388, + 484, + 647 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 127, + 388, + 484, + 647 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 127, + 388, + 484, + 647 + ], + "spans": [ + { + "bbox": [ + 127, + 388, + 484, + 647 + ], + "score": 0.967, + "type": "image", + "image_path": "c3c2bcb78752a1c696fd47a0788206292074dab31e50f5bcc13a4530d4914cd3.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 127, + 388, + 484, + 474.3333333333333 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 127, + 474.3333333333333, + 484, + 560.6666666666666 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 127, + 560.6666666666666, + 484, + 647.0 + ], + "spans": [], + "index": 9 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 659, + 505, + 693 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 658, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 658, + 505, + 672 + ], + "score": 1.0, + "content": "Figure 21: Ablation study of quality boosting with Asyrp. Our quality boosting enhances fine", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 669, + 505, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 669, + 505, + 684 + ], + "score": 1.0, + "content": "details and prevents images from being noisy. Note that the source of degradation is DDIM process,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 681, + 410, + 694 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 410, + 694 + ], + "score": 1.0, + "content": "not Asyrp, confirmed in Figure 20. Please zoom in for detailed comparison.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11 + } + ], + "index": 9.5 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 129, + 82, + 483, + 228 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 129, + 82, + 483, + 228 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 129, + 82, + 483, + 228 + ], + "spans": [ + { + "bbox": [ + 129, + 82, + 483, + 228 + ], + "score": 0.967, + "type": "image", + "image_path": "9105ae3ff594358c3dd1ad873a025ed2c74b6b492fe058518f574f1ebe0cc84f.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 129, + 82, + 483, + 130.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 129, + 130.66666666666666, + 483, + 179.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 129, + 179.33333333333331, + 483, + 227.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 240, + 505, + 275 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 240, + 506, + 253 + ], + "spans": [ + { + "bbox": [ + 106, + 240, + 506, + 253 + ], + "score": 1.0, + "content": "Figure 22: Analyzing the effect of the hyperparameters The figure shows that the calculated", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 252, + 505, + 264 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 505, + 264 + ], + "score": 1.0, + "content": "parameter works effectively as the maximum boundary of editing while maintaining the quality of", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 262, + 151, + 276 + ], + "spans": [ + { + "bbox": [ + 105, + 262, + 151, + 276 + ], + "score": 1.0, + "content": "the image.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 127, + 297, + 483, + 446 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 127, + 297, + 483, + 446 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 127, + 297, + 483, + 446 + ], + "spans": [ + { + "bbox": [ + 127, + 297, + 483, + 446 + ], + "score": 0.962, + "type": "image", + "image_path": "067bbe65a525c781959e6f88a9e6ed2322b6ae14dd2301e761c08b57fdce7806.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 127, + 297, + 483, + 346.6666666666667 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 127, + 346.6666666666667, + 483, + 396.33333333333337 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 127, + 396.33333333333337, + 483, + 446.00000000000006 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 104, + 457, + 505, + 481 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 457, + 505, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 457, + 505, + 470 + ], + "score": 1.0, + "content": "Figure 23: Analyzing the effect of the hyperparameters We observe that the quality of results has", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 104, + 466, + 303, + 484 + ], + "spans": [ + { + "bbox": [ + 104, + 466, + 160, + 484 + ], + "score": 1.0, + "content": "robustness to", + "type": "text" + }, + { + "bbox": [ + 161, + 470, + 187, + 480 + ], + "score": 0.9, + "content": "\\gamma _ { t _ { \\mathrm { b o o s t } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 187, + 466, + 271, + 484 + ], + "score": 1.0, + "content": ", except for too large", + "type": "text" + }, + { + "bbox": [ + 271, + 470, + 298, + 481 + ], + "score": 0.89, + "content": "\\gamma _ { t _ { \\mathrm { b o o s t } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 298, + 466, + 303, + 484 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + } + ], + "index": 8.25 + }, + { + "type": "text", + "bbox": [ + 108, + 511, + 504, + 534 + ], + "lines": [ + { + "bbox": [ + 106, + 510, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 106, + 510, + 282, + 524 + ], + "score": 1.0, + "content": "perfect inversion property. We observe that", + "type": "text" + }, + { + "bbox": [ + 282, + 513, + 307, + 523 + ], + "score": 0.9, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 307, + 510, + 505, + 524 + ], + "score": 1.0, + "content": "is not sensitive, but the larger interval brings the", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 523, + 177, + 534 + ], + "spans": [ + { + "bbox": [ + 106, + 523, + 177, + 534 + ], + "score": 1.0, + "content": "less preservation.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5 + }, + { + "type": "text", + "bbox": [ + 107, + 539, + 398, + 551 + ], + "lines": [ + { + "bbox": [ + 105, + 536, + 398, + 555 + ], + "spans": [ + { + "bbox": [ + 105, + 536, + 398, + 555 + ], + "score": 1.0, + "content": "Figure 21 shows quality improvements by our quality boosting in Asyrp.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 13 + }, + { + "type": "title", + "bbox": [ + 107, + 573, + 189, + 586 + ], + "lines": [ + { + "bbox": [ + 104, + 572, + 190, + 587 + ], + "spans": [ + { + "bbox": [ + 104, + 572, + 190, + 587 + ], + "score": 1.0, + "content": "I ALGORITHM", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14 + }, + { + "type": "text", + "bbox": [ + 106, + 602, + 504, + 624 + ], + "lines": [ + { + "bbox": [ + 105, + 600, + 506, + 615 + ], + "spans": [ + { + "bbox": [ + 105, + 600, + 506, + 615 + ], + "score": 1.0, + "content": "Figure 24 illustrates generative process. Algorithm 1 and 2 describe training algorithm and inference", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 612, + 240, + 626 + ], + "spans": [ + { + "bbox": [ + 106, + 612, + 240, + 626 + ], + "score": 1.0, + "content": "algorithm of Asyrp, respectively.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5 + }, + { + "type": "title", + "bbox": [ + 108, + 646, + 223, + 659 + ], + "lines": [ + { + "bbox": [ + 105, + 646, + 225, + 661 + ], + "spans": [ + { + "bbox": [ + 105, + 646, + 225, + 661 + ], + "score": 1.0, + "content": "J TRAINING DETAILS", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 17 + }, + { + "type": "title", + "bbox": [ + 108, + 675, + 216, + 686 + ], + "lines": [ + { + "bbox": [ + 105, + 675, + 217, + 688 + ], + "spans": [ + { + "bbox": [ + 105, + 675, + 217, + 688 + ], + "score": 1.0, + "content": "J.1 LOSS COEFFICIENTS", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 309, + 711 + ], + "score": 1.0, + "content": "Table 3 reports loss coefficients for each attribute.", + "type": "text" + }, + { + "bbox": [ + 309, + 699, + 340, + 710 + ], + "score": 0.75, + "content": "\\lambda _ { \\mathrm { { C L I P } } } \\mathrm { { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 340, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "near 0.8 are suitable for most in-domain", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "attributes. For unseen domains, higher coefficient leads to more noticeable changes. Note that we", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 104, + 720, + 492, + 734 + ], + "spans": [ + { + "bbox": [ + 104, + 720, + 122, + 734 + ], + "score": 1.0, + "content": "use", + "type": "text" + }, + { + "bbox": [ + 122, + 721, + 184, + 732 + ], + "score": 0.85, + "content": "\\lambda _ { \\mathrm { r e c o n } } = \\mathbf { C } \\mathbf { L } \\mathbf { I } \\mathbf { P }", + "type": "inline_equation" + }, + { + "bbox": [ + 184, + 720, + 225, + 734 + ], + "score": 1.0, + "content": "similarity", + "type": "text" + }, + { + "bbox": [ + 225, + 721, + 239, + 731 + ], + "score": 0.83, + "content": "^ \\ast 3", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 720, + 492, + 734 + ], + "score": 1.0, + "content": "which reduces L1 loss when an attribute needs a lot of change.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20 + } + ], + "page_idx": 22, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 129, + 82, + 483, + 228 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 129, + 82, + 483, + 228 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 129, + 82, + 483, + 228 + ], + "spans": [ + { + "bbox": [ + 129, + 82, + 483, + 228 + ], + "score": 0.967, + "type": "image", + "image_path": "9105ae3ff594358c3dd1ad873a025ed2c74b6b492fe058518f574f1ebe0cc84f.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 129, + 82, + 483, + 130.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 129, + 130.66666666666666, + 483, + 179.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 129, + 179.33333333333331, + 483, + 227.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 240, + 505, + 275 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 240, + 506, + 253 + ], + "spans": [ + { + "bbox": [ + 106, + 240, + 506, + 253 + ], + "score": 1.0, + "content": "Figure 22: Analyzing the effect of the hyperparameters The figure shows that the calculated", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 252, + 505, + 264 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 505, + 264 + ], + "score": 1.0, + "content": "parameter works effectively as the maximum boundary of editing while maintaining the quality of", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 262, + 151, + 276 + ], + "spans": [ + { + "bbox": [ + 105, + 262, + 151, + 276 + ], + "score": 1.0, + "content": "the image.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 127, + 297, + 483, + 446 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 127, + 297, + 483, + 446 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 127, + 297, + 483, + 446 + ], + "spans": [ + { + "bbox": [ + 127, + 297, + 483, + 446 + ], + "score": 0.962, + "type": "image", + "image_path": "067bbe65a525c781959e6f88a9e6ed2322b6ae14dd2301e761c08b57fdce7806.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 127, + 297, + 483, + 346.6666666666667 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 127, + 346.6666666666667, + 483, + 396.33333333333337 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 127, + 396.33333333333337, + 483, + 446.00000000000006 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 104, + 457, + 505, + 481 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 457, + 505, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 457, + 505, + 470 + ], + "score": 1.0, + "content": "Figure 23: Analyzing the effect of the hyperparameters We observe that the quality of results has", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 104, + 466, + 303, + 484 + ], + "spans": [ + { + "bbox": [ + 104, + 466, + 160, + 484 + ], + "score": 1.0, + "content": "robustness to", + "type": "text" + }, + { + "bbox": [ + 161, + 470, + 187, + 480 + ], + "score": 0.9, + "content": "\\gamma _ { t _ { \\mathrm { b o o s t } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 187, + 466, + 271, + 484 + ], + "score": 1.0, + "content": ", except for too large", + "type": "text" + }, + { + "bbox": [ + 271, + 470, + 298, + 481 + ], + "score": 0.89, + "content": "\\gamma _ { t _ { \\mathrm { b o o s t } } }", + "type": "inline_equation" + }, + { + "bbox": [ + 298, + 466, + 303, + 484 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + } + ], + "index": 8.25 + }, + { + "type": "text", + "bbox": [ + 108, + 511, + 504, + 534 + ], + "lines": [ + { + "bbox": [ + 106, + 510, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 106, + 510, + 282, + 524 + ], + "score": 1.0, + "content": "perfect inversion property. We observe that", + "type": "text" + }, + { + "bbox": [ + 282, + 513, + 307, + 523 + ], + "score": 0.9, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 307, + 510, + 505, + 524 + ], + "score": 1.0, + "content": "is not sensitive, but the larger interval brings the", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 523, + 177, + 534 + ], + "spans": [ + { + "bbox": [ + 106, + 523, + 177, + 534 + ], + "score": 1.0, + "content": "less preservation.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5, + "bbox_fs": [ + 106, + 510, + 505, + 534 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 539, + 398, + 551 + ], + "lines": [ + { + "bbox": [ + 105, + 536, + 398, + 555 + ], + "spans": [ + { + "bbox": [ + 105, + 536, + 398, + 555 + ], + "score": 1.0, + "content": "Figure 21 shows quality improvements by our quality boosting in Asyrp.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 13, + "bbox_fs": [ + 105, + 536, + 398, + 555 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 573, + 189, + 586 + ], + "lines": [ + { + "bbox": [ + 104, + 572, + 190, + 587 + ], + "spans": [ + { + "bbox": [ + 104, + 572, + 190, + 587 + ], + "score": 1.0, + "content": "I ALGORITHM", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14 + }, + { + "type": "text", + "bbox": [ + 106, + 602, + 504, + 624 + ], + "lines": [ + { + "bbox": [ + 105, + 600, + 506, + 615 + ], + "spans": [ + { + "bbox": [ + 105, + 600, + 506, + 615 + ], + "score": 1.0, + "content": "Figure 24 illustrates generative process. Algorithm 1 and 2 describe training algorithm and inference", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 612, + 240, + 626 + ], + "spans": [ + { + "bbox": [ + 106, + 612, + 240, + 626 + ], + "score": 1.0, + "content": "algorithm of Asyrp, respectively.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5, + "bbox_fs": [ + 105, + 600, + 506, + 626 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 646, + 223, + 659 + ], + "lines": [ + { + "bbox": [ + 105, + 646, + 225, + 661 + ], + "spans": [ + { + "bbox": [ + 105, + 646, + 225, + 661 + ], + "score": 1.0, + "content": "J TRAINING DETAILS", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 17 + }, + { + "type": "title", + "bbox": [ + 108, + 675, + 216, + 686 + ], + "lines": [ + { + "bbox": [ + 105, + 675, + 217, + 688 + ], + "spans": [ + { + "bbox": [ + 105, + 675, + 217, + 688 + ], + "score": 1.0, + "content": "J.1 LOSS COEFFICIENTS", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 309, + 711 + ], + "score": 1.0, + "content": "Table 3 reports loss coefficients for each attribute.", + "type": "text" + }, + { + "bbox": [ + 309, + 699, + 340, + 710 + ], + "score": 0.75, + "content": "\\lambda _ { \\mathrm { { C L I P } } } \\mathrm { { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 340, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "near 0.8 are suitable for most in-domain", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "attributes. For unseen domains, higher coefficient leads to more noticeable changes. Note that we", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 104, + 720, + 492, + 734 + ], + "spans": [ + { + "bbox": [ + 104, + 720, + 122, + 734 + ], + "score": 1.0, + "content": "use", + "type": "text" + }, + { + "bbox": [ + 122, + 721, + 184, + 732 + ], + "score": 0.85, + "content": "\\lambda _ { \\mathrm { r e c o n } } = \\mathbf { C } \\mathbf { L } \\mathbf { I } \\mathbf { P }", + "type": "inline_equation" + }, + { + "bbox": [ + 184, + 720, + 225, + 734 + ], + "score": 1.0, + "content": "similarity", + "type": "text" + }, + { + "bbox": [ + 225, + 721, + 239, + 731 + ], + "score": 0.83, + "content": "^ \\ast 3", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 720, + 492, + 734 + ], + "score": 1.0, + "content": "which reduces L1 loss when an attribute needs a lot of change.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20, + "bbox_fs": [ + 104, + 698, + 505, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 81, + 504, + 233 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 81, + 504, + 233 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 81, + 504, + 233 + ], + "spans": [ + { + "bbox": [ + 107, + 81, + 504, + 233 + ], + "score": 0.971, + "type": "image", + "image_path": "4a4dfe0ce3ef126a67f1e86d022d8d15a8d17d09cceb878b5663ee16091daf9b.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 81, + 504, + 131.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 131.66666666666666, + 504, + 182.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 182.33333333333331, + 504, + 232.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 203, + 246, + 405, + 258 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 203, + 244, + 405, + 260 + ], + "spans": [ + { + "bbox": [ + 203, + 244, + 405, + 260 + ], + "score": 1.0, + "content": "Figure 24: Overview of our generative process.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 3 + } + ], + "index": 2.0 + }, + { + "type": "text", + "bbox": [ + 107, + 290, + 451, + 301 + ], + "lines": [ + { + "bbox": [ + 105, + 290, + 452, + 302 + ], + "spans": [ + { + "bbox": [ + 105, + 290, + 452, + 302 + ], + "score": 1.0, + "content": "J.2 TRAINING WITH RANDOM SAMPLING INSTEAD OF THE TRAINING DATASETS", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 106, + 313, + 505, + 416 + ], + "lines": [ + { + "bbox": [ + 105, + 313, + 505, + 328 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 505, + 328 + ], + "score": 1.0, + "content": "Apparently, for training, inverting real-images can be replaced by random sampling. It refers to using", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 325, + 506, + 342 + ], + "spans": [ + { + "bbox": [ + 106, + 326, + 167, + 339 + ], + "score": 0.93, + "content": "\\mathbf { \\boldsymbol { x } } _ { T } \\sim \\mathcal { N } ( 0 , \\mathbf { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 326, + 211, + 342 + ], + "score": 1.0, + "content": "instead of", + "type": "text" + }, + { + "bbox": [ + 212, + 325, + 349, + 340 + ], + "score": 0.92, + "content": "\\begin{array} { r } { \\pmb { x } _ { T } = \\bar { q } ( \\pmb { x } _ { 0 } ) \\cdot \\bar { \\prod } _ { t = 1 } ^ { T } q \\left( \\pmb { x } _ { t } \\mid \\bar { \\pmb { x } } _ { t - 1 } \\right) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 350, + 326, + 378, + 342 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 379, + 327, + 441, + 339 + ], + "score": 0.93, + "content": "\\pmb { x } _ { 0 } \\sim p _ { d a t a } ( x )", + "type": "inline_equation" + }, + { + "bbox": [ + 441, + 326, + 506, + 342 + ], + "score": 1.0, + "content": ". It allows us to", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 338, + 505, + 350 + ], + "spans": [ + { + "bbox": [ + 106, + 338, + 505, + 350 + ], + "score": 1.0, + "content": "train Asyrp only with the pretrained network and without extra dataset. Using random samples has", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 349, + 505, + 361 + ], + "spans": [ + { + "bbox": [ + 106, + 349, + 505, + 361 + ], + "score": 1.0, + "content": "tradeoff between preservation of contents and possible amount in editing. It take advantages when", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 104, + 360, + 505, + 372 + ], + "spans": [ + { + "bbox": [ + 104, + 360, + 505, + 372 + ], + "score": 1.0, + "content": "a target attribute requires large amount changes. We assume that the inversion of real-image is in", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 371, + 505, + 382 + ], + "spans": [ + { + "bbox": [ + 106, + 371, + 505, + 382 + ], + "score": 1.0, + "content": "the long tail of a Gaussian distribution because of realistic background or detailed clothes. On the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 381, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 381, + 506, + 394 + ], + "score": 1.0, + "content": "other hand, random noise is considered to be closer to the mean of the normal, so it is easier to", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 393, + 505, + 406 + ], + "spans": [ + { + "bbox": [ + 105, + 393, + 505, + 406 + ], + "score": 1.0, + "content": "find directions. It can easily bring larger changes but also easily alter the contents. On the contrary,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 404, + 315, + 417 + ], + "spans": [ + { + "bbox": [ + 105, + 404, + 315, + 417 + ], + "score": 1.0, + "content": "training with inversion shows the opposite property.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 106, + 421, + 504, + 443 + ], + "lines": [ + { + "bbox": [ + 106, + 420, + 505, + 433 + ], + "spans": [ + { + "bbox": [ + 106, + 420, + 142, + 433 + ], + "score": 1.0, + "content": "We train", + "type": "text" + }, + { + "bbox": [ + 143, + 421, + 153, + 433 + ], + "score": 0.86, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 153, + 420, + 505, + 433 + ], + "score": 1.0, + "content": "with random sampling for attributes whose identity preservation is not important to take", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 432, + 437, + 444 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 437, + 444 + ], + "score": 1.0, + "content": "advantage of these properties. The rightmost column in Table 3 shows the choices.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5 + }, + { + "type": "title", + "bbox": [ + 108, + 468, + 196, + 480 + ], + "lines": [ + { + "bbox": [ + 105, + 465, + 198, + 483 + ], + "spans": [ + { + "bbox": [ + 105, + 465, + 198, + 483 + ], + "score": 1.0, + "content": "K EVALUATION", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "title", + "bbox": [ + 108, + 497, + 189, + 508 + ], + "lines": [ + { + "bbox": [ + 105, + 496, + 190, + 510 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 190, + 510 + ], + "score": 1.0, + "content": "K.1 USER STUDY", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 17 + }, + { + "type": "text", + "bbox": [ + 107, + 521, + 505, + 577 + ], + "lines": [ + { + "bbox": [ + 107, + 522, + 505, + 533 + ], + "spans": [ + { + "bbox": [ + 107, + 522, + 505, + 533 + ], + "score": 1.0, + "content": "We conduct user study to compare the performance of Asyrp and DiffusionCLIP (Kim & Ye, 2021)", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 532, + 505, + 545 + ], + "spans": [ + { + "bbox": [ + 106, + 532, + 505, + 545 + ], + "score": 1.0, + "content": "on Celeba-HQ (Liu et al., 2015) and LSUN-church (Yu et al., 2015). We use official checkpoints", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "score": 1.0, + "content": "provided by DiffusionCLIP except for some facial attributes whose checkpoint do not exist. We", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 554, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 554, + 505, + 567 + ], + "score": 1.0, + "content": "tried our best to tune their hyperparameters following the manual for fair comparison. Example", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 564, + 247, + 578 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 247, + 578 + ], + "score": 1.0, + "content": "images are shown in Figure 25-27.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 582, + 505, + 615 + ], + "lines": [ + { + "bbox": [ + 106, + 582, + 505, + 594 + ], + "spans": [ + { + "bbox": [ + 106, + 582, + 505, + 594 + ], + "score": 1.0, + "content": "We use smiling and sad for in-domain CelebA-HQ attributes, Pixar and Neanderthal for", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 593, + 505, + 605 + ], + "spans": [ + { + "bbox": [ + 106, + 593, + 505, + 605 + ], + "score": 1.0, + "content": "unseen-domain CelebA-HQ attributes, and department store, ancient, and wooden for", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 604, + 168, + 615 + ], + "spans": [ + { + "bbox": [ + 105, + 604, + 168, + 615 + ], + "score": 1.0, + "content": "LSUN-church.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 108, + 621, + 505, + 654 + ], + "lines": [ + { + "bbox": [ + 105, + 620, + 505, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 620, + 505, + 633 + ], + "score": 1.0, + "content": "In unseen domain and Lsun-church, we use official checkpoints provided by DiffusionCLIP. We", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 632, + 505, + 644 + ], + "spans": [ + { + "bbox": [ + 106, + 632, + 505, + 644 + ], + "score": 1.0, + "content": "also randomly select 8 images for each CelebA-HQ attribute and 12 images for each LSUN-church", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 642, + 146, + 656 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 146, + 656 + ], + "score": 1.0, + "content": "attribute.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 108, + 659, + 505, + 693 + ], + "lines": [ + { + "bbox": [ + 106, + 659, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 659, + 505, + 672 + ], + "score": 1.0, + "content": "We observe that DiffusionCLIP works better in changing the holistic style of images. At the same", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 671, + 506, + 683 + ], + "spans": [ + { + "bbox": [ + 106, + 671, + 506, + 683 + ], + "score": 1.0, + "content": "time, it is short of the ability to bring semantic changes and suffers noisy results and a lack of", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 681, + 440, + 694 + ], + "spans": [ + { + "bbox": [ + 106, + 681, + 440, + 694 + ], + "score": 1.0, + "content": "diversity. The problems would be caused by fine-tuning the whole diffusion model.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "We use the following questions for the survey. 1) Quality: Which image quality do you think is", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "better? (clear and less noisy) 2) Attribute: Which image do you think is “Attribute(e.g., Smiling)", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 720, + 506, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 734 + ], + "score": 1.0, + "content": "naturally”? 3) Overall: Which image do you think is better considering the above evaluation criteria?", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33 + } + ], + "page_idx": 23, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 81, + 504, + 233 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 81, + 504, + 233 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 81, + 504, + 233 + ], + "spans": [ + { + "bbox": [ + 107, + 81, + 504, + 233 + ], + "score": 0.971, + "type": "image", + "image_path": "4a4dfe0ce3ef126a67f1e86d022d8d15a8d17d09cceb878b5663ee16091daf9b.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 81, + 504, + 131.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 131.66666666666666, + 504, + 182.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 182.33333333333331, + 504, + 232.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 203, + 246, + 405, + 258 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 203, + 244, + 405, + 260 + ], + "spans": [ + { + "bbox": [ + 203, + 244, + 405, + 260 + ], + "score": 1.0, + "content": "Figure 24: Overview of our generative process.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 3 + } + ], + "index": 2.0 + }, + { + "type": "text", + "bbox": [ + 107, + 290, + 451, + 301 + ], + "lines": [ + { + "bbox": [ + 105, + 290, + 452, + 302 + ], + "spans": [ + { + "bbox": [ + 105, + 290, + 452, + 302 + ], + "score": 1.0, + "content": "J.2 TRAINING WITH RANDOM SAMPLING INSTEAD OF THE TRAINING DATASETS", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4, + "bbox_fs": [ + 105, + 290, + 452, + 302 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 313, + 505, + 416 + ], + "lines": [ + { + "bbox": [ + 105, + 313, + 505, + 328 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 505, + 328 + ], + "score": 1.0, + "content": "Apparently, for training, inverting real-images can be replaced by random sampling. It refers to using", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 325, + 506, + 342 + ], + "spans": [ + { + "bbox": [ + 106, + 326, + 167, + 339 + ], + "score": 0.93, + "content": "\\mathbf { \\boldsymbol { x } } _ { T } \\sim \\mathcal { N } ( 0 , \\mathbf { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 326, + 211, + 342 + ], + "score": 1.0, + "content": "instead of", + "type": "text" + }, + { + "bbox": [ + 212, + 325, + 349, + 340 + ], + "score": 0.92, + "content": "\\begin{array} { r } { \\pmb { x } _ { T } = \\bar { q } ( \\pmb { x } _ { 0 } ) \\cdot \\bar { \\prod } _ { t = 1 } ^ { T } q \\left( \\pmb { x } _ { t } \\mid \\bar { \\pmb { x } } _ { t - 1 } \\right) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 350, + 326, + 378, + 342 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 379, + 327, + 441, + 339 + ], + "score": 0.93, + "content": "\\pmb { x } _ { 0 } \\sim p _ { d a t a } ( x )", + "type": "inline_equation" + }, + { + "bbox": [ + 441, + 326, + 506, + 342 + ], + "score": 1.0, + "content": ". It allows us to", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 338, + 505, + 350 + ], + "spans": [ + { + "bbox": [ + 106, + 338, + 505, + 350 + ], + "score": 1.0, + "content": "train Asyrp only with the pretrained network and without extra dataset. Using random samples has", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 349, + 505, + 361 + ], + "spans": [ + { + "bbox": [ + 106, + 349, + 505, + 361 + ], + "score": 1.0, + "content": "tradeoff between preservation of contents and possible amount in editing. It take advantages when", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 104, + 360, + 505, + 372 + ], + "spans": [ + { + "bbox": [ + 104, + 360, + 505, + 372 + ], + "score": 1.0, + "content": "a target attribute requires large amount changes. We assume that the inversion of real-image is in", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 371, + 505, + 382 + ], + "spans": [ + { + "bbox": [ + 106, + 371, + 505, + 382 + ], + "score": 1.0, + "content": "the long tail of a Gaussian distribution because of realistic background or detailed clothes. On the", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 381, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 381, + 506, + 394 + ], + "score": 1.0, + "content": "other hand, random noise is considered to be closer to the mean of the normal, so it is easier to", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 393, + 505, + 406 + ], + "spans": [ + { + "bbox": [ + 105, + 393, + 505, + 406 + ], + "score": 1.0, + "content": "find directions. It can easily bring larger changes but also easily alter the contents. On the contrary,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 404, + 315, + 417 + ], + "spans": [ + { + "bbox": [ + 105, + 404, + 315, + 417 + ], + "score": 1.0, + "content": "training with inversion shows the opposite property.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 9, + "bbox_fs": [ + 104, + 313, + 506, + 417 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 421, + 504, + 443 + ], + "lines": [ + { + "bbox": [ + 106, + 420, + 505, + 433 + ], + "spans": [ + { + "bbox": [ + 106, + 420, + 142, + 433 + ], + "score": 1.0, + "content": "We train", + "type": "text" + }, + { + "bbox": [ + 143, + 421, + 153, + 433 + ], + "score": 0.86, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 153, + 420, + 505, + 433 + ], + "score": 1.0, + "content": "with random sampling for attributes whose identity preservation is not important to take", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 432, + 437, + 444 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 437, + 444 + ], + "score": 1.0, + "content": "advantage of these properties. The rightmost column in Table 3 shows the choices.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5, + "bbox_fs": [ + 105, + 420, + 505, + 444 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 468, + 196, + 480 + ], + "lines": [ + { + "bbox": [ + 105, + 465, + 198, + 483 + ], + "spans": [ + { + "bbox": [ + 105, + 465, + 198, + 483 + ], + "score": 1.0, + "content": "K EVALUATION", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "title", + "bbox": [ + 108, + 497, + 189, + 508 + ], + "lines": [ + { + "bbox": [ + 105, + 496, + 190, + 510 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 190, + 510 + ], + "score": 1.0, + "content": "K.1 USER STUDY", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 17 + }, + { + "type": "text", + "bbox": [ + 107, + 521, + 505, + 577 + ], + "lines": [ + { + "bbox": [ + 107, + 522, + 505, + 533 + ], + "spans": [ + { + "bbox": [ + 107, + 522, + 505, + 533 + ], + "score": 1.0, + "content": "We conduct user study to compare the performance of Asyrp and DiffusionCLIP (Kim & Ye, 2021)", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 532, + 505, + 545 + ], + "spans": [ + { + "bbox": [ + 106, + 532, + 505, + 545 + ], + "score": 1.0, + "content": "on Celeba-HQ (Liu et al., 2015) and LSUN-church (Yu et al., 2015). We use official checkpoints", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 506, + 556 + ], + "score": 1.0, + "content": "provided by DiffusionCLIP except for some facial attributes whose checkpoint do not exist. We", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 554, + 505, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 554, + 505, + 567 + ], + "score": 1.0, + "content": "tried our best to tune their hyperparameters following the manual for fair comparison. Example", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 564, + 247, + 578 + ], + "spans": [ + { + "bbox": [ + 105, + 564, + 247, + 578 + ], + "score": 1.0, + "content": "images are shown in Figure 25-27.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 20, + "bbox_fs": [ + 105, + 522, + 506, + 578 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 582, + 505, + 615 + ], + "lines": [ + { + "bbox": [ + 106, + 582, + 505, + 594 + ], + "spans": [ + { + "bbox": [ + 106, + 582, + 505, + 594 + ], + "score": 1.0, + "content": "We use smiling and sad for in-domain CelebA-HQ attributes, Pixar and Neanderthal for", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 593, + 505, + 605 + ], + "spans": [ + { + "bbox": [ + 106, + 593, + 505, + 605 + ], + "score": 1.0, + "content": "unseen-domain CelebA-HQ attributes, and department store, ancient, and wooden for", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 604, + 168, + 615 + ], + "spans": [ + { + "bbox": [ + 105, + 604, + 168, + 615 + ], + "score": 1.0, + "content": "LSUN-church.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24, + "bbox_fs": [ + 105, + 582, + 505, + 615 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 621, + 505, + 654 + ], + "lines": [ + { + "bbox": [ + 105, + 620, + 505, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 620, + 505, + 633 + ], + "score": 1.0, + "content": "In unseen domain and Lsun-church, we use official checkpoints provided by DiffusionCLIP. We", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 632, + 505, + 644 + ], + "spans": [ + { + "bbox": [ + 106, + 632, + 505, + 644 + ], + "score": 1.0, + "content": "also randomly select 8 images for each CelebA-HQ attribute and 12 images for each LSUN-church", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 642, + 146, + 656 + ], + "spans": [ + { + "bbox": [ + 105, + 642, + 146, + 656 + ], + "score": 1.0, + "content": "attribute.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27, + "bbox_fs": [ + 105, + 620, + 505, + 656 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 659, + 505, + 693 + ], + "lines": [ + { + "bbox": [ + 106, + 659, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 659, + 505, + 672 + ], + "score": 1.0, + "content": "We observe that DiffusionCLIP works better in changing the holistic style of images. At the same", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 671, + 506, + 683 + ], + "spans": [ + { + "bbox": [ + 106, + 671, + 506, + 683 + ], + "score": 1.0, + "content": "time, it is short of the ability to bring semantic changes and suffers noisy results and a lack of", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 681, + 440, + 694 + ], + "spans": [ + { + "bbox": [ + 106, + 681, + 440, + 694 + ], + "score": 1.0, + "content": "diversity. The problems would be caused by fine-tuning the whole diffusion model.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30, + "bbox_fs": [ + 106, + 659, + 506, + 694 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 106, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "We use the following questions for the survey. 1) Quality: Which image quality do you think is", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "better? (clear and less noisy) 2) Attribute: Which image do you think is “Attribute(e.g., Smiling)", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 720, + 506, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 734 + ], + "score": 1.0, + "content": "naturally”? 3) Overall: Which image do you think is better considering the above evaluation criteria?", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33, + "bbox_fs": [ + 105, + 698, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 85, + 240, + 97 + ], + "lines": [ + { + "bbox": [ + 106, + 83, + 240, + 101 + ], + "spans": [ + { + "bbox": [ + 106, + 83, + 240, + 101 + ], + "score": 1.0, + "content": "Algorithm 1: Editing(Inference)", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 96, + 98, + 497, + 509 + ], + "lines": [ + { + "bbox": [ + 105, + 97, + 482, + 114 + ], + "spans": [ + { + "bbox": [ + 105, + 97, + 136, + 114 + ], + "score": 1.0, + "content": "Input:", + "type": "text" + }, + { + "bbox": [ + 136, + 102, + 147, + 112 + ], + "score": 0.73, + "content": "x _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 148, + 97, + 206, + 114 + ], + "score": 1.0, + "content": "(Input image),", + "type": "text" + }, + { + "bbox": [ + 206, + 100, + 239, + 113 + ], + "score": 0.35, + "content": "\\{ f _ { t } ^ { i } \\} _ { i = 1 } ^ { M }", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 97, + 426, + 114 + ], + "score": 1.0, + "content": "(M Neural implicit functions for M attributes),", + "type": "text" + }, + { + "bbox": [ + 426, + 100, + 458, + 113 + ], + "score": 0.83, + "content": "\\{ c _ { i } \\} _ { i = 1 } ^ { M }", + "type": "inline_equation" + }, + { + "bbox": [ + 459, + 97, + 482, + 114 + ], + "score": 1.0, + "content": "(user", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 135, + 110, + 491, + 124 + ], + "spans": [ + { + "bbox": [ + 135, + 110, + 328, + 124 + ], + "score": 1.0, + "content": "defined M scaling coefficients for M attributes),", + "type": "text" + }, + { + "bbox": [ + 329, + 113, + 338, + 123 + ], + "score": 0.8, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 339, + 110, + 447, + 124 + ], + "score": 1.0, + "content": "(frozen pretrained model),", + "type": "text" + }, + { + "bbox": [ + 447, + 113, + 467, + 124 + ], + "score": 0.89, + "content": "S _ { f o r }", + "type": "inline_equation" + }, + { + "bbox": [ + 468, + 110, + 491, + 124 + ], + "score": 1.0, + "content": "(# of", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 135, + 120, + 453, + 137 + ], + "spans": [ + { + "bbox": [ + 135, + 120, + 203, + 137 + ], + "score": 1.0, + "content": "inversion steps),", + "type": "text" + }, + { + "bbox": [ + 204, + 123, + 223, + 134 + ], + "score": 0.88, + "content": "S _ { g e n }", + "type": "inline_equation" + }, + { + "bbox": [ + 224, + 120, + 312, + 137 + ], + "score": 1.0, + "content": "(# of inference steps),", + "type": "text" + }, + { + "bbox": [ + 312, + 123, + 331, + 134 + ], + "score": 0.88, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 120, + 399, + 137 + ], + "score": 1.0, + "content": "(computed from", + "type": "text" + }, + { + "bbox": [ + 399, + 123, + 420, + 134 + ], + "score": 0.72, + "content": "\\ S 4 . 1", + "type": "inline_equation" + }, + { + "bbox": [ + 420, + 120, + 453, + 137 + ], + "score": 1.0, + "content": "), tboost", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 136, + 133, + 228, + 145 + ], + "spans": [ + { + "bbox": [ + 136, + 133, + 203, + 145 + ], + "score": 1.0, + "content": "(computed from", + "type": "text" + }, + { + "bbox": [ + 203, + 134, + 224, + 145 + ], + "score": 0.56, + "content": "\\ S 4 . 2", + "type": "inline_equation" + }, + { + "bbox": [ + 224, + 133, + 228, + 145 + ], + "score": 1.0, + "content": ")", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 96, + 147, + 317, + 163 + ], + "spans": [ + { + "bbox": [ + 96, + 147, + 185, + 163 + ], + "score": 1.0, + "content": "1 Function Editor", + "type": "text" + }, + { + "bbox": [ + 185, + 147, + 312, + 162 + ], + "score": 0.49, + "content": "( x _ { 0 } , \\epsilon _ { \\theta } , \\{ c _ { i } \\} _ { i = 1 } ^ { M } , S _ { f o r } , S _ { g e n } , * )", + "type": "inline_equation" + }, + { + "bbox": [ + 313, + 147, + 317, + 163 + ], + "score": 1.0, + "content": ":", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 122, + 163, + 292, + 177 + ], + "spans": [ + { + "bbox": [ + 122, + 163, + 292, + 177 + ], + "score": 1.0, + "content": "// step 1: Semantic encoding", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 99, + 171, + 276, + 198 + ], + "spans": [ + { + "bbox": [ + 99, + 182, + 104, + 189 + ], + "score": 1.0, + "content": "2", + "type": "text" + }, + { + "bbox": [ + 117, + 171, + 151, + 198 + ], + "score": 1.0, + "content": "Define", + "type": "text" + }, + { + "bbox": [ + 151, + 176, + 187, + 191 + ], + "score": 0.89, + "content": "\\{ \\tau _ { s } \\} _ { s = 1 } ^ { S _ { f o r } }", + "type": "inline_equation" + }, + { + "bbox": [ + 187, + 171, + 200, + 198 + ], + "score": 1.0, + "content": "s.t", + "type": "text" + }, + { + "bbox": [ + 200, + 178, + 229, + 190 + ], + "score": 0.75, + "content": "\\tau _ { 1 } = 0", + "type": "inline_equation" + }, + { + "bbox": [ + 229, + 171, + 232, + 198 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 233, + 178, + 276, + 192 + ], + "score": 0.76, + "content": "\\tau _ { S _ { f o r } } = T", + "type": "inline_equation" + } + ], + "index": 7 + }, + { + "bbox": [ + 98, + 189, + 239, + 205 + ], + "spans": [ + { + "bbox": [ + 98, + 193, + 105, + 201 + ], + "score": 1.0, + "content": "3", + "type": "text" + }, + { + "bbox": [ + 120, + 189, + 137, + 205 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 137, + 191, + 224, + 203 + ], + "score": 0.8, + "content": "s = 1 , 2 , . . . , S _ { f o r } - 1", + "type": "inline_equation" + }, + { + "bbox": [ + 224, + 189, + 239, + 205 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 99, + 203, + 201, + 214 + ], + "spans": [ + { + "bbox": [ + 99, + 205, + 105, + 212 + ], + "score": 1.0, + "content": "4", + "type": "text" + }, + { + "bbox": [ + 136, + 203, + 201, + 214 + ], + "score": 0.59, + "content": "\\epsilon \\epsilon _ { \\theta } ( x _ { \\tau _ { s } } , \\tau _ { s } )", + "type": "inline_equation" + } + ], + "index": 9 + }, + { + "bbox": [ + 98, + 214, + 267, + 228 + ], + "spans": [ + { + "bbox": [ + 98, + 217, + 105, + 225 + ], + "score": 1.0, + "content": "5", + "type": "text" + }, + { + "bbox": [ + 127, + 214, + 134, + 227 + ], + "score": 0.0, + "content": "", + "type": "text" + }, + { + "bbox": [ + 134, + 214, + 267, + 228 + ], + "score": 0.51, + "content": "x _ { \\tau _ { s + 1 } } = \\sqrt { \\alpha _ { \\tau _ { s } } } x _ { \\tau _ { s } } + \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon", + "type": "inline_equation" + } + ], + "index": 10 + }, + { + "bbox": [ + 98, + 232, + 425, + 272 + ], + "spans": [ + { + "bbox": [ + 98, + 251, + 105, + 259 + ], + "score": 1.0, + "content": "6", + "type": "text" + }, + { + "bbox": [ + 151, + 245, + 189, + 261 + ], + "score": 0.72, + "content": "\\{ \\tilde { \\tau } _ { s } \\} _ { s = 1 } ^ { S _ { g e n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 189, + 232, + 201, + 272 + ], + "score": 1.0, + "content": "Mas.t", + "type": "text" + }, + { + "bbox": [ + 202, + 248, + 231, + 259 + ], + "score": 0.71, + "content": "\\tilde { \\tau } _ { 1 } = 0", + "type": "inline_equation" + }, + { + "bbox": [ + 232, + 232, + 235, + 272 + ], + "score": 1.0, + "content": "a,", + "type": "text" + }, + { + "bbox": [ + 235, + 248, + 291, + 260 + ], + "score": 0.8, + "content": "\\tilde { \\tau } _ { S _ { e d i t } } = t _ { e d i t }", + "type": "inline_equation" + }, + { + "bbox": [ + 291, + 232, + 295, + 272 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 295, + 249, + 361, + 261 + ], + "score": 0.81, + "content": "\\tilde { \\tau } _ { S _ { n o i s e } } = t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 232, + 379, + 272 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 380, + 248, + 425, + 261 + ], + "score": 0.91, + "content": "\\tilde { \\tau } _ { S _ { g e n } } = T", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 98, + 261, + 188, + 274 + ], + "spans": [ + { + "bbox": [ + 98, + 263, + 105, + 271 + ], + "score": 1.0, + "content": "7", + "type": "text" + }, + { + "bbox": [ + 122, + 261, + 188, + 274 + ], + "score": 0.75, + "content": "\\tilde { x } _ { \\tilde { \\tau } _ { S _ { g e n } } } = x _ { \\tilde { \\tau } _ { S _ { f o r } } }", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 98, + 269, + 256, + 291 + ], + "spans": [ + { + "bbox": [ + 98, + 277, + 105, + 285 + ], + "score": 1.0, + "content": "8", + "type": "text" + }, + { + "bbox": [ + 117, + 269, + 137, + 291 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 137, + 275, + 239, + 286 + ], + "score": 0.67, + "content": "s = S _ { g e n } , S _ { g e n } - 1 , . . . , 2", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 269, + 256, + 291 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 136, + 283, + 254, + 298 + ], + "spans": [ + { + "bbox": [ + 136, + 283, + 254, + 298 + ], + "score": 1.0, + "content": "// phase 1: editing", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 98, + 295, + 209, + 308 + ], + "spans": [ + { + "bbox": [ + 98, + 298, + 106, + 307 + ], + "score": 1.0, + "content": "9", + "type": "text" + }, + { + "bbox": [ + 136, + 295, + 146, + 308 + ], + "score": 1.0, + "content": "if", + "type": "text" + }, + { + "bbox": [ + 146, + 297, + 185, + 307 + ], + "score": 0.84, + "content": "s \\geq S _ { e d i t }", + "type": "inline_equation" + }, + { + "bbox": [ + 186, + 295, + 209, + 308 + ], + "score": 1.0, + "content": "then", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 96, + 305, + 304, + 320 + ], + "spans": [ + { + "bbox": [ + 96, + 309, + 106, + 318 + ], + "score": 1.0, + "content": "10", + "type": "text" + }, + { + "bbox": [ + 150, + 305, + 233, + 320 + ], + "score": 1.0, + "content": "Extract feature map", + "type": "text" + }, + { + "bbox": [ + 234, + 307, + 249, + 318 + ], + "score": 0.88, + "content": "h _ { \\tilde { \\tau } _ { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 305, + 272, + 320 + ], + "score": 1.0, + "content": "from", + "type": "text" + }, + { + "bbox": [ + 273, + 307, + 304, + 319 + ], + "score": 0.89, + "content": "\\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } )", + "type": "inline_equation" + } + ], + "index": 16 + }, + { + "bbox": [ + 95, + 318, + 288, + 370 + ], + "spans": [ + { + "bbox": [ + 95, + 323, + 107, + 334 + ], + "score": 1.0, + "content": "11", + "type": "text" + }, + { + "bbox": [ + 149, + 318, + 288, + 370 + ], + "score": 0.32, + "content": "\\begin{array} { r l } & { \\Delta h _ { \\tilde { \\tau } _ { s } } = \\frac { S _ { f o r } } { S _ { g e n } } ( \\sum _ { i = 1 } ^ { M } c _ { i } f _ { \\tilde { \\tau } _ { s } } ^ { i } ( h _ { \\tilde { \\tau } _ { s } } ) ) } \\\\ & { \\tilde { \\epsilon } = \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } | \\Delta h _ { \\tilde { \\tau } _ { s } } ) } \\\\ & { \\epsilon = \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } ) } \\\\ & { \\sigma _ { \\tilde { \\tau } _ { s } } = 0 } \\end{array}", + "type": "inline_equation", + "image_path": "04afa24a0133c9be96d6d2df48a1abeb0876ad72b210390ee102ec52a597df14.jpg" + } + ], + "index": 18 + }, + { + "bbox": [ + 96, + 337, + 106, + 347 + ], + "spans": [ + { + "bbox": [ + 96, + 337, + 106, + 347 + ], + "score": 1.0, + "content": "12", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 95, + 347, + 107, + 358 + ], + "spans": [ + { + "bbox": [ + 95, + 347, + 107, + 358 + ], + "score": 1.0, + "content": "13", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 95, + 359, + 106, + 369 + ], + "spans": [ + { + "bbox": [ + 95, + 359, + 106, + 369 + ], + "score": 1.0, + "content": "14", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 137, + 370, + 266, + 385 + ], + "spans": [ + { + "bbox": [ + 137, + 370, + 266, + 385 + ], + "score": 1.0, + "content": "// phase 2: denoising", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 95, + 381, + 232, + 396 + ], + "spans": [ + { + "bbox": [ + 95, + 384, + 107, + 395 + ], + "score": 1.0, + "content": "15", + "type": "text" + }, + { + "bbox": [ + 136, + 381, + 164, + 396 + ], + "score": 1.0, + "content": "else if", + "type": "text" + }, + { + "bbox": [ + 164, + 383, + 209, + 394 + ], + "score": 0.84, + "content": "s \\geq S _ { n o i s e }", + "type": "inline_equation" + }, + { + "bbox": [ + 209, + 381, + 232, + 396 + ], + "score": 1.0, + "content": "then", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 95, + 393, + 221, + 407 + ], + "spans": [ + { + "bbox": [ + 95, + 395, + 107, + 406 + ], + "score": 1.0, + "content": "16", + "type": "text" + }, + { + "bbox": [ + 150, + 393, + 221, + 407 + ], + "score": 1.0, + "content": "ϵ˜ = ϵ = ϵθ(˜xτ˜s )", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 95, + 405, + 190, + 417 + ], + "spans": [ + { + "bbox": [ + 95, + 406, + 107, + 417 + ], + "score": 1.0, + "content": "17", + "type": "text" + }, + { + "bbox": [ + 148, + 405, + 190, + 417 + ], + "score": 1.0, + "content": "στ˜ = 0", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 136, + 417, + 308, + 433 + ], + "spans": [ + { + "bbox": [ + 136, + 417, + 308, + 433 + ], + "score": 1.0, + "content": "// phase 3: quality boosting", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 95, + 429, + 158, + 443 + ], + "spans": [ + { + "bbox": [ + 95, + 432, + 107, + 443 + ], + "score": 1.0, + "content": "18", + "type": "text" + }, + { + "bbox": [ + 135, + 429, + 158, + 443 + ], + "score": 1.0, + "content": "else", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 95, + 441, + 402, + 505 + ], + "spans": [ + { + "bbox": [ + 95, + 443, + 107, + 455 + ], + "score": 1.0, + "content": "19", + "type": "text" + }, + { + "bbox": [ + 133, + 441, + 402, + 505 + ], + "score": 0.9, + "content": "\\begin{array} { r l } & { \\quad \\Big \\lfloor \\tilde { \\epsilon } = \\epsilon = \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } ) } \\\\ & { \\quad \\Big \\lfloor \\sigma _ { \\tilde { \\tau } _ { s } } = \\sqrt { \\left( 1 - \\alpha _ { \\tilde { \\tau } _ { s } - 1 } \\right) / \\left( 1 - \\alpha _ { \\tilde { \\tau } _ { s } } \\right) } \\sqrt { 1 - \\alpha _ { \\tilde { \\tau } _ { s } } / \\alpha _ { \\tilde { \\tau } _ { s } - 1 } } } \\\\ & { \\quad z \\sim N ( 0 , 1 ) } \\\\ & { \\quad \\tilde { x } _ { \\tilde { \\tau } _ { s - 1 } } = \\sqrt { \\alpha _ { \\tilde { \\tau } _ { s - 1 } } } \\big ( \\frac { \\tilde { x } _ { \\tilde { \\tau } _ { s } } - \\sqrt { 1 - \\alpha _ { \\tilde { \\tau } _ { s } } } \\tilde { \\epsilon } } { \\sqrt { \\alpha _ { \\tilde { \\tau } _ { s } } } } \\big ) + \\sqrt { 1 - \\alpha _ { \\tilde { \\tau } _ { s - 1 } } - \\sigma _ { \\tilde { \\tau } _ { s } } ^ { 2 } } \\epsilon + \\sigma _ { \\tilde { \\tau } _ { s } } z } \\end{array}", + "type": "inline_equation", + "image_path": "4b60cb648c2f2bfda28ffaa77201d9eefa496bc6e449393cb5a00d4074a5187e.jpg" + } + ], + "index": 30 + }, + { + "bbox": [ + 95, + 456, + 107, + 468 + ], + "spans": [ + { + "bbox": [ + 95, + 456, + 107, + 468 + ], + "score": 1.0, + "content": "20", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 95, + 471, + 106, + 482 + ], + "spans": [ + { + "bbox": [ + 95, + 471, + 106, + 482 + ], + "score": 1.0, + "content": "21", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 95, + 488, + 107, + 500 + ], + "spans": [ + { + "bbox": [ + 95, + 488, + 107, + 500 + ], + "score": 1.0, + "content": "22", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 15.5 + }, + { + "type": "text", + "bbox": [ + 108, + 506, + 249, + 518 + ], + "lines": [ + { + "bbox": [ + 110, + 504, + 250, + 521 + ], + "spans": [ + { + "bbox": [ + 110, + 504, + 151, + 521 + ], + "score": 1.0, + "content": "return", + "type": "text" + }, + { + "bbox": [ + 151, + 506, + 163, + 517 + ], + "score": 0.64, + "content": "\\tilde { x } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 163, + 504, + 250, + 521 + ], + "score": 1.0, + "content": "(manipulated image)", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 108, + 548, + 504, + 582 + ], + "lines": [ + { + "bbox": [ + 105, + 547, + 504, + 561 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 504, + 561 + ], + "score": 1.0, + "content": "As for LSUN-church, we provide a set of four images at once and add a question: 3) Diversity:", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 558, + 506, + 572 + ], + "spans": [ + { + "bbox": [ + 105, + 558, + 506, + 572 + ], + "score": 1.0, + "content": "Which group do you think has a more diverse style? 4) Overall: Which image do you think is better", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 570, + 275, + 582 + ], + "spans": [ + { + "bbox": [ + 106, + 570, + 275, + 582 + ], + "score": 1.0, + "content": "considering the above evaluation criteria?", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33 + }, + { + "type": "title", + "bbox": [ + 108, + 595, + 426, + 607 + ], + "lines": [ + { + "bbox": [ + 105, + 595, + 428, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 595, + 428, + 609 + ], + "score": 1.0, + "content": "K.2 SEGMENTATION CONSISTENCY AND DIRECTIONAL CLIP SIMILARITY", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 35 + }, + { + "type": "text", + "bbox": [ + 106, + 615, + 505, + 693 + ], + "lines": [ + { + "bbox": [ + 106, + 615, + 505, + 629 + ], + "spans": [ + { + "bbox": [ + 106, + 615, + 405, + 629 + ], + "score": 1.0, + "content": "We compare Asyrp and DiffusionCLIP using directional CLIP similarity", + "type": "text" + }, + { + "bbox": [ + 405, + 616, + 427, + 627 + ], + "score": 0.81, + "content": "( S _ { \\mathrm { d i r } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 427, + 615, + 505, + 629 + ], + "score": 1.0, + "content": "and segmentation-", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 627, + 505, + 639 + ], + "spans": [ + { + "bbox": [ + 106, + 627, + 505, + 639 + ], + "score": 1.0, + "content": "consistency (SC) following the protocols in DiffusionCLIP in Table 4 and Table 5. A pretrained", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 638, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 106, + 638, + 505, + 650 + ], + "score": 1.0, + "content": "CLIP (Radford et al., 2021) and segmentation models (Yu et al. (2018); Zhou et al. (2019; 2017); Lee", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 648, + 506, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 243, + 662 + ], + "score": 1.0, + "content": "et al. (2020)) are used to compute", + "type": "text" + }, + { + "bbox": [ + 243, + 649, + 261, + 660 + ], + "score": 0.91, + "content": "S _ { \\mathrm { d i r } }", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 648, + 506, + 662 + ], + "score": 1.0, + "content": "and SC, respectively. We choose three attributes (smiling,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "score": 1.0, + "content": "sad, tanned) for CelebA-HQ-in-domain, two attributes (Pixar,Neaderthal) for CelebA-HQ-", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 671, + 506, + 683 + ], + "spans": [ + { + "bbox": [ + 105, + 671, + 506, + 683 + ], + "score": 1.0, + "content": "unseen-domain and three attributes (department store, ancient, red brick) for LSUN-", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 681, + 138, + 693 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 138, + 693 + ], + "score": 1.0, + "content": "church.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 39 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "For a fair comparison, we use official checkpoints of DiffusionCLIP and provide scores of the at-", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 710, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 505, + 722 + ], + "score": 1.0, + "content": "tributes (tanned, red brick) following Kim & Ye (2021). Regarding attributes without the", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 720, + 506, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 734 + ], + "score": 1.0, + "content": "official checkpoints (smiling,sad), we train DiffusionCLIP by ourselves with the official code.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44 + } + ], + "page_idx": 24, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 85, + 240, + 97 + ], + "lines": [ + { + "bbox": [ + 106, + 83, + 240, + 101 + ], + "spans": [ + { + "bbox": [ + 106, + 83, + 240, + 101 + ], + "score": 1.0, + "content": "Algorithm 1: Editing(Inference)", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 96, + 98, + 497, + 509 + ], + "lines": [ + { + "bbox": [ + 105, + 97, + 482, + 114 + ], + "spans": [ + { + "bbox": [ + 105, + 97, + 136, + 114 + ], + "score": 1.0, + "content": "Input:", + "type": "text" + }, + { + "bbox": [ + 136, + 102, + 147, + 112 + ], + "score": 0.73, + "content": "x _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 148, + 97, + 206, + 114 + ], + "score": 1.0, + "content": "(Input image),", + "type": "text" + }, + { + "bbox": [ + 206, + 100, + 239, + 113 + ], + "score": 0.35, + "content": "\\{ f _ { t } ^ { i } \\} _ { i = 1 } ^ { M }", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 97, + 426, + 114 + ], + "score": 1.0, + "content": "(M Neural implicit functions for M attributes),", + "type": "text" + }, + { + "bbox": [ + 426, + 100, + 458, + 113 + ], + "score": 0.83, + "content": "\\{ c _ { i } \\} _ { i = 1 } ^ { M }", + "type": "inline_equation" + }, + { + "bbox": [ + 459, + 97, + 482, + 114 + ], + "score": 1.0, + "content": "(user", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 135, + 110, + 491, + 124 + ], + "spans": [ + { + "bbox": [ + 135, + 110, + 328, + 124 + ], + "score": 1.0, + "content": "defined M scaling coefficients for M attributes),", + "type": "text" + }, + { + "bbox": [ + 329, + 113, + 338, + 123 + ], + "score": 0.8, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 339, + 110, + 447, + 124 + ], + "score": 1.0, + "content": "(frozen pretrained model),", + "type": "text" + }, + { + "bbox": [ + 447, + 113, + 467, + 124 + ], + "score": 0.89, + "content": "S _ { f o r }", + "type": "inline_equation" + }, + { + "bbox": [ + 468, + 110, + 491, + 124 + ], + "score": 1.0, + "content": "(# of", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 135, + 120, + 453, + 137 + ], + "spans": [ + { + "bbox": [ + 135, + 120, + 203, + 137 + ], + "score": 1.0, + "content": "inversion steps),", + "type": "text" + }, + { + "bbox": [ + 204, + 123, + 223, + 134 + ], + "score": 0.88, + "content": "S _ { g e n }", + "type": "inline_equation" + }, + { + "bbox": [ + 224, + 120, + 312, + 137 + ], + "score": 1.0, + "content": "(# of inference steps),", + "type": "text" + }, + { + "bbox": [ + 312, + 123, + 331, + 134 + ], + "score": 0.88, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 120, + 399, + 137 + ], + "score": 1.0, + "content": "(computed from", + "type": "text" + }, + { + "bbox": [ + 399, + 123, + 420, + 134 + ], + "score": 0.72, + "content": "\\ S 4 . 1", + "type": "inline_equation" + }, + { + "bbox": [ + 420, + 120, + 453, + 137 + ], + "score": 1.0, + "content": "), tboost", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 136, + 133, + 228, + 145 + ], + "spans": [ + { + "bbox": [ + 136, + 133, + 203, + 145 + ], + "score": 1.0, + "content": "(computed from", + "type": "text" + }, + { + "bbox": [ + 203, + 134, + 224, + 145 + ], + "score": 0.56, + "content": "\\ S 4 . 2", + "type": "inline_equation" + }, + { + "bbox": [ + 224, + 133, + 228, + 145 + ], + "score": 1.0, + "content": ")", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 96, + 147, + 317, + 163 + ], + "spans": [ + { + "bbox": [ + 96, + 147, + 185, + 163 + ], + "score": 1.0, + "content": "1 Function Editor", + "type": "text" + }, + { + "bbox": [ + 185, + 147, + 312, + 162 + ], + "score": 0.49, + "content": "( x _ { 0 } , \\epsilon _ { \\theta } , \\{ c _ { i } \\} _ { i = 1 } ^ { M } , S _ { f o r } , S _ { g e n } , * )", + "type": "inline_equation" + }, + { + "bbox": [ + 313, + 147, + 317, + 163 + ], + "score": 1.0, + "content": ":", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 122, + 163, + 292, + 177 + ], + "spans": [ + { + "bbox": [ + 122, + 163, + 292, + 177 + ], + "score": 1.0, + "content": "// step 1: Semantic encoding", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 99, + 171, + 276, + 198 + ], + "spans": [ + { + "bbox": [ + 99, + 182, + 104, + 189 + ], + "score": 1.0, + "content": "2", + "type": "text" + }, + { + "bbox": [ + 117, + 171, + 151, + 198 + ], + "score": 1.0, + "content": "Define", + "type": "text" + }, + { + "bbox": [ + 151, + 176, + 187, + 191 + ], + "score": 0.89, + "content": "\\{ \\tau _ { s } \\} _ { s = 1 } ^ { S _ { f o r } }", + "type": "inline_equation" + }, + { + "bbox": [ + 187, + 171, + 200, + 198 + ], + "score": 1.0, + "content": "s.t", + "type": "text" + }, + { + "bbox": [ + 200, + 178, + 229, + 190 + ], + "score": 0.75, + "content": "\\tau _ { 1 } = 0", + "type": "inline_equation" + }, + { + "bbox": [ + 229, + 171, + 232, + 198 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 233, + 178, + 276, + 192 + ], + "score": 0.76, + "content": "\\tau _ { S _ { f o r } } = T", + "type": "inline_equation" + } + ], + "index": 7 + }, + { + "bbox": [ + 98, + 189, + 239, + 205 + ], + "spans": [ + { + "bbox": [ + 98, + 193, + 105, + 201 + ], + "score": 1.0, + "content": "3", + "type": "text" + }, + { + "bbox": [ + 120, + 189, + 137, + 205 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 137, + 191, + 224, + 203 + ], + "score": 0.8, + "content": "s = 1 , 2 , . . . , S _ { f o r } - 1", + "type": "inline_equation" + }, + { + "bbox": [ + 224, + 189, + 239, + 205 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 99, + 203, + 201, + 214 + ], + "spans": [ + { + "bbox": [ + 99, + 205, + 105, + 212 + ], + "score": 1.0, + "content": "4", + "type": "text" + }, + { + "bbox": [ + 136, + 203, + 201, + 214 + ], + "score": 0.59, + "content": "\\epsilon \\epsilon _ { \\theta } ( x _ { \\tau _ { s } } , \\tau _ { s } )", + "type": "inline_equation" + } + ], + "index": 9 + }, + { + "bbox": [ + 98, + 214, + 267, + 228 + ], + "spans": [ + { + "bbox": [ + 98, + 217, + 105, + 225 + ], + "score": 1.0, + "content": "5", + "type": "text" + }, + { + "bbox": [ + 127, + 214, + 134, + 227 + ], + "score": 0.0, + "content": "", + "type": "text" + }, + { + "bbox": [ + 134, + 214, + 267, + 228 + ], + "score": 0.51, + "content": "x _ { \\tau _ { s + 1 } } = \\sqrt { \\alpha _ { \\tau _ { s } } } x _ { \\tau _ { s } } + \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon", + "type": "inline_equation" + } + ], + "index": 10 + }, + { + "bbox": [ + 98, + 232, + 425, + 272 + ], + "spans": [ + { + "bbox": [ + 98, + 251, + 105, + 259 + ], + "score": 1.0, + "content": "6", + "type": "text" + }, + { + "bbox": [ + 151, + 245, + 189, + 261 + ], + "score": 0.72, + "content": "\\{ \\tilde { \\tau } _ { s } \\} _ { s = 1 } ^ { S _ { g e n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 189, + 232, + 201, + 272 + ], + "score": 1.0, + "content": "Mas.t", + "type": "text" + }, + { + "bbox": [ + 202, + 248, + 231, + 259 + ], + "score": 0.71, + "content": "\\tilde { \\tau } _ { 1 } = 0", + "type": "inline_equation" + }, + { + "bbox": [ + 232, + 232, + 235, + 272 + ], + "score": 1.0, + "content": "a,", + "type": "text" + }, + { + "bbox": [ + 235, + 248, + 291, + 260 + ], + "score": 0.8, + "content": "\\tilde { \\tau } _ { S _ { e d i t } } = t _ { e d i t }", + "type": "inline_equation" + }, + { + "bbox": [ + 291, + 232, + 295, + 272 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 295, + 249, + 361, + 261 + ], + "score": 0.81, + "content": "\\tilde { \\tau } _ { S _ { n o i s e } } = t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 232, + 379, + 272 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 380, + 248, + 425, + 261 + ], + "score": 0.91, + "content": "\\tilde { \\tau } _ { S _ { g e n } } = T", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 98, + 261, + 188, + 274 + ], + "spans": [ + { + "bbox": [ + 98, + 263, + 105, + 271 + ], + "score": 1.0, + "content": "7", + "type": "text" + }, + { + "bbox": [ + 122, + 261, + 188, + 274 + ], + "score": 0.75, + "content": "\\tilde { x } _ { \\tilde { \\tau } _ { S _ { g e n } } } = x _ { \\tilde { \\tau } _ { S _ { f o r } } }", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 98, + 269, + 256, + 291 + ], + "spans": [ + { + "bbox": [ + 98, + 277, + 105, + 285 + ], + "score": 1.0, + "content": "8", + "type": "text" + }, + { + "bbox": [ + 117, + 269, + 137, + 291 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 137, + 275, + 239, + 286 + ], + "score": 0.67, + "content": "s = S _ { g e n } , S _ { g e n } - 1 , . . . , 2", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 269, + 256, + 291 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 136, + 283, + 254, + 298 + ], + "spans": [ + { + "bbox": [ + 136, + 283, + 254, + 298 + ], + "score": 1.0, + "content": "// phase 1: editing", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 98, + 295, + 209, + 308 + ], + "spans": [ + { + "bbox": [ + 98, + 298, + 106, + 307 + ], + "score": 1.0, + "content": "9", + "type": "text" + }, + { + "bbox": [ + 136, + 295, + 146, + 308 + ], + "score": 1.0, + "content": "if", + "type": "text" + }, + { + "bbox": [ + 146, + 297, + 185, + 307 + ], + "score": 0.84, + "content": "s \\geq S _ { e d i t }", + "type": "inline_equation" + }, + { + "bbox": [ + 186, + 295, + 209, + 308 + ], + "score": 1.0, + "content": "then", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 96, + 305, + 304, + 320 + ], + "spans": [ + { + "bbox": [ + 96, + 309, + 106, + 318 + ], + "score": 1.0, + "content": "10", + "type": "text" + }, + { + "bbox": [ + 150, + 305, + 233, + 320 + ], + "score": 1.0, + "content": "Extract feature map", + "type": "text" + }, + { + "bbox": [ + 234, + 307, + 249, + 318 + ], + "score": 0.88, + "content": "h _ { \\tilde { \\tau } _ { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 305, + 272, + 320 + ], + "score": 1.0, + "content": "from", + "type": "text" + }, + { + "bbox": [ + 273, + 307, + 304, + 319 + ], + "score": 0.89, + "content": "\\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } )", + "type": "inline_equation" + } + ], + "index": 16 + }, + { + "bbox": [ + 95, + 318, + 288, + 370 + ], + "spans": [ + { + "bbox": [ + 95, + 323, + 107, + 334 + ], + "score": 1.0, + "content": "11", + "type": "text" + }, + { + "bbox": [ + 149, + 318, + 288, + 370 + ], + "score": 0.32, + "content": "\\begin{array} { r l } & { \\Delta h _ { \\tilde { \\tau } _ { s } } = \\frac { S _ { f o r } } { S _ { g e n } } ( \\sum _ { i = 1 } ^ { M } c _ { i } f _ { \\tilde { \\tau } _ { s } } ^ { i } ( h _ { \\tilde { \\tau } _ { s } } ) ) } \\\\ & { \\tilde { \\epsilon } = \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } | \\Delta h _ { \\tilde { \\tau } _ { s } } ) } \\\\ & { \\epsilon = \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } ) } \\\\ & { \\sigma _ { \\tilde { \\tau } _ { s } } = 0 } \\end{array}", + "type": "inline_equation", + "image_path": "04afa24a0133c9be96d6d2df48a1abeb0876ad72b210390ee102ec52a597df14.jpg" + } + ], + "index": 18 + }, + { + "bbox": [ + 96, + 337, + 106, + 347 + ], + "spans": [ + { + "bbox": [ + 96, + 337, + 106, + 347 + ], + "score": 1.0, + "content": "12", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 95, + 347, + 107, + 358 + ], + "spans": [ + { + "bbox": [ + 95, + 347, + 107, + 358 + ], + "score": 1.0, + "content": "13", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 95, + 359, + 106, + 369 + ], + "spans": [ + { + "bbox": [ + 95, + 359, + 106, + 369 + ], + "score": 1.0, + "content": "14", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 137, + 370, + 266, + 385 + ], + "spans": [ + { + "bbox": [ + 137, + 370, + 266, + 385 + ], + "score": 1.0, + "content": "// phase 2: denoising", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 95, + 381, + 232, + 396 + ], + "spans": [ + { + "bbox": [ + 95, + 384, + 107, + 395 + ], + "score": 1.0, + "content": "15", + "type": "text" + }, + { + "bbox": [ + 136, + 381, + 164, + 396 + ], + "score": 1.0, + "content": "else if", + "type": "text" + }, + { + "bbox": [ + 164, + 383, + 209, + 394 + ], + "score": 0.84, + "content": "s \\geq S _ { n o i s e }", + "type": "inline_equation" + }, + { + "bbox": [ + 209, + 381, + 232, + 396 + ], + "score": 1.0, + "content": "then", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 95, + 393, + 221, + 407 + ], + "spans": [ + { + "bbox": [ + 95, + 395, + 107, + 406 + ], + "score": 1.0, + "content": "16", + "type": "text" + }, + { + "bbox": [ + 150, + 393, + 221, + 407 + ], + "score": 1.0, + "content": "ϵ˜ = ϵ = ϵθ(˜xτ˜s )", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 95, + 405, + 190, + 417 + ], + "spans": [ + { + "bbox": [ + 95, + 406, + 107, + 417 + ], + "score": 1.0, + "content": "17", + "type": "text" + }, + { + "bbox": [ + 148, + 405, + 190, + 417 + ], + "score": 1.0, + "content": "στ˜ = 0", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 136, + 417, + 308, + 433 + ], + "spans": [ + { + "bbox": [ + 136, + 417, + 308, + 433 + ], + "score": 1.0, + "content": "// phase 3: quality boosting", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 95, + 429, + 158, + 443 + ], + "spans": [ + { + "bbox": [ + 95, + 432, + 107, + 443 + ], + "score": 1.0, + "content": "18", + "type": "text" + }, + { + "bbox": [ + 135, + 429, + 158, + 443 + ], + "score": 1.0, + "content": "else", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 95, + 441, + 402, + 505 + ], + "spans": [ + { + "bbox": [ + 95, + 443, + 107, + 455 + ], + "score": 1.0, + "content": "19", + "type": "text" + }, + { + "bbox": [ + 133, + 441, + 402, + 505 + ], + "score": 0.9, + "content": "\\begin{array} { r l } & { \\quad \\Big \\lfloor \\tilde { \\epsilon } = \\epsilon = \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tilde { \\tau } _ { s } } ) } \\\\ & { \\quad \\Big \\lfloor \\sigma _ { \\tilde { \\tau } _ { s } } = \\sqrt { \\left( 1 - \\alpha _ { \\tilde { \\tau } _ { s } - 1 } \\right) / \\left( 1 - \\alpha _ { \\tilde { \\tau } _ { s } } \\right) } \\sqrt { 1 - \\alpha _ { \\tilde { \\tau } _ { s } } / \\alpha _ { \\tilde { \\tau } _ { s } - 1 } } } \\\\ & { \\quad z \\sim N ( 0 , 1 ) } \\\\ & { \\quad \\tilde { x } _ { \\tilde { \\tau } _ { s - 1 } } = \\sqrt { \\alpha _ { \\tilde { \\tau } _ { s - 1 } } } \\big ( \\frac { \\tilde { x } _ { \\tilde { \\tau } _ { s } } - \\sqrt { 1 - \\alpha _ { \\tilde { \\tau } _ { s } } } \\tilde { \\epsilon } } { \\sqrt { \\alpha _ { \\tilde { \\tau } _ { s } } } } \\big ) + \\sqrt { 1 - \\alpha _ { \\tilde { \\tau } _ { s - 1 } } - \\sigma _ { \\tilde { \\tau } _ { s } } ^ { 2 } } \\epsilon + \\sigma _ { \\tilde { \\tau } _ { s } } z } \\end{array}", + "type": "inline_equation", + "image_path": "4b60cb648c2f2bfda28ffaa77201d9eefa496bc6e449393cb5a00d4074a5187e.jpg" + } + ], + "index": 30 + }, + { + "bbox": [ + 95, + 456, + 107, + 468 + ], + "spans": [ + { + "bbox": [ + 95, + 456, + 107, + 468 + ], + "score": 1.0, + "content": "20", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 95, + 471, + 106, + 482 + ], + "spans": [ + { + "bbox": [ + 95, + 471, + 106, + 482 + ], + "score": 1.0, + "content": "21", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 95, + 488, + 107, + 500 + ], + "spans": [ + { + "bbox": [ + 95, + 488, + 107, + 500 + ], + "score": 1.0, + "content": "22", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 15.5, + "bbox_fs": [ + 95, + 97, + 491, + 505 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 506, + 249, + 518 + ], + "lines": [ + { + "bbox": [ + 110, + 504, + 250, + 521 + ], + "spans": [ + { + "bbox": [ + 110, + 504, + 151, + 521 + ], + "score": 1.0, + "content": "return", + "type": "text" + }, + { + "bbox": [ + 151, + 506, + 163, + 517 + ], + "score": 0.64, + "content": "\\tilde { x } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 163, + 504, + 250, + 521 + ], + "score": 1.0, + "content": "(manipulated image)", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31, + "bbox_fs": [ + 110, + 504, + 250, + 521 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 548, + 504, + 582 + ], + "lines": [ + { + "bbox": [ + 105, + 547, + 504, + 561 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 504, + 561 + ], + "score": 1.0, + "content": "As for LSUN-church, we provide a set of four images at once and add a question: 3) Diversity:", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 558, + 506, + 572 + ], + "spans": [ + { + "bbox": [ + 105, + 558, + 506, + 572 + ], + "score": 1.0, + "content": "Which group do you think has a more diverse style? 4) Overall: Which image do you think is better", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 570, + 275, + 582 + ], + "spans": [ + { + "bbox": [ + 106, + 570, + 275, + 582 + ], + "score": 1.0, + "content": "considering the above evaluation criteria?", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33, + "bbox_fs": [ + 105, + 547, + 506, + 582 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 595, + 426, + 607 + ], + "lines": [ + { + "bbox": [ + 105, + 595, + 428, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 595, + 428, + 609 + ], + "score": 1.0, + "content": "K.2 SEGMENTATION CONSISTENCY AND DIRECTIONAL CLIP SIMILARITY", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 35 + }, + { + "type": "text", + "bbox": [ + 106, + 615, + 505, + 693 + ], + "lines": [ + { + "bbox": [ + 106, + 615, + 505, + 629 + ], + "spans": [ + { + "bbox": [ + 106, + 615, + 405, + 629 + ], + "score": 1.0, + "content": "We compare Asyrp and DiffusionCLIP using directional CLIP similarity", + "type": "text" + }, + { + "bbox": [ + 405, + 616, + 427, + 627 + ], + "score": 0.81, + "content": "( S _ { \\mathrm { d i r } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 427, + 615, + 505, + 629 + ], + "score": 1.0, + "content": "and segmentation-", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 627, + 505, + 639 + ], + "spans": [ + { + "bbox": [ + 106, + 627, + 505, + 639 + ], + "score": 1.0, + "content": "consistency (SC) following the protocols in DiffusionCLIP in Table 4 and Table 5. A pretrained", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 638, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 106, + 638, + 505, + 650 + ], + "score": 1.0, + "content": "CLIP (Radford et al., 2021) and segmentation models (Yu et al. (2018); Zhou et al. (2019; 2017); Lee", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 648, + 506, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 243, + 662 + ], + "score": 1.0, + "content": "et al. (2020)) are used to compute", + "type": "text" + }, + { + "bbox": [ + 243, + 649, + 261, + 660 + ], + "score": 0.91, + "content": "S _ { \\mathrm { d i r } }", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 648, + 506, + 662 + ], + "score": 1.0, + "content": "and SC, respectively. We choose three attributes (smiling,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "score": 1.0, + "content": "sad, tanned) for CelebA-HQ-in-domain, two attributes (Pixar,Neaderthal) for CelebA-HQ-", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 671, + 506, + 683 + ], + "spans": [ + { + "bbox": [ + 105, + 671, + 506, + 683 + ], + "score": 1.0, + "content": "unseen-domain and three attributes (department store, ancient, red brick) for LSUN-", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 681, + 138, + 693 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 138, + 693 + ], + "score": 1.0, + "content": "church.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 39, + "bbox_fs": [ + 105, + 615, + 506, + 693 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "For a fair comparison, we use official checkpoints of DiffusionCLIP and provide scores of the at-", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 710, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 505, + 722 + ], + "score": 1.0, + "content": "tributes (tanned, red brick) following Kim & Ye (2021). Regarding attributes without the", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 720, + 506, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 734 + ], + "score": 1.0, + "content": "official checkpoints (smiling,sad), we train DiffusionCLIP by ourselves with the official code.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44, + "bbox_fs": [ + 105, + 698, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 106, + 85, + 309, + 97 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 309, + 100 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 299, + 100 + ], + "score": 1.0, + "content": "Algorithm 2: Training Neural implicit function", + "type": "text" + }, + { + "bbox": [ + 299, + 86, + 309, + 97 + ], + "score": 0.81, + "content": "f _ { t }", + "type": "inline_equation" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 109, + 101, + 488, + 137 + ], + "lines": [ + { + "bbox": [ + 102, + 93, + 489, + 126 + ], + "spans": [ + { + "bbox": [ + 102, + 93, + 136, + 126 + ], + "score": 1.0, + "content": "Input:", + "type": "text" + }, + { + "bbox": [ + 136, + 104, + 145, + 114 + ], + "score": 0.77, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 93, + 189, + 126 + ], + "score": 1.0, + "content": "(pretrained attribute),", + "type": "text" + }, + { + "bbox": [ + 208, + 93, + 224, + 126 + ], + "score": 1.0, + "content": "el), (sou", + "type": "text" + }, + { + "bbox": [ + 225, + 100, + 264, + 115 + ], + "score": 0.87, + "content": "\\{ x _ { 0 } ^ { ( i ) } \\} _ { i = 1 } ^ { N }", + "type": "inline_equation" + }, + { + "bbox": [ + 279, + 93, + 329, + 126 + ], + "score": 1.0, + "content": "ages to preco(target text),", + "type": "text" + }, + { + "bbox": [ + 349, + 93, + 362, + 126 + ], + "score": 1.0, + "content": "te), (#", + "type": "text" + }, + { + "bbox": [ + 362, + 103, + 372, + 114 + ], + "score": 0.83, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 372, + 93, + 437, + 126 + ], + "score": 1.0, + "content": "(Neural implicitinversion steps),", + "type": "text" + }, + { + "bbox": [ + 447, + 93, + 489, + 126 + ], + "score": 1.0, + "content": "nction of(# of", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 189, + 114, + 447, + 126 + ], + "spans": [ + { + "bbox": [ + 189, + 114, + 207, + 125 + ], + "score": 0.85, + "content": "y _ { s r c }", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 115, + 278, + 125 + ], + "score": 0.87, + "content": "y _ { t a r }", + "type": "inline_equation" + }, + { + "bbox": [ + 329, + 114, + 348, + 126 + ], + "score": 0.89, + "content": "S _ { f o r }", + "type": "inline_equation" + }, + { + "bbox": [ + 437, + 114, + 447, + 124 + ], + "score": 0.26, + "content": "K", + "type": "inline_equation" + } + ], + "index": 2 + }, + { + "bbox": [ + 136, + 125, + 438, + 137 + ], + "spans": [ + { + "bbox": [ + 136, + 125, + 205, + 137 + ], + "score": 1.0, + "content": "training epochs),", + "type": "text" + }, + { + "bbox": [ + 205, + 126, + 224, + 136 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 125, + 292, + 137 + ], + "score": 1.0, + "content": "(computed from", + "type": "text" + }, + { + "bbox": [ + 293, + 125, + 315, + 136 + ], + "score": 0.8, + "content": "\\ S 4 . 1 \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 125, + 320, + 137 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 320, + 126, + 344, + 136 + ], + "score": 0.81, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 125, + 412, + 137 + ], + "score": 1.0, + "content": "(computed from", + "type": "text" + }, + { + "bbox": [ + 413, + 125, + 435, + 136 + ], + "score": 0.61, + "content": "\\ S 4 . 2 )", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 125, + 438, + 137 + ], + "score": 1.0, + "content": ")", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2 + }, + { + "type": "text", + "bbox": [ + 98, + 140, + 282, + 239 + ], + "lines": [ + { + "bbox": [ + 107, + 140, + 282, + 152 + ], + "spans": [ + { + "bbox": [ + 107, + 140, + 282, + 152 + ], + "score": 1.0, + "content": "// step 1: Precompute latents", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 93, + 146, + 258, + 172 + ], + "spans": [ + { + "bbox": [ + 93, + 146, + 136, + 172 + ], + "score": 1.0, + "content": "1 Define", + "type": "text" + }, + { + "bbox": [ + 136, + 152, + 172, + 167 + ], + "score": 0.61, + "content": "\\{ \\tau _ { s } \\} _ { s = 1 } ^ { S _ { f o r } }", + "type": "inline_equation" + }, + { + "bbox": [ + 172, + 146, + 185, + 172 + ], + "score": 1.0, + "content": "s.t", + "type": "text" + }, + { + "bbox": [ + 185, + 154, + 258, + 167 + ], + "score": 0.71, + "content": "\\tau _ { 1 } = 0 , \\tau _ { S _ { f o r } } = T", + "type": "inline_equation" + } + ], + "index": 5 + }, + { + "bbox": [ + 97, + 166, + 196, + 178 + ], + "spans": [ + { + "bbox": [ + 97, + 166, + 121, + 178 + ], + "score": 1.0, + "content": "2 for", + "type": "text" + }, + { + "bbox": [ + 122, + 167, + 180, + 178 + ], + "score": 0.82, + "content": "i = 1 , 2 , . . . , N", + "type": "inline_equation" + }, + { + "bbox": [ + 181, + 166, + 196, + 178 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 97, + 176, + 239, + 190 + ], + "spans": [ + { + "bbox": [ + 97, + 179, + 105, + 189 + ], + "score": 1.0, + "content": "3", + "type": "text" + }, + { + "bbox": [ + 119, + 176, + 137, + 190 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 137, + 178, + 225, + 189 + ], + "score": 0.78, + "content": "s = 1 , 2 , . . . , S _ { f o r } - 1", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 176, + 239, + 190 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 98, + 189, + 203, + 203 + ], + "spans": [ + { + "bbox": [ + 98, + 193, + 106, + 203 + ], + "score": 1.0, + "content": "4", + "type": "text" + }, + { + "bbox": [ + 136, + 189, + 203, + 203 + ], + "score": 0.54, + "content": "\\epsilon \\epsilon _ { \\theta } ( x _ { \\tau _ { s } } ^ { ( i ) } , \\bar { \\tau _ { s } } )", + "type": "inline_equation" + } + ], + "index": 8 + }, + { + "bbox": [ + 98, + 203, + 269, + 220 + ], + "spans": [ + { + "bbox": [ + 98, + 208, + 106, + 218 + ], + "score": 1.0, + "content": "5", + "type": "text" + }, + { + "bbox": [ + 136, + 203, + 269, + 220 + ], + "score": 0.55, + "content": "\\begin{array} { r } { x _ { \\tau _ { s + 1 } } ^ { ( i ) } = \\sqrt { \\alpha _ { \\tau _ { s } } } x _ { \\tau _ { s } } ^ { ( i ) } + \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon } \\end{array}", + "type": "inline_equation" + } + ], + "index": 9 + }, + { + "bbox": [ + 99, + 220, + 213, + 240 + ], + "spans": [ + { + "bbox": [ + 99, + 227, + 104, + 234 + ], + "score": 1.0, + "content": "6", + "type": "text" + }, + { + "bbox": [ + 119, + 220, + 213, + 240 + ], + "score": 1.0, + "content": "Save the latent x(i)τSf or", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 97, + 246, + 226, + 278 + ], + "lines": [ + { + "bbox": [ + 96, + 244, + 162, + 258 + ], + "spans": [ + { + "bbox": [ + 96, + 244, + 106, + 258 + ], + "score": 1.0, + "content": "7", + "type": "text" + }, + { + "bbox": [ + 106, + 246, + 162, + 257 + ], + "score": 0.83, + "content": "\\tau _ { S _ { e d i t } } = t _ { e d i t }", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 255, + 224, + 268 + ], + "spans": [ + { + "bbox": [ + 106, + 255, + 214, + 268 + ], + "score": 1.0, + "content": "// step 2: Update", + "type": "text" + }, + { + "bbox": [ + 214, + 256, + 224, + 267 + ], + "score": 0.51, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 97, + 266, + 212, + 279 + ], + "spans": [ + { + "bbox": [ + 97, + 266, + 134, + 279 + ], + "score": 1.0, + "content": "8 for ep", + "type": "text" + }, + { + "bbox": [ + 135, + 267, + 197, + 278 + ], + "score": 0.64, + "content": "{ \\varkappa } h { = 1 , 2 , . . . , K }", + "type": "inline_equation" + }, + { + "bbox": [ + 198, + 266, + 212, + 279 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12 + }, + { + "type": "interline_equation", + "bbox": [ + 151, + 341, + 393, + 363 + ], + "lines": [ + { + "bbox": [ + 151, + 341, + 393, + 363 + ], + "spans": [ + { + "bbox": [ + 151, + 341, + 393, + 363 + ], + "score": 0.56, + "content": "\\begin{array} { r } { P = \\frac { \\tilde { x } _ { \\tau _ { s } } ^ { ( i ) } - \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon _ { \\theta } \\left( \\tilde { x } _ { \\tau _ { s } } ^ { ( i ) } | \\Delta h _ { \\tau _ { s } } \\right) } { \\sqrt { \\alpha _ { \\tau _ { s } } } } ; P _ { s r c } = \\frac { x _ { \\tau _ { s } } ^ { ( i ) } - \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon _ { \\theta } \\left( x _ { \\tau _ { s } } ^ { ( i ) } \\right) } { \\sqrt { \\alpha _ { \\tau _ { s } } } } } \\end{array}", + "type": "interline_equation", + "image_path": "0ba6b4a72d84a3eab40bb6fb5efa2526ffdd9ef453ce460351eeb798a49902e7.jpg" + } + ] + } + ], + "index": 14, + "virtual_lines": [ + { + "bbox": [ + 151, + 341, + 393, + 363 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 151, + 366, + 332, + 399 + ], + "lines": [ + { + "bbox": [ + 151, + 366, + 332, + 399 + ], + "spans": [ + { + "bbox": [ + 151, + 366, + 332, + 399 + ], + "score": 0.25, + "content": "\\begin{array} { r l } & { \\tilde { x } _ { \\tau _ { s - 1 } } ^ { ( i ) } = \\sqrt { \\alpha _ { \\tau _ { s - 1 } } } P + \\sqrt { 1 - \\alpha _ { \\tau _ { s - 1 } } } \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tau _ { s } } ^ { ( i ) } ) } \\\\ & { x _ { \\tau _ { s - 1 } } ^ { ( i ) } = \\sqrt { \\alpha _ { \\tau _ { s - 1 } } } P _ { s r c } + \\sqrt { 1 - \\alpha _ { \\tau _ { s - 1 } } } \\epsilon _ { \\theta } ( x _ { \\tau _ { s } } ^ { ( i ) } ) } \\end{array}", + "type": "interline_equation", + "image_path": "bc6689a31c48e5761d9ac0c043ce0d1e0ee55fb1266965a19fd3791bbbf44a8e.jpg" + } + ] + } + ], + "index": 15.5, + "virtual_lines": [ + { + "bbox": [ + 151, + 366, + 332, + 382.5 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 151, + 382.5, + 332, + 399.0 + ], + "spans": [], + "index": 16 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 152, + 401, + 424, + 414 + ], + "lines": [ + { + "bbox": [ + 152, + 401, + 424, + 414 + ], + "spans": [ + { + "bbox": [ + 152, + 401, + 424, + 414 + ], + "score": 0.62, + "content": "E _ { t o t a l } \\lambda _ { C L I P } L _ { d i r e c t i o n } ( P , y _ { t a r } , P _ { s r c } , y _ { s r c } ) + \\lambda _ { r e c o n } | P - P _ { s r c } |", + "type": "interline_equation", + "image_path": "4af7d31768be7cc1ed82bdab507bf120173b9d97531e91f8e08dfa2efa05090c.jpg" + } + ] + } + ], + "index": 17, + "virtual_lines": [ + { + "bbox": [ + 152, + 401, + 424, + 414 + ], + "spans": [], + "index": 17 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 467, + 505, + 557 + ], + "lines": [ + { + "bbox": [ + 105, + 468, + 505, + 481 + ], + "spans": [ + { + "bbox": [ + 105, + 468, + 401, + 481 + ], + "score": 1.0, + "content": "We use 100 samples per attribute. Asyrp outperforms DiffusionCLIP on", + "type": "text" + }, + { + "bbox": [ + 401, + 468, + 419, + 479 + ], + "score": 0.92, + "content": "S _ { \\mathrm { d i r } }", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 468, + 505, + 481 + ], + "score": 1.0, + "content": "for all attributes. On", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 479, + 505, + 492 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 505, + 492 + ], + "score": 1.0, + "content": "SC, DiffusionCLIP achieves better or competitive scores. Because DiffusionCLIP manipulates im-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 491, + 505, + 503 + ], + "spans": [ + { + "bbox": [ + 106, + 491, + 505, + 503 + ], + "score": 1.0, + "content": "ages mostly by focusing on texture or color while preserving structure and shape, it takes advantage", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 502, + 505, + 513 + ], + "spans": [ + { + "bbox": [ + 106, + 502, + 505, + 513 + ], + "score": 1.0, + "content": "of getting higher SC scores. However, in Figure 28, results of smiling show that it is not proper to", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "score": 1.0, + "content": "edit attributes which require structural manipulation. Note that better SC scores do not guarantee bet-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 523, + 505, + 535 + ], + "spans": [ + { + "bbox": [ + 105, + 523, + 505, + 535 + ], + "score": 1.0, + "content": "ter qualitative performance. Results of Pixar also show the similar tendency of each method. We", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 534, + 506, + 546 + ], + "spans": [ + { + "bbox": [ + 105, + 534, + 506, + 546 + ], + "score": 1.0, + "content": "allow more structural changes than DiffusionCLIP while editing. Lower SC of our method comes", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 545, + 335, + 557 + ], + "spans": [ + { + "bbox": [ + 105, + 545, + 335, + 557 + ], + "score": 1.0, + "content": "from desirable structural changes as shown in Figure 28.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 21.5 + }, + { + "type": "interline_equation", + "bbox": [ + 212, + 577, + 379, + 604 + ], + "lines": [ + { + "bbox": [ + 212, + 577, + 379, + 604 + ], + "spans": [ + { + "bbox": [ + 212, + 577, + 379, + 604 + ], + "score": 0.94, + "content": "S _ { \\mathrm { d i r } } \\left( x _ { \\mathrm { g e n } } , y _ { \\mathrm { t a r } } ; x _ { \\mathrm { r e f } } , y _ { \\mathrm { r e f } } \\right) : = \\frac { \\left. \\Delta I , \\Delta T \\right. } { \\left\\| \\Delta I \\right\\| \\left\\| \\Delta T \\right\\| } ,", + "type": "interline_equation", + "image_path": "53ce9b39242e61d513e443f453e9d9b59f2f8d6c59eb55f50e784775b71557e9.jpg" + } + ] + } + ], + "index": 26.5, + "virtual_lines": [ + { + "bbox": [ + 212, + 577, + 379, + 590.5 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 212, + 590.5, + 379, + 604.0 + ], + "spans": [], + "index": 27 + } + ] + }, + { + "type": "title", + "bbox": [ + 107, + 624, + 192, + 637 + ], + "lines": [ + { + "bbox": [ + 105, + 623, + 193, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 623, + 193, + 640 + ], + "score": 1.0, + "content": "L DIRECTIONS", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "title", + "bbox": [ + 107, + 649, + 219, + 661 + ], + "lines": [ + { + "bbox": [ + 105, + 649, + 221, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 221, + 662 + ], + "score": 1.0, + "content": "L.1 GLOBAL DIRECTION", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 670, + 505, + 704 + ], + "lines": [ + { + "bbox": [ + 105, + 670, + 505, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 505, + 684 + ], + "score": 1.0, + "content": "Figure 29 and Figure 30 show that the effects of mean direction and global direction are quite", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 681, + 505, + 694 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 157, + 694 + ], + "score": 1.0, + "content": "similar with", + "type": "text" + }, + { + "bbox": [ + 157, + 682, + 176, + 693 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 176, + 681, + 190, + 694 + ], + "score": 1.0, + "content": "by", + "type": "text" + }, + { + "bbox": [ + 190, + 682, + 200, + 693 + ], + "score": 0.87, + "content": "\\mathbf { \\Delta } f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 201, + 681, + 505, + 694 + ], + "score": 1.0, + "content": "in various attributes. We compute mean direction and global direction from", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 691, + 189, + 706 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 189, + 706 + ], + "score": 1.0, + "content": "20 different images.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 107, + 709, + 503, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 504, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 167, + 722 + ], + "score": 1.0, + "content": "We argue that", + "type": "text" + }, + { + "bbox": [ + 167, + 710, + 174, + 720 + ], + "score": 0.74, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 174, + 709, + 504, + 722 + ], + "score": 1.0, + "content": "-space is roughly homogeneous across samples and timesteps. However, we ob-", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 721, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 149, + 734 + ], + "score": 1.0, + "content": "serve that", + "type": "text" + }, + { + "bbox": [ + 150, + 721, + 156, + 731 + ], + "score": 0.75, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 156, + 721, + 505, + 734 + ], + "score": 1.0, + "content": "-space is not completely independent to the conditions especially on unseen-domain", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5 + } + ], + "page_idx": 25, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 106, + 85, + 309, + 97 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 309, + 100 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 299, + 100 + ], + "score": 1.0, + "content": "Algorithm 2: Training Neural implicit function", + "type": "text" + }, + { + "bbox": [ + 299, + 86, + 309, + 97 + ], + "score": 0.81, + "content": "f _ { t }", + "type": "inline_equation" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 109, + 101, + 488, + 137 + ], + "lines": [ + { + "bbox": [ + 102, + 93, + 489, + 126 + ], + "spans": [ + { + "bbox": [ + 102, + 93, + 136, + 126 + ], + "score": 1.0, + "content": "Input:", + "type": "text" + }, + { + "bbox": [ + 136, + 104, + 145, + 114 + ], + "score": 0.77, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 93, + 189, + 126 + ], + "score": 1.0, + "content": "(pretrained attribute),", + "type": "text" + }, + { + "bbox": [ + 208, + 93, + 224, + 126 + ], + "score": 1.0, + "content": "el), (sou", + "type": "text" + }, + { + "bbox": [ + 225, + 100, + 264, + 115 + ], + "score": 0.87, + "content": "\\{ x _ { 0 } ^ { ( i ) } \\} _ { i = 1 } ^ { N }", + "type": "inline_equation" + }, + { + "bbox": [ + 279, + 93, + 329, + 126 + ], + "score": 1.0, + "content": "ages to preco(target text),", + "type": "text" + }, + { + "bbox": [ + 349, + 93, + 362, + 126 + ], + "score": 1.0, + "content": "te), (#", + "type": "text" + }, + { + "bbox": [ + 362, + 103, + 372, + 114 + ], + "score": 0.83, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 372, + 93, + 437, + 126 + ], + "score": 1.0, + "content": "(Neural implicitinversion steps),", + "type": "text" + }, + { + "bbox": [ + 447, + 93, + 489, + 126 + ], + "score": 1.0, + "content": "nction of(# of", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 189, + 114, + 447, + 126 + ], + "spans": [ + { + "bbox": [ + 189, + 114, + 207, + 125 + ], + "score": 0.85, + "content": "y _ { s r c }", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 115, + 278, + 125 + ], + "score": 0.87, + "content": "y _ { t a r }", + "type": "inline_equation" + }, + { + "bbox": [ + 329, + 114, + 348, + 126 + ], + "score": 0.89, + "content": "S _ { f o r }", + "type": "inline_equation" + }, + { + "bbox": [ + 437, + 114, + 447, + 124 + ], + "score": 0.26, + "content": "K", + "type": "inline_equation" + } + ], + "index": 2 + }, + { + "bbox": [ + 136, + 125, + 438, + 137 + ], + "spans": [ + { + "bbox": [ + 136, + 125, + 205, + 137 + ], + "score": 1.0, + "content": "training epochs),", + "type": "text" + }, + { + "bbox": [ + 205, + 126, + 224, + 136 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 125, + 292, + 137 + ], + "score": 1.0, + "content": "(computed from", + "type": "text" + }, + { + "bbox": [ + 293, + 125, + 315, + 136 + ], + "score": 0.8, + "content": "\\ S 4 . 1 \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 125, + 320, + 137 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 320, + 126, + 344, + 136 + ], + "score": 0.81, + "content": "t _ { \\mathrm { b o o s t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 125, + 412, + 137 + ], + "score": 1.0, + "content": "(computed from", + "type": "text" + }, + { + "bbox": [ + 413, + 125, + 435, + 136 + ], + "score": 0.61, + "content": "\\ S 4 . 2 )", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 125, + 438, + 137 + ], + "score": 1.0, + "content": ")", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2, + "bbox_fs": [ + 102, + 93, + 489, + 137 + ] + }, + { + "type": "list", + "bbox": [ + 98, + 140, + 282, + 239 + ], + "lines": [ + { + "bbox": [ + 107, + 140, + 282, + 152 + ], + "spans": [ + { + "bbox": [ + 107, + 140, + 282, + 152 + ], + "score": 1.0, + "content": "// step 1: Precompute latents", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 93, + 146, + 258, + 172 + ], + "spans": [ + { + "bbox": [ + 93, + 146, + 136, + 172 + ], + "score": 1.0, + "content": "1 Define", + "type": "text" + }, + { + "bbox": [ + 136, + 152, + 172, + 167 + ], + "score": 0.61, + "content": "\\{ \\tau _ { s } \\} _ { s = 1 } ^ { S _ { f o r } }", + "type": "inline_equation" + }, + { + "bbox": [ + 172, + 146, + 185, + 172 + ], + "score": 1.0, + "content": "s.t", + "type": "text" + }, + { + "bbox": [ + 185, + 154, + 258, + 167 + ], + "score": 0.71, + "content": "\\tau _ { 1 } = 0 , \\tau _ { S _ { f o r } } = T", + "type": "inline_equation" + } + ], + "index": 5, + "is_list_start_line": true, + "is_list_end_line": true + }, + { + "bbox": [ + 97, + 166, + 196, + 178 + ], + "spans": [ + { + "bbox": [ + 97, + 166, + 121, + 178 + ], + "score": 1.0, + "content": "2 for", + "type": "text" + }, + { + "bbox": [ + 122, + 167, + 180, + 178 + ], + "score": 0.82, + "content": "i = 1 , 2 , . . . , N", + "type": "inline_equation" + }, + { + "bbox": [ + 181, + 166, + 196, + 178 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 6, + "is_list_start_line": true, + "is_list_end_line": true + }, + { + "bbox": [ + 97, + 176, + 239, + 190 + ], + "spans": [ + { + "bbox": [ + 97, + 179, + 105, + 189 + ], + "score": 1.0, + "content": "3", + "type": "text" + }, + { + "bbox": [ + 119, + 176, + 137, + 190 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 137, + 178, + 225, + 189 + ], + "score": 0.78, + "content": "s = 1 , 2 , . . . , S _ { f o r } - 1", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 176, + 239, + 190 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 7, + "is_list_start_line": true, + "is_list_end_line": true + }, + { + "bbox": [ + 98, + 189, + 203, + 203 + ], + "spans": [ + { + "bbox": [ + 98, + 193, + 106, + 203 + ], + "score": 1.0, + "content": "4", + "type": "text" + }, + { + "bbox": [ + 136, + 189, + 203, + 203 + ], + "score": 0.54, + "content": "\\epsilon \\epsilon _ { \\theta } ( x _ { \\tau _ { s } } ^ { ( i ) } , \\bar { \\tau _ { s } } )", + "type": "inline_equation" + } + ], + "index": 8, + "is_list_start_line": true, + "is_list_end_line": true + }, + { + "bbox": [ + 98, + 203, + 269, + 220 + ], + "spans": [ + { + "bbox": [ + 98, + 208, + 106, + 218 + ], + "score": 1.0, + "content": "5", + "type": "text" + }, + { + "bbox": [ + 136, + 203, + 269, + 220 + ], + "score": 0.55, + "content": "\\begin{array} { r } { x _ { \\tau _ { s + 1 } } ^ { ( i ) } = \\sqrt { \\alpha _ { \\tau _ { s } } } x _ { \\tau _ { s } } ^ { ( i ) } + \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon } \\end{array}", + "type": "inline_equation" + } + ], + "index": 9, + "is_list_start_line": true, + "is_list_end_line": true + }, + { + "bbox": [ + 99, + 220, + 213, + 240 + ], + "spans": [ + { + "bbox": [ + 99, + 227, + 104, + 234 + ], + "score": 1.0, + "content": "6", + "type": "text" + }, + { + "bbox": [ + 119, + 220, + 213, + 240 + ], + "score": 1.0, + "content": "Save the latent x(i)τSf or", + "type": "text" + } + ], + "index": 10, + "is_list_end_line": true + } + ], + "index": 7, + "bbox_fs": [ + 93, + 140, + 282, + 240 + ] + }, + { + "type": "text", + "bbox": [ + 97, + 246, + 226, + 278 + ], + "lines": [ + { + "bbox": [ + 96, + 244, + 162, + 258 + ], + "spans": [ + { + "bbox": [ + 96, + 244, + 106, + 258 + ], + "score": 1.0, + "content": "7", + "type": "text" + }, + { + "bbox": [ + 106, + 246, + 162, + 257 + ], + "score": 0.83, + "content": "\\tau _ { S _ { e d i t } } = t _ { e d i t }", + "type": "inline_equation" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 255, + 224, + 268 + ], + "spans": [ + { + "bbox": [ + 106, + 255, + 214, + 268 + ], + "score": 1.0, + "content": "// step 2: Update", + "type": "text" + }, + { + "bbox": [ + 214, + 256, + 224, + 267 + ], + "score": 0.51, + "content": "\\pmb { f } _ { t }", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 97, + 266, + 212, + 279 + ], + "spans": [ + { + "bbox": [ + 97, + 266, + 134, + 279 + ], + "score": 1.0, + "content": "8 for ep", + "type": "text" + }, + { + "bbox": [ + 135, + 267, + 197, + 278 + ], + "score": 0.64, + "content": "{ \\varkappa } h { = 1 , 2 , . . . , K }", + "type": "inline_equation" + }, + { + "bbox": [ + 198, + 266, + 212, + 279 + ], + "score": 1.0, + "content": "do", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12, + "bbox_fs": [ + 96, + 244, + 224, + 279 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 151, + 341, + 393, + 363 + ], + "lines": [ + { + "bbox": [ + 151, + 341, + 393, + 363 + ], + "spans": [ + { + "bbox": [ + 151, + 341, + 393, + 363 + ], + "score": 0.56, + "content": "\\begin{array} { r } { P = \\frac { \\tilde { x } _ { \\tau _ { s } } ^ { ( i ) } - \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon _ { \\theta } \\left( \\tilde { x } _ { \\tau _ { s } } ^ { ( i ) } | \\Delta h _ { \\tau _ { s } } \\right) } { \\sqrt { \\alpha _ { \\tau _ { s } } } } ; P _ { s r c } = \\frac { x _ { \\tau _ { s } } ^ { ( i ) } - \\sqrt { 1 - \\alpha _ { \\tau _ { s } } } \\epsilon _ { \\theta } \\left( x _ { \\tau _ { s } } ^ { ( i ) } \\right) } { \\sqrt { \\alpha _ { \\tau _ { s } } } } } \\end{array}", + "type": "interline_equation", + "image_path": "0ba6b4a72d84a3eab40bb6fb5efa2526ffdd9ef453ce460351eeb798a49902e7.jpg" + } + ] + } + ], + "index": 14, + "virtual_lines": [ + { + "bbox": [ + 151, + 341, + 393, + 363 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 151, + 366, + 332, + 399 + ], + "lines": [ + { + "bbox": [ + 151, + 366, + 332, + 399 + ], + "spans": [ + { + "bbox": [ + 151, + 366, + 332, + 399 + ], + "score": 0.25, + "content": "\\begin{array} { r l } & { \\tilde { x } _ { \\tau _ { s - 1 } } ^ { ( i ) } = \\sqrt { \\alpha _ { \\tau _ { s - 1 } } } P + \\sqrt { 1 - \\alpha _ { \\tau _ { s - 1 } } } \\epsilon _ { \\theta } ( \\tilde { x } _ { \\tau _ { s } } ^ { ( i ) } ) } \\\\ & { x _ { \\tau _ { s - 1 } } ^ { ( i ) } = \\sqrt { \\alpha _ { \\tau _ { s - 1 } } } P _ { s r c } + \\sqrt { 1 - \\alpha _ { \\tau _ { s - 1 } } } \\epsilon _ { \\theta } ( x _ { \\tau _ { s } } ^ { ( i ) } ) } \\end{array}", + "type": "interline_equation", + "image_path": "bc6689a31c48e5761d9ac0c043ce0d1e0ee55fb1266965a19fd3791bbbf44a8e.jpg" + } + ] + } + ], + "index": 15.5, + "virtual_lines": [ + { + "bbox": [ + 151, + 366, + 332, + 382.5 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 151, + 382.5, + 332, + 399.0 + ], + "spans": [], + "index": 16 + } + ] + }, + { + "type": "interline_equation", + "bbox": [ + 152, + 401, + 424, + 414 + ], + "lines": [ + { + "bbox": [ + 152, + 401, + 424, + 414 + ], + "spans": [ + { + "bbox": [ + 152, + 401, + 424, + 414 + ], + "score": 0.62, + "content": "E _ { t o t a l } \\lambda _ { C L I P } L _ { d i r e c t i o n } ( P , y _ { t a r } , P _ { s r c } , y _ { s r c } ) + \\lambda _ { r e c o n } | P - P _ { s r c } |", + "type": "interline_equation", + "image_path": "4af7d31768be7cc1ed82bdab507bf120173b9d97531e91f8e08dfa2efa05090c.jpg" + } + ] + } + ], + "index": 17, + "virtual_lines": [ + { + "bbox": [ + 152, + 401, + 424, + 414 + ], + "spans": [], + "index": 17 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 467, + 505, + 557 + ], + "lines": [ + { + "bbox": [ + 105, + 468, + 505, + 481 + ], + "spans": [ + { + "bbox": [ + 105, + 468, + 401, + 481 + ], + "score": 1.0, + "content": "We use 100 samples per attribute. Asyrp outperforms DiffusionCLIP on", + "type": "text" + }, + { + "bbox": [ + 401, + 468, + 419, + 479 + ], + "score": 0.92, + "content": "S _ { \\mathrm { d i r } }", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 468, + 505, + 481 + ], + "score": 1.0, + "content": "for all attributes. On", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 479, + 505, + 492 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 505, + 492 + ], + "score": 1.0, + "content": "SC, DiffusionCLIP achieves better or competitive scores. Because DiffusionCLIP manipulates im-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 491, + 505, + 503 + ], + "spans": [ + { + "bbox": [ + 106, + 491, + 505, + 503 + ], + "score": 1.0, + "content": "ages mostly by focusing on texture or color while preserving structure and shape, it takes advantage", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 502, + 505, + 513 + ], + "spans": [ + { + "bbox": [ + 106, + 502, + 505, + 513 + ], + "score": 1.0, + "content": "of getting higher SC scores. However, in Figure 28, results of smiling show that it is not proper to", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "score": 1.0, + "content": "edit attributes which require structural manipulation. Note that better SC scores do not guarantee bet-", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 523, + 505, + 535 + ], + "spans": [ + { + "bbox": [ + 105, + 523, + 505, + 535 + ], + "score": 1.0, + "content": "ter qualitative performance. Results of Pixar also show the similar tendency of each method. We", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 534, + 506, + 546 + ], + "spans": [ + { + "bbox": [ + 105, + 534, + 506, + 546 + ], + "score": 1.0, + "content": "allow more structural changes than DiffusionCLIP while editing. Lower SC of our method comes", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 545, + 335, + 557 + ], + "spans": [ + { + "bbox": [ + 105, + 545, + 335, + 557 + ], + "score": 1.0, + "content": "from desirable structural changes as shown in Figure 28.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 21.5, + "bbox_fs": [ + 105, + 468, + 506, + 557 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 212, + 577, + 379, + 604 + ], + "lines": [ + { + "bbox": [ + 212, + 577, + 379, + 604 + ], + "spans": [ + { + "bbox": [ + 212, + 577, + 379, + 604 + ], + "score": 0.94, + "content": "S _ { \\mathrm { d i r } } \\left( x _ { \\mathrm { g e n } } , y _ { \\mathrm { t a r } } ; x _ { \\mathrm { r e f } } , y _ { \\mathrm { r e f } } \\right) : = \\frac { \\left. \\Delta I , \\Delta T \\right. } { \\left\\| \\Delta I \\right\\| \\left\\| \\Delta T \\right\\| } ,", + "type": "interline_equation", + "image_path": "53ce9b39242e61d513e443f453e9d9b59f2f8d6c59eb55f50e784775b71557e9.jpg" + } + ] + } + ], + "index": 26.5, + "virtual_lines": [ + { + "bbox": [ + 212, + 577, + 379, + 590.5 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 212, + 590.5, + 379, + 604.0 + ], + "spans": [], + "index": 27 + } + ] + }, + { + "type": "title", + "bbox": [ + 107, + 624, + 192, + 637 + ], + "lines": [ + { + "bbox": [ + 105, + 623, + 193, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 623, + 193, + 640 + ], + "score": 1.0, + "content": "L DIRECTIONS", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "title", + "bbox": [ + 107, + 649, + 219, + 661 + ], + "lines": [ + { + "bbox": [ + 105, + 649, + 221, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 221, + 662 + ], + "score": 1.0, + "content": "L.1 GLOBAL DIRECTION", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 670, + 505, + 704 + ], + "lines": [ + { + "bbox": [ + 105, + 670, + 505, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 505, + 684 + ], + "score": 1.0, + "content": "Figure 29 and Figure 30 show that the effects of mean direction and global direction are quite", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 681, + 505, + 694 + ], + "spans": [ + { + "bbox": [ + 105, + 681, + 157, + 694 + ], + "score": 1.0, + "content": "similar with", + "type": "text" + }, + { + "bbox": [ + 157, + 682, + 176, + 693 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 176, + 681, + 190, + 694 + ], + "score": 1.0, + "content": "by", + "type": "text" + }, + { + "bbox": [ + 190, + 682, + 200, + 693 + ], + "score": 0.87, + "content": "\\mathbf { \\Delta } f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 201, + 681, + 505, + 694 + ], + "score": 1.0, + "content": "in various attributes. We compute mean direction and global direction from", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 691, + 189, + 706 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 189, + 706 + ], + "score": 1.0, + "content": "20 different images.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31, + "bbox_fs": [ + 105, + 670, + 505, + 706 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 709, + 503, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 504, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 167, + 722 + ], + "score": 1.0, + "content": "We argue that", + "type": "text" + }, + { + "bbox": [ + 167, + 710, + 174, + 720 + ], + "score": 0.74, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 174, + 709, + 504, + 722 + ], + "score": 1.0, + "content": "-space is roughly homogeneous across samples and timesteps. However, we ob-", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 721, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 149, + 734 + ], + "score": 1.0, + "content": "serve that", + "type": "text" + }, + { + "bbox": [ + 150, + 721, + 156, + 731 + ], + "score": 0.75, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 156, + 721, + 505, + 734 + ], + "score": 1.0, + "content": "-space is not completely independent to the conditions especially on unseen-domain", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 453, + 690 + ], + "score": 1.0, + "content": "(See Figure 29). Note that unseen-domains require a longer editing interval with small", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 453, + 678, + 472, + 688 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 472, + 676, + 505, + 690 + ], + "score": 1.0, + "content": ". There-", + "type": "text", + "cross_page": true + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 277, + 700 + ], + "score": 1.0, + "content": "fore, we conjecture that the consistency of", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 277, + 688, + 284, + 698 + ], + "score": 0.74, + "content": "h", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 284, + 687, + 505, + 700 + ], + "score": 1.0, + "content": "-space decreases at the end of the generative process. It", + "type": "text", + "cross_page": true + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 699, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 363, + 712 + ], + "score": 1.0, + "content": "is supported by additional experiments that L2 distance between", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 363, + 699, + 382, + 710 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 383, + 699, + 505, + 712 + ], + "score": 1.0, + "content": "and global direction gradually", + "type": "text", + "cross_page": true + } + ], + "index": 17 + }, + { + "bbox": [ + 104, + 708, + 505, + 724 + ], + "spans": [ + { + "bbox": [ + 104, + 708, + 384, + 724 + ], + "score": 1.0, + "content": "increases along with timesteps. We leave a more detailed analysis on", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 385, + 711, + 391, + 720 + ], + "score": 0.77, + "content": "h", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 391, + 708, + 505, + 724 + ], + "score": 1.0, + "content": "-space at different timesteps", + "type": "text", + "cross_page": true + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 721, + 176, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 176, + 732 + ], + "score": 1.0, + "content": "as a future work.", + "type": "text", + "cross_page": true + } + ], + "index": 19 + } + ], + "index": 33.5, + "bbox_fs": [ + 105, + 709, + 505, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 139, + 505, + 425 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 139, + 505, + 425 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 139, + 505, + 425 + ], + "spans": [ + { + "bbox": [ + 106, + 139, + 505, + 425 + ], + "score": 0.984, + "html": "
Datasetsrc_txttrg_txtXCLIPXreconfrom random noise
CelebA-HQPersonYoung person0.80.905*3X
FaceSmiling face0.80.899*3X
FaceSad face0.80.894*3X
FaceAngry face0.80.892*3X
FaceTanned face0.80.886*3X
FaceDisgusted face0.80.880*3X
PersonPerson with makeup0.80.875*3X
Person PhotoPerson with curly hair0.80.835*3X
Self-portrait by Frida Kahlo0.50.44330
HumanZombie0.60.868*30
PersonNicolas Cage0.80.710*30
Human PhotoPainting in the style of Pixar0.80.667*30
HumanPainting in Modigliani style0.80.565*30
LSUN-churchChurchNeanderthal Gothic church1.20.802*30
ChurchTemple0.80.912*30
ChurchDepartment store0.80.898*30
ChurchWooden house0.80.841*30
ChurchAncient traditional Asian tower0.8 0.80.793*30
ChurchRed brick wall Church0.80.784*3 0.774*30
ChurchFactory0.80.702*30
LSUN-bedroomBedroomPrincess bedroom1.50.912*30 0
BedroomHotel bedroom1.50.917*3
AFHQ-dogDogHappy dog0
DogSleepy dog1.50.883*30
Dog1.50.866*30
DogWolf Yorkshire terrier1.70.850*30
METFACESPainting of a personPainting of a Sad person2 1.50.690*3 0.921*30 X
Painting of a personPainting of a Smiling person1.50.908*3X
Painting of a personPainting of a Disgusted person1.50.879*3X
", + "type": "table", + "image_path": "93e57daef13c9d000d22d16cc5b1feb897cc08af451e893d40410c75fce7b88a.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 139, + 505, + 234.33333333333331 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 234.33333333333331, + 505, + 329.66666666666663 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 329.66666666666663, + 505, + 424.99999999999994 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 106, + 434, + 505, + 479 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 434, + 505, + 446 + ], + "spans": [ + { + "bbox": [ + 106, + 434, + 505, + 446 + ], + "score": 1.0, + "content": "Table 3: The coefficients range from 0.5 to 0.8 for in-domain attributes. Unseen domains need", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 444, + 505, + 458 + ], + "spans": [ + { + "bbox": [ + 105, + 444, + 176, + 458 + ], + "score": 1.0, + "content": "slightly stronger", + "type": "text" + }, + { + "bbox": [ + 176, + 445, + 207, + 456 + ], + "score": 0.85, + "content": "\\lambda _ { \\mathrm { { C L I P } } } \\mathrm { { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 207, + 444, + 505, + 458 + ], + "score": 1.0, + "content": ". We also report which attributes we train with random noise sampling.", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 454, + 505, + 469 + ], + "spans": [ + { + "bbox": [ + 105, + 454, + 465, + 469 + ], + "score": 1.0, + "content": "The criterion is which attributes require relatively less maintenance of identity. We use", + "type": "text" + }, + { + "bbox": [ + 465, + 456, + 505, + 468 + ], + "score": 0.87, + "content": "\\lambda _ { \\mathrm { { r e c o n } } } =", + "type": "inline_equation" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 466, + 438, + 480 + ], + "spans": [ + { + "bbox": [ + 105, + 466, + 171, + 480 + ], + "score": 1.0, + "content": "CLIP similarity", + "type": "text" + }, + { + "bbox": [ + 171, + 468, + 185, + 477 + ], + "score": 0.81, + "content": "^ \\ast 3", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 466, + 438, + 480 + ], + "score": 1.0, + "content": "which reduces L1 loss when an attribute needs a lot of change.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "table", + "bbox": [ + 106, + 602, + 503, + 652 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 602, + 503, + 652 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 602, + 503, + 652 + ], + "spans": [ + { + "bbox": [ + 106, + 602, + 503, + 652 + ], + "score": 0.966, + "html": "
CelebA-HQ-in-domainCelebA-HQ-unseen-domain
SmilingSadTannedPixarNeanderthal
SdirSCSdirSCSdirSCSdirSCSdirSC
Asyrp (ours)0.92189.02%0.96488.90%0.99185.71%0.95676.51%0.80579.03
DiffusionCLIP0.81391.41%0.76089.93%0.88892.85%0.81189.91%0.66181.23%
", + "type": "table", + "image_path": "39128b22784c462f3d60b7d120d611eaf69db40d8f68d9dcfcebcfddaf3e23f5.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 106, + 602, + 503, + 618.6666666666666 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 106, + 618.6666666666666, + 503, + 635.3333333333333 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 106, + 635.3333333333333, + 503, + 651.9999999999999 + ], + "spans": [], + "index": 9 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 203, + 659, + 408, + 671 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 202, + 658, + 408, + 673 + ], + "spans": [ + { + "bbox": [ + 202, + 658, + 408, + 673 + ], + "score": 1.0, + "content": "Table 4: Quantitative evaluation on CelebA-HQ.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 10 + } + ], + "index": 9.0 + } + ], + "page_idx": 26, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 765 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 765 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 139, + 505, + 425 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 139, + 505, + 425 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 139, + 505, + 425 + ], + "spans": [ + { + "bbox": [ + 106, + 139, + 505, + 425 + ], + "score": 0.984, + "html": "
Datasetsrc_txttrg_txtXCLIPXreconfrom random noise
CelebA-HQPersonYoung person0.80.905*3X
FaceSmiling face0.80.899*3X
FaceSad face0.80.894*3X
FaceAngry face0.80.892*3X
FaceTanned face0.80.886*3X
FaceDisgusted face0.80.880*3X
PersonPerson with makeup0.80.875*3X
Person PhotoPerson with curly hair0.80.835*3X
Self-portrait by Frida Kahlo0.50.44330
HumanZombie0.60.868*30
PersonNicolas Cage0.80.710*30
Human PhotoPainting in the style of Pixar0.80.667*30
HumanPainting in Modigliani style0.80.565*30
LSUN-churchChurchNeanderthal Gothic church1.20.802*30
ChurchTemple0.80.912*30
ChurchDepartment store0.80.898*30
ChurchWooden house0.80.841*30
ChurchAncient traditional Asian tower0.8 0.80.793*30
ChurchRed brick wall Church0.80.784*3 0.774*30
ChurchFactory0.80.702*30
LSUN-bedroomBedroomPrincess bedroom1.50.912*30 0
BedroomHotel bedroom1.50.917*3
AFHQ-dogDogHappy dog0
DogSleepy dog1.50.883*30
Dog1.50.866*30
DogWolf Yorkshire terrier1.70.850*30
METFACESPainting of a personPainting of a Sad person2 1.50.690*3 0.921*30 X
Painting of a personPainting of a Smiling person1.50.908*3X
Painting of a personPainting of a Disgusted person1.50.879*3X
", + "type": "table", + "image_path": "93e57daef13c9d000d22d16cc5b1feb897cc08af451e893d40410c75fce7b88a.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 139, + 505, + 234.33333333333331 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 234.33333333333331, + 505, + 329.66666666666663 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 329.66666666666663, + 505, + 424.99999999999994 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 106, + 434, + 505, + 479 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 434, + 505, + 446 + ], + "spans": [ + { + "bbox": [ + 106, + 434, + 505, + 446 + ], + "score": 1.0, + "content": "Table 3: The coefficients range from 0.5 to 0.8 for in-domain attributes. Unseen domains need", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 444, + 505, + 458 + ], + "spans": [ + { + "bbox": [ + 105, + 444, + 176, + 458 + ], + "score": 1.0, + "content": "slightly stronger", + "type": "text" + }, + { + "bbox": [ + 176, + 445, + 207, + 456 + ], + "score": 0.85, + "content": "\\lambda _ { \\mathrm { { C L I P } } } \\mathrm { { s } }", + "type": "inline_equation" + }, + { + "bbox": [ + 207, + 444, + 505, + 458 + ], + "score": 1.0, + "content": ". We also report which attributes we train with random noise sampling.", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 454, + 505, + 469 + ], + "spans": [ + { + "bbox": [ + 105, + 454, + 465, + 469 + ], + "score": 1.0, + "content": "The criterion is which attributes require relatively less maintenance of identity. We use", + "type": "text" + }, + { + "bbox": [ + 465, + 456, + 505, + 468 + ], + "score": 0.87, + "content": "\\lambda _ { \\mathrm { { r e c o n } } } =", + "type": "inline_equation" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 466, + 438, + 480 + ], + "spans": [ + { + "bbox": [ + 105, + 466, + 171, + 480 + ], + "score": 1.0, + "content": "CLIP similarity", + "type": "text" + }, + { + "bbox": [ + 171, + 468, + 185, + 477 + ], + "score": 0.81, + "content": "^ \\ast 3", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 466, + 438, + 480 + ], + "score": 1.0, + "content": "which reduces L1 loss when an attribute needs a lot of change.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "table", + "bbox": [ + 106, + 602, + 503, + 652 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 602, + 503, + 652 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 602, + 503, + 652 + ], + "spans": [ + { + "bbox": [ + 106, + 602, + 503, + 652 + ], + "score": 0.966, + "html": "
CelebA-HQ-in-domainCelebA-HQ-unseen-domain
SmilingSadTannedPixarNeanderthal
SdirSCSdirSCSdirSCSdirSCSdirSC
Asyrp (ours)0.92189.02%0.96488.90%0.99185.71%0.95676.51%0.80579.03
DiffusionCLIP0.81391.41%0.76089.93%0.88892.85%0.81189.91%0.66181.23%
", + "type": "table", + "image_path": "39128b22784c462f3d60b7d120d611eaf69db40d8f68d9dcfcebcfddaf3e23f5.jpg" + } + ] + } + ], + "index": 8, + "virtual_lines": [ + { + "bbox": [ + 106, + 602, + 503, + 618.6666666666666 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 106, + 618.6666666666666, + 503, + 635.3333333333333 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 106, + 635.3333333333333, + 503, + 651.9999999999999 + ], + "spans": [], + "index": 9 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 203, + 659, + 408, + 671 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 202, + 658, + 408, + 673 + ], + "spans": [ + { + "bbox": [ + 202, + 658, + 408, + 673 + ], + "score": 1.0, + "content": "Table 4: Quantitative evaluation on CelebA-HQ.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 10 + } + ], + "index": 9.0 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 108, + 81, + 502, + 275 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 81, + 502, + 275 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 81, + 502, + 275 + ], + "spans": [ + { + "bbox": [ + 108, + 81, + 502, + 275 + ], + "score": 0.966, + "type": "image", + "image_path": "c2cf22d51b44b51faa69e1a51e8e027ba84a40344d50621d1a284856a9eb758c.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 108, + 81, + 502, + 145.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 108, + 145.66666666666669, + 502, + 210.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 210.33333333333337, + 502, + 275.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 288, + 505, + 322 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 288, + 505, + 300 + ], + "spans": [ + { + "bbox": [ + 105, + 288, + 505, + 300 + ], + "score": 1.0, + "content": "Figure 25: Asyrp vs. DiffusionCLIP on CelebA-HQ in-domain attributes. We observe that Dif-", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 299, + 505, + 312 + ], + "spans": [ + { + "bbox": [ + 106, + 299, + 505, + 312 + ], + "score": 1.0, + "content": "fusionCLIP struggles to change semantic facial attributes. Their official checkpoints do not exist and", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 312, + 309, + 324 + ], + "spans": [ + { + "bbox": [ + 106, + 312, + 309, + 324 + ], + "score": 1.0, + "content": "we ran careful hyperparameter tuning for training.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 110, + 328, + 502, + 521 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 110, + 328, + 502, + 521 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 110, + 328, + 502, + 521 + ], + "spans": [ + { + "bbox": [ + 110, + 328, + 502, + 521 + ], + "score": 0.964, + "type": "image", + "image_path": "06f561db0257bb962e8c406afb37da80494f09913966d224a12c97e0991f5899.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 110, + 328, + 502, + 392.3333333333333 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 110, + 392.3333333333333, + 502, + 456.66666666666663 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 110, + 456.66666666666663, + 502, + 521.0 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 534, + 504, + 558 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 534, + 504, + 547 + ], + "spans": [ + { + "bbox": [ + 106, + 534, + 504, + 547 + ], + "score": 1.0, + "content": "Figure 26: Asyrp vs. DiffusionCLIP on CelebA-HQ unseen-domain attributes. We use the offi-", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 547, + 475, + 558 + ], + "spans": [ + { + "bbox": [ + 106, + 547, + 475, + 558 + ], + "score": 1.0, + "content": "cial checkpoint provided by DiffusionCLIP. Asyrp works well even for the unseen-domains.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + } + ], + "index": 8.25 + }, + { + "type": "table", + "bbox": [ + 149, + 576, + 462, + 634 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 149, + 576, + 462, + 634 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 149, + 576, + 462, + 634 + ], + "spans": [ + { + "bbox": [ + 149, + 576, + 462, + 634 + ], + "score": 0.969, + "html": "
LSUN_church
Department storeancientred brick
SdirSCSdirSCSdirSC
Asyrp (ours)0.77857.62%0.94362.65%0.98965.83%
DIffusionCLIP0.66154.50%0.90764.82%0.96465.02%
", + "type": "table", + "image_path": "63d4f8f633262547023dff557121d24f1c8c7ecb03ea93fc468a9540dcddfdf3.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 149, + 576, + 462, + 595.3333333333334 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 149, + 595.3333333333334, + 462, + 614.6666666666667 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 149, + 614.6666666666667, + 462, + 634.0000000000001 + ], + "spans": [], + "index": 13 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 199, + 641, + 412, + 654 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 198, + 641, + 413, + 654 + ], + "spans": [ + { + "bbox": [ + 198, + 641, + 413, + 654 + ], + "score": 1.0, + "content": "Table 5: Quantitative evaluation on LSUN-church.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14 + } + ], + "index": 13.0 + }, + { + "type": "text", + "bbox": [ + 106, + 676, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 453, + 690 + ], + "score": 1.0, + "content": "(See Figure 29). Note that unseen-domains require a longer editing interval with small", + "type": "text" + }, + { + "bbox": [ + 453, + 678, + 472, + 688 + ], + "score": 0.89, + "content": "t _ { \\mathrm { e d i t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 472, + 676, + 505, + 690 + ], + "score": 1.0, + "content": ". There-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 277, + 700 + ], + "score": 1.0, + "content": "fore, we conjecture that the consistency of", + "type": "text" + }, + { + "bbox": [ + 277, + 688, + 284, + 698 + ], + "score": 0.74, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 284, + 687, + 505, + 700 + ], + "score": 1.0, + "content": "-space decreases at the end of the generative process. It", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 699, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 363, + 712 + ], + "score": 1.0, + "content": "is supported by additional experiments that L2 distance between", + "type": "text" + }, + { + "bbox": [ + 363, + 699, + 382, + 710 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 383, + 699, + 505, + 712 + ], + "score": 1.0, + "content": "and global direction gradually", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 104, + 708, + 505, + 724 + ], + "spans": [ + { + "bbox": [ + 104, + 708, + 384, + 724 + ], + "score": 1.0, + "content": "increases along with timesteps. We leave a more detailed analysis on", + "type": "text" + }, + { + "bbox": [ + 385, + 711, + 391, + 720 + ], + "score": 0.77, + "content": "h", + "type": "inline_equation" + }, + { + "bbox": [ + 391, + 708, + 505, + 724 + ], + "score": 1.0, + "content": "-space at different timesteps", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 721, + 176, + 732 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 176, + 732 + ], + "score": 1.0, + "content": "as a future work.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 17 + } + ], + "page_idx": 27, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 312, + 763 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 312, + 763 + ], + "score": 1.0, + "content": "28", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 108, + 81, + 502, + 275 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 81, + 502, + 275 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 108, + 81, + 502, + 275 + ], + "spans": [ + { + "bbox": [ + 108, + 81, + 502, + 275 + ], + "score": 0.966, + "type": "image", + "image_path": "c2cf22d51b44b51faa69e1a51e8e027ba84a40344d50621d1a284856a9eb758c.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 108, + 81, + 502, + 145.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 108, + 145.66666666666669, + 502, + 210.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 108, + 210.33333333333337, + 502, + 275.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 288, + 505, + 322 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 288, + 505, + 300 + ], + "spans": [ + { + "bbox": [ + 105, + 288, + 505, + 300 + ], + "score": 1.0, + "content": "Figure 25: Asyrp vs. DiffusionCLIP on CelebA-HQ in-domain attributes. We observe that Dif-", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 299, + 505, + 312 + ], + "spans": [ + { + "bbox": [ + 106, + 299, + 505, + 312 + ], + "score": 1.0, + "content": "fusionCLIP struggles to change semantic facial attributes. Their official checkpoints do not exist and", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 312, + 309, + 324 + ], + "spans": [ + { + "bbox": [ + 106, + 312, + 309, + 324 + ], + "score": 1.0, + "content": "we ran careful hyperparameter tuning for training.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 110, + 328, + 502, + 521 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 110, + 328, + 502, + 521 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 110, + 328, + 502, + 521 + ], + "spans": [ + { + "bbox": [ + 110, + 328, + 502, + 521 + ], + "score": 0.964, + "type": "image", + "image_path": "06f561db0257bb962e8c406afb37da80494f09913966d224a12c97e0991f5899.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 110, + 328, + 502, + 392.3333333333333 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 110, + 392.3333333333333, + 502, + 456.66666666666663 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 110, + 456.66666666666663, + 502, + 521.0 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 534, + 504, + 558 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 534, + 504, + 547 + ], + "spans": [ + { + "bbox": [ + 106, + 534, + 504, + 547 + ], + "score": 1.0, + "content": "Figure 26: Asyrp vs. DiffusionCLIP on CelebA-HQ unseen-domain attributes. We use the offi-", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 547, + 475, + 558 + ], + "spans": [ + { + "bbox": [ + 106, + 547, + 475, + 558 + ], + "score": 1.0, + "content": "cial checkpoint provided by DiffusionCLIP. Asyrp works well even for the unseen-domains.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + } + ], + "index": 8.25 + }, + { + "type": "table", + "bbox": [ + 149, + 576, + 462, + 634 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 149, + 576, + 462, + 634 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 149, + 576, + 462, + 634 + ], + "spans": [ + { + "bbox": [ + 149, + 576, + 462, + 634 + ], + "score": 0.969, + "html": "
LSUN_church
Department storeancientred brick
SdirSCSdirSCSdirSC
Asyrp (ours)0.77857.62%0.94362.65%0.98965.83%
DIffusionCLIP0.66154.50%0.90764.82%0.96465.02%
", + "type": "table", + "image_path": "63d4f8f633262547023dff557121d24f1c8c7ecb03ea93fc468a9540dcddfdf3.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 149, + 576, + 462, + 595.3333333333334 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 149, + 595.3333333333334, + 462, + 614.6666666666667 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 149, + 614.6666666666667, + 462, + 634.0000000000001 + ], + "spans": [], + "index": 13 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 199, + 641, + 412, + 654 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 198, + 641, + 413, + 654 + ], + "spans": [ + { + "bbox": [ + 198, + 641, + 413, + 654 + ], + "score": 1.0, + "content": "Table 5: Quantitative evaluation on LSUN-church.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14 + } + ], + "index": 13.0 + }, + { + "type": "text", + "bbox": [ + 106, + 676, + 505, + 732 + ], + "lines": [], + "index": 17, + "bbox_fs": [ + 104, + 676, + 505, + 732 + ], + "lines_deleted": true + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 105, + 74, + 502, + 476 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 105, + 74, + 502, + 476 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 74, + 502, + 476 + ], + "spans": [ + { + "bbox": [ + 105, + 74, + 502, + 476 + ], + "score": 0.979, + "type": "image", + "image_path": "e6959cd9188841bfcc00b46aeed46d7800c43970a4f909834f9efccdda44607a.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 105, + 74, + 502, + 208.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 105, + 208.0, + 502, + 342.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 105, + 342.0, + 502, + 476.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 487, + 504, + 522 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 488, + 505, + 500 + ], + "spans": [ + { + "bbox": [ + 106, + 488, + 505, + 500 + ], + "score": 1.0, + "content": "Figure 27: Asyrp vs. DiffusionCLIP on LSUN-church. Note that we use the official checkpoint", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 498, + 505, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 498, + 505, + 513 + ], + "score": 1.0, + "content": "provided by DiffusionCLIP. We observe DiffusionCLIP produces narrow range of styles while Asyrp", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 509, + 204, + 522 + ], + "spans": [ + { + "bbox": [ + 105, + 509, + 204, + 522 + ], + "score": 1.0, + "content": "produces diverse styles.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "title", + "bbox": [ + 108, + 547, + 251, + 559 + ], + "lines": [ + { + "bbox": [ + 106, + 547, + 252, + 559 + ], + "spans": [ + { + "bbox": [ + 106, + 547, + 252, + 559 + ], + "score": 1.0, + "content": "L.2 COMPARE THREE METHODS", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 105, + 568, + 504, + 591 + ], + "lines": [ + { + "bbox": [ + 105, + 568, + 504, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 568, + 385, + 581 + ], + "score": 1.0, + "content": "In this section, we compare three methods: implicit neural direction", + "type": "text" + }, + { + "bbox": [ + 386, + 569, + 395, + 581 + ], + "score": 0.88, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 396, + 568, + 443, + 581 + ], + "score": 1.0, + "content": ", optimized", + "type": "text" + }, + { + "bbox": [ + 443, + 569, + 461, + 580 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 462, + 568, + 504, + 581 + ], + "score": 1.0, + "content": ", and opti-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 104, + 578, + 172, + 591 + ], + "spans": [ + { + "bbox": [ + 104, + 578, + 172, + 591 + ], + "score": 1.0, + "content": "mized ∆hglobal.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 108, + 603, + 264, + 617 + ], + "lines": [ + { + "bbox": [ + 105, + 601, + 264, + 619 + ], + "spans": [ + { + "bbox": [ + 105, + 601, + 173, + 619 + ], + "score": 1.0, + "content": "Training time", + "type": "text" + }, + { + "bbox": [ + 173, + 603, + 264, + 616 + ], + "score": 0.74, + "content": "f _ { t } \\approx \\Delta h ^ { g l o b a l } < \\Delta h _ { t }", + "type": "inline_equation", + "image_path": "1d299396c6f675caeb1cd1048f4afd5a14f276f0f482d2835097051d9c431a05.jpg" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 107, + 621, + 505, + 655 + ], + "lines": [ + { + "bbox": [ + 106, + 620, + 505, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 620, + 212, + 634 + ], + "score": 1.0, + "content": "We have to optimize each", + "type": "text" + }, + { + "bbox": [ + 212, + 622, + 231, + 632 + ], + "score": 0.91, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 231, + 620, + 306, + 634 + ], + "score": 1.0, + "content": "for each time step", + "type": "text" + }, + { + "bbox": [ + 306, + 623, + 311, + 631 + ], + "score": 0.63, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 312, + 620, + 505, + 634 + ], + "score": 1.0, + "content": ". Additionally, it needs specific hyperparameters", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 631, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 631, + 144, + 645 + ], + "score": 1.0, + "content": "for each", + "type": "text" + }, + { + "bbox": [ + 144, + 632, + 162, + 643 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 163, + 631, + 318, + 645 + ], + "score": 1.0, + "content": ", e.g., higher learning rates for larger", + "type": "text" + }, + { + "bbox": [ + 319, + 633, + 324, + 642 + ], + "score": 0.67, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 631, + 483, + 645 + ], + "score": 1.0, + "content": ". On the contrary, time-consuming for", + "type": "text" + }, + { + "bbox": [ + 484, + 632, + 493, + 644 + ], + "score": 0.88, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 494, + 631, + 505, + 645 + ], + "score": 1.0, + "content": "is", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 641, + 232, + 656 + ], + "spans": [ + { + "bbox": [ + 104, + 641, + 232, + 656 + ], + "score": 1.0, + "content": "similar to optimizing ∆hglobal.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11 + }, + { + "type": "title", + "bbox": [ + 107, + 666, + 239, + 680 + ], + "lines": [ + { + "bbox": [ + 105, + 666, + 147, + 681 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 147, + 681 + ], + "score": 1.0, + "content": "Quality", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 107, + 685, + 505, + 708 + ], + "lines": [ + { + "bbox": [ + 107, + 684, + 505, + 698 + ], + "spans": [ + { + "bbox": [ + 107, + 685, + 116, + 696 + ], + "score": 0.85, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 117, + 684, + 136, + 698 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 136, + 685, + 154, + 696 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 684, + 505, + 698 + ], + "score": 1.0, + "content": ", where directions can be obtained for each timestep, have the best quality. As can be", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 104, + 693, + 467, + 709 + ], + "spans": [ + { + "bbox": [ + 104, + 693, + 188, + 709 + ], + "score": 1.0, + "content": "shown in Figure 10,", + "type": "text" + }, + { + "bbox": [ + 189, + 695, + 225, + 706 + ], + "score": 0.87, + "content": "\\Delta h ^ { g l o b a l }", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 693, + 467, + 709 + ], + "score": 1.0, + "content": "is sometimes accompanied by slight differences in hair, etc.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5 + }, + { + "type": "text", + "bbox": [ + 107, + 719, + 259, + 732 + ], + "lines": [ + { + "bbox": [ + 104, + 718, + 259, + 733 + ], + "spans": [ + { + "bbox": [ + 104, + 718, + 169, + 733 + ], + "score": 1.0, + "content": "Extensibility", + "type": "text" + }, + { + "bbox": [ + 169, + 719, + 259, + 732 + ], + "score": 0.65, + "content": "f _ { t } > \\Delta h _ { t } > \\Delta h ^ { g l o b a l }", + "type": "inline_equation", + "image_path": "80d14b77af4cb47296b42376a79464bbeb5a78c0f58be4baa3bcf4247056cc52.jpg" + } + ], + "index": 16 + } + ], + "index": 16 + } + ], + "page_idx": 28, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 108, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 294, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 105, + 74, + 502, + 476 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 105, + 74, + 502, + 476 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 74, + 502, + 476 + ], + "spans": [ + { + "bbox": [ + 105, + 74, + 502, + 476 + ], + "score": 0.979, + "type": "image", + "image_path": "e6959cd9188841bfcc00b46aeed46d7800c43970a4f909834f9efccdda44607a.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 105, + 74, + 502, + 208.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 105, + 208.0, + 502, + 342.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 105, + 342.0, + 502, + 476.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 487, + 504, + 522 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 488, + 505, + 500 + ], + "spans": [ + { + "bbox": [ + 106, + 488, + 505, + 500 + ], + "score": 1.0, + "content": "Figure 27: Asyrp vs. DiffusionCLIP on LSUN-church. Note that we use the official checkpoint", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 498, + 505, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 498, + 505, + 513 + ], + "score": 1.0, + "content": "provided by DiffusionCLIP. We observe DiffusionCLIP produces narrow range of styles while Asyrp", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 509, + 204, + 522 + ], + "spans": [ + { + "bbox": [ + 105, + 509, + 204, + 522 + ], + "score": 1.0, + "content": "produces diverse styles.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "title", + "bbox": [ + 108, + 547, + 251, + 559 + ], + "lines": [ + { + "bbox": [ + 106, + 547, + 252, + 559 + ], + "spans": [ + { + "bbox": [ + 106, + 547, + 252, + 559 + ], + "score": 1.0, + "content": "L.2 COMPARE THREE METHODS", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 105, + 568, + 504, + 591 + ], + "lines": [ + { + "bbox": [ + 105, + 568, + 504, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 568, + 385, + 581 + ], + "score": 1.0, + "content": "In this section, we compare three methods: implicit neural direction", + "type": "text" + }, + { + "bbox": [ + 386, + 569, + 395, + 581 + ], + "score": 0.88, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 396, + 568, + 443, + 581 + ], + "score": 1.0, + "content": ", optimized", + "type": "text" + }, + { + "bbox": [ + 443, + 569, + 461, + 580 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 462, + 568, + 504, + 581 + ], + "score": 1.0, + "content": ", and opti-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 104, + 578, + 172, + 591 + ], + "spans": [ + { + "bbox": [ + 104, + 578, + 172, + 591 + ], + "score": 1.0, + "content": "mized ∆hglobal.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5, + "bbox_fs": [ + 104, + 568, + 504, + 591 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 603, + 264, + 617 + ], + "lines": [ + { + "bbox": [ + 105, + 601, + 264, + 619 + ], + "spans": [ + { + "bbox": [ + 105, + 601, + 173, + 619 + ], + "score": 1.0, + "content": "Training time", + "type": "text" + }, + { + "bbox": [ + 173, + 603, + 264, + 616 + ], + "score": 0.74, + "content": "f _ { t } \\approx \\Delta h ^ { g l o b a l } < \\Delta h _ { t }", + "type": "inline_equation", + "image_path": "1d299396c6f675caeb1cd1048f4afd5a14f276f0f482d2835097051d9c431a05.jpg" + } + ], + "index": 9 + } + ], + "index": 9, + "bbox_fs": [ + 105, + 601, + 264, + 619 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 621, + 505, + 655 + ], + "lines": [ + { + "bbox": [ + 106, + 620, + 505, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 620, + 212, + 634 + ], + "score": 1.0, + "content": "We have to optimize each", + "type": "text" + }, + { + "bbox": [ + 212, + 622, + 231, + 632 + ], + "score": 0.91, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 231, + 620, + 306, + 634 + ], + "score": 1.0, + "content": "for each time step", + "type": "text" + }, + { + "bbox": [ + 306, + 623, + 311, + 631 + ], + "score": 0.63, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 312, + 620, + 505, + 634 + ], + "score": 1.0, + "content": ". Additionally, it needs specific hyperparameters", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 631, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 105, + 631, + 144, + 645 + ], + "score": 1.0, + "content": "for each", + "type": "text" + }, + { + "bbox": [ + 144, + 632, + 162, + 643 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 163, + 631, + 318, + 645 + ], + "score": 1.0, + "content": ", e.g., higher learning rates for larger", + "type": "text" + }, + { + "bbox": [ + 319, + 633, + 324, + 642 + ], + "score": 0.67, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 631, + 483, + 645 + ], + "score": 1.0, + "content": ". On the contrary, time-consuming for", + "type": "text" + }, + { + "bbox": [ + 484, + 632, + 493, + 644 + ], + "score": 0.88, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 494, + 631, + 505, + 645 + ], + "score": 1.0, + "content": "is", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 641, + 232, + 656 + ], + "spans": [ + { + "bbox": [ + 104, + 641, + 232, + 656 + ], + "score": 1.0, + "content": "similar to optimizing ∆hglobal.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11, + "bbox_fs": [ + 104, + 620, + 505, + 656 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 666, + 239, + 680 + ], + "lines": [ + { + "bbox": [ + 105, + 666, + 147, + 681 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 147, + 681 + ], + "score": 1.0, + "content": "Quality", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 107, + 685, + 505, + 708 + ], + "lines": [ + { + "bbox": [ + 107, + 684, + 505, + 698 + ], + "spans": [ + { + "bbox": [ + 107, + 685, + 116, + 696 + ], + "score": 0.85, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 117, + 684, + 136, + 698 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 136, + 685, + 154, + 696 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 684, + 505, + 698 + ], + "score": 1.0, + "content": ", where directions can be obtained for each timestep, have the best quality. As can be", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 104, + 693, + 467, + 709 + ], + "spans": [ + { + "bbox": [ + 104, + 693, + 188, + 709 + ], + "score": 1.0, + "content": "shown in Figure 10,", + "type": "text" + }, + { + "bbox": [ + 189, + 695, + 225, + 706 + ], + "score": 0.87, + "content": "\\Delta h ^ { g l o b a l }", + "type": "inline_equation" + }, + { + "bbox": [ + 225, + 693, + 467, + 709 + ], + "score": 1.0, + "content": "is sometimes accompanied by slight differences in hair, etc.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5, + "bbox_fs": [ + 104, + 684, + 505, + 709 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 719, + 259, + 732 + ], + "lines": [ + { + "bbox": [ + 104, + 718, + 259, + 733 + ], + "spans": [ + { + "bbox": [ + 104, + 718, + 169, + 733 + ], + "score": 1.0, + "content": "Extensibility", + "type": "text" + }, + { + "bbox": [ + 169, + 719, + 259, + 732 + ], + "score": 0.65, + "content": "f _ { t } > \\Delta h _ { t } > \\Delta h ^ { g l o b a l }", + "type": "inline_equation", + "image_path": "80d14b77af4cb47296b42376a79464bbeb5a78c0f58be4baa3bcf4247056cc52.jpg" + } + ], + "index": 16 + } + ], + "index": 16, + "bbox_fs": [ + 104, + 718, + 259, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 159, + 83, + 424, + 499 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 159, + 83, + 424, + 499 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 159, + 83, + 424, + 499 + ], + "spans": [ + { + "bbox": [ + 159, + 83, + 424, + 499 + ], + "score": 0.952, + "type": "image", + "image_path": "4bcaa6683a66fe5dc47e975b0ea521a3cef4e214da2c1f2d293b268e362cc228.jpg" + } + ] + } + ], + "index": 13.5, + "virtual_lines": [ + { + "bbox": [ + 159, + 83, + 424, + 97.85714285714286 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 159, + 97.85714285714286, + 424, + 112.71428571428572 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 159, + 112.71428571428572, + 424, + 127.57142857142858 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 159, + 127.57142857142858, + 424, + 142.42857142857144 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 159, + 142.42857142857144, + 424, + 157.2857142857143 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 159, + 157.2857142857143, + 424, + 172.14285714285717 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 159, + 172.14285714285717, + 424, + 187.00000000000003 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 159, + 187.00000000000003, + 424, + 201.8571428571429 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 159, + 201.8571428571429, + 424, + 216.71428571428575 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 159, + 216.71428571428575, + 424, + 231.5714285714286 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 159, + 231.5714285714286, + 424, + 246.42857142857147 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 159, + 246.42857142857147, + 424, + 261.28571428571433 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 159, + 261.28571428571433, + 424, + 276.14285714285717 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 159, + 276.14285714285717, + 424, + 291.0 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 159, + 291.0, + 424, + 305.85714285714283 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 159, + 305.85714285714283, + 424, + 320.71428571428567 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 159, + 320.71428571428567, + 424, + 335.5714285714285 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 159, + 335.5714285714285, + 424, + 350.42857142857133 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 159, + 350.42857142857133, + 424, + 365.28571428571416 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 159, + 365.28571428571416, + 424, + 380.142857142857 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 159, + 380.142857142857, + 424, + 394.99999999999983 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 159, + 394.99999999999983, + 424, + 409.85714285714266 + ], + "spans": [], + "index": 21 + }, + { + "bbox": [ + 159, + 409.85714285714266, + 424, + 424.7142857142855 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 159, + 424.7142857142855, + 424, + 439.5714285714283 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 159, + 439.5714285714283, + 424, + 454.42857142857116 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 159, + 454.42857142857116, + 424, + 469.285714285714 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 159, + 469.285714285714, + 424, + 484.1428571428568 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 159, + 484.1428571428568, + 424, + 498.99999999999966 + ], + "spans": [], + "index": 27 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 132, + 511, + 475, + 524 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 130, + 510, + 477, + 526 + ], + "spans": [ + { + "bbox": [ + 130, + 510, + 477, + 526 + ], + "score": 1.0, + "content": "Figure 28: Example segmentations for computing segmentation-consistency (SC).", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + } + ], + "index": 20.75 + }, + { + "type": "text", + "bbox": [ + 107, + 569, + 430, + 599 + ], + "lines": [ + { + "bbox": [ + 107, + 566, + 432, + 585 + ], + "spans": [ + { + "bbox": [ + 107, + 570, + 125, + 581 + ], + "score": 0.88, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 566, + 213, + 585 + ], + "score": 1.0, + "content": "can be obtained from", + "type": "text" + }, + { + "bbox": [ + 214, + 570, + 223, + 582 + ], + "score": 0.86, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 223, + 566, + 244, + 585 + ], + "score": 1.0, + "content": ", and", + "type": "text" + }, + { + "bbox": [ + 244, + 569, + 280, + 581 + ], + "score": 0.89, + "content": "\\Delta h ^ { g l o b a l }", + "type": "inline_equation" + }, + { + "bbox": [ + 281, + 566, + 408, + 585 + ], + "score": 1.0, + "content": "can be obtained by aggregating", + "type": "text" + }, + { + "bbox": [ + 409, + 570, + 427, + 581 + ], + "score": 0.89, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 427, + 566, + 432, + 585 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 586, + 286, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 586, + 163, + 600 + ], + "score": 1.0, + "content": "We opt to use", + "type": "text" + }, + { + "bbox": [ + 163, + 587, + 173, + 599 + ], + "score": 0.87, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 173, + 586, + 286, + 600 + ], + "score": 1.0, + "content": "for above three advantages.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5 + }, + { + "type": "title", + "bbox": [ + 107, + 644, + 236, + 657 + ], + "lines": [ + { + "bbox": [ + 105, + 644, + 236, + 659 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 236, + 659 + ], + "score": 1.0, + "content": "M RANDOM SAMPLING", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "score": 1.0, + "content": "We conduct extra experiments: generating images with target attributes using Asyrp not from inver-", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "sion but from random Gaussian noises. As a consequence, the generative process can be used for", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "conditional random sampling. We provide the results in Figure 31. However, it is beyond the scope", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 719, + 161, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 161, + 734 + ], + "score": 1.0, + "content": "of this paper.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 33.5 + } + ], + "page_idx": 29, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 763 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 763 + ], + "score": 1.0, + "content": "30", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 159, + 83, + 424, + 499 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 159, + 83, + 424, + 499 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 159, + 83, + 424, + 499 + ], + "spans": [ + { + "bbox": [ + 159, + 83, + 424, + 499 + ], + "score": 0.952, + "type": "image", + "image_path": "4bcaa6683a66fe5dc47e975b0ea521a3cef4e214da2c1f2d293b268e362cc228.jpg" + } + ] + } + ], + "index": 13.5, + "virtual_lines": [ + { + "bbox": [ + 159, + 83, + 424, + 97.85714285714286 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 159, + 97.85714285714286, + 424, + 112.71428571428572 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 159, + 112.71428571428572, + 424, + 127.57142857142858 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 159, + 127.57142857142858, + 424, + 142.42857142857144 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 159, + 142.42857142857144, + 424, + 157.2857142857143 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 159, + 157.2857142857143, + 424, + 172.14285714285717 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 159, + 172.14285714285717, + 424, + 187.00000000000003 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 159, + 187.00000000000003, + 424, + 201.8571428571429 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 159, + 201.8571428571429, + 424, + 216.71428571428575 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 159, + 216.71428571428575, + 424, + 231.5714285714286 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 159, + 231.5714285714286, + 424, + 246.42857142857147 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 159, + 246.42857142857147, + 424, + 261.28571428571433 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 159, + 261.28571428571433, + 424, + 276.14285714285717 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 159, + 276.14285714285717, + 424, + 291.0 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 159, + 291.0, + 424, + 305.85714285714283 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 159, + 305.85714285714283, + 424, + 320.71428571428567 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 159, + 320.71428571428567, + 424, + 335.5714285714285 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 159, + 335.5714285714285, + 424, + 350.42857142857133 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 159, + 350.42857142857133, + 424, + 365.28571428571416 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 159, + 365.28571428571416, + 424, + 380.142857142857 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 159, + 380.142857142857, + 424, + 394.99999999999983 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 159, + 394.99999999999983, + 424, + 409.85714285714266 + ], + "spans": [], + "index": 21 + }, + { + "bbox": [ + 159, + 409.85714285714266, + 424, + 424.7142857142855 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 159, + 424.7142857142855, + 424, + 439.5714285714283 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 159, + 439.5714285714283, + 424, + 454.42857142857116 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 159, + 454.42857142857116, + 424, + 469.285714285714 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 159, + 469.285714285714, + 424, + 484.1428571428568 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 159, + 484.1428571428568, + 424, + 498.99999999999966 + ], + "spans": [], + "index": 27 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 132, + 511, + 475, + 524 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 130, + 510, + 477, + 526 + ], + "spans": [ + { + "bbox": [ + 130, + 510, + 477, + 526 + ], + "score": 1.0, + "content": "Figure 28: Example segmentations for computing segmentation-consistency (SC).", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + } + ], + "index": 20.75 + }, + { + "type": "list", + "bbox": [ + 107, + 569, + 430, + 599 + ], + "lines": [ + { + "bbox": [ + 107, + 566, + 432, + 585 + ], + "spans": [ + { + "bbox": [ + 107, + 570, + 125, + 581 + ], + "score": 0.88, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 566, + 213, + 585 + ], + "score": 1.0, + "content": "can be obtained from", + "type": "text" + }, + { + "bbox": [ + 214, + 570, + 223, + 582 + ], + "score": 0.86, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 223, + 566, + 244, + 585 + ], + "score": 1.0, + "content": ", and", + "type": "text" + }, + { + "bbox": [ + 244, + 569, + 280, + 581 + ], + "score": 0.89, + "content": "\\Delta h ^ { g l o b a l }", + "type": "inline_equation" + }, + { + "bbox": [ + 281, + 566, + 408, + 585 + ], + "score": 1.0, + "content": "can be obtained by aggregating", + "type": "text" + }, + { + "bbox": [ + 409, + 570, + 427, + 581 + ], + "score": 0.89, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 427, + 566, + 432, + 585 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 29, + "is_list_end_line": true + }, + { + "bbox": [ + 106, + 586, + 286, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 586, + 163, + 600 + ], + "score": 1.0, + "content": "We opt to use", + "type": "text" + }, + { + "bbox": [ + 163, + 587, + 173, + 599 + ], + "score": 0.87, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 173, + 586, + 286, + 600 + ], + "score": 1.0, + "content": "for above three advantages.", + "type": "text" + } + ], + "index": 30, + "is_list_start_line": true, + "is_list_end_line": true + } + ], + "index": 29.5, + "bbox_fs": [ + 106, + 566, + 432, + 600 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 644, + 236, + 657 + ], + "lines": [ + { + "bbox": [ + 105, + 644, + 236, + 659 + ], + "spans": [ + { + "bbox": [ + 105, + 644, + 236, + 659 + ], + "score": 1.0, + "content": "M RANDOM SAMPLING", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 701 + ], + "score": 1.0, + "content": "We conduct extra experiments: generating images with target attributes using Asyrp not from inver-", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "sion but from random Gaussian noises. As a consequence, the generative process can be used for", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "conditional random sampling. We provide the results in Figure 31. However, it is beyond the scope", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 719, + 161, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 161, + 734 + ], + "score": 1.0, + "content": "of this paper.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 33.5, + "bbox_fs": [ + 105, + 687, + 505, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 115, + 82, + 501, + 322 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 82, + 501, + 322 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 115, + 82, + 501, + 322 + ], + "spans": [ + { + "bbox": [ + 115, + 82, + 501, + 322 + ], + "score": 0.97, + "type": "image", + "image_path": "03e1b521e960c51738ecc1ae3c6697bcf45be2371c9a8ab319c4542be8d594bf.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 115, + 82, + 501, + 162.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 115, + 162.0, + 501, + 242.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 115, + 242.0, + 501, + 322.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 334, + 505, + 367 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 334, + 505, + 346 + ], + "spans": [ + { + "bbox": [ + 106, + 334, + 505, + 346 + ], + "score": 1.0, + "content": "Figure 29: We compute mean direction and global direction from 20 other images on CelebA-", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 345, + 505, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 345, + 419, + 358 + ], + "score": 1.0, + "content": "HQ. The effect of mean direction and global direction are quite similar with", + "type": "text" + }, + { + "bbox": [ + 419, + 346, + 437, + 357 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 438, + 345, + 452, + 358 + ], + "score": 1.0, + "content": "by", + "type": "text" + }, + { + "bbox": [ + 452, + 346, + 462, + 357 + ], + "score": 0.86, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 462, + 345, + 505, + 358 + ], + "score": 1.0, + "content": "at diverse", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 355, + 149, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 355, + 149, + 369 + ], + "score": 1.0, + "content": "attributes.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 108, + 380, + 502, + 530 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 380, + 502, + 530 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 108, + 380, + 502, + 530 + ], + "spans": [ + { + "bbox": [ + 108, + 380, + 502, + 530 + ], + "score": 0.96, + "type": "image", + "image_path": "94795dd3dafb5d37e50068f7124646c932ba91d53b723f68b82fcc14beab1bf8.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 108, + 380, + 502, + 430.0 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 108, + 430.0, + 502, + 480.0 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 108, + 480.0, + 502, + 530.0 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 544, + 505, + 577 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 544, + 504, + 556 + ], + "spans": [ + { + "bbox": [ + 106, + 544, + 504, + 556 + ], + "score": 1.0, + "content": "Figure 30: We obtain mean direction and global direction from 20 other images on LSUN-church.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 554, + 505, + 568 + ], + "spans": [ + { + "bbox": [ + 106, + 554, + 403, + 568 + ], + "score": 1.0, + "content": "The effect of mean direction and global direction are quite similar with", + "type": "text" + }, + { + "bbox": [ + 403, + 555, + 421, + 566 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 422, + 554, + 437, + 568 + ], + "score": 1.0, + "content": "by", + "type": "text" + }, + { + "bbox": [ + 437, + 556, + 447, + 567 + ], + "score": 0.86, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 448, + 554, + 505, + 568 + ], + "score": 1.0, + "content": "at diverse at-", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 565, + 142, + 579 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 142, + 579 + ], + "score": 1.0, + "content": "tributes.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + } + ], + "index": 8.5 + }, + { + "type": "image", + "bbox": [ + 108, + 591, + 503, + 666 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 591, + 503, + 666 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 108, + 591, + 503, + 666 + ], + "spans": [ + { + "bbox": [ + 108, + 591, + 503, + 666 + ], + "score": 0.958, + "type": "image", + "image_path": "6eb1c92e476116c392feb38047e039fb60ec3d76570c709ea5ec47a32b19d996.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 108, + 591, + 503, + 616.0 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 108, + 616.0, + 503, + 641.0 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 108, + 641.0, + 503, + 666.0 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 105, + 678, + 504, + 701 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 105, + 677, + 505, + 692 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 505, + 692 + ], + "score": 1.0, + "content": "Figure 31: Uncurated random sampling. We generate images from random noise with Asyrp.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 687, + 458, + 703 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 458, + 703 + ], + "score": 1.0, + "content": "Although we do not focus on these results, Asyrp can be used for conditional sampling.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5 + } + ], + "index": 14.25 + } + ], + "page_idx": 30, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 115, + 82, + 501, + 322 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 115, + 82, + 501, + 322 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 115, + 82, + 501, + 322 + ], + "spans": [ + { + "bbox": [ + 115, + 82, + 501, + 322 + ], + "score": 0.97, + "type": "image", + "image_path": "03e1b521e960c51738ecc1ae3c6697bcf45be2371c9a8ab319c4542be8d594bf.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 115, + 82, + 501, + 162.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 115, + 162.0, + 501, + 242.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 115, + 242.0, + 501, + 322.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 334, + 505, + 367 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 334, + 505, + 346 + ], + "spans": [ + { + "bbox": [ + 106, + 334, + 505, + 346 + ], + "score": 1.0, + "content": "Figure 29: We compute mean direction and global direction from 20 other images on CelebA-", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 345, + 505, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 345, + 419, + 358 + ], + "score": 1.0, + "content": "HQ. The effect of mean direction and global direction are quite similar with", + "type": "text" + }, + { + "bbox": [ + 419, + 346, + 437, + 357 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 438, + 345, + 452, + 358 + ], + "score": 1.0, + "content": "by", + "type": "text" + }, + { + "bbox": [ + 452, + 346, + 462, + 357 + ], + "score": 0.86, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 462, + 345, + 505, + 358 + ], + "score": 1.0, + "content": "at diverse", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 355, + 149, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 355, + 149, + 369 + ], + "score": 1.0, + "content": "attributes.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 108, + 380, + 502, + 530 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 380, + 502, + 530 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 108, + 380, + 502, + 530 + ], + "spans": [ + { + "bbox": [ + 108, + 380, + 502, + 530 + ], + "score": 0.96, + "type": "image", + "image_path": "94795dd3dafb5d37e50068f7124646c932ba91d53b723f68b82fcc14beab1bf8.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 108, + 380, + 502, + 430.0 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 108, + 430.0, + 502, + 480.0 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 108, + 480.0, + 502, + 530.0 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 544, + 505, + 577 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 544, + 504, + 556 + ], + "spans": [ + { + "bbox": [ + 106, + 544, + 504, + 556 + ], + "score": 1.0, + "content": "Figure 30: We obtain mean direction and global direction from 20 other images on LSUN-church.", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 554, + 505, + 568 + ], + "spans": [ + { + "bbox": [ + 106, + 554, + 403, + 568 + ], + "score": 1.0, + "content": "The effect of mean direction and global direction are quite similar with", + "type": "text" + }, + { + "bbox": [ + 403, + 555, + 421, + 566 + ], + "score": 0.9, + "content": "\\Delta h _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 422, + 554, + 437, + 568 + ], + "score": 1.0, + "content": "by", + "type": "text" + }, + { + "bbox": [ + 437, + 556, + 447, + 567 + ], + "score": 0.86, + "content": "f _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 448, + 554, + 505, + 568 + ], + "score": 1.0, + "content": "at diverse at-", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 565, + 142, + 579 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 142, + 579 + ], + "score": 1.0, + "content": "tributes.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + } + ], + "index": 8.5 + }, + { + "type": "image", + "bbox": [ + 108, + 591, + 503, + 666 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 108, + 591, + 503, + 666 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 108, + 591, + 503, + 666 + ], + "spans": [ + { + "bbox": [ + 108, + 591, + 503, + 666 + ], + "score": 0.958, + "type": "image", + "image_path": "6eb1c92e476116c392feb38047e039fb60ec3d76570c709ea5ec47a32b19d996.jpg" + } + ] + } + ], + "index": 13, + "virtual_lines": [ + { + "bbox": [ + 108, + 591, + 503, + 616.0 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 108, + 616.0, + 503, + 641.0 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 108, + 641.0, + 503, + 666.0 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 105, + 678, + 504, + 701 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 105, + 677, + 505, + 692 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 505, + 692 + ], + "score": 1.0, + "content": "Figure 31: Uncurated random sampling. We generate images from random noise with Asyrp.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 687, + 458, + 703 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 458, + 703 + ], + "score": 1.0, + "content": "Although we do not focus on these results, Asyrp can be used for conditional sampling.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5 + } + ], + "index": 14.25 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 81, + 211, + 94 + ], + "lines": [ + { + "bbox": [ + 105, + 80, + 213, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 213, + 95 + ], + "score": 1.0, + "content": "N MORE SAMPLES", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "title", + "bbox": [ + 107, + 106, + 182, + 118 + ], + "lines": [ + { + "bbox": [ + 105, + 105, + 183, + 119 + ], + "spans": [ + { + "bbox": [ + 105, + 105, + 183, + 119 + ], + "score": 1.0, + "content": "N.1 IMAGENET", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 107, + 126, + 506, + 192 + ], + "lines": [ + { + "bbox": [ + 106, + 127, + 505, + 139 + ], + "spans": [ + { + "bbox": [ + 106, + 127, + 505, + 139 + ], + "score": 1.0, + "content": "We conduct extra experiments: editing images with target class using Asyrp with ImageNet pre-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 137, + 505, + 149 + ], + "spans": [ + { + "bbox": [ + 106, + 137, + 505, + 149 + ], + "score": 1.0, + "content": "trained model. We verified that models trained on large datasets, such as ImageNet, can be edited", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 149, + 505, + 161 + ], + "spans": [ + { + "bbox": [ + 106, + 149, + 505, + 161 + ], + "score": 1.0, + "content": "using Asyrp. However, we also observed that in this case the latent space is not partitioned by classes.", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "score": 1.0, + "content": "For an orange, we have different latents for a single orange, for many oranges, for a cross-section of", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 169, + 506, + 185 + ], + "spans": [ + { + "bbox": [ + 105, + 169, + 506, + 185 + ], + "score": 1.0, + "content": "cut orange, and for a single piece of orange. Therefore, we learned the implicit function by collecting", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 182, + 250, + 194 + ], + "spans": [ + { + "bbox": [ + 106, + 182, + 250, + 194 + ], + "score": 1.0, + "content": "similar images to find the direction.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 4.5 + }, + { + "type": "image", + "bbox": [ + 159, + 205, + 452, + 263 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 159, + 205, + 452, + 263 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 159, + 205, + 452, + 263 + ], + "spans": [ + { + "bbox": [ + 159, + 205, + 452, + 263 + ], + "score": 0.962, + "type": "image", + "image_path": "774272204ff4ad42b215ddad0f822037c3a70d6b9b02276bd89a0e47fea88d21.jpg" + } + ] + } + ], + "index": 9, + "virtual_lines": [ + { + "bbox": [ + 159, + 205, + 452, + 224.33333333333334 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 159, + 224.33333333333334, + 452, + 243.66666666666669 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 159, + 243.66666666666669, + 452, + 263.0 + ], + "spans": [], + "index": 10 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 277, + 504, + 301 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "score": 1.0, + "content": "Figure 32: Result of Asyrp in ImageNet. The result shows that Asyrp works even in ImageNet", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 288, + 139, + 300 + ], + "spans": [ + { + "bbox": [ + 105, + 288, + 139, + 300 + ], + "score": 1.0, + "content": "dataset.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5 + } + ], + "index": 10.25 + }, + { + "type": "title", + "bbox": [ + 107, + 322, + 236, + 334 + ], + "lines": [ + { + "bbox": [ + 106, + 323, + 236, + 335 + ], + "spans": [ + { + "bbox": [ + 106, + 323, + 236, + 335 + ], + "score": 1.0, + "content": "N.2 MULTI-INTERPOLATION", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 107, + 343, + 503, + 366 + ], + "lines": [ + { + "bbox": [ + 106, + 343, + 504, + 356 + ], + "spans": [ + { + "bbox": [ + 106, + 343, + 504, + 356 + ], + "score": 1.0, + "content": "Figure 33 provides mixed interpolation between multiple attributes. We observe that any interpola-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 354, + 243, + 367 + ], + "spans": [ + { + "bbox": [ + 105, + 354, + 243, + 367 + ], + "score": 1.0, + "content": "tion with any attribute is possible.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5 + }, + { + "type": "title", + "bbox": [ + 108, + 379, + 280, + 390 + ], + "lines": [ + { + "bbox": [ + 105, + 378, + 281, + 392 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 281, + 392 + ], + "score": 1.0, + "content": "N.3 MORE RESULTS ON ALL DATASETS", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 106, + 399, + 504, + 423 + ], + "lines": [ + { + "bbox": [ + 105, + 398, + 505, + 412 + ], + "spans": [ + { + "bbox": [ + 105, + 398, + 505, + 412 + ], + "score": 1.0, + "content": "We provide more results on CelebA-HQ (Figure 34), LSUN-church (Figure 35), AFHQ, LSUN-", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 410, + 247, + 423 + ], + "spans": [ + { + "bbox": [ + 105, + 410, + 247, + 423 + ], + "score": 1.0, + "content": "bedroom, METFACES (Figure 36).", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5 + } + ], + "page_idx": 31, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 765 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 765 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 81, + 211, + 94 + ], + "lines": [ + { + "bbox": [ + 105, + 80, + 213, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 80, + 213, + 95 + ], + "score": 1.0, + "content": "N MORE SAMPLES", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "title", + "bbox": [ + 107, + 106, + 182, + 118 + ], + "lines": [ + { + "bbox": [ + 105, + 105, + 183, + 119 + ], + "spans": [ + { + "bbox": [ + 105, + 105, + 183, + 119 + ], + "score": 1.0, + "content": "N.1 IMAGENET", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 107, + 126, + 506, + 192 + ], + "lines": [ + { + "bbox": [ + 106, + 127, + 505, + 139 + ], + "spans": [ + { + "bbox": [ + 106, + 127, + 505, + 139 + ], + "score": 1.0, + "content": "We conduct extra experiments: editing images with target class using Asyrp with ImageNet pre-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 137, + 505, + 149 + ], + "spans": [ + { + "bbox": [ + 106, + 137, + 505, + 149 + ], + "score": 1.0, + "content": "trained model. We verified that models trained on large datasets, such as ImageNet, can be edited", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 149, + 505, + 161 + ], + "spans": [ + { + "bbox": [ + 106, + 149, + 505, + 161 + ], + "score": 1.0, + "content": "using Asyrp. However, we also observed that in this case the latent space is not partitioned by classes.", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 506, + 172 + ], + "score": 1.0, + "content": "For an orange, we have different latents for a single orange, for many oranges, for a cross-section of", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 169, + 506, + 185 + ], + "spans": [ + { + "bbox": [ + 105, + 169, + 506, + 185 + ], + "score": 1.0, + "content": "cut orange, and for a single piece of orange. Therefore, we learned the implicit function by collecting", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 182, + 250, + 194 + ], + "spans": [ + { + "bbox": [ + 106, + 182, + 250, + 194 + ], + "score": 1.0, + "content": "similar images to find the direction.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 4.5, + "bbox_fs": [ + 105, + 127, + 506, + 194 + ] + }, + { + "type": "image", + "bbox": [ + 159, + 205, + 452, + 263 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 159, + 205, + 452, + 263 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 159, + 205, + 452, + 263 + ], + "spans": [ + { + "bbox": [ + 159, + 205, + 452, + 263 + ], + "score": 0.962, + "type": "image", + "image_path": "774272204ff4ad42b215ddad0f822037c3a70d6b9b02276bd89a0e47fea88d21.jpg" + } + ] + } + ], + "index": 9, + "virtual_lines": [ + { + "bbox": [ + 159, + 205, + 452, + 224.33333333333334 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 159, + 224.33333333333334, + 452, + 243.66666666666669 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 159, + 243.66666666666669, + 452, + 263.0 + ], + "spans": [], + "index": 10 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 277, + 504, + 301 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "score": 1.0, + "content": "Figure 32: Result of Asyrp in ImageNet. The result shows that Asyrp works even in ImageNet", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 288, + 139, + 300 + ], + "spans": [ + { + "bbox": [ + 105, + 288, + 139, + 300 + ], + "score": 1.0, + "content": "dataset.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5 + } + ], + "index": 10.25 + }, + { + "type": "title", + "bbox": [ + 107, + 322, + 236, + 334 + ], + "lines": [ + { + "bbox": [ + 106, + 323, + 236, + 335 + ], + "spans": [ + { + "bbox": [ + 106, + 323, + 236, + 335 + ], + "score": 1.0, + "content": "N.2 MULTI-INTERPOLATION", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 107, + 343, + 503, + 366 + ], + "lines": [ + { + "bbox": [ + 106, + 343, + 504, + 356 + ], + "spans": [ + { + "bbox": [ + 106, + 343, + 504, + 356 + ], + "score": 1.0, + "content": "Figure 33 provides mixed interpolation between multiple attributes. We observe that any interpola-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 354, + 243, + 367 + ], + "spans": [ + { + "bbox": [ + 105, + 354, + 243, + 367 + ], + "score": 1.0, + "content": "tion with any attribute is possible.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5, + "bbox_fs": [ + 105, + 343, + 504, + 367 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 379, + 280, + 390 + ], + "lines": [ + { + "bbox": [ + 105, + 378, + 281, + 392 + ], + "spans": [ + { + "bbox": [ + 105, + 378, + 281, + 392 + ], + "score": 1.0, + "content": "N.3 MORE RESULTS ON ALL DATASETS", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 106, + 399, + 504, + 423 + ], + "lines": [ + { + "bbox": [ + 105, + 398, + 505, + 412 + ], + "spans": [ + { + "bbox": [ + 105, + 398, + 505, + 412 + ], + "score": 1.0, + "content": "We provide more results on CelebA-HQ (Figure 34), LSUN-church (Figure 35), AFHQ, LSUN-", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 410, + 247, + 423 + ], + "spans": [ + { + "bbox": [ + 105, + 410, + 247, + 423 + ], + "score": 1.0, + "content": "bedroom, METFACES (Figure 36).", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5, + "bbox_fs": [ + 105, + 398, + 505, + 423 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 162, + 92, + 446, + 711 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 162, + 92, + 446, + 711 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 162, + 92, + 446, + 711 + ], + "spans": [ + { + "bbox": [ + 162, + 92, + 446, + 711 + ], + "score": 0.875, + "type": "image", + "image_path": "b9bf01755a1cba0600b8b0a808d391a7b78c8c875350410552a6a08c44d09b8a.jpg" + } + ] + } + ], + "index": 20, + "virtual_lines": [ + { + "bbox": [ + 162, + 92, + 446, + 107.09756097560975 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 162, + 107.09756097560975, + 446, + 122.1951219512195 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 162, + 122.1951219512195, + 446, + 137.29268292682926 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 162, + 137.29268292682926, + 446, + 152.390243902439 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 162, + 152.390243902439, + 446, + 167.48780487804876 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 162, + 167.48780487804876, + 446, + 182.58536585365852 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 162, + 182.58536585365852, + 446, + 197.68292682926827 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 162, + 197.68292682926827, + 446, + 212.78048780487802 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 162, + 212.78048780487802, + 446, + 227.87804878048777 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 162, + 227.87804878048777, + 446, + 242.97560975609753 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 162, + 242.97560975609753, + 446, + 258.0731707317073 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 162, + 258.0731707317073, + 446, + 273.1707317073171 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 162, + 273.1707317073171, + 446, + 288.26829268292687 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 162, + 288.26829268292687, + 446, + 303.36585365853665 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 162, + 303.36585365853665, + 446, + 318.46341463414643 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 162, + 318.46341463414643, + 446, + 333.5609756097562 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 162, + 333.5609756097562, + 446, + 348.658536585366 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 162, + 348.658536585366, + 446, + 363.7560975609758 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 162, + 363.7560975609758, + 446, + 378.85365853658556 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 162, + 378.85365853658556, + 446, + 393.95121951219534 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 162, + 393.95121951219534, + 446, + 409.0487804878051 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 162, + 409.0487804878051, + 446, + 424.1463414634149 + ], + "spans": [], + "index": 21 + }, + { + "bbox": [ + 162, + 424.1463414634149, + 446, + 439.2439024390247 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 162, + 439.2439024390247, + 446, + 454.34146341463446 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 162, + 454.34146341463446, + 446, + 469.43902439024424 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 162, + 469.43902439024424, + 446, + 484.536585365854 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 162, + 484.536585365854, + 446, + 499.6341463414638 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 162, + 499.6341463414638, + 446, + 514.7317073170735 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 162, + 514.7317073170735, + 446, + 529.8292682926833 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 162, + 529.8292682926833, + 446, + 544.9268292682931 + ], + "spans": [], + "index": 29 + }, + { + "bbox": [ + 162, + 544.9268292682931, + 446, + 560.0243902439029 + ], + "spans": [], + "index": 30 + }, + { + "bbox": [ + 162, + 560.0243902439029, + 446, + 575.1219512195127 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 162, + 575.1219512195127, + 446, + 590.2195121951224 + ], + "spans": [], + "index": 32 + }, + { + "bbox": [ + 162, + 590.2195121951224, + 446, + 605.3170731707322 + ], + "spans": [], + "index": 33 + }, + { + "bbox": [ + 162, + 605.3170731707322, + 446, + 620.414634146342 + ], + "spans": [], + "index": 34 + }, + { + "bbox": [ + 162, + 620.414634146342, + 446, + 635.5121951219518 + ], + "spans": [], + "index": 35 + }, + { + "bbox": [ + 162, + 635.5121951219518, + 446, + 650.6097560975616 + ], + "spans": [], + "index": 36 + }, + { + "bbox": [ + 162, + 650.6097560975616, + 446, + 665.7073170731713 + ], + "spans": [], + "index": 37 + }, + { + "bbox": [ + 162, + 665.7073170731713, + 446, + 680.8048780487811 + ], + "spans": [], + "index": 38 + }, + { + "bbox": [ + 162, + 680.8048780487811, + 446, + 695.9024390243909 + ], + "spans": [], + "index": 39 + }, + { + "bbox": [ + 162, + 695.9024390243909, + 446, + 711.0000000000007 + ], + "spans": [], + "index": 40 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 694, + 504, + 717 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 704, + 230, + 719 + ], + "spans": [ + { + "bbox": [ + 107, + 705, + 122, + 716 + ], + "score": 0.57, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 122, + 704, + 230, + 719 + ], + "score": 1.0, + "content": "of smiling and young.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 41 + } + ], + "index": 30.5 + } + ], + "page_idx": 32, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 312, + 763 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 312, + 763 + ], + "score": 1.0, + "content": "33", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 162, + 92, + 446, + 711 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 162, + 92, + 446, + 711 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 162, + 92, + 446, + 711 + ], + "spans": [ + { + "bbox": [ + 162, + 92, + 446, + 711 + ], + "score": 0.875, + "type": "image", + "image_path": "b9bf01755a1cba0600b8b0a808d391a7b78c8c875350410552a6a08c44d09b8a.jpg" + } + ] + } + ], + "index": 20, + "virtual_lines": [ + { + "bbox": [ + 162, + 92, + 446, + 107.09756097560975 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 162, + 107.09756097560975, + 446, + 122.1951219512195 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 162, + 122.1951219512195, + 446, + 137.29268292682926 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 162, + 137.29268292682926, + 446, + 152.390243902439 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 162, + 152.390243902439, + 446, + 167.48780487804876 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 162, + 167.48780487804876, + 446, + 182.58536585365852 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 162, + 182.58536585365852, + 446, + 197.68292682926827 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 162, + 197.68292682926827, + 446, + 212.78048780487802 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 162, + 212.78048780487802, + 446, + 227.87804878048777 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 162, + 227.87804878048777, + 446, + 242.97560975609753 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 162, + 242.97560975609753, + 446, + 258.0731707317073 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 162, + 258.0731707317073, + 446, + 273.1707317073171 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 162, + 273.1707317073171, + 446, + 288.26829268292687 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 162, + 288.26829268292687, + 446, + 303.36585365853665 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 162, + 303.36585365853665, + 446, + 318.46341463414643 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 162, + 318.46341463414643, + 446, + 333.5609756097562 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 162, + 333.5609756097562, + 446, + 348.658536585366 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 162, + 348.658536585366, + 446, + 363.7560975609758 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 162, + 363.7560975609758, + 446, + 378.85365853658556 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 162, + 378.85365853658556, + 446, + 393.95121951219534 + ], + "spans": [], + "index": 19 + }, + { + "bbox": [ + 162, + 393.95121951219534, + 446, + 409.0487804878051 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 162, + 409.0487804878051, + 446, + 424.1463414634149 + ], + "spans": [], + "index": 21 + }, + { + "bbox": [ + 162, + 424.1463414634149, + 446, + 439.2439024390247 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 162, + 439.2439024390247, + 446, + 454.34146341463446 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 162, + 454.34146341463446, + 446, + 469.43902439024424 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 162, + 469.43902439024424, + 446, + 484.536585365854 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 162, + 484.536585365854, + 446, + 499.6341463414638 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 162, + 499.6341463414638, + 446, + 514.7317073170735 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 162, + 514.7317073170735, + 446, + 529.8292682926833 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 162, + 529.8292682926833, + 446, + 544.9268292682931 + ], + "spans": [], + "index": 29 + }, + { + "bbox": [ + 162, + 544.9268292682931, + 446, + 560.0243902439029 + ], + "spans": [], + "index": 30 + }, + { + "bbox": [ + 162, + 560.0243902439029, + 446, + 575.1219512195127 + ], + "spans": [], + "index": 31 + }, + { + "bbox": [ + 162, + 575.1219512195127, + 446, + 590.2195121951224 + ], + "spans": [], + "index": 32 + }, + { + "bbox": [ + 162, + 590.2195121951224, + 446, + 605.3170731707322 + ], + "spans": [], + "index": 33 + }, + { + "bbox": [ + 162, + 605.3170731707322, + 446, + 620.414634146342 + ], + "spans": [], + "index": 34 + }, + { + "bbox": [ + 162, + 620.414634146342, + 446, + 635.5121951219518 + ], + "spans": [], + "index": 35 + }, + { + "bbox": [ + 162, + 635.5121951219518, + 446, + 650.6097560975616 + ], + "spans": [], + "index": 36 + }, + { + "bbox": [ + 162, + 650.6097560975616, + 446, + 665.7073170731713 + ], + "spans": [], + "index": 37 + }, + { + "bbox": [ + 162, + 665.7073170731713, + 446, + 680.8048780487811 + ], + "spans": [], + "index": 38 + }, + { + "bbox": [ + 162, + 680.8048780487811, + 446, + 695.9024390243909 + ], + "spans": [], + "index": 39 + }, + { + "bbox": [ + 162, + 695.9024390243909, + 446, + 711.0000000000007 + ], + "spans": [], + "index": 40 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 694, + 504, + 717 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 704, + 230, + 719 + ], + "spans": [ + { + "bbox": [ + 107, + 705, + 122, + 716 + ], + "score": 0.57, + "content": "\\Delta h", + "type": "inline_equation" + }, + { + "bbox": [ + 122, + 704, + 230, + 719 + ], + "score": 1.0, + "content": "of smiling and young.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 41 + } + ], + "index": 30.5 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 169, + 504, + 385 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 169, + 504, + 385 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 169, + 504, + 385 + ], + "spans": [ + { + "bbox": [ + 106, + 169, + 504, + 385 + ], + "score": 0.942, + "type": "image", + "image_path": "e936bdacea2d549661e4189967aeb6d92e07c0baab9c7e9b2fce8aea9a2adbf3.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 169, + 504, + 241.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 241.0, + 504, + 313.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 313.0, + 504, + 385.0 + ], + "spans": [], + "index": 2 + } + ] + } + ], + "index": 1 + }, + { + "type": "image", + "bbox": [ + 107, + 394, + 504, + 609 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 394, + 504, + 609 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 394, + 504, + 609 + ], + "spans": [ + { + "bbox": [ + 107, + 394, + 504, + 609 + ], + "score": 0.947, + "type": "image", + "image_path": "377b9520a7229fe39d63b39397e4df0928ea29df4c7531e407e9ec506cc84359.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 107, + 394, + 504, + 465.6666666666667 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 107, + 465.6666666666667, + 504, + 537.3333333333334 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 107, + 537.3333333333334, + 504, + 609.0 + ], + "spans": [], + "index": 5 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 184, + 626, + 424, + 639 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 183, + 625, + 424, + 641 + ], + "spans": [ + { + "bbox": [ + 183, + 625, + 424, + 641 + ], + "score": 1.0, + "content": "Figure 34: We provide more results on CelebA-HQ dataset.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + } + ], + "index": 5.0 + } + ], + "page_idx": 33, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 169, + 504, + 385 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 169, + 504, + 385 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 169, + 504, + 385 + ], + "spans": [ + { + "bbox": [ + 106, + 169, + 504, + 385 + ], + "score": 0.942, + "type": "image", + "image_path": "e936bdacea2d549661e4189967aeb6d92e07c0baab9c7e9b2fce8aea9a2adbf3.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 169, + 504, + 241.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 241.0, + 504, + 313.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 313.0, + 504, + 385.0 + ], + "spans": [], + "index": 2 + } + ] + } + ], + "index": 1 + }, + { + "type": "image", + "bbox": [ + 107, + 394, + 504, + 609 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 394, + 504, + 609 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 394, + 504, + 609 + ], + "spans": [ + { + "bbox": [ + 107, + 394, + 504, + 609 + ], + "score": 0.947, + "type": "image", + "image_path": "377b9520a7229fe39d63b39397e4df0928ea29df4c7531e407e9ec506cc84359.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 107, + 394, + 504, + 465.6666666666667 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 107, + 465.6666666666667, + 504, + 537.3333333333334 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 107, + 537.3333333333334, + 504, + 609.0 + ], + "spans": [], + "index": 5 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 184, + 626, + 424, + 639 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 183, + 625, + 424, + 641 + ], + "spans": [ + { + "bbox": [ + 183, + 625, + 424, + 641 + ], + "score": 1.0, + "content": "Figure 34: We provide more results on CelebA-HQ dataset.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + } + ], + "index": 5.0 + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 106, + 502, + 366 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 106, + 502, + 366 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 106, + 502, + 366 + ], + "spans": [ + { + "bbox": [ + 107, + 106, + 502, + 366 + ], + "score": 0.974, + "type": "image", + "image_path": "23c8ad00b208a9a83c86351a6572870e2502164c093437e52becf10d8cf7b5ae.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 106, + 502, + 192.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 192.66666666666669, + 502, + 279.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 279.33333333333337, + 502, + 366.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 180, + 379, + 427, + 391 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 180, + 378, + 429, + 393 + ], + "spans": [ + { + "bbox": [ + 180, + 378, + 429, + 393 + ], + "score": 1.0, + "content": "Figure 35: We provide more results on LSUN-church dataset.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 3 + } + ], + "index": 2.0 + }, + { + "type": "image", + "bbox": [ + 116, + 453, + 497, + 666 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 116, + 453, + 497, + 666 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 116, + 453, + 497, + 666 + ], + "spans": [ + { + "bbox": [ + 116, + 453, + 497, + 666 + ], + "score": 0.971, + "type": "image", + "image_path": "01bb6a5ce5257428dc08aae084bd10e960340eda2d2141d0c31b80453e60a4df.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 116, + 453, + 497, + 524.0 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 116, + 524.0, + 497, + 595.0 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 116, + 595.0, + 497, + 666.0 + ], + "spans": [], + "index": 6 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 680, + 504, + 704 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 104, + 679, + 505, + 694 + ], + "spans": [ + { + "bbox": [ + 104, + 679, + 505, + 694 + ], + "score": 1.0, + "content": "Figure 36: We provide more results on AFHQ, LSUN-bedroom and METFACES datasets respec-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 104, + 690, + 135, + 706 + ], + "spans": [ + { + "bbox": [ + 104, + 690, + 135, + 706 + ], + "score": 1.0, + "content": "tively.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + } + ], + "index": 6.25 + } + ], + "page_idx": 34, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 293, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 293, + 38 + ], + "score": 1.0, + "content": "Published as a conference paper at ICLR 2023", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 107, + 106, + 502, + 366 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 106, + 502, + 366 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 106, + 502, + 366 + ], + "spans": [ + { + "bbox": [ + 107, + 106, + 502, + 366 + ], + "score": 0.974, + "type": "image", + "image_path": "23c8ad00b208a9a83c86351a6572870e2502164c093437e52becf10d8cf7b5ae.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 107, + 106, + 502, + 192.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 107, + 192.66666666666669, + 502, + 279.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 279.33333333333337, + 502, + 366.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 180, + 379, + 427, + 391 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 180, + 378, + 429, + 393 + ], + "spans": [ + { + "bbox": [ + 180, + 378, + 429, + 393 + ], + "score": 1.0, + "content": "Figure 35: We provide more results on LSUN-church dataset.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 3 + } + ], + "index": 2.0 + }, + { + "type": "image", + "bbox": [ + 116, + 453, + 497, + 666 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 116, + 453, + 497, + 666 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 116, + 453, + 497, + 666 + ], + "spans": [ + { + "bbox": [ + 116, + 453, + 497, + 666 + ], + "score": 0.971, + "type": "image", + "image_path": "01bb6a5ce5257428dc08aae084bd10e960340eda2d2141d0c31b80453e60a4df.jpg" + } + ] + } + ], + "index": 5, + "virtual_lines": [ + { + "bbox": [ + 116, + 453, + 497, + 524.0 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 116, + 524.0, + 497, + 595.0 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 116, + 595.0, + 497, + 666.0 + ], + "spans": [], + "index": 6 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 680, + 504, + 704 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 104, + 679, + 505, + 694 + ], + "spans": [ + { + "bbox": [ + 104, + 679, + 505, + 694 + ], + "score": 1.0, + "content": "Figure 36: We provide more results on AFHQ, LSUN-bedroom and METFACES datasets respec-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 104, + 690, + 135, + 706 + ], + "spans": [ + { + "bbox": [ + 104, + 690, + 135, + 706 + ], + "score": 1.0, + "content": "tively.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + } + ], + "index": 6.25 + } + ] + } + ], + "_backend": "pipeline", + "_version_name": "2.2.2" +} \ No newline at end of file diff --git a/parse/dev/ptbePrczhRt/images/02d730e18fc781423258f212000f172ebab663d2c7bff3dea87e7f7272aada70.jpg b/parse/dev/ptbePrczhRt/images/02d730e18fc781423258f212000f172ebab663d2c7bff3dea87e7f7272aada70.jpg new file mode 100644 index 0000000000000000000000000000000000000000..335b520955cf5cc0375a3aeb6fada32e6aa7d3f2 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/02d730e18fc781423258f212000f172ebab663d2c7bff3dea87e7f7272aada70.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bc8f1b57771e0788545f55d6b80a7f6f2e8a3a3b6c64ccaec30333a6e82ee68b +size 5694 diff --git a/parse/dev/ptbePrczhRt/images/04f1922acd4aeb6ae8534c4f7304cb1c2798c6b34b92a5b8f8751e5892f0ef19.jpg b/parse/dev/ptbePrczhRt/images/04f1922acd4aeb6ae8534c4f7304cb1c2798c6b34b92a5b8f8751e5892f0ef19.jpg new file mode 100644 index 0000000000000000000000000000000000000000..31f71435f93cc0101396ac9821cc33f58ff0e349 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/04f1922acd4aeb6ae8534c4f7304cb1c2798c6b34b92a5b8f8751e5892f0ef19.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6fa61ba3fca20dac53b7f30028add939d7d65f12813581ba46944dcd5864ed79 +size 23745 diff --git a/parse/dev/ptbePrczhRt/images/190c4acb644673ed1399c0b8e6129a230a6fa94a6c06e0fc3be37554a99bac1c.jpg b/parse/dev/ptbePrczhRt/images/190c4acb644673ed1399c0b8e6129a230a6fa94a6c06e0fc3be37554a99bac1c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9d4b185f550a2a6dcba8d9ef5edd183103828e2f --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/190c4acb644673ed1399c0b8e6129a230a6fa94a6c06e0fc3be37554a99bac1c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:680b390a5173ab518c36a31172bdad1999d71f72b70661c79ad4046949eb125d +size 22527 diff --git a/parse/dev/ptbePrczhRt/images/486b95c9fb62408c391bb4c2ddddcedb7dae55ccf5c8ea49fdbdade56fc3e41a.jpg b/parse/dev/ptbePrczhRt/images/486b95c9fb62408c391bb4c2ddddcedb7dae55ccf5c8ea49fdbdade56fc3e41a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..97ac0e7586fae4cb1ad445492da1d9483a2b9f71 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/486b95c9fb62408c391bb4c2ddddcedb7dae55ccf5c8ea49fdbdade56fc3e41a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3469dfd3572d92a79815882a56cd993b466b8149fe7487373cbb5c08f81a37c5 +size 98623 diff --git a/parse/dev/ptbePrczhRt/images/4876b40230a538e668b3c7cbd2ebde43a847873dec4b796661c14ff95013c0b6.jpg b/parse/dev/ptbePrczhRt/images/4876b40230a538e668b3c7cbd2ebde43a847873dec4b796661c14ff95013c0b6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cdff0771ea3d0630873f2502d27e7cd8580f01bb --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/4876b40230a538e668b3c7cbd2ebde43a847873dec4b796661c14ff95013c0b6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:555f79dd0c4d62daf6f94f3dbc939c8664c611dc1be83abf10ceb5e6a714b320 +size 16056 diff --git a/parse/dev/ptbePrczhRt/images/4f2c7cf0477e3f2822fce351f39b8feea65b926f2d434c544bc8b46686defc8a.jpg b/parse/dev/ptbePrczhRt/images/4f2c7cf0477e3f2822fce351f39b8feea65b926f2d434c544bc8b46686defc8a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..094877140fe11662e0c9b48b1736c39d2f4c2bce --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/4f2c7cf0477e3f2822fce351f39b8feea65b926f2d434c544bc8b46686defc8a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc5255cc945a670bc948052fe93b75d47af8e297f27aba4e53304d665b0409c2 +size 6381 diff --git a/parse/dev/ptbePrczhRt/images/52de95440a92256f2bbe60a615fe59cb90bd3502400fbab178a243c641dc1c30.jpg b/parse/dev/ptbePrczhRt/images/52de95440a92256f2bbe60a615fe59cb90bd3502400fbab178a243c641dc1c30.jpg new file mode 100644 index 0000000000000000000000000000000000000000..30974f17a2d9754f9457bd26167e65152bd62d59 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/52de95440a92256f2bbe60a615fe59cb90bd3502400fbab178a243c641dc1c30.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:84455eabddf620b82a6f7eef34f5e66d72b67571279b702a0ebd518ea4ca220c +size 20832 diff --git a/parse/dev/ptbePrczhRt/images/60cd095a36b0b05186f18be02fe946bc3b66ded119d5e4e33090acb9cfa1c933.jpg b/parse/dev/ptbePrczhRt/images/60cd095a36b0b05186f18be02fe946bc3b66ded119d5e4e33090acb9cfa1c933.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0be01e7e0c340cef65d7995ce88c4f5f4504967c --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/60cd095a36b0b05186f18be02fe946bc3b66ded119d5e4e33090acb9cfa1c933.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:68a3c83ef7279a1dd2eb7b8d3eac7ebe75b8e826d6446bdc6edaeb8ded9755ab +size 56218 diff --git a/parse/dev/ptbePrczhRt/images/75d6e874bbb1df35b3e6ca039f345099875f0c7f97f8c8effe15ffbb05a3ae8d.jpg b/parse/dev/ptbePrczhRt/images/75d6e874bbb1df35b3e6ca039f345099875f0c7f97f8c8effe15ffbb05a3ae8d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9b7cb26f2356cc8ca988bd03e7dc971a4df2b7ba --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/75d6e874bbb1df35b3e6ca039f345099875f0c7f97f8c8effe15ffbb05a3ae8d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc192194abb247bfd25290aa3cd6b696c742695557061f8965472db77f9380a5 +size 31651 diff --git a/parse/dev/ptbePrczhRt/images/81e70e9281e2977f9e5b92dd4cf5368ad7429753b260ab47663881daefce956d.jpg b/parse/dev/ptbePrczhRt/images/81e70e9281e2977f9e5b92dd4cf5368ad7429753b260ab47663881daefce956d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d46958343e5fc49d551778c533e7d67d63f4dbf0 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/81e70e9281e2977f9e5b92dd4cf5368ad7429753b260ab47663881daefce956d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3f182bdf91c565b4cffc1010b94b79d120de54a7a57f63439f437451fb4c50bb +size 2932 diff --git a/parse/dev/ptbePrczhRt/images/99bf0d4cca3a10ab3278fcaf6e24d64aaa376387d392dd87f32c21ff9bbd092a.jpg b/parse/dev/ptbePrczhRt/images/99bf0d4cca3a10ab3278fcaf6e24d64aaa376387d392dd87f32c21ff9bbd092a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cd8f63cea2c2b133023cdf77235c39d4840c02ac --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/99bf0d4cca3a10ab3278fcaf6e24d64aaa376387d392dd87f32c21ff9bbd092a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:553c6d79f3b33c90deef23acffc60a9d01f96cddf94fa706842c6d2b2536951a +size 2782 diff --git a/parse/dev/ptbePrczhRt/images/9ecdad3b6fb531e6ef4e60a84f7680f1b112fd0215b98ac5788044896d809b2d.jpg b/parse/dev/ptbePrczhRt/images/9ecdad3b6fb531e6ef4e60a84f7680f1b112fd0215b98ac5788044896d809b2d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..aba2af8128f78fabcf7368fc1e14b846492beb80 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/9ecdad3b6fb531e6ef4e60a84f7680f1b112fd0215b98ac5788044896d809b2d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:854b38b0aecb6331847dcae3aa2bbb368ba589ed92d7a9669abfdbbff140d106 +size 22323 diff --git a/parse/dev/ptbePrczhRt/images/a7b7a7c382a8bd9b23abbdced6cdc33aac7d5e36684acf2ea3018aeb3a7eca1f.jpg b/parse/dev/ptbePrczhRt/images/a7b7a7c382a8bd9b23abbdced6cdc33aac7d5e36684acf2ea3018aeb3a7eca1f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..745b53f1e37fbe4caa9feb852626bc834cca69c1 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/a7b7a7c382a8bd9b23abbdced6cdc33aac7d5e36684acf2ea3018aeb3a7eca1f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d13ffb575b3df7508eb1550437b656f4fc3cebb5f02c5092b6429777d930d15 +size 5946 diff --git a/parse/dev/ptbePrczhRt/images/a904152358629d821cf19f8635f8db9545e953a3ef23fa37800693409a6be337.jpg b/parse/dev/ptbePrczhRt/images/a904152358629d821cf19f8635f8db9545e953a3ef23fa37800693409a6be337.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a0ee622ef7ad40f01ffe3a842e5a6344011e07dc --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/a904152358629d821cf19f8635f8db9545e953a3ef23fa37800693409a6be337.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b720115e36b2fadf7ab67f80fff281a819e93d75d7c24ef0eb84c3cd1f2783f +size 55568 diff --git a/parse/dev/ptbePrczhRt/images/cba77e657c623d3cea9bf9ed955e7538b770196402d28e694bbbced728ff0da2.jpg b/parse/dev/ptbePrczhRt/images/cba77e657c623d3cea9bf9ed955e7538b770196402d28e694bbbced728ff0da2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8e509d31bf24eab202d247f42002da92e5a3893e --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/cba77e657c623d3cea9bf9ed955e7538b770196402d28e694bbbced728ff0da2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af8ac4a273f41254e848b01211942c6bb464a0aa45c169f682e81665945d65fd +size 22237 diff --git a/parse/dev/ptbePrczhRt/images/d7919433f00d4ff1559497170c573a432e8da7eb7944b07f73ea6d3292e9b6df.jpg b/parse/dev/ptbePrczhRt/images/d7919433f00d4ff1559497170c573a432e8da7eb7944b07f73ea6d3292e9b6df.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7fb5ece55cc9b3a0bcadf9e5a0ca5385b5bcab99 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/d7919433f00d4ff1559497170c573a432e8da7eb7944b07f73ea6d3292e9b6df.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f03279dca91e86fe78e6bdeefb80c6ab26567b8006169dd32c3c7fbf78e68c0 +size 73110 diff --git a/parse/dev/ptbePrczhRt/images/dd6e6b80ca895bde5440cc2306b0796e690dafe2c85d8cccc90f0b24b710f6b4.jpg b/parse/dev/ptbePrczhRt/images/dd6e6b80ca895bde5440cc2306b0796e690dafe2c85d8cccc90f0b24b710f6b4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c9e5fd3a850738354b7aeb1c9ee9645dab210487 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/dd6e6b80ca895bde5440cc2306b0796e690dafe2c85d8cccc90f0b24b710f6b4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87faaa13a82fe0e91b3cace7cc4530b7627fd5ef3df0545a056f5f3117068f16 +size 168771 diff --git a/parse/dev/ptbePrczhRt/images/e47a56c5e91371513c49aa17762025bfcfc53b2d89bc114a88503f564b606c50.jpg b/parse/dev/ptbePrczhRt/images/e47a56c5e91371513c49aa17762025bfcfc53b2d89bc114a88503f564b606c50.jpg new file mode 100644 index 0000000000000000000000000000000000000000..da2a3caf17d53b05c3fb5549d707d2a4bc8a6c85 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/e47a56c5e91371513c49aa17762025bfcfc53b2d89bc114a88503f564b606c50.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e850ff4c39a6a9ba0d71c7f6e243c0e92daeae1c5c1d17dc0deac3ae495c11b +size 104161 diff --git a/parse/dev/ptbePrczhRt/images/f2b22add9b2978b99ed5b9807e3c5e99eaf6487c80ab95b9281d677da1556879.jpg b/parse/dev/ptbePrczhRt/images/f2b22add9b2978b99ed5b9807e3c5e99eaf6487c80ab95b9281d677da1556879.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0ddaf049bd357c4cdeef8637c2abe800e11a0b47 --- /dev/null +++ b/parse/dev/ptbePrczhRt/images/f2b22add9b2978b99ed5b9807e3c5e99eaf6487c80ab95b9281d677da1556879.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f74f95062caafc4b0f1af3df4ffef92b9c10d6515ca9906b3004f7e0dcaa1e7 +size 15941 diff --git a/parse/dev/q0nmYciuuZN/images/17a837ab5526c5f7aa07495259d3aeb8824a22629f761f08501aa45a4bb3eafd.jpg b/parse/dev/q0nmYciuuZN/images/17a837ab5526c5f7aa07495259d3aeb8824a22629f761f08501aa45a4bb3eafd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8de0933df4c94e018cd6067d50d03992ef1c650c --- /dev/null +++ b/parse/dev/q0nmYciuuZN/images/17a837ab5526c5f7aa07495259d3aeb8824a22629f761f08501aa45a4bb3eafd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a36e58e85daeb807d4346799662e5f649016a36dd277f8d474db62a749228050 +size 10313 diff --git a/parse/dev/q0nmYciuuZN/images/4078c9eecf86fe1485868510ad2c608724e4c9c5423123dfd3172bf5f10cef43.jpg b/parse/dev/q0nmYciuuZN/images/4078c9eecf86fe1485868510ad2c608724e4c9c5423123dfd3172bf5f10cef43.jpg new file mode 100644 index 0000000000000000000000000000000000000000..defc62ee26c1ccf90f307051c2a409c4c087b912 --- /dev/null +++ b/parse/dev/q0nmYciuuZN/images/4078c9eecf86fe1485868510ad2c608724e4c9c5423123dfd3172bf5f10cef43.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:caa656b9d74a104bdaf4ccede50418bdbae5555c03c2fada6414430e6561f206 +size 11374 diff --git a/parse/dev/q0nmYciuuZN/images/45d356e78d43f516c46c98386617ff868143eb4b5ded2f4bddd250e0b61816f7.jpg b/parse/dev/q0nmYciuuZN/images/45d356e78d43f516c46c98386617ff868143eb4b5ded2f4bddd250e0b61816f7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..992478962668140e83135130bf39bb1b72c6ead9 --- /dev/null +++ b/parse/dev/q0nmYciuuZN/images/45d356e78d43f516c46c98386617ff868143eb4b5ded2f4bddd250e0b61816f7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b88169df0a96c81e4f09eb4b758b3491463a2ab92d6b01e5d34e42b48a394a0 +size 44976 diff --git a/parse/dev/q0nmYciuuZN/images/87a0594f4a0c739edbe8cdf6c948dc006e2e3dc4883678fa5e57f93909c5a1a1.jpg b/parse/dev/q0nmYciuuZN/images/87a0594f4a0c739edbe8cdf6c948dc006e2e3dc4883678fa5e57f93909c5a1a1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..78561d45d744c07f96c248e856a0f05d41234411 --- /dev/null +++ b/parse/dev/q0nmYciuuZN/images/87a0594f4a0c739edbe8cdf6c948dc006e2e3dc4883678fa5e57f93909c5a1a1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b21332e4997e356c86a5b55f896b92c652e66e014e36d6e3db64cb0f5a7c46be +size 15737 diff --git a/parse/dev/q0nmYciuuZN/images/b431908e9cb9e6694fb6da8b93c9d8b41ea4263723f75bc1dd54297907baf201.jpg b/parse/dev/q0nmYciuuZN/images/b431908e9cb9e6694fb6da8b93c9d8b41ea4263723f75bc1dd54297907baf201.jpg new file mode 100644 index 0000000000000000000000000000000000000000..37d1c52508d2706d62db83bb223d4f56e8985f1c --- /dev/null +++ b/parse/dev/q0nmYciuuZN/images/b431908e9cb9e6694fb6da8b93c9d8b41ea4263723f75bc1dd54297907baf201.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2b1e7ba5b24d8403efc98878e90bacc46f1450beb120e615ecf8c27f47ecab0e +size 101243 diff --git a/parse/dev/q0nmYciuuZN/images/e643869e68389e735e1170957ebb1437fef959d85d8aa5c6b138dd3e1f9aa3e2.jpg b/parse/dev/q0nmYciuuZN/images/e643869e68389e735e1170957ebb1437fef959d85d8aa5c6b138dd3e1f9aa3e2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fa03e133a8b182ea5c1f0b00320b331c5ea8435c --- /dev/null +++ b/parse/dev/q0nmYciuuZN/images/e643869e68389e735e1170957ebb1437fef959d85d8aa5c6b138dd3e1f9aa3e2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:132e0b6e1d453334b73881a2068b737ea39acdb05794d119933e1badb797204e +size 10082 diff --git a/parse/dev/qHGCH75usg/qHGCH75usg_model.json b/parse/dev/qHGCH75usg/qHGCH75usg_model.json new file mode 100644 index 0000000000000000000000000000000000000000..d8d2cc554fdb5fbf1035facaae4147fb959cba56 --- /dev/null +++ b/parse/dev/qHGCH75usg/qHGCH75usg_model.json @@ -0,0 +1,20692 @@ +[ + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 398, + 1093, + 1304, + 1093, + 1304, + 1459, + 398, + 1459 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 1589, + 1404, + 1589, + 1404, + 1925, + 298, + 1925 + ], + "score": 0.978 + }, + { + "category_id": 0, + "poly": [ + 434, + 271, + 1266, + 271, + 1266, + 379, + 434, + 379 + ], + "score": 0.934 + }, + { + "category_id": 1, + "poly": [ + 788, + 494, + 1026, + 494, + 1026, + 557, + 788, + 557 + ], + "score": 0.914 + }, + { + "category_id": 1, + "poly": [ + 383, + 494, + 689, + 494, + 689, + 588, + 383, + 588 + ], + "score": 0.913 + }, + { + "category_id": 1, + "poly": [ + 460, + 851, + 824, + 851, + 824, + 944, + 460, + 944 + ], + "score": 0.908 + }, + { + "category_id": 0, + "poly": [ + 300, + 1518, + 530, + 1518, + 530, + 1556, + 300, + 1556 + ], + "score": 0.905 + }, + { + "category_id": 1, + "poly": [ + 683, + 634, + 871, + 634, + 871, + 695, + 683, + 695 + ], + "score": 0.891 + }, + { + "category_id": 0, + "poly": [ + 787, + 1022, + 913, + 1022, + 913, + 1058, + 787, + 1058 + ], + "score": 0.889 + }, + { + "category_id": 2, + "poly": [ + 297, + 2033, + 1071, + 2033, + 1071, + 2062, + 297, + 2062 + ], + "score": 0.888 + }, + { + "category_id": 1, + "poly": [ + 1129, + 494, + 1310, + 494, + 1310, + 557, + 1129, + 557 + ], + "score": 0.887 + }, + { + "category_id": 2, + "poly": [ + 334, + 1951, + 536, + 1951, + 536, + 1978, + 334, + 1978 + ], + "score": 0.882 + }, + { + "category_id": 1, + "poly": [ + 596, + 742, + 812, + 742, + 812, + 805, + 596, + 805 + ], + "score": 0.88 + }, + { + "category_id": 1, + "poly": [ + 939, + 634, + 1139, + 634, + 1139, + 695, + 939, + 695 + ], + "score": 0.873 + }, + { + "category_id": 1, + "poly": [ + 1001, + 851, + 1236, + 851, + 1236, + 944, + 1001, + 944 + ], + "score": 0.857 + }, + { + "category_id": 1, + "poly": [ + 840, + 743, + 1006, + 743, + 1006, + 805, + 840, + 805 + ], + "score": 0.777 + }, + { + "category_id": 1, + "poly": [ + 1209, + 635, + 1345, + 635, + 1345, + 695, + 1209, + 695 + ], + "score": 0.773 + }, + { + "category_id": 1, + "poly": [ + 1221, + 743, + 1383, + 743, + 1383, + 805, + 1221, + 805 + ], + "score": 0.725 + }, + { + "category_id": 1, + "poly": [ + 353, + 634, + 616, + 634, + 616, + 696, + 353, + 696 + ], + "score": 0.719 + }, + { + "category_id": 1, + "poly": [ + 1028, + 742, + 1198, + 742, + 1198, + 804, + 1028, + 804 + ], + "score": 0.619 + }, + { + "category_id": 1, + "poly": [ + 314, + 742, + 560, + 742, + 560, + 805, + 314, + 805 + ], + "score": 0.559 + }, + { + "category_id": 15, + "poly": [ + 688.0, + 268.0, + 1020.0, + 268.0, + 1020.0, + 328.0, + 688.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 435.0, + 322.0, + 1266.0, + 322.0, + 1266.0, + 383.0, + 435.0, + 383.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1515.0, + 536.0, + 1515.0, + 536.0, + 1563.0, + 292.0, + 1563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 782.0, + 1019.0, + 920.0, + 1019.0, + 920.0, + 1063.0, + 782.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 2030.0, + 1071.0, + 2030.0, + 1071.0, + 2066.0, + 295.0, + 2066.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 1946.0, + 542.0, + 1946.0, + 542.0, + 1983.0, + 329.0, + 1983.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 396.0, + 1094.0, + 1305.0, + 1094.0, + 1305.0, + 1129.0, + 396.0, + 1129.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1125.0, + 1305.0, + 1125.0, + 1305.0, + 1161.0, + 394.0, + 1161.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1154.0, + 1309.0, + 1154.0, + 1309.0, + 1190.0, + 393.0, + 1190.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1186.0, + 1306.0, + 1186.0, + 1306.0, + 1218.0, + 393.0, + 1218.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1216.0, + 1308.0, + 1216.0, + 1308.0, + 1250.0, + 392.0, + 1250.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1246.0, + 1307.0, + 1246.0, + 1307.0, + 1281.0, + 394.0, + 1281.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1275.0, + 1308.0, + 1275.0, + 1308.0, + 1310.0, + 393.0, + 1310.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1307.0, + 1305.0, + 1307.0, + 1305.0, + 1339.0, + 393.0, + 1339.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1335.0, + 1305.0, + 1335.0, + 1305.0, + 1373.0, + 392.0, + 1373.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1366.0, + 1304.0, + 1366.0, + 1304.0, + 1401.0, + 394.0, + 1401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1397.0, + 1304.0, + 1397.0, + 1304.0, + 1431.0, + 394.0, + 1431.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 1427.0, + 1223.0, + 1427.0, + 1223.0, + 1464.0, + 393.0, + 1464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1591.0, + 1402.0, + 1591.0, + 1402.0, + 1626.0, + 294.0, + 1626.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1621.0, + 1405.0, + 1621.0, + 1405.0, + 1657.0, + 293.0, + 1657.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1650.0, + 1404.0, + 1650.0, + 1404.0, + 1688.0, + 293.0, + 1688.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1680.0, + 1405.0, + 1680.0, + 1405.0, + 1717.0, + 293.0, + 1717.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1707.0, + 1406.0, + 1707.0, + 1406.0, + 1748.0, + 292.0, + 1748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1740.0, + 1408.0, + 1740.0, + 1408.0, + 1779.0, + 292.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1774.0, + 1404.0, + 1774.0, + 1404.0, + 1805.0, + 296.0, + 1805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1803.0, + 1405.0, + 1803.0, + 1405.0, + 1838.0, + 294.0, + 1838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1833.0, + 1405.0, + 1833.0, + 1405.0, + 1866.0, + 293.0, + 1866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1864.0, + 1405.0, + 1864.0, + 1405.0, + 1898.0, + 294.0, + 1898.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1892.0, + 1404.0, + 1892.0, + 1404.0, + 1930.0, + 293.0, + 1930.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 786.0, + 494.0, + 1030.0, + 494.0, + 1030.0, + 527.0, + 786.0, + 527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 843.0, + 522.0, + 975.0, + 522.0, + 975.0, + 561.0, + 843.0, + 561.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 408.0, + 491.0, + 675.0, + 491.0, + 675.0, + 529.0, + 408.0, + 529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 472.0, + 524.0, + 601.0, + 524.0, + 601.0, + 557.0, + 472.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 385.0, + 556.0, + 689.0, + 556.0, + 689.0, + 592.0, + 385.0, + 592.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 457.0, + 848.0, + 826.0, + 848.0, + 826.0, + 886.0, + 457.0, + 886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 578.0, + 881.0, + 710.0, + 881.0, + 710.0, + 914.0, + 578.0, + 914.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 509.0, + 911.0, + 778.0, + 911.0, + 778.0, + 949.0, + 509.0, + 949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 682.0, + 631.0, + 874.0, + 631.0, + 874.0, + 667.0, + 682.0, + 667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 713.0, + 660.0, + 846.0, + 660.0, + 846.0, + 700.0, + 713.0, + 700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1127.0, + 492.0, + 1315.0, + 492.0, + 1315.0, + 528.0, + 1127.0, + 528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1157.0, + 523.0, + 1288.0, + 523.0, + 1288.0, + 560.0, + 1157.0, + 560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 589.0, + 741.0, + 816.0, + 741.0, + 816.0, + 774.0, + 589.0, + 774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 640.0, + 771.0, + 769.0, + 771.0, + 769.0, + 808.0, + 640.0, + 808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 940.0, + 631.0, + 1141.0, + 631.0, + 1141.0, + 667.0, + 940.0, + 667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 660.0, + 1109.0, + 660.0, + 1109.0, + 700.0, + 974.0, + 700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1054.0, + 847.0, + 1187.0, + 847.0, + 1187.0, + 883.0, + 1054.0, + 883.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1056.0, + 878.0, + 1187.0, + 878.0, + 1187.0, + 915.0, + 1056.0, + 915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 997.0, + 911.0, + 1243.0, + 911.0, + 1243.0, + 947.0, + 997.0, + 947.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 843.0, + 737.0, + 1011.0, + 737.0, + 1011.0, + 778.0, + 843.0, + 778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 859.0, + 771.0, + 989.0, + 771.0, + 989.0, + 807.0, + 859.0, + 807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1206.0, + 631.0, + 1350.0, + 631.0, + 1350.0, + 667.0, + 1206.0, + 667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1211.0, + 661.0, + 1344.0, + 661.0, + 1344.0, + 700.0, + 1211.0, + 700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1226.0, + 738.0, + 1387.0, + 738.0, + 1387.0, + 776.0, + 1226.0, + 776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1241.0, + 772.0, + 1372.0, + 772.0, + 1372.0, + 808.0, + 1241.0, + 808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 634.0, + 618.0, + 634.0, + 618.0, + 666.0, + 351.0, + 666.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 421.0, + 664.0, + 551.0, + 664.0, + 551.0, + 697.0, + 421.0, + 697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1032.0, + 738.0, + 1202.0, + 738.0, + 1202.0, + 774.0, + 1032.0, + 774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1050.0, + 769.0, + 1184.0, + 769.0, + 1184.0, + 811.0, + 1050.0, + 811.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 314.0, + 742.0, + 564.0, + 742.0, + 564.0, + 775.0, + 314.0, + 775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 374.0, + 770.0, + 505.0, + 770.0, + 505.0, + 807.0, + 374.0, + 807.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 0, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1197, + 1406, + 1197, + 1406, + 1471, + 297, + 1471 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 297, + 628, + 1405, + 628, + 1405, + 1083, + 297, + 1083 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 297, + 279, + 1405, + 279, + 1405, + 613, + 297, + 613 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 298, + 1486, + 1405, + 1486, + 1405, + 1700, + 298, + 1700 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 1714, + 1404, + 1714, + 1404, + 1868, + 298, + 1868 + ], + "score": 0.971 + }, + { + "category_id": 2, + "poly": [ + 298, + 1894, + 1402, + 1894, + 1402, + 2006, + 298, + 2006 + ], + "score": 0.958 + }, + { + "category_id": 1, + "poly": [ + 296, + 203, + 1401, + 203, + 1401, + 264, + 296, + 264 + ], + "score": 0.947 + }, + { + "category_id": 0, + "poly": [ + 299, + 1127, + 548, + 1127, + 548, + 1165, + 299, + 1165 + ], + "score": 0.915 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.756 + }, + { + "category_id": 15, + "poly": [ + 328.0, + 1888.0, + 1407.0, + 1888.0, + 1407.0, + 1933.0, + 328.0, + 1933.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1921.0, + 1404.0, + 1921.0, + 1404.0, + 1955.0, + 293.0, + 1955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1947.0, + 1404.0, + 1947.0, + 1404.0, + 1983.0, + 294.0, + 1983.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1979.0, + 544.0, + 1979.0, + 544.0, + 2008.0, + 296.0, + 2008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1124.0, + 552.0, + 1124.0, + 552.0, + 1170.0, + 292.0, + 1170.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2093.0, + 838.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1199.0, + 1407.0, + 1199.0, + 1407.0, + 1231.0, + 296.0, + 1231.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1229.0, + 1408.0, + 1229.0, + 1408.0, + 1265.0, + 294.0, + 1265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1257.0, + 1409.0, + 1257.0, + 1409.0, + 1296.0, + 292.0, + 1296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1287.0, + 1406.0, + 1287.0, + 1406.0, + 1326.0, + 292.0, + 1326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1321.0, + 1408.0, + 1321.0, + 1408.0, + 1357.0, + 294.0, + 1357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1350.0, + 1406.0, + 1350.0, + 1406.0, + 1387.0, + 292.0, + 1387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1381.0, + 1403.0, + 1381.0, + 1403.0, + 1413.0, + 295.0, + 1413.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1408.0, + 1409.0, + 1408.0, + 1409.0, + 1447.0, + 292.0, + 1447.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1443.0, + 449.0, + 1443.0, + 449.0, + 1470.0, + 295.0, + 1470.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 629.0, + 1403.0, + 629.0, + 1403.0, + 660.0, + 296.0, + 660.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 657.0, + 1405.0, + 657.0, + 1405.0, + 692.0, + 295.0, + 692.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 686.0, + 1407.0, + 686.0, + 1407.0, + 725.0, + 291.0, + 725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 718.0, + 1405.0, + 718.0, + 1405.0, + 756.0, + 294.0, + 756.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 748.0, + 1406.0, + 748.0, + 1406.0, + 786.0, + 294.0, + 786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 780.0, + 1406.0, + 780.0, + 1406.0, + 815.0, + 294.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 810.0, + 1406.0, + 810.0, + 1406.0, + 844.0, + 295.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 839.0, + 1407.0, + 839.0, + 1407.0, + 874.0, + 294.0, + 874.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 870.0, + 1405.0, + 870.0, + 1405.0, + 905.0, + 294.0, + 905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 901.0, + 1406.0, + 901.0, + 1406.0, + 936.0, + 294.0, + 936.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 929.0, + 1406.0, + 929.0, + 1406.0, + 968.0, + 291.0, + 968.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 958.0, + 1408.0, + 958.0, + 1408.0, + 998.0, + 292.0, + 998.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 992.0, + 1406.0, + 992.0, + 1406.0, + 1026.0, + 295.0, + 1026.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1019.0, + 1408.0, + 1019.0, + 1408.0, + 1060.0, + 294.0, + 1060.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1050.0, + 871.0, + 1050.0, + 871.0, + 1086.0, + 292.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 280.0, + 1405.0, + 280.0, + 1405.0, + 314.0, + 295.0, + 314.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 310.0, + 1405.0, + 310.0, + 1405.0, + 345.0, + 294.0, + 345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 337.0, + 1406.0, + 337.0, + 1406.0, + 377.0, + 292.0, + 377.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 370.0, + 1406.0, + 370.0, + 1406.0, + 405.0, + 292.0, + 405.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 399.0, + 1407.0, + 399.0, + 1407.0, + 436.0, + 291.0, + 436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 431.0, + 1406.0, + 431.0, + 1406.0, + 466.0, + 295.0, + 466.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 462.0, + 1407.0, + 462.0, + 1407.0, + 495.0, + 292.0, + 495.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 488.0, + 1409.0, + 488.0, + 1409.0, + 529.0, + 292.0, + 529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 520.0, + 1407.0, + 520.0, + 1407.0, + 556.0, + 292.0, + 556.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 551.0, + 1405.0, + 551.0, + 1405.0, + 585.0, + 294.0, + 585.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 582.0, + 890.0, + 582.0, + 890.0, + 617.0, + 294.0, + 617.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1485.0, + 1407.0, + 1485.0, + 1407.0, + 1522.0, + 295.0, + 1522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1518.0, + 1408.0, + 1518.0, + 1408.0, + 1553.0, + 293.0, + 1553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1548.0, + 1407.0, + 1548.0, + 1407.0, + 1582.0, + 293.0, + 1582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1577.0, + 1405.0, + 1577.0, + 1405.0, + 1611.0, + 295.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1608.0, + 1407.0, + 1608.0, + 1407.0, + 1642.0, + 295.0, + 1642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1637.0, + 1403.0, + 1637.0, + 1403.0, + 1674.0, + 293.0, + 1674.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1669.0, + 836.0, + 1669.0, + 836.0, + 1703.0, + 296.0, + 1703.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1712.0, + 1408.0, + 1712.0, + 1408.0, + 1750.0, + 293.0, + 1750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1741.0, + 1405.0, + 1741.0, + 1405.0, + 1783.0, + 293.0, + 1783.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1772.0, + 1407.0, + 1772.0, + 1407.0, + 1812.0, + 292.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1803.0, + 1406.0, + 1803.0, + 1406.0, + 1843.0, + 293.0, + 1843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1833.0, + 1405.0, + 1833.0, + 1405.0, + 1873.0, + 292.0, + 1873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 200.0, + 1406.0, + 200.0, + 1406.0, + 240.0, + 294.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 233.0, + 677.0, + 233.0, + 677.0, + 269.0, + 295.0, + 269.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 1, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 924, + 1406, + 924, + 1406, + 1290, + 297, + 1290 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 579, + 1405, + 579, + 1405, + 825, + 297, + 825 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 299, + 388, + 1406, + 388, + 1406, + 479, + 299, + 479 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 295, + 203, + 1403, + 203, + 1403, + 266, + 295, + 266 + ], + "score": 0.948 + }, + { + "category_id": 3, + "poly": [ + 389, + 1331, + 1313, + 1331, + 1313, + 1838, + 389, + 1838 + ], + "score": 0.937 + }, + { + "category_id": 4, + "poly": [ + 431, + 1884, + 1263, + 1884, + 1263, + 1918, + 431, + 1918 + ], + "score": 0.924 + }, + { + "category_id": 0, + "poly": [ + 299, + 521, + 671, + 521, + 671, + 554, + 299, + 554 + ], + "score": 0.917 + }, + { + "category_id": 2, + "poly": [ + 331, + 1977, + 957, + 1977, + 957, + 2008, + 331, + 2008 + ], + "score": 0.916 + }, + { + "category_id": 0, + "poly": [ + 299, + 866, + 732, + 866, + 732, + 899, + 299, + 899 + ], + "score": 0.914 + }, + { + "category_id": 0, + "poly": [ + 298, + 313, + 462, + 313, + 462, + 350, + 298, + 350 + ], + "score": 0.905 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.696 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.387 + }, + { + "category_id": 13, + "poly": [ + 1063, + 701, + 1225, + 701, + 1225, + 737, + 1063, + 737 + ], + "score": 0.92, + "latex": "\\textstyle { \\mathcal { H } } = \\bigcup _ { t \\in \\mathbb { N } } { \\mathcal { H } } _ { t }" + }, + { + "category_id": 13, + "poly": [ + 298, + 642, + 386, + 642, + 386, + 672, + 298, + 672 + ], + "score": 0.92, + "latex": "a _ { t } \\in \\mathcal A" + }, + { + "category_id": 13, + "poly": [ + 948, + 733, + 1099, + 733, + 1099, + 763, + 948, + 763 + ], + "score": 0.92, + "latex": "\\pi : \\mathcal { H } \\to \\Delta _ { \\mathcal { A } }" + }, + { + "category_id": 13, + "poly": [ + 1302, + 702, + 1401, + 702, + 1401, + 733, + 1302, + 733 + ], + "score": 0.92, + "latex": "\\mathcal { H } _ { 0 } = \\mathcal { O }" + }, + { + "category_id": 13, + "poly": [ + 1032, + 641, + 1261, + 641, + 1261, + 674, + 1032, + 674 + ], + "score": 0.92, + "latex": "p : \\mathcal { H } \\times \\mathcal { A } \\to \\Delta _ { \\mathcal { O } } { } ^ { 3 }" + }, + { + "category_id": 13, + "poly": [ + 298, + 793, + 491, + 793, + 491, + 823, + 298, + 823 + ], + "score": 0.91, + "latex": "r _ { e } : \\mathcal { H } \\times \\mathcal { A } \\to \\mathbb { R }" + }, + { + "category_id": 13, + "poly": [ + 1042, + 611, + 1127, + 611, + 1127, + 641, + 1042, + 641 + ], + "score": 0.91, + "latex": "o _ { t } \\in \\mathcal { O }" + }, + { + "category_id": 13, + "poly": [ + 578, + 611, + 647, + 611, + 647, + 640, + 578, + 640 + ], + "score": 0.89, + "latex": "t \\in \\mathbb { N }" + }, + { + "category_id": 14, + "poly": [ + 586, + 1481, + 1312, + 1481, + 1312, + 1572, + 586, + 1572 + ], + "score": 0.87, + "latex": "\\mathcal { L } _ { \\mathtt { B Y 0 L - E x p l o r e } } ( \\theta , t ) = \\frac { 1 } { K } \\sum _ { k = 1 } ^ { K } \\left| \\left| \\frac { g _ { \\theta } ( b _ { t , k } ) } { \\| g _ { \\theta } ( b _ { t , k } ) \\| _ { 2 } } - \\mathbf { s } \\mathbf { g } \\left( \\frac { f _ { \\phi } ( o _ { t + k } ) } { \\| f _ { \\phi } ( o _ { t + k } ) \\| _ { 2 } } \\right) \\right| \\right| _ { 2 } ^ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 928, + 1984, + 948, + 1984, + 948, + 2005, + 928, + 2005 + ], + "score": 0.84, + "latex": "_ { \\mathcal { V } }" + }, + { + "category_id": 13, + "poly": [ + 439, + 1978, + 479, + 1978, + 479, + 2007, + 439, + 2007 + ], + "score": 0.84, + "latex": "\\Delta _ { y }" + }, + { + "category_id": 13, + "poly": [ + 449, + 732, + 695, + 732, + 695, + 765, + 449, + 765 + ], + "score": 0.78, + "latex": "\\mathcal { H } _ { t + 1 } = \\mathcal { H } _ { t } \\times \\mathcal { A } \\times \\mathcal { O }" + }, + { + "category_id": 13, + "poly": [ + 344, + 732, + 438, + 732, + 438, + 762, + 344, + 762 + ], + "score": 0.74, + "latex": "\\forall t \\in \\mathbb { N } ^ { * }" + }, + { + "category_id": 13, + "poly": [ + 395, + 1756, + 418, + 1756, + 418, + 1775, + 395, + 1775 + ], + "score": 0.43, + "latex": "O _ { t }" + }, + { + "category_id": 13, + "poly": [ + 519, + 1657, + 576, + 1657, + 576, + 1679, + 519, + 1679 + ], + "score": 0.26, + "latex": "a _ { t - 1 }" + }, + { + "category_id": 15, + "poly": [ + 782.0, + 1328.0, + 850.0, + 1328.0, + 850.0, + 1369.0, + 782.0, + 1369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 910.0, + 1330.0, + 978.0, + 1330.0, + 978.0, + 1372.0, + 910.0, + 1372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1144.0, + 1328.0, + 1219.0, + 1328.0, + 1219.0, + 1367.0, + 1144.0, + 1367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1029.0, + 1399.0, + 1063.0, + 1399.0, + 1063.0, + 1410.0, + 1029.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1065.0, + 1399.0, + 1109.0, + 1399.0, + 1109.0, + 1410.0, + 1065.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 780.0, + 1408.0, + 846.0, + 1408.0, + 846.0, + 1432.0, + 780.0, + 1432.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 908.0, + 1406.0, + 977.0, + 1406.0, + 977.0, + 1433.0, + 908.0, + 1433.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1151.0, + 1406.0, + 1217.0, + 1406.0, + 1217.0, + 1434.0, + 1151.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 786.0, + 1422.0, + 839.0, + 1422.0, + 839.0, + 1477.0, + 786.0, + 1477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 915.0, + 1422.0, + 969.0, + 1422.0, + 969.0, + 1478.0, + 915.0, + 1478.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1157.0, + 1422.0, + 1210.0, + 1422.0, + 1210.0, + 1478.0, + 1157.0, + 1478.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 783.0, + 1590.0, + 838.0, + 1590.0, + 838.0, + 1611.0, + 783.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 921.0, + 1590.0, + 977.0, + 1590.0, + 977.0, + 1611.0, + 921.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1161.0, + 1590.0, + 1218.0, + 1590.0, + 1218.0, + 1611.0, + 1161.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 790.0, + 1611.0, + 832.0, + 1611.0, + 832.0, + 1645.0, + 790.0, + 1645.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 926.0, + 1609.0, + 972.0, + 1609.0, + 972.0, + 1646.0, + 926.0, + 1646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1172.0, + 1609.0, + 1217.0, + 1609.0, + 1217.0, + 1646.0, + 1172.0, + 1646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 513.0, + 1645.0, + 518.0, + 1645.0, + 518.0, + 1687.0, + 513.0, + 1687.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 577.0, + 1645.0, + 585.0, + 1645.0, + 585.0, + 1687.0, + 577.0, + 1687.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 642.0, + 1649.0, + 654.0, + 1649.0, + 654.0, + 1672.0, + 642.0, + 1672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 752.0, + 1644.0, + 826.0, + 1644.0, + 826.0, + 1690.0, + 752.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 893.0, + 1641.0, + 963.0, + 1641.0, + 963.0, + 1691.0, + 893.0, + 1691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1129.0, + 1639.0, + 1207.0, + 1639.0, + 1207.0, + 1688.0, + 1129.0, + 1688.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 641.0, + 1668.0, + 654.0, + 1668.0, + 654.0, + 1681.0, + 641.0, + 1681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 612.0, + 1692.0, + 683.0, + 1692.0, + 683.0, + 1715.0, + 612.0, + 1715.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 698.0, + 1682.0, + 739.0, + 1682.0, + 739.0, + 1728.0, + 698.0, + 1728.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 778.0, + 1692.0, + 846.0, + 1692.0, + 846.0, + 1719.0, + 778.0, + 1719.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 916.0, + 1695.0, + 981.0, + 1695.0, + 981.0, + 1716.0, + 916.0, + 1716.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1159.0, + 1694.0, + 1224.0, + 1694.0, + 1224.0, + 1715.0, + 1159.0, + 1715.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 474.0, + 1740.0, + 528.0, + 1740.0, + 528.0, + 1762.0, + 474.0, + 1762.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 621.0, + 1708.0, + 674.0, + 1708.0, + 674.0, + 1769.0, + 621.0, + 1769.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 784.0, + 1708.0, + 844.0, + 1708.0, + 844.0, + 1772.0, + 784.0, + 1772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 922.0, + 1705.0, + 982.0, + 1705.0, + 982.0, + 1772.0, + 922.0, + 1772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1164.0, + 1706.0, + 1223.0, + 1706.0, + 1223.0, + 1772.0, + 1164.0, + 1772.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 482.0, + 1754.0, + 528.0, + 1754.0, + 528.0, + 1798.0, + 482.0, + 1798.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 632.0, + 1768.0, + 688.0, + 1768.0, + 688.0, + 1835.0, + 632.0, + 1835.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 806.0, + 1768.0, + 815.0, + 1768.0, + 815.0, + 1776.0, + 806.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 1766.0, + 957.0, + 1766.0, + 957.0, + 1784.0, + 941.0, + 1784.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1186.0, + 1768.0, + 1194.0, + 1768.0, + 1194.0, + 1776.0, + 1186.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 918.0, + 1796.0, + 989.0, + 1796.0, + 989.0, + 1838.0, + 918.0, + 1838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1145.0, + 1793.0, + 1252.0, + 1793.0, + 1252.0, + 1834.0, + 1145.0, + 1834.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 790.0, + 1517.5, + 805.0, + 1517.5, + 805.0, + 1538.5, + 790.0, + 1538.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 818.75, + 1519.0, + 843.75, + 1519.0, + 843.75, + 1557.0, + 818.75, + 1557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1289.75, + 1541.0, + 1303.75, + 1541.0, + 1303.75, + 1558.5, + 1289.75, + 1558.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 647.25, + 1610.0, + 707.25, + 1610.0, + 707.25, + 1652.5, + 647.25, + 1652.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 859.25, + 1685.5, + 897.25, + 1685.5, + 897.25, + 1724.5, + 859.25, + 1724.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1057.0, + 1686.0, + 1142.0, + 1686.0, + 1142.0, + 1725.0, + 1057.0, + 1725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 1748.5, + 433.0, + 1748.5, + 433.0, + 1779.0, + 392.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 793.75, + 1799.5, + 831.75, + 1799.5, + 831.75, + 1835.0, + 793.75, + 1835.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 431.0, + 1882.0, + 1267.0, + 1882.0, + 1267.0, + 1921.0, + 431.0, + 1921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 520.0, + 673.0, + 520.0, + 673.0, + 559.0, + 294.0, + 559.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 329.0, + 1969.0, + 438.0, + 1969.0, + 438.0, + 2015.0, + 329.0, + 2015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 480.0, + 1969.0, + 927.0, + 1969.0, + 927.0, + 2015.0, + 480.0, + 2015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 949.0, + 1969.0, + 961.0, + 1969.0, + 961.0, + 2015.0, + 949.0, + 2015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 866.0, + 737.0, + 866.0, + 737.0, + 902.0, + 295.0, + 902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 308.0, + 468.0, + 308.0, + 468.0, + 356.0, + 289.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 926.0, + 1406.0, + 926.0, + 1406.0, + 960.0, + 295.0, + 960.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 957.0, + 1406.0, + 957.0, + 1406.0, + 990.0, + 295.0, + 990.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 986.0, + 1406.0, + 986.0, + 1406.0, + 1020.0, + 295.0, + 1020.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1017.0, + 1406.0, + 1017.0, + 1406.0, + 1052.0, + 294.0, + 1052.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1045.0, + 1409.0, + 1045.0, + 1409.0, + 1084.0, + 292.0, + 1084.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1078.0, + 1406.0, + 1078.0, + 1406.0, + 1112.0, + 294.0, + 1112.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1107.0, + 1404.0, + 1107.0, + 1404.0, + 1142.0, + 294.0, + 1142.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1137.0, + 1406.0, + 1137.0, + 1406.0, + 1171.0, + 292.0, + 1171.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1168.0, + 1404.0, + 1168.0, + 1404.0, + 1202.0, + 295.0, + 1202.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1197.0, + 1406.0, + 1197.0, + 1406.0, + 1232.0, + 294.0, + 1232.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1231.0, + 1404.0, + 1231.0, + 1404.0, + 1261.0, + 296.0, + 1261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1259.0, + 1408.0, + 1259.0, + 1408.0, + 1293.0, + 295.0, + 1293.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 579.0, + 1406.0, + 579.0, + 1406.0, + 617.0, + 294.0, + 617.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 609.0, + 577.0, + 609.0, + 577.0, + 644.0, + 294.0, + 644.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 648.0, + 609.0, + 1041.0, + 609.0, + 1041.0, + 644.0, + 648.0, + 644.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1128.0, + 609.0, + 1405.0, + 609.0, + 1405.0, + 644.0, + 1128.0, + 644.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 641.0, + 297.0, + 641.0, + 297.0, + 678.0, + 292.0, + 678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 641.0, + 1031.0, + 641.0, + 1031.0, + 678.0, + 387.0, + 678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1262.0, + 641.0, + 1408.0, + 641.0, + 1408.0, + 678.0, + 1262.0, + 678.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 672.0, + 1405.0, + 672.0, + 1405.0, + 706.0, + 294.0, + 706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 700.0, + 1062.0, + 700.0, + 1062.0, + 740.0, + 292.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1226.0, + 700.0, + 1301.0, + 700.0, + 1301.0, + 740.0, + 1226.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 700.0, + 1407.0, + 700.0, + 1407.0, + 740.0, + 1402.0, + 740.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 730.0, + 343.0, + 730.0, + 343.0, + 767.0, + 292.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 439.0, + 730.0, + 448.0, + 730.0, + 448.0, + 767.0, + 439.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 696.0, + 730.0, + 947.0, + 730.0, + 947.0, + 767.0, + 696.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 730.0, + 1407.0, + 730.0, + 1407.0, + 767.0, + 1100.0, + 767.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 764.0, + 1403.0, + 764.0, + 1403.0, + 795.0, + 296.0, + 795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 793.0, + 297.0, + 793.0, + 297.0, + 826.0, + 294.0, + 826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 492.0, + 793.0, + 1171.0, + 793.0, + 1171.0, + 826.0, + 492.0, + 826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 387.0, + 1406.0, + 387.0, + 1406.0, + 420.0, + 294.0, + 420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 418.0, + 1406.0, + 418.0, + 1406.0, + 451.0, + 296.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 447.0, + 1343.0, + 447.0, + 1343.0, + 484.0, + 294.0, + 484.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 203.0, + 1406.0, + 203.0, + 1406.0, + 238.0, + 293.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 233.0, + 1338.0, + 233.0, + 1338.0, + 268.0, + 294.0, + 268.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 2, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1505, + 1405, + 1505, + 1405, + 1781, + 297, + 1781 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 297, + 570, + 1404, + 570, + 1404, + 730, + 297, + 730 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 299, + 1868, + 1406, + 1868, + 1406, + 2009, + 299, + 2009 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 756, + 1404, + 756, + 1404, + 925, + 298, + 925 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 298, + 1379, + 1405, + 1379, + 1405, + 1477, + 298, + 1477 + ], + "score": 0.954 + }, + { + "category_id": 8, + "poly": [ + 401, + 932, + 1301, + 932, + 1301, + 1133, + 401, + 1133 + ], + "score": 0.949 + }, + { + "category_id": 8, + "poly": [ + 633, + 1303, + 1067, + 1303, + 1067, + 1374, + 633, + 1374 + ], + "score": 0.945 + }, + { + "category_id": 1, + "poly": [ + 297, + 1232, + 1401, + 1232, + 1401, + 1297, + 297, + 1297 + ], + "score": 0.945 + }, + { + "category_id": 1, + "poly": [ + 299, + 1137, + 1401, + 1137, + 1401, + 1202, + 299, + 1202 + ], + "score": 0.94 + }, + { + "category_id": 0, + "poly": [ + 297, + 1812, + 953, + 1812, + 953, + 1846, + 297, + 1846 + ], + "score": 0.933 + }, + { + "category_id": 1, + "poly": [ + 297, + 201, + 1405, + 201, + 1405, + 542, + 297, + 542 + ], + "score": 0.929 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 858, + 2061, + 858, + 2084, + 841, + 2084 + ], + "score": 0.763 + }, + { + "category_id": 13, + "poly": [ + 809, + 386, + 999, + 386, + 999, + 423, + 809, + 423 + ], + "score": 0.94, + "latex": "( b _ { t , k } \\in \\mathbb { R } ^ { M } ) _ { k = 1 } ^ { K - 1 }" + }, + { + "category_id": 13, + "poly": [ + 1184, + 1230, + 1334, + 1230, + 1334, + 1271, + 1184, + 1271 + ], + "score": 0.94, + "latex": "( o _ { t } ^ { j } , a _ { t } ^ { j } , o _ { t + 1 } ^ { j } )" + }, + { + "category_id": 13, + "poly": [ + 898, + 884, + 1002, + 884, + 1002, + 927, + 898, + 927 + ], + "score": 0.94, + "latex": "f _ { \\phi } ( o _ { t + k } ^ { j } )" + }, + { + "category_id": 13, + "poly": [ + 500, + 885, + 591, + 885, + 591, + 929, + 500, + 929 + ], + "score": 0.93, + "latex": "g _ { \\theta } ( b _ { t , k } ^ { j } )" + }, + { + "category_id": 13, + "poly": [ + 976, + 478, + 1139, + 478, + 1139, + 513, + 976, + 513 + ], + "score": 0.93, + "latex": "g _ { \\theta } ( b _ { t , k } ) \\dot { } \\in \\mathbb { R } ^ { N }" + }, + { + "category_id": 14, + "poly": [ + 632, + 1299, + 1063, + 1299, + 1063, + 1376, + 632, + 1376 + ], + "score": 0.93, + "latex": "\\ell _ { t } ^ { j } = \\sum _ { p + q = t + 1 } \\mathcal { L } _ { \\mathtt { B Y O L - E x p l o r e } } ( \\theta , j , p , q ) ," + }, + { + "category_id": 13, + "poly": [ + 862, + 232, + 1011, + 232, + 1011, + 267, + 862, + 267 + ], + "score": 0.93, + "latex": "f _ { \\theta } ( o _ { t } ) \\in \\mathbb { R } ^ { N }" + }, + { + "category_id": 13, + "poly": [ + 971, + 1931, + 1135, + 1931, + 1135, + 1973, + 971, + 1973 + ], + "score": 0.93, + "latex": "( ( \\ell _ { t } ^ { j } ) _ { t = 0 } ^ { T - 2 } ) _ { j = 0 } ^ { B - 1 }" + }, + { + "category_id": 13, + "poly": [ + 925, + 1973, + 995, + 1973, + 995, + 2009, + 925, + 2009 + ], + "score": 0.93, + "latex": "\\ell _ { t } ^ { j } / \\sigma _ { r }" + }, + { + "category_id": 13, + "poly": [ + 696, + 324, + 1009, + 324, + 1009, + 359, + 696, + 359 + ], + "score": 0.92, + "latex": "b _ { t } = h _ { \\theta } ^ { c } ( { \\bar { b } } _ { t - 1 } , a _ { t - 1 } , f _ { \\theta } ( o _ { t } ) )" + }, + { + "category_id": 13, + "poly": [ + 730, + 263, + 803, + 263, + 803, + 296, + 730, + 296 + ], + "score": 0.92, + "latex": "f _ { \\theta } ( o _ { t } )" + }, + { + "category_id": 13, + "poly": [ + 955, + 1409, + 1009, + 1409, + 1009, + 1450, + 955, + 1450 + ], + "score": 0.92, + "latex": "o _ { t + 1 } ^ { j }" + }, + { + "category_id": 13, + "poly": [ + 517, + 667, + 748, + 667, + 748, + 699, + 517, + 699 + ], + "score": 0.92, + "latex": "\\phi \\alpha \\phi + ( 1 - \\alpha ) \\theta" + }, + { + "category_id": 13, + "poly": [ + 299, + 788, + 510, + 788, + 510, + 854, + 299, + 854 + ], + "score": 0.92, + "latex": "\\left( ( o _ { t } ^ { j } , a _ { t } ^ { j } ) _ { t = 0 } ^ { T - 1 } \\right) _ { j = 0 } ^ { B - 1 }" + }, + { + "category_id": 13, + "poly": [ + 427, + 326, + 527, + 326, + 527, + 356, + 427, + 356 + ], + "score": 0.92, + "latex": "h _ { t } \\in \\mathcal { H } _ { t }" + }, + { + "category_id": 13, + "poly": [ + 1034, + 387, + 1326, + 387, + 1326, + 423, + 1034, + 423 + ], + "score": 0.91, + "latex": "b _ { t , k } = h _ { \\theta } ^ { o } ( b _ { t , k - 1 } , a _ { t + k - 1 } )" + }, + { + "category_id": 13, + "poly": [ + 1101, + 235, + 1198, + 235, + 1198, + 262, + 1101, + 262 + ], + "score": 0.91, + "latex": "N \\in \\mathbb { N } ^ { * }" + }, + { + "category_id": 13, + "poly": [ + 1100, + 325, + 1204, + 325, + 1204, + 353, + 1100, + 353 + ], + "score": 0.91, + "latex": "M \\in \\mathbb { N } ^ { * }" + }, + { + "category_id": 13, + "poly": [ + 1130, + 604, + 1308, + 604, + 1308, + 640, + 1130, + 640 + ], + "score": 0.91, + "latex": "f _ { \\phi } ( o _ { t + k } ) \\in \\mathbb { R } ^ { N }" + }, + { + "category_id": 13, + "poly": [ + 298, + 421, + 398, + 421, + 398, + 452, + 298, + 452 + ], + "score": 0.91, + "latex": "b _ { t , 0 } = b _ { t }" + }, + { + "category_id": 13, + "poly": [ + 726, + 1380, + 891, + 1380, + 891, + 1411, + 726, + 1411 + ], + "score": 0.91, + "latex": "0 \\leq t \\leq T - 2" + }, + { + "category_id": 13, + "poly": [ + 298, + 480, + 341, + 480, + 341, + 513, + 298, + 513 + ], + "score": 0.91, + "latex": "b _ { t , k }" + }, + { + "category_id": 13, + "poly": [ + 370, + 1138, + 681, + 1138, + 681, + 1172, + 370, + 1172 + ], + "score": 0.91, + "latex": "K ( t ) = \\operatorname* { m i n } ( K , T - 1 - t )" + }, + { + "category_id": 13, + "poly": [ + 1000, + 806, + 1091, + 806, + 1091, + 834, + 1000, + 834 + ], + "score": 0.91, + "latex": "B \\in \\mathbb { N } ^ { * }" + }, + { + "category_id": 13, + "poly": [ + 737, + 511, + 799, + 511, + 799, + 539, + 737, + 539 + ], + "score": 0.9, + "latex": "t + k" + }, + { + "category_id": 13, + "poly": [ + 1266, + 294, + 1369, + 294, + 1369, + 325, + 1266, + 325 + ], + "score": 0.9, + "latex": "b _ { t } \\in \\mathbb { R } ^ { M }" + }, + { + "category_id": 13, + "poly": [ + 1226, + 480, + 1289, + 480, + 1289, + 509, + 1226, + 509 + ], + "score": 0.9, + "latex": "t + k" + }, + { + "category_id": 13, + "poly": [ + 1026, + 1865, + 1055, + 1865, + 1055, + 1903, + 1026, + 1903 + ], + "score": 0.9, + "latex": "\\ell _ { t } ^ { j }" + }, + { + "category_id": 13, + "poly": [ + 598, + 806, + 688, + 806, + 688, + 834, + 598, + 834 + ], + "score": 0.9, + "latex": "T \\in \\mathbb { N } ^ { * }" + }, + { + "category_id": 13, + "poly": [ + 369, + 1380, + 677, + 1380, + 677, + 1412, + 369, + 1412 + ], + "score": 0.9, + "latex": "0 \\leq p \\leq T - 2 , 1 \\leq q \\leq K" + }, + { + "category_id": 13, + "poly": [ + 1089, + 890, + 1152, + 890, + 1152, + 919, + 1089, + 919 + ], + "score": 0.89, + "latex": "t + k" + }, + { + "category_id": 13, + "poly": [ + 1211, + 573, + 1242, + 573, + 1242, + 605, + 1211, + 605 + ], + "score": 0.88, + "latex": "f _ { \\phi }" + }, + { + "category_id": 13, + "poly": [ + 412, + 295, + 443, + 295, + 443, + 327, + 412, + 327 + ], + "score": 0.87, + "latex": "h _ { \\theta } ^ { c }" + }, + { + "category_id": 13, + "poly": [ + 344, + 389, + 375, + 389, + 375, + 422, + 344, + 422 + ], + "score": 0.87, + "latex": "h _ { \\theta } ^ { o }" + }, + { + "category_id": 13, + "poly": [ + 1154, + 205, + 1182, + 205, + 1182, + 234, + 1154, + 234 + ], + "score": 0.86, + "latex": "f _ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 932, + 358, + 956, + 358, + 956, + 385, + 932, + 385 + ], + "score": 0.86, + "latex": "b _ { t }" + }, + { + "category_id": 13, + "poly": [ + 565, + 1981, + 595, + 1981, + 595, + 2006, + 565, + 2006 + ], + "score": 0.86, + "latex": "\\sigma _ { r }" + }, + { + "category_id": 13, + "poly": [ + 564, + 485, + 592, + 485, + 592, + 511, + 564, + 511 + ], + "score": 0.84, + "latex": "g _ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 436, + 238, + 462, + 238, + 462, + 264, + 436, + 264 + ], + "score": 0.84, + "latex": "o _ { t }" + }, + { + "category_id": 13, + "poly": [ + 1295, + 1140, + 1319, + 1140, + 1319, + 1166, + 1295, + 1166 + ], + "score": 0.84, + "latex": "T" + }, + { + "category_id": 13, + "poly": [ + 349, + 854, + 531, + 854, + 531, + 886, + 349, + 886 + ], + "score": 0.83, + "latex": "\\mathcal { L } _ { \\mathtt { B Y O L - E x p l o r e } } ( \\theta )" + }, + { + "category_id": 13, + "poly": [ + 449, + 420, + 478, + 420, + 478, + 447, + 449, + 447 + ], + "score": 0.82, + "latex": "K" + }, + { + "category_id": 13, + "poly": [ + 825, + 672, + 847, + 672, + 847, + 694, + 825, + 694 + ], + "score": 0.8, + "latex": "\\alpha" + }, + { + "category_id": 13, + "poly": [ + 1290, + 269, + 1346, + 269, + 1346, + 294, + 1290, + 294 + ], + "score": 0.79, + "latex": "a _ { t - 1 }" + }, + { + "category_id": 13, + "poly": [ + 902, + 606, + 919, + 606, + 919, + 633, + 902, + 633 + ], + "score": 0.74, + "latex": "\\theta" + }, + { + "category_id": 14, + "poly": [ + 395, + 933, + 1297, + 933, + 1297, + 1137, + 395, + 1137 + ], + "score": 0.74, + "latex": "\\begin{array} { r l } & { \\mathcal { L } _ { \\mathtt { B Y 0 L - E x p l o r e } } ( \\theta , j , t , k ) = \\bigg \\| \\frac { g _ { \\theta } ( b _ { t , k } ^ { j } ) } { \\| g _ { \\theta } ( b _ { t , k } ^ { j } ) \\| _ { 2 } } - \\mathrm { s g } \\left( \\frac { f _ { \\phi } ( o _ { t + k } ^ { j } ) } { \\| f _ { \\phi } ( o _ { t + k } ^ { j } ) \\| _ { 2 } } \\right) \\bigg \\| _ { 2 } ^ { 2 } , } \\\\ & { \\mathcal { L } _ { \\mathtt { B Y 0 L - E x p l o r e } } ( \\theta ) = \\frac { 1 } { B ( T - 1 ) } \\displaystyle \\sum _ { j = 0 } ^ { B - 1 } \\sum _ { t = 0 } ^ { T - 2 } \\frac { 1 } { K ( t ) } \\sum _ { k = 1 } ^ { K ( t ) } \\mathcal { L } _ { \\mathtt { B Y 0 L - E x p l o r e } } ( \\theta , j , t , k ) , } \\end{array}" + }, + { + "category_id": 14, + "poly": [ + 475, + 1038, + 1294, + 1038, + 1294, + 1135, + 475, + 1135 + ], + "score": 0.39, + "latex": "\\mathcal { L } _ { \\mathtt { B Y O L - E x p l o r e } } ( \\theta ) = \\frac { 1 } { B ( T - 1 ) } \\sum _ { j = 0 } ^ { B - 1 } \\sum _ { t = 0 } ^ { T - 2 } \\frac { 1 } { K ( t ) } \\sum _ { k = 1 } ^ { K ( t ) } \\mathcal { L } _ { \\mathtt { B Y O L - E x p l o r e } } ( \\theta , j , t , k ) ," + }, + { + "category_id": 14, + "poly": [ + 397, + 933, + 1176, + 933, + 1176, + 1031, + 397, + 1031 + ], + "score": 0.31, + "latex": "\\mathcal { L } _ { \\mathtt { B Y 0 L - E x p l o r e } } ( \\theta , j , t , k ) = \\left\\| \\frac { g _ { \\theta } ( b _ { t , k } ^ { j } ) } { \\| g _ { \\theta } ( b _ { t , k } ^ { j } ) \\| _ { 2 } } - \\mathbf { s g } \\left( \\frac { f _ { \\phi } ( o _ { t + k } ^ { j } ) } { \\| f _ { \\phi } ( o _ { t + k } ^ { j } ) \\| _ { 2 } } \\right) \\right\\| _ { 2 } ^ { 2 } ," + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1811.0, + 954.0, + 1811.0, + 954.0, + 1851.0, + 293.0, + 1851.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2059.0, + 862.0, + 2059.0, + 862.0, + 2091.0, + 839.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1506.0, + 1406.0, + 1506.0, + 1406.0, + 1542.0, + 295.0, + 1542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1535.0, + 1406.0, + 1535.0, + 1406.0, + 1573.0, + 292.0, + 1573.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1567.0, + 1405.0, + 1567.0, + 1405.0, + 1600.0, + 295.0, + 1600.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1595.0, + 1407.0, + 1595.0, + 1407.0, + 1634.0, + 291.0, + 1634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1627.0, + 1406.0, + 1627.0, + 1406.0, + 1661.0, + 292.0, + 1661.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1657.0, + 1405.0, + 1657.0, + 1405.0, + 1694.0, + 295.0, + 1694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1689.0, + 1402.0, + 1689.0, + 1402.0, + 1721.0, + 296.0, + 1721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1717.0, + 1405.0, + 1717.0, + 1405.0, + 1754.0, + 295.0, + 1754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1749.0, + 960.0, + 1749.0, + 960.0, + 1785.0, + 295.0, + 1785.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 570.0, + 1210.0, + 570.0, + 1210.0, + 604.0, + 297.0, + 604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1243.0, + 570.0, + 1405.0, + 570.0, + 1405.0, + 604.0, + 1243.0, + 604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 601.0, + 901.0, + 601.0, + 901.0, + 642.0, + 292.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 920.0, + 601.0, + 1129.0, + 601.0, + 1129.0, + 642.0, + 920.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1309.0, + 601.0, + 1405.0, + 601.0, + 1405.0, + 642.0, + 1309.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 636.0, + 1404.0, + 636.0, + 1404.0, + 670.0, + 296.0, + 670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 666.0, + 516.0, + 666.0, + 516.0, + 700.0, + 296.0, + 700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 749.0, + 666.0, + 824.0, + 666.0, + 824.0, + 700.0, + 749.0, + 700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 848.0, + 666.0, + 1405.0, + 666.0, + 1405.0, + 700.0, + 848.0, + 700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 695.0, + 1110.0, + 695.0, + 1110.0, + 731.0, + 295.0, + 731.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1868.0, + 1025.0, + 1868.0, + 1025.0, + 1903.0, + 296.0, + 1903.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1056.0, + 1868.0, + 1408.0, + 1868.0, + 1408.0, + 1903.0, + 1056.0, + 1903.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 282.0, + 1901.0, + 970.0, + 1901.0, + 970.0, + 1990.0, + 282.0, + 1990.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1136.0, + 1901.0, + 1406.0, + 1901.0, + 1406.0, + 1990.0, + 1136.0, + 1990.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1972.0, + 564.0, + 1972.0, + 564.0, + 2011.0, + 294.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 596.0, + 1972.0, + 924.0, + 1972.0, + 924.0, + 2011.0, + 596.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 996.0, + 1972.0, + 1386.0, + 1972.0, + 1386.0, + 2011.0, + 996.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 750.0, + 1406.0, + 750.0, + 1406.0, + 800.0, + 292.0, + 800.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 282.0, + 782.0, + 298.0, + 782.0, + 298.0, + 866.0, + 282.0, + 866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 511.0, + 785.0, + 514.0, + 785.0, + 514.0, + 814.0, + 511.0, + 814.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 511.0, + 803.0, + 597.0, + 803.0, + 597.0, + 844.0, + 511.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 689.0, + 803.0, + 999.0, + 803.0, + 999.0, + 844.0, + 689.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1092.0, + 803.0, + 1405.0, + 803.0, + 1405.0, + 844.0, + 1092.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 844.0, + 348.0, + 844.0, + 348.0, + 896.0, + 288.0, + 896.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 532.0, + 844.0, + 1408.0, + 844.0, + 1408.0, + 896.0, + 532.0, + 896.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 886.0, + 499.0, + 886.0, + 499.0, + 927.0, + 293.0, + 927.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 592.0, + 886.0, + 897.0, + 886.0, + 897.0, + 927.0, + 592.0, + 927.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1003.0, + 886.0, + 1088.0, + 886.0, + 1088.0, + 927.0, + 1003.0, + 927.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1153.0, + 886.0, + 1166.0, + 886.0, + 1166.0, + 927.0, + 1153.0, + 927.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1375.0, + 368.0, + 1375.0, + 368.0, + 1417.0, + 292.0, + 1417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 678.0, + 1375.0, + 725.0, + 1375.0, + 725.0, + 1417.0, + 678.0, + 1417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 892.0, + 1375.0, + 1405.0, + 1375.0, + 1405.0, + 1417.0, + 892.0, + 1417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 1404.0, + 954.0, + 1404.0, + 954.0, + 1459.0, + 288.0, + 1459.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1010.0, + 1404.0, + 1411.0, + 1404.0, + 1411.0, + 1459.0, + 1010.0, + 1459.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1443.0, + 1260.0, + 1443.0, + 1260.0, + 1479.0, + 293.0, + 1479.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1228.0, + 1183.0, + 1228.0, + 1183.0, + 1272.0, + 293.0, + 1272.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1335.0, + 1228.0, + 1405.0, + 1228.0, + 1405.0, + 1272.0, + 1335.0, + 1272.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1266.0, + 786.0, + 1266.0, + 786.0, + 1302.0, + 295.0, + 1302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1129.0, + 369.0, + 1129.0, + 369.0, + 1180.0, + 292.0, + 1180.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 682.0, + 1129.0, + 1294.0, + 1129.0, + 1294.0, + 1180.0, + 682.0, + 1180.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1320.0, + 1129.0, + 1407.0, + 1129.0, + 1407.0, + 1180.0, + 1320.0, + 1180.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1166.0, + 623.0, + 1166.0, + 623.0, + 1206.0, + 293.0, + 1206.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 201.0, + 1153.0, + 201.0, + 1153.0, + 239.0, + 294.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1183.0, + 201.0, + 1405.0, + 201.0, + 1405.0, + 239.0, + 1183.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 226.0, + 435.0, + 226.0, + 435.0, + 273.0, + 291.0, + 273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 463.0, + 226.0, + 861.0, + 226.0, + 861.0, + 273.0, + 463.0, + 273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1012.0, + 226.0, + 1100.0, + 226.0, + 1100.0, + 273.0, + 1012.0, + 273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1199.0, + 226.0, + 1406.0, + 226.0, + 1406.0, + 273.0, + 1199.0, + 273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 264.0, + 729.0, + 264.0, + 729.0, + 299.0, + 294.0, + 299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 804.0, + 264.0, + 1289.0, + 264.0, + 1289.0, + 299.0, + 804.0, + 299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1347.0, + 264.0, + 1406.0, + 264.0, + 1406.0, + 299.0, + 1347.0, + 299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 291.0, + 411.0, + 291.0, + 411.0, + 329.0, + 292.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 444.0, + 291.0, + 1265.0, + 291.0, + 1265.0, + 329.0, + 444.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1370.0, + 291.0, + 1408.0, + 291.0, + 1408.0, + 329.0, + 1370.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 320.0, + 426.0, + 320.0, + 426.0, + 361.0, + 292.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 528.0, + 320.0, + 695.0, + 320.0, + 695.0, + 361.0, + 528.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1010.0, + 320.0, + 1099.0, + 320.0, + 1099.0, + 361.0, + 1010.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1205.0, + 320.0, + 1406.0, + 320.0, + 1406.0, + 361.0, + 1205.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 351.0, + 931.0, + 351.0, + 931.0, + 394.0, + 290.0, + 394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 957.0, + 351.0, + 1407.0, + 351.0, + 1407.0, + 394.0, + 957.0, + 394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 383.0, + 343.0, + 383.0, + 343.0, + 426.0, + 288.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 376.0, + 383.0, + 808.0, + 383.0, + 808.0, + 426.0, + 376.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1000.0, + 383.0, + 1033.0, + 383.0, + 1033.0, + 426.0, + 1000.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1327.0, + 383.0, + 1409.0, + 383.0, + 1409.0, + 426.0, + 1327.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 417.0, + 297.0, + 417.0, + 297.0, + 454.0, + 292.0, + 454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 399.0, + 417.0, + 448.0, + 417.0, + 448.0, + 454.0, + 399.0, + 454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 479.0, + 417.0, + 1406.0, + 417.0, + 1406.0, + 454.0, + 479.0, + 454.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 448.0, + 1405.0, + 448.0, + 1405.0, + 484.0, + 292.0, + 484.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 475.0, + 297.0, + 475.0, + 297.0, + 518.0, + 292.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 342.0, + 475.0, + 563.0, + 475.0, + 563.0, + 518.0, + 342.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 593.0, + 475.0, + 975.0, + 475.0, + 975.0, + 518.0, + 593.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1140.0, + 475.0, + 1225.0, + 475.0, + 1225.0, + 518.0, + 1140.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1290.0, + 475.0, + 1407.0, + 475.0, + 1407.0, + 518.0, + 1290.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 509.0, + 736.0, + 509.0, + 736.0, + 544.0, + 295.0, + 544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 800.0, + 509.0, + 812.0, + 509.0, + 812.0, + 544.0, + 800.0, + 544.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 3, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1361, + 1406, + 1361, + 1406, + 1698, + 297, + 1698 + ], + "score": 0.984 + }, + { + "category_id": 1, + "poly": [ + 298, + 1793, + 1404, + 1793, + 1404, + 2007, + 298, + 2007 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 298, + 618, + 1405, + 618, + 1405, + 901, + 298, + 901 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 202, + 1404, + 202, + 1404, + 524, + 297, + 524 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 1143, + 1404, + 1143, + 1404, + 1327, + 298, + 1327 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 298, + 1016, + 1404, + 1016, + 1404, + 1108, + 298, + 1108 + ], + "score": 0.975 + }, + { + "category_id": 0, + "poly": [ + 298, + 561, + 974, + 561, + 974, + 595, + 298, + 595 + ], + "score": 0.933 + }, + { + "category_id": 0, + "poly": [ + 298, + 945, + 532, + 945, + 532, + 983, + 298, + 983 + ], + "score": 0.912 + }, + { + "category_id": 0, + "poly": [ + 299, + 1735, + 599, + 1735, + 599, + 1769, + 299, + 1769 + ], + "score": 0.911 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 858, + 2061, + 858, + 2085, + 841, + 2085 + ], + "score": 0.715 + }, + { + "category_id": 13, + "poly": [ + 655, + 805, + 756, + 805, + 756, + 842, + 655, + 842 + ], + "score": 0.93, + "latex": "h _ { \\theta } ^ { c } = h _ { \\psi } ^ { c }" + }, + { + "category_id": 13, + "poly": [ + 508, + 805, + 604, + 805, + 604, + 839, + 508, + 839 + ], + "score": 0.93, + "latex": "f _ { \\psi } = f _ { \\theta }" + }, + { + "category_id": 13, + "poly": [ + 593, + 484, + 934, + 484, + 934, + 527, + 593, + 527 + ], + "score": 0.92, + "latex": "r _ { i , t } ^ { j } = \\mathrm { \\ i n a x } ( \\ell _ { t } ^ { j } / \\sigma _ { r } - \\mu _ { \\ell / \\sigma _ { r } } , 0 )" + }, + { + "category_id": 13, + "poly": [ + 779, + 418, + 982, + 418, + 982, + 462, + 779, + 462 + ], + "score": 0.91, + "latex": "( ( \\boldsymbol { \\underline { { \\ell } } } _ { t } ^ { j } / \\sigma _ { r } ) _ { t = 0 } ^ { T - 2 } ) _ { j = 0 } ^ { B - 1 }" + }, + { + "category_id": 13, + "poly": [ + 1010, + 1236, + 1103, + 1236, + 1103, + 1265, + 1010, + 1265 + ], + "score": 0.91, + "latex": "8 4 \\times 8 4" + }, + { + "category_id": 13, + "poly": [ + 945, + 388, + 1010, + 388, + 1010, + 420, + 945, + 420 + ], + "score": 0.9, + "latex": "\\mu _ { \\ell / \\sigma _ { r } }" + }, + { + "category_id": 13, + "poly": [ + 1149, + 712, + 1181, + 712, + 1181, + 745, + 1149, + 745 + ], + "score": 0.89, + "latex": "f _ { \\psi }" + }, + { + "category_id": 13, + "poly": [ + 1081, + 429, + 1147, + 429, + 1147, + 459, + 1081, + 459 + ], + "score": 0.88, + "latex": "\\mu _ { \\ell / \\sigma _ { r } }" + }, + { + "category_id": 13, + "poly": [ + 1338, + 711, + 1373, + 711, + 1373, + 748, + 1338, + 748 + ], + "score": 0.88, + "latex": "h _ { \\psi } ^ { c }" + }, + { + "category_id": 13, + "poly": [ + 925, + 1946, + 1008, + 1946, + 1008, + 1973, + 925, + 1973 + ], + "score": 0.87, + "latex": "K = 1" + }, + { + "category_id": 13, + "poly": [ + 394, + 1947, + 469, + 1947, + 469, + 1974, + 394, + 1974 + ], + "score": 0.87, + "latex": "\\alpha = 1" + }, + { + "category_id": 13, + "poly": [ + 886, + 776, + 909, + 776, + 909, + 806, + 886, + 806 + ], + "score": 0.86, + "latex": "\\psi" + }, + { + "category_id": 13, + "poly": [ + 430, + 748, + 466, + 748, + 466, + 777, + 430, + 777 + ], + "score": 0.86, + "latex": "\\pi _ { \\psi }" + }, + { + "category_id": 13, + "poly": [ + 659, + 748, + 691, + 748, + 691, + 778, + 659, + 778 + ], + "score": 0.86, + "latex": "v _ { \\psi }" + }, + { + "category_id": 13, + "poly": [ + 1384, + 746, + 1402, + 746, + 1402, + 772, + 1384, + 772 + ], + "score": 0.76, + "latex": "\\theta" + }, + { + "category_id": 13, + "poly": [ + 575, + 1947, + 611, + 1947, + 611, + 1974, + 575, + 1974 + ], + "score": 0.74, + "latex": "= I" + }, + { + "category_id": 13, + "poly": [ + 553, + 1825, + 581, + 1825, + 581, + 1852, + 553, + 1852 + ], + "score": 0.74, + "latex": "K" + }, + { + "category_id": 13, + "poly": [ + 1376, + 1801, + 1397, + 1801, + 1397, + 1821, + 1376, + 1821 + ], + "score": 0.65, + "latex": "\\alpha" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 555.0, + 977.0, + 555.0, + 977.0, + 604.0, + 292.0, + 604.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 940.0, + 537.0, + 940.0, + 537.0, + 991.0, + 289.0, + 991.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1728.0, + 604.0, + 1728.0, + 604.0, + 1779.0, + 291.0, + 1779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2058.0, + 861.0, + 2058.0, + 861.0, + 2093.0, + 839.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1361.0, + 1406.0, + 1361.0, + 1406.0, + 1398.0, + 294.0, + 1398.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1394.0, + 1408.0, + 1394.0, + 1408.0, + 1429.0, + 295.0, + 1429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1423.0, + 1406.0, + 1423.0, + 1406.0, + 1460.0, + 292.0, + 1460.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1453.0, + 1406.0, + 1453.0, + 1406.0, + 1489.0, + 294.0, + 1489.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1484.0, + 1408.0, + 1484.0, + 1408.0, + 1522.0, + 292.0, + 1522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1515.0, + 1407.0, + 1515.0, + 1407.0, + 1551.0, + 292.0, + 1551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1544.0, + 1406.0, + 1544.0, + 1406.0, + 1578.0, + 294.0, + 1578.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1574.0, + 1409.0, + 1574.0, + 1409.0, + 1611.0, + 292.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1601.0, + 1408.0, + 1601.0, + 1408.0, + 1643.0, + 292.0, + 1643.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1634.0, + 1406.0, + 1634.0, + 1406.0, + 1671.0, + 294.0, + 1671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1665.0, + 1373.0, + 1665.0, + 1373.0, + 1702.0, + 294.0, + 1702.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1790.0, + 1375.0, + 1790.0, + 1375.0, + 1831.0, + 293.0, + 1831.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1398.0, + 1790.0, + 1408.0, + 1790.0, + 1408.0, + 1831.0, + 1398.0, + 1831.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1823.0, + 552.0, + 1823.0, + 552.0, + 1857.0, + 294.0, + 1857.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 582.0, + 1823.0, + 1404.0, + 1823.0, + 1404.0, + 1857.0, + 582.0, + 1857.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1852.0, + 1406.0, + 1852.0, + 1406.0, + 1889.0, + 293.0, + 1889.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1885.0, + 1406.0, + 1885.0, + 1406.0, + 1920.0, + 293.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1913.0, + 1405.0, + 1913.0, + 1405.0, + 1948.0, + 294.0, + 1948.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1944.0, + 393.0, + 1944.0, + 393.0, + 1980.0, + 292.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 470.0, + 1944.0, + 574.0, + 1944.0, + 574.0, + 1980.0, + 470.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 612.0, + 1944.0, + 924.0, + 1944.0, + 924.0, + 1980.0, + 612.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1009.0, + 1944.0, + 1405.0, + 1944.0, + 1405.0, + 1980.0, + 1009.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1976.0, + 1404.0, + 1976.0, + 1404.0, + 2010.0, + 293.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 618.0, + 1405.0, + 618.0, + 1405.0, + 655.0, + 292.0, + 655.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 649.0, + 1403.0, + 649.0, + 1403.0, + 683.0, + 293.0, + 683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 682.0, + 1405.0, + 682.0, + 1405.0, + 715.0, + 295.0, + 715.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 708.0, + 1148.0, + 708.0, + 1148.0, + 748.0, + 292.0, + 748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1182.0, + 708.0, + 1337.0, + 708.0, + 1337.0, + 748.0, + 1182.0, + 748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1374.0, + 708.0, + 1406.0, + 708.0, + 1406.0, + 748.0, + 1374.0, + 748.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 744.0, + 429.0, + 744.0, + 429.0, + 779.0, + 293.0, + 779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 467.0, + 744.0, + 658.0, + 744.0, + 658.0, + 779.0, + 467.0, + 779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 692.0, + 744.0, + 1383.0, + 744.0, + 1383.0, + 779.0, + 692.0, + 779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 774.0, + 885.0, + 774.0, + 885.0, + 807.0, + 295.0, + 807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 910.0, + 774.0, + 1406.0, + 774.0, + 1406.0, + 807.0, + 910.0, + 807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 803.0, + 507.0, + 803.0, + 507.0, + 843.0, + 292.0, + 843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 605.0, + 803.0, + 654.0, + 803.0, + 654.0, + 843.0, + 605.0, + 843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 757.0, + 803.0, + 1408.0, + 803.0, + 1408.0, + 843.0, + 757.0, + 843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 837.0, + 1405.0, + 837.0, + 1405.0, + 871.0, + 293.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 868.0, + 952.0, + 868.0, + 952.0, + 902.0, + 296.0, + 902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 202.0, + 1404.0, + 202.0, + 1404.0, + 238.0, + 295.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 232.0, + 1406.0, + 232.0, + 1406.0, + 268.0, + 294.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 264.0, + 1405.0, + 264.0, + 1405.0, + 297.0, + 295.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 294.0, + 1405.0, + 294.0, + 1405.0, + 327.0, + 296.0, + 327.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 323.0, + 1406.0, + 323.0, + 1406.0, + 359.0, + 292.0, + 359.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 354.0, + 1402.0, + 354.0, + 1402.0, + 387.0, + 295.0, + 387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 379.0, + 944.0, + 379.0, + 944.0, + 426.0, + 290.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1011.0, + 379.0, + 1409.0, + 379.0, + 1409.0, + 426.0, + 1011.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 454.0, + 1405.0, + 454.0, + 1405.0, + 490.0, + 293.0, + 490.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 483.0, + 592.0, + 483.0, + 592.0, + 529.0, + 291.0, + 529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 935.0, + 483.0, + 947.0, + 483.0, + 947.0, + 529.0, + 935.0, + 529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 279.0, + 394.5, + 1419.0, + 394.5, + 1419.0, + 480.5, + 279.0, + 480.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1141.0, + 1405.0, + 1141.0, + 1405.0, + 1181.0, + 293.0, + 1181.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1174.0, + 1405.0, + 1174.0, + 1405.0, + 1210.0, + 294.0, + 1210.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1203.0, + 1406.0, + 1203.0, + 1406.0, + 1240.0, + 292.0, + 1240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1235.0, + 1009.0, + 1235.0, + 1009.0, + 1270.0, + 293.0, + 1270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1104.0, + 1235.0, + 1404.0, + 1235.0, + 1404.0, + 1270.0, + 1104.0, + 1270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1264.0, + 1409.0, + 1264.0, + 1409.0, + 1302.0, + 292.0, + 1302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1295.0, + 1234.0, + 1295.0, + 1234.0, + 1330.0, + 293.0, + 1330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1013.0, + 1408.0, + 1013.0, + 1408.0, + 1052.0, + 293.0, + 1052.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1045.0, + 1405.0, + 1045.0, + 1405.0, + 1081.0, + 295.0, + 1081.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1078.0, + 1308.0, + 1078.0, + 1308.0, + 1111.0, + 296.0, + 1111.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 4, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1793, + 1404, + 1793, + 1404, + 2008, + 298, + 2008 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 808, + 1405, + 808, + 1405, + 1023, + 297, + 1023 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 1223, + 1405, + 1223, + 1405, + 1378, + 298, + 1378 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 298, + 1392, + 1405, + 1392, + 1405, + 1556, + 298, + 1556 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 299, + 1061, + 1404, + 1061, + 1404, + 1185, + 299, + 1185 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 298, + 1656, + 1403, + 1656, + 1403, + 1780, + 298, + 1780 + ], + "score": 0.977 + }, + { + "category_id": 3, + "poly": [ + 356, + 197, + 1346, + 197, + 1346, + 584, + 356, + 584 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 299, + 733, + 1399, + 733, + 1399, + 795, + 299, + 795 + ], + "score": 0.946 + }, + { + "category_id": 4, + "poly": [ + 298, + 605, + 1402, + 605, + 1402, + 667, + 298, + 667 + ], + "score": 0.945 + }, + { + "category_id": 0, + "poly": [ + 298, + 1597, + 520, + 1597, + 520, + 1630, + 298, + 1630 + ], + "score": 0.914 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 858, + 2062, + 858, + 2085, + 840, + 2085 + ], + "score": 0.789 + }, + { + "category_id": 13, + "poly": [ + 479, + 841, + 667, + 841, + 667, + 873, + 479, + 873 + ], + "score": 0.93, + "latex": "\\boldsymbol { r } _ { t } = \\boldsymbol { r } _ { e , t } + \\lambda \\boldsymbol { r } _ { i , t }" + }, + { + "category_id": 13, + "poly": [ + 296, + 1452, + 651, + 1452, + 651, + 1498, + 296, + 1498 + ], + "score": 0.92, + "latex": "\\begin{array} { r } { \\mathtt { H N S } ( t ) = \\frac { \\mathtt { A g e n t } _ { \\mathtt { s c o r e } } ( t ) - \\mathtt { R a n d o m } _ { \\mathtt { s c o r e } } } { \\mathtt { H u m a n } _ { \\mathtt { s c o r e } } - \\mathtt { R a n d o m } _ { \\mathtt { s c o r e } } } } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 1140, + 1459, + 1365, + 1459, + 1365, + 1492, + 1140, + 1492 + ], + "score": 0.91, + "latex": "\\mathrm { H N S } = \\operatorname* { m a x } _ { t } \\mathrm { H N S } ( t )" + }, + { + "category_id": 13, + "poly": [ + 832, + 874, + 872, + 874, + 872, + 903, + 832, + 903 + ], + "score": 0.91, + "latex": "{ r } _ { i , t }" + }, + { + "category_id": 13, + "poly": [ + 344, + 1688, + 444, + 1688, + 444, + 1717, + 344, + 1717 + ], + "score": 0.9, + "latex": "\\lambda = 0 . 1" + }, + { + "category_id": 13, + "poly": [ + 1257, + 1658, + 1343, + 1658, + 1343, + 1686, + 1257, + 1686 + ], + "score": 0.9, + "latex": "K = 8" + }, + { + "category_id": 13, + "poly": [ + 870, + 1659, + 985, + 1659, + 985, + 1687, + 870, + 1687 + ], + "score": 0.89, + "latex": "\\alpha = 0 . 9 9" + }, + { + "category_id": 13, + "poly": [ + 814, + 964, + 854, + 964, + 854, + 994, + 814, + 994 + ], + "score": 0.88, + "latex": "{ \\boldsymbol { r } } _ { i , t }" + }, + { + "category_id": 13, + "poly": [ + 297, + 874, + 340, + 874, + 340, + 903, + 297, + 903 + ], + "score": 0.87, + "latex": "{ r } _ { e , t }" + }, + { + "category_id": 13, + "poly": [ + 908, + 1392, + 1300, + 1392, + 1300, + 1426, + 908, + 1426 + ], + "score": 0.87, + "latex": "\\begin{array} { r } { \\mathsf { A g e n t } _ { \\mathsf { s c o r e } } = \\operatorname* { m a x } _ { t } \\mathsf { A g e n t } _ { \\mathsf { s c o r e } } ( t ) } \\end{array}" + }, + { + "category_id": 13, + "poly": [ + 552, + 1255, + 711, + 1255, + 711, + 1289, + 552, + 1289 + ], + "score": 0.87, + "latex": "\\mathtt { A g e n t } _ { \\mathtt { s c o r e } } ( t )" + }, + { + "category_id": 13, + "poly": [ + 542, + 1916, + 576, + 1916, + 576, + 1943, + 542, + 1943 + ], + "score": 0.72, + "latex": "= I" + }, + { + "category_id": 13, + "poly": [ + 1383, + 1425, + 1397, + 1425, + 1397, + 1449, + 1383, + 1449 + ], + "score": 0.72, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 377, + 1288, + 393, + 1288, + 393, + 1313, + 377, + 1313 + ], + "score": 0.69, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 1131, + 871, + 1150, + 871, + 1150, + 897, + 1131, + 897 + ], + "score": 0.66, + "latex": "\\lambda" + }, + { + "category_id": 13, + "poly": [ + 527, + 1258, + 542, + 1258, + 542, + 1284, + 527, + 1284 + ], + "score": 0.51, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 376, + 604, + 438, + 604, + 438, + 634, + 376, + 634 + ], + "score": 0.49, + "latex": "2 \\colon 1 ^ { \\mathrm { s t } }" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 203.0, + 514.0, + 203.0, + 514.0, + 223.0, + 450.0, + 223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 683.0, + 201.0, + 770.0, + 201.0, + 770.0, + 222.0, + 683.0, + 222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 945.0, + 203.0, + 1004.0, + 203.0, + 1004.0, + 223.0, + 945.0, + 223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1167.0, + 200.0, + 1280.0, + 200.0, + 1280.0, + 222.0, + 1167.0, + 222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 448.0, + 395.0, + 513.0, + 395.0, + 513.0, + 415.0, + 448.0, + 415.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 657.0, + 395.0, + 799.0, + 395.0, + 799.0, + 417.0, + 657.0, + 417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 859.0, + 397.0, + 1022.0, + 397.0, + 1022.0, + 479.0, + 859.0, + 479.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1144.0, + 393.0, + 1297.0, + 393.0, + 1297.0, + 419.0, + 1144.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 603.0, + 375.0, + 603.0, + 375.0, + 639.0, + 296.0, + 639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 439.0, + 603.0, + 1404.0, + 603.0, + 1404.0, + 639.0, + 439.0, + 639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 635.0, + 1403.0, + 635.0, + 1403.0, + 670.0, + 296.0, + 670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1593.0, + 524.0, + 1593.0, + 524.0, + 1634.0, + 294.0, + 1634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2060.0, + 861.0, + 2060.0, + 861.0, + 2091.0, + 840.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1793.0, + 1404.0, + 1793.0, + 1404.0, + 1828.0, + 294.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1822.0, + 1404.0, + 1822.0, + 1404.0, + 1861.0, + 293.0, + 1861.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1854.0, + 1405.0, + 1854.0, + 1405.0, + 1891.0, + 292.0, + 1891.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1886.0, + 1407.0, + 1886.0, + 1407.0, + 1919.0, + 293.0, + 1919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1913.0, + 541.0, + 1913.0, + 541.0, + 1951.0, + 295.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 577.0, + 1913.0, + 1405.0, + 1913.0, + 1405.0, + 1951.0, + 577.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1946.0, + 1405.0, + 1946.0, + 1405.0, + 1980.0, + 294.0, + 1980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1974.0, + 1403.0, + 1974.0, + 1403.0, + 2011.0, + 295.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 807.0, + 1405.0, + 807.0, + 1405.0, + 844.0, + 294.0, + 844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 839.0, + 478.0, + 839.0, + 478.0, + 874.0, + 292.0, + 874.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 668.0, + 839.0, + 1405.0, + 839.0, + 1405.0, + 874.0, + 668.0, + 874.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 869.0, + 296.0, + 869.0, + 296.0, + 906.0, + 291.0, + 906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 341.0, + 869.0, + 831.0, + 869.0, + 831.0, + 906.0, + 341.0, + 906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 873.0, + 869.0, + 1130.0, + 869.0, + 1130.0, + 906.0, + 873.0, + 906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1151.0, + 869.0, + 1407.0, + 869.0, + 1407.0, + 906.0, + 1151.0, + 906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 898.0, + 1410.0, + 898.0, + 1410.0, + 939.0, + 292.0, + 939.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 930.0, + 1405.0, + 930.0, + 1405.0, + 966.0, + 292.0, + 966.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 960.0, + 813.0, + 960.0, + 813.0, + 995.0, + 294.0, + 995.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 855.0, + 960.0, + 1406.0, + 960.0, + 1406.0, + 995.0, + 855.0, + 995.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 992.0, + 977.0, + 992.0, + 977.0, + 1025.0, + 294.0, + 1025.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1223.0, + 1406.0, + 1223.0, + 1406.0, + 1261.0, + 293.0, + 1261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1250.0, + 526.0, + 1250.0, + 526.0, + 1299.0, + 291.0, + 1299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 543.0, + 1250.0, + 551.0, + 1250.0, + 551.0, + 1299.0, + 543.0, + 1299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 712.0, + 1250.0, + 1411.0, + 1250.0, + 1411.0, + 1299.0, + 712.0, + 1299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1283.0, + 376.0, + 1283.0, + 376.0, + 1320.0, + 293.0, + 1320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 1283.0, + 1407.0, + 1283.0, + 1407.0, + 1320.0, + 394.0, + 1320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1315.0, + 1405.0, + 1315.0, + 1405.0, + 1351.0, + 293.0, + 1351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1347.0, + 806.0, + 1347.0, + 806.0, + 1380.0, + 295.0, + 1380.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1301.0, + 1392.0, + 1407.0, + 1392.0, + 1407.0, + 1428.0, + 1301.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1423.0, + 1382.0, + 1423.0, + 1382.0, + 1453.0, + 296.0, + 1453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1398.0, + 1423.0, + 1407.0, + 1423.0, + 1407.0, + 1453.0, + 1398.0, + 1453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 286.0, + 1442.0, + 295.0, + 1442.0, + 295.0, + 1507.0, + 286.0, + 1507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 652.0, + 1442.0, + 1139.0, + 1442.0, + 1139.0, + 1507.0, + 652.0, + 1507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1366.0, + 1442.0, + 1414.0, + 1442.0, + 1414.0, + 1507.0, + 1366.0, + 1507.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1493.0, + 1406.0, + 1493.0, + 1406.0, + 1529.0, + 293.0, + 1529.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1524.0, + 731.0, + 1524.0, + 731.0, + 1555.0, + 294.0, + 1555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.25, + 1386.0, + 1047.25, + 1386.0, + 1047.25, + 1430.0, + 291.25, + 1430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1059.0, + 1405.0, + 1059.0, + 1405.0, + 1099.0, + 293.0, + 1099.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1094.0, + 1405.0, + 1094.0, + 1405.0, + 1127.0, + 295.0, + 1127.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1123.0, + 1408.0, + 1123.0, + 1408.0, + 1158.0, + 293.0, + 1158.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1155.0, + 1036.0, + 1155.0, + 1036.0, + 1187.0, + 294.0, + 1187.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1657.0, + 869.0, + 1657.0, + 869.0, + 1690.0, + 294.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 986.0, + 1657.0, + 1256.0, + 1657.0, + 1256.0, + 1690.0, + 986.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1344.0, + 1657.0, + 1405.0, + 1657.0, + 1405.0, + 1690.0, + 1344.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1688.0, + 343.0, + 1688.0, + 343.0, + 1720.0, + 294.0, + 1720.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 445.0, + 1688.0, + 1405.0, + 1688.0, + 1405.0, + 1720.0, + 445.0, + 1720.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1718.0, + 1405.0, + 1718.0, + 1405.0, + 1751.0, + 293.0, + 1751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1750.0, + 862.0, + 1750.0, + 862.0, + 1782.0, + 294.0, + 1782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 734.0, + 1403.0, + 734.0, + 1403.0, + 769.0, + 295.0, + 769.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 763.0, + 1392.0, + 763.0, + 1392.0, + 799.0, + 294.0, + 799.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 5, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1702, + 1404, + 1702, + 1404, + 2008, + 298, + 2008 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 298, + 966, + 913, + 966, + 913, + 1605, + 298, + 1605 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 298, + 810, + 1405, + 810, + 1405, + 932, + 298, + 932 + ], + "score": 0.969 + }, + { + "category_id": 3, + "poly": [ + 304, + 199, + 1392, + 199, + 1392, + 586, + 304, + 586 + ], + "score": 0.961 + }, + { + "category_id": 3, + "poly": [ + 939, + 1008, + 1397, + 1008, + 1397, + 1403, + 939, + 1403 + ], + "score": 0.961 + }, + { + "category_id": 4, + "poly": [ + 935, + 1425, + 1403, + 1425, + 1403, + 1517, + 935, + 1517 + ], + "score": 0.96 + }, + { + "category_id": 1, + "poly": [ + 296, + 1606, + 1400, + 1606, + 1400, + 1667, + 296, + 1667 + ], + "score": 0.948 + }, + { + "category_id": 4, + "poly": [ + 297, + 609, + 1404, + 609, + 1404, + 672, + 297, + 672 + ], + "score": 0.942 + }, + { + "category_id": 1, + "poly": [ + 299, + 734, + 1399, + 734, + 1399, + 796, + 299, + 796 + ], + "score": 0.934 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 858, + 2061, + 858, + 2084, + 841, + 2084 + ], + "score": 0.709 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2084, + 841, + 2084 + ], + "score": 0.092 + }, + { + "category_id": 13, + "poly": [ + 764, + 1887, + 798, + 1887, + 798, + 1913, + 764, + 1913 + ], + "score": 0.75, + "latex": "= I" + }, + { + "category_id": 13, + "poly": [ + 631, + 812, + 667, + 812, + 667, + 839, + 631, + 839 + ], + "score": 0.67, + "latex": "= I" + }, + { + "category_id": 13, + "poly": [ + 538, + 610, + 568, + 610, + 568, + 642, + 538, + 642 + ], + "score": 0.4, + "latex": "( t )" + }, + { + "category_id": 13, + "poly": [ + 476, + 609, + 569, + 609, + 569, + 642, + 476, + 642 + ], + "score": 0.31, + "latex": "{ \\mathrm { C H N S } } ( t )" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 205.0, + 375.0, + 205.0, + 375.0, + 238.0, + 331.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 923.0, + 205.0, + 967.0, + 205.0, + 967.0, + 238.0, + 923.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1342.0, + 227.0, + 1377.0, + 227.0, + 1377.0, + 250.0, + 1342.0, + 250.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1300.0, + 231.0, + 1350.0, + 231.0, + 1350.0, + 248.0, + 1300.0, + 248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 303.0, + 267.0, + 373.0, + 267.0, + 373.0, + 446.0, + 303.0, + 446.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 427.0, + 291.0, + 440.0, + 291.0, + 440.0, + 301.0, + 427.0, + 301.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 599.0, + 286.0, + 791.0, + 286.0, + 791.0, + 310.0, + 599.0, + 310.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 894.0, + 266.0, + 966.0, + 266.0, + 966.0, + 446.0, + 894.0, + 446.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1074.0, + 254.0, + 1124.0, + 254.0, + 1124.0, + 302.0, + 1074.0, + 302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1133.0, + 290.0, + 1146.0, + 290.0, + 1146.0, + 299.0, + 1133.0, + 299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 343.0, + 330.0, + 372.0, + 330.0, + 372.0, + 357.0, + 343.0, + 357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 933.0, + 325.0, + 966.0, + 325.0, + 966.0, + 354.0, + 933.0, + 354.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1270.0, + 334.0, + 1388.0, + 334.0, + 1388.0, + 367.0, + 1270.0, + 367.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 497.0, + 363.0, + 530.0, + 363.0, + 530.0, + 380.0, + 497.0, + 380.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 544.0, + 362.0, + 566.0, + 362.0, + 566.0, + 371.0, + 544.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 582.0, + 363.0, + 593.0, + 363.0, + 593.0, + 372.0, + 582.0, + 372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 638.0, + 364.0, + 757.0, + 364.0, + 757.0, + 400.0, + 638.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1269.0, + 356.0, + 1373.0, + 356.0, + 1373.0, + 401.0, + 1269.0, + 401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 343.0, + 393.0, + 372.0, + 393.0, + 372.0, + 419.0, + 343.0, + 419.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 638.0, + 394.0, + 793.0, + 394.0, + 793.0, + 429.0, + 638.0, + 429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 932.0, + 384.0, + 966.0, + 384.0, + 966.0, + 414.0, + 932.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1269.0, + 390.0, + 1384.0, + 390.0, + 1384.0, + 426.0, + 1269.0, + 426.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 638.0, + 422.0, + 690.0, + 422.0, + 690.0, + 458.0, + 638.0, + 458.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1270.0, + 415.0, + 1368.0, + 415.0, + 1368.0, + 455.0, + 1270.0, + 455.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 341.0, + 451.0, + 374.0, + 451.0, + 374.0, + 481.0, + 341.0, + 481.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 639.0, + 452.0, + 686.0, + 452.0, + 686.0, + 485.0, + 639.0, + 485.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 932.0, + 441.0, + 967.0, + 441.0, + 967.0, + 472.0, + 932.0, + 472.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1270.0, + 450.0, + 1360.0, + 450.0, + 1360.0, + 479.0, + 1270.0, + 479.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 638.0, + 479.0, + 676.0, + 479.0, + 676.0, + 514.0, + 638.0, + 514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 942.0, + 502.0, + 967.0, + 502.0, + 967.0, + 531.0, + 942.0, + 531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 359.0, + 532.0, + 401.0, + 532.0, + 401.0, + 565.0, + 359.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 532.0, + 490.0, + 532.0, + 490.0, + 565.0, + 450.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 540.0, + 531.0, + 581.0, + 531.0, + 581.0, + 565.0, + 540.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 628.0, + 531.0, + 670.0, + 531.0, + 670.0, + 565.0, + 628.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 717.0, + 531.0, + 759.0, + 531.0, + 759.0, + 565.0, + 717.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 952.0, + 532.0, + 993.0, + 532.0, + 993.0, + 565.0, + 952.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1041.0, + 532.0, + 1083.0, + 532.0, + 1083.0, + 565.0, + 1041.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1131.0, + 531.0, + 1172.0, + 531.0, + 1172.0, + 564.0, + 1131.0, + 564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1220.0, + 531.0, + 1261.0, + 531.0, + 1261.0, + 565.0, + 1220.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1310.0, + 531.0, + 1352.0, + 531.0, + 1352.0, + 565.0, + 1310.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 504.0, + 557.0, + 670.0, + 557.0, + 670.0, + 589.0, + 504.0, + 589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 735.0, + 555.0, + 800.0, + 555.0, + 800.0, + 588.0, + 735.0, + 588.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1096.0, + 555.0, + 1262.0, + 555.0, + 1262.0, + 589.0, + 1096.0, + 589.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1328.0, + 552.0, + 1394.0, + 552.0, + 1394.0, + 592.0, + 1328.0, + 592.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 447.0, + 213.5, + 568.0, + 213.5, + 568.0, + 250.5, + 447.0, + 250.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1102.0, + 230.0, + 1320.0, + 230.0, + 1320.0, + 293.0, + 1102.0, + 293.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 419.0, + 230.0, + 802.0, + 230.0, + 802.0, + 291.0, + 419.0, + 291.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.25, + 267.0, + 1082.25, + 267.0, + 1082.25, + 289.0, + 1028.25, + 289.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1018.0, + 277.0, + 1094.0, + 277.0, + 1094.0, + 317.0, + 1018.0, + 317.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1084.0, + 1006.0, + 1309.0, + 1006.0, + 1309.0, + 1042.0, + 1084.0, + 1042.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 963.0, + 1065.0, + 994.0, + 1065.0, + 994.0, + 1094.0, + 963.0, + 1094.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 939.0, + 1089.0, + 969.0, + 1089.0, + 969.0, + 1296.0, + 939.0, + 1296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 965.0, + 1134.0, + 993.0, + 1134.0, + 993.0, + 1159.0, + 965.0, + 1159.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 964.0, + 1200.0, + 992.0, + 1200.0, + 992.0, + 1225.0, + 964.0, + 1225.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 975.0, + 1268.0, + 991.0, + 1268.0, + 991.0, + 1290.0, + 975.0, + 1290.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1069.0, + 1270.0, + 1180.0, + 1270.0, + 1180.0, + 1304.0, + 1069.0, + 1304.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1252.0, + 1274.0, + 1294.0, + 1274.0, + 1294.0, + 1300.0, + 1252.0, + 1300.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1071.0, + 1300.0, + 1113.0, + 1300.0, + 1113.0, + 1327.0, + 1071.0, + 1327.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1252.0, + 1300.0, + 1390.0, + 1300.0, + 1390.0, + 1330.0, + 1252.0, + 1330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 974.0, + 1335.0, + 991.0, + 1335.0, + 991.0, + 1356.0, + 974.0, + 1356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 981.0, + 1351.0, + 1019.0, + 1351.0, + 1019.0, + 1382.0, + 981.0, + 1382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1049.0, + 1351.0, + 1088.0, + 1351.0, + 1088.0, + 1382.0, + 1049.0, + 1382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1117.0, + 1351.0, + 1157.0, + 1351.0, + 1157.0, + 1382.0, + 1117.0, + 1382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1184.0, + 1351.0, + 1225.0, + 1351.0, + 1225.0, + 1382.0, + 1184.0, + 1382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1254.0, + 1351.0, + 1293.0, + 1351.0, + 1293.0, + 1382.0, + 1254.0, + 1382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1322.0, + 1351.0, + 1362.0, + 1351.0, + 1362.0, + 1382.0, + 1322.0, + 1382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1117.0, + 1372.0, + 1277.0, + 1372.0, + 1277.0, + 1406.0, + 1117.0, + 1406.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1339.0, + 1372.0, + 1401.0, + 1372.0, + 1401.0, + 1405.0, + 1339.0, + 1405.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 933.0, + 1424.0, + 1405.0, + 1424.0, + 1405.0, + 1458.0, + 933.0, + 1458.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 932.0, + 1452.0, + 1405.0, + 1452.0, + 1405.0, + 1491.0, + 932.0, + 1491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 932.0, + 1484.0, + 1408.0, + 1484.0, + 1408.0, + 1520.0, + 932.0, + 1520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 607.0, + 475.0, + 607.0, + 475.0, + 643.0, + 296.0, + 643.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 570.0, + 607.0, + 1402.0, + 607.0, + 1402.0, + 643.0, + 570.0, + 643.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 639.0, + 1308.0, + 639.0, + 1308.0, + 675.0, + 295.0, + 675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2059.0, + 860.0, + 2059.0, + 860.0, + 2091.0, + 839.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2059.0, + 860.0, + 2059.0, + 860.0, + 2092.0, + 840.0, + 2092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1700.0, + 1405.0, + 1700.0, + 1405.0, + 1737.0, + 293.0, + 1737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1731.0, + 1408.0, + 1731.0, + 1408.0, + 1771.0, + 292.0, + 1771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1762.0, + 1405.0, + 1762.0, + 1405.0, + 1799.0, + 293.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1793.0, + 1404.0, + 1793.0, + 1404.0, + 1828.0, + 294.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1826.0, + 1405.0, + 1826.0, + 1405.0, + 1858.0, + 296.0, + 1858.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1855.0, + 1406.0, + 1855.0, + 1406.0, + 1891.0, + 294.0, + 1891.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1884.0, + 763.0, + 1884.0, + 763.0, + 1920.0, + 294.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 799.0, + 1884.0, + 1406.0, + 1884.0, + 1406.0, + 1920.0, + 799.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1915.0, + 1404.0, + 1915.0, + 1404.0, + 1950.0, + 294.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1946.0, + 1404.0, + 1946.0, + 1404.0, + 1982.0, + 296.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1976.0, + 704.0, + 1976.0, + 704.0, + 2011.0, + 294.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 967.0, + 916.0, + 967.0, + 916.0, + 1001.0, + 296.0, + 1001.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 997.0, + 915.0, + 997.0, + 915.0, + 1031.0, + 295.0, + 1031.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1028.0, + 915.0, + 1028.0, + 915.0, + 1063.0, + 294.0, + 1063.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1058.0, + 830.0, + 1058.0, + 830.0, + 1095.0, + 294.0, + 1095.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 863.0, + 1060.0, + 916.0, + 1060.0, + 916.0, + 1092.0, + 863.0, + 1092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1090.0, + 916.0, + 1090.0, + 916.0, + 1122.0, + 294.0, + 1122.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1120.0, + 914.0, + 1120.0, + 914.0, + 1154.0, + 295.0, + 1154.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1150.0, + 916.0, + 1150.0, + 916.0, + 1185.0, + 295.0, + 1185.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1181.0, + 915.0, + 1181.0, + 915.0, + 1215.0, + 293.0, + 1215.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1212.0, + 915.0, + 1212.0, + 915.0, + 1246.0, + 296.0, + 1246.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1241.0, + 916.0, + 1241.0, + 916.0, + 1275.0, + 296.0, + 1275.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1271.0, + 915.0, + 1271.0, + 915.0, + 1306.0, + 295.0, + 1306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1300.0, + 915.0, + 1300.0, + 915.0, + 1335.0, + 294.0, + 1335.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1334.0, + 915.0, + 1334.0, + 915.0, + 1365.0, + 294.0, + 1365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1361.0, + 915.0, + 1361.0, + 915.0, + 1397.0, + 294.0, + 1397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1394.0, + 914.0, + 1394.0, + 914.0, + 1425.0, + 296.0, + 1425.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1423.0, + 914.0, + 1423.0, + 914.0, + 1457.0, + 296.0, + 1457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1453.0, + 915.0, + 1453.0, + 915.0, + 1487.0, + 295.0, + 1487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1481.0, + 916.0, + 1481.0, + 916.0, + 1520.0, + 293.0, + 1520.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1512.0, + 916.0, + 1512.0, + 916.0, + 1550.0, + 294.0, + 1550.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1545.0, + 915.0, + 1545.0, + 915.0, + 1579.0, + 295.0, + 1579.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1575.0, + 915.0, + 1575.0, + 915.0, + 1609.0, + 295.0, + 1609.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 811.0, + 630.0, + 811.0, + 630.0, + 843.0, + 295.0, + 843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 668.0, + 811.0, + 1405.0, + 811.0, + 1405.0, + 843.0, + 668.0, + 843.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 836.0, + 1406.0, + 836.0, + 1406.0, + 878.0, + 291.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 871.0, + 1406.0, + 871.0, + 1406.0, + 907.0, + 293.0, + 907.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 900.0, + 1024.0, + 900.0, + 1024.0, + 938.0, + 293.0, + 938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1603.0, + 1405.0, + 1603.0, + 1405.0, + 1640.0, + 294.0, + 1640.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1635.0, + 1131.0, + 1635.0, + 1131.0, + 1670.0, + 295.0, + 1670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 733.0, + 1404.0, + 733.0, + 1404.0, + 769.0, + 296.0, + 769.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 762.0, + 1404.0, + 762.0, + 1404.0, + 800.0, + 293.0, + 800.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 6, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 258, + 1406, + 258, + 1406, + 715, + 298, + 715 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 1379, + 1404, + 1379, + 1404, + 1592, + 298, + 1592 + ], + "score": 0.979 + }, + { + "category_id": 3, + "poly": [ + 356, + 746, + 1349, + 746, + 1349, + 1231, + 356, + 1231 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 1687, + 1405, + 1687, + 1405, + 1810, + 298, + 1810 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 298, + 1822, + 1403, + 1822, + 1403, + 2008, + 298, + 2008 + ], + "score": 0.975 + }, + { + "category_id": 4, + "poly": [ + 299, + 1249, + 1398, + 1249, + 1398, + 1341, + 299, + 1341 + ], + "score": 0.955 + }, + { + "category_id": 0, + "poly": [ + 297, + 202, + 1012, + 202, + 1012, + 235, + 297, + 235 + ], + "score": 0.91 + }, + { + "category_id": 0, + "poly": [ + 299, + 1631, + 587, + 1631, + 587, + 1662, + 299, + 1662 + ], + "score": 0.903 + }, + { + "category_id": 2, + "poly": [ + 841, + 2062, + 858, + 2062, + 858, + 2085, + 841, + 2085 + ], + "score": 0.794 + }, + { + "category_id": 13, + "poly": [ + 993, + 624, + 1086, + 624, + 1086, + 653, + 993, + 653 + ], + "score": 0.92, + "latex": "1 4 \\times 1 4" + }, + { + "category_id": 13, + "poly": [ + 1196, + 625, + 1261, + 625, + 1261, + 653, + 1196, + 653 + ], + "score": 0.9, + "latex": "6 \\times 6" + }, + { + "category_id": 13, + "poly": [ + 844, + 1689, + 951, + 1689, + 951, + 1716, + 844, + 1716 + ], + "score": 0.89, + "latex": "\\alpha = 0 . 9 9" + }, + { + "category_id": 13, + "poly": [ + 297, + 1718, + 403, + 1718, + 403, + 1747, + 297, + 1747 + ], + "score": 0.89, + "latex": "\\lambda = 0 . 0 1" + }, + { + "category_id": 13, + "poly": [ + 1212, + 1688, + 1305, + 1688, + 1305, + 1716, + 1212, + 1716 + ], + "score": 0.89, + "latex": "K = 1 0" + }, + { + "category_id": 13, + "poly": [ + 1061, + 563, + 1166, + 563, + 1166, + 593, + 1061, + 593 + ], + "score": 0.88, + "latex": "( 8 4 \\times 8 4 )" + }, + { + "category_id": 13, + "poly": [ + 572, + 1947, + 606, + 1947, + 606, + 1974, + 572, + 1974 + ], + "score": 0.73, + "latex": "= I" + }, + { + "category_id": 13, + "poly": [ + 700, + 1824, + 793, + 1824, + 793, + 1856, + 700, + 1856 + ], + "score": 0.48, + "latex": "{ \\mathrm { C H N S } } ( t )" + }, + { + "category_id": 15, + "poly": [ + 1046.0, + 748.0, + 1217.0, + 748.0, + 1217.0, + 777.0, + 1046.0, + 777.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 872.0, + 806.0, + 916.0, + 806.0, + 916.0, + 829.0, + 872.0, + 829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1110.0, + 822.0, + 1119.0, + 822.0, + 1119.0, + 830.0, + 1110.0, + 830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 871.0, + 856.0, + 916.0, + 856.0, + 916.0, + 880.0, + 871.0, + 880.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1110.0, + 831.0, + 1209.0, + 831.0, + 1209.0, + 878.0, + 1110.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 872.0, + 906.0, + 916.0, + 906.0, + 916.0, + 929.0, + 872.0, + 929.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 852.0, + 930.0, + 879.0, + 930.0, + 879.0, + 1030.0, + 852.0, + 1030.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 876.0, + 955.0, + 916.0, + 955.0, + 916.0, + 979.0, + 876.0, + 979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 878.0, + 1007.0, + 915.0, + 1007.0, + 915.0, + 1027.0, + 878.0, + 1027.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 879.0, + 1058.0, + 915.0, + 1058.0, + 915.0, + 1079.0, + 879.0, + 1079.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 878.0, + 1106.0, + 916.0, + 1106.0, + 916.0, + 1130.0, + 878.0, + 1130.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 949.0, + 1107.0, + 970.0, + 1107.0, + 970.0, + 1119.0, + 949.0, + 1119.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1032.0, + 1123.0, + 1114.0, + 1123.0, + 1114.0, + 1146.0, + 1032.0, + 1146.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1212.0, + 1123.0, + 1245.0, + 1123.0, + 1245.0, + 1145.0, + 1212.0, + 1145.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 1141.0, + 1157.0, + 1141.0, + 1157.0, + 1169.0, + 1028.0, + 1169.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1209.0, + 1143.0, + 1289.0, + 1143.0, + 1289.0, + 1169.0, + 1209.0, + 1169.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 900.0, + 1160.0, + 914.0, + 1160.0, + 914.0, + 1175.0, + 900.0, + 1175.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 906.0, + 1191.0, + 934.0, + 1191.0, + 934.0, + 1212.0, + 906.0, + 1212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 980.0, + 1191.0, + 1008.0, + 1191.0, + 1008.0, + 1212.0, + 980.0, + 1212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1054.0, + 1190.0, + 1082.0, + 1190.0, + 1082.0, + 1212.0, + 1054.0, + 1212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1128.0, + 1190.0, + 1156.0, + 1190.0, + 1156.0, + 1212.0, + 1128.0, + 1212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1202.0, + 1191.0, + 1230.0, + 1191.0, + 1230.0, + 1212.0, + 1202.0, + 1212.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1274.0, + 1188.0, + 1305.0, + 1188.0, + 1305.0, + 1215.0, + 1274.0, + 1215.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1071.0, + 1207.0, + 1193.0, + 1207.0, + 1193.0, + 1232.0, + 1071.0, + 1232.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1305.0, + 1204.0, + 1351.0, + 1204.0, + 1351.0, + 1232.0, + 1305.0, + 1232.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 427.25, + 759.0, + 522.25, + 759.0, + 522.25, + 799.0, + 427.25, + 799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 998.0, + 847.0, + 1119.0, + 847.0, + 1119.0, + 892.5, + 998.0, + 892.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1002.0, + 884.5, + 1132.0, + 884.5, + 1132.0, + 925.5, + 1002.0, + 925.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1248.0, + 1406.0, + 1248.0, + 1406.0, + 1285.0, + 294.0, + 1285.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1282.0, + 1402.0, + 1282.0, + 1402.0, + 1312.0, + 295.0, + 1312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1312.0, + 852.0, + 1312.0, + 852.0, + 1344.0, + 294.0, + 1344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 201.0, + 1014.0, + 201.0, + 1014.0, + 238.0, + 294.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1629.0, + 590.0, + 1629.0, + 590.0, + 1665.0, + 295.0, + 1665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2059.0, + 860.0, + 2059.0, + 860.0, + 2089.0, + 839.0, + 2089.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 259.0, + 1404.0, + 259.0, + 1404.0, + 294.0, + 295.0, + 294.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 291.0, + 1406.0, + 291.0, + 1406.0, + 322.0, + 296.0, + 322.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 320.0, + 1407.0, + 320.0, + 1407.0, + 355.0, + 295.0, + 355.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 352.0, + 1406.0, + 352.0, + 1406.0, + 383.0, + 296.0, + 383.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 377.0, + 1409.0, + 377.0, + 1409.0, + 418.0, + 292.0, + 418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 412.0, + 1407.0, + 412.0, + 1407.0, + 447.0, + 295.0, + 447.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 442.0, + 1406.0, + 442.0, + 1406.0, + 475.0, + 292.0, + 475.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 473.0, + 1404.0, + 473.0, + 1404.0, + 504.0, + 295.0, + 504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 500.0, + 1408.0, + 500.0, + 1408.0, + 538.0, + 292.0, + 538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 531.0, + 1403.0, + 531.0, + 1403.0, + 566.0, + 295.0, + 566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 562.0, + 1060.0, + 562.0, + 1060.0, + 597.0, + 295.0, + 597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1167.0, + 562.0, + 1404.0, + 562.0, + 1404.0, + 597.0, + 1167.0, + 597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 591.0, + 1404.0, + 591.0, + 1404.0, + 627.0, + 292.0, + 627.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 623.0, + 992.0, + 623.0, + 992.0, + 658.0, + 295.0, + 658.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1087.0, + 623.0, + 1195.0, + 623.0, + 1195.0, + 658.0, + 1087.0, + 658.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1262.0, + 623.0, + 1404.0, + 623.0, + 1404.0, + 658.0, + 1262.0, + 658.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 652.0, + 1406.0, + 652.0, + 1406.0, + 689.0, + 293.0, + 689.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 684.0, + 1055.0, + 684.0, + 1055.0, + 718.0, + 293.0, + 718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1377.0, + 1405.0, + 1377.0, + 1405.0, + 1416.0, + 293.0, + 1416.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1411.0, + 1405.0, + 1411.0, + 1405.0, + 1445.0, + 293.0, + 1445.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1440.0, + 1404.0, + 1440.0, + 1404.0, + 1473.0, + 293.0, + 1473.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1470.0, + 1406.0, + 1470.0, + 1406.0, + 1506.0, + 291.0, + 1506.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1499.0, + 1407.0, + 1499.0, + 1407.0, + 1536.0, + 292.0, + 1536.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1531.0, + 1404.0, + 1531.0, + 1404.0, + 1565.0, + 293.0, + 1565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1561.0, + 959.0, + 1561.0, + 959.0, + 1594.0, + 295.0, + 1594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1688.0, + 843.0, + 1688.0, + 843.0, + 1721.0, + 295.0, + 1721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 952.0, + 1688.0, + 1211.0, + 1688.0, + 1211.0, + 1721.0, + 952.0, + 1721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1306.0, + 1688.0, + 1405.0, + 1688.0, + 1405.0, + 1721.0, + 1306.0, + 1721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1717.0, + 296.0, + 1717.0, + 296.0, + 1751.0, + 293.0, + 1751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 1717.0, + 1403.0, + 1717.0, + 1403.0, + 1751.0, + 404.0, + 1751.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1749.0, + 1410.0, + 1749.0, + 1410.0, + 1783.0, + 291.0, + 1783.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1780.0, + 1249.0, + 1780.0, + 1249.0, + 1812.0, + 295.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1823.0, + 699.0, + 1823.0, + 699.0, + 1859.0, + 294.0, + 1859.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 794.0, + 1823.0, + 1407.0, + 1823.0, + 1407.0, + 1859.0, + 794.0, + 1859.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1855.0, + 1405.0, + 1855.0, + 1405.0, + 1891.0, + 293.0, + 1891.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1882.0, + 1405.0, + 1882.0, + 1405.0, + 1923.0, + 291.0, + 1923.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1912.0, + 1405.0, + 1912.0, + 1405.0, + 1949.0, + 292.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1943.0, + 571.0, + 1943.0, + 571.0, + 1981.0, + 292.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 607.0, + 1943.0, + 1405.0, + 1943.0, + 1405.0, + 1981.0, + 607.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1975.0, + 1405.0, + 1975.0, + 1405.0, + 2009.0, + 294.0, + 2009.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 7, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1248, + 1405, + 1248, + 1405, + 1553, + 298, + 1553 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 1021, + 1404, + 1021, + 1404, + 1234, + 298, + 1234 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 299, + 1587, + 1403, + 1587, + 1403, + 1740, + 299, + 1740 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 299, + 1854, + 1404, + 1854, + 1404, + 2007, + 299, + 2007 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 854, + 1404, + 854, + 1404, + 1007, + 298, + 1007 + ], + "score": 0.973 + }, + { + "category_id": 3, + "poly": [ + 308, + 331, + 1393, + 331, + 1393, + 731, + 308, + 731 + ], + "score": 0.969 + }, + { + "category_id": 1, + "poly": [ + 295, + 203, + 1404, + 203, + 1404, + 296, + 295, + 296 + ], + "score": 0.956 + }, + { + "category_id": 4, + "poly": [ + 295, + 755, + 1402, + 755, + 1402, + 819, + 295, + 819 + ], + "score": 0.934 + }, + { + "category_id": 0, + "poly": [ + 298, + 1784, + 509, + 1784, + 509, + 1821, + 298, + 1821 + ], + "score": 0.905 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 858, + 2062, + 858, + 2084, + 840, + 2084 + ], + "score": 0.797 + }, + { + "category_id": 13, + "poly": [ + 533, + 756, + 564, + 756, + 564, + 788, + 533, + 788 + ], + "score": 0.47, + "latex": "( t )" + }, + { + "category_id": 15, + "poly": [ + 428.0, + 343.0, + 552.0, + 343.0, + 552.0, + 381.0, + 428.0, + 381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1021.0, + 342.0, + 1145.0, + 342.0, + 1145.0, + 382.0, + 1021.0, + 382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 335.0, + 375.0, + 354.0, + 375.0, + 354.0, + 400.0, + 335.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 427.0, + 375.0, + 480.0, + 375.0, + 480.0, + 410.0, + 427.0, + 410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 927.0, + 375.0, + 947.0, + 375.0, + 947.0, + 400.0, + 927.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1020.0, + 372.0, + 1128.0, + 372.0, + 1128.0, + 414.0, + 1020.0, + 414.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 426.0, + 404.0, + 479.0, + 404.0, + 479.0, + 438.0, + 426.0, + 438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1021.0, + 405.0, + 1118.0, + 405.0, + 1118.0, + 437.0, + 1021.0, + 437.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 301.0, + 430.0, + 356.0, + 430.0, + 356.0, + 577.0, + 301.0, + 577.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 428.0, + 435.0, + 464.0, + 435.0, + 464.0, + 469.0, + 428.0, + 469.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 894.0, + 430.0, + 947.0, + 430.0, + 947.0, + 577.0, + 894.0, + 577.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1021.0, + 435.0, + 1141.0, + 435.0, + 1141.0, + 470.0, + 1021.0, + 470.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 336.0, + 509.0, + 354.0, + 509.0, + 354.0, + 532.0, + 336.0, + 532.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 927.0, + 509.0, + 947.0, + 509.0, + 947.0, + 532.0, + 927.0, + 532.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 574.0, + 357.0, + 574.0, + 357.0, + 602.0, + 333.0, + 602.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 927.0, + 575.0, + 947.0, + 575.0, + 947.0, + 601.0, + 927.0, + 601.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 336.0, + 644.0, + 354.0, + 644.0, + 354.0, + 667.0, + 336.0, + 667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 927.0, + 644.0, + 948.0, + 644.0, + 948.0, + 667.0, + 927.0, + 667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 957.0, + 651.0, + 973.0, + 651.0, + 973.0, + 664.0, + 957.0, + 664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 643.0, + 1027.0, + 643.0, + 1027.0, + 656.0, + 1006.0, + 656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 682.0, + 374.0, + 682.0, + 374.0, + 705.0, + 354.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 441.0, + 681.0, + 460.0, + 681.0, + 460.0, + 706.0, + 441.0, + 706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 525.0, + 680.0, + 548.0, + 680.0, + 548.0, + 706.0, + 525.0, + 706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 614.0, + 681.0, + 633.0, + 681.0, + 633.0, + 706.0, + 614.0, + 706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 699.0, + 681.0, + 718.0, + 681.0, + 718.0, + 706.0, + 699.0, + 706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 947.0, + 682.0, + 967.0, + 682.0, + 967.0, + 705.0, + 947.0, + 705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1031.0, + 680.0, + 1054.0, + 680.0, + 1054.0, + 709.0, + 1031.0, + 709.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1116.0, + 680.0, + 1141.0, + 680.0, + 1141.0, + 706.0, + 1116.0, + 706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1204.0, + 680.0, + 1227.0, + 680.0, + 1227.0, + 708.0, + 1204.0, + 708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1290.0, + 681.0, + 1310.0, + 681.0, + 1310.0, + 706.0, + 1290.0, + 706.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 492.0, + 699.0, + 667.0, + 699.0, + 667.0, + 737.0, + 492.0, + 737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 752.0, + 700.0, + 804.0, + 700.0, + 804.0, + 735.0, + 752.0, + 735.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1084.0, + 699.0, + 1258.0, + 699.0, + 1258.0, + 737.0, + 1084.0, + 737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1344.0, + 700.0, + 1395.0, + 700.0, + 1395.0, + 737.0, + 1344.0, + 737.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1167.0, + 578.0, + 1211.0, + 578.0, + 1211.0, + 595.0, + 1167.0, + 595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1131.25, + 589.0, + 1167.25, + 589.0, + 1167.25, + 607.0, + 1131.25, + 607.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 752.0, + 532.0, + 752.0, + 532.0, + 791.0, + 294.0, + 791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 565.0, + 752.0, + 1405.0, + 752.0, + 1405.0, + 791.0, + 565.0, + 791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 784.0, + 1226.0, + 784.0, + 1226.0, + 822.0, + 294.0, + 822.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1780.0, + 513.0, + 1780.0, + 513.0, + 1828.0, + 290.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2060.0, + 860.0, + 2060.0, + 860.0, + 2090.0, + 839.0, + 2090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1250.0, + 1403.0, + 1250.0, + 1403.0, + 1282.0, + 296.0, + 1282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1280.0, + 1405.0, + 1280.0, + 1405.0, + 1312.0, + 295.0, + 1312.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1309.0, + 1406.0, + 1309.0, + 1406.0, + 1345.0, + 293.0, + 1345.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1338.0, + 1408.0, + 1338.0, + 1408.0, + 1378.0, + 293.0, + 1378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1369.0, + 1405.0, + 1369.0, + 1405.0, + 1407.0, + 292.0, + 1407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1397.0, + 1406.0, + 1397.0, + 1406.0, + 1439.0, + 291.0, + 1439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1428.0, + 1408.0, + 1428.0, + 1408.0, + 1466.0, + 293.0, + 1466.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1460.0, + 1405.0, + 1460.0, + 1405.0, + 1495.0, + 292.0, + 1495.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1490.0, + 1405.0, + 1490.0, + 1405.0, + 1526.0, + 295.0, + 1526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1522.0, + 1377.0, + 1522.0, + 1377.0, + 1557.0, + 293.0, + 1557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1021.0, + 1405.0, + 1021.0, + 1405.0, + 1054.0, + 293.0, + 1054.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1052.0, + 1405.0, + 1052.0, + 1405.0, + 1083.0, + 294.0, + 1083.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1082.0, + 1404.0, + 1082.0, + 1404.0, + 1116.0, + 294.0, + 1116.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1113.0, + 1406.0, + 1113.0, + 1406.0, + 1146.0, + 292.0, + 1146.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1141.0, + 1406.0, + 1141.0, + 1406.0, + 1176.0, + 293.0, + 1176.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1172.0, + 1406.0, + 1172.0, + 1406.0, + 1209.0, + 293.0, + 1209.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1203.0, + 743.0, + 1203.0, + 743.0, + 1238.0, + 294.0, + 1238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1586.0, + 1405.0, + 1586.0, + 1405.0, + 1623.0, + 294.0, + 1623.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1616.0, + 1405.0, + 1616.0, + 1405.0, + 1652.0, + 293.0, + 1652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1646.0, + 1404.0, + 1646.0, + 1404.0, + 1682.0, + 294.0, + 1682.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1674.0, + 1404.0, + 1674.0, + 1404.0, + 1718.0, + 292.0, + 1718.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1710.0, + 724.0, + 1710.0, + 724.0, + 1743.0, + 295.0, + 1743.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1851.0, + 1409.0, + 1851.0, + 1409.0, + 1890.0, + 294.0, + 1890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1884.0, + 1404.0, + 1884.0, + 1404.0, + 1917.0, + 295.0, + 1917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1913.0, + 1406.0, + 1913.0, + 1406.0, + 1950.0, + 294.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1942.0, + 1409.0, + 1942.0, + 1409.0, + 1985.0, + 293.0, + 1985.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1970.0, + 1405.0, + 1970.0, + 1405.0, + 2016.0, + 292.0, + 2016.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 851.0, + 1405.0, + 851.0, + 1405.0, + 891.0, + 295.0, + 891.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 884.0, + 1404.0, + 884.0, + 1404.0, + 917.0, + 294.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 915.0, + 1405.0, + 915.0, + 1405.0, + 950.0, + 293.0, + 950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 942.0, + 1408.0, + 942.0, + 1408.0, + 983.0, + 293.0, + 983.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 975.0, + 1408.0, + 975.0, + 1408.0, + 1012.0, + 294.0, + 1012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 202.0, + 1406.0, + 202.0, + 1406.0, + 240.0, + 294.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 233.0, + 1407.0, + 233.0, + 1407.0, + 270.0, + 294.0, + 270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 262.0, + 1230.0, + 262.0, + 1230.0, + 301.0, + 292.0, + 301.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 8, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1095, + 1405, + 1095, + 1405, + 1339, + 297, + 1339 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 297, + 837, + 1404, + 837, + 1404, + 1080, + 297, + 1080 + ], + "score": 0.979 + }, + { + "category_id": 1, + "poly": [ + 299, + 1450, + 1405, + 1450, + 1405, + 1573, + 299, + 1573 + ], + "score": 0.977 + }, + { + "category_id": 3, + "poly": [ + 299, + 195, + 1346, + 195, + 1346, + 696, + 299, + 696 + ], + "score": 0.975 + }, + { + "category_id": 4, + "poly": [ + 299, + 715, + 1400, + 715, + 1400, + 777, + 299, + 777 + ], + "score": 0.943 + }, + { + "category_id": 0, + "poly": [ + 299, + 1380, + 944, + 1380, + 944, + 1420, + 299, + 1420 + ], + "score": 0.935 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.846 + }, + { + "category_id": 15, + "poly": [ + 416.0, + 198.0, + 518.0, + 198.0, + 518.0, + 228.0, + 416.0, + 228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 643.0, + 198.0, + 797.0, + 198.0, + 797.0, + 227.0, + 643.0, + 227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 909.0, + 198.0, + 1041.0, + 198.0, + 1041.0, + 228.0, + 909.0, + 228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1177.0, + 199.0, + 1279.0, + 199.0, + 1279.0, + 226.0, + 1177.0, + 226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 240.0, + 349.0, + 240.0, + 349.0, + 264.0, + 324.0, + 264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 578.0, + 240.0, + 603.0, + 240.0, + 603.0, + 264.0, + 578.0, + 264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 723.0, + 233.0, + 824.0, + 233.0, + 824.0, + 260.0, + 723.0, + 260.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 239.0, + 857.0, + 239.0, + 857.0, + 264.0, + 832.0, + 264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1084.0, + 240.0, + 1110.0, + 240.0, + 1110.0, + 264.0, + 1084.0, + 264.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 301.0, + 252.0, + 332.0, + 252.0, + 332.0, + 369.0, + 301.0, + 369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 257.0, + 765.0, + 257.0, + 765.0, + 286.0, + 721.0, + 286.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 281.0, + 764.0, + 281.0, + 764.0, + 311.0, + 721.0, + 311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 305.0, + 349.0, + 305.0, + 349.0, + 328.0, + 331.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 585.0, + 306.0, + 603.0, + 306.0, + 603.0, + 328.0, + 585.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 307.0, + 753.0, + 307.0, + 753.0, + 334.0, + 721.0, + 334.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 305.0, + 857.0, + 305.0, + 857.0, + 328.0, + 839.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1093.0, + 306.0, + 1112.0, + 306.0, + 1112.0, + 328.0, + 1093.0, + 328.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 370.0, + 349.0, + 370.0, + 349.0, + 393.0, + 331.0, + 393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 587.0, + 374.0, + 600.0, + 374.0, + 600.0, + 389.0, + 587.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 842.0, + 374.0, + 855.0, + 374.0, + 855.0, + 389.0, + 842.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1095.0, + 374.0, + 1108.0, + 374.0, + 1108.0, + 389.0, + 1095.0, + 389.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 406.0, + 365.0, + 406.0, + 365.0, + 427.0, + 347.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 391.0, + 406.0, + 410.0, + 406.0, + 410.0, + 427.0, + 391.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 435.0, + 406.0, + 453.0, + 406.0, + 453.0, + 427.0, + 435.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 480.0, + 406.0, + 499.0, + 406.0, + 499.0, + 427.0, + 480.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 524.0, + 404.0, + 543.0, + 404.0, + 543.0, + 427.0, + 524.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 406.0, + 619.0, + 406.0, + 619.0, + 427.0, + 600.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 646.0, + 404.0, + 664.0, + 404.0, + 664.0, + 427.0, + 646.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 689.0, + 404.0, + 708.0, + 404.0, + 708.0, + 427.0, + 689.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 734.0, + 404.0, + 752.0, + 404.0, + 752.0, + 427.0, + 734.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 777.0, + 404.0, + 796.0, + 404.0, + 796.0, + 427.0, + 777.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 854.0, + 404.0, + 874.0, + 404.0, + 874.0, + 427.0, + 854.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 897.0, + 403.0, + 919.0, + 403.0, + 919.0, + 429.0, + 897.0, + 429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 942.0, + 406.0, + 961.0, + 406.0, + 961.0, + 427.0, + 942.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 987.0, + 406.0, + 1006.0, + 406.0, + 1006.0, + 427.0, + 987.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1030.0, + 404.0, + 1050.0, + 404.0, + 1050.0, + 427.0, + 1030.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1108.0, + 406.0, + 1127.0, + 406.0, + 1127.0, + 427.0, + 1108.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1152.0, + 406.0, + 1171.0, + 406.0, + 1171.0, + 427.0, + 1152.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1196.0, + 406.0, + 1216.0, + 406.0, + 1216.0, + 427.0, + 1196.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1241.0, + 406.0, + 1261.0, + 406.0, + 1261.0, + 427.0, + 1241.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1285.0, + 406.0, + 1304.0, + 406.0, + 1304.0, + 427.0, + 1285.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 542.0, + 421.0, + 584.0, + 421.0, + 584.0, + 451.0, + 542.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 795.0, + 421.0, + 838.0, + 421.0, + 838.0, + 451.0, + 795.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1050.0, + 421.0, + 1091.0, + 421.0, + 1091.0, + 451.0, + 1050.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1303.0, + 421.0, + 1346.0, + 421.0, + 1346.0, + 451.0, + 1303.0, + 451.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 445.0, + 543.0, + 445.0, + 543.0, + 472.0, + 392.0, + 472.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 665.0, + 442.0, + 776.0, + 442.0, + 776.0, + 475.0, + 665.0, + 475.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 916.0, + 443.0, + 1034.0, + 443.0, + 1034.0, + 474.0, + 916.0, + 474.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1188.0, + 443.0, + 1268.0, + 443.0, + 1268.0, + 474.0, + 1188.0, + 474.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 485.0, + 351.0, + 485.0, + 351.0, + 510.0, + 324.0, + 510.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 578.0, + 485.0, + 604.0, + 485.0, + 604.0, + 509.0, + 578.0, + 509.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 720.0, + 476.0, + 825.0, + 476.0, + 825.0, + 508.0, + 720.0, + 508.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 831.0, + 485.0, + 858.0, + 485.0, + 858.0, + 510.0, + 831.0, + 510.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1083.0, + 483.0, + 1113.0, + 483.0, + 1113.0, + 511.0, + 1083.0, + 511.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 301.0, + 498.0, + 332.0, + 498.0, + 332.0, + 616.0, + 301.0, + 616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 721.0, + 502.0, + 765.0, + 502.0, + 765.0, + 531.0, + 721.0, + 531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 723.0, + 529.0, + 762.0, + 529.0, + 762.0, + 555.0, + 723.0, + 555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 551.0, + 349.0, + 551.0, + 349.0, + 573.0, + 331.0, + 573.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 585.0, + 551.0, + 603.0, + 551.0, + 603.0, + 573.0, + 585.0, + 573.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 722.0, + 552.0, + 753.0, + 552.0, + 753.0, + 580.0, + 722.0, + 580.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 551.0, + 857.0, + 551.0, + 857.0, + 575.0, + 839.0, + 575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1093.0, + 551.0, + 1112.0, + 551.0, + 1112.0, + 575.0, + 1093.0, + 575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 333.0, + 619.0, + 347.0, + 619.0, + 347.0, + 637.0, + 333.0, + 637.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 585.0, + 618.0, + 603.0, + 618.0, + 603.0, + 638.0, + 585.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 842.0, + 620.0, + 855.0, + 620.0, + 855.0, + 636.0, + 842.0, + 636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1094.0, + 620.0, + 1108.0, + 620.0, + 1108.0, + 637.0, + 1094.0, + 637.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 652.0, + 363.0, + 652.0, + 363.0, + 671.0, + 348.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 390.0, + 651.0, + 413.0, + 651.0, + 413.0, + 673.0, + 390.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 435.0, + 651.0, + 455.0, + 651.0, + 455.0, + 672.0, + 435.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 480.0, + 651.0, + 500.0, + 651.0, + 500.0, + 672.0, + 480.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 523.0, + 650.0, + 543.0, + 650.0, + 543.0, + 672.0, + 523.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 651.0, + 619.0, + 651.0, + 619.0, + 672.0, + 600.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 646.0, + 651.0, + 665.0, + 651.0, + 665.0, + 673.0, + 646.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 689.0, + 651.0, + 708.0, + 651.0, + 708.0, + 671.0, + 689.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 733.0, + 651.0, + 753.0, + 651.0, + 753.0, + 672.0, + 733.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 777.0, + 650.0, + 796.0, + 650.0, + 796.0, + 672.0, + 777.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 852.0, + 649.0, + 875.0, + 649.0, + 875.0, + 674.0, + 852.0, + 674.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 899.0, + 651.0, + 919.0, + 651.0, + 919.0, + 672.0, + 899.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 941.0, + 651.0, + 963.0, + 651.0, + 963.0, + 673.0, + 941.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 987.0, + 651.0, + 1008.0, + 651.0, + 1008.0, + 672.0, + 987.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1028.0, + 649.0, + 1052.0, + 649.0, + 1052.0, + 673.0, + 1028.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1108.0, + 651.0, + 1127.0, + 651.0, + 1127.0, + 672.0, + 1108.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1152.0, + 651.0, + 1173.0, + 651.0, + 1173.0, + 672.0, + 1152.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 651.0, + 1217.0, + 651.0, + 1217.0, + 671.0, + 1195.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1239.0, + 651.0, + 1262.0, + 651.0, + 1262.0, + 671.0, + 1239.0, + 671.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1285.0, + 651.0, + 1304.0, + 651.0, + 1304.0, + 672.0, + 1285.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 396.0, + 666.0, + 584.0, + 666.0, + 584.0, + 697.0, + 396.0, + 697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 648.0, + 666.0, + 838.0, + 666.0, + 838.0, + 697.0, + 648.0, + 697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 901.0, + 667.0, + 1091.0, + 667.0, + 1091.0, + 697.0, + 901.0, + 697.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1157.0, + 670.0, + 1343.0, + 670.0, + 1343.0, + 694.0, + 1157.0, + 694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 714.0, + 1402.0, + 714.0, + 1402.0, + 750.0, + 297.0, + 750.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 745.0, + 1219.0, + 745.0, + 1219.0, + 779.0, + 294.0, + 779.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1376.0, + 949.0, + 1376.0, + 949.0, + 1430.0, + 293.0, + 1430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1096.0, + 1403.0, + 1096.0, + 1403.0, + 1130.0, + 295.0, + 1130.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1126.0, + 1406.0, + 1126.0, + 1406.0, + 1160.0, + 295.0, + 1160.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1156.0, + 1405.0, + 1156.0, + 1405.0, + 1190.0, + 294.0, + 1190.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1186.0, + 1406.0, + 1186.0, + 1406.0, + 1222.0, + 292.0, + 1222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1217.0, + 1405.0, + 1217.0, + 1405.0, + 1250.0, + 294.0, + 1250.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1247.0, + 1405.0, + 1247.0, + 1405.0, + 1280.0, + 295.0, + 1280.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1278.0, + 1405.0, + 1278.0, + 1405.0, + 1310.0, + 294.0, + 1310.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1306.0, + 535.0, + 1306.0, + 535.0, + 1344.0, + 294.0, + 1344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 838.0, + 1402.0, + 838.0, + 1402.0, + 871.0, + 296.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 869.0, + 1406.0, + 869.0, + 1406.0, + 902.0, + 294.0, + 902.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 896.0, + 1409.0, + 896.0, + 1409.0, + 934.0, + 292.0, + 934.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 929.0, + 1404.0, + 929.0, + 1404.0, + 963.0, + 296.0, + 963.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 957.0, + 1406.0, + 957.0, + 1406.0, + 995.0, + 294.0, + 995.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 987.0, + 1406.0, + 987.0, + 1406.0, + 1027.0, + 292.0, + 1027.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1019.0, + 1406.0, + 1019.0, + 1406.0, + 1056.0, + 295.0, + 1056.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1050.0, + 620.0, + 1050.0, + 620.0, + 1086.0, + 292.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1448.0, + 1409.0, + 1448.0, + 1409.0, + 1485.0, + 293.0, + 1485.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1481.0, + 1405.0, + 1481.0, + 1405.0, + 1514.0, + 295.0, + 1514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1509.0, + 1406.0, + 1509.0, + 1406.0, + 1546.0, + 292.0, + 1546.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1541.0, + 1378.0, + 1541.0, + 1378.0, + 1577.0, + 294.0, + 1577.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 9, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 0, + "poly": [ + 299, + 200, + 455, + 200, + 455, + 235, + 299, + 235 + ], + "score": 0.898 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 863, + 2061, + 863, + 2086, + 836, + 2086 + ], + "score": 0.74 + }, + { + "category_id": 1, + "poly": [ + 295, + 128, + 1409, + 128, + 1409, + 2019, + 295, + 2019 + ], + "score": 0.699 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 863, + 2061, + 863, + 2086, + 836, + 2086 + ], + "score": 0.227 + }, + { + "category_id": 15, + "poly": [ + 295.0, + 196.0, + 460.0, + 196.0, + 460.0, + 241.0, + 295.0, + 241.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2098.0, + 832.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 868.0, + 2058.0, + 868.0, + 2098.0, + 832.0, + 2098.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 196.0, + 463.0, + 196.0, + 463.0, + 240.0, + 294.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 254.0, + 1408.0, + 254.0, + 1408.0, + 287.0, + 308.0, + 287.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 277.0, + 1411.0, + 277.0, + 1411.0, + 325.0, + 350.0, + 325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 312.0, + 1246.0, + 312.0, + 1246.0, + 352.0, + 355.0, + 352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 364.0, + 1406.0, + 364.0, + 1406.0, + 404.0, + 308.0, + 404.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 395.0, + 1389.0, + 395.0, + 1389.0, + 435.0, + 352.0, + 435.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 447.0, + 1408.0, + 447.0, + 1408.0, + 487.0, + 308.0, + 487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 478.0, + 1406.0, + 478.0, + 1406.0, + 518.0, + 352.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 509.0, + 1406.0, + 509.0, + 1406.0, + 549.0, + 355.0, + 549.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 561.0, + 1406.0, + 561.0, + 1406.0, + 601.0, + 308.0, + 601.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 594.0, + 1404.0, + 594.0, + 1404.0, + 628.0, + 355.0, + 628.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 619.0, + 1406.0, + 619.0, + 1406.0, + 661.0, + 350.0, + 661.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 653.0, + 763.0, + 653.0, + 763.0, + 690.0, + 348.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 704.0, + 1406.0, + 704.0, + 1406.0, + 744.0, + 308.0, + 744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 738.0, + 1404.0, + 738.0, + 1404.0, + 771.0, + 355.0, + 771.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 767.0, + 995.0, + 767.0, + 995.0, + 806.0, + 355.0, + 806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 304.0, + 814.0, + 1408.0, + 814.0, + 1408.0, + 862.0, + 304.0, + 862.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 850.0, + 1406.0, + 850.0, + 1406.0, + 889.0, + 352.0, + 889.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 879.0, + 690.0, + 879.0, + 690.0, + 918.0, + 352.0, + 918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 935.0, + 1408.0, + 935.0, + 1408.0, + 968.0, + 308.0, + 968.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 962.0, + 1330.0, + 962.0, + 1330.0, + 1001.0, + 352.0, + 1001.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 1018.0, + 1404.0, + 1018.0, + 1404.0, + 1051.0, + 308.0, + 1051.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1042.0, + 1406.0, + 1042.0, + 1406.0, + 1084.0, + 348.0, + 1084.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1074.0, + 498.0, + 1074.0, + 498.0, + 1113.0, + 352.0, + 1113.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 308.0, + 1132.0, + 1404.0, + 1132.0, + 1404.0, + 1165.0, + 308.0, + 1165.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1156.0, + 1301.0, + 1156.0, + 1301.0, + 1196.0, + 352.0, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1215.0, + 1404.0, + 1215.0, + 1404.0, + 1248.0, + 296.0, + 1248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1244.0, + 1402.0, + 1244.0, + 1402.0, + 1277.0, + 355.0, + 1277.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1275.0, + 629.0, + 1275.0, + 629.0, + 1308.0, + 355.0, + 1308.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1324.0, + 1406.0, + 1324.0, + 1406.0, + 1364.0, + 293.0, + 1364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1356.0, + 1408.0, + 1356.0, + 1408.0, + 1395.0, + 352.0, + 1395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1387.0, + 430.0, + 1387.0, + 430.0, + 1422.0, + 352.0, + 1422.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1441.0, + 1406.0, + 1441.0, + 1406.0, + 1474.0, + 296.0, + 1474.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1472.0, + 1404.0, + 1472.0, + 1404.0, + 1505.0, + 355.0, + 1505.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1499.0, + 1094.0, + 1499.0, + 1094.0, + 1538.0, + 352.0, + 1538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1555.0, + 1406.0, + 1555.0, + 1406.0, + 1588.0, + 296.0, + 1588.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1580.0, + 1358.0, + 1580.0, + 1358.0, + 1623.0, + 348.0, + 1623.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1633.0, + 1411.0, + 1633.0, + 1411.0, + 1673.0, + 293.0, + 1673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1665.0, + 1339.0, + 1665.0, + 1339.0, + 1704.0, + 350.0, + 1704.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1714.0, + 1411.0, + 1714.0, + 1411.0, + 1758.0, + 291.0, + 1758.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1745.0, + 909.0, + 1745.0, + 909.0, + 1787.0, + 350.0, + 1787.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1804.0, + 1406.0, + 1804.0, + 1406.0, + 1837.0, + 296.0, + 1837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1830.0, + 1408.0, + 1830.0, + 1408.0, + 1870.0, + 352.0, + 1870.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 350.0, + 1860.0, + 433.0, + 1860.0, + 433.0, + 1901.0, + 350.0, + 1901.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1913.0, + 1406.0, + 1913.0, + 1406.0, + 1953.0, + 291.0, + 1953.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1938.0, + 1408.0, + 1938.0, + 1408.0, + 1988.0, + 348.0, + 1988.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1969.0, + 566.0, + 1969.0, + 566.0, + 2013.0, + 349.0, + 2013.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 10, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.675 + }, + { + "category_id": 1, + "poly": [ + 297, + 842, + 1402, + 842, + 1402, + 939, + 297, + 939 + ], + "score": 0.574 + }, + { + "category_id": 1, + "poly": [ + 296, + 696, + 1406, + 696, + 1406, + 820, + 296, + 820 + ], + "score": 0.569 + }, + { + "category_id": 1, + "poly": [ + 296, + 1362, + 1400, + 1362, + 1400, + 1458, + 296, + 1458 + ], + "score": 0.561 + }, + { + "category_id": 1, + "poly": [ + 296, + 1595, + 1403, + 1595, + 1403, + 1690, + 296, + 1690 + ], + "score": 0.558 + }, + { + "category_id": 1, + "poly": [ + 293, + 579, + 1404, + 579, + 1404, + 675, + 293, + 675 + ], + "score": 0.551 + }, + { + "category_id": 1, + "poly": [ + 292, + 1160, + 1404, + 1160, + 1404, + 1226, + 292, + 1226 + ], + "score": 0.534 + }, + { + "category_id": 1, + "poly": [ + 293, + 202, + 1401, + 202, + 1401, + 297, + 293, + 297 + ], + "score": 0.53 + }, + { + "category_id": 1, + "poly": [ + 297, + 1711, + 1401, + 1711, + 1401, + 1806, + 297, + 1806 + ], + "score": 0.521 + }, + { + "category_id": 1, + "poly": [ + 295, + 1247, + 1402, + 1247, + 1402, + 1342, + 295, + 1342 + ], + "score": 0.519 + }, + { + "category_id": 1, + "poly": [ + 295, + 317, + 1398, + 317, + 1398, + 412, + 295, + 412 + ], + "score": 0.509 + }, + { + "category_id": 1, + "poly": [ + 296, + 1944, + 1399, + 1944, + 1399, + 2007, + 296, + 2007 + ], + "score": 0.503 + }, + { + "category_id": 1, + "poly": [ + 295, + 1828, + 1398, + 1828, + 1398, + 1923, + 295, + 1923 + ], + "score": 0.488 + }, + { + "category_id": 1, + "poly": [ + 294, + 1104, + 1379, + 1104, + 1379, + 1142, + 294, + 1142 + ], + "score": 0.47 + }, + { + "category_id": 1, + "poly": [ + 297, + 434, + 1405, + 434, + 1405, + 559, + 297, + 559 + ], + "score": 0.441 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.411 + }, + { + "category_id": 1, + "poly": [ + 298, + 957, + 1406, + 957, + 1406, + 1082, + 298, + 1082 + ], + "score": 0.329 + }, + { + "category_id": 1, + "poly": [ + 295, + 1478, + 1405, + 1478, + 1405, + 1572, + 295, + 1572 + ], + "score": 0.255 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 843.0, + 1405.0, + 843.0, + 1405.0, + 878.0, + 293.0, + 878.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 873.0, + 1409.0, + 873.0, + 1409.0, + 909.0, + 354.0, + 909.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 905.0, + 1172.0, + 905.0, + 1172.0, + 940.0, + 356.0, + 940.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 696.0, + 1409.0, + 696.0, + 1409.0, + 735.0, + 294.0, + 735.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 725.0, + 1406.0, + 725.0, + 1406.0, + 765.0, + 352.0, + 765.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 757.0, + 1406.0, + 757.0, + 1406.0, + 795.0, + 353.0, + 795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 790.0, + 745.0, + 790.0, + 745.0, + 822.0, + 353.0, + 822.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1361.0, + 1405.0, + 1361.0, + 1405.0, + 1401.0, + 294.0, + 1401.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1394.0, + 1404.0, + 1394.0, + 1404.0, + 1430.0, + 354.0, + 1430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1426.0, + 723.0, + 1426.0, + 723.0, + 1460.0, + 355.0, + 1460.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1594.0, + 1404.0, + 1594.0, + 1404.0, + 1634.0, + 294.0, + 1634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1627.0, + 1408.0, + 1627.0, + 1408.0, + 1663.0, + 353.0, + 1663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1657.0, + 608.0, + 1657.0, + 608.0, + 1691.0, + 353.0, + 1691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 579.0, + 1406.0, + 579.0, + 1406.0, + 616.0, + 293.0, + 616.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 612.0, + 1406.0, + 612.0, + 1406.0, + 647.0, + 355.0, + 647.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 642.0, + 1338.0, + 642.0, + 1338.0, + 680.0, + 354.0, + 680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1162.0, + 1407.0, + 1162.0, + 1407.0, + 1197.0, + 295.0, + 1197.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1192.0, + 1281.0, + 1192.0, + 1281.0, + 1228.0, + 354.0, + 1228.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 202.0, + 1406.0, + 202.0, + 1406.0, + 239.0, + 293.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 234.0, + 1404.0, + 234.0, + 1404.0, + 268.0, + 356.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 264.0, + 951.0, + 264.0, + 951.0, + 297.0, + 355.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1709.0, + 1407.0, + 1709.0, + 1407.0, + 1749.0, + 292.0, + 1749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1741.0, + 1405.0, + 1741.0, + 1405.0, + 1778.0, + 355.0, + 1778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1774.0, + 740.0, + 1774.0, + 740.0, + 1806.0, + 352.0, + 1806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1247.0, + 1405.0, + 1247.0, + 1405.0, + 1284.0, + 294.0, + 1284.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1280.0, + 1404.0, + 1280.0, + 1404.0, + 1314.0, + 356.0, + 1314.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1310.0, + 967.0, + 1310.0, + 967.0, + 1342.0, + 354.0, + 1342.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 316.0, + 1405.0, + 316.0, + 1405.0, + 356.0, + 293.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 350.0, + 1404.0, + 350.0, + 1404.0, + 384.0, + 355.0, + 384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 378.0, + 920.0, + 378.0, + 920.0, + 416.0, + 353.0, + 416.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1942.0, + 1404.0, + 1942.0, + 1404.0, + 1982.0, + 295.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1973.0, + 620.0, + 1973.0, + 620.0, + 2007.0, + 352.0, + 2007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1829.0, + 1403.0, + 1829.0, + 1403.0, + 1863.0, + 296.0, + 1863.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1858.0, + 1403.0, + 1858.0, + 1403.0, + 1893.0, + 355.0, + 1893.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1886.0, + 1184.0, + 1886.0, + 1184.0, + 1925.0, + 354.0, + 1925.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1104.0, + 1383.0, + 1104.0, + 1383.0, + 1142.0, + 295.0, + 1142.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 434.0, + 1405.0, + 434.0, + 1405.0, + 471.0, + 295.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 465.0, + 1408.0, + 465.0, + 1408.0, + 501.0, + 354.0, + 501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 495.0, + 1407.0, + 495.0, + 1407.0, + 535.0, + 351.0, + 535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 527.0, + 996.0, + 527.0, + 996.0, + 560.0, + 354.0, + 560.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 959.0, + 1408.0, + 959.0, + 1408.0, + 995.0, + 293.0, + 995.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 989.0, + 1404.0, + 989.0, + 1404.0, + 1025.0, + 353.0, + 1025.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1018.0, + 1406.0, + 1018.0, + 1406.0, + 1059.0, + 353.0, + 1059.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1051.0, + 654.0, + 1051.0, + 654.0, + 1084.0, + 355.0, + 1084.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1479.0, + 1405.0, + 1479.0, + 1405.0, + 1515.0, + 294.0, + 1515.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1509.0, + 1410.0, + 1509.0, + 1410.0, + 1547.0, + 352.0, + 1547.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1535.0, + 435.0, + 1535.0, + 435.0, + 1576.0, + 352.0, + 1576.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 11, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.7 + }, + { + "category_id": 1, + "poly": [ + 291, + 137, + 1409, + 137, + 1409, + 2007, + 291, + 2007 + ], + "score": 0.609 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.372 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 202.0, + 1409.0, + 202.0, + 1409.0, + 241.0, + 293.0, + 241.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 231.0, + 1407.0, + 231.0, + 1407.0, + 274.0, + 353.0, + 274.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 263.0, + 610.0, + 263.0, + 610.0, + 296.0, + 355.0, + 296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 317.0, + 1407.0, + 317.0, + 1407.0, + 356.0, + 295.0, + 356.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 350.0, + 978.0, + 350.0, + 978.0, + 384.0, + 349.0, + 384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 401.0, + 1405.0, + 401.0, + 1405.0, + 440.0, + 293.0, + 440.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 432.0, + 501.0, + 432.0, + 501.0, + 471.0, + 355.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 485.0, + 1407.0, + 485.0, + 1407.0, + 524.0, + 295.0, + 524.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 516.0, + 1407.0, + 516.0, + 1407.0, + 555.0, + 353.0, + 555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 544.0, + 514.0, + 544.0, + 514.0, + 584.0, + 354.0, + 584.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 602.0, + 1405.0, + 602.0, + 1405.0, + 635.0, + 297.0, + 635.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 629.0, + 1405.0, + 629.0, + 1405.0, + 672.0, + 351.0, + 672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 662.0, + 609.0, + 662.0, + 609.0, + 698.0, + 351.0, + 698.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 715.0, + 1407.0, + 715.0, + 1407.0, + 754.0, + 295.0, + 754.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 739.0, + 1411.0, + 739.0, + 1411.0, + 789.0, + 349.0, + 789.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 772.0, + 700.0, + 772.0, + 700.0, + 815.0, + 349.0, + 815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 828.0, + 1407.0, + 828.0, + 1407.0, + 867.0, + 293.0, + 867.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 861.0, + 1407.0, + 861.0, + 1407.0, + 900.0, + 355.0, + 900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 889.0, + 970.0, + 889.0, + 970.0, + 926.0, + 353.0, + 926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 943.0, + 1409.0, + 943.0, + 1409.0, + 982.0, + 293.0, + 982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 971.0, + 1407.0, + 971.0, + 1407.0, + 1015.0, + 351.0, + 1015.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 357.0, + 1006.0, + 652.0, + 1006.0, + 652.0, + 1039.0, + 357.0, + 1039.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1058.0, + 1409.0, + 1058.0, + 1409.0, + 1097.0, + 293.0, + 1097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1090.0, + 1407.0, + 1090.0, + 1407.0, + 1129.0, + 355.0, + 1129.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1119.0, + 970.0, + 1119.0, + 970.0, + 1156.0, + 353.0, + 1156.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1173.0, + 1407.0, + 1173.0, + 1407.0, + 1209.0, + 291.0, + 1209.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1203.0, + 1405.0, + 1203.0, + 1405.0, + 1242.0, + 353.0, + 1242.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1232.0, + 1379.0, + 1232.0, + 1379.0, + 1273.0, + 349.0, + 1273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1287.0, + 1407.0, + 1287.0, + 1407.0, + 1326.0, + 295.0, + 1326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1318.0, + 1233.0, + 1318.0, + 1233.0, + 1357.0, + 353.0, + 1357.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1372.0, + 1407.0, + 1372.0, + 1407.0, + 1411.0, + 293.0, + 1411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1396.0, + 1411.0, + 1396.0, + 1411.0, + 1446.0, + 349.0, + 1446.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1431.0, + 1291.0, + 1431.0, + 1291.0, + 1470.0, + 353.0, + 1470.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1484.0, + 1407.0, + 1484.0, + 1407.0, + 1523.0, + 293.0, + 1523.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 357.0, + 1519.0, + 1407.0, + 1519.0, + 1407.0, + 1552.0, + 357.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1548.0, + 1403.0, + 1548.0, + 1403.0, + 1587.0, + 355.0, + 1587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1601.0, + 1409.0, + 1601.0, + 1409.0, + 1640.0, + 293.0, + 1640.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1626.0, + 1407.0, + 1626.0, + 1407.0, + 1675.0, + 351.0, + 1675.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1663.0, + 700.0, + 1663.0, + 700.0, + 1700.0, + 351.0, + 1700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1714.0, + 1409.0, + 1714.0, + 1409.0, + 1753.0, + 293.0, + 1753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1747.0, + 1407.0, + 1747.0, + 1407.0, + 1786.0, + 353.0, + 1786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1778.0, + 693.0, + 1778.0, + 693.0, + 1813.0, + 349.0, + 1813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1829.0, + 1409.0, + 1829.0, + 1409.0, + 1868.0, + 295.0, + 1868.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1860.0, + 1082.0, + 1860.0, + 1082.0, + 1899.0, + 353.0, + 1899.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1913.0, + 1407.0, + 1913.0, + 1407.0, + 1952.0, + 295.0, + 1952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1942.0, + 1405.0, + 1942.0, + 1405.0, + 1985.0, + 349.0, + 1985.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1973.0, + 431.0, + 1973.0, + 431.0, + 2014.0, + 353.0, + 2014.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 12, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.836 + }, + { + "category_id": 1, + "poly": [ + 298, + 1172, + 1398, + 1172, + 1398, + 1236, + 298, + 1236 + ], + "score": 0.582 + }, + { + "category_id": 1, + "poly": [ + 294, + 202, + 1402, + 202, + 1402, + 297, + 294, + 297 + ], + "score": 0.54 + }, + { + "category_id": 1, + "poly": [ + 295, + 1256, + 1403, + 1256, + 1403, + 1352, + 295, + 1352 + ], + "score": 0.535 + }, + { + "category_id": 1, + "poly": [ + 295, + 1371, + 1401, + 1371, + 1401, + 1465, + 295, + 1465 + ], + "score": 0.534 + }, + { + "category_id": 1, + "poly": [ + 299, + 1088, + 1400, + 1088, + 1400, + 1153, + 299, + 1153 + ], + "score": 0.518 + }, + { + "category_id": 1, + "poly": [ + 295, + 1484, + 1399, + 1484, + 1399, + 1580, + 295, + 1580 + ], + "score": 0.511 + }, + { + "category_id": 1, + "poly": [ + 299, + 773, + 1402, + 773, + 1402, + 898, + 299, + 898 + ], + "score": 0.465 + }, + { + "category_id": 1, + "poly": [ + 296, + 1003, + 1403, + 1003, + 1403, + 1069, + 296, + 1069 + ], + "score": 0.442 + }, + { + "category_id": 1, + "poly": [ + 294, + 918, + 1405, + 918, + 1405, + 985, + 294, + 985 + ], + "score": 0.44 + }, + { + "category_id": 1, + "poly": [ + 296, + 515, + 1402, + 515, + 1402, + 610, + 296, + 610 + ], + "score": 0.428 + }, + { + "category_id": 1, + "poly": [ + 293, + 315, + 1405, + 315, + 1405, + 381, + 293, + 381 + ], + "score": 0.425 + }, + { + "category_id": 1, + "poly": [ + 295, + 629, + 1402, + 629, + 1402, + 755, + 295, + 755 + ], + "score": 0.407 + }, + { + "category_id": 1, + "poly": [ + 294, + 1745, + 1401, + 1745, + 1401, + 1840, + 294, + 1840 + ], + "score": 0.396 + }, + { + "category_id": 1, + "poly": [ + 299, + 399, + 1398, + 399, + 1398, + 495, + 299, + 495 + ], + "score": 0.393 + }, + { + "category_id": 1, + "poly": [ + 295, + 1859, + 1399, + 1859, + 1399, + 1954, + 295, + 1954 + ], + "score": 0.365 + }, + { + "category_id": 1, + "poly": [ + 297, + 1975, + 1403, + 1975, + 1403, + 2009, + 297, + 2009 + ], + "score": 0.358 + }, + { + "category_id": 1, + "poly": [ + 298, + 1600, + 1406, + 1600, + 1406, + 1723, + 298, + 1723 + ], + "score": 0.336 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2095.0, + 832.0, + 2095.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1172.0, + 1402.0, + 1172.0, + 1402.0, + 1208.0, + 295.0, + 1208.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1200.0, + 1212.0, + 1200.0, + 1212.0, + 1239.0, + 353.0, + 1239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 202.0, + 1404.0, + 202.0, + 1404.0, + 240.0, + 296.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 235.0, + 1405.0, + 235.0, + 1405.0, + 269.0, + 356.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 262.0, + 942.0, + 262.0, + 942.0, + 298.0, + 352.0, + 298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1257.0, + 1406.0, + 1257.0, + 1406.0, + 1292.0, + 295.0, + 1292.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1286.0, + 1406.0, + 1286.0, + 1406.0, + 1326.0, + 352.0, + 1326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1319.0, + 1406.0, + 1319.0, + 1406.0, + 1353.0, + 356.0, + 1353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1368.0, + 1405.0, + 1368.0, + 1405.0, + 1410.0, + 294.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1401.0, + 1405.0, + 1401.0, + 1405.0, + 1436.0, + 353.0, + 1436.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1435.0, + 738.0, + 1435.0, + 738.0, + 1465.0, + 353.0, + 1465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1088.0, + 1405.0, + 1088.0, + 1405.0, + 1124.0, + 297.0, + 1124.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1121.0, + 992.0, + 1121.0, + 992.0, + 1153.0, + 354.0, + 1153.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1484.0, + 1404.0, + 1484.0, + 1404.0, + 1522.0, + 295.0, + 1522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1517.0, + 1405.0, + 1517.0, + 1405.0, + 1552.0, + 354.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1545.0, + 942.0, + 1545.0, + 942.0, + 1583.0, + 353.0, + 1583.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 775.0, + 1407.0, + 775.0, + 1407.0, + 810.0, + 293.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 806.0, + 1404.0, + 806.0, + 1404.0, + 839.0, + 354.0, + 839.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 834.0, + 1406.0, + 834.0, + 1406.0, + 873.0, + 352.0, + 873.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 867.0, + 581.0, + 867.0, + 581.0, + 899.0, + 354.0, + 899.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1005.0, + 1405.0, + 1005.0, + 1405.0, + 1041.0, + 296.0, + 1041.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1036.0, + 1296.0, + 1036.0, + 1296.0, + 1068.0, + 353.0, + 1068.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 919.0, + 1408.0, + 919.0, + 1408.0, + 955.0, + 297.0, + 955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 950.0, + 1280.0, + 950.0, + 1280.0, + 985.0, + 354.0, + 985.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 515.0, + 1407.0, + 515.0, + 1407.0, + 553.0, + 295.0, + 553.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 545.0, + 1404.0, + 545.0, + 1404.0, + 582.0, + 354.0, + 582.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 577.0, + 976.0, + 577.0, + 976.0, + 611.0, + 355.0, + 611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 317.0, + 1408.0, + 317.0, + 1408.0, + 353.0, + 296.0, + 353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 347.0, + 1367.0, + 347.0, + 1367.0, + 382.0, + 354.0, + 382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 631.0, + 1405.0, + 631.0, + 1405.0, + 666.0, + 293.0, + 666.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 661.0, + 1404.0, + 661.0, + 1404.0, + 694.0, + 354.0, + 694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 692.0, + 1407.0, + 692.0, + 1407.0, + 726.0, + 352.0, + 726.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 721.0, + 1002.0, + 721.0, + 1002.0, + 757.0, + 356.0, + 757.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1744.0, + 1406.0, + 1744.0, + 1406.0, + 1783.0, + 294.0, + 1783.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1778.0, + 1404.0, + 1778.0, + 1404.0, + 1812.0, + 356.0, + 1812.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1805.0, + 1265.0, + 1805.0, + 1265.0, + 1842.0, + 353.0, + 1842.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 398.0, + 1404.0, + 398.0, + 1404.0, + 438.0, + 294.0, + 438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 434.0, + 1404.0, + 434.0, + 1404.0, + 468.0, + 355.0, + 468.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 463.0, + 941.0, + 463.0, + 941.0, + 497.0, + 355.0, + 497.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1859.0, + 1404.0, + 1859.0, + 1404.0, + 1897.0, + 295.0, + 1897.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1892.0, + 1403.0, + 1892.0, + 1403.0, + 1926.0, + 355.0, + 1926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1924.0, + 767.0, + 1924.0, + 767.0, + 1955.0, + 354.0, + 1955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1972.0, + 1408.0, + 1972.0, + 1408.0, + 2010.0, + 295.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1599.0, + 1408.0, + 1599.0, + 1408.0, + 1637.0, + 293.0, + 1637.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1629.0, + 1406.0, + 1629.0, + 1406.0, + 1667.0, + 351.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1664.0, + 1406.0, + 1664.0, + 1406.0, + 1696.0, + 354.0, + 1696.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1693.0, + 642.0, + 1693.0, + 642.0, + 1724.0, + 354.0, + 1724.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 13, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.817 + }, + { + "category_id": 1, + "poly": [ + 297, + 1663, + 1404, + 1663, + 1404, + 1757, + 297, + 1757 + ], + "score": 0.643 + }, + { + "category_id": 1, + "poly": [ + 294, + 1579, + 1405, + 1579, + 1405, + 1646, + 294, + 1646 + ], + "score": 0.551 + }, + { + "category_id": 1, + "poly": [ + 298, + 1271, + 1403, + 1271, + 1403, + 1367, + 298, + 1367 + ], + "score": 0.479 + }, + { + "category_id": 1, + "poly": [ + 294, + 1497, + 1404, + 1497, + 1404, + 1563, + 294, + 1563 + ], + "score": 0.471 + }, + { + "category_id": 1, + "poly": [ + 299, + 1384, + 1401, + 1384, + 1401, + 1479, + 299, + 1479 + ], + "score": 0.46 + }, + { + "category_id": 1, + "poly": [ + 293, + 1106, + 1402, + 1106, + 1402, + 1170, + 293, + 1170 + ], + "score": 0.44 + }, + { + "category_id": 1, + "poly": [ + 294, + 1189, + 1402, + 1189, + 1402, + 1254, + 294, + 1254 + ], + "score": 0.438 + }, + { + "category_id": 1, + "poly": [ + 298, + 683, + 1401, + 683, + 1401, + 779, + 298, + 779 + ], + "score": 0.437 + }, + { + "category_id": 1, + "poly": [ + 298, + 345, + 1402, + 345, + 1402, + 438, + 298, + 438 + ], + "score": 0.432 + }, + { + "category_id": 1, + "poly": [ + 297, + 879, + 1401, + 879, + 1401, + 975, + 297, + 975 + ], + "score": 0.427 + }, + { + "category_id": 1, + "poly": [ + 297, + 570, + 1402, + 570, + 1402, + 667, + 297, + 667 + ], + "score": 0.402 + }, + { + "category_id": 1, + "poly": [ + 297, + 992, + 1402, + 992, + 1402, + 1089, + 297, + 1089 + ], + "score": 0.402 + }, + { + "category_id": 1, + "poly": [ + 297, + 796, + 1402, + 796, + 1402, + 863, + 297, + 863 + ], + "score": 0.401 + }, + { + "category_id": 1, + "poly": [ + 297, + 458, + 1399, + 458, + 1399, + 554, + 297, + 554 + ], + "score": 0.393 + }, + { + "category_id": 1, + "poly": [ + 296, + 202, + 1408, + 202, + 1408, + 325, + 296, + 325 + ], + "score": 0.281 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1663.0, + 1407.0, + 1663.0, + 1407.0, + 1700.0, + 295.0, + 1700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1692.0, + 1407.0, + 1692.0, + 1407.0, + 1729.0, + 353.0, + 1729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1722.0, + 431.0, + 1722.0, + 431.0, + 1759.0, + 351.0, + 1759.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1577.0, + 1410.0, + 1577.0, + 1410.0, + 1620.0, + 294.0, + 1620.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1611.0, + 1323.0, + 1611.0, + 1323.0, + 1646.0, + 354.0, + 1646.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1271.0, + 1408.0, + 1271.0, + 1408.0, + 1307.0, + 293.0, + 1307.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1304.0, + 1405.0, + 1304.0, + 1405.0, + 1337.0, + 353.0, + 1337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1334.0, + 1021.0, + 1334.0, + 1021.0, + 1368.0, + 355.0, + 1368.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1499.0, + 1405.0, + 1499.0, + 1405.0, + 1535.0, + 297.0, + 1535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 1529.0, + 1407.0, + 1529.0, + 1407.0, + 1565.0, + 356.0, + 1565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1383.0, + 1407.0, + 1383.0, + 1407.0, + 1423.0, + 293.0, + 1423.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 1413.0, + 1405.0, + 1413.0, + 1405.0, + 1455.0, + 352.0, + 1455.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1446.0, + 570.0, + 1446.0, + 570.0, + 1481.0, + 355.0, + 1481.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1106.0, + 1404.0, + 1106.0, + 1404.0, + 1142.0, + 296.0, + 1142.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 1137.0, + 914.0, + 1137.0, + 914.0, + 1170.0, + 354.0, + 1170.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1187.0, + 1404.0, + 1187.0, + 1404.0, + 1229.0, + 294.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 1222.0, + 888.0, + 1222.0, + 888.0, + 1254.0, + 353.0, + 1254.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 681.0, + 1406.0, + 681.0, + 1406.0, + 725.0, + 292.0, + 725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 713.0, + 1406.0, + 713.0, + 1406.0, + 753.0, + 352.0, + 753.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 747.0, + 680.0, + 747.0, + 680.0, + 778.0, + 356.0, + 778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 344.0, + 1408.0, + 344.0, + 1408.0, + 382.0, + 294.0, + 382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 375.0, + 1407.0, + 375.0, + 1407.0, + 411.0, + 351.0, + 411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 402.0, + 432.0, + 402.0, + 432.0, + 443.0, + 352.0, + 443.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 879.0, + 1406.0, + 879.0, + 1406.0, + 917.0, + 293.0, + 917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 912.0, + 1405.0, + 912.0, + 1405.0, + 947.0, + 355.0, + 947.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 943.0, + 1022.0, + 943.0, + 1022.0, + 977.0, + 355.0, + 977.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 568.0, + 1409.0, + 568.0, + 1409.0, + 611.0, + 293.0, + 611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 601.0, + 1406.0, + 601.0, + 1406.0, + 638.0, + 354.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 632.0, + 1286.0, + 632.0, + 1286.0, + 669.0, + 354.0, + 669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 992.0, + 1406.0, + 992.0, + 1406.0, + 1029.0, + 293.0, + 1029.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 357.0, + 1024.0, + 1404.0, + 1024.0, + 1404.0, + 1059.0, + 357.0, + 1059.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 1055.0, + 1047.0, + 1055.0, + 1047.0, + 1090.0, + 355.0, + 1090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 796.0, + 1403.0, + 796.0, + 1403.0, + 832.0, + 296.0, + 832.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 354.0, + 828.0, + 1404.0, + 828.0, + 1404.0, + 865.0, + 354.0, + 865.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 456.0, + 1404.0, + 456.0, + 1404.0, + 496.0, + 293.0, + 496.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 488.0, + 1404.0, + 488.0, + 1404.0, + 526.0, + 353.0, + 526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 356.0, + 521.0, + 1072.0, + 521.0, + 1072.0, + 555.0, + 356.0, + 555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 201.0, + 1409.0, + 201.0, + 1409.0, + 239.0, + 294.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 353.0, + 232.0, + 1406.0, + 232.0, + 1406.0, + 267.0, + 353.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 261.0, + 1406.0, + 261.0, + 1406.0, + 302.0, + 352.0, + 302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 355.0, + 294.0, + 567.0, + 294.0, + 567.0, + 327.0, + 355.0, + 327.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 14, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 406, + 730, + 1404, + 730, + 1404, + 1111, + 406, + 1111 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 407, + 1161, + 1403, + 1161, + 1403, + 1428, + 407, + 1428 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 406, + 300, + 1404, + 300, + 1404, + 499, + 406, + 499 + ], + "score": 0.969 + }, + { + "category_id": 1, + "poly": [ + 407, + 1478, + 1404, + 1478, + 1404, + 1676, + 407, + 1676 + ], + "score": 0.967 + }, + { + "category_id": 1, + "poly": [ + 407, + 549, + 1404, + 549, + 1404, + 678, + 407, + 678 + ], + "score": 0.964 + }, + { + "category_id": 1, + "poly": [ + 363, + 688, + 672, + 688, + 672, + 720, + 363, + 720 + ], + "score": 0.92 + }, + { + "category_id": 1, + "poly": [ + 365, + 509, + 841, + 509, + 841, + 541, + 365, + 541 + ], + "score": 0.917 + }, + { + "category_id": 1, + "poly": [ + 364, + 257, + 580, + 257, + 580, + 288, + 364, + 288 + ], + "score": 0.898 + }, + { + "category_id": 0, + "poly": [ + 298, + 199, + 435, + 199, + 435, + 235, + 298, + 235 + ], + "score": 0.897 + }, + { + "category_id": 1, + "poly": [ + 366, + 1437, + 1200, + 1437, + 1200, + 1469, + 366, + 1469 + ], + "score": 0.848 + }, + { + "category_id": 1, + "poly": [ + 354, + 1121, + 1401, + 1121, + 1401, + 1153, + 354, + 1153 + ], + "score": 0.844 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.769 + }, + { + "category_id": 2, + "poly": [ + 836, + 2062, + 865, + 2062, + 865, + 2086, + 836, + 2086 + ], + "score": 0.208 + }, + { + "category_id": 15, + "poly": [ + 296.0, + 197.0, + 440.0, + 197.0, + 440.0, + 238.0, + 296.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 831.0, + 2058.0, + 871.0, + 2058.0, + 871.0, + 2095.0, + 831.0, + 2095.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 727.0, + 1409.0, + 727.0, + 1409.0, + 766.0, + 404.0, + 766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 759.0, + 1405.0, + 759.0, + 1405.0, + 794.0, + 450.0, + 794.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 792.0, + 581.0, + 792.0, + 581.0, + 826.0, + 450.0, + 826.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 406.0, + 825.0, + 1403.0, + 825.0, + 1403.0, + 862.0, + 406.0, + 862.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 855.0, + 1406.0, + 855.0, + 1406.0, + 890.0, + 449.0, + 890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 448.0, + 885.0, + 1073.0, + 885.0, + 1073.0, + 922.0, + 448.0, + 922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 409.0, + 924.0, + 1406.0, + 924.0, + 1406.0, + 956.0, + 409.0, + 956.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 951.0, + 1405.0, + 951.0, + 1405.0, + 989.0, + 449.0, + 989.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 983.0, + 715.0, + 983.0, + 715.0, + 1019.0, + 449.0, + 1019.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 406.0, + 1016.0, + 1405.0, + 1016.0, + 1405.0, + 1055.0, + 406.0, + 1055.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 451.0, + 1049.0, + 1404.0, + 1049.0, + 1404.0, + 1081.0, + 451.0, + 1081.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 451.0, + 1080.0, + 891.0, + 1080.0, + 891.0, + 1115.0, + 451.0, + 1115.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 1160.0, + 1404.0, + 1160.0, + 1404.0, + 1196.0, + 405.0, + 1196.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 1191.0, + 865.0, + 1191.0, + 865.0, + 1227.0, + 449.0, + 1227.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 1228.0, + 1000.0, + 1228.0, + 1000.0, + 1261.0, + 404.0, + 1261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 1262.0, + 1399.0, + 1262.0, + 1399.0, + 1298.0, + 405.0, + 1298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 1300.0, + 447.0, + 1300.0, + 447.0, + 1335.0, + 404.0, + 1335.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 1300.0, + 1403.0, + 1300.0, + 1403.0, + 1332.0, + 449.0, + 1332.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 451.0, + 1330.0, + 705.0, + 1330.0, + 705.0, + 1363.0, + 451.0, + 1363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 407.0, + 1364.0, + 1402.0, + 1364.0, + 1402.0, + 1400.0, + 407.0, + 1400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 1395.0, + 904.0, + 1395.0, + 904.0, + 1430.0, + 450.0, + 1430.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 296.0, + 1405.0, + 296.0, + 1405.0, + 335.0, + 404.0, + 335.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 328.0, + 802.0, + 328.0, + 802.0, + 364.0, + 449.0, + 364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 363.0, + 1046.0, + 363.0, + 1046.0, + 400.0, + 404.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 401.0, + 1295.0, + 401.0, + 1295.0, + 437.0, + 405.0, + 437.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 435.0, + 1405.0, + 435.0, + 1405.0, + 472.0, + 404.0, + 472.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 448.0, + 462.0, + 597.0, + 462.0, + 597.0, + 504.0, + 448.0, + 504.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 406.0, + 1478.0, + 1407.0, + 1478.0, + 1407.0, + 1514.0, + 406.0, + 1514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 1508.0, + 1011.0, + 1508.0, + 1011.0, + 1548.0, + 450.0, + 1548.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 1544.0, + 1406.0, + 1544.0, + 1406.0, + 1580.0, + 405.0, + 1580.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 448.0, + 1573.0, + 1299.0, + 1573.0, + 1299.0, + 1613.0, + 448.0, + 1613.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 407.0, + 1612.0, + 1406.0, + 1612.0, + 1406.0, + 1648.0, + 407.0, + 1648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 448.0, + 1641.0, + 1275.0, + 1641.0, + 1275.0, + 1681.0, + 448.0, + 1681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 403.0, + 546.0, + 1407.0, + 546.0, + 1407.0, + 586.0, + 403.0, + 586.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 580.0, + 772.0, + 580.0, + 772.0, + 613.0, + 450.0, + 613.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 402.0, + 613.0, + 1405.0, + 613.0, + 1405.0, + 654.0, + 402.0, + 654.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 451.0, + 646.0, + 661.0, + 646.0, + 661.0, + 679.0, + 451.0, + 679.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 359.0, + 685.0, + 674.0, + 685.0, + 674.0, + 725.0, + 359.0, + 725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 360.0, + 507.0, + 845.0, + 507.0, + 845.0, + 544.0, + 360.0, + 544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 361.0, + 253.0, + 586.0, + 253.0, + 586.0, + 292.0, + 361.0, + 292.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 357.0, + 1431.0, + 1205.0, + 1431.0, + 1205.0, + 1477.0, + 357.0, + 1477.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 359.0, + 1116.0, + 1408.0, + 1116.0, + 1408.0, + 1159.0, + 359.0, + 1159.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 15, + "width": 1700, + "height": 2200 + } + } +] \ No newline at end of file diff --git a/parse/dev/sP1fo2K9DFG/sP1fo2K9DFG_middle.json b/parse/dev/sP1fo2K9DFG/sP1fo2K9DFG_middle.json new file mode 100644 index 0000000000000000000000000000000000000000..6f05db69b3802c623c918c01a2479dfc3ce93988 --- /dev/null +++ b/parse/dev/sP1fo2K9DFG/sP1fo2K9DFG_middle.json @@ -0,0 +1,61222 @@ +{ + "pdf_info": [ + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 79, + 505, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 77, + 505, + 98 + ], + "spans": [ + { + "bbox": [ + 105, + 77, + 505, + 98 + ], + "score": 1.0, + "content": "IS CONDITIONAL GENERATIVE MODELING ALL YOU", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 99, + 341, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 99, + 341, + 117 + ], + "score": 1.0, + "content": "NEED FOR DECISION-MAKING?", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 115, + 134, + 512, + 158 + ], + "lines": [ + { + "bbox": [ + 113, + 132, + 514, + 149 + ], + "spans": [ + { + "bbox": [ + 113, + 132, + 308, + 149 + ], + "score": 1.0, + "content": "Anurag Ajay∗ †§¶, Yilun Du *§¶, Abhi Gupta", + "type": "text" + }, + { + "bbox": [ + 308, + 134, + 325, + 145 + ], + "score": 0.47, + "content": "\\ast \\ddag \\mathfrak { S } \\ P", + "type": "inline_equation" + }, + { + "bbox": [ + 325, + 132, + 514, + 149 + ], + "score": 1.0, + "content": ", Joshua Tenenbaum¶, Tommi Jaakkola‡§¶,", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 113, + 145, + 195, + 159 + ], + "spans": [ + { + "bbox": [ + 113, + 145, + 195, + 159 + ], + "score": 1.0, + "content": "Pulkit Agrawal†§¶", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 115, + 159, + 316, + 204 + ], + "lines": [ + { + "bbox": [ + 114, + 158, + 197, + 169 + ], + "spans": [ + { + "bbox": [ + 114, + 158, + 197, + 169 + ], + "score": 1.0, + "content": "Improbable AI Lab†", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 114, + 169, + 233, + 181 + ], + "spans": [ + { + "bbox": [ + 114, + 169, + 233, + 181 + ], + "score": 1.0, + "content": "Operations Research Center‡", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 114, + 181, + 317, + 193 + ], + "spans": [ + { + "bbox": [ + 114, + 181, + 317, + 193 + ], + "score": 1.0, + "content": "Computer Science and Artificial Intelligence Lab§", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 114, + 192, + 275, + 206 + ], + "spans": [ + { + "bbox": [ + 114, + 192, + 275, + 206 + ], + "score": 1.0, + "content": "Massachusetts Institute of Technology¶", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5.5 + }, + { + "type": "title", + "bbox": [ + 278, + 234, + 333, + 245 + ], + "lines": [ + { + "bbox": [ + 276, + 232, + 336, + 248 + ], + "spans": [ + { + "bbox": [ + 276, + 232, + 336, + 248 + ], + "score": 1.0, + "content": "ABSTRACT", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 143, + 261, + 469, + 425 + ], + "lines": [ + { + "bbox": [ + 142, + 262, + 469, + 273 + ], + "spans": [ + { + "bbox": [ + 142, + 262, + 469, + 273 + ], + "score": 1.0, + "content": "Recent improvements in conditional generative modeling have made it possible", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 141, + 272, + 469, + 285 + ], + "spans": [ + { + "bbox": [ + 141, + 272, + 469, + 285 + ], + "score": 1.0, + "content": "to generate high-quality images from language descriptions alone. We investigate", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 282, + 470, + 295 + ], + "spans": [ + { + "bbox": [ + 141, + 282, + 470, + 295 + ], + "score": 1.0, + "content": "whether these methods can directly address the problem of sequential decision-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 293, + 470, + 308 + ], + "spans": [ + { + "bbox": [ + 141, + 293, + 470, + 308 + ], + "score": 1.0, + "content": "making. We view decision-making not through the lens of reinforcement learning", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 304, + 470, + 318 + ], + "spans": [ + { + "bbox": [ + 141, + 304, + 470, + 318 + ], + "score": 1.0, + "content": "(RL), but rather through conditional generative modeling. To our surprise, we", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 315, + 469, + 329 + ], + "spans": [ + { + "bbox": [ + 141, + 315, + 469, + 329 + ], + "score": 1.0, + "content": "find that our formulation leads to policies that can outperform existing offline", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 327, + 470, + 339 + ], + "spans": [ + { + "bbox": [ + 141, + 327, + 470, + 339 + ], + "score": 1.0, + "content": "RL approaches across standard benchmarks. By modeling a policy as a return-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 338, + 470, + 350 + ], + "spans": [ + { + "bbox": [ + 141, + 338, + 470, + 350 + ], + "score": 1.0, + "content": "conditional diffusion model, we illustrate how we may circumvent the need for", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 349, + 470, + 361 + ], + "spans": [ + { + "bbox": [ + 141, + 349, + 470, + 361 + ], + "score": 1.0, + "content": "dynamic programming and subsequently eliminate many of the complexities", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 359, + 470, + 373 + ], + "spans": [ + { + "bbox": [ + 141, + 359, + 470, + 373 + ], + "score": 1.0, + "content": "that come with traditional offline RL. We further demonstrate the advantages", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 371, + 469, + 383 + ], + "spans": [ + { + "bbox": [ + 141, + 371, + 469, + 383 + ], + "score": 1.0, + "content": "of modeling policies as conditional diffusion models by considering two other", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 142, + 382, + 469, + 394 + ], + "spans": [ + { + "bbox": [ + 142, + 382, + 469, + 394 + ], + "score": 1.0, + "content": "conditioning variables: constraints and skills. Conditioning on a single constraint", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 141, + 393, + 469, + 405 + ], + "spans": [ + { + "bbox": [ + 141, + 393, + 469, + 405 + ], + "score": 1.0, + "content": "or skill during training leads to behaviors at test-time that can satisfy several", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 141, + 404, + 469, + 415 + ], + "spans": [ + { + "bbox": [ + 141, + 404, + 469, + 415 + ], + "score": 1.0, + "content": "constraints together or demonstrate a composition of skills. Our results illustrate", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 141, + 413, + 449, + 428 + ], + "spans": [ + { + "bbox": [ + 141, + 413, + 449, + 428 + ], + "score": 1.0, + "content": "that conditional generative modeling is a powerful tool for decision-making.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 16 + }, + { + "type": "title", + "bbox": [ + 108, + 453, + 206, + 465 + ], + "lines": [ + { + "bbox": [ + 105, + 452, + 208, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 452, + 208, + 468 + ], + "score": 1.0, + "content": "1 INTRODUCTION", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 107, + 471, + 505, + 592 + ], + "lines": [ + { + "bbox": [ + 106, + 470, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 506, + 484 + ], + "score": 1.0, + "content": "Over the last few years, conditional generative modeling has yielded impressive results in a", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 481, + 506, + 494 + ], + "spans": [ + { + "bbox": [ + 106, + 481, + 506, + 494 + ], + "score": 1.0, + "content": "range of domains, including high-resolution image generation from text descriptions (DALL-E,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 493, + 506, + 505 + ], + "spans": [ + { + "bbox": [ + 106, + 493, + 506, + 505 + ], + "score": 1.0, + "content": "ImageGen) (Ramesh et al., 2022; Saharia et al., 2022), language generation (GPT) (Brown et al.,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 502, + 506, + 516 + ], + "spans": [ + { + "bbox": [ + 105, + 502, + 506, + 516 + ], + "score": 1.0, + "content": "2020), and step-by-step solutions to math problems (Minerva) (Lewkowycz et al., 2022). The", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 104, + 514, + 506, + 528 + ], + "spans": [ + { + "bbox": [ + 104, + 514, + 506, + 528 + ], + "score": 1.0, + "content": "success of generative models in countless domains motivates us to apply them to decision-making.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 525, + 506, + 538 + ], + "spans": [ + { + "bbox": [ + 105, + 525, + 506, + 538 + ], + "score": 1.0, + "content": "Conveniently, there exists a wide body of research on recovering high-performing policies from data", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 537, + 506, + 549 + ], + "spans": [ + { + "bbox": [ + 106, + 537, + 506, + 549 + ], + "score": 1.0, + "content": "logged by already operational systems (Kostrikov et al., 2022; Kumar et al., 2020; Walke et al., 2022).", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 547, + 505, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 505, + 560 + ], + "score": 1.0, + "content": "This is particularly useful in real-world settings where interacting with the environment is not always", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 559, + 505, + 570 + ], + "spans": [ + { + "bbox": [ + 105, + 559, + 505, + 570 + ], + "score": 1.0, + "content": "possible, and exploratory decisions can have fatal consequences (Dulac-Arnold et al., 2021). With", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 570, + 505, + 582 + ], + "spans": [ + { + "bbox": [ + 106, + 570, + 505, + 582 + ], + "score": 1.0, + "content": "access to such offline datasets, the problem of decision-making reduces to learning a probabilistic", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 581, + 444, + 592 + ], + "spans": [ + { + "bbox": [ + 106, + 581, + 444, + 592 + ], + "score": 1.0, + "content": "model of trajectories, a setting where generative models have already found success.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 107, + 597, + 505, + 707 + ], + "lines": [ + { + "bbox": [ + 105, + 596, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 105, + 596, + 505, + 610 + ], + "score": 1.0, + "content": "In offline decision-making, we aim to recover optimal reward-maximizing trajectories by stitching", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 608, + 506, + 621 + ], + "spans": [ + { + "bbox": [ + 105, + 608, + 506, + 621 + ], + "score": 1.0, + "content": "together sub-optimal reward-labeled trajectories in the training dataset. Prior works (Kumar et al.,", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 617, + 505, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 505, + 633 + ], + "score": 1.0, + "content": "2020; Kostrikov et al., 2022; Wu et al., 2019; Kostrikov et al., 2021; Dadashi et al., 2021; Ajay", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 630, + 505, + 642 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 505, + 642 + ], + "score": 1.0, + "content": "et al., 2020; Ghosh et al., 2022) have tackled this problem with reinforcement learning (RL) that uses", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 641, + 505, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 505, + 654 + ], + "score": 1.0, + "content": "dynamic programming for trajectory stitching. To enable dynamic programming, these works learn", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 653, + 504, + 664 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 504, + 664 + ], + "score": 1.0, + "content": "a value function that estimates the discounted sum of rewards from a given state. However, value", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 663, + 506, + 676 + ], + "spans": [ + { + "bbox": [ + 105, + 663, + 506, + 676 + ], + "score": 1.0, + "content": "function estimation is prone to instabilities due to function approximation, off-policy learning, and", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 674, + 505, + 687 + ], + "spans": [ + { + "bbox": [ + 105, + 674, + 505, + 687 + ], + "score": 1.0, + "content": "bootstrapping together, together known as the deadly triad (Sutton & Barto, 2018). Furthermore, to", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 684, + 505, + 698 + ], + "spans": [ + { + "bbox": [ + 105, + 684, + 505, + 698 + ], + "score": 1.0, + "content": "stabilize value estimation in offline regime, these works rely on heuristics to keep the policy within", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 696, + 497, + 708 + ], + "spans": [ + { + "bbox": [ + 106, + 696, + 497, + 708 + ], + "score": 1.0, + "content": "the dataset distribution. These challenges make it difficult to scale existing offline RL algorithms.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 40.5 + } + ], + "page_idx": 0, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 115, + 722, + 453, + 731 + ], + "lines": [ + { + "bbox": [ + 118, + 720, + 456, + 733 + ], + "spans": [ + { + "bbox": [ + 118, + 720, + 456, + 733 + ], + "score": 1.0, + "content": "∗ denotes equal contribution. Correspondence to aajay@mit.edu, yilundu@mit.edu, abhig@mit.edu", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 308, + 761 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 308, + 761 + ], + "score": 1.0, + "content": "1", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 79, + 505, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 77, + 505, + 98 + ], + "spans": [ + { + "bbox": [ + 105, + 77, + 505, + 98 + ], + "score": 1.0, + "content": "IS CONDITIONAL GENERATIVE MODELING ALL YOU", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 99, + 341, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 99, + 341, + 117 + ], + "score": 1.0, + "content": "NEED FOR DECISION-MAKING?", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 115, + 134, + 512, + 158 + ], + "lines": [ + { + "bbox": [ + 113, + 132, + 514, + 149 + ], + "spans": [ + { + "bbox": [ + 113, + 132, + 308, + 149 + ], + "score": 1.0, + "content": "Anurag Ajay∗ †§¶, Yilun Du *§¶, Abhi Gupta", + "type": "text" + }, + { + "bbox": [ + 308, + 134, + 325, + 145 + ], + "score": 0.47, + "content": "\\ast \\ddag \\mathfrak { S } \\ P", + "type": "inline_equation" + }, + { + "bbox": [ + 325, + 132, + 514, + 149 + ], + "score": 1.0, + "content": ", Joshua Tenenbaum¶, Tommi Jaakkola‡§¶,", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 113, + 145, + 195, + 159 + ], + "spans": [ + { + "bbox": [ + 113, + 145, + 195, + 159 + ], + "score": 1.0, + "content": "Pulkit Agrawal†§¶", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5, + "bbox_fs": [ + 113, + 132, + 514, + 159 + ] + }, + { + "type": "list", + "bbox": [ + 115, + 159, + 316, + 204 + ], + "lines": [ + { + "bbox": [ + 114, + 158, + 197, + 169 + ], + "spans": [ + { + "bbox": [ + 114, + 158, + 197, + 169 + ], + "score": 1.0, + "content": "Improbable AI Lab†", + "type": "text" + } + ], + "index": 4, + "is_list_start_line": true + }, + { + "bbox": [ + 114, + 169, + 233, + 181 + ], + "spans": [ + { + "bbox": [ + 114, + 169, + 233, + 181 + ], + "score": 1.0, + "content": "Operations Research Center‡", + "type": "text" + } + ], + "index": 5, + "is_list_start_line": true + }, + { + "bbox": [ + 114, + 181, + 317, + 193 + ], + "spans": [ + { + "bbox": [ + 114, + 181, + 317, + 193 + ], + "score": 1.0, + "content": "Computer Science and Artificial Intelligence Lab§", + "type": "text" + } + ], + "index": 6, + "is_list_start_line": true + }, + { + "bbox": [ + 114, + 192, + 275, + 206 + ], + "spans": [ + { + "bbox": [ + 114, + 192, + 275, + 206 + ], + "score": 1.0, + "content": "Massachusetts Institute of Technology¶", + "type": "text" + } + ], + "index": 7, + "is_list_start_line": true + } + ], + "index": 5.5, + "bbox_fs": [ + 114, + 158, + 317, + 206 + ] + }, + { + "type": "title", + "bbox": [ + 278, + 234, + 333, + 245 + ], + "lines": [ + { + "bbox": [ + 276, + 232, + 336, + 248 + ], + "spans": [ + { + "bbox": [ + 276, + 232, + 336, + 248 + ], + "score": 1.0, + "content": "ABSTRACT", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 143, + 261, + 469, + 425 + ], + "lines": [ + { + "bbox": [ + 142, + 262, + 469, + 273 + ], + "spans": [ + { + "bbox": [ + 142, + 262, + 469, + 273 + ], + "score": 1.0, + "content": "Recent improvements in conditional generative modeling have made it possible", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 141, + 272, + 469, + 285 + ], + "spans": [ + { + "bbox": [ + 141, + 272, + 469, + 285 + ], + "score": 1.0, + "content": "to generate high-quality images from language descriptions alone. We investigate", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 282, + 470, + 295 + ], + "spans": [ + { + "bbox": [ + 141, + 282, + 470, + 295 + ], + "score": 1.0, + "content": "whether these methods can directly address the problem of sequential decision-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 293, + 470, + 308 + ], + "spans": [ + { + "bbox": [ + 141, + 293, + 470, + 308 + ], + "score": 1.0, + "content": "making. We view decision-making not through the lens of reinforcement learning", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 304, + 470, + 318 + ], + "spans": [ + { + "bbox": [ + 141, + 304, + 470, + 318 + ], + "score": 1.0, + "content": "(RL), but rather through conditional generative modeling. To our surprise, we", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 315, + 469, + 329 + ], + "spans": [ + { + "bbox": [ + 141, + 315, + 469, + 329 + ], + "score": 1.0, + "content": "find that our formulation leads to policies that can outperform existing offline", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 327, + 470, + 339 + ], + "spans": [ + { + "bbox": [ + 141, + 327, + 470, + 339 + ], + "score": 1.0, + "content": "RL approaches across standard benchmarks. By modeling a policy as a return-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 338, + 470, + 350 + ], + "spans": [ + { + "bbox": [ + 141, + 338, + 470, + 350 + ], + "score": 1.0, + "content": "conditional diffusion model, we illustrate how we may circumvent the need for", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 349, + 470, + 361 + ], + "spans": [ + { + "bbox": [ + 141, + 349, + 470, + 361 + ], + "score": 1.0, + "content": "dynamic programming and subsequently eliminate many of the complexities", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 359, + 470, + 373 + ], + "spans": [ + { + "bbox": [ + 141, + 359, + 470, + 373 + ], + "score": 1.0, + "content": "that come with traditional offline RL. We further demonstrate the advantages", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 371, + 469, + 383 + ], + "spans": [ + { + "bbox": [ + 141, + 371, + 469, + 383 + ], + "score": 1.0, + "content": "of modeling policies as conditional diffusion models by considering two other", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 142, + 382, + 469, + 394 + ], + "spans": [ + { + "bbox": [ + 142, + 382, + 469, + 394 + ], + "score": 1.0, + "content": "conditioning variables: constraints and skills. Conditioning on a single constraint", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 141, + 393, + 469, + 405 + ], + "spans": [ + { + "bbox": [ + 141, + 393, + 469, + 405 + ], + "score": 1.0, + "content": "or skill during training leads to behaviors at test-time that can satisfy several", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 141, + 404, + 469, + 415 + ], + "spans": [ + { + "bbox": [ + 141, + 404, + 469, + 415 + ], + "score": 1.0, + "content": "constraints together or demonstrate a composition of skills. Our results illustrate", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 141, + 413, + 449, + 428 + ], + "spans": [ + { + "bbox": [ + 141, + 413, + 449, + 428 + ], + "score": 1.0, + "content": "that conditional generative modeling is a powerful tool for decision-making.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 16, + "bbox_fs": [ + 141, + 262, + 470, + 428 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 453, + 206, + 465 + ], + "lines": [ + { + "bbox": [ + 105, + 452, + 208, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 452, + 208, + 468 + ], + "score": 1.0, + "content": "1 INTRODUCTION", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 107, + 471, + 505, + 592 + ], + "lines": [ + { + "bbox": [ + 106, + 470, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 470, + 506, + 484 + ], + "score": 1.0, + "content": "Over the last few years, conditional generative modeling has yielded impressive results in a", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 481, + 506, + 494 + ], + "spans": [ + { + "bbox": [ + 106, + 481, + 506, + 494 + ], + "score": 1.0, + "content": "range of domains, including high-resolution image generation from text descriptions (DALL-E,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 493, + 506, + 505 + ], + "spans": [ + { + "bbox": [ + 106, + 493, + 506, + 505 + ], + "score": 1.0, + "content": "ImageGen) (Ramesh et al., 2022; Saharia et al., 2022), language generation (GPT) (Brown et al.,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 502, + 506, + 516 + ], + "spans": [ + { + "bbox": [ + 105, + 502, + 506, + 516 + ], + "score": 1.0, + "content": "2020), and step-by-step solutions to math problems (Minerva) (Lewkowycz et al., 2022). The", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 104, + 514, + 506, + 528 + ], + "spans": [ + { + "bbox": [ + 104, + 514, + 506, + 528 + ], + "score": 1.0, + "content": "success of generative models in countless domains motivates us to apply them to decision-making.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 525, + 506, + 538 + ], + "spans": [ + { + "bbox": [ + 105, + 525, + 506, + 538 + ], + "score": 1.0, + "content": "Conveniently, there exists a wide body of research on recovering high-performing policies from data", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 537, + 506, + 549 + ], + "spans": [ + { + "bbox": [ + 106, + 537, + 506, + 549 + ], + "score": 1.0, + "content": "logged by already operational systems (Kostrikov et al., 2022; Kumar et al., 2020; Walke et al., 2022).", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 547, + 505, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 505, + 560 + ], + "score": 1.0, + "content": "This is particularly useful in real-world settings where interacting with the environment is not always", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 559, + 505, + 570 + ], + "spans": [ + { + "bbox": [ + 105, + 559, + 505, + 570 + ], + "score": 1.0, + "content": "possible, and exploratory decisions can have fatal consequences (Dulac-Arnold et al., 2021). With", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 570, + 505, + 582 + ], + "spans": [ + { + "bbox": [ + 106, + 570, + 505, + 582 + ], + "score": 1.0, + "content": "access to such offline datasets, the problem of decision-making reduces to learning a probabilistic", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 581, + 444, + 592 + ], + "spans": [ + { + "bbox": [ + 106, + 581, + 444, + 592 + ], + "score": 1.0, + "content": "model of trajectories, a setting where generative models have already found success.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 30, + "bbox_fs": [ + 104, + 470, + 506, + 592 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 597, + 505, + 707 + ], + "lines": [ + { + "bbox": [ + 105, + 596, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 105, + 596, + 505, + 610 + ], + "score": 1.0, + "content": "In offline decision-making, we aim to recover optimal reward-maximizing trajectories by stitching", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 608, + 506, + 621 + ], + "spans": [ + { + "bbox": [ + 105, + 608, + 506, + 621 + ], + "score": 1.0, + "content": "together sub-optimal reward-labeled trajectories in the training dataset. Prior works (Kumar et al.,", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 617, + 505, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 505, + 633 + ], + "score": 1.0, + "content": "2020; Kostrikov et al., 2022; Wu et al., 2019; Kostrikov et al., 2021; Dadashi et al., 2021; Ajay", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 630, + 505, + 642 + ], + "spans": [ + { + "bbox": [ + 105, + 630, + 505, + 642 + ], + "score": 1.0, + "content": "et al., 2020; Ghosh et al., 2022) have tackled this problem with reinforcement learning (RL) that uses", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 641, + 505, + 654 + ], + "spans": [ + { + "bbox": [ + 105, + 641, + 505, + 654 + ], + "score": 1.0, + "content": "dynamic programming for trajectory stitching. To enable dynamic programming, these works learn", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 653, + 504, + 664 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 504, + 664 + ], + "score": 1.0, + "content": "a value function that estimates the discounted sum of rewards from a given state. However, value", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 663, + 506, + 676 + ], + "spans": [ + { + "bbox": [ + 105, + 663, + 506, + 676 + ], + "score": 1.0, + "content": "function estimation is prone to instabilities due to function approximation, off-policy learning, and", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 674, + 505, + 687 + ], + "spans": [ + { + "bbox": [ + 105, + 674, + 505, + 687 + ], + "score": 1.0, + "content": "bootstrapping together, together known as the deadly triad (Sutton & Barto, 2018). Furthermore, to", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 684, + 505, + 698 + ], + "spans": [ + { + "bbox": [ + 105, + 684, + 505, + 698 + ], + "score": 1.0, + "content": "stabilize value estimation in offline regime, these works rely on heuristics to keep the policy within", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 696, + 497, + 708 + ], + "spans": [ + { + "bbox": [ + 106, + 696, + 497, + 708 + ], + "score": 1.0, + "content": "the dataset distribution. These challenges make it difficult to scale existing offline RL algorithms.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 40.5, + "bbox_fs": [ + 105, + 596, + 506, + 708 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 185, + 79, + 424, + 140 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 185, + 79, + 424, + 140 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 185, + 79, + 424, + 140 + ], + "spans": [ + { + "bbox": [ + 185, + 79, + 424, + 140 + ], + "score": 0.963, + "type": "image", + "image_path": "9dac5e5f723a941f045f73b69f622ccae89b7291e5b8f4eed92c65b334c4c613.jpg" + } + ] + } + ], + "index": 1.5, + "virtual_lines": [ + { + "bbox": [ + 185, + 79, + 424, + 94.25 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 185, + 94.25, + 424, + 109.5 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 185, + 109.5, + 424, + 124.75 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 185, + 124.75, + 424, + 140.0 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 150, + 504, + 171 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 149, + 506, + 162 + ], + "spans": [ + { + "bbox": [ + 105, + 149, + 506, + 162 + ], + "score": 1.0, + "content": "Figure 1: Decision Making using Conditional Generative Modeling. Framing decision making as a", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 160, + 505, + 171 + ], + "spans": [ + { + "bbox": [ + 106, + 160, + 505, + 171 + ], + "score": 1.0, + "content": "conditional generative modeling problem allows us to maximize rewards, satisfy constraints and compose skills.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4.5 + } + ], + "index": 3.0 + }, + { + "type": "text", + "bbox": [ + 106, + 176, + 505, + 286 + ], + "lines": [ + { + "bbox": [ + 105, + 176, + 505, + 189 + ], + "spans": [ + { + "bbox": [ + 105, + 176, + 505, + 189 + ], + "score": 1.0, + "content": "In this paper, we ask if we can perform dynamic programming to stitch together sub-optimal", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 187, + 505, + 199 + ], + "spans": [ + { + "bbox": [ + 105, + 187, + 505, + 199 + ], + "score": 1.0, + "content": "trajectories to obtain an optimal trajectory without relying on value estimation. Since conditional", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 198, + 506, + 211 + ], + "spans": [ + { + "bbox": [ + 106, + 198, + 506, + 211 + ], + "score": 1.0, + "content": "diffusion generative models can generate novel data points by composing training data (Saharia", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 209, + 505, + 221 + ], + "spans": [ + { + "bbox": [ + 106, + 209, + 505, + 221 + ], + "score": 1.0, + "content": "et al., 2022), we leverage it for trajectory stitching in offline decision-making. Given a dataset", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 219, + 506, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 219, + 506, + 233 + ], + "score": 1.0, + "content": "of reward-labeled trajectories, we adapt diffusion models (Sohl-Dickstein et al., 2015) to learn a", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 231, + 506, + 243 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 506, + 243 + ], + "score": 1.0, + "content": "return-conditional trajectory model. During inference, we use classifier-free guidance with low-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 242, + 506, + 255 + ], + "spans": [ + { + "bbox": [ + 105, + 242, + 506, + 255 + ], + "score": 1.0, + "content": "temperature sampling, which we hypothesize to implicitly perform dynamics programming, to", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 253, + 506, + 265 + ], + "spans": [ + { + "bbox": [ + 105, + 253, + 506, + 265 + ], + "score": 1.0, + "content": "capture the best behaviors in the dataset and glean return maximizing trajectories (see Appendix A).", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 263, + 505, + 277 + ], + "spans": [ + { + "bbox": [ + 105, + 263, + 505, + 277 + ], + "score": 1.0, + "content": "Our straightforward conditional generative modeling formulation outperforms existing approaches", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 275, + 274, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 275, + 274, + 287 + ], + "score": 1.0, + "content": "on standard D4RL tasks (Fu et al., 2020).", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 10.5 + }, + { + "type": "text", + "bbox": [ + 106, + 291, + 506, + 401 + ], + "lines": [ + { + "bbox": [ + 105, + 289, + 506, + 306 + ], + "spans": [ + { + "bbox": [ + 105, + 289, + 506, + 306 + ], + "score": 1.0, + "content": "Viewing offline decision-making through the lens of conditional generative modeling allows going", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 303, + 506, + 316 + ], + "spans": [ + { + "bbox": [ + 106, + 303, + 506, + 316 + ], + "score": 1.0, + "content": "beyond conditioning on returns (Figure 1). Consider an example (detailed in Appendix A) where a", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 314, + 506, + 326 + ], + "spans": [ + { + "bbox": [ + 105, + 314, + 506, + 326 + ], + "score": 1.0, + "content": "robot with linear dynamics navigates an environment containing two concentric circles (Figure 2). We", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 325, + 506, + 337 + ], + "spans": [ + { + "bbox": [ + 105, + 325, + 506, + 337 + ], + "score": 1.0, + "content": "are given a dataset of state-action trajectories of the robot, each satisfying one of two constraints: (i)", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 336, + 505, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 336, + 505, + 347 + ], + "score": 1.0, + "content": "the final position of the robot is within the larger circle, and (ii) the final position of the robot is outside", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 347, + 507, + 359 + ], + "spans": [ + { + "bbox": [ + 106, + 347, + 507, + 359 + ], + "score": 1.0, + "content": "the smaller circle. With conditional diffusion modeling, we can use the datasets to learn a constraint-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 356, + 507, + 371 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 507, + 371 + ], + "score": 1.0, + "content": "conditioned model that can generate trajectories satisfying any set of constraints. During inference,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 368, + 506, + 381 + ], + "spans": [ + { + "bbox": [ + 105, + 368, + 506, + 381 + ], + "score": 1.0, + "content": "the learned trajectory model can merge constraints from the dataset and generate trajectories that", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 380, + 505, + 392 + ], + "spans": [ + { + "bbox": [ + 106, + 380, + 505, + 392 + ], + "score": 1.0, + "content": "satisfy the combined constraint. Figure 2 shows that the constraint-conditioned model can generate", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 391, + 452, + 403 + ], + "spans": [ + { + "bbox": [ + 106, + 391, + 452, + 403 + ], + "score": 1.0, + "content": "trajectories such that the final position of the robot lies between the concentric circles.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 20.5 + }, + { + "type": "image", + "bbox": [ + 206, + 402, + 405, + 460 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 206, + 402, + 405, + 460 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 206, + 402, + 405, + 460 + ], + "spans": [ + { + "bbox": [ + 206, + 402, + 405, + 460 + ], + "score": 0.965, + "type": "image", + "image_path": "3f638c6b9477f25a868e5f8e34ac18c10ed64b5789def600b6b4750a38198356.jpg" + } + ] + } + ], + "index": 27.5, + "virtual_lines": [ + { + "bbox": [ + 206, + 402, + 405, + 416.5 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 206, + 416.5, + 405, + 431.0 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 206, + 431.0, + 405, + 445.5 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 206, + 445.5, + 405, + 460.0 + ], + "spans": [], + "index": 29 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 469, + 504, + 499 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 468, + 506, + 480 + ], + "spans": [ + { + "bbox": [ + 106, + 468, + 506, + 480 + ], + "score": 1.0, + "content": "Figure 2: Illustrative example. We visualize the 2d robot navigation environment and the constraints satisfied", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 479, + 506, + 491 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 506, + 491 + ], + "score": 1.0, + "content": "by the trajectories in the dataset derived from the environment. We show the ability of the conditional diffusion", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 489, + 351, + 500 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 351, + 500 + ], + "score": 1.0, + "content": "model to generate trajectories that satisfy the combined constraints.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31 + } + ], + "index": 29.25 + }, + { + "type": "text", + "bbox": [ + 106, + 502, + 506, + 612 + ], + "lines": [ + { + "bbox": [ + 106, + 502, + 506, + 515 + ], + "spans": [ + { + "bbox": [ + 106, + 502, + 506, + 515 + ], + "score": 1.0, + "content": "Here, we demonstrate the benefits of modeling policies as conditional generative models. First,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 513, + 506, + 526 + ], + "spans": [ + { + "bbox": [ + 106, + 513, + 506, + 526 + ], + "score": 1.0, + "content": "conditioning on constraints allows policies to not only generate behaviors satisfying individual", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 524, + 506, + 537 + ], + "spans": [ + { + "bbox": [ + 106, + 524, + 506, + 537 + ], + "score": 1.0, + "content": "constraints but also generate novel behaviors by flexibly combining constraints at test time. Further,", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 535, + 505, + 548 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 505, + 548 + ], + "score": 1.0, + "content": "conditioning on skills allows policies to not only imitate individual skills but also generate novel", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 104, + 545, + 506, + 559 + ], + "spans": [ + { + "bbox": [ + 104, + 545, + 506, + 559 + ], + "score": 1.0, + "content": "behaviors by composing those skills. We instantiate this idea with a state-sequence based diffusion", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 104, + 556, + 507, + 571 + ], + "spans": [ + { + "bbox": [ + 104, + 556, + 507, + 571 + ], + "score": 1.0, + "content": "probabilistic model (Ho et al., 2020) called Decision Diffuser, visualized in Figure 1. In summary,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 567, + 506, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 567, + 506, + 581 + ], + "score": 1.0, + "content": "our contributions include (i) illustrating conditional generative modeling as an effective tool in", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 578, + 506, + 592 + ], + "spans": [ + { + "bbox": [ + 105, + 578, + 506, + 592 + ], + "score": 1.0, + "content": "offline decision making, (ii) using classifier-free guidance with low-temperature sampling, instead of", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 589, + 506, + 603 + ], + "spans": [ + { + "bbox": [ + 105, + 589, + 506, + 603 + ], + "score": 1.0, + "content": "dynamic programming, to get return-maximizing trajectories and, (iii) leveraging the framework of", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 601, + 501, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 601, + 501, + 613 + ], + "score": 1.0, + "content": "conditional generative modeling to combine constraints and compose skills during inference flexibly.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 37.5 + }, + { + "type": "title", + "bbox": [ + 107, + 618, + 200, + 630 + ], + "lines": [ + { + "bbox": [ + 104, + 616, + 202, + 633 + ], + "spans": [ + { + "bbox": [ + 104, + 616, + 202, + 633 + ], + "score": 1.0, + "content": "2 BACKGROUND", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 43 + }, + { + "type": "title", + "bbox": [ + 108, + 634, + 255, + 645 + ], + "lines": [ + { + "bbox": [ + 106, + 634, + 255, + 647 + ], + "spans": [ + { + "bbox": [ + 106, + 634, + 255, + 647 + ], + "score": 1.0, + "content": "2.1 REINFORCEMENT LEARNING", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 44 + }, + { + "type": "text", + "bbox": [ + 106, + 651, + 505, + 733 + ], + "lines": [ + { + "bbox": [ + 105, + 651, + 506, + 664 + ], + "spans": [ + { + "bbox": [ + 105, + 651, + 506, + 664 + ], + "score": 1.0, + "content": "We formulate the sequential decision-making problem as a discounted Markov Decision Process", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 663, + 505, + 676 + ], + "spans": [ + { + "bbox": [ + 105, + 663, + 216, + 676 + ], + "score": 1.0, + "content": "(MDP) defined by the tuple", + "type": "text" + }, + { + "bbox": [ + 217, + 663, + 293, + 674 + ], + "score": 0.91, + "content": "\\langle \\rho _ { 0 } , S , { \\mathcal { A } } , { \\mathcal { T } } , { \\mathcal { R } } , { \\gamma } \\rangle", + "type": "inline_equation" + }, + { + "bbox": [ + 294, + 663, + 323, + 676 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 324, + 664, + 335, + 675 + ], + "score": 0.86, + "content": "\\rho _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 335, + 663, + 454, + 676 + ], + "score": 1.0, + "content": "is the initial state distribution,", + "type": "text" + }, + { + "bbox": [ + 455, + 663, + 462, + 673 + ], + "score": 0.79, + "content": "s", + "type": "inline_equation" + }, + { + "bbox": [ + 463, + 663, + 480, + 676 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 480, + 663, + 489, + 673 + ], + "score": 0.78, + "content": "\\mathcal { A }", + "type": "inline_equation" + }, + { + "bbox": [ + 490, + 663, + 505, + 676 + ], + "score": 1.0, + "content": "are", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 673, + 505, + 686 + ], + "spans": [ + { + "bbox": [ + 105, + 673, + 204, + 686 + ], + "score": 1.0, + "content": "state and action spaces,", + "type": "text" + }, + { + "bbox": [ + 204, + 675, + 272, + 684 + ], + "score": 0.88, + "content": "\\mathcal { T } : \\mathcal { S } \\times \\mathcal { A } \\mathcal { S }", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 673, + 378, + 686 + ], + "score": 1.0, + "content": "is the transition function,", + "type": "text" + }, + { + "bbox": [ + 378, + 673, + 465, + 684 + ], + "score": 0.91, + "content": "\\mathcal { R } : \\mathcal { S } \\times \\mathcal { A } \\times \\mathcal { S } \\mathbb { R }", + "type": "inline_equation" + }, + { + "bbox": [ + 465, + 673, + 505, + 686 + ], + "score": 1.0, + "content": "gives the", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 684, + 506, + 698 + ], + "spans": [ + { + "bbox": [ + 105, + 684, + 218, + 698 + ], + "score": 1.0, + "content": "reward at any transition and", + "type": "text" + }, + { + "bbox": [ + 218, + 685, + 258, + 697 + ], + "score": 0.9, + "content": "\\gamma \\in [ 0 , 1 )", + "type": "inline_equation" + }, + { + "bbox": [ + 258, + 684, + 492, + 698 + ], + "score": 1.0, + "content": "is a discount factor. The agent acts with a stochastic policy", + "type": "text" + }, + { + "bbox": [ + 492, + 686, + 506, + 695 + ], + "score": 0.5, + "content": "\\pi :", + "type": "inline_equation" + } + ], + "index": 48 + }, + { + "bbox": [ + 106, + 693, + 504, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 696, + 145, + 707 + ], + "score": 0.9, + "content": "{ \\cal S } \\to \\Delta _ { \\cal A }", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 693, + 421, + 711 + ], + "score": 1.0, + "content": ", generating a sequence of state-action-reward transitions or trajectory", + "type": "text" + }, + { + "bbox": [ + 421, + 696, + 504, + 708 + ], + "score": 0.92, + "content": "\\tau : = ( s _ { k } , a _ { k } , r _ { k } ) _ { k \\geq 0 }", + "type": "inline_equation" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 707, + 507, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 707, + 174, + 722 + ], + "score": 1.0, + "content": "with probability", + "type": "text" + }, + { + "bbox": [ + 174, + 708, + 199, + 720 + ], + "score": 0.93, + "content": "p _ { \\pi } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 199, + 707, + 245, + 722 + ], + "score": 1.0, + "content": "and return", + "type": "text" + }, + { + "bbox": [ + 246, + 707, + 331, + 721 + ], + "score": 0.95, + "content": "\\begin{array} { r } { R ( \\tau ) : = \\sum _ { k \\geq 0 } \\gamma ^ { k } r _ { k } } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 332, + 707, + 507, + 722 + ], + "score": 1.0, + "content": ". The standard objective in RL is to find a", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 720, + 335, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 211, + 734 + ], + "score": 1.0, + "content": "return-maximizing policy", + "type": "text" + }, + { + "bbox": [ + 211, + 721, + 330, + 733 + ], + "score": 0.89, + "content": "\\begin{array} { r } { \\pi ^ { * } = \\arg \\operatorname* { m a x } _ { \\pi } \\mathbb { E } _ { \\tau \\sim p _ { \\pi } } \\overline { { [ R ( \\tau ) ] } } } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 720, + 335, + 734 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 48 + } + ], + "page_idx": 1, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 763 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 763 + ], + "score": 1.0, + "content": "2", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 185, + 79, + 424, + 140 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 185, + 79, + 424, + 140 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 185, + 79, + 424, + 140 + ], + "spans": [ + { + "bbox": [ + 185, + 79, + 424, + 140 + ], + "score": 0.963, + "type": "image", + "image_path": "9dac5e5f723a941f045f73b69f622ccae89b7291e5b8f4eed92c65b334c4c613.jpg" + } + ] + } + ], + "index": 1.5, + "virtual_lines": [ + { + "bbox": [ + 185, + 79, + 424, + 94.25 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 185, + 94.25, + 424, + 109.5 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 185, + 109.5, + 424, + 124.75 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 185, + 124.75, + 424, + 140.0 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 150, + 504, + 171 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 149, + 506, + 162 + ], + "spans": [ + { + "bbox": [ + 105, + 149, + 506, + 162 + ], + "score": 1.0, + "content": "Figure 1: Decision Making using Conditional Generative Modeling. Framing decision making as a", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 160, + 505, + 171 + ], + "spans": [ + { + "bbox": [ + 106, + 160, + 505, + 171 + ], + "score": 1.0, + "content": "conditional generative modeling problem allows us to maximize rewards, satisfy constraints and compose skills.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4.5 + } + ], + "index": 3.0 + }, + { + "type": "text", + "bbox": [ + 106, + 176, + 505, + 286 + ], + "lines": [ + { + "bbox": [ + 105, + 176, + 505, + 189 + ], + "spans": [ + { + "bbox": [ + 105, + 176, + 505, + 189 + ], + "score": 1.0, + "content": "In this paper, we ask if we can perform dynamic programming to stitch together sub-optimal", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 187, + 505, + 199 + ], + "spans": [ + { + "bbox": [ + 105, + 187, + 505, + 199 + ], + "score": 1.0, + "content": "trajectories to obtain an optimal trajectory without relying on value estimation. Since conditional", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 198, + 506, + 211 + ], + "spans": [ + { + "bbox": [ + 106, + 198, + 506, + 211 + ], + "score": 1.0, + "content": "diffusion generative models can generate novel data points by composing training data (Saharia", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 209, + 505, + 221 + ], + "spans": [ + { + "bbox": [ + 106, + 209, + 505, + 221 + ], + "score": 1.0, + "content": "et al., 2022), we leverage it for trajectory stitching in offline decision-making. Given a dataset", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 219, + 506, + 233 + ], + "spans": [ + { + "bbox": [ + 105, + 219, + 506, + 233 + ], + "score": 1.0, + "content": "of reward-labeled trajectories, we adapt diffusion models (Sohl-Dickstein et al., 2015) to learn a", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 231, + 506, + 243 + ], + "spans": [ + { + "bbox": [ + 106, + 231, + 506, + 243 + ], + "score": 1.0, + "content": "return-conditional trajectory model. During inference, we use classifier-free guidance with low-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 242, + 506, + 255 + ], + "spans": [ + { + "bbox": [ + 105, + 242, + 506, + 255 + ], + "score": 1.0, + "content": "temperature sampling, which we hypothesize to implicitly perform dynamics programming, to", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 253, + 506, + 265 + ], + "spans": [ + { + "bbox": [ + 105, + 253, + 506, + 265 + ], + "score": 1.0, + "content": "capture the best behaviors in the dataset and glean return maximizing trajectories (see Appendix A).", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 263, + 505, + 277 + ], + "spans": [ + { + "bbox": [ + 105, + 263, + 505, + 277 + ], + "score": 1.0, + "content": "Our straightforward conditional generative modeling formulation outperforms existing approaches", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 275, + 274, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 275, + 274, + 287 + ], + "score": 1.0, + "content": "on standard D4RL tasks (Fu et al., 2020).", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 10.5, + "bbox_fs": [ + 105, + 176, + 506, + 287 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 291, + 506, + 401 + ], + "lines": [ + { + "bbox": [ + 105, + 289, + 506, + 306 + ], + "spans": [ + { + "bbox": [ + 105, + 289, + 506, + 306 + ], + "score": 1.0, + "content": "Viewing offline decision-making through the lens of conditional generative modeling allows going", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 303, + 506, + 316 + ], + "spans": [ + { + "bbox": [ + 106, + 303, + 506, + 316 + ], + "score": 1.0, + "content": "beyond conditioning on returns (Figure 1). Consider an example (detailed in Appendix A) where a", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 314, + 506, + 326 + ], + "spans": [ + { + "bbox": [ + 105, + 314, + 506, + 326 + ], + "score": 1.0, + "content": "robot with linear dynamics navigates an environment containing two concentric circles (Figure 2). We", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 325, + 506, + 337 + ], + "spans": [ + { + "bbox": [ + 105, + 325, + 506, + 337 + ], + "score": 1.0, + "content": "are given a dataset of state-action trajectories of the robot, each satisfying one of two constraints: (i)", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 336, + 505, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 336, + 505, + 347 + ], + "score": 1.0, + "content": "the final position of the robot is within the larger circle, and (ii) the final position of the robot is outside", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 347, + 507, + 359 + ], + "spans": [ + { + "bbox": [ + 106, + 347, + 507, + 359 + ], + "score": 1.0, + "content": "the smaller circle. With conditional diffusion modeling, we can use the datasets to learn a constraint-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 356, + 507, + 371 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 507, + 371 + ], + "score": 1.0, + "content": "conditioned model that can generate trajectories satisfying any set of constraints. During inference,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 368, + 506, + 381 + ], + "spans": [ + { + "bbox": [ + 105, + 368, + 506, + 381 + ], + "score": 1.0, + "content": "the learned trajectory model can merge constraints from the dataset and generate trajectories that", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 380, + 505, + 392 + ], + "spans": [ + { + "bbox": [ + 106, + 380, + 505, + 392 + ], + "score": 1.0, + "content": "satisfy the combined constraint. Figure 2 shows that the constraint-conditioned model can generate", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 391, + 452, + 403 + ], + "spans": [ + { + "bbox": [ + 106, + 391, + 452, + 403 + ], + "score": 1.0, + "content": "trajectories such that the final position of the robot lies between the concentric circles.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 20.5, + "bbox_fs": [ + 105, + 289, + 507, + 403 + ] + }, + { + "type": "image", + "bbox": [ + 206, + 402, + 405, + 460 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 206, + 402, + 405, + 460 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 206, + 402, + 405, + 460 + ], + "spans": [ + { + "bbox": [ + 206, + 402, + 405, + 460 + ], + "score": 0.965, + "type": "image", + "image_path": "3f638c6b9477f25a868e5f8e34ac18c10ed64b5789def600b6b4750a38198356.jpg" + } + ] + } + ], + "index": 27.5, + "virtual_lines": [ + { + "bbox": [ + 206, + 402, + 405, + 416.5 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 206, + 416.5, + 405, + 431.0 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 206, + 431.0, + 405, + 445.5 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 206, + 445.5, + 405, + 460.0 + ], + "spans": [], + "index": 29 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 469, + 504, + 499 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 468, + 506, + 480 + ], + "spans": [ + { + "bbox": [ + 106, + 468, + 506, + 480 + ], + "score": 1.0, + "content": "Figure 2: Illustrative example. We visualize the 2d robot navigation environment and the constraints satisfied", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 479, + 506, + 491 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 506, + 491 + ], + "score": 1.0, + "content": "by the trajectories in the dataset derived from the environment. We show the ability of the conditional diffusion", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 489, + 351, + 500 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 351, + 500 + ], + "score": 1.0, + "content": "model to generate trajectories that satisfy the combined constraints.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31 + } + ], + "index": 29.25 + }, + { + "type": "text", + "bbox": [ + 106, + 502, + 506, + 612 + ], + "lines": [ + { + "bbox": [ + 106, + 502, + 506, + 515 + ], + "spans": [ + { + "bbox": [ + 106, + 502, + 506, + 515 + ], + "score": 1.0, + "content": "Here, we demonstrate the benefits of modeling policies as conditional generative models. First,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 513, + 506, + 526 + ], + "spans": [ + { + "bbox": [ + 106, + 513, + 506, + 526 + ], + "score": 1.0, + "content": "conditioning on constraints allows policies to not only generate behaviors satisfying individual", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 524, + 506, + 537 + ], + "spans": [ + { + "bbox": [ + 106, + 524, + 506, + 537 + ], + "score": 1.0, + "content": "constraints but also generate novel behaviors by flexibly combining constraints at test time. Further,", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 535, + 505, + 548 + ], + "spans": [ + { + "bbox": [ + 105, + 535, + 505, + 548 + ], + "score": 1.0, + "content": "conditioning on skills allows policies to not only imitate individual skills but also generate novel", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 104, + 545, + 506, + 559 + ], + "spans": [ + { + "bbox": [ + 104, + 545, + 506, + 559 + ], + "score": 1.0, + "content": "behaviors by composing those skills. We instantiate this idea with a state-sequence based diffusion", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 104, + 556, + 507, + 571 + ], + "spans": [ + { + "bbox": [ + 104, + 556, + 507, + 571 + ], + "score": 1.0, + "content": "probabilistic model (Ho et al., 2020) called Decision Diffuser, visualized in Figure 1. In summary,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 567, + 506, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 567, + 506, + 581 + ], + "score": 1.0, + "content": "our contributions include (i) illustrating conditional generative modeling as an effective tool in", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 578, + 506, + 592 + ], + "spans": [ + { + "bbox": [ + 105, + 578, + 506, + 592 + ], + "score": 1.0, + "content": "offline decision making, (ii) using classifier-free guidance with low-temperature sampling, instead of", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 589, + 506, + 603 + ], + "spans": [ + { + "bbox": [ + 105, + 589, + 506, + 603 + ], + "score": 1.0, + "content": "dynamic programming, to get return-maximizing trajectories and, (iii) leveraging the framework of", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 601, + 501, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 601, + 501, + 613 + ], + "score": 1.0, + "content": "conditional generative modeling to combine constraints and compose skills during inference flexibly.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 37.5, + "bbox_fs": [ + 104, + 502, + 507, + 613 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 618, + 200, + 630 + ], + "lines": [ + { + "bbox": [ + 104, + 616, + 202, + 633 + ], + "spans": [ + { + "bbox": [ + 104, + 616, + 202, + 633 + ], + "score": 1.0, + "content": "2 BACKGROUND", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 43 + }, + { + "type": "title", + "bbox": [ + 108, + 634, + 255, + 645 + ], + "lines": [ + { + "bbox": [ + 106, + 634, + 255, + 647 + ], + "spans": [ + { + "bbox": [ + 106, + 634, + 255, + 647 + ], + "score": 1.0, + "content": "2.1 REINFORCEMENT LEARNING", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 44 + }, + { + "type": "text", + "bbox": [ + 106, + 651, + 505, + 733 + ], + "lines": [ + { + "bbox": [ + 105, + 651, + 506, + 664 + ], + "spans": [ + { + "bbox": [ + 105, + 651, + 506, + 664 + ], + "score": 1.0, + "content": "We formulate the sequential decision-making problem as a discounted Markov Decision Process", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 663, + 505, + 676 + ], + "spans": [ + { + "bbox": [ + 105, + 663, + 216, + 676 + ], + "score": 1.0, + "content": "(MDP) defined by the tuple", + "type": "text" + }, + { + "bbox": [ + 217, + 663, + 293, + 674 + ], + "score": 0.91, + "content": "\\langle \\rho _ { 0 } , S , { \\mathcal { A } } , { \\mathcal { T } } , { \\mathcal { R } } , { \\gamma } \\rangle", + "type": "inline_equation" + }, + { + "bbox": [ + 294, + 663, + 323, + 676 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 324, + 664, + 335, + 675 + ], + "score": 0.86, + "content": "\\rho _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 335, + 663, + 454, + 676 + ], + "score": 1.0, + "content": "is the initial state distribution,", + "type": "text" + }, + { + "bbox": [ + 455, + 663, + 462, + 673 + ], + "score": 0.79, + "content": "s", + "type": "inline_equation" + }, + { + "bbox": [ + 463, + 663, + 480, + 676 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 480, + 663, + 489, + 673 + ], + "score": 0.78, + "content": "\\mathcal { A }", + "type": "inline_equation" + }, + { + "bbox": [ + 490, + 663, + 505, + 676 + ], + "score": 1.0, + "content": "are", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 673, + 505, + 686 + ], + "spans": [ + { + "bbox": [ + 105, + 673, + 204, + 686 + ], + "score": 1.0, + "content": "state and action spaces,", + "type": "text" + }, + { + "bbox": [ + 204, + 675, + 272, + 684 + ], + "score": 0.88, + "content": "\\mathcal { T } : \\mathcal { S } \\times \\mathcal { A } \\mathcal { S }", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 673, + 378, + 686 + ], + "score": 1.0, + "content": "is the transition function,", + "type": "text" + }, + { + "bbox": [ + 378, + 673, + 465, + 684 + ], + "score": 0.91, + "content": "\\mathcal { R } : \\mathcal { S } \\times \\mathcal { A } \\times \\mathcal { S } \\mathbb { R }", + "type": "inline_equation" + }, + { + "bbox": [ + 465, + 673, + 505, + 686 + ], + "score": 1.0, + "content": "gives the", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 684, + 506, + 698 + ], + "spans": [ + { + "bbox": [ + 105, + 684, + 218, + 698 + ], + "score": 1.0, + "content": "reward at any transition and", + "type": "text" + }, + { + "bbox": [ + 218, + 685, + 258, + 697 + ], + "score": 0.9, + "content": "\\gamma \\in [ 0 , 1 )", + "type": "inline_equation" + }, + { + "bbox": [ + 258, + 684, + 492, + 698 + ], + "score": 1.0, + "content": "is a discount factor. The agent acts with a stochastic policy", + "type": "text" + }, + { + "bbox": [ + 492, + 686, + 506, + 695 + ], + "score": 0.5, + "content": "\\pi :", + "type": "inline_equation" + } + ], + "index": 48 + }, + { + "bbox": [ + 106, + 693, + 504, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 696, + 145, + 707 + ], + "score": 0.9, + "content": "{ \\cal S } \\to \\Delta _ { \\cal A }", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 693, + 421, + 711 + ], + "score": 1.0, + "content": ", generating a sequence of state-action-reward transitions or trajectory", + "type": "text" + }, + { + "bbox": [ + 421, + 696, + 504, + 708 + ], + "score": 0.92, + "content": "\\tau : = ( s _ { k } , a _ { k } , r _ { k } ) _ { k \\geq 0 }", + "type": "inline_equation" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 707, + 507, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 707, + 174, + 722 + ], + "score": 1.0, + "content": "with probability", + "type": "text" + }, + { + "bbox": [ + 174, + 708, + 199, + 720 + ], + "score": 0.93, + "content": "p _ { \\pi } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 199, + 707, + 245, + 722 + ], + "score": 1.0, + "content": "and return", + "type": "text" + }, + { + "bbox": [ + 246, + 707, + 331, + 721 + ], + "score": 0.95, + "content": "\\begin{array} { r } { R ( \\tau ) : = \\sum _ { k \\geq 0 } \\gamma ^ { k } r _ { k } } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 332, + 707, + 507, + 722 + ], + "score": 1.0, + "content": ". The standard objective in RL is to find a", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 720, + 335, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 211, + 734 + ], + "score": 1.0, + "content": "return-maximizing policy", + "type": "text" + }, + { + "bbox": [ + 211, + 721, + 330, + 733 + ], + "score": 0.89, + "content": "\\begin{array} { r } { \\pi ^ { * } = \\arg \\operatorname* { m a x } _ { \\pi } \\mathbb { E } _ { \\tau \\sim p _ { \\pi } } \\overline { { [ R ( \\tau ) ] } } } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 331, + 720, + 335, + 734 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 48, + "bbox_fs": [ + 105, + 651, + 507, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 505, + 126 + ], + "lines": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "score": 1.0, + "content": "Temporal Difference Learning TD methods (Fujimoto et al., 2018; Lillicrap et al., 2015) estimate", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 92, + 506, + 108 + ], + "spans": [ + { + "bbox": [ + 106, + 93, + 280, + 106 + ], + "score": 0.9, + "content": "Q ^ { * } ( s , a ) : = \\mathbb { E } _ { \\tau \\sim p _ { \\pi ^ { * } } } [ R ( \\tau ) | s _ { 0 } = s , a _ { 0 } = a ]", + "type": "inline_equation" + }, + { + "bbox": [ + 280, + 92, + 467, + 108 + ], + "score": 1.0, + "content": ", the return achieved under the optimal policy", + "type": "text" + }, + { + "bbox": [ + 467, + 94, + 479, + 104 + ], + "score": 0.86, + "content": "\\pi ^ { * }", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 92, + 506, + 108 + ], + "score": 1.0, + "content": "when", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 105, + 505, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 105, + 169, + 117 + ], + "score": 1.0, + "content": "starting in state", + "type": "text" + }, + { + "bbox": [ + 169, + 107, + 175, + 114 + ], + "score": 0.74, + "content": "s", + "type": "inline_equation" + }, + { + "bbox": [ + 175, + 105, + 245, + 117 + ], + "score": 1.0, + "content": "and taking action", + "type": "text" + }, + { + "bbox": [ + 246, + 108, + 252, + 114 + ], + "score": 0.68, + "content": "a", + "type": "inline_equation" + }, + { + "bbox": [ + 252, + 105, + 340, + 117 + ], + "score": 1.0, + "content": ", with a parameterized", + "type": "text" + }, + { + "bbox": [ + 340, + 105, + 349, + 116 + ], + "score": 0.86, + "content": "Q", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 105, + 505, + 117 + ], + "score": 1.0, + "content": "-function. This requires minimizing the", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 114, + 184, + 127 + ], + "spans": [ + { + "bbox": [ + 105, + 114, + 184, + 127 + ], + "score": 1.0, + "content": "following TD loss:", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 1.5 + }, + { + "type": "interline_equation", + "bbox": [ + 181, + 129, + 429, + 148 + ], + "lines": [ + { + "bbox": [ + 181, + 129, + 429, + 148 + ], + "spans": [ + { + "bbox": [ + 181, + 129, + 429, + 148 + ], + "score": 0.91, + "content": "\\mathcal { L } _ { \\mathrm { T D } } ( \\theta ) : = \\mathbb { E } _ { ( s , a , r , s ^ { \\prime } ) \\in \\mathcal { D } } [ ( r + \\gamma \\operatorname* { m a x } _ { a ^ { \\prime } \\in \\mathcal { A } } Q _ { \\theta } ( s ^ { \\prime } , a ^ { \\prime } ) - Q _ { \\theta } ( s , a ) ) ^ { 2 } ]", + "type": "interline_equation", + "image_path": "2bf532cfc6885cbf314f179956efa8d115c87c8434197ca19e0caa295e22639a.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 181, + 129, + 429, + 148 + ], + "spans": [], + "index": 4 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 151, + 504, + 174 + ], + "lines": [ + { + "bbox": [ + 106, + 150, + 506, + 164 + ], + "spans": [ + { + "bbox": [ + 106, + 150, + 386, + 164 + ], + "score": 1.0, + "content": "Continuous action spaces further require learning a parametric policy", + "type": "text" + }, + { + "bbox": [ + 387, + 151, + 419, + 164 + ], + "score": 0.92, + "content": "\\pi _ { \\phi } ( a | s )", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 150, + 506, + 164 + ], + "score": 1.0, + "content": "that plays the role of", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 162, + 488, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 162, + 488, + 174 + ], + "score": 1.0, + "content": "the maximizing action in equation 1. This results in a policy objective that must be maximized:", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + }, + { + "type": "interline_equation", + "bbox": [ + 238, + 176, + 373, + 191 + ], + "lines": [ + { + "bbox": [ + 238, + 176, + 373, + 191 + ], + "spans": [ + { + "bbox": [ + 238, + 176, + 373, + 191 + ], + "score": 0.93, + "content": "\\mathcal { I } ( \\phi ) : = \\mathbb { E } _ { s \\in \\mathcal { D } , a \\sim \\pi _ { \\phi } ( . | s ) } [ Q ( s , a ) ]", + "type": "interline_equation", + "image_path": "04b40983b847436144e6f160dc15a472f7e55a555678def1dbc3c0712123ce7d.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 238, + 176, + 373, + 191 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 199, + 506, + 232 + ], + "lines": [ + { + "bbox": [ + 105, + 197, + 504, + 212 + ], + "spans": [ + { + "bbox": [ + 105, + 197, + 232, + 212 + ], + "score": 1.0, + "content": "Here, the dataset of transitions", + "type": "text" + }, + { + "bbox": [ + 233, + 199, + 242, + 209 + ], + "score": 0.83, + "content": "\\mathcal { D }", + "type": "inline_equation" + }, + { + "bbox": [ + 243, + 197, + 490, + 212 + ], + "score": 1.0, + "content": "evolves as the agent interacts with the environment and both", + "type": "text" + }, + { + "bbox": [ + 491, + 200, + 504, + 210 + ], + "score": 0.86, + "content": "Q _ { \\theta }", + "type": "inline_equation" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 208, + 507, + 223 + ], + "spans": [ + { + "bbox": [ + 105, + 208, + 123, + 223 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 123, + 211, + 135, + 222 + ], + "score": 0.88, + "content": "\\pi _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 136, + 208, + 507, + 223 + ], + "score": 1.0, + "content": "are trained together. These methods make use of function approximation, off-policy learning,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 221, + 507, + 233 + ], + "spans": [ + { + "bbox": [ + 106, + 221, + 507, + 233 + ], + "score": 1.0, + "content": "and bootstrapping, leading to several instabilities in practice (Sutton, 1988; Van Hasselt et al., 2018).", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 106, + 234, + 505, + 356 + ], + "lines": [ + { + "bbox": [ + 106, + 235, + 505, + 247 + ], + "spans": [ + { + "bbox": [ + 106, + 235, + 505, + 247 + ], + "score": 1.0, + "content": "Offline RL requires finding a return-maximizing policy from a fixed dataset of transitions collected", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 246, + 505, + 258 + ], + "spans": [ + { + "bbox": [ + 106, + 246, + 242, + 258 + ], + "score": 1.0, + "content": "by an unknown behavior policy", + "type": "text" + }, + { + "bbox": [ + 242, + 248, + 249, + 257 + ], + "score": 0.77, + "content": "\\mu", + "type": "inline_equation" + }, + { + "bbox": [ + 250, + 246, + 505, + 258 + ], + "score": 1.0, + "content": "(Levine et al., 2020). Using TD-learning naively causes the", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 104, + 255, + 506, + 270 + ], + "spans": [ + { + "bbox": [ + 104, + 255, + 220, + 270 + ], + "score": 1.0, + "content": "state visitation distribution", + "type": "text" + }, + { + "bbox": [ + 221, + 258, + 249, + 269 + ], + "score": 0.92, + "content": "d ^ { \\pi _ { \\phi } } ( s )", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 255, + 462, + 270 + ], + "score": 1.0, + "content": "to move away from the distribution of the dataset", + "type": "text" + }, + { + "bbox": [ + 462, + 257, + 486, + 269 + ], + "score": 0.92, + "content": "d ^ { \\mu } ( s )", + "type": "inline_equation" + }, + { + "bbox": [ + 487, + 255, + 506, + 270 + ], + "score": 1.0, + "content": ". In", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 268, + 506, + 280 + ], + "spans": [ + { + "bbox": [ + 106, + 268, + 172, + 280 + ], + "score": 1.0, + "content": "turn, the policy", + "type": "text" + }, + { + "bbox": [ + 172, + 269, + 184, + 280 + ], + "score": 0.84, + "content": "\\pi _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 268, + 506, + 280 + ], + "score": 1.0, + "content": "begins to take actions that are substantially different from those already seen", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "score": 1.0, + "content": "in the data. Offline RL algorithms resolve this distribution-shift by imposing a constraint of the", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 289, + 505, + 302 + ], + "spans": [ + { + "bbox": [ + 106, + 289, + 129, + 301 + ], + "score": 1.0, + "content": "form", + "type": "text" + }, + { + "bbox": [ + 129, + 291, + 177, + 302 + ], + "score": 0.92, + "content": "D ( d ^ { \\pi _ { \\phi } } | | d ^ { \\mu } )", + "type": "inline_equation" + }, + { + "bbox": [ + 177, + 289, + 208, + 301 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 208, + 290, + 218, + 299 + ], + "score": 0.83, + "content": "D", + "type": "inline_equation" + }, + { + "bbox": [ + 218, + 289, + 505, + 301 + ], + "score": 1.0, + "content": "is some divergence metric, directly in the TD-learning procedure. The", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 300, + 506, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 300, + 506, + 314 + ], + "score": 1.0, + "content": "constrained optimization problem now demands additional implementation heuristics to achieve any", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 104, + 311, + 506, + 325 + ], + "spans": [ + { + "bbox": [ + 104, + 311, + 506, + 325 + ], + "score": 1.0, + "content": "reasonable performance (Kumar et al., 2021). The Decision Diffuser, in comparison, doesn’t have any", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 104, + 320, + 506, + 336 + ], + "spans": [ + { + "bbox": [ + 104, + 320, + 370, + 336 + ], + "score": 1.0, + "content": "of these disadvantages. It does not require estimating any kind of", + "type": "text" + }, + { + "bbox": [ + 370, + 323, + 379, + 334 + ], + "score": 0.85, + "content": "Q", + "type": "inline_equation" + }, + { + "bbox": [ + 380, + 320, + 506, + 336 + ], + "score": 1.0, + "content": "-function, thereby sidestepping", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 334, + 505, + 345 + ], + "spans": [ + { + "bbox": [ + 106, + 334, + 505, + 345 + ], + "score": 1.0, + "content": "TD methods altogether. It also does not face the risk of distribution-shift as generative models are", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 345, + 290, + 356 + ], + "spans": [ + { + "bbox": [ + 106, + 345, + 290, + 356 + ], + "score": 1.0, + "content": "trained with maximum-likelihood estimation.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 16 + }, + { + "type": "title", + "bbox": [ + 109, + 363, + 287, + 374 + ], + "lines": [ + { + "bbox": [ + 106, + 363, + 289, + 376 + ], + "spans": [ + { + "bbox": [ + 106, + 363, + 289, + 376 + ], + "score": 1.0, + "content": "2.2 DIFFUSION PROBABILISTIC MODELS", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 106, + 378, + 505, + 479 + ], + "lines": [ + { + "bbox": [ + 105, + 377, + 505, + 391 + ], + "spans": [ + { + "bbox": [ + 105, + 377, + 505, + 391 + ], + "score": 1.0, + "content": "Diffusion models (Sohl-Dickstein et al., 2015; Ho et al., 2020) are a specific type of generative model", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 389, + 506, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 232, + 403 + ], + "score": 1.0, + "content": "that learn the data distribution", + "type": "text" + }, + { + "bbox": [ + 232, + 390, + 252, + 402 + ], + "score": 0.91, + "content": "q ( { \\pmb x } )", + "type": "inline_equation" + }, + { + "bbox": [ + 253, + 389, + 315, + 403 + ], + "score": 1.0, + "content": "from a dataset", + "type": "text" + }, + { + "bbox": [ + 315, + 389, + 389, + 402 + ], + "score": 0.93, + "content": "\\mathcal { D } : = \\{ \\pmb { x } ^ { i } \\} _ { 0 \\leq i < M }", + "type": "inline_equation" + }, + { + "bbox": [ + 389, + 389, + 506, + 403 + ], + "score": 1.0, + "content": ". They have been used most", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 104, + 399, + 506, + 413 + ], + "spans": [ + { + "bbox": [ + 104, + 399, + 506, + 413 + ], + "score": 1.0, + "content": "notably for synthesizing high-quality images from text descriptions (Saharia et al., 2022; Nichol et al.,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 410, + 506, + 425 + ], + "spans": [ + { + "bbox": [ + 105, + 410, + 506, + 425 + ], + "score": 1.0, + "content": "2021). Here, the data-generating procedure is modelled with a predefined forward noising process", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 107, + 422, + 505, + 435 + ], + "spans": [ + { + "bbox": [ + 107, + 423, + 300, + 434 + ], + "score": 0.72, + "content": "q ( \\pmb { x } _ { k + 1 } | \\pmb { x } _ { k } ) : = \\mathcal { N } ( \\pmb { x } _ { k + 1 } ; \\sqrt { \\alpha _ { k } } \\pmb { x } _ { k } , ( 1 - \\alpha _ { k } ) \\pmb { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 301, + 422, + 434, + 435 + ], + "score": 1.0, + "content": "and a trainable reverse process", + "type": "text" + }, + { + "bbox": [ + 435, + 423, + 505, + 435 + ], + "score": 0.9, + "content": "p _ { \\theta } ( \\pmb { x } _ { k - 1 } | \\pmb { x } _ { k } ) : =", + "type": "inline_equation" + } + ], + "index": 27 + }, + { + "bbox": [ + 107, + 432, + 506, + 447 + ], + "spans": [ + { + "bbox": [ + 107, + 434, + 205, + 445 + ], + "score": 0.88, + "content": "\\mathcal { N } ( \\pmb { x } _ { k - 1 } | \\mu _ { \\theta } ( \\pmb { x } _ { k } , k ) , \\Sigma _ { k } )", + "type": "inline_equation" + }, + { + "bbox": [ + 205, + 432, + 236, + 447 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 236, + 434, + 272, + 446 + ], + "score": 0.91, + "content": "\\mathcal { N } ( \\boldsymbol { \\mu } , \\boldsymbol { \\Sigma } )", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 432, + 444, + 447 + ], + "score": 1.0, + "content": "denotes a Gaussian distribution with mean", + "type": "text" + }, + { + "bbox": [ + 444, + 436, + 451, + 445 + ], + "score": 0.83, + "content": "\\mu", + "type": "inline_equation" + }, + { + "bbox": [ + 452, + 432, + 506, + 447 + ], + "score": 1.0, + "content": "and variance", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 443, + 507, + 458 + ], + "spans": [ + { + "bbox": [ + 106, + 446, + 115, + 455 + ], + "score": 0.55, + "content": "\\Sigma", + "type": "inline_equation" + }, + { + "bbox": [ + 115, + 443, + 118, + 458 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 118, + 446, + 150, + 456 + ], + "score": 0.85, + "content": "\\alpha _ { k } \\in \\mathbb { R }", + "type": "inline_equation" + }, + { + "bbox": [ + 151, + 443, + 289, + 458 + ], + "score": 1.0, + "content": "determines the variance schedule,", + "type": "text" + }, + { + "bbox": [ + 290, + 446, + 324, + 456 + ], + "score": 0.89, + "content": "\\scriptstyle { \\pmb { x } } _ { 0 } : = { \\pmb { x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 443, + 375, + 458 + ], + "score": 1.0, + "content": "is a sample,", + "type": "text" + }, + { + "bbox": [ + 375, + 446, + 444, + 456 + ], + "score": 0.89, + "content": "{ \\pmb x } _ { 1 } , { \\pmb x } _ { 2 } , . . . , { \\pmb x } _ { K - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 444, + 443, + 507, + 458 + ], + "score": 1.0, + "content": "are the latents,", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 454, + 507, + 469 + ], + "spans": [ + { + "bbox": [ + 105, + 454, + 124, + 469 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 125, + 455, + 189, + 467 + ], + "score": 0.93, + "content": "\\pmb { x } _ { K } \\sim \\mathcal { N } ( \\mathbf { 0 } , \\pmb { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 454, + 277, + 469 + ], + "score": 1.0, + "content": "for carefully chosen", + "type": "text" + }, + { + "bbox": [ + 278, + 457, + 290, + 466 + ], + "score": 0.86, + "content": "\\alpha _ { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 291, + 454, + 365, + 469 + ], + "score": 1.0, + "content": "and long enough", + "type": "text" + }, + { + "bbox": [ + 365, + 456, + 375, + 465 + ], + "score": 0.83, + "content": "K", + "type": "inline_equation" + }, + { + "bbox": [ + 376, + 454, + 507, + 469 + ], + "score": 1.0, + "content": ". Starting with Gaussian noise,", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 466, + 411, + 480 + ], + "spans": [ + { + "bbox": [ + 106, + 466, + 411, + 480 + ], + "score": 1.0, + "content": "samples are then iteratively generated through a series of ”denoising” steps.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 106, + 483, + 504, + 505 + ], + "lines": [ + { + "bbox": [ + 105, + 482, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 303, + 496 + ], + "score": 1.0, + "content": "Although a tractable variational lower-bound on", + "type": "text" + }, + { + "bbox": [ + 303, + 484, + 329, + 495 + ], + "score": 0.87, + "content": "\\log p _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 329, + 482, + 506, + 496 + ], + "score": 1.0, + "content": "can be optimized to train diffusion models,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 494, + 315, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 494, + 315, + 507 + ], + "score": 1.0, + "content": "Ho et al. (2020) propose a simplified surrogate loss:", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32.5 + }, + { + "type": "interline_equation", + "bbox": [ + 191, + 507, + 420, + 523 + ], + "lines": [ + { + "bbox": [ + 191, + 507, + 420, + 523 + ], + "spans": [ + { + "bbox": [ + 191, + 507, + 420, + 523 + ], + "score": 0.88, + "content": "\\mathcal { L } _ { \\mathrm { d e n o i s e } } ( \\theta ) : = \\mathbb { E } _ { k \\sim [ 1 , K ] , \\boldsymbol { x } _ { 0 } \\sim q , \\epsilon \\sim \\mathcal { N } ( \\mathbf { 0 } , I ) } [ | | \\epsilon - \\epsilon _ { \\theta } ( \\boldsymbol { x } _ { k } , k ) | | ^ { 2 } ]", + "type": "interline_equation", + "image_path": "bd78f527a1f1f8d43ca1aecd368174189ff5bd78b1018ae19a032b1c0623c57a.jpg" + } + ] + } + ], + "index": 34, + "virtual_lines": [ + { + "bbox": [ + 191, + 507, + 420, + 523 + ], + "spans": [], + "index": 34 + } + ] + }, + { + "type": "text", + "bbox": [ + 109, + 526, + 504, + 560 + ], + "lines": [ + { + "bbox": [ + 107, + 525, + 505, + 539 + ], + "spans": [ + { + "bbox": [ + 107, + 525, + 189, + 539 + ], + "score": 1.0, + "content": "The predicted noise", + "type": "text" + }, + { + "bbox": [ + 189, + 526, + 228, + 538 + ], + "score": 0.93, + "content": "\\boldsymbol { \\epsilon } _ { \\boldsymbol { \\theta } } ( \\boldsymbol { x } _ { k } , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 525, + 487, + 539 + ], + "score": 1.0, + "content": ", parameterized with a deep neural network, estimates the noise", + "type": "text" + }, + { + "bbox": [ + 488, + 528, + 505, + 537 + ], + "score": 0.79, + "content": "\\epsilon \\sim", + "type": "inline_equation" + } + ], + "index": 35 + }, + { + "bbox": [ + 107, + 535, + 506, + 550 + ], + "spans": [ + { + "bbox": [ + 107, + 537, + 140, + 549 + ], + "score": 0.92, + "content": "\\mathcal { N } ( 0 , I )", + "type": "inline_equation" + }, + { + "bbox": [ + 140, + 535, + 257, + 550 + ], + "score": 1.0, + "content": "added to the dataset sample", + "type": "text" + }, + { + "bbox": [ + 257, + 538, + 269, + 548 + ], + "score": 0.86, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 535, + 342, + 550 + ], + "score": 1.0, + "content": "to produce noisy", + "type": "text" + }, + { + "bbox": [ + 342, + 538, + 354, + 548 + ], + "score": 0.87, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 355, + 535, + 506, + 550 + ], + "score": 1.0, + "content": ". This is equivalent to predicting the", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 547, + 506, + 561 + ], + "spans": [ + { + "bbox": [ + 106, + 547, + 141, + 561 + ], + "score": 1.0, + "content": "mean of", + "type": "text" + }, + { + "bbox": [ + 141, + 548, + 195, + 560 + ], + "score": 0.93, + "content": "p _ { \\theta } ( \\pmb { x } _ { k - 1 } | \\pmb { x } _ { k } )", + "type": "inline_equation" + }, + { + "bbox": [ + 195, + 547, + 219, + 561 + ], + "score": 1.0, + "content": "since", + "type": "text" + }, + { + "bbox": [ + 220, + 548, + 261, + 560 + ], + "score": 0.93, + "content": "\\mu _ { \\boldsymbol { \\theta } } ( \\mathbf { \\boldsymbol { x } } _ { k } , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 547, + 396, + 561 + ], + "score": 1.0, + "content": "can be calculated as a function of", + "type": "text" + }, + { + "bbox": [ + 397, + 548, + 435, + 560 + ], + "score": 0.93, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 547, + 506, + 561 + ], + "score": 1.0, + "content": "(Ho et al., 2020).", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36 + }, + { + "type": "text", + "bbox": [ + 106, + 561, + 506, + 685 + ], + "lines": [ + { + "bbox": [ + 106, + 560, + 505, + 575 + ], + "spans": [ + { + "bbox": [ + 106, + 560, + 360, + 575 + ], + "score": 1.0, + "content": "Guided Diffusion Modelling the conditional data distribution", + "type": "text" + }, + { + "bbox": [ + 360, + 561, + 389, + 574 + ], + "score": 0.94, + "content": "q ( { \\pmb x } | { \\pmb y } )", + "type": "inline_equation" + }, + { + "bbox": [ + 389, + 560, + 505, + 575 + ], + "score": 1.0, + "content": "makes it possible to generate", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 572, + 506, + 585 + ], + "spans": [ + { + "bbox": [ + 106, + 572, + 259, + 585 + ], + "score": 1.0, + "content": "samples with attributes of the label", + "type": "text" + }, + { + "bbox": [ + 260, + 574, + 268, + 585 + ], + "score": 0.77, + "content": "\\textbf { { y } }", + "type": "inline_equation" + }, + { + "bbox": [ + 268, + 572, + 506, + 585 + ], + "score": 1.0, + "content": ". The equivalence between diffusion models and score-", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 583, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 106, + 583, + 273, + 596 + ], + "score": 1.0, + "content": "matching (Song et al., 2021), which shows", + "type": "text" + }, + { + "bbox": [ + 273, + 583, + 384, + 596 + ], + "score": 0.9, + "content": "\\epsilon _ { \\theta } ( x _ { k } , \\bar { k } ) \\propto \\nabla _ { x _ { k } } \\log p ( \\pmb { x } _ { k } )", + "type": "inline_equation" + }, + { + "bbox": [ + 385, + 583, + 506, + 596 + ], + "score": 1.0, + "content": ", leads to two kinds of methods", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 595, + 506, + 607 + ], + "spans": [ + { + "bbox": [ + 106, + 595, + 506, + 607 + ], + "score": 1.0, + "content": "for conditioning: classifier-guided (Nichol & Dhariwal, 2021) and classifier-free (Ho & Salimans,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 604, + 506, + 619 + ], + "spans": [ + { + "bbox": [ + 105, + 604, + 344, + 619 + ], + "score": 1.0, + "content": "2022). The former requires training an additional classifier", + "type": "text" + }, + { + "bbox": [ + 344, + 606, + 384, + 618 + ], + "score": 0.95, + "content": "p _ { \\phi } ( \\pmb { y } | \\pmb { x } _ { k } )", + "type": "inline_equation" + }, + { + "bbox": [ + 384, + 604, + 506, + 619 + ], + "score": 1.0, + "content": "on noisy data so that samples", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 617, + 505, + 630 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 339, + 630 + ], + "score": 1.0, + "content": "may be generated at test-time with the perturbed noise", + "type": "text" + }, + { + "bbox": [ + 339, + 617, + 502, + 630 + ], + "score": 0.9, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } , k ) - \\omega \\sqrt { 1 - \\bar { \\alpha } _ { k } } \\nabla _ { \\pmb { x } _ { k } } \\log p ( \\pmb { y } | \\pmb { x } _ { k } )", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 617, + 505, + 630 + ], + "score": 1.0, + "content": ",", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 629, + 505, + 640 + ], + "spans": [ + { + "bbox": [ + 106, + 629, + 133, + 640 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 134, + 631, + 142, + 639 + ], + "score": 0.77, + "content": "\\omega", + "type": "inline_equation" + }, + { + "bbox": [ + 142, + 629, + 505, + 640 + ], + "score": 1.0, + "content": "is referred to as the guidance scale. The latter does not separately train a classifier but", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 640, + 506, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 640, + 365, + 653 + ], + "score": 1.0, + "content": "modifies the original training setup to learn both a conditional", + "type": "text" + }, + { + "bbox": [ + 365, + 640, + 415, + 652 + ], + "score": 0.93, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } , \\pmb { y } , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 415, + 640, + 506, + 653 + ], + "score": 1.0, + "content": "and an unconditional", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 650, + 505, + 663 + ], + "spans": [ + { + "bbox": [ + 106, + 650, + 145, + 662 + ], + "score": 0.92, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 650, + 505, + 663 + ], + "score": 1.0, + "content": "model for the noise. The unconditional noise is represented, in practice, as the", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 661, + 506, + 674 + ], + "spans": [ + { + "bbox": [ + 105, + 661, + 179, + 674 + ], + "score": 1.0, + "content": "conditional noise", + "type": "text" + }, + { + "bbox": [ + 180, + 662, + 230, + 673 + ], + "score": 0.9, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } , 0 , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 230, + 661, + 324, + 674 + ], + "score": 1.0, + "content": "where a dummy value", + "type": "text" + }, + { + "bbox": [ + 325, + 662, + 334, + 672 + ], + "score": 0.65, + "content": "\\varnothing", + "type": "inline_equation" + }, + { + "bbox": [ + 334, + 661, + 408, + 674 + ], + "score": 1.0, + "content": "takes the place of", + "type": "text" + }, + { + "bbox": [ + 409, + 663, + 416, + 673 + ], + "score": 0.8, + "content": "\\textbf { { y } }", + "type": "inline_equation" + }, + { + "bbox": [ + 416, + 661, + 506, + 674 + ], + "score": 1.0, + "content": ". The perturbed noise", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 106, + 673, + 405, + 685 + ], + "spans": [ + { + "bbox": [ + 106, + 673, + 269, + 685 + ], + "score": 0.9, + "content": "\\epsilon _ { \\theta } ( { \\pmb x } _ { k } , k ) + \\omega ( \\epsilon _ { \\theta } ( { \\pmb x } _ { k } , { \\pmb y } , k ) - \\epsilon _ { \\theta } ( { \\pmb x } _ { k } , k ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 673, + 405, + 685 + ], + "score": 1.0, + "content": "is used to later generate samples.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 43 + }, + { + "type": "title", + "bbox": [ + 108, + 689, + 417, + 702 + ], + "lines": [ + { + "bbox": [ + 104, + 687, + 417, + 704 + ], + "spans": [ + { + "bbox": [ + 104, + 687, + 417, + 704 + ], + "score": 1.0, + "content": "3 GENERATIVE MODELING WITH THE DECISION DIFFUSER", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 49 + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 503, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 708, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 505, + 723 + ], + "score": 1.0, + "content": "It is useful to solve RL from offline data, both without relying on TD-learning and without risking", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 719, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 505, + 733 + ], + "score": 1.0, + "content": "distribution-shift. To this end, we formulate sequential decision-making as the standard problem of", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 50.5 + } + ], + "page_idx": 2, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "score": 1.0, + "content": "3", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 505, + 126 + ], + "lines": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "score": 1.0, + "content": "Temporal Difference Learning TD methods (Fujimoto et al., 2018; Lillicrap et al., 2015) estimate", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 92, + 506, + 108 + ], + "spans": [ + { + "bbox": [ + 106, + 93, + 280, + 106 + ], + "score": 0.9, + "content": "Q ^ { * } ( s , a ) : = \\mathbb { E } _ { \\tau \\sim p _ { \\pi ^ { * } } } [ R ( \\tau ) | s _ { 0 } = s , a _ { 0 } = a ]", + "type": "inline_equation" + }, + { + "bbox": [ + 280, + 92, + 467, + 108 + ], + "score": 1.0, + "content": ", the return achieved under the optimal policy", + "type": "text" + }, + { + "bbox": [ + 467, + 94, + 479, + 104 + ], + "score": 0.86, + "content": "\\pi ^ { * }", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 92, + 506, + 108 + ], + "score": 1.0, + "content": "when", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 105, + 505, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 105, + 169, + 117 + ], + "score": 1.0, + "content": "starting in state", + "type": "text" + }, + { + "bbox": [ + 169, + 107, + 175, + 114 + ], + "score": 0.74, + "content": "s", + "type": "inline_equation" + }, + { + "bbox": [ + 175, + 105, + 245, + 117 + ], + "score": 1.0, + "content": "and taking action", + "type": "text" + }, + { + "bbox": [ + 246, + 108, + 252, + 114 + ], + "score": 0.68, + "content": "a", + "type": "inline_equation" + }, + { + "bbox": [ + 252, + 105, + 340, + 117 + ], + "score": 1.0, + "content": ", with a parameterized", + "type": "text" + }, + { + "bbox": [ + 340, + 105, + 349, + 116 + ], + "score": 0.86, + "content": "Q", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 105, + 505, + 117 + ], + "score": 1.0, + "content": "-function. This requires minimizing the", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 114, + 184, + 127 + ], + "spans": [ + { + "bbox": [ + 105, + 114, + 184, + 127 + ], + "score": 1.0, + "content": "following TD loss:", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 1.5, + "bbox_fs": [ + 105, + 82, + 506, + 127 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 181, + 129, + 429, + 148 + ], + "lines": [ + { + "bbox": [ + 181, + 129, + 429, + 148 + ], + "spans": [ + { + "bbox": [ + 181, + 129, + 429, + 148 + ], + "score": 0.91, + "content": "\\mathcal { L } _ { \\mathrm { T D } } ( \\theta ) : = \\mathbb { E } _ { ( s , a , r , s ^ { \\prime } ) \\in \\mathcal { D } } [ ( r + \\gamma \\operatorname* { m a x } _ { a ^ { \\prime } \\in \\mathcal { A } } Q _ { \\theta } ( s ^ { \\prime } , a ^ { \\prime } ) - Q _ { \\theta } ( s , a ) ) ^ { 2 } ]", + "type": "interline_equation", + "image_path": "2bf532cfc6885cbf314f179956efa8d115c87c8434197ca19e0caa295e22639a.jpg" + } + ] + } + ], + "index": 4, + "virtual_lines": [ + { + "bbox": [ + 181, + 129, + 429, + 148 + ], + "spans": [], + "index": 4 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 151, + 504, + 174 + ], + "lines": [ + { + "bbox": [ + 106, + 150, + 506, + 164 + ], + "spans": [ + { + "bbox": [ + 106, + 150, + 386, + 164 + ], + "score": 1.0, + "content": "Continuous action spaces further require learning a parametric policy", + "type": "text" + }, + { + "bbox": [ + 387, + 151, + 419, + 164 + ], + "score": 0.92, + "content": "\\pi _ { \\phi } ( a | s )", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 150, + 506, + 164 + ], + "score": 1.0, + "content": "that plays the role of", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 162, + 488, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 162, + 488, + 174 + ], + "score": 1.0, + "content": "the maximizing action in equation 1. This results in a policy objective that must be maximized:", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5, + "bbox_fs": [ + 105, + 150, + 506, + 174 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 238, + 176, + 373, + 191 + ], + "lines": [ + { + "bbox": [ + 238, + 176, + 373, + 191 + ], + "spans": [ + { + "bbox": [ + 238, + 176, + 373, + 191 + ], + "score": 0.93, + "content": "\\mathcal { I } ( \\phi ) : = \\mathbb { E } _ { s \\in \\mathcal { D } , a \\sim \\pi _ { \\phi } ( . | s ) } [ Q ( s , a ) ]", + "type": "interline_equation", + "image_path": "04b40983b847436144e6f160dc15a472f7e55a555678def1dbc3c0712123ce7d.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 238, + 176, + 373, + 191 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 199, + 506, + 232 + ], + "lines": [ + { + "bbox": [ + 105, + 197, + 504, + 212 + ], + "spans": [ + { + "bbox": [ + 105, + 197, + 232, + 212 + ], + "score": 1.0, + "content": "Here, the dataset of transitions", + "type": "text" + }, + { + "bbox": [ + 233, + 199, + 242, + 209 + ], + "score": 0.83, + "content": "\\mathcal { D }", + "type": "inline_equation" + }, + { + "bbox": [ + 243, + 197, + 490, + 212 + ], + "score": 1.0, + "content": "evolves as the agent interacts with the environment and both", + "type": "text" + }, + { + "bbox": [ + 491, + 200, + 504, + 210 + ], + "score": 0.86, + "content": "Q _ { \\theta }", + "type": "inline_equation" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 208, + 507, + 223 + ], + "spans": [ + { + "bbox": [ + 105, + 208, + 123, + 223 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 123, + 211, + 135, + 222 + ], + "score": 0.88, + "content": "\\pi _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 136, + 208, + 507, + 223 + ], + "score": 1.0, + "content": "are trained together. These methods make use of function approximation, off-policy learning,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 221, + 507, + 233 + ], + "spans": [ + { + "bbox": [ + 106, + 221, + 507, + 233 + ], + "score": 1.0, + "content": "and bootstrapping, leading to several instabilities in practice (Sutton, 1988; Van Hasselt et al., 2018).", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9, + "bbox_fs": [ + 105, + 197, + 507, + 233 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 234, + 505, + 356 + ], + "lines": [ + { + "bbox": [ + 106, + 235, + 505, + 247 + ], + "spans": [ + { + "bbox": [ + 106, + 235, + 505, + 247 + ], + "score": 1.0, + "content": "Offline RL requires finding a return-maximizing policy from a fixed dataset of transitions collected", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 246, + 505, + 258 + ], + "spans": [ + { + "bbox": [ + 106, + 246, + 242, + 258 + ], + "score": 1.0, + "content": "by an unknown behavior policy", + "type": "text" + }, + { + "bbox": [ + 242, + 248, + 249, + 257 + ], + "score": 0.77, + "content": "\\mu", + "type": "inline_equation" + }, + { + "bbox": [ + 250, + 246, + 505, + 258 + ], + "score": 1.0, + "content": "(Levine et al., 2020). Using TD-learning naively causes the", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 104, + 255, + 506, + 270 + ], + "spans": [ + { + "bbox": [ + 104, + 255, + 220, + 270 + ], + "score": 1.0, + "content": "state visitation distribution", + "type": "text" + }, + { + "bbox": [ + 221, + 258, + 249, + 269 + ], + "score": 0.92, + "content": "d ^ { \\pi _ { \\phi } } ( s )", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 255, + 462, + 270 + ], + "score": 1.0, + "content": "to move away from the distribution of the dataset", + "type": "text" + }, + { + "bbox": [ + 462, + 257, + 486, + 269 + ], + "score": 0.92, + "content": "d ^ { \\mu } ( s )", + "type": "inline_equation" + }, + { + "bbox": [ + 487, + 255, + 506, + 270 + ], + "score": 1.0, + "content": ". In", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 268, + 506, + 280 + ], + "spans": [ + { + "bbox": [ + 106, + 268, + 172, + 280 + ], + "score": 1.0, + "content": "turn, the policy", + "type": "text" + }, + { + "bbox": [ + 172, + 269, + 184, + 280 + ], + "score": 0.84, + "content": "\\pi _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 268, + 506, + 280 + ], + "score": 1.0, + "content": "begins to take actions that are substantially different from those already seen", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 506, + 291 + ], + "score": 1.0, + "content": "in the data. Offline RL algorithms resolve this distribution-shift by imposing a constraint of the", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 289, + 505, + 302 + ], + "spans": [ + { + "bbox": [ + 106, + 289, + 129, + 301 + ], + "score": 1.0, + "content": "form", + "type": "text" + }, + { + "bbox": [ + 129, + 291, + 177, + 302 + ], + "score": 0.92, + "content": "D ( d ^ { \\pi _ { \\phi } } | | d ^ { \\mu } )", + "type": "inline_equation" + }, + { + "bbox": [ + 177, + 289, + 208, + 301 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 208, + 290, + 218, + 299 + ], + "score": 0.83, + "content": "D", + "type": "inline_equation" + }, + { + "bbox": [ + 218, + 289, + 505, + 301 + ], + "score": 1.0, + "content": "is some divergence metric, directly in the TD-learning procedure. The", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 300, + 506, + 314 + ], + "spans": [ + { + "bbox": [ + 105, + 300, + 506, + 314 + ], + "score": 1.0, + "content": "constrained optimization problem now demands additional implementation heuristics to achieve any", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 104, + 311, + 506, + 325 + ], + "spans": [ + { + "bbox": [ + 104, + 311, + 506, + 325 + ], + "score": 1.0, + "content": "reasonable performance (Kumar et al., 2021). The Decision Diffuser, in comparison, doesn’t have any", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 104, + 320, + 506, + 336 + ], + "spans": [ + { + "bbox": [ + 104, + 320, + 370, + 336 + ], + "score": 1.0, + "content": "of these disadvantages. It does not require estimating any kind of", + "type": "text" + }, + { + "bbox": [ + 370, + 323, + 379, + 334 + ], + "score": 0.85, + "content": "Q", + "type": "inline_equation" + }, + { + "bbox": [ + 380, + 320, + 506, + 336 + ], + "score": 1.0, + "content": "-function, thereby sidestepping", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 334, + 505, + 345 + ], + "spans": [ + { + "bbox": [ + 106, + 334, + 505, + 345 + ], + "score": 1.0, + "content": "TD methods altogether. It also does not face the risk of distribution-shift as generative models are", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 345, + 290, + 356 + ], + "spans": [ + { + "bbox": [ + 106, + 345, + 290, + 356 + ], + "score": 1.0, + "content": "trained with maximum-likelihood estimation.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 16, + "bbox_fs": [ + 104, + 235, + 506, + 356 + ] + }, + { + "type": "title", + "bbox": [ + 109, + 363, + 287, + 374 + ], + "lines": [ + { + "bbox": [ + 106, + 363, + 289, + 376 + ], + "spans": [ + { + "bbox": [ + 106, + 363, + 289, + 376 + ], + "score": 1.0, + "content": "2.2 DIFFUSION PROBABILISTIC MODELS", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 106, + 378, + 505, + 479 + ], + "lines": [ + { + "bbox": [ + 105, + 377, + 505, + 391 + ], + "spans": [ + { + "bbox": [ + 105, + 377, + 505, + 391 + ], + "score": 1.0, + "content": "Diffusion models (Sohl-Dickstein et al., 2015; Ho et al., 2020) are a specific type of generative model", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 389, + 506, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 232, + 403 + ], + "score": 1.0, + "content": "that learn the data distribution", + "type": "text" + }, + { + "bbox": [ + 232, + 390, + 252, + 402 + ], + "score": 0.91, + "content": "q ( { \\pmb x } )", + "type": "inline_equation" + }, + { + "bbox": [ + 253, + 389, + 315, + 403 + ], + "score": 1.0, + "content": "from a dataset", + "type": "text" + }, + { + "bbox": [ + 315, + 389, + 389, + 402 + ], + "score": 0.93, + "content": "\\mathcal { D } : = \\{ \\pmb { x } ^ { i } \\} _ { 0 \\leq i < M }", + "type": "inline_equation" + }, + { + "bbox": [ + 389, + 389, + 506, + 403 + ], + "score": 1.0, + "content": ". They have been used most", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 104, + 399, + 506, + 413 + ], + "spans": [ + { + "bbox": [ + 104, + 399, + 506, + 413 + ], + "score": 1.0, + "content": "notably for synthesizing high-quality images from text descriptions (Saharia et al., 2022; Nichol et al.,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 410, + 506, + 425 + ], + "spans": [ + { + "bbox": [ + 105, + 410, + 506, + 425 + ], + "score": 1.0, + "content": "2021). Here, the data-generating procedure is modelled with a predefined forward noising process", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 107, + 422, + 505, + 435 + ], + "spans": [ + { + "bbox": [ + 107, + 423, + 300, + 434 + ], + "score": 0.72, + "content": "q ( \\pmb { x } _ { k + 1 } | \\pmb { x } _ { k } ) : = \\mathcal { N } ( \\pmb { x } _ { k + 1 } ; \\sqrt { \\alpha _ { k } } \\pmb { x } _ { k } , ( 1 - \\alpha _ { k } ) \\pmb { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 301, + 422, + 434, + 435 + ], + "score": 1.0, + "content": "and a trainable reverse process", + "type": "text" + }, + { + "bbox": [ + 435, + 423, + 505, + 435 + ], + "score": 0.9, + "content": "p _ { \\theta } ( \\pmb { x } _ { k - 1 } | \\pmb { x } _ { k } ) : =", + "type": "inline_equation" + } + ], + "index": 27 + }, + { + "bbox": [ + 107, + 432, + 506, + 447 + ], + "spans": [ + { + "bbox": [ + 107, + 434, + 205, + 445 + ], + "score": 0.88, + "content": "\\mathcal { N } ( \\pmb { x } _ { k - 1 } | \\mu _ { \\theta } ( \\pmb { x } _ { k } , k ) , \\Sigma _ { k } )", + "type": "inline_equation" + }, + { + "bbox": [ + 205, + 432, + 236, + 447 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 236, + 434, + 272, + 446 + ], + "score": 0.91, + "content": "\\mathcal { N } ( \\boldsymbol { \\mu } , \\boldsymbol { \\Sigma } )", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 432, + 444, + 447 + ], + "score": 1.0, + "content": "denotes a Gaussian distribution with mean", + "type": "text" + }, + { + "bbox": [ + 444, + 436, + 451, + 445 + ], + "score": 0.83, + "content": "\\mu", + "type": "inline_equation" + }, + { + "bbox": [ + 452, + 432, + 506, + 447 + ], + "score": 1.0, + "content": "and variance", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 443, + 507, + 458 + ], + "spans": [ + { + "bbox": [ + 106, + 446, + 115, + 455 + ], + "score": 0.55, + "content": "\\Sigma", + "type": "inline_equation" + }, + { + "bbox": [ + 115, + 443, + 118, + 458 + ], + "score": 1.0, + "content": ",", + "type": "text" + }, + { + "bbox": [ + 118, + 446, + 150, + 456 + ], + "score": 0.85, + "content": "\\alpha _ { k } \\in \\mathbb { R }", + "type": "inline_equation" + }, + { + "bbox": [ + 151, + 443, + 289, + 458 + ], + "score": 1.0, + "content": "determines the variance schedule,", + "type": "text" + }, + { + "bbox": [ + 290, + 446, + 324, + 456 + ], + "score": 0.89, + "content": "\\scriptstyle { \\pmb { x } } _ { 0 } : = { \\pmb { x } }", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 443, + 375, + 458 + ], + "score": 1.0, + "content": "is a sample,", + "type": "text" + }, + { + "bbox": [ + 375, + 446, + 444, + 456 + ], + "score": 0.89, + "content": "{ \\pmb x } _ { 1 } , { \\pmb x } _ { 2 } , . . . , { \\pmb x } _ { K - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 444, + 443, + 507, + 458 + ], + "score": 1.0, + "content": "are the latents,", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 454, + 507, + 469 + ], + "spans": [ + { + "bbox": [ + 105, + 454, + 124, + 469 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 125, + 455, + 189, + 467 + ], + "score": 0.93, + "content": "\\pmb { x } _ { K } \\sim \\mathcal { N } ( \\mathbf { 0 } , \\pmb { I } )", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 454, + 277, + 469 + ], + "score": 1.0, + "content": "for carefully chosen", + "type": "text" + }, + { + "bbox": [ + 278, + 457, + 290, + 466 + ], + "score": 0.86, + "content": "\\alpha _ { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 291, + 454, + 365, + 469 + ], + "score": 1.0, + "content": "and long enough", + "type": "text" + }, + { + "bbox": [ + 365, + 456, + 375, + 465 + ], + "score": 0.83, + "content": "K", + "type": "inline_equation" + }, + { + "bbox": [ + 376, + 454, + 507, + 469 + ], + "score": 1.0, + "content": ". Starting with Gaussian noise,", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 466, + 411, + 480 + ], + "spans": [ + { + "bbox": [ + 106, + 466, + 411, + 480 + ], + "score": 1.0, + "content": "samples are then iteratively generated through a series of ”denoising” steps.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 27, + "bbox_fs": [ + 104, + 377, + 507, + 480 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 483, + 504, + 505 + ], + "lines": [ + { + "bbox": [ + 105, + 482, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 303, + 496 + ], + "score": 1.0, + "content": "Although a tractable variational lower-bound on", + "type": "text" + }, + { + "bbox": [ + 303, + 484, + 329, + 495 + ], + "score": 0.87, + "content": "\\log p _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 329, + 482, + 506, + 496 + ], + "score": 1.0, + "content": "can be optimized to train diffusion models,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 494, + 315, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 494, + 315, + 507 + ], + "score": 1.0, + "content": "Ho et al. (2020) propose a simplified surrogate loss:", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32.5, + "bbox_fs": [ + 105, + 482, + 506, + 507 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 191, + 507, + 420, + 523 + ], + "lines": [ + { + "bbox": [ + 191, + 507, + 420, + 523 + ], + "spans": [ + { + "bbox": [ + 191, + 507, + 420, + 523 + ], + "score": 0.88, + "content": "\\mathcal { L } _ { \\mathrm { d e n o i s e } } ( \\theta ) : = \\mathbb { E } _ { k \\sim [ 1 , K ] , \\boldsymbol { x } _ { 0 } \\sim q , \\epsilon \\sim \\mathcal { N } ( \\mathbf { 0 } , I ) } [ | | \\epsilon - \\epsilon _ { \\theta } ( \\boldsymbol { x } _ { k } , k ) | | ^ { 2 } ]", + "type": "interline_equation", + "image_path": "bd78f527a1f1f8d43ca1aecd368174189ff5bd78b1018ae19a032b1c0623c57a.jpg" + } + ] + } + ], + "index": 34, + "virtual_lines": [ + { + "bbox": [ + 191, + 507, + 420, + 523 + ], + "spans": [], + "index": 34 + } + ] + }, + { + "type": "text", + "bbox": [ + 109, + 526, + 504, + 560 + ], + "lines": [ + { + "bbox": [ + 107, + 525, + 505, + 539 + ], + "spans": [ + { + "bbox": [ + 107, + 525, + 189, + 539 + ], + "score": 1.0, + "content": "The predicted noise", + "type": "text" + }, + { + "bbox": [ + 189, + 526, + 228, + 538 + ], + "score": 0.93, + "content": "\\boldsymbol { \\epsilon } _ { \\boldsymbol { \\theta } } ( \\boldsymbol { x } _ { k } , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 525, + 487, + 539 + ], + "score": 1.0, + "content": ", parameterized with a deep neural network, estimates the noise", + "type": "text" + }, + { + "bbox": [ + 488, + 528, + 505, + 537 + ], + "score": 0.79, + "content": "\\epsilon \\sim", + "type": "inline_equation" + } + ], + "index": 35 + }, + { + "bbox": [ + 107, + 535, + 506, + 550 + ], + "spans": [ + { + "bbox": [ + 107, + 537, + 140, + 549 + ], + "score": 0.92, + "content": "\\mathcal { N } ( 0 , I )", + "type": "inline_equation" + }, + { + "bbox": [ + 140, + 535, + 257, + 550 + ], + "score": 1.0, + "content": "added to the dataset sample", + "type": "text" + }, + { + "bbox": [ + 257, + 538, + 269, + 548 + ], + "score": 0.86, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { 0 }", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 535, + 342, + 550 + ], + "score": 1.0, + "content": "to produce noisy", + "type": "text" + }, + { + "bbox": [ + 342, + 538, + 354, + 548 + ], + "score": 0.87, + "content": "\\scriptstyle { \\mathbf { { \\mathit { x } } } } _ { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 355, + 535, + 506, + 550 + ], + "score": 1.0, + "content": ". This is equivalent to predicting the", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 547, + 506, + 561 + ], + "spans": [ + { + "bbox": [ + 106, + 547, + 141, + 561 + ], + "score": 1.0, + "content": "mean of", + "type": "text" + }, + { + "bbox": [ + 141, + 548, + 195, + 560 + ], + "score": 0.93, + "content": "p _ { \\theta } ( \\pmb { x } _ { k - 1 } | \\pmb { x } _ { k } )", + "type": "inline_equation" + }, + { + "bbox": [ + 195, + 547, + 219, + 561 + ], + "score": 1.0, + "content": "since", + "type": "text" + }, + { + "bbox": [ + 220, + 548, + 261, + 560 + ], + "score": 0.93, + "content": "\\mu _ { \\boldsymbol { \\theta } } ( \\mathbf { \\boldsymbol { x } } _ { k } , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 547, + 396, + 561 + ], + "score": 1.0, + "content": "can be calculated as a function of", + "type": "text" + }, + { + "bbox": [ + 397, + 548, + 435, + 560 + ], + "score": 0.93, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 547, + 506, + 561 + ], + "score": 1.0, + "content": "(Ho et al., 2020).", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36, + "bbox_fs": [ + 106, + 525, + 506, + 561 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 561, + 506, + 685 + ], + "lines": [ + { + "bbox": [ + 106, + 560, + 505, + 575 + ], + "spans": [ + { + "bbox": [ + 106, + 560, + 360, + 575 + ], + "score": 1.0, + "content": "Guided Diffusion Modelling the conditional data distribution", + "type": "text" + }, + { + "bbox": [ + 360, + 561, + 389, + 574 + ], + "score": 0.94, + "content": "q ( { \\pmb x } | { \\pmb y } )", + "type": "inline_equation" + }, + { + "bbox": [ + 389, + 560, + 505, + 575 + ], + "score": 1.0, + "content": "makes it possible to generate", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 572, + 506, + 585 + ], + "spans": [ + { + "bbox": [ + 106, + 572, + 259, + 585 + ], + "score": 1.0, + "content": "samples with attributes of the label", + "type": "text" + }, + { + "bbox": [ + 260, + 574, + 268, + 585 + ], + "score": 0.77, + "content": "\\textbf { { y } }", + "type": "inline_equation" + }, + { + "bbox": [ + 268, + 572, + 506, + 585 + ], + "score": 1.0, + "content": ". The equivalence between diffusion models and score-", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 583, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 106, + 583, + 273, + 596 + ], + "score": 1.0, + "content": "matching (Song et al., 2021), which shows", + "type": "text" + }, + { + "bbox": [ + 273, + 583, + 384, + 596 + ], + "score": 0.9, + "content": "\\epsilon _ { \\theta } ( x _ { k } , \\bar { k } ) \\propto \\nabla _ { x _ { k } } \\log p ( \\pmb { x } _ { k } )", + "type": "inline_equation" + }, + { + "bbox": [ + 385, + 583, + 506, + 596 + ], + "score": 1.0, + "content": ", leads to two kinds of methods", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 595, + 506, + 607 + ], + "spans": [ + { + "bbox": [ + 106, + 595, + 506, + 607 + ], + "score": 1.0, + "content": "for conditioning: classifier-guided (Nichol & Dhariwal, 2021) and classifier-free (Ho & Salimans,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 604, + 506, + 619 + ], + "spans": [ + { + "bbox": [ + 105, + 604, + 344, + 619 + ], + "score": 1.0, + "content": "2022). The former requires training an additional classifier", + "type": "text" + }, + { + "bbox": [ + 344, + 606, + 384, + 618 + ], + "score": 0.95, + "content": "p _ { \\phi } ( \\pmb { y } | \\pmb { x } _ { k } )", + "type": "inline_equation" + }, + { + "bbox": [ + 384, + 604, + 506, + 619 + ], + "score": 1.0, + "content": "on noisy data so that samples", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 617, + 505, + 630 + ], + "spans": [ + { + "bbox": [ + 105, + 617, + 339, + 630 + ], + "score": 1.0, + "content": "may be generated at test-time with the perturbed noise", + "type": "text" + }, + { + "bbox": [ + 339, + 617, + 502, + 630 + ], + "score": 0.9, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } , k ) - \\omega \\sqrt { 1 - \\bar { \\alpha } _ { k } } \\nabla _ { \\pmb { x } _ { k } } \\log p ( \\pmb { y } | \\pmb { x } _ { k } )", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 617, + 505, + 630 + ], + "score": 1.0, + "content": ",", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 629, + 505, + 640 + ], + "spans": [ + { + "bbox": [ + 106, + 629, + 133, + 640 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 134, + 631, + 142, + 639 + ], + "score": 0.77, + "content": "\\omega", + "type": "inline_equation" + }, + { + "bbox": [ + 142, + 629, + 505, + 640 + ], + "score": 1.0, + "content": "is referred to as the guidance scale. The latter does not separately train a classifier but", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 640, + 506, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 640, + 365, + 653 + ], + "score": 1.0, + "content": "modifies the original training setup to learn both a conditional", + "type": "text" + }, + { + "bbox": [ + 365, + 640, + 415, + 652 + ], + "score": 0.93, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } , \\pmb { y } , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 415, + 640, + 506, + 653 + ], + "score": 1.0, + "content": "and an unconditional", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 650, + 505, + 663 + ], + "spans": [ + { + "bbox": [ + 106, + 650, + 145, + 662 + ], + "score": 0.92, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 146, + 650, + 505, + 663 + ], + "score": 1.0, + "content": "model for the noise. The unconditional noise is represented, in practice, as the", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 661, + 506, + 674 + ], + "spans": [ + { + "bbox": [ + 105, + 661, + 179, + 674 + ], + "score": 1.0, + "content": "conditional noise", + "type": "text" + }, + { + "bbox": [ + 180, + 662, + 230, + 673 + ], + "score": 0.9, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } , 0 , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 230, + 661, + 324, + 674 + ], + "score": 1.0, + "content": "where a dummy value", + "type": "text" + }, + { + "bbox": [ + 325, + 662, + 334, + 672 + ], + "score": 0.65, + "content": "\\varnothing", + "type": "inline_equation" + }, + { + "bbox": [ + 334, + 661, + 408, + 674 + ], + "score": 1.0, + "content": "takes the place of", + "type": "text" + }, + { + "bbox": [ + 409, + 663, + 416, + 673 + ], + "score": 0.8, + "content": "\\textbf { { y } }", + "type": "inline_equation" + }, + { + "bbox": [ + 416, + 661, + 506, + 674 + ], + "score": 1.0, + "content": ". The perturbed noise", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 106, + 673, + 405, + 685 + ], + "spans": [ + { + "bbox": [ + 106, + 673, + 269, + 685 + ], + "score": 0.9, + "content": "\\epsilon _ { \\theta } ( { \\pmb x } _ { k } , k ) + \\omega ( \\epsilon _ { \\theta } ( { \\pmb x } _ { k } , { \\pmb y } , k ) - \\epsilon _ { \\theta } ( { \\pmb x } _ { k } , k ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 270, + 673, + 405, + 685 + ], + "score": 1.0, + "content": "is used to later generate samples.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 43, + "bbox_fs": [ + 105, + 560, + 506, + 685 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 689, + 417, + 702 + ], + "lines": [ + { + "bbox": [ + 104, + 687, + 417, + 704 + ], + "spans": [ + { + "bbox": [ + 104, + 687, + 417, + 704 + ], + "score": 1.0, + "content": "3 GENERATIVE MODELING WITH THE DECISION DIFFUSER", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 49 + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 503, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 708, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 505, + 723 + ], + "score": 1.0, + "content": "It is useful to solve RL from offline data, both without relying on TD-learning and without risking", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 719, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 505, + 733 + ], + "score": 1.0, + "content": "distribution-shift. To this end, we formulate sequential decision-making as the standard problem of", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 50.5, + "bbox_fs": [ + 105, + 708, + 505, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 166, + 79, + 444, + 194 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 166, + 79, + 444, + 194 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 166, + 79, + 444, + 194 + ], + "spans": [ + { + "bbox": [ + 166, + 79, + 444, + 194 + ], + "score": 0.972, + "type": "image", + "image_path": "5193fd80ab094ae9c4154b23239d26a5d2b68886d0126e16c31679f9afe8be80.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 166, + 79, + 444, + 117.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 166, + 117.33333333333334, + 444, + 155.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 166, + 155.66666666666669, + 444, + 194.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 203, + 505, + 234 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 203, + 505, + 215 + ], + "spans": [ + { + "bbox": [ + 105, + 203, + 361, + 215 + ], + "score": 1.0, + "content": "Figure 3: Planning with Decision Diffuser. Given the current state", + "type": "text" + }, + { + "bbox": [ + 361, + 205, + 370, + 213 + ], + "score": 0.84, + "content": "s _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 370, + 203, + 505, + 215 + ], + "score": 1.0, + "content": "and conditioning, Decision Diffuser", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 213, + 506, + 225 + ], + "spans": [ + { + "bbox": [ + 105, + 213, + 506, + 225 + ], + "score": 1.0, + "content": "uses classifier-free guidance with low-temperature sampling to generate a sequence of future states. It then uses", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 222, + 462, + 236 + ], + "spans": [ + { + "bbox": [ + 105, + 222, + 288, + 236 + ], + "score": 1.0, + "content": "inverse dynamics to extract and execute the action", + "type": "text" + }, + { + "bbox": [ + 289, + 225, + 298, + 233 + ], + "score": 0.84, + "content": "a _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 298, + 222, + 440, + 236 + ], + "score": 1.0, + "content": "that leads to the immediate future state", + "type": "text" + }, + { + "bbox": [ + 440, + 225, + 458, + 234 + ], + "score": 0.88, + "content": "s _ { t + 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 458, + 222, + 462, + 236 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 107, + 243, + 240, + 254 + ], + "lines": [ + { + "bbox": [ + 105, + 240, + 240, + 258 + ], + "spans": [ + { + "bbox": [ + 105, + 240, + 240, + 258 + ], + "score": 1.0, + "content": "conditional generative modeling:", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6 + }, + { + "type": "interline_equation", + "bbox": [ + 239, + 262, + 372, + 280 + ], + "lines": [ + { + "bbox": [ + 239, + 262, + 372, + 280 + ], + "spans": [ + { + "bbox": [ + 239, + 262, + 372, + 280 + ], + "score": 0.91, + "content": "\\operatorname* { m a x } _ { \\theta } \\mathbb { E } _ { \\tau \\sim \\mathcal { D } } [ \\log p _ { \\theta } ( \\pmb { x } _ { 0 } ( \\tau ) | \\pmb { y } ( \\tau ) ) ]", + "type": "interline_equation", + "image_path": "5839cb65e52e459da7b8716946f055e40e525790aaf7cc5eb376c9f196100d62.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 239, + 262, + 372, + 280 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 288, + 505, + 332 + ], + "lines": [ + { + "bbox": [ + 106, + 288, + 506, + 300 + ], + "spans": [ + { + "bbox": [ + 106, + 288, + 344, + 300 + ], + "score": 1.0, + "content": "Our goal is to estimate the conditional data distribution with", + "type": "text" + }, + { + "bbox": [ + 344, + 290, + 355, + 299 + ], + "score": 0.83, + "content": "p _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 356, + 288, + 506, + 300 + ], + "score": 1.0, + "content": "so we can later generate portions of a", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 299, + 505, + 312 + ], + "spans": [ + { + "bbox": [ + 105, + 299, + 147, + 312 + ], + "score": 1.0, + "content": "trajectory", + "type": "text" + }, + { + "bbox": [ + 148, + 299, + 173, + 311 + ], + "score": 0.92, + "content": "\\pmb { x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 173, + 299, + 246, + 312 + ], + "score": 1.0, + "content": "from information", + "type": "text" + }, + { + "bbox": [ + 247, + 299, + 267, + 311 + ], + "score": 0.92, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 267, + 299, + 357, + 312 + ], + "score": 1.0, + "content": "about it. Examples of", + "type": "text" + }, + { + "bbox": [ + 358, + 301, + 365, + 311 + ], + "score": 0.83, + "content": "\\textbf { { y } }", + "type": "inline_equation" + }, + { + "bbox": [ + 366, + 299, + 505, + 312 + ], + "score": 1.0, + "content": "could include the return under the", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 310, + 505, + 323 + ], + "spans": [ + { + "bbox": [ + 106, + 310, + 505, + 323 + ], + "score": 1.0, + "content": "trajectory, the constraints satisfied by the trajectory, or the skill demonstrated in the trajectory. We", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 321, + 420, + 334 + ], + "spans": [ + { + "bbox": [ + 105, + 321, + 420, + 334 + ], + "score": 1.0, + "content": "construct our generative model according to the conditional diffusion process:", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 9.5 + }, + { + "type": "interline_equation", + "bbox": [ + 207, + 340, + 404, + 354 + ], + "lines": [ + { + "bbox": [ + 207, + 340, + 404, + 354 + ], + "spans": [ + { + "bbox": [ + 207, + 340, + 404, + 354 + ], + "score": 0.87, + "content": "\\begin{array} { r l } { q ( \\pmb { x } _ { k + 1 } ( \\tau ) | \\pmb { x } _ { k } ( \\tau ) ) , } & { { } p _ { \\theta } ( \\pmb { x } _ { k - 1 } ( \\tau ) | \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ( \\tau ) ) } \\end{array}", + "type": "interline_equation", + "image_path": "4a2aedb6b54591ec48f654546dd7beae0401fc8c1a4821d171893e3898154bc3.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 207, + 340, + 404, + 354 + ], + "spans": [], + "index": 12 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 361, + 505, + 428 + ], + "lines": [ + { + "bbox": [ + 105, + 361, + 506, + 374 + ], + "spans": [ + { + "bbox": [ + 105, + 361, + 147, + 374 + ], + "score": 1.0, + "content": "As usual,", + "type": "text" + }, + { + "bbox": [ + 147, + 364, + 153, + 374 + ], + "score": 0.8, + "content": "q", + "type": "inline_equation" + }, + { + "bbox": [ + 154, + 361, + 340, + 374 + ], + "score": 1.0, + "content": "represents the forward noising process while", + "type": "text" + }, + { + "bbox": [ + 340, + 364, + 351, + 373 + ], + "score": 0.85, + "content": "p _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 351, + 361, + 506, + 374 + ], + "score": 1.0, + "content": "the reverse denoising process. In the", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 371, + 505, + 386 + ], + "spans": [ + { + "bbox": [ + 105, + 371, + 505, + 386 + ], + "score": 1.0, + "content": "following, we discuss how we may use diffusion for decision making. First, we discuss the modeling", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 383, + 505, + 396 + ], + "spans": [ + { + "bbox": [ + 106, + 383, + 505, + 396 + ], + "score": 1.0, + "content": "choices for diffusion in Section 3.1. Next, we discuss how we may utilize classifier-free guidance to", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 394, + 506, + 406 + ], + "spans": [ + { + "bbox": [ + 105, + 394, + 506, + 406 + ], + "score": 1.0, + "content": "capture the best aspects of trajectories in Section 3.2. We then discuss the different behaviors that", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 405, + 505, + 418 + ], + "spans": [ + { + "bbox": [ + 105, + 405, + 505, + 418 + ], + "score": 1.0, + "content": "may be implemented with conditional diffusion models in Section 3.3. Finally, we discuss practical", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 416, + 294, + 428 + ], + "spans": [ + { + "bbox": [ + 105, + 416, + 294, + 428 + ], + "score": 1.0, + "content": "training details of our approach in Section 3.4.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 15.5 + }, + { + "type": "title", + "bbox": [ + 107, + 434, + 239, + 445 + ], + "lines": [ + { + "bbox": [ + 105, + 433, + 240, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 433, + 240, + 447 + ], + "score": 1.0, + "content": "3.1 DIFFUSING OVER STATES", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 106, + 447, + 506, + 536 + ], + "lines": [ + { + "bbox": [ + 105, + 448, + 505, + 461 + ], + "spans": [ + { + "bbox": [ + 105, + 448, + 505, + 461 + ], + "score": 1.0, + "content": "In images, the diffusion process is applied across all pixel values in an image. Na¨ıvely, it would", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 458, + 506, + 472 + ], + "spans": [ + { + "bbox": [ + 105, + 458, + 506, + 472 + ], + "score": 1.0, + "content": "therefore be natural to apply a similar process to model the state and actions of a trajectory. However,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 470, + 506, + 483 + ], + "spans": [ + { + "bbox": [ + 105, + 470, + 506, + 483 + ], + "score": 1.0, + "content": "in the reinforcement learning setting, directly modeling actions using a diffusion process has several", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 482, + 506, + 493 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 506, + 493 + ], + "score": 1.0, + "content": "practical issues. First, while states are typically continuous in nature in RL, actions are more varied,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 492, + 506, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 506, + 504 + ], + "score": 1.0, + "content": "and are often discrete in nature. Furthermore, sequences over actions, which are often represented as", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 104, + 503, + 506, + 515 + ], + "spans": [ + { + "bbox": [ + 104, + 503, + 506, + 515 + ], + "score": 1.0, + "content": "joint torques, tend to be more high-frequency and less smooth, making them much harder to predict", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 514, + 505, + 526 + ], + "spans": [ + { + "bbox": [ + 105, + 514, + 505, + 526 + ], + "score": 1.0, + "content": "and model (Tedrake, 2022). Due to these practical issues, we choose to diffuse only over states, as", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 524, + 168, + 537 + ], + "spans": [ + { + "bbox": [ + 105, + 524, + 168, + 537 + ], + "score": 1.0, + "content": "defined below:", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 23.5 + }, + { + "type": "interline_equation", + "bbox": [ + 238, + 538, + 373, + 552 + ], + "lines": [ + { + "bbox": [ + 238, + 538, + 373, + 552 + ], + "spans": [ + { + "bbox": [ + 238, + 538, + 373, + 552 + ], + "score": 0.92, + "content": "\\pmb { x } _ { k } ( \\tau ) : = ( s _ { t } , s _ { t + 1 } , . . . , s _ { t + H - 1 } ) _ { k }", + "type": "interline_equation", + "image_path": "efff52266db41ee034265912e4c7f6655709a26573a8d48807a02ed9a034d46d.jpg" + } + ] + } + ], + "index": 28, + "virtual_lines": [ + { + "bbox": [ + 238, + 538, + 373, + 552 + ], + "spans": [], + "index": 28 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 558, + 505, + 603 + ], + "lines": [ + { + "bbox": [ + 105, + 557, + 506, + 570 + ], + "spans": [ + { + "bbox": [ + 105, + 557, + 130, + 570 + ], + "score": 1.0, + "content": "Here,", + "type": "text" + }, + { + "bbox": [ + 130, + 559, + 137, + 568 + ], + "score": 0.79, + "content": "k", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 557, + 325, + 570 + ], + "score": 1.0, + "content": "denotes the timestep in the forward process and", + "type": "text" + }, + { + "bbox": [ + 325, + 559, + 330, + 568 + ], + "score": 0.79, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 330, + 557, + 506, + 570 + ], + "score": 1.0, + "content": "denotes the time at which a state was visited", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 104, + 568, + 506, + 582 + ], + "spans": [ + { + "bbox": [ + 104, + 568, + 157, + 582 + ], + "score": 1.0, + "content": "in trajectory", + "type": "text" + }, + { + "bbox": [ + 158, + 571, + 164, + 579 + ], + "score": 0.76, + "content": "\\tau", + "type": "inline_equation" + }, + { + "bbox": [ + 164, + 568, + 292, + 582 + ], + "score": 1.0, + "content": ". Moving forward, we will view", + "type": "text" + }, + { + "bbox": [ + 293, + 569, + 318, + 581 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 318, + 568, + 506, + 582 + ], + "score": 1.0, + "content": "as a noisy sequence of states from a trajectory", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 579, + 506, + 593 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 145, + 593 + ], + "score": 1.0, + "content": "of length", + "type": "text" + }, + { + "bbox": [ + 145, + 581, + 155, + 590 + ], + "score": 0.78, + "content": "H", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 579, + 214, + 593 + ], + "score": 1.0, + "content": ". We represent", + "type": "text" + }, + { + "bbox": [ + 215, + 580, + 240, + 592 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 241, + 579, + 506, + 593 + ], + "score": 1.0, + "content": "as a two-dimensional array with one column for each timestep of", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 590, + 163, + 605 + ], + "spans": [ + { + "bbox": [ + 105, + 590, + 163, + 605 + ], + "score": 1.0, + "content": "the sequence.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 30.5 + }, + { + "type": "text", + "bbox": [ + 106, + 608, + 505, + 653 + ], + "lines": [ + { + "bbox": [ + 105, + 607, + 505, + 621 + ], + "spans": [ + { + "bbox": [ + 105, + 607, + 505, + 621 + ], + "score": 1.0, + "content": "Acting with Inverse-Dynamics. Sampling states from a diffusion model is not enough for defining", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 619, + 506, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 619, + 405, + 631 + ], + "score": 1.0, + "content": "a controller. A policy can, however, be inferred from estimating the action", + "type": "text" + }, + { + "bbox": [ + 405, + 620, + 415, + 630 + ], + "score": 0.84, + "content": "a _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 416, + 619, + 484, + 631 + ], + "score": 1.0, + "content": "that led the state", + "type": "text" + }, + { + "bbox": [ + 484, + 621, + 493, + 630 + ], + "score": 0.84, + "content": "s _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 494, + 619, + 506, + 631 + ], + "score": 1.0, + "content": "to", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 107, + 629, + 506, + 643 + ], + "spans": [ + { + "bbox": [ + 107, + 631, + 126, + 642 + ], + "score": 0.88, + "content": "s _ { t + 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 629, + 195, + 643 + ], + "score": 1.0, + "content": "for any timestep", + "type": "text" + }, + { + "bbox": [ + 195, + 631, + 200, + 640 + ], + "score": 0.65, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 629, + 211, + 643 + ], + "score": 1.0, + "content": "in", + "type": "text" + }, + { + "bbox": [ + 212, + 630, + 236, + 642 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 237, + 629, + 506, + 643 + ], + "score": 1.0, + "content": ". Given two consecutive states, we generate an action according to", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 641, + 392, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 641, + 392, + 653 + ], + "score": 1.0, + "content": "the inverse dynamics model (Agrawal et al., 2016; Pathak et al., 2018):", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34.5 + }, + { + "type": "interline_equation", + "bbox": [ + 268, + 660, + 343, + 674 + ], + "lines": [ + { + "bbox": [ + 268, + 660, + 343, + 674 + ], + "spans": [ + { + "bbox": [ + 268, + 660, + 343, + 674 + ], + "score": 0.93, + "content": "a _ { t } : = f _ { \\phi } ( s _ { t } , s _ { t + 1 } )", + "type": "interline_equation", + "image_path": "d6688a1fe6c467d6c5495ff3e7909d6a6aaedc5b107b9bc358307fbdf2656291.jpg" + } + ] + } + ], + "index": 37, + "virtual_lines": [ + { + "bbox": [ + 268, + 660, + 343, + 674 + ], + "spans": [], + "index": 37 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 681, + 505, + 738 + ], + "lines": [ + { + "bbox": [ + 105, + 680, + 505, + 694 + ], + "spans": [ + { + "bbox": [ + 105, + 680, + 361, + 694 + ], + "score": 1.0, + "content": "Note that the same offline data used to train the reverse process", + "type": "text" + }, + { + "bbox": [ + 361, + 685, + 371, + 693 + ], + "score": 0.7, + "content": "p _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 372, + 680, + 474, + 694 + ], + "score": 1.0, + "content": "can also be used to learn", + "type": "text" + }, + { + "bbox": [ + 474, + 682, + 485, + 694 + ], + "score": 0.88, + "content": "f _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 485, + 680, + 505, + 694 + ], + "score": 1.0, + "content": ". We", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 691, + 505, + 705 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 505, + 705 + ], + "score": 1.0, + "content": "illustrate in Table 2 how the design choice of directly diffusing state distributions, with an inverse", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 703, + 505, + 717 + ], + "spans": [ + { + "bbox": [ + 106, + 703, + 505, + 717 + ], + "score": 1.0, + "content": "dynamics model to predict action, significantly improves performance over diffusing across both", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 714, + 505, + 727 + ], + "spans": [ + { + "bbox": [ + 105, + 714, + 505, + 727 + ], + "score": 1.0, + "content": "states and actions jointly. Furthermore, we empirically compare and analyze when to use inverse", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 725, + 341, + 737 + ], + "spans": [ + { + "bbox": [ + 106, + 725, + 341, + 737 + ], + "score": 1.0, + "content": "dynamics and when to diffuse over actions in Appendix F.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 40 + } + ], + "page_idx": 3, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 309, + 762 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 12, + "width": 8 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 166, + 79, + 444, + 194 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 166, + 79, + 444, + 194 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 166, + 79, + 444, + 194 + ], + "spans": [ + { + "bbox": [ + 166, + 79, + 444, + 194 + ], + "score": 0.972, + "type": "image", + "image_path": "5193fd80ab094ae9c4154b23239d26a5d2b68886d0126e16c31679f9afe8be80.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 166, + 79, + 444, + 117.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 166, + 117.33333333333334, + 444, + 155.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 166, + 155.66666666666669, + 444, + 194.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 203, + 505, + 234 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 203, + 505, + 215 + ], + "spans": [ + { + "bbox": [ + 105, + 203, + 361, + 215 + ], + "score": 1.0, + "content": "Figure 3: Planning with Decision Diffuser. Given the current state", + "type": "text" + }, + { + "bbox": [ + 361, + 205, + 370, + 213 + ], + "score": 0.84, + "content": "s _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 370, + 203, + 505, + 215 + ], + "score": 1.0, + "content": "and conditioning, Decision Diffuser", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 213, + 506, + 225 + ], + "spans": [ + { + "bbox": [ + 105, + 213, + 506, + 225 + ], + "score": 1.0, + "content": "uses classifier-free guidance with low-temperature sampling to generate a sequence of future states. It then uses", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 222, + 462, + 236 + ], + "spans": [ + { + "bbox": [ + 105, + 222, + 288, + 236 + ], + "score": 1.0, + "content": "inverse dynamics to extract and execute the action", + "type": "text" + }, + { + "bbox": [ + 289, + 225, + 298, + 233 + ], + "score": 0.84, + "content": "a _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 298, + 222, + 440, + 236 + ], + "score": 1.0, + "content": "that leads to the immediate future state", + "type": "text" + }, + { + "bbox": [ + 440, + 225, + 458, + 234 + ], + "score": 0.88, + "content": "s _ { t + 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 458, + 222, + 462, + 236 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 107, + 243, + 240, + 254 + ], + "lines": [ + { + "bbox": [ + 105, + 240, + 240, + 258 + ], + "spans": [ + { + "bbox": [ + 105, + 240, + 240, + 258 + ], + "score": 1.0, + "content": "conditional generative modeling:", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 6, + "bbox_fs": [ + 105, + 240, + 240, + 258 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 239, + 262, + 372, + 280 + ], + "lines": [ + { + "bbox": [ + 239, + 262, + 372, + 280 + ], + "spans": [ + { + "bbox": [ + 239, + 262, + 372, + 280 + ], + "score": 0.91, + "content": "\\operatorname* { m a x } _ { \\theta } \\mathbb { E } _ { \\tau \\sim \\mathcal { D } } [ \\log p _ { \\theta } ( \\pmb { x } _ { 0 } ( \\tau ) | \\pmb { y } ( \\tau ) ) ]", + "type": "interline_equation", + "image_path": "5839cb65e52e459da7b8716946f055e40e525790aaf7cc5eb376c9f196100d62.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 239, + 262, + 372, + 280 + ], + "spans": [], + "index": 7 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 288, + 505, + 332 + ], + "lines": [ + { + "bbox": [ + 106, + 288, + 506, + 300 + ], + "spans": [ + { + "bbox": [ + 106, + 288, + 344, + 300 + ], + "score": 1.0, + "content": "Our goal is to estimate the conditional data distribution with", + "type": "text" + }, + { + "bbox": [ + 344, + 290, + 355, + 299 + ], + "score": 0.83, + "content": "p _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 356, + 288, + 506, + 300 + ], + "score": 1.0, + "content": "so we can later generate portions of a", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 299, + 505, + 312 + ], + "spans": [ + { + "bbox": [ + 105, + 299, + 147, + 312 + ], + "score": 1.0, + "content": "trajectory", + "type": "text" + }, + { + "bbox": [ + 148, + 299, + 173, + 311 + ], + "score": 0.92, + "content": "\\pmb { x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 173, + 299, + 246, + 312 + ], + "score": 1.0, + "content": "from information", + "type": "text" + }, + { + "bbox": [ + 247, + 299, + 267, + 311 + ], + "score": 0.92, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 267, + 299, + 357, + 312 + ], + "score": 1.0, + "content": "about it. Examples of", + "type": "text" + }, + { + "bbox": [ + 358, + 301, + 365, + 311 + ], + "score": 0.83, + "content": "\\textbf { { y } }", + "type": "inline_equation" + }, + { + "bbox": [ + 366, + 299, + 505, + 312 + ], + "score": 1.0, + "content": "could include the return under the", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 310, + 505, + 323 + ], + "spans": [ + { + "bbox": [ + 106, + 310, + 505, + 323 + ], + "score": 1.0, + "content": "trajectory, the constraints satisfied by the trajectory, or the skill demonstrated in the trajectory. We", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 321, + 420, + 334 + ], + "spans": [ + { + "bbox": [ + 105, + 321, + 420, + 334 + ], + "score": 1.0, + "content": "construct our generative model according to the conditional diffusion process:", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 9.5, + "bbox_fs": [ + 105, + 288, + 506, + 334 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 207, + 340, + 404, + 354 + ], + "lines": [ + { + "bbox": [ + 207, + 340, + 404, + 354 + ], + "spans": [ + { + "bbox": [ + 207, + 340, + 404, + 354 + ], + "score": 0.87, + "content": "\\begin{array} { r l } { q ( \\pmb { x } _ { k + 1 } ( \\tau ) | \\pmb { x } _ { k } ( \\tau ) ) , } & { { } p _ { \\theta } ( \\pmb { x } _ { k - 1 } ( \\tau ) | \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ( \\tau ) ) } \\end{array}", + "type": "interline_equation", + "image_path": "4a2aedb6b54591ec48f654546dd7beae0401fc8c1a4821d171893e3898154bc3.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 207, + 340, + 404, + 354 + ], + "spans": [], + "index": 12 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 361, + 505, + 428 + ], + "lines": [ + { + "bbox": [ + 105, + 361, + 506, + 374 + ], + "spans": [ + { + "bbox": [ + 105, + 361, + 147, + 374 + ], + "score": 1.0, + "content": "As usual,", + "type": "text" + }, + { + "bbox": [ + 147, + 364, + 153, + 374 + ], + "score": 0.8, + "content": "q", + "type": "inline_equation" + }, + { + "bbox": [ + 154, + 361, + 340, + 374 + ], + "score": 1.0, + "content": "represents the forward noising process while", + "type": "text" + }, + { + "bbox": [ + 340, + 364, + 351, + 373 + ], + "score": 0.85, + "content": "p _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 351, + 361, + 506, + 374 + ], + "score": 1.0, + "content": "the reverse denoising process. In the", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 371, + 505, + 386 + ], + "spans": [ + { + "bbox": [ + 105, + 371, + 505, + 386 + ], + "score": 1.0, + "content": "following, we discuss how we may use diffusion for decision making. First, we discuss the modeling", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 383, + 505, + 396 + ], + "spans": [ + { + "bbox": [ + 106, + 383, + 505, + 396 + ], + "score": 1.0, + "content": "choices for diffusion in Section 3.1. Next, we discuss how we may utilize classifier-free guidance to", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 394, + 506, + 406 + ], + "spans": [ + { + "bbox": [ + 105, + 394, + 506, + 406 + ], + "score": 1.0, + "content": "capture the best aspects of trajectories in Section 3.2. We then discuss the different behaviors that", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 405, + 505, + 418 + ], + "spans": [ + { + "bbox": [ + 105, + 405, + 505, + 418 + ], + "score": 1.0, + "content": "may be implemented with conditional diffusion models in Section 3.3. Finally, we discuss practical", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 416, + 294, + 428 + ], + "spans": [ + { + "bbox": [ + 105, + 416, + 294, + 428 + ], + "score": 1.0, + "content": "training details of our approach in Section 3.4.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 15.5, + "bbox_fs": [ + 105, + 361, + 506, + 428 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 434, + 239, + 445 + ], + "lines": [ + { + "bbox": [ + 105, + 433, + 240, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 433, + 240, + 447 + ], + "score": 1.0, + "content": "3.1 DIFFUSING OVER STATES", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 106, + 447, + 506, + 536 + ], + "lines": [ + { + "bbox": [ + 105, + 448, + 505, + 461 + ], + "spans": [ + { + "bbox": [ + 105, + 448, + 505, + 461 + ], + "score": 1.0, + "content": "In images, the diffusion process is applied across all pixel values in an image. Na¨ıvely, it would", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 458, + 506, + 472 + ], + "spans": [ + { + "bbox": [ + 105, + 458, + 506, + 472 + ], + "score": 1.0, + "content": "therefore be natural to apply a similar process to model the state and actions of a trajectory. However,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 470, + 506, + 483 + ], + "spans": [ + { + "bbox": [ + 105, + 470, + 506, + 483 + ], + "score": 1.0, + "content": "in the reinforcement learning setting, directly modeling actions using a diffusion process has several", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 482, + 506, + 493 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 506, + 493 + ], + "score": 1.0, + "content": "practical issues. First, while states are typically continuous in nature in RL, actions are more varied,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 492, + 506, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 492, + 506, + 504 + ], + "score": 1.0, + "content": "and are often discrete in nature. Furthermore, sequences over actions, which are often represented as", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 104, + 503, + 506, + 515 + ], + "spans": [ + { + "bbox": [ + 104, + 503, + 506, + 515 + ], + "score": 1.0, + "content": "joint torques, tend to be more high-frequency and less smooth, making them much harder to predict", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 514, + 505, + 526 + ], + "spans": [ + { + "bbox": [ + 105, + 514, + 505, + 526 + ], + "score": 1.0, + "content": "and model (Tedrake, 2022). Due to these practical issues, we choose to diffuse only over states, as", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 524, + 168, + 537 + ], + "spans": [ + { + "bbox": [ + 105, + 524, + 168, + 537 + ], + "score": 1.0, + "content": "defined below:", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 23.5, + "bbox_fs": [ + 104, + 448, + 506, + 537 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 238, + 538, + 373, + 552 + ], + "lines": [ + { + "bbox": [ + 238, + 538, + 373, + 552 + ], + "spans": [ + { + "bbox": [ + 238, + 538, + 373, + 552 + ], + "score": 0.92, + "content": "\\pmb { x } _ { k } ( \\tau ) : = ( s _ { t } , s _ { t + 1 } , . . . , s _ { t + H - 1 } ) _ { k }", + "type": "interline_equation", + "image_path": "efff52266db41ee034265912e4c7f6655709a26573a8d48807a02ed9a034d46d.jpg" + } + ] + } + ], + "index": 28, + "virtual_lines": [ + { + "bbox": [ + 238, + 538, + 373, + 552 + ], + "spans": [], + "index": 28 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 558, + 505, + 603 + ], + "lines": [ + { + "bbox": [ + 105, + 557, + 506, + 570 + ], + "spans": [ + { + "bbox": [ + 105, + 557, + 130, + 570 + ], + "score": 1.0, + "content": "Here,", + "type": "text" + }, + { + "bbox": [ + 130, + 559, + 137, + 568 + ], + "score": 0.79, + "content": "k", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 557, + 325, + 570 + ], + "score": 1.0, + "content": "denotes the timestep in the forward process and", + "type": "text" + }, + { + "bbox": [ + 325, + 559, + 330, + 568 + ], + "score": 0.79, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 330, + 557, + 506, + 570 + ], + "score": 1.0, + "content": "denotes the time at which a state was visited", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 104, + 568, + 506, + 582 + ], + "spans": [ + { + "bbox": [ + 104, + 568, + 157, + 582 + ], + "score": 1.0, + "content": "in trajectory", + "type": "text" + }, + { + "bbox": [ + 158, + 571, + 164, + 579 + ], + "score": 0.76, + "content": "\\tau", + "type": "inline_equation" + }, + { + "bbox": [ + 164, + 568, + 292, + 582 + ], + "score": 1.0, + "content": ". Moving forward, we will view", + "type": "text" + }, + { + "bbox": [ + 293, + 569, + 318, + 581 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 318, + 568, + 506, + 582 + ], + "score": 1.0, + "content": "as a noisy sequence of states from a trajectory", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 579, + 506, + 593 + ], + "spans": [ + { + "bbox": [ + 105, + 579, + 145, + 593 + ], + "score": 1.0, + "content": "of length", + "type": "text" + }, + { + "bbox": [ + 145, + 581, + 155, + 590 + ], + "score": 0.78, + "content": "H", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 579, + 214, + 593 + ], + "score": 1.0, + "content": ". We represent", + "type": "text" + }, + { + "bbox": [ + 215, + 580, + 240, + 592 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 241, + 579, + 506, + 593 + ], + "score": 1.0, + "content": "as a two-dimensional array with one column for each timestep of", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 590, + 163, + 605 + ], + "spans": [ + { + "bbox": [ + 105, + 590, + 163, + 605 + ], + "score": 1.0, + "content": "the sequence.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 30.5, + "bbox_fs": [ + 104, + 557, + 506, + 605 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 608, + 505, + 653 + ], + "lines": [ + { + "bbox": [ + 105, + 607, + 505, + 621 + ], + "spans": [ + { + "bbox": [ + 105, + 607, + 505, + 621 + ], + "score": 1.0, + "content": "Acting with Inverse-Dynamics. Sampling states from a diffusion model is not enough for defining", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 619, + 506, + 631 + ], + "spans": [ + { + "bbox": [ + 105, + 619, + 405, + 631 + ], + "score": 1.0, + "content": "a controller. A policy can, however, be inferred from estimating the action", + "type": "text" + }, + { + "bbox": [ + 405, + 620, + 415, + 630 + ], + "score": 0.84, + "content": "a _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 416, + 619, + 484, + 631 + ], + "score": 1.0, + "content": "that led the state", + "type": "text" + }, + { + "bbox": [ + 484, + 621, + 493, + 630 + ], + "score": 0.84, + "content": "s _ { t }", + "type": "inline_equation" + }, + { + "bbox": [ + 494, + 619, + 506, + 631 + ], + "score": 1.0, + "content": "to", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 107, + 629, + 506, + 643 + ], + "spans": [ + { + "bbox": [ + 107, + 631, + 126, + 642 + ], + "score": 0.88, + "content": "s _ { t + 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 629, + 195, + 643 + ], + "score": 1.0, + "content": "for any timestep", + "type": "text" + }, + { + "bbox": [ + 195, + 631, + 200, + 640 + ], + "score": 0.65, + "content": "t", + "type": "inline_equation" + }, + { + "bbox": [ + 200, + 629, + 211, + 643 + ], + "score": 1.0, + "content": "in", + "type": "text" + }, + { + "bbox": [ + 212, + 630, + 236, + 642 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 237, + 629, + 506, + 643 + ], + "score": 1.0, + "content": ". Given two consecutive states, we generate an action according to", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 641, + 392, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 641, + 392, + 653 + ], + "score": 1.0, + "content": "the inverse dynamics model (Agrawal et al., 2016; Pathak et al., 2018):", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34.5, + "bbox_fs": [ + 105, + 607, + 506, + 653 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 268, + 660, + 343, + 674 + ], + "lines": [ + { + "bbox": [ + 268, + 660, + 343, + 674 + ], + "spans": [ + { + "bbox": [ + 268, + 660, + 343, + 674 + ], + "score": 0.93, + "content": "a _ { t } : = f _ { \\phi } ( s _ { t } , s _ { t + 1 } )", + "type": "interline_equation", + "image_path": "d6688a1fe6c467d6c5495ff3e7909d6a6aaedc5b107b9bc358307fbdf2656291.jpg" + } + ] + } + ], + "index": 37, + "virtual_lines": [ + { + "bbox": [ + 268, + 660, + 343, + 674 + ], + "spans": [], + "index": 37 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 681, + 505, + 738 + ], + "lines": [ + { + "bbox": [ + 105, + 680, + 505, + 694 + ], + "spans": [ + { + "bbox": [ + 105, + 680, + 361, + 694 + ], + "score": 1.0, + "content": "Note that the same offline data used to train the reverse process", + "type": "text" + }, + { + "bbox": [ + 361, + 685, + 371, + 693 + ], + "score": 0.7, + "content": "p _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 372, + 680, + 474, + 694 + ], + "score": 1.0, + "content": "can also be used to learn", + "type": "text" + }, + { + "bbox": [ + 474, + 682, + 485, + 694 + ], + "score": 0.88, + "content": "f _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 485, + 680, + 505, + 694 + ], + "score": 1.0, + "content": ". We", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 691, + 505, + 705 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 505, + 705 + ], + "score": 1.0, + "content": "illustrate in Table 2 how the design choice of directly diffusing state distributions, with an inverse", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 703, + 505, + 717 + ], + "spans": [ + { + "bbox": [ + 106, + 703, + 505, + 717 + ], + "score": 1.0, + "content": "dynamics model to predict action, significantly improves performance over diffusing across both", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 714, + 505, + 727 + ], + "spans": [ + { + "bbox": [ + 105, + 714, + 505, + 727 + ], + "score": 1.0, + "content": "states and actions jointly. Furthermore, we empirically compare and analyze when to use inverse", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 725, + 341, + 737 + ], + "spans": [ + { + "bbox": [ + 106, + 725, + 341, + 737 + ], + "score": 1.0, + "content": "dynamics and when to diffuse over actions in Appendix F.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 40, + "bbox_fs": [ + 105, + 680, + 505, + 737 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 106, + 82, + 332, + 93 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 333, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 333, + 95 + ], + "score": 1.0, + "content": "3.2 PLANNING WITH CLASSIFIER-FREE GUIDANCE", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 106, + 96, + 505, + 163 + ], + "lines": [ + { + "bbox": [ + 106, + 95, + 505, + 109 + ], + "spans": [ + { + "bbox": [ + 106, + 95, + 505, + 109 + ], + "score": 1.0, + "content": "Given a diffusion model representing the different trajectories in a dataset, we next discuss how", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 104, + 106, + 506, + 121 + ], + "spans": [ + { + "bbox": [ + 104, + 106, + 506, + 121 + ], + "score": 1.0, + "content": "we may utilize the diffusion model for planning. To use the model for planning, it is necessary", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 118, + 506, + 131 + ], + "spans": [ + { + "bbox": [ + 105, + 118, + 371, + 131 + ], + "score": 1.0, + "content": "to additionally condition the diffusion process on characteristics", + "type": "text" + }, + { + "bbox": [ + 371, + 118, + 392, + 129 + ], + "score": 0.92, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 392, + 118, + 506, + 131 + ], + "score": 1.0, + "content": ". One approach could be to", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 128, + 505, + 142 + ], + "spans": [ + { + "bbox": [ + 105, + 128, + 171, + 142 + ], + "score": 1.0, + "content": "train a classifier", + "type": "text" + }, + { + "bbox": [ + 172, + 129, + 237, + 141 + ], + "score": 0.93, + "content": "p _ { \\phi } ( \\pmb { y } ( \\tau ) | \\pmb { x } _ { k } ( \\tau ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 237, + 128, + 279, + 142 + ], + "score": 1.0, + "content": "to predict", + "type": "text" + }, + { + "bbox": [ + 279, + 129, + 299, + 141 + ], + "score": 0.92, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 299, + 128, + 392, + 142 + ], + "score": 1.0, + "content": "from noisy trajectories", + "type": "text" + }, + { + "bbox": [ + 392, + 129, + 418, + 141 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 418, + 128, + 484, + 142 + ], + "score": 1.0, + "content": ". In the case that", + "type": "text" + }, + { + "bbox": [ + 484, + 129, + 505, + 141 + ], + "score": 0.91, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 140, + 506, + 153 + ], + "spans": [ + { + "bbox": [ + 105, + 140, + 388, + 153 + ], + "score": 1.0, + "content": "represents the return under a trajectory, this would require estimating a", + "type": "text" + }, + { + "bbox": [ + 388, + 141, + 397, + 151 + ], + "score": 0.85, + "content": "Q", + "type": "inline_equation" + }, + { + "bbox": [ + 398, + 140, + 506, + 153 + ], + "score": 1.0, + "content": "-function, which requires a", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 151, + 319, + 164 + ], + "spans": [ + { + "bbox": [ + 105, + 151, + 319, + 164 + ], + "score": 1.0, + "content": "separate, complex dynamic programming procedure.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 106, + 167, + 505, + 267 + ], + "lines": [ + { + "bbox": [ + 106, + 167, + 505, + 181 + ], + "spans": [ + { + "bbox": [ + 106, + 167, + 505, + 181 + ], + "score": 1.0, + "content": "One approach to avoid dynamic programming is to directly train a conditional diffusion model", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 179, + 505, + 191 + ], + "spans": [ + { + "bbox": [ + 106, + 179, + 218, + 191 + ], + "score": 1.0, + "content": "conditioned on the returns", + "type": "text" + }, + { + "bbox": [ + 218, + 179, + 239, + 191 + ], + "score": 0.91, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 179, + 505, + 191 + ], + "score": 1.0, + "content": "in the offline dataset. However, as our dataset consists of a set", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 189, + 505, + 203 + ], + "spans": [ + { + "bbox": [ + 106, + 189, + 505, + 203 + ], + "score": 1.0, + "content": "of sub-optimal trajectories, the conditional diffusion model will be polluted by such sub-optimal", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 200, + 505, + 212 + ], + "spans": [ + { + "bbox": [ + 106, + 200, + 505, + 212 + ], + "score": 1.0, + "content": "behaviors. To circumvent this issue, we utilize classifier-free guidance (Ho & Salimans, 2022) with", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 212, + 505, + 224 + ], + "spans": [ + { + "bbox": [ + 105, + 212, + 505, + 224 + ], + "score": 1.0, + "content": "low-temperature sampling, to extract high-likelihood trajectories in the dataset. We find that such", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 221, + 505, + 236 + ], + "spans": [ + { + "bbox": [ + 105, + 221, + 505, + 236 + ], + "score": 1.0, + "content": "trajectories correspond to the best set of behaviors in the dataset. For a detailed discussion comparing", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 234, + 505, + 245 + ], + "spans": [ + { + "bbox": [ + 106, + 234, + 505, + 245 + ], + "score": 1.0, + "content": "Q-function guidance and classifier-free guidance, please refer to Appendix K. Formally, to implement", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 243, + 506, + 258 + ], + "spans": [ + { + "bbox": [ + 105, + 243, + 215, + 258 + ], + "score": 1.0, + "content": "classifier free guidance, a", + "type": "text" + }, + { + "bbox": [ + 215, + 244, + 240, + 256 + ], + "score": 0.92, + "content": "\\pmb { x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 240, + 243, + 422, + 258 + ], + "score": 1.0, + "content": "is sampled by starting with Gaussian noise", + "type": "text" + }, + { + "bbox": [ + 423, + 244, + 451, + 257 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { K } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 243, + 506, + 258 + ], + "score": 1.0, + "content": "and refining", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 107, + 255, + 410, + 268 + ], + "spans": [ + { + "bbox": [ + 107, + 255, + 132, + 267 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 132, + 255, + 151, + 268 + ], + "score": 1.0, + "content": "into", + "type": "text" + }, + { + "bbox": [ + 152, + 255, + 187, + 267 + ], + "score": 0.93, + "content": "\\pmb { x } _ { k - 1 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 188, + 255, + 410, + 268 + ], + "score": 1.0, + "content": "at each intermediate timestep with the perturbed noise:", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 11 + }, + { + "type": "interline_equation", + "bbox": [ + 173, + 272, + 437, + 286 + ], + "lines": [ + { + "bbox": [ + 173, + 272, + 437, + 286 + ], + "spans": [ + { + "bbox": [ + 173, + 272, + 437, + 286 + ], + "score": 0.87, + "content": "\\hat { \\epsilon } : = \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , \\mathcal { O } , k ) + \\omega ( \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , { \\pmb y } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , \\mathcal { O } , k ) ) ,", + "type": "interline_equation", + "image_path": "25d4496eecae08b1222f19972ee9ea51012b605922dabff02464df366d671eed.jpg" + } + ] + } + ], + "index": 16, + "virtual_lines": [ + { + "bbox": [ + 173, + 272, + 437, + 286 + ], + "spans": [], + "index": 16 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 291, + 505, + 369 + ], + "lines": [ + { + "bbox": [ + 105, + 290, + 506, + 306 + ], + "spans": [ + { + "bbox": [ + 105, + 290, + 174, + 306 + ], + "score": 1.0, + "content": "where the scalar", + "type": "text" + }, + { + "bbox": [ + 174, + 294, + 182, + 302 + ], + "score": 0.78, + "content": "\\omega", + "type": "inline_equation" + }, + { + "bbox": [ + 182, + 290, + 226, + 306 + ], + "score": 1.0, + "content": "applied to", + "type": "text" + }, + { + "bbox": [ + 227, + 291, + 384, + 304 + ], + "score": 0.91, + "content": "( \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , { \\pmb y } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , \\emptyset , k ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 384, + 290, + 506, + 306 + ], + "score": 1.0, + "content": "seeks to augment and extract", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 303, + 505, + 315 + ], + "spans": [ + { + "bbox": [ + 106, + 303, + 329, + 315 + ], + "score": 1.0, + "content": "the best portions of trajectories in the dataset that exhibit", + "type": "text" + }, + { + "bbox": [ + 329, + 304, + 349, + 315 + ], + "score": 0.89, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 303, + 505, + 315 + ], + "score": 1.0, + "content": ". With these ingredients, sampling from", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 313, + 506, + 326 + ], + "spans": [ + { + "bbox": [ + 106, + 313, + 506, + 326 + ], + "score": 1.0, + "content": "the Decision Diffuser becomes similar to planning in RL. First, we observe a state in the environment.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 324, + 505, + 338 + ], + "spans": [ + { + "bbox": [ + 105, + 324, + 450, + 338 + ], + "score": 1.0, + "content": "Next, we sample states later into the horizon with our diffusion process conditioned on", + "type": "text" + }, + { + "bbox": [ + 450, + 327, + 458, + 336 + ], + "score": 0.81, + "content": "\\textbf { { y } }", + "type": "inline_equation" + }, + { + "bbox": [ + 458, + 324, + 505, + 338 + ], + "score": 1.0, + "content": "and history", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 335, + 506, + 348 + ], + "spans": [ + { + "bbox": [ + 106, + 335, + 135, + 348 + ], + "score": 1.0, + "content": "of last", + "type": "text" + }, + { + "bbox": [ + 136, + 336, + 145, + 345 + ], + "score": 0.82, + "content": "C", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 335, + 506, + 348 + ], + "score": 1.0, + "content": "states observed. Finally, we identify the action that should be taken to reach the most", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 347, + 506, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 506, + 358 + ], + "score": 1.0, + "content": "immediate predicted state with our inverse dynamics model. This procedure repeats in a standard", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 358, + 439, + 370 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 439, + 370 + ], + "score": 1.0, + "content": "receding-horizon control loop described in Algorithm 1 and visualized in Figure 3.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 20 + }, + { + "type": "title", + "bbox": [ + 107, + 375, + 280, + 387 + ], + "lines": [ + { + "bbox": [ + 105, + 374, + 281, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 374, + 281, + 388 + ], + "score": 1.0, + "content": "3.3 CONDITIONING BEYOND RETURNS", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 107, + 389, + 504, + 422 + ], + "lines": [ + { + "bbox": [ + 105, + 389, + 506, + 402 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 352, + 402 + ], + "score": 1.0, + "content": "So far we have not explicitly defined the conditioning variable", + "type": "text" + }, + { + "bbox": [ + 352, + 389, + 372, + 401 + ], + "score": 0.91, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 373, + 389, + 506, + 402 + ], + "score": 1.0, + "content": ". Though we have mentioned that", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 401, + 505, + 412 + ], + "spans": [ + { + "bbox": [ + 106, + 401, + 505, + 412 + ], + "score": 1.0, + "content": "it can be the return under a trajectory, we may also consider guiding our diffusion process towards", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 411, + 445, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 445, + 424 + ], + "score": 1.0, + "content": "sequences of states that satisfy relevant constraints or demonstrate specific behavior.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26 + }, + { + "type": "text", + "bbox": [ + 106, + 424, + 505, + 480 + ], + "lines": [ + { + "bbox": [ + 106, + 425, + 505, + 437 + ], + "spans": [ + { + "bbox": [ + 106, + 425, + 505, + 437 + ], + "score": 1.0, + "content": "Maximizing Returns To generate trajectories that maximize return, we condition the noise", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 435, + 505, + 449 + ], + "spans": [ + { + "bbox": [ + 106, + 435, + 260, + 449 + ], + "score": 1.0, + "content": "model on the return of a trajectory so", + "type": "text" + }, + { + "bbox": [ + 260, + 435, + 428, + 448 + ], + "score": 0.92, + "content": "\\begin{array} { r } { \\epsilon _ { \\theta } ( \\bar { \\mathbf { x } } _ { k } ( \\tau ) , \\mathbf { \\boldsymbol { y } } ( \\tau ) , k ) : = \\epsilon _ { \\theta } ( \\mathbf { \\Delta x } _ { k } ( \\tau ) , R ( \\tau ) , k ) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 435, + 505, + 449 + ], + "score": 1.0, + "content": ". These returns are", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 446, + 505, + 460 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 188, + 460 + ], + "score": 1.0, + "content": "normalized to keep", + "type": "text" + }, + { + "bbox": [ + 189, + 447, + 244, + 459 + ], + "score": 0.92, + "content": "R ( \\tau ) \\in [ 0 , 1 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 244, + 446, + 505, + 460 + ], + "score": 1.0, + "content": ". Sampling a high return trajectory amounts to conditioning on", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 107, + 456, + 505, + 471 + ], + "spans": [ + { + "bbox": [ + 107, + 457, + 150, + 470 + ], + "score": 0.92, + "content": "R ( \\tau ) = 1", + "type": "inline_equation" + }, + { + "bbox": [ + 150, + 456, + 315, + 471 + ], + "score": 1.0, + "content": ". Note that we do not make use of any", + "type": "text" + }, + { + "bbox": [ + 315, + 458, + 324, + 469 + ], + "score": 0.86, + "content": "Q", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 456, + 505, + 471 + ], + "score": 1.0, + "content": "-values, which would then require dynamic", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 469, + 166, + 483 + ], + "spans": [ + { + "bbox": [ + 105, + 469, + 166, + 483 + ], + "score": 1.0, + "content": "programming.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 106, + 483, + 505, + 549 + ], + "lines": [ + { + "bbox": [ + 106, + 483, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 106, + 483, + 505, + 495 + ], + "score": 1.0, + "content": "Satisfying Constraints Trajectories may satisfy a variety of constraints, each represented by the", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 493, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 106, + 493, + 120, + 506 + ], + "score": 1.0, + "content": "set", + "type": "text" + }, + { + "bbox": [ + 120, + 494, + 130, + 505 + ], + "score": 0.88, + "content": "\\mathcal { C } _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 130, + 493, + 505, + 506 + ], + "score": 1.0, + "content": ", such as reaching a specific goal, visiting states in a particular order, or avoiding parts of the", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 505, + 505, + 518 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 361, + 518 + ], + "score": 1.0, + "content": "state space. To generate trajectories satisfying a given constraint", + "type": "text" + }, + { + "bbox": [ + 361, + 505, + 370, + 515 + ], + "score": 0.84, + "content": "\\mathcal { C } _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 371, + 505, + 505, + 518 + ], + "score": 1.0, + "content": ", we condition the noise model on", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 515, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 216, + 528 + ], + "score": 1.0, + "content": "a one-hot encoding so that", + "type": "text" + }, + { + "bbox": [ + 217, + 515, + 403, + 528 + ], + "score": 0.92, + "content": "\\begin{array} { r } { \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ( \\tau ) , \\bar { k } ) : = \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\mathbb { 1 } ( \\tau \\in \\mathcal { C } _ { i } ) , k ) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 403, + 515, + 505, + 528 + ], + "score": 1.0, + "content": ". Although we train with", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 527, + 506, + 539 + ], + "spans": [ + { + "bbox": [ + 106, + 527, + 506, + 539 + ], + "score": 1.0, + "content": "an offline dataset in which trajectories satisfy only one of the available constraints, at inference we", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 537, + 264, + 550 + ], + "spans": [ + { + "bbox": [ + 106, + 537, + 264, + 550 + ], + "score": 1.0, + "content": "can satisfy several constraints together.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 35.5 + }, + { + "type": "text", + "bbox": [ + 107, + 551, + 505, + 596 + ], + "lines": [ + { + "bbox": [ + 106, + 551, + 505, + 564 + ], + "spans": [ + { + "bbox": [ + 106, + 551, + 229, + 564 + ], + "score": 1.0, + "content": "Composing Skills A skill", + "type": "text" + }, + { + "bbox": [ + 230, + 552, + 235, + 561 + ], + "score": 0.65, + "content": "i", + "type": "inline_equation" + }, + { + "bbox": [ + 235, + 551, + 433, + 564 + ], + "score": 1.0, + "content": "can be specified from a set of demonstrations", + "type": "text" + }, + { + "bbox": [ + 434, + 552, + 445, + 563 + ], + "score": 0.88, + "content": "B _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 445, + 551, + 505, + 564 + ], + "score": 1.0, + "content": ". To generate", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 562, + 505, + 575 + ], + "spans": [ + { + "bbox": [ + 105, + 562, + 505, + 575 + ], + "score": 1.0, + "content": "trajectories that demonstrate a given skill, we condition the noise model on a one-hot encoding so", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 573, + 505, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 573, + 124, + 586 + ], + "score": 1.0, + "content": "that", + "type": "text" + }, + { + "bbox": [ + 124, + 573, + 315, + 585 + ], + "score": 0.91, + "content": "\\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , { \\pmb y } ( \\tau ) , k ) : = \\epsilon _ { \\theta } ( \\bar { { \\pmb x } } _ { k } ( \\tau ) , \\mathbb { 1 } ( \\tau \\in \\mathcal { B } _ { i } ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 316, + 573, + 505, + 586 + ], + "score": 1.0, + "content": ". Although we train with individual skills, we", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 585, + 346, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 346, + 597 + ], + "score": 1.0, + "content": "may further compose these skills together during inference.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 40.5 + }, + { + "type": "text", + "bbox": [ + 107, + 601, + 505, + 635 + ], + "lines": [ + { + "bbox": [ + 105, + 600, + 506, + 614 + ], + "spans": [ + { + "bbox": [ + 105, + 600, + 334, + 614 + ], + "score": 1.0, + "content": "Assuming we have learned the data distributions", + "type": "text" + }, + { + "bbox": [ + 334, + 600, + 486, + 614 + ], + "score": 0.91, + "content": "q ( { \\pmb x } _ { 0 } ( \\tau ) | { \\pmb y } ^ { 1 } ( \\tau ) ) , \\dots , q ( { \\pmb x } _ { 0 } ( \\tau ) | { \\pmb y } ^ { n } ( \\tau ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 486, + 600, + 506, + 614 + ], + "score": 1.0, + "content": "for", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 107, + 612, + 505, + 624 + ], + "spans": [ + { + "bbox": [ + 107, + 614, + 114, + 622 + ], + "score": 0.74, + "content": "n", + "type": "inline_equation" + }, + { + "bbox": [ + 114, + 612, + 505, + 624 + ], + "score": 1.0, + "content": "different conditioning variables, we can sample from the composed data distribution", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 107, + 622, + 396, + 636 + ], + "spans": [ + { + "bbox": [ + 107, + 623, + 218, + 636 + ], + "score": 0.91, + "content": "q ( \\pmb { x } _ { 0 } ( \\tau ) | \\pmb { y } ^ { 1 } ( \\tau ) , \\dots , \\pmb { y } ^ { n } ( \\bar { \\tau } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 218, + 622, + 396, + 636 + ], + "score": 1.0, + "content": "using the perturbed noise (Liu et al., 2022):", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44 + }, + { + "type": "interline_equation", + "bbox": [ + 164, + 640, + 445, + 673 + ], + "lines": [ + { + "bbox": [ + 164, + 640, + 445, + 673 + ], + "spans": [ + { + "bbox": [ + 164, + 640, + 445, + 673 + ], + "score": 0.94, + "content": "\\hat { \\epsilon } : = \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\emptyset , k ) + \\omega \\sum _ { i = 1 } ^ { n } ( \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ^ { i } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\emptyset , k ) )", + "type": "interline_equation", + "image_path": "387eca885b4e7a9ca4a4a7e4e308a743d03ad7f9ae1155a2384c8112667bea94.jpg" + } + ] + } + ], + "index": 47, + "virtual_lines": [ + { + "bbox": [ + 164, + 640, + 445, + 651.0 + ], + "spans": [], + "index": 46 + }, + { + "bbox": [ + 164, + 651.0, + 445, + 662.0 + ], + "spans": [], + "index": 47 + }, + { + "bbox": [ + 164, + 662.0, + 445, + 673.0 + ], + "spans": [], + "index": 48 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 678, + 506, + 735 + ], + "lines": [ + { + "bbox": [ + 104, + 677, + 508, + 695 + ], + "spans": [ + { + "bbox": [ + 104, + 677, + 213, + 695 + ], + "score": 1.0, + "content": "This property assumes that", + "type": "text" + }, + { + "bbox": [ + 214, + 678, + 260, + 692 + ], + "score": 0.93, + "content": "\\{ \\pmb { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 677, + 477, + 695 + ], + "score": 1.0, + "content": "are conditionally independent given the state trajectory", + "type": "text" + }, + { + "bbox": [ + 478, + 679, + 502, + 691 + ], + "score": 0.92, + "content": "\\pmb { x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 677, + 508, + 695 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 689, + 506, + 704 + ], + "spans": [ + { + "bbox": [ + 105, + 689, + 506, + 704 + ], + "score": 1.0, + "content": "However, we empirically observe that this assumption doesn’t have to be strictly satisfied as long as", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 702, + 505, + 713 + ], + "spans": [ + { + "bbox": [ + 106, + 702, + 505, + 713 + ], + "score": 1.0, + "content": "the composition of conditioning variables is feasible. For more detailed discussion, please refer to", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 106, + 712, + 506, + 726 + ], + "spans": [ + { + "bbox": [ + 106, + 712, + 506, + 726 + ], + "score": 1.0, + "content": "Appendix D. We use this property to compose more than one constraint or skill together at test-time.", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 105, + 722, + 501, + 736 + ], + "spans": [ + { + "bbox": [ + 105, + 722, + 501, + 736 + ], + "score": 1.0, + "content": "We also show how Decision Diffuser can avoid particular constraint or skill (NOT) in Appendix J.", + "type": "text" + } + ], + "index": 53 + } + ], + "index": 51 + } + ], + "page_idx": 4, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "score": 1.0, + "content": "5", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 106, + 82, + 332, + 93 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 333, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 333, + 95 + ], + "score": 1.0, + "content": "3.2 PLANNING WITH CLASSIFIER-FREE GUIDANCE", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 106, + 96, + 505, + 163 + ], + "lines": [ + { + "bbox": [ + 106, + 95, + 505, + 109 + ], + "spans": [ + { + "bbox": [ + 106, + 95, + 505, + 109 + ], + "score": 1.0, + "content": "Given a diffusion model representing the different trajectories in a dataset, we next discuss how", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 104, + 106, + 506, + 121 + ], + "spans": [ + { + "bbox": [ + 104, + 106, + 506, + 121 + ], + "score": 1.0, + "content": "we may utilize the diffusion model for planning. To use the model for planning, it is necessary", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 118, + 506, + 131 + ], + "spans": [ + { + "bbox": [ + 105, + 118, + 371, + 131 + ], + "score": 1.0, + "content": "to additionally condition the diffusion process on characteristics", + "type": "text" + }, + { + "bbox": [ + 371, + 118, + 392, + 129 + ], + "score": 0.92, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 392, + 118, + 506, + 131 + ], + "score": 1.0, + "content": ". One approach could be to", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 128, + 505, + 142 + ], + "spans": [ + { + "bbox": [ + 105, + 128, + 171, + 142 + ], + "score": 1.0, + "content": "train a classifier", + "type": "text" + }, + { + "bbox": [ + 172, + 129, + 237, + 141 + ], + "score": 0.93, + "content": "p _ { \\phi } ( \\pmb { y } ( \\tau ) | \\pmb { x } _ { k } ( \\tau ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 237, + 128, + 279, + 142 + ], + "score": 1.0, + "content": "to predict", + "type": "text" + }, + { + "bbox": [ + 279, + 129, + 299, + 141 + ], + "score": 0.92, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 299, + 128, + 392, + 142 + ], + "score": 1.0, + "content": "from noisy trajectories", + "type": "text" + }, + { + "bbox": [ + 392, + 129, + 418, + 141 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 418, + 128, + 484, + 142 + ], + "score": 1.0, + "content": ". In the case that", + "type": "text" + }, + { + "bbox": [ + 484, + 129, + 505, + 141 + ], + "score": 0.91, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 140, + 506, + 153 + ], + "spans": [ + { + "bbox": [ + 105, + 140, + 388, + 153 + ], + "score": 1.0, + "content": "represents the return under a trajectory, this would require estimating a", + "type": "text" + }, + { + "bbox": [ + 388, + 141, + 397, + 151 + ], + "score": 0.85, + "content": "Q", + "type": "inline_equation" + }, + { + "bbox": [ + 398, + 140, + 506, + 153 + ], + "score": 1.0, + "content": "-function, which requires a", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 151, + 319, + 164 + ], + "spans": [ + { + "bbox": [ + 105, + 151, + 319, + 164 + ], + "score": 1.0, + "content": "separate, complex dynamic programming procedure.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 3.5, + "bbox_fs": [ + 104, + 95, + 506, + 164 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 167, + 505, + 267 + ], + "lines": [ + { + "bbox": [ + 106, + 167, + 505, + 181 + ], + "spans": [ + { + "bbox": [ + 106, + 167, + 505, + 181 + ], + "score": 1.0, + "content": "One approach to avoid dynamic programming is to directly train a conditional diffusion model", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 179, + 505, + 191 + ], + "spans": [ + { + "bbox": [ + 106, + 179, + 218, + 191 + ], + "score": 1.0, + "content": "conditioned on the returns", + "type": "text" + }, + { + "bbox": [ + 218, + 179, + 239, + 191 + ], + "score": 0.91, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 179, + 505, + 191 + ], + "score": 1.0, + "content": "in the offline dataset. However, as our dataset consists of a set", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 189, + 505, + 203 + ], + "spans": [ + { + "bbox": [ + 106, + 189, + 505, + 203 + ], + "score": 1.0, + "content": "of sub-optimal trajectories, the conditional diffusion model will be polluted by such sub-optimal", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 200, + 505, + 212 + ], + "spans": [ + { + "bbox": [ + 106, + 200, + 505, + 212 + ], + "score": 1.0, + "content": "behaviors. To circumvent this issue, we utilize classifier-free guidance (Ho & Salimans, 2022) with", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 212, + 505, + 224 + ], + "spans": [ + { + "bbox": [ + 105, + 212, + 505, + 224 + ], + "score": 1.0, + "content": "low-temperature sampling, to extract high-likelihood trajectories in the dataset. We find that such", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 221, + 505, + 236 + ], + "spans": [ + { + "bbox": [ + 105, + 221, + 505, + 236 + ], + "score": 1.0, + "content": "trajectories correspond to the best set of behaviors in the dataset. For a detailed discussion comparing", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 234, + 505, + 245 + ], + "spans": [ + { + "bbox": [ + 106, + 234, + 505, + 245 + ], + "score": 1.0, + "content": "Q-function guidance and classifier-free guidance, please refer to Appendix K. Formally, to implement", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 243, + 506, + 258 + ], + "spans": [ + { + "bbox": [ + 105, + 243, + 215, + 258 + ], + "score": 1.0, + "content": "classifier free guidance, a", + "type": "text" + }, + { + "bbox": [ + 215, + 244, + 240, + 256 + ], + "score": 0.92, + "content": "\\pmb { x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 240, + 243, + 422, + 258 + ], + "score": 1.0, + "content": "is sampled by starting with Gaussian noise", + "type": "text" + }, + { + "bbox": [ + 423, + 244, + 451, + 257 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { K } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 451, + 243, + 506, + 258 + ], + "score": 1.0, + "content": "and refining", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 107, + 255, + 410, + 268 + ], + "spans": [ + { + "bbox": [ + 107, + 255, + 132, + 267 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 132, + 255, + 151, + 268 + ], + "score": 1.0, + "content": "into", + "type": "text" + }, + { + "bbox": [ + 152, + 255, + 187, + 267 + ], + "score": 0.93, + "content": "\\pmb { x } _ { k - 1 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 188, + 255, + 410, + 268 + ], + "score": 1.0, + "content": "at each intermediate timestep with the perturbed noise:", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 167, + 506, + 268 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 173, + 272, + 437, + 286 + ], + "lines": [ + { + "bbox": [ + 173, + 272, + 437, + 286 + ], + "spans": [ + { + "bbox": [ + 173, + 272, + 437, + 286 + ], + "score": 0.87, + "content": "\\hat { \\epsilon } : = \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , \\mathcal { O } , k ) + \\omega ( \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , { \\pmb y } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , \\mathcal { O } , k ) ) ,", + "type": "interline_equation", + "image_path": "25d4496eecae08b1222f19972ee9ea51012b605922dabff02464df366d671eed.jpg" + } + ] + } + ], + "index": 16, + "virtual_lines": [ + { + "bbox": [ + 173, + 272, + 437, + 286 + ], + "spans": [], + "index": 16 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 291, + 505, + 369 + ], + "lines": [ + { + "bbox": [ + 105, + 290, + 506, + 306 + ], + "spans": [ + { + "bbox": [ + 105, + 290, + 174, + 306 + ], + "score": 1.0, + "content": "where the scalar", + "type": "text" + }, + { + "bbox": [ + 174, + 294, + 182, + 302 + ], + "score": 0.78, + "content": "\\omega", + "type": "inline_equation" + }, + { + "bbox": [ + 182, + 290, + 226, + 306 + ], + "score": 1.0, + "content": "applied to", + "type": "text" + }, + { + "bbox": [ + 227, + 291, + 384, + 304 + ], + "score": 0.91, + "content": "( \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , { \\pmb y } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , \\emptyset , k ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 384, + 290, + 506, + 306 + ], + "score": 1.0, + "content": "seeks to augment and extract", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 303, + 505, + 315 + ], + "spans": [ + { + "bbox": [ + 106, + 303, + 329, + 315 + ], + "score": 1.0, + "content": "the best portions of trajectories in the dataset that exhibit", + "type": "text" + }, + { + "bbox": [ + 329, + 304, + 349, + 315 + ], + "score": 0.89, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 349, + 303, + 505, + 315 + ], + "score": 1.0, + "content": ". With these ingredients, sampling from", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 313, + 506, + 326 + ], + "spans": [ + { + "bbox": [ + 106, + 313, + 506, + 326 + ], + "score": 1.0, + "content": "the Decision Diffuser becomes similar to planning in RL. First, we observe a state in the environment.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 324, + 505, + 338 + ], + "spans": [ + { + "bbox": [ + 105, + 324, + 450, + 338 + ], + "score": 1.0, + "content": "Next, we sample states later into the horizon with our diffusion process conditioned on", + "type": "text" + }, + { + "bbox": [ + 450, + 327, + 458, + 336 + ], + "score": 0.81, + "content": "\\textbf { { y } }", + "type": "inline_equation" + }, + { + "bbox": [ + 458, + 324, + 505, + 338 + ], + "score": 1.0, + "content": "and history", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 335, + 506, + 348 + ], + "spans": [ + { + "bbox": [ + 106, + 335, + 135, + 348 + ], + "score": 1.0, + "content": "of last", + "type": "text" + }, + { + "bbox": [ + 136, + 336, + 145, + 345 + ], + "score": 0.82, + "content": "C", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 335, + 506, + 348 + ], + "score": 1.0, + "content": "states observed. Finally, we identify the action that should be taken to reach the most", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 347, + 506, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 506, + 358 + ], + "score": 1.0, + "content": "immediate predicted state with our inverse dynamics model. This procedure repeats in a standard", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 358, + 439, + 370 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 439, + 370 + ], + "score": 1.0, + "content": "receding-horizon control loop described in Algorithm 1 and visualized in Figure 3.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 20, + "bbox_fs": [ + 105, + 290, + 506, + 370 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 375, + 280, + 387 + ], + "lines": [ + { + "bbox": [ + 105, + 374, + 281, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 374, + 281, + 388 + ], + "score": 1.0, + "content": "3.3 CONDITIONING BEYOND RETURNS", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 107, + 389, + 504, + 422 + ], + "lines": [ + { + "bbox": [ + 105, + 389, + 506, + 402 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 352, + 402 + ], + "score": 1.0, + "content": "So far we have not explicitly defined the conditioning variable", + "type": "text" + }, + { + "bbox": [ + 352, + 389, + 372, + 401 + ], + "score": 0.91, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 373, + 389, + 506, + 402 + ], + "score": 1.0, + "content": ". Though we have mentioned that", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 401, + 505, + 412 + ], + "spans": [ + { + "bbox": [ + 106, + 401, + 505, + 412 + ], + "score": 1.0, + "content": "it can be the return under a trajectory, we may also consider guiding our diffusion process towards", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 411, + 445, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 445, + 424 + ], + "score": 1.0, + "content": "sequences of states that satisfy relevant constraints or demonstrate specific behavior.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26, + "bbox_fs": [ + 105, + 389, + 506, + 424 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 424, + 505, + 480 + ], + "lines": [ + { + "bbox": [ + 106, + 425, + 505, + 437 + ], + "spans": [ + { + "bbox": [ + 106, + 425, + 505, + 437 + ], + "score": 1.0, + "content": "Maximizing Returns To generate trajectories that maximize return, we condition the noise", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 435, + 505, + 449 + ], + "spans": [ + { + "bbox": [ + 106, + 435, + 260, + 449 + ], + "score": 1.0, + "content": "model on the return of a trajectory so", + "type": "text" + }, + { + "bbox": [ + 260, + 435, + 428, + 448 + ], + "score": 0.92, + "content": "\\begin{array} { r } { \\epsilon _ { \\theta } ( \\bar { \\mathbf { x } } _ { k } ( \\tau ) , \\mathbf { \\boldsymbol { y } } ( \\tau ) , k ) : = \\epsilon _ { \\theta } ( \\mathbf { \\Delta x } _ { k } ( \\tau ) , R ( \\tau ) , k ) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 435, + 505, + 449 + ], + "score": 1.0, + "content": ". These returns are", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 446, + 505, + 460 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 188, + 460 + ], + "score": 1.0, + "content": "normalized to keep", + "type": "text" + }, + { + "bbox": [ + 189, + 447, + 244, + 459 + ], + "score": 0.92, + "content": "R ( \\tau ) \\in [ 0 , 1 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 244, + 446, + 505, + 460 + ], + "score": 1.0, + "content": ". Sampling a high return trajectory amounts to conditioning on", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 107, + 456, + 505, + 471 + ], + "spans": [ + { + "bbox": [ + 107, + 457, + 150, + 470 + ], + "score": 0.92, + "content": "R ( \\tau ) = 1", + "type": "inline_equation" + }, + { + "bbox": [ + 150, + 456, + 315, + 471 + ], + "score": 1.0, + "content": ". Note that we do not make use of any", + "type": "text" + }, + { + "bbox": [ + 315, + 458, + 324, + 469 + ], + "score": 0.86, + "content": "Q", + "type": "inline_equation" + }, + { + "bbox": [ + 324, + 456, + 505, + 471 + ], + "score": 1.0, + "content": "-values, which would then require dynamic", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 469, + 166, + 483 + ], + "spans": [ + { + "bbox": [ + 105, + 469, + 166, + 483 + ], + "score": 1.0, + "content": "programming.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 30, + "bbox_fs": [ + 105, + 425, + 505, + 483 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 483, + 505, + 549 + ], + "lines": [ + { + "bbox": [ + 106, + 483, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 106, + 483, + 505, + 495 + ], + "score": 1.0, + "content": "Satisfying Constraints Trajectories may satisfy a variety of constraints, each represented by the", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 493, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 106, + 493, + 120, + 506 + ], + "score": 1.0, + "content": "set", + "type": "text" + }, + { + "bbox": [ + 120, + 494, + 130, + 505 + ], + "score": 0.88, + "content": "\\mathcal { C } _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 130, + 493, + 505, + 506 + ], + "score": 1.0, + "content": ", such as reaching a specific goal, visiting states in a particular order, or avoiding parts of the", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 505, + 505, + 518 + ], + "spans": [ + { + "bbox": [ + 106, + 505, + 361, + 518 + ], + "score": 1.0, + "content": "state space. To generate trajectories satisfying a given constraint", + "type": "text" + }, + { + "bbox": [ + 361, + 505, + 370, + 515 + ], + "score": 0.84, + "content": "\\mathcal { C } _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 371, + 505, + 505, + 518 + ], + "score": 1.0, + "content": ", we condition the noise model on", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 515, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 216, + 528 + ], + "score": 1.0, + "content": "a one-hot encoding so that", + "type": "text" + }, + { + "bbox": [ + 217, + 515, + 403, + 528 + ], + "score": 0.92, + "content": "\\begin{array} { r } { \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ( \\tau ) , \\bar { k } ) : = \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\mathbb { 1 } ( \\tau \\in \\mathcal { C } _ { i } ) , k ) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 403, + 515, + 505, + 528 + ], + "score": 1.0, + "content": ". Although we train with", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 527, + 506, + 539 + ], + "spans": [ + { + "bbox": [ + 106, + 527, + 506, + 539 + ], + "score": 1.0, + "content": "an offline dataset in which trajectories satisfy only one of the available constraints, at inference we", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 537, + 264, + 550 + ], + "spans": [ + { + "bbox": [ + 106, + 537, + 264, + 550 + ], + "score": 1.0, + "content": "can satisfy several constraints together.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 35.5, + "bbox_fs": [ + 105, + 483, + 506, + 550 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 551, + 505, + 596 + ], + "lines": [ + { + "bbox": [ + 106, + 551, + 505, + 564 + ], + "spans": [ + { + "bbox": [ + 106, + 551, + 229, + 564 + ], + "score": 1.0, + "content": "Composing Skills A skill", + "type": "text" + }, + { + "bbox": [ + 230, + 552, + 235, + 561 + ], + "score": 0.65, + "content": "i", + "type": "inline_equation" + }, + { + "bbox": [ + 235, + 551, + 433, + 564 + ], + "score": 1.0, + "content": "can be specified from a set of demonstrations", + "type": "text" + }, + { + "bbox": [ + 434, + 552, + 445, + 563 + ], + "score": 0.88, + "content": "B _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 445, + 551, + 505, + 564 + ], + "score": 1.0, + "content": ". To generate", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 562, + 505, + 575 + ], + "spans": [ + { + "bbox": [ + 105, + 562, + 505, + 575 + ], + "score": 1.0, + "content": "trajectories that demonstrate a given skill, we condition the noise model on a one-hot encoding so", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 573, + 505, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 573, + 124, + 586 + ], + "score": 1.0, + "content": "that", + "type": "text" + }, + { + "bbox": [ + 124, + 573, + 315, + 585 + ], + "score": 0.91, + "content": "\\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , { \\pmb y } ( \\tau ) , k ) : = \\epsilon _ { \\theta } ( \\bar { { \\pmb x } } _ { k } ( \\tau ) , \\mathbb { 1 } ( \\tau \\in \\mathcal { B } _ { i } ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 316, + 573, + 505, + 586 + ], + "score": 1.0, + "content": ". Although we train with individual skills, we", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 585, + 346, + 597 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 346, + 597 + ], + "score": 1.0, + "content": "may further compose these skills together during inference.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 40.5, + "bbox_fs": [ + 105, + 551, + 505, + 597 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 601, + 505, + 635 + ], + "lines": [ + { + "bbox": [ + 105, + 600, + 506, + 614 + ], + "spans": [ + { + "bbox": [ + 105, + 600, + 334, + 614 + ], + "score": 1.0, + "content": "Assuming we have learned the data distributions", + "type": "text" + }, + { + "bbox": [ + 334, + 600, + 486, + 614 + ], + "score": 0.91, + "content": "q ( { \\pmb x } _ { 0 } ( \\tau ) | { \\pmb y } ^ { 1 } ( \\tau ) ) , \\dots , q ( { \\pmb x } _ { 0 } ( \\tau ) | { \\pmb y } ^ { n } ( \\tau ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 486, + 600, + 506, + 614 + ], + "score": 1.0, + "content": "for", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 107, + 612, + 505, + 624 + ], + "spans": [ + { + "bbox": [ + 107, + 614, + 114, + 622 + ], + "score": 0.74, + "content": "n", + "type": "inline_equation" + }, + { + "bbox": [ + 114, + 612, + 505, + 624 + ], + "score": 1.0, + "content": "different conditioning variables, we can sample from the composed data distribution", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 107, + 622, + 396, + 636 + ], + "spans": [ + { + "bbox": [ + 107, + 623, + 218, + 636 + ], + "score": 0.91, + "content": "q ( \\pmb { x } _ { 0 } ( \\tau ) | \\pmb { y } ^ { 1 } ( \\tau ) , \\dots , \\pmb { y } ^ { n } ( \\bar { \\tau } ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 218, + 622, + 396, + 636 + ], + "score": 1.0, + "content": "using the perturbed noise (Liu et al., 2022):", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44, + "bbox_fs": [ + 105, + 600, + 506, + 636 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 164, + 640, + 445, + 673 + ], + "lines": [ + { + "bbox": [ + 164, + 640, + 445, + 673 + ], + "spans": [ + { + "bbox": [ + 164, + 640, + 445, + 673 + ], + "score": 0.94, + "content": "\\hat { \\epsilon } : = \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\emptyset , k ) + \\omega \\sum _ { i = 1 } ^ { n } ( \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ^ { i } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\emptyset , k ) )", + "type": "interline_equation", + "image_path": "387eca885b4e7a9ca4a4a7e4e308a743d03ad7f9ae1155a2384c8112667bea94.jpg" + } + ] + } + ], + "index": 47, + "virtual_lines": [ + { + "bbox": [ + 164, + 640, + 445, + 651.0 + ], + "spans": [], + "index": 46 + }, + { + "bbox": [ + 164, + 651.0, + 445, + 662.0 + ], + "spans": [], + "index": 47 + }, + { + "bbox": [ + 164, + 662.0, + 445, + 673.0 + ], + "spans": [], + "index": 48 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 678, + 506, + 735 + ], + "lines": [ + { + "bbox": [ + 104, + 677, + 508, + 695 + ], + "spans": [ + { + "bbox": [ + 104, + 677, + 213, + 695 + ], + "score": 1.0, + "content": "This property assumes that", + "type": "text" + }, + { + "bbox": [ + 214, + 678, + 260, + 692 + ], + "score": 0.93, + "content": "\\{ \\pmb { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 261, + 677, + 477, + 695 + ], + "score": 1.0, + "content": "are conditionally independent given the state trajectory", + "type": "text" + }, + { + "bbox": [ + 478, + 679, + 502, + 691 + ], + "score": 0.92, + "content": "\\pmb { x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 677, + 508, + 695 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 689, + 506, + 704 + ], + "spans": [ + { + "bbox": [ + 105, + 689, + 506, + 704 + ], + "score": 1.0, + "content": "However, we empirically observe that this assumption doesn’t have to be strictly satisfied as long as", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 702, + 505, + 713 + ], + "spans": [ + { + "bbox": [ + 106, + 702, + 505, + 713 + ], + "score": 1.0, + "content": "the composition of conditioning variables is feasible. For more detailed discussion, please refer to", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 106, + 712, + 506, + 726 + ], + "spans": [ + { + "bbox": [ + 106, + 712, + 506, + 726 + ], + "score": 1.0, + "content": "Appendix D. We use this property to compose more than one constraint or skill together at test-time.", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 105, + 722, + 501, + 736 + ], + "spans": [ + { + "bbox": [ + 105, + 722, + 501, + 736 + ], + "score": 1.0, + "content": "We also show how Decision Diffuser can avoid particular constraint or skill (NOT) in Appendix J.", + "type": "text" + } + ], + "index": 53 + } + ], + "index": 51, + "bbox_fs": [ + 104, + 677, + 508, + 736 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 145, + 91, + 344, + 100 + ], + "lines": [ + { + "bbox": [ + 144, + 90, + 345, + 103 + ], + "spans": [ + { + "bbox": [ + 144, + 90, + 345, + 103 + ], + "score": 1.0, + "content": "Algorithm 1 Conditional Planning with the Decision Diffuser", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "image", + "bbox": [ + 146, + 101, + 464, + 276 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 146, + 101, + 464, + 276 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 146, + 101, + 464, + 276 + ], + "spans": [ + { + "bbox": [ + 146, + 101, + 464, + 276 + ], + "score": 0.855, + "type": "image", + "image_path": "978ee750ffe38aa619ac11373331980f674e6924fad8991ee08cfb832205fc9a.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 146, + 101, + 464, + 159.33333333333334 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 146, + 159.33333333333334, + 464, + 217.66666666666669 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 146, + 217.66666666666669, + 464, + 276.0 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 276, + 505, + 316 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 275, + 505, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 275, + 505, + 287 + ], + "score": 1.0, + "content": "Figure 4: Results Overview. Decision Diffuser performs better than both TD learning (CQL) and Behavorial", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 285, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 285, + 505, + 298 + ], + "score": 1.0, + "content": "Cloning (BC) across D4RL locomotion tasks, D4RL Kitchen tasks and Kuka Block Stacking tasks (single", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 296, + 505, + 307 + ], + "spans": [ + { + "bbox": [ + 106, + 296, + 505, + 307 + ], + "score": 1.0, + "content": "constraint) using only a conditional generative modeling objective. For performance metric, we use normalized", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 305, + 505, + 318 + ], + "spans": [ + { + "bbox": [ + 105, + 305, + 505, + 318 + ], + "score": 1.0, + "content": "average returns (Fu et al., 2020) for D4RL tasks (Locomotion and Kitchen) and success rate for Block Stacking.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5.5 + } + ], + "index": 3.75 + }, + { + "type": "title", + "bbox": [ + 109, + 321, + 285, + 333 + ], + "lines": [ + { + "bbox": [ + 106, + 321, + 286, + 334 + ], + "spans": [ + { + "bbox": [ + 106, + 321, + 286, + 334 + ], + "score": 1.0, + "content": "3.4 TRAINING THE DECISION DIFFUSER", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 107, + 338, + 505, + 394 + ], + "lines": [ + { + "bbox": [ + 105, + 338, + 506, + 351 + ], + "spans": [ + { + "bbox": [ + 105, + 338, + 506, + 351 + ], + "score": 1.0, + "content": "The Decision Diffuser, our conditional generative model for decision-making, is trained in a", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 349, + 507, + 362 + ], + "spans": [ + { + "bbox": [ + 106, + 349, + 261, + 362 + ], + "score": 1.0, + "content": "supervised manner. Given a dataset", + "type": "text" + }, + { + "bbox": [ + 261, + 350, + 270, + 360 + ], + "score": 0.8, + "content": "\\mathcal { D }", + "type": "inline_equation" + }, + { + "bbox": [ + 271, + 349, + 507, + 362 + ], + "score": 1.0, + "content": "of trajectories, each labeled with the return it achieves,", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 361, + 506, + 373 + ], + "spans": [ + { + "bbox": [ + 106, + 361, + 506, + 373 + ], + "score": 1.0, + "content": "the constraint that it satisfies, or the skill that it demonstrates, we simultaneously train the reverse", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 370, + 504, + 386 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 176, + 386 + ], + "score": 1.0, + "content": "diffusion process", + "type": "text" + }, + { + "bbox": [ + 177, + 374, + 187, + 383 + ], + "score": 0.85, + "content": "p _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 188, + 370, + 348, + 386 + ], + "score": 1.0, + "content": ", parameterized through the noise model", + "type": "text" + }, + { + "bbox": [ + 349, + 373, + 358, + 383 + ], + "score": 0.85, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 359, + 370, + 492, + 386 + ], + "score": 1.0, + "content": ", and the inverse dynamics model", + "type": "text" + }, + { + "bbox": [ + 493, + 372, + 504, + 384 + ], + "score": 0.88, + "content": "f _ { \\phi }", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 382, + 203, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 382, + 203, + 396 + ], + "score": 1.0, + "content": "with the following loss:", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 107, + 396, + 506, + 460 + ], + "lines": [ + { + "bbox": [ + 106, + 396, + 506, + 412 + ], + "spans": [ + { + "bbox": [ + 106, + 397, + 111, + 412 + ], + "score": 0.0, + "content": "", + "type": "text" + }, + { + "bbox": [ + 111, + 396, + 506, + 412 + ], + "score": 0.89, + "content": "\\begin{array} { r } { \\natural ( \\theta , \\phi ) : = \\mathbb { E } _ { k , \\tau \\in \\mathcal { D } , \\beta \\sim \\mathrm { B e n } ( p ) } [ \\| \\epsilon - \\epsilon _ { \\theta } ( { x } _ { k } ( \\tau ) , ( 1 - \\beta ) y ( \\tau ) + \\beta \\mathcal { O } , k ) \\| ^ { 2 } ] + \\mathbb { E } _ { ( s , a , s ^ { \\prime } ) \\in \\mathcal { D } } [ \\| a - f _ { \\phi } ( s , s ^ { \\prime } ) \\| ^ { 2 } ] } \\end{array}", + "type": "inline_equation", + "image_path": "b010e505706cc7edf96a41a2186d23368bdbfa71d912b0c59fe8316c81fe6c0b.jpg" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 415, + 505, + 428 + ], + "spans": [ + { + "bbox": [ + 105, + 415, + 182, + 428 + ], + "score": 1.0, + "content": "For each trajectory", + "type": "text" + }, + { + "bbox": [ + 182, + 418, + 189, + 426 + ], + "score": 0.77, + "content": "\\tau", + "type": "inline_equation" + }, + { + "bbox": [ + 189, + 415, + 277, + 428 + ], + "score": 1.0, + "content": ", we first sample noise", + "type": "text" + }, + { + "bbox": [ + 277, + 415, + 329, + 428 + ], + "score": 0.93, + "content": "\\epsilon \\sim \\mathcal { N } ( 0 , I )", + "type": "inline_equation" + }, + { + "bbox": [ + 329, + 415, + 388, + 428 + ], + "score": 1.0, + "content": "and a timestep", + "type": "text" + }, + { + "bbox": [ + 389, + 415, + 462, + 427 + ], + "score": 0.9, + "content": "k \\sim \\mathcal { U } \\{ 1 , \\dots , K \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 462, + 415, + 505, + 428 + ], + "score": 1.0, + "content": ". Then, we", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 426, + 505, + 439 + ], + "spans": [ + { + "bbox": [ + 106, + 426, + 231, + 439 + ], + "score": 1.0, + "content": "construct a noisy array of states", + "type": "text" + }, + { + "bbox": [ + 232, + 426, + 257, + 438 + ], + "score": 0.93, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 258, + 426, + 380, + 439 + ], + "score": 1.0, + "content": "and finally predict the noise as", + "type": "text" + }, + { + "bbox": [ + 380, + 427, + 479, + 438 + ], + "score": 0.9, + "content": "\\hat { \\epsilon } _ { \\boldsymbol { \\theta } } : = \\epsilon _ { \\boldsymbol { \\theta } } ( \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ( \\tau ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 426, + 505, + 439 + ], + "score": 1.0, + "content": ". Note", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 436, + 506, + 450 + ], + "spans": [ + { + "bbox": [ + 105, + 436, + 191, + 450 + ], + "score": 1.0, + "content": "that with probability", + "type": "text" + }, + { + "bbox": [ + 192, + 439, + 198, + 449 + ], + "score": 0.79, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 199, + 436, + 506, + 450 + ], + "score": 1.0, + "content": "we ignore the conditioning information and the inverse dynamics is trained", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 447, + 307, + 461 + ], + "spans": [ + { + "bbox": [ + 105, + 447, + 307, + 461 + ], + "score": 1.0, + "content": "with individual transitions rather than trajectories.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 106, + 462, + 505, + 540 + ], + "lines": [ + { + "bbox": [ + 105, + 461, + 505, + 475 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 240, + 475 + ], + "score": 1.0, + "content": "Architecture We parameterize", + "type": "text" + }, + { + "bbox": [ + 240, + 464, + 250, + 473 + ], + "score": 0.85, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 250, + 461, + 505, + 475 + ], + "score": 1.0, + "content": "with a temporal U-Net architecture, a neural network consisting", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 473, + 506, + 486 + ], + "spans": [ + { + "bbox": [ + 106, + 473, + 506, + 486 + ], + "score": 1.0, + "content": "of repeated convolutional residual blocks (Janner et al., 2022). This effectively treats a sequence of", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 484, + 505, + 497 + ], + "spans": [ + { + "bbox": [ + 106, + 484, + 131, + 497 + ], + "score": 1.0, + "content": "states", + "type": "text" + }, + { + "bbox": [ + 132, + 484, + 157, + 496 + ], + "score": 0.93, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 158, + 484, + 505, + 497 + ], + "score": 1.0, + "content": "as an image where the height represents the dimension of a single state and the width", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 495, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 414, + 508 + ], + "score": 1.0, + "content": "denotes the length of the trajectory. We encode the conditioning information", + "type": "text" + }, + { + "bbox": [ + 415, + 495, + 435, + 507 + ], + "score": 0.92, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 495, + 505, + 508 + ], + "score": 1.0, + "content": "as either a scalar", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 104, + 505, + 506, + 520 + ], + "spans": [ + { + "bbox": [ + 104, + 505, + 323, + 520 + ], + "score": 1.0, + "content": "or a one-hot vector and project it into a latent variable", + "type": "text" + }, + { + "bbox": [ + 324, + 505, + 354, + 516 + ], + "score": 0.92, + "content": "z \\in \\mathbb { R } ^ { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 354, + 505, + 506, + 520 + ], + "score": 1.0, + "content": "with a multi-layer perceptron (MLP).", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 516, + 506, + 530 + ], + "spans": [ + { + "bbox": [ + 105, + 516, + 132, + 530 + ], + "score": 1.0, + "content": "When", + "type": "text" + }, + { + "bbox": [ + 132, + 517, + 173, + 529 + ], + "score": 0.92, + "content": "\\pmb { y } ( \\tau ) = 0", + "type": "inline_equation" + }, + { + "bbox": [ + 174, + 516, + 279, + 530 + ], + "score": 1.0, + "content": ", we zero out the entries of", + "type": "text" + }, + { + "bbox": [ + 279, + 519, + 285, + 527 + ], + "score": 0.76, + "content": "z", + "type": "inline_equation" + }, + { + "bbox": [ + 286, + 516, + 461, + 530 + ], + "score": 1.0, + "content": ". We also parameterize the inverse dynamics", + "type": "text" + }, + { + "bbox": [ + 461, + 517, + 472, + 529 + ], + "score": 0.89, + "content": "f _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 472, + 516, + 506, + 530 + ], + "score": 1.0, + "content": "with an", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 528, + 370, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 370, + 541 + ], + "score": 1.0, + "content": "MLP. For implementation details, please refer to the Appendix B.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 106, + 542, + 505, + 597 + ], + "lines": [ + { + "bbox": [ + 105, + 541, + 506, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 541, + 462, + 556 + ], + "score": 1.0, + "content": "Low-temperature Sampling In the denoising step of Algorithm 1, we compute", + "type": "text" + }, + { + "bbox": [ + 462, + 544, + 485, + 554 + ], + "score": 0.87, + "content": "\\mu _ { k - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 485, + 541, + 506, + 556 + ], + "score": 1.0, + "content": "and", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 551, + 504, + 567 + ], + "spans": [ + { + "bbox": [ + 106, + 553, + 129, + 564 + ], + "score": 0.88, + "content": "\\Sigma _ { k - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 130, + 551, + 468, + 567 + ], + "score": 1.0, + "content": "from a noisy sequence of states and a predicted noise. We find that sampling", + "type": "text" + }, + { + "bbox": [ + 469, + 555, + 504, + 565 + ], + "score": 0.88, + "content": "x _ { k - 1 } \\sim", + "type": "inline_equation" + } + ], + "index": 27 + }, + { + "bbox": [ + 107, + 563, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 107, + 564, + 180, + 576 + ], + "score": 0.91, + "content": "\\mathcal { N } ( \\mu _ { k - 1 } , \\alpha \\Sigma _ { k - 1 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 180, + 563, + 318, + 577 + ], + "score": 1.0, + "content": "where the variance is scaled by", + "type": "text" + }, + { + "bbox": [ + 318, + 564, + 363, + 576 + ], + "score": 0.92, + "content": "\\alpha \\in [ 0 , 1 )", + "type": "inline_equation" + }, + { + "bbox": [ + 363, + 563, + 506, + 577 + ], + "score": 1.0, + "content": "leads to better quality sequences", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 575, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 575, + 505, + 588 + ], + "score": 1.0, + "content": "(corresponding to sampling lower temperature samples). For a proper ablation study, please refer to", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 104, + 586, + 157, + 598 + ], + "spans": [ + { + "bbox": [ + 104, + 586, + 157, + 598 + ], + "score": 1.0, + "content": "Appendix C.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 28 + }, + { + "type": "title", + "bbox": [ + 107, + 603, + 201, + 615 + ], + "lines": [ + { + "bbox": [ + 105, + 602, + 201, + 617 + ], + "spans": [ + { + "bbox": [ + 105, + 602, + 201, + 617 + ], + "score": 1.0, + "content": "4 EXPERIMENTS", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 107, + 623, + 505, + 689 + ], + "lines": [ + { + "bbox": [ + 106, + 623, + 504, + 635 + ], + "spans": [ + { + "bbox": [ + 106, + 623, + 504, + 635 + ], + "score": 1.0, + "content": "In this section, we explore the efficacy of the Decision Diffuser on a variety of decision-making tasks", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 634, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 634, + 506, + 646 + ], + "score": 1.0, + "content": "(performance illustrated in Figure 4). In particular, we evaluate (1) the ability to recover effective RL", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 645, + 506, + 658 + ], + "spans": [ + { + "bbox": [ + 105, + 645, + 506, + 658 + ], + "score": 1.0, + "content": "policies from offline data, (2) the ability to generate behavior that satisfies multiple sets of constraints,", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 657, + 505, + 668 + ], + "spans": [ + { + "bbox": [ + 106, + 657, + 505, + 668 + ], + "score": 1.0, + "content": "(3) the ability compose multiple different skills together. In addition, we empirically justify use of", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 666, + 506, + 680 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 506, + 680 + ], + "score": 1.0, + "content": "classifier-free guidance, low-temperature sampling (Appendix C), and inverse dynamics (Appendix F)", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 677, + 436, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 436, + 691 + ], + "score": 1.0, + "content": "and test the robustness of Decision Diffuser to stochastic dynamics (Appendix G).", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 34.5 + }, + { + "type": "title", + "bbox": [ + 107, + 692, + 294, + 703 + ], + "lines": [ + { + "bbox": [ + 106, + 692, + 294, + 704 + ], + "spans": [ + { + "bbox": [ + 106, + 692, + 294, + 704 + ], + "score": 1.0, + "content": "4.1 OFFLINE REINFORCEMENT LEARNING", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 104, + 708, + 507, + 723 + ], + "spans": [ + { + "bbox": [ + 104, + 708, + 507, + 723 + ], + "score": 1.0, + "content": "Setup We first test whether the Decision Diffuser can generate return-maximizing trajectories.", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "To test this, we train a state diffusion process and inverse dynamics model on publicly available", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5 + } + ], + "page_idx": 5, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "score": 1.0, + "content": "6", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 145, + 91, + 344, + 100 + ], + "lines": [ + { + "bbox": [ + 144, + 90, + 345, + 103 + ], + "spans": [ + { + "bbox": [ + 144, + 90, + 345, + 103 + ], + "score": 1.0, + "content": "Algorithm 1 Conditional Planning with the Decision Diffuser", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "image", + "bbox": [ + 146, + 101, + 464, + 276 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 146, + 101, + 464, + 276 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 146, + 101, + 464, + 276 + ], + "spans": [ + { + "bbox": [ + 146, + 101, + 464, + 276 + ], + "score": 0.855, + "type": "image", + "image_path": "978ee750ffe38aa619ac11373331980f674e6924fad8991ee08cfb832205fc9a.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 146, + 101, + 464, + 159.33333333333334 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 146, + 159.33333333333334, + 464, + 217.66666666666669 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 146, + 217.66666666666669, + 464, + 276.0 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 276, + 505, + 316 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 275, + 505, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 275, + 505, + 287 + ], + "score": 1.0, + "content": "Figure 4: Results Overview. Decision Diffuser performs better than both TD learning (CQL) and Behavorial", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 285, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 285, + 505, + 298 + ], + "score": 1.0, + "content": "Cloning (BC) across D4RL locomotion tasks, D4RL Kitchen tasks and Kuka Block Stacking tasks (single", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 296, + 505, + 307 + ], + "spans": [ + { + "bbox": [ + 106, + 296, + 505, + 307 + ], + "score": 1.0, + "content": "constraint) using only a conditional generative modeling objective. For performance metric, we use normalized", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 305, + 505, + 318 + ], + "spans": [ + { + "bbox": [ + 105, + 305, + 505, + 318 + ], + "score": 1.0, + "content": "average returns (Fu et al., 2020) for D4RL tasks (Locomotion and Kitchen) and success rate for Block Stacking.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5.5 + } + ], + "index": 3.75 + }, + { + "type": "title", + "bbox": [ + 109, + 321, + 285, + 333 + ], + "lines": [ + { + "bbox": [ + 106, + 321, + 286, + 334 + ], + "spans": [ + { + "bbox": [ + 106, + 321, + 286, + 334 + ], + "score": 1.0, + "content": "3.4 TRAINING THE DECISION DIFFUSER", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 107, + 338, + 505, + 394 + ], + "lines": [ + { + "bbox": [ + 105, + 338, + 506, + 351 + ], + "spans": [ + { + "bbox": [ + 105, + 338, + 506, + 351 + ], + "score": 1.0, + "content": "The Decision Diffuser, our conditional generative model for decision-making, is trained in a", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 349, + 507, + 362 + ], + "spans": [ + { + "bbox": [ + 106, + 349, + 261, + 362 + ], + "score": 1.0, + "content": "supervised manner. Given a dataset", + "type": "text" + }, + { + "bbox": [ + 261, + 350, + 270, + 360 + ], + "score": 0.8, + "content": "\\mathcal { D }", + "type": "inline_equation" + }, + { + "bbox": [ + 271, + 349, + 507, + 362 + ], + "score": 1.0, + "content": "of trajectories, each labeled with the return it achieves,", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 361, + 506, + 373 + ], + "spans": [ + { + "bbox": [ + 106, + 361, + 506, + 373 + ], + "score": 1.0, + "content": "the constraint that it satisfies, or the skill that it demonstrates, we simultaneously train the reverse", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 370, + 504, + 386 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 176, + 386 + ], + "score": 1.0, + "content": "diffusion process", + "type": "text" + }, + { + "bbox": [ + 177, + 374, + 187, + 383 + ], + "score": 0.85, + "content": "p _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 188, + 370, + 348, + 386 + ], + "score": 1.0, + "content": ", parameterized through the noise model", + "type": "text" + }, + { + "bbox": [ + 349, + 373, + 358, + 383 + ], + "score": 0.85, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 359, + 370, + 492, + 386 + ], + "score": 1.0, + "content": ", and the inverse dynamics model", + "type": "text" + }, + { + "bbox": [ + 493, + 372, + 504, + 384 + ], + "score": 0.88, + "content": "f _ { \\phi }", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 382, + 203, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 382, + 203, + 396 + ], + "score": 1.0, + "content": "with the following loss:", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 338, + 507, + 396 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 396, + 506, + 460 + ], + "lines": [ + { + "bbox": [ + 106, + 396, + 506, + 412 + ], + "spans": [ + { + "bbox": [ + 106, + 397, + 111, + 412 + ], + "score": 0.0, + "content": "", + "type": "text" + }, + { + "bbox": [ + 111, + 396, + 506, + 412 + ], + "score": 0.89, + "content": "\\begin{array} { r } { \\natural ( \\theta , \\phi ) : = \\mathbb { E } _ { k , \\tau \\in \\mathcal { D } , \\beta \\sim \\mathrm { B e n } ( p ) } [ \\| \\epsilon - \\epsilon _ { \\theta } ( { x } _ { k } ( \\tau ) , ( 1 - \\beta ) y ( \\tau ) + \\beta \\mathcal { O } , k ) \\| ^ { 2 } ] + \\mathbb { E } _ { ( s , a , s ^ { \\prime } ) \\in \\mathcal { D } } [ \\| a - f _ { \\phi } ( s , s ^ { \\prime } ) \\| ^ { 2 } ] } \\end{array}", + "type": "inline_equation", + "image_path": "b010e505706cc7edf96a41a2186d23368bdbfa71d912b0c59fe8316c81fe6c0b.jpg" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 415, + 505, + 428 + ], + "spans": [ + { + "bbox": [ + 105, + 415, + 182, + 428 + ], + "score": 1.0, + "content": "For each trajectory", + "type": "text" + }, + { + "bbox": [ + 182, + 418, + 189, + 426 + ], + "score": 0.77, + "content": "\\tau", + "type": "inline_equation" + }, + { + "bbox": [ + 189, + 415, + 277, + 428 + ], + "score": 1.0, + "content": ", we first sample noise", + "type": "text" + }, + { + "bbox": [ + 277, + 415, + 329, + 428 + ], + "score": 0.93, + "content": "\\epsilon \\sim \\mathcal { N } ( 0 , I )", + "type": "inline_equation" + }, + { + "bbox": [ + 329, + 415, + 388, + 428 + ], + "score": 1.0, + "content": "and a timestep", + "type": "text" + }, + { + "bbox": [ + 389, + 415, + 462, + 427 + ], + "score": 0.9, + "content": "k \\sim \\mathcal { U } \\{ 1 , \\dots , K \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 462, + 415, + 505, + 428 + ], + "score": 1.0, + "content": ". Then, we", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 426, + 505, + 439 + ], + "spans": [ + { + "bbox": [ + 106, + 426, + 231, + 439 + ], + "score": 1.0, + "content": "construct a noisy array of states", + "type": "text" + }, + { + "bbox": [ + 232, + 426, + 257, + 438 + ], + "score": 0.93, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 258, + 426, + 380, + 439 + ], + "score": 1.0, + "content": "and finally predict the noise as", + "type": "text" + }, + { + "bbox": [ + 380, + 427, + 479, + 438 + ], + "score": 0.9, + "content": "\\hat { \\epsilon } _ { \\boldsymbol { \\theta } } : = \\epsilon _ { \\boldsymbol { \\theta } } ( \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ( \\tau ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 426, + 505, + 439 + ], + "score": 1.0, + "content": ". Note", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 436, + 506, + 450 + ], + "spans": [ + { + "bbox": [ + 105, + 436, + 191, + 450 + ], + "score": 1.0, + "content": "that with probability", + "type": "text" + }, + { + "bbox": [ + 192, + 439, + 198, + 449 + ], + "score": 0.79, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 199, + 436, + 506, + 450 + ], + "score": 1.0, + "content": "we ignore the conditioning information and the inverse dynamics is trained", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 447, + 307, + 461 + ], + "spans": [ + { + "bbox": [ + 105, + 447, + 307, + 461 + ], + "score": 1.0, + "content": "with individual transitions rather than trajectories.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 16, + "bbox_fs": [ + 105, + 396, + 506, + 461 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 462, + 505, + 540 + ], + "lines": [ + { + "bbox": [ + 105, + 461, + 505, + 475 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 240, + 475 + ], + "score": 1.0, + "content": "Architecture We parameterize", + "type": "text" + }, + { + "bbox": [ + 240, + 464, + 250, + 473 + ], + "score": 0.85, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 250, + 461, + 505, + 475 + ], + "score": 1.0, + "content": "with a temporal U-Net architecture, a neural network consisting", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 473, + 506, + 486 + ], + "spans": [ + { + "bbox": [ + 106, + 473, + 506, + 486 + ], + "score": 1.0, + "content": "of repeated convolutional residual blocks (Janner et al., 2022). This effectively treats a sequence of", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 484, + 505, + 497 + ], + "spans": [ + { + "bbox": [ + 106, + 484, + 131, + 497 + ], + "score": 1.0, + "content": "states", + "type": "text" + }, + { + "bbox": [ + 132, + 484, + 157, + 496 + ], + "score": 0.93, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 158, + 484, + 505, + 497 + ], + "score": 1.0, + "content": "as an image where the height represents the dimension of a single state and the width", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 495, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 414, + 508 + ], + "score": 1.0, + "content": "denotes the length of the trajectory. We encode the conditioning information", + "type": "text" + }, + { + "bbox": [ + 415, + 495, + 435, + 507 + ], + "score": 0.92, + "content": "\\pmb { y } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 495, + 505, + 508 + ], + "score": 1.0, + "content": "as either a scalar", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 104, + 505, + 506, + 520 + ], + "spans": [ + { + "bbox": [ + 104, + 505, + 323, + 520 + ], + "score": 1.0, + "content": "or a one-hot vector and project it into a latent variable", + "type": "text" + }, + { + "bbox": [ + 324, + 505, + 354, + 516 + ], + "score": 0.92, + "content": "z \\in \\mathbb { R } ^ { h }", + "type": "inline_equation" + }, + { + "bbox": [ + 354, + 505, + 506, + 520 + ], + "score": 1.0, + "content": "with a multi-layer perceptron (MLP).", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 516, + 506, + 530 + ], + "spans": [ + { + "bbox": [ + 105, + 516, + 132, + 530 + ], + "score": 1.0, + "content": "When", + "type": "text" + }, + { + "bbox": [ + 132, + 517, + 173, + 529 + ], + "score": 0.92, + "content": "\\pmb { y } ( \\tau ) = 0", + "type": "inline_equation" + }, + { + "bbox": [ + 174, + 516, + 279, + 530 + ], + "score": 1.0, + "content": ", we zero out the entries of", + "type": "text" + }, + { + "bbox": [ + 279, + 519, + 285, + 527 + ], + "score": 0.76, + "content": "z", + "type": "inline_equation" + }, + { + "bbox": [ + 286, + 516, + 461, + 530 + ], + "score": 1.0, + "content": ". We also parameterize the inverse dynamics", + "type": "text" + }, + { + "bbox": [ + 461, + 517, + 472, + 529 + ], + "score": 0.89, + "content": "f _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 472, + 516, + 506, + 530 + ], + "score": 1.0, + "content": "with an", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 528, + 370, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 370, + 541 + ], + "score": 1.0, + "content": "MLP. For implementation details, please refer to the Appendix B.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 22, + "bbox_fs": [ + 104, + 461, + 506, + 541 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 542, + 505, + 597 + ], + "lines": [ + { + "bbox": [ + 105, + 541, + 506, + 556 + ], + "spans": [ + { + "bbox": [ + 105, + 541, + 462, + 556 + ], + "score": 1.0, + "content": "Low-temperature Sampling In the denoising step of Algorithm 1, we compute", + "type": "text" + }, + { + "bbox": [ + 462, + 544, + 485, + 554 + ], + "score": 0.87, + "content": "\\mu _ { k - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 485, + 541, + 506, + 556 + ], + "score": 1.0, + "content": "and", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 551, + 504, + 567 + ], + "spans": [ + { + "bbox": [ + 106, + 553, + 129, + 564 + ], + "score": 0.88, + "content": "\\Sigma _ { k - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 130, + 551, + 468, + 567 + ], + "score": 1.0, + "content": "from a noisy sequence of states and a predicted noise. We find that sampling", + "type": "text" + }, + { + "bbox": [ + 469, + 555, + 504, + 565 + ], + "score": 0.88, + "content": "x _ { k - 1 } \\sim", + "type": "inline_equation" + } + ], + "index": 27 + }, + { + "bbox": [ + 107, + 563, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 107, + 564, + 180, + 576 + ], + "score": 0.91, + "content": "\\mathcal { N } ( \\mu _ { k - 1 } , \\alpha \\Sigma _ { k - 1 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 180, + 563, + 318, + 577 + ], + "score": 1.0, + "content": "where the variance is scaled by", + "type": "text" + }, + { + "bbox": [ + 318, + 564, + 363, + 576 + ], + "score": 0.92, + "content": "\\alpha \\in [ 0 , 1 )", + "type": "inline_equation" + }, + { + "bbox": [ + 363, + 563, + 506, + 577 + ], + "score": 1.0, + "content": "leads to better quality sequences", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 575, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 575, + 505, + 588 + ], + "score": 1.0, + "content": "(corresponding to sampling lower temperature samples). For a proper ablation study, please refer to", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 104, + 586, + 157, + 598 + ], + "spans": [ + { + "bbox": [ + 104, + 586, + 157, + 598 + ], + "score": 1.0, + "content": "Appendix C.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 28, + "bbox_fs": [ + 104, + 541, + 506, + 598 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 603, + 201, + 615 + ], + "lines": [ + { + "bbox": [ + 105, + 602, + 201, + 617 + ], + "spans": [ + { + "bbox": [ + 105, + 602, + 201, + 617 + ], + "score": 1.0, + "content": "4 EXPERIMENTS", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 107, + 623, + 505, + 689 + ], + "lines": [ + { + "bbox": [ + 106, + 623, + 504, + 635 + ], + "spans": [ + { + "bbox": [ + 106, + 623, + 504, + 635 + ], + "score": 1.0, + "content": "In this section, we explore the efficacy of the Decision Diffuser on a variety of decision-making tasks", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 634, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 634, + 506, + 646 + ], + "score": 1.0, + "content": "(performance illustrated in Figure 4). In particular, we evaluate (1) the ability to recover effective RL", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 645, + 506, + 658 + ], + "spans": [ + { + "bbox": [ + 105, + 645, + 506, + 658 + ], + "score": 1.0, + "content": "policies from offline data, (2) the ability to generate behavior that satisfies multiple sets of constraints,", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 657, + 505, + 668 + ], + "spans": [ + { + "bbox": [ + 106, + 657, + 505, + 668 + ], + "score": 1.0, + "content": "(3) the ability compose multiple different skills together. In addition, we empirically justify use of", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 666, + 506, + 680 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 506, + 680 + ], + "score": 1.0, + "content": "classifier-free guidance, low-temperature sampling (Appendix C), and inverse dynamics (Appendix F)", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 677, + 436, + 691 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 436, + 691 + ], + "score": 1.0, + "content": "and test the robustness of Decision Diffuser to stochastic dynamics (Appendix G).", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 34.5, + "bbox_fs": [ + 105, + 623, + 506, + 691 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 692, + 294, + 703 + ], + "lines": [ + { + "bbox": [ + 106, + 692, + 294, + 704 + ], + "spans": [ + { + "bbox": [ + 106, + 692, + 294, + 704 + ], + "score": 1.0, + "content": "4.1 OFFLINE REINFORCEMENT LEARNING", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 104, + 708, + 507, + 723 + ], + "spans": [ + { + "bbox": [ + 104, + 708, + 507, + 723 + ], + "score": 1.0, + "content": "Setup We first test whether the Decision Diffuser can generate return-maximizing trajectories.", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "To test this, we train a state diffusion process and inverse dynamics model on publicly available", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 39.5, + "bbox_fs": [ + 104, + 708, + 507, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 138 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 507, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 507, + 95 + ], + "score": 1.0, + "content": "D4RL datasets (Fu et al., 2020). We compare with existing offline RL methods, including model-", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 93, + 505, + 105 + ], + "spans": [ + { + "bbox": [ + 106, + 93, + 505, + 105 + ], + "score": 1.0, + "content": "free algorithms like CQL (Kumar et al., 2020) and IQL (Kostrikov et al., 2022), and model-based", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 506, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 506, + 117 + ], + "score": 1.0, + "content": "algorithms such as trajectory transformer (TT, Janner et al. (2021)) and MoReL (Kidambi et al., 2020).", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 115, + 506, + 127 + ], + "spans": [ + { + "bbox": [ + 106, + 115, + 506, + 127 + ], + "score": 1.0, + "content": "We also compare with sequence-models like the Decision Transformer (DT) (Chen et al. (2021) and", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 126, + 312, + 138 + ], + "spans": [ + { + "bbox": [ + 105, + 126, + 312, + 138 + ], + "score": 1.0, + "content": "diffusion models like Diffuser (Janner et al., 2022).", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 2 + }, + { + "type": "table", + "bbox": [ + 145, + 150, + 465, + 292 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 145, + 150, + 465, + 292 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 145, + 150, + 465, + 292 + ], + "spans": [ + { + "bbox": [ + 145, + 150, + 465, + 292 + ], + "score": 0.977, + "html": "
DatasetEnvironmentBCCQLIQLDTTTMOReLDiffuserDD
Med-ExpertHalfCheetah55.291.686.786.89553.379.890.6 ±1.3
Med-ExpertHopper52.5105.491.5107.6110.0108.7107.2111.8 ±1.8
Med-ExpertWalker2d107.5108.8109.6108.1101.995.6108.4108.8 ±1.7
MediumHalfCheetah42.644.047.442.646.942.144.249.1 ±1.0
MediumHopper52.958.566.367.661.195.458.579.3 ±3.6
MediumWalker2d75.372.578.374.07977.879.782.5±1.4
Med-ReplayHalfCheetah36.645.544.236.641.940.242.239.3 ±4.1
Med-ReplayHopper18.19594.782.791.593.696.8100±0.7
Med-ReplayWalker2d26.077.273.966.682.649.861.275±4.3
Average51.977.67774.778.972.975.381.8
MixedKitchen51.552.4511、1165 ±2.8
PartialKitchen3850.146.3--==57 ±2.5
Average44.851.248.7-1--61
", + "type": "table", + "image_path": "0a2d0b726961f161d351afaa00b88a1d4c6293180230e7ffd5e79853da686c2d.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 145, + 150, + 465, + 197.33333333333334 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 145, + 197.33333333333334, + 465, + 244.66666666666669 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 145, + 244.66666666666669, + 465, + 292.0 + ], + "spans": [], + "index": 7 + } + ] + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 107, + 343, + 505, + 399 + ], + "lines": [ + { + "bbox": [ + 105, + 342, + 505, + 357 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 505, + 357 + ], + "score": 1.0, + "content": "Results Across different offline RL tasks, we find that the Decision Diffuser is either competitive or", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 353, + 505, + 367 + ], + "spans": [ + { + "bbox": [ + 105, + 353, + 505, + 367 + ], + "score": 1.0, + "content": "outperforms many offline RL baselines (Table 1). It also outperforms Diffuser and sequence modeling", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 365, + 506, + 379 + ], + "spans": [ + { + "bbox": [ + 105, + 365, + 506, + 379 + ], + "score": 1.0, + "content": "approaches, such as Decision Transformer and Trajectory Transformer. The difference between", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 376, + 505, + 389 + ], + "spans": [ + { + "bbox": [ + 106, + 376, + 505, + 389 + ], + "score": 1.0, + "content": "Decision Diffuser and other methods becomes even more significant on harder D4RL Kitchen tasks", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 387, + 280, + 401 + ], + "spans": [ + { + "bbox": [ + 105, + 387, + 280, + 401 + ], + "score": 1.0, + "content": "which require long-term credit assignment.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 10 + }, + { + "type": "text", + "bbox": [ + 106, + 404, + 505, + 547 + ], + "lines": [ + { + "bbox": [ + 106, + 405, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 106, + 405, + 505, + 416 + ], + "score": 1.0, + "content": "To convey the importance of classifier-free guidance, we also compare with the baseline", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 415, + 506, + 427 + ], + "spans": [ + { + "bbox": [ + 106, + 415, + 506, + 427 + ], + "score": 1.0, + "content": "CondDiffuser, which diffuses over both state and action sequences as in Diffuser without classifier-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 425, + 506, + 439 + ], + "spans": [ + { + "bbox": [ + 105, + 425, + 506, + 439 + ], + "score": 1.0, + "content": "guidance. In Table 2, we observe that CondDiffuser improves over Diffuser in 2 out of 3", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 437, + 505, + 450 + ], + "spans": [ + { + "bbox": [ + 105, + 437, + 505, + 450 + ], + "score": 1.0, + "content": "environments. Decision Diffuser further improves over CondDiffuser, performing better across all", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 104, + 446, + 506, + 462 + ], + "spans": [ + { + "bbox": [ + 104, + 446, + 506, + 462 + ], + "score": 1.0, + "content": "3 environments. We conclude that learning the inverse dynamics is a good alternative to diffusing", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 460, + 505, + 471 + ], + "spans": [ + { + "bbox": [ + 106, + 460, + 505, + 471 + ], + "score": 1.0, + "content": "over actions. We further empirically analyze when to use inverse dynamics and when to diffuse", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 470, + 506, + 483 + ], + "spans": [ + { + "bbox": [ + 105, + 470, + 506, + 483 + ], + "score": 1.0, + "content": "over actions in Appendix F. We also compare against CondMLPDiffuser, a policy where the current", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 480, + 505, + 494 + ], + "spans": [ + { + "bbox": [ + 105, + 480, + 505, + 494 + ], + "score": 1.0, + "content": "action is denoised according to a diffusion process conditioned on both the state and return. We see", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 491, + 505, + 505 + ], + "spans": [ + { + "bbox": [ + 105, + 491, + 505, + 505 + ], + "score": 1.0, + "content": "that CondMLPDiffuser performs the worst amongst diffusion models. Till now, we mainly tested on", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 501, + 506, + 517 + ], + "spans": [ + { + "bbox": [ + 105, + 501, + 506, + 517 + ], + "score": 1.0, + "content": "offline RL tasks that have deterministic (or near deterministic) environment dynamics. Hence, we", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 514, + 505, + 526 + ], + "spans": [ + { + "bbox": [ + 106, + 514, + 505, + 526 + ], + "score": 1.0, + "content": "test the robustness of Decision Diffuser to stochastic dynamics and compare it to Diffuser and CQL as", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 524, + 505, + 539 + ], + "spans": [ + { + "bbox": [ + 105, + 524, + 505, + 539 + ], + "score": 1.0, + "content": "we vary the stochasticity in environment dynamics, in Appendix G. Finally, we analyze the runtime", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 536, + 313, + 548 + ], + "spans": [ + { + "bbox": [ + 105, + 536, + 313, + 548 + ], + "score": 1.0, + "content": "characteristics of Decision Diffuser in Appendix E.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 19 + }, + { + "type": "title", + "bbox": [ + 108, + 548, + 253, + 559 + ], + "lines": [ + { + "bbox": [ + 105, + 546, + 255, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 546, + 255, + 560 + ], + "score": 1.0, + "content": "4.2 CONSTRAINT SATISFACTION", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 26 + }, + { + "type": "text", + "bbox": [ + 106, + 564, + 505, + 674 + ], + "lines": [ + { + "bbox": [ + 105, + 561, + 506, + 578 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 506, + 578 + ], + "score": 1.0, + "content": "Setup We next evaluate how well we can generate trajectories that satisfy a set of constraints using", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 574, + 506, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 574, + 506, + 588 + ], + "score": 1.0, + "content": "the Kuka Block Stacking environment (Janner et al., 2022) visualized in Figure 5. In this domain,", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 585, + 506, + 598 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 506, + 598 + ], + "score": 1.0, + "content": "there are four blocks which can be stacked as a single tower or rearranged into several towers. A", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 597, + 506, + 610 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 227, + 610 + ], + "score": 1.0, + "content": "constraint like BlockHeigh", + "type": "text" + }, + { + "bbox": [ + 228, + 597, + 334, + 609 + ], + "score": 0.3, + "content": "\\mathsf { \\Lambda } = ( i ) > \\mathsf { B l o c k H e i g h t } ( j )", + "type": "inline_equation" + }, + { + "bbox": [ + 334, + 597, + 410, + 610 + ], + "score": 1.0, + "content": "requires that block", + "type": "text" + }, + { + "bbox": [ + 410, + 598, + 415, + 607 + ], + "score": 0.61, + "content": "i", + "type": "inline_equation" + }, + { + "bbox": [ + 415, + 597, + 506, + 610 + ], + "score": 1.0, + "content": "be placed above block", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 607, + 506, + 621 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 112, + 619 + ], + "score": 0.63, + "content": "j", + "type": "inline_equation" + }, + { + "bbox": [ + 113, + 607, + 506, + 621 + ], + "score": 1.0, + "content": ". We train the Decision Diffuser from 10, 000 expert demonstrations each satisfying one of these", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 618, + 505, + 632 + ], + "spans": [ + { + "bbox": [ + 105, + 618, + 505, + 632 + ], + "score": 1.0, + "content": "constraints. We randomize the positions of these blocks and consider two tasks at inference: sampling", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 630, + 505, + 641 + ], + "spans": [ + { + "bbox": [ + 106, + 630, + 505, + 641 + ], + "score": 1.0, + "content": "trajectories that satisfy a single constraint seen before in the dataset or satisfy a group of constraints", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "score": 1.0, + "content": "for which demonstrations were never provided. In the latter, we ask the Decision Diffuser to generate", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 651, + 506, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 651, + 229, + 664 + ], + "score": 1.0, + "content": "trajectories so BlockHeight", + "type": "text" + }, + { + "bbox": [ + 230, + 651, + 253, + 664 + ], + "score": 0.38, + "content": "( i ) >", + "type": "inline_equation" + }, + { + "bbox": [ + 253, + 651, + 320, + 664 + ], + "score": 1.0, + "content": "BlockHeight", + "type": "text" + }, + { + "bbox": [ + 320, + 651, + 344, + 664 + ], + "score": 0.74, + "content": "( j ) >", + "type": "inline_equation" + }, + { + "bbox": [ + 344, + 651, + 411, + 664 + ], + "score": 1.0, + "content": "BlockHeight", + "type": "text" + }, + { + "bbox": [ + 411, + 651, + 425, + 664 + ], + "score": 0.7, + "content": "( k )", + "type": "inline_equation" + }, + { + "bbox": [ + 425, + 651, + 506, + 664 + ], + "score": 1.0, + "content": "for three of the four", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 662, + 346, + 675 + ], + "spans": [ + { + "bbox": [ + 106, + 662, + 135, + 675 + ], + "score": 1.0, + "content": "blocks", + "type": "text" + }, + { + "bbox": [ + 135, + 663, + 158, + 674 + ], + "score": 0.88, + "content": "i , j , k", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 662, + 346, + 675 + ], + "score": 1.0, + "content": ". For more details, please refer to Appendix H.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 107, + 677, + 505, + 743 + ], + "lines": [ + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "Results In both the stacking and rearrangement settings, Decision Diffuser satisfies single", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "constraints with greater success rate than Diffuser (Table 3). We also compare with BCQ (Fujimoto", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "et al., 2019) and CQL (Kumar et al., 2020), but they consistently fail to stack or rearrange the blocks", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 710, + 505, + 721 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 721 + ], + "score": 1.0, + "content": "leading to a 0.0 success rate. Unlike these baselines, our method can just as effectively satisfy several", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 719, + 507, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 507, + 733 + ], + "score": 1.0, + "content": "constraints together according to Equation 9. For a visualization of these generated trajectories,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 731, + 383, + 744 + ], + "spans": [ + { + "bbox": [ + 105, + 731, + 383, + 744 + ], + "score": 1.0, + "content": "please see the website https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 39.5 + } + ], + "page_idx": 6, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 309, + 762 + ], + "score": 1.0, + "content": "7", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 138 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 507, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 507, + 95 + ], + "score": 1.0, + "content": "D4RL datasets (Fu et al., 2020). We compare with existing offline RL methods, including model-", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 93, + 505, + 105 + ], + "spans": [ + { + "bbox": [ + 106, + 93, + 505, + 105 + ], + "score": 1.0, + "content": "free algorithms like CQL (Kumar et al., 2020) and IQL (Kostrikov et al., 2022), and model-based", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 506, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 506, + 117 + ], + "score": 1.0, + "content": "algorithms such as trajectory transformer (TT, Janner et al. (2021)) and MoReL (Kidambi et al., 2020).", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 115, + 506, + 127 + ], + "spans": [ + { + "bbox": [ + 106, + 115, + 506, + 127 + ], + "score": 1.0, + "content": "We also compare with sequence-models like the Decision Transformer (DT) (Chen et al. (2021) and", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 126, + 312, + 138 + ], + "spans": [ + { + "bbox": [ + 105, + 126, + 312, + 138 + ], + "score": 1.0, + "content": "diffusion models like Diffuser (Janner et al., 2022).", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 2, + "bbox_fs": [ + 105, + 81, + 507, + 138 + ] + }, + { + "type": "table", + "bbox": [ + 145, + 150, + 465, + 292 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 145, + 150, + 465, + 292 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 145, + 150, + 465, + 292 + ], + "spans": [ + { + "bbox": [ + 145, + 150, + 465, + 292 + ], + "score": 0.977, + "html": "
DatasetEnvironmentBCCQLIQLDTTTMOReLDiffuserDD
Med-ExpertHalfCheetah55.291.686.786.89553.379.890.6 ±1.3
Med-ExpertHopper52.5105.491.5107.6110.0108.7107.2111.8 ±1.8
Med-ExpertWalker2d107.5108.8109.6108.1101.995.6108.4108.8 ±1.7
MediumHalfCheetah42.644.047.442.646.942.144.249.1 ±1.0
MediumHopper52.958.566.367.661.195.458.579.3 ±3.6
MediumWalker2d75.372.578.374.07977.879.782.5±1.4
Med-ReplayHalfCheetah36.645.544.236.641.940.242.239.3 ±4.1
Med-ReplayHopper18.19594.782.791.593.696.8100±0.7
Med-ReplayWalker2d26.077.273.966.682.649.861.275±4.3
Average51.977.67774.778.972.975.381.8
MixedKitchen51.552.4511、1165 ±2.8
PartialKitchen3850.146.3--==57 ±2.5
Average44.851.248.7-1--61
", + "type": "table", + "image_path": "0a2d0b726961f161d351afaa00b88a1d4c6293180230e7ffd5e79853da686c2d.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 145, + 150, + 465, + 197.33333333333334 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 145, + 197.33333333333334, + 465, + 244.66666666666669 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 145, + 244.66666666666669, + 465, + 292.0 + ], + "spans": [], + "index": 7 + } + ] + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 107, + 343, + 505, + 399 + ], + "lines": [ + { + "bbox": [ + 105, + 342, + 505, + 357 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 505, + 357 + ], + "score": 1.0, + "content": "Results Across different offline RL tasks, we find that the Decision Diffuser is either competitive or", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 353, + 505, + 367 + ], + "spans": [ + { + "bbox": [ + 105, + 353, + 505, + 367 + ], + "score": 1.0, + "content": "outperforms many offline RL baselines (Table 1). It also outperforms Diffuser and sequence modeling", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 365, + 506, + 379 + ], + "spans": [ + { + "bbox": [ + 105, + 365, + 506, + 379 + ], + "score": 1.0, + "content": "approaches, such as Decision Transformer and Trajectory Transformer. The difference between", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 376, + 505, + 389 + ], + "spans": [ + { + "bbox": [ + 106, + 376, + 505, + 389 + ], + "score": 1.0, + "content": "Decision Diffuser and other methods becomes even more significant on harder D4RL Kitchen tasks", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 387, + 280, + 401 + ], + "spans": [ + { + "bbox": [ + 105, + 387, + 280, + 401 + ], + "score": 1.0, + "content": "which require long-term credit assignment.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 10, + "bbox_fs": [ + 105, + 342, + 506, + 401 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 404, + 505, + 547 + ], + "lines": [ + { + "bbox": [ + 106, + 405, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 106, + 405, + 505, + 416 + ], + "score": 1.0, + "content": "To convey the importance of classifier-free guidance, we also compare with the baseline", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 415, + 506, + 427 + ], + "spans": [ + { + "bbox": [ + 106, + 415, + 506, + 427 + ], + "score": 1.0, + "content": "CondDiffuser, which diffuses over both state and action sequences as in Diffuser without classifier-", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 425, + 506, + 439 + ], + "spans": [ + { + "bbox": [ + 105, + 425, + 506, + 439 + ], + "score": 1.0, + "content": "guidance. In Table 2, we observe that CondDiffuser improves over Diffuser in 2 out of 3", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 437, + 505, + 450 + ], + "spans": [ + { + "bbox": [ + 105, + 437, + 505, + 450 + ], + "score": 1.0, + "content": "environments. Decision Diffuser further improves over CondDiffuser, performing better across all", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 104, + 446, + 506, + 462 + ], + "spans": [ + { + "bbox": [ + 104, + 446, + 506, + 462 + ], + "score": 1.0, + "content": "3 environments. We conclude that learning the inverse dynamics is a good alternative to diffusing", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 460, + 505, + 471 + ], + "spans": [ + { + "bbox": [ + 106, + 460, + 505, + 471 + ], + "score": 1.0, + "content": "over actions. We further empirically analyze when to use inverse dynamics and when to diffuse", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 470, + 506, + 483 + ], + "spans": [ + { + "bbox": [ + 105, + 470, + 506, + 483 + ], + "score": 1.0, + "content": "over actions in Appendix F. We also compare against CondMLPDiffuser, a policy where the current", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 480, + 505, + 494 + ], + "spans": [ + { + "bbox": [ + 105, + 480, + 505, + 494 + ], + "score": 1.0, + "content": "action is denoised according to a diffusion process conditioned on both the state and return. We see", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 491, + 505, + 505 + ], + "spans": [ + { + "bbox": [ + 105, + 491, + 505, + 505 + ], + "score": 1.0, + "content": "that CondMLPDiffuser performs the worst amongst diffusion models. Till now, we mainly tested on", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 501, + 506, + 517 + ], + "spans": [ + { + "bbox": [ + 105, + 501, + 506, + 517 + ], + "score": 1.0, + "content": "offline RL tasks that have deterministic (or near deterministic) environment dynamics. Hence, we", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 514, + 505, + 526 + ], + "spans": [ + { + "bbox": [ + 106, + 514, + 505, + 526 + ], + "score": 1.0, + "content": "test the robustness of Decision Diffuser to stochastic dynamics and compare it to Diffuser and CQL as", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 524, + 505, + 539 + ], + "spans": [ + { + "bbox": [ + 105, + 524, + 505, + 539 + ], + "score": 1.0, + "content": "we vary the stochasticity in environment dynamics, in Appendix G. Finally, we analyze the runtime", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 536, + 313, + 548 + ], + "spans": [ + { + "bbox": [ + 105, + 536, + 313, + 548 + ], + "score": 1.0, + "content": "characteristics of Decision Diffuser in Appendix E.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 19, + "bbox_fs": [ + 104, + 405, + 506, + 548 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 548, + 253, + 559 + ], + "lines": [ + { + "bbox": [ + 105, + 546, + 255, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 546, + 255, + 560 + ], + "score": 1.0, + "content": "4.2 CONSTRAINT SATISFACTION", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 26 + }, + { + "type": "text", + "bbox": [ + 106, + 564, + 505, + 674 + ], + "lines": [ + { + "bbox": [ + 105, + 561, + 506, + 578 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 506, + 578 + ], + "score": 1.0, + "content": "Setup We next evaluate how well we can generate trajectories that satisfy a set of constraints using", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 574, + 506, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 574, + 506, + 588 + ], + "score": 1.0, + "content": "the Kuka Block Stacking environment (Janner et al., 2022) visualized in Figure 5. In this domain,", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 585, + 506, + 598 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 506, + 598 + ], + "score": 1.0, + "content": "there are four blocks which can be stacked as a single tower or rearranged into several towers. A", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 597, + 506, + 610 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 227, + 610 + ], + "score": 1.0, + "content": "constraint like BlockHeigh", + "type": "text" + }, + { + "bbox": [ + 228, + 597, + 334, + 609 + ], + "score": 0.3, + "content": "\\mathsf { \\Lambda } = ( i ) > \\mathsf { B l o c k H e i g h t } ( j )", + "type": "inline_equation" + }, + { + "bbox": [ + 334, + 597, + 410, + 610 + ], + "score": 1.0, + "content": "requires that block", + "type": "text" + }, + { + "bbox": [ + 410, + 598, + 415, + 607 + ], + "score": 0.61, + "content": "i", + "type": "inline_equation" + }, + { + "bbox": [ + 415, + 597, + 506, + 610 + ], + "score": 1.0, + "content": "be placed above block", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 607, + 506, + 621 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 112, + 619 + ], + "score": 0.63, + "content": "j", + "type": "inline_equation" + }, + { + "bbox": [ + 113, + 607, + 506, + 621 + ], + "score": 1.0, + "content": ". We train the Decision Diffuser from 10, 000 expert demonstrations each satisfying one of these", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 618, + 505, + 632 + ], + "spans": [ + { + "bbox": [ + 105, + 618, + 505, + 632 + ], + "score": 1.0, + "content": "constraints. We randomize the positions of these blocks and consider two tasks at inference: sampling", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 630, + 505, + 641 + ], + "spans": [ + { + "bbox": [ + 106, + 630, + 505, + 641 + ], + "score": 1.0, + "content": "trajectories that satisfy a single constraint seen before in the dataset or satisfy a group of constraints", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "spans": [ + { + "bbox": [ + 106, + 641, + 505, + 653 + ], + "score": 1.0, + "content": "for which demonstrations were never provided. In the latter, we ask the Decision Diffuser to generate", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 651, + 506, + 664 + ], + "spans": [ + { + "bbox": [ + 106, + 651, + 229, + 664 + ], + "score": 1.0, + "content": "trajectories so BlockHeight", + "type": "text" + }, + { + "bbox": [ + 230, + 651, + 253, + 664 + ], + "score": 0.38, + "content": "( i ) >", + "type": "inline_equation" + }, + { + "bbox": [ + 253, + 651, + 320, + 664 + ], + "score": 1.0, + "content": "BlockHeight", + "type": "text" + }, + { + "bbox": [ + 320, + 651, + 344, + 664 + ], + "score": 0.74, + "content": "( j ) >", + "type": "inline_equation" + }, + { + "bbox": [ + 344, + 651, + 411, + 664 + ], + "score": 1.0, + "content": "BlockHeight", + "type": "text" + }, + { + "bbox": [ + 411, + 651, + 425, + 664 + ], + "score": 0.7, + "content": "( k )", + "type": "inline_equation" + }, + { + "bbox": [ + 425, + 651, + 506, + 664 + ], + "score": 1.0, + "content": "for three of the four", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 662, + 346, + 675 + ], + "spans": [ + { + "bbox": [ + 106, + 662, + 135, + 675 + ], + "score": 1.0, + "content": "blocks", + "type": "text" + }, + { + "bbox": [ + 135, + 663, + 158, + 674 + ], + "score": 0.88, + "content": "i , j , k", + "type": "inline_equation" + }, + { + "bbox": [ + 159, + 662, + 346, + 675 + ], + "score": 1.0, + "content": ". For more details, please refer to Appendix H.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 31.5, + "bbox_fs": [ + 105, + 561, + 506, + 675 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 677, + 505, + 743 + ], + "lines": [ + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "Results In both the stacking and rearrangement settings, Decision Diffuser satisfies single", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "constraints with greater success rate than Diffuser (Table 3). We also compare with BCQ (Fujimoto", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "et al., 2019) and CQL (Kumar et al., 2020), but they consistently fail to stack or rearrange the blocks", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 710, + 505, + 721 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 721 + ], + "score": 1.0, + "content": "leading to a 0.0 success rate. Unlike these baselines, our method can just as effectively satisfy several", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 719, + 507, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 507, + 733 + ], + "score": 1.0, + "content": "constraints together according to Equation 9. For a visualization of these generated trajectories,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 731, + 383, + 744 + ], + "spans": [ + { + "bbox": [ + 105, + 731, + 383, + 744 + ], + "score": 1.0, + "content": "please see the website https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 39.5, + "bbox_fs": [ + 105, + 676, + 507, + 744 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 200, + 81, + 411, + 118 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 200, + 81, + 411, + 118 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 200, + 81, + 411, + 118 + ], + "spans": [ + { + "bbox": [ + 200, + 81, + 411, + 118 + ], + "score": 0.962, + "html": "
Hopper-*Diffuser CondDiffuser CondMLPDiffuser Decision Diffuser
Med-Expert107.6111.3105.6111.8 ±1.6
Medium58.566.354.179.3 ±3.6
Med-Replay96.876.566.5100 ±0.7
", + "type": "table", + "image_path": "257506580b52a6d5d4748e37a00af4828bf9dfd022b9e7bd55dc454a5663de22.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 200, + 81, + 411, + 93.33333333333333 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 200, + 93.33333333333333, + 411, + 105.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 200, + 105.66666666666666, + 411, + 117.99999999999999 + ], + "spans": [], + "index": 2 + } + ] + } + ], + "index": 1 + }, + { + "type": "table", + "bbox": [ + 264, + 174, + 460, + 256 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 126, + 505, + 167 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 126, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 106, + 126, + 505, + 138 + ], + "score": 1.0, + "content": "Table 2: Ablations. Using classifier-free guidance with Diffuser, resulting in CondDiffuser, improves", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 136, + 505, + 148 + ], + "spans": [ + { + "bbox": [ + 106, + 136, + 505, + 148 + ], + "score": 1.0, + "content": "performance in 2 (out of 3) environments. Additionally, using inverse dynamics for action prediction in Decision", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 147, + 505, + 157 + ], + "spans": [ + { + "bbox": [ + 106, + 147, + 505, + 157 + ], + "score": 1.0, + "content": "Diffuser improves performance in all 3 environments. CondMLPDiffuser, that diffuses over current action", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 156, + 354, + 168 + ], + "spans": [ + { + "bbox": [ + 105, + 156, + 354, + 168 + ], + "score": 1.0, + "content": "given the current state and the target return, doesn’t perform as well.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + }, + { + "type": "table_body", + "bbox": [ + 264, + 174, + 460, + 256 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 264, + 174, + 460, + 256 + ], + "spans": [ + { + "bbox": [ + 264, + 174, + 460, + 256 + ], + "score": 0.951, + "html": "
EnvironmentDiffuserDD
Single Constraint - Stacking45.6 ±3.158.0 ±3.1
Single Constraint - Rearrangement58.9 ±3.462.7 ±3.1
Single Constraint Average52.360.4
Multiple Constraints - Stacking160.3 ±3.1
Multiple Constraints - Rearrangement67.2 ±3.1
Multiple Constraints Average163.8
", + "type": "table", + "image_path": "fb01f37be7d7902c1eccd3141e1effc0540e8f4dcdc17360b00d08dd994e4fd5.jpg" + } + ] + } + ], + "index": 13.5, + "virtual_lines": [ + { + "bbox": [ + 264, + 174, + 460, + 187.66666666666666 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 264, + 187.66666666666666, + 460, + 201.33333333333331 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 264, + 201.33333333333331, + 460, + 214.99999999999997 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 264, + 214.99999999999997, + 460, + 228.66666666666663 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 264, + 228.66666666666663, + 460, + 242.3333333333333 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 264, + 242.3333333333333, + 460, + 255.99999999999994 + ], + "spans": [], + "index": 16 + } + ] + } + ], + "index": 9.0 + }, + { + "type": "image", + "bbox": [ + 151, + 207, + 228, + 280 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 151, + 207, + 228, + 280 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 151, + 207, + 228, + 280 + ], + "spans": [ + { + "bbox": [ + 151, + 207, + 228, + 280 + ], + "score": 0.706, + "type": "image", + "image_path": "52d2870f8dae13bbc0f1462a3db00c970f194d62fa9583c2df125b201342dd4a.jpg" + } + ] + } + ], + "index": 7.5, + "virtual_lines": [ + { + "bbox": [ + 151, + 207, + 228, + 243.5 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 151, + 243.5, + 228, + 280.0 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 148, + 285, + 229, + 303 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 148, + 285, + 231, + 295 + ], + "spans": [ + { + "bbox": [ + 148, + 285, + 231, + 295 + ], + "score": 1.0, + "content": "Figure 5: Kuka Block", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 148, + 294, + 194, + 303 + ], + "spans": [ + { + "bbox": [ + 148, + 294, + 194, + 303 + ], + "score": 1.0, + "content": "Stacking task.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "image_caption", + "bbox": [ + 252, + 262, + 460, + 303 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 252, + 262, + 461, + 271 + ], + "spans": [ + { + "bbox": [ + 252, + 262, + 461, + 271 + ], + "score": 1.0, + "content": "Table 3: Block Stacking through Constraint Minimization. Decision", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 251, + 270, + 461, + 280 + ], + "spans": [ + { + "bbox": [ + 251, + 270, + 461, + 280 + ], + "score": 1.0, + "content": "Diffuser (DD) improves over Diffuser in terms of the success rate of", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 251, + 279, + 461, + 287 + ], + "spans": [ + { + "bbox": [ + 251, + 279, + 461, + 287 + ], + "score": 1.0, + "content": "generating trajectories satisfying a set of block-stacking constraints. It", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 252, + 287, + 460, + 295 + ], + "spans": [ + { + "bbox": [ + 252, + 287, + 460, + 295 + ], + "score": 1.0, + "content": "can also flexibly combine multiple constraints during test time. We report", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 252, + 295, + 443, + 303 + ], + "spans": [ + { + "bbox": [ + 252, + 295, + 443, + 303 + ], + "score": 1.0, + "content": "the mean success rate and the standard error over 5 random seeds.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 19 + } + ], + "index": 9.5 + }, + { + "type": "title", + "bbox": [ + 107, + 311, + 222, + 322 + ], + "lines": [ + { + "bbox": [ + 105, + 309, + 223, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 309, + 223, + 324 + ], + "score": 1.0, + "content": "4.3 SKILL COMPOSITION", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 106, + 326, + 505, + 381 + ], + "lines": [ + { + "bbox": [ + 105, + 325, + 506, + 339 + ], + "spans": [ + { + "bbox": [ + 105, + 325, + 135, + 339 + ], + "score": 1.0, + "content": "Setup", + "type": "text" + }, + { + "bbox": [ + 142, + 326, + 506, + 339 + ], + "score": 1.0, + "content": "Finally, we look at how to compose different skills together. We consider the Unitree-go-", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 104, + 336, + 506, + 350 + ], + "spans": [ + { + "bbox": [ + 104, + 336, + 506, + 350 + ], + "score": 1.0, + "content": "running environment (Margolis & Agrawal, 2022), where a quadruped robot can be found running", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 349, + 505, + 361 + ], + "spans": [ + { + "bbox": [ + 106, + 349, + 505, + 361 + ], + "score": 1.0, + "content": "with various gaits, like bounding, pacing, and trotting. We explore if it is possible to generate", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "score": 1.0, + "content": "trajectories that transition between these gaits after only training on individual gaits. For each gait,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 370, + 432, + 382 + ], + "spans": [ + { + "bbox": [ + 106, + 370, + 432, + 382 + ], + "score": 1.0, + "content": "we collect a dataset of 2500 demonstrations on which we train Decision Diffuser.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 106, + 384, + 505, + 505 + ], + "lines": [ + { + "bbox": [ + 105, + 383, + 505, + 397 + ], + "spans": [ + { + "bbox": [ + 105, + 383, + 505, + 397 + ], + "score": 1.0, + "content": "Results During testing, we use the noise model of our reverse diffusion process according to", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 395, + 506, + 408 + ], + "spans": [ + { + "bbox": [ + 105, + 395, + 506, + 408 + ], + "score": 1.0, + "content": "equation 9 to sample trajectories of the quadruped robot with entirely new running behavior. Figure 6", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 406, + 505, + 418 + ], + "spans": [ + { + "bbox": [ + 106, + 406, + 505, + 418 + ], + "score": 1.0, + "content": "shows a trajectory that begins with bounding but ends with pacing. Appendix I provides additional", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 418, + 505, + 429 + ], + "spans": [ + { + "bbox": [ + 106, + 418, + 505, + 429 + ], + "score": 1.0, + "content": "visualizations of running gaits being composed together. Although it visually appears that trajectories", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 427, + 505, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 427, + 505, + 441 + ], + "score": 1.0, + "content": "generated with the Decision Diffuser contain more than one gait, we would like to quantify exactly", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 104, + 438, + 506, + 452 + ], + "spans": [ + { + "bbox": [ + 104, + 438, + 506, + 452 + ], + "score": 1.0, + "content": "how well different gaits can be composed. To this end, we train a classifier to predict at every", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "spans": [ + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "score": 1.0, + "content": "time-step or frame in a trajectory the running gait of the quadruped (i.e. bound, pace, or trott). We", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 461, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 472 + ], + "score": 1.0, + "content": "reuse the demonstrations collected for training the Decision Diffuser to also train this classifier, where", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "our inputs are defined as robot joint states over a fixed period of time (i.e. state sub-sequences of", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 483, + 506, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 506, + 495 + ], + "score": 1.0, + "content": "length 10) and the label is the gait demonstrated in this sequence. The complete details of our gait", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 495, + 317, + 506 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 317, + 506 + ], + "score": 1.0, + "content": "classification procedure can be found in Appendix I.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 33 + }, + { + "type": "table", + "bbox": [ + 337, + 526, + 501, + 580 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 337, + 526, + 501, + 580 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 337, + 526, + 501, + 580 + ], + "spans": [ + { + "bbox": [ + 337, + 526, + 501, + 580 + ], + "score": 0.956, + "html": "
ConditionTrottPaceBound
Only Bound0.81.098.2
Only Pace1.497.70.9
Bound + Pace1.438.560.1
", + "type": "table", + "image_path": "32c5686d1598622b99760125a23a8c4046cff1a0b7089a79a049bd8f9292a57e.jpg" + } + ] + } + ], + "index": 44.0, + "virtual_lines": [ + { + "bbox": [ + 337, + 526, + 501, + 539.5 + ], + "spans": [], + "index": 41 + }, + { + "bbox": [ + 337, + 539.5, + 501, + 553.0 + ], + "spans": [], + "index": 43 + }, + { + "bbox": [ + 337, + 553.0, + 501, + 566.5 + ], + "spans": [], + "index": 45 + }, + { + "bbox": [ + 337, + 566.5, + 501, + 580.0 + ], + "spans": [], + "index": 47 + } + ] + } + ], + "index": 44.0 + }, + { + "type": "image", + "bbox": [ + 106, + 516, + 329, + 592 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 516, + 329, + 592 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 516, + 329, + 592 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 329, + 592 + ], + "score": 0.907, + "type": "image", + "image_path": "8874639fd20e50d9304875bc9a9d40220fd33875e237e4b977f7e557544d6fc5.jpg" + } + ] + } + ], + "index": 43.0, + "virtual_lines": [ + { + "bbox": [ + 106, + 516, + 329, + 528.6666666666666 + ], + "spans": [], + "index": 39 + }, + { + "bbox": [ + 106, + 528.6666666666666, + 329, + 541.3333333333333 + ], + "spans": [], + "index": 40 + }, + { + "bbox": [ + 106, + 541.3333333333333, + 329, + 553.9999999999999 + ], + "spans": [], + "index": 42 + }, + { + "bbox": [ + 106, + 553.9999999999999, + 329, + 566.6666666666665 + ], + "spans": [], + "index": 44 + }, + { + "bbox": [ + 106, + 566.6666666666665, + 329, + 579.3333333333331 + ], + "spans": [], + "index": 46 + }, + { + "bbox": [ + 106, + 579.3333333333331, + 329, + 591.9999999999998 + ], + "spans": [], + "index": 48 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 599, + 506, + 640 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 598, + 506, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 598, + 506, + 613 + ], + "score": 1.0, + "content": "Figure 7: Classifying Running Gaits. A classifier predicts the running gait of the quadruped at every timestep.", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 106, + 609, + 505, + 621 + ], + "spans": [ + { + "bbox": [ + 106, + 609, + 505, + 621 + ], + "score": 1.0, + "content": "On trajectories generated by conditioning on a single skill, like only bounding or pacing, the classifier predicts", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 619, + 505, + 630 + ], + "spans": [ + { + "bbox": [ + 106, + 619, + 505, + 630 + ], + "score": 1.0, + "content": "the respective gait with largest probability. When conditioned on both skills, some timesteps are classified as", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 105, + 627, + 227, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 227, + 643 + ], + "score": 1.0, + "content": "bounding while others as pacing.", + "type": "text" + } + ], + "index": 52 + } + ], + "index": 50.5 + } + ], + "index": 46.75 + }, + { + "type": "text", + "bbox": [ + 106, + 644, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "score": 1.0, + "content": "We use our running gait classifier in two ways: to evaluate how the behavior of the quadruped changes", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 106, + 655, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 506, + 668 + ], + "score": 1.0, + "content": "over the course of a single, generated trajectory and to measure how often each gait emerges over", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 105, + 666, + 504, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 504, + 678 + ], + "score": 1.0, + "content": "several generated trajectories. In the former, we first sample three trajectories from the Decision", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "score": 1.0, + "content": "Diffuser conditioned either on the bounding gait, the pacing gait, or both. For every trajectory, we", + "type": "text" + } + ], + "index": 56 + }, + { + "bbox": [ + 104, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 104, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "separately plot the classification probability of each gait over the length of the sequence. As shown in", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 105, + 698, + 505, + 713 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 713 + ], + "score": 1.0, + "content": "the plots of Figure 7, the classifier predicts bound and pace respectively to be the most likely running", + "type": "text" + } + ], + "index": 58 + }, + { + "bbox": [ + 105, + 710, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 505, + 723 + ], + "score": 1.0, + "content": "gait in trajectories sampled with this condition. When the trajectory is generated by conditioning", + "type": "text" + } + ], + "index": 59 + }, + { + "bbox": [ + 105, + 720, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 505, + 733 + ], + "score": 1.0, + "content": "on both gaits, the classifier transitions between predicting one gait with largest probability to the", + "type": "text" + } + ], + "index": 60 + } + ], + "index": 56.5 + } + ], + "page_idx": 7, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 300, + 750, + 309, + 761 + ], + "spans": [ + { + "bbox": [ + 300, + 750, + 309, + 761 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 11, + "width": 9 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 200, + 81, + 411, + 118 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 200, + 81, + 411, + 118 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 200, + 81, + 411, + 118 + ], + "spans": [ + { + "bbox": [ + 200, + 81, + 411, + 118 + ], + "score": 0.962, + "html": "
Hopper-*Diffuser CondDiffuser CondMLPDiffuser Decision Diffuser
Med-Expert107.6111.3105.6111.8 ±1.6
Medium58.566.354.179.3 ±3.6
Med-Replay96.876.566.5100 ±0.7
", + "type": "table", + "image_path": "257506580b52a6d5d4748e37a00af4828bf9dfd022b9e7bd55dc454a5663de22.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 200, + 81, + 411, + 93.33333333333333 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 200, + 93.33333333333333, + 411, + 105.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 200, + 105.66666666666666, + 411, + 117.99999999999999 + ], + "spans": [], + "index": 2 + } + ] + } + ], + "index": 1 + }, + { + "type": "table", + "bbox": [ + 264, + 174, + 460, + 256 + ], + "blocks": [ + { + "type": "table_caption", + "bbox": [ + 106, + 126, + 505, + 167 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 126, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 106, + 126, + 505, + 138 + ], + "score": 1.0, + "content": "Table 2: Ablations. Using classifier-free guidance with Diffuser, resulting in CondDiffuser, improves", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 136, + 505, + 148 + ], + "spans": [ + { + "bbox": [ + 106, + 136, + 505, + 148 + ], + "score": 1.0, + "content": "performance in 2 (out of 3) environments. Additionally, using inverse dynamics for action prediction in Decision", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 147, + 505, + 157 + ], + "spans": [ + { + "bbox": [ + 106, + 147, + 505, + 157 + ], + "score": 1.0, + "content": "Diffuser improves performance in all 3 environments. CondMLPDiffuser, that diffuses over current action", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 156, + 354, + 168 + ], + "spans": [ + { + "bbox": [ + 105, + 156, + 354, + 168 + ], + "score": 1.0, + "content": "given the current state and the target return, doesn’t perform as well.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + }, + { + "type": "table_body", + "bbox": [ + 264, + 174, + 460, + 256 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 264, + 174, + 460, + 256 + ], + "spans": [ + { + "bbox": [ + 264, + 174, + 460, + 256 + ], + "score": 0.951, + "html": "
EnvironmentDiffuserDD
Single Constraint - Stacking45.6 ±3.158.0 ±3.1
Single Constraint - Rearrangement58.9 ±3.462.7 ±3.1
Single Constraint Average52.360.4
Multiple Constraints - Stacking160.3 ±3.1
Multiple Constraints - Rearrangement67.2 ±3.1
Multiple Constraints Average163.8
", + "type": "table", + "image_path": "fb01f37be7d7902c1eccd3141e1effc0540e8f4dcdc17360b00d08dd994e4fd5.jpg" + } + ] + } + ], + "index": 13.5, + "virtual_lines": [ + { + "bbox": [ + 264, + 174, + 460, + 187.66666666666666 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 264, + 187.66666666666666, + 460, + 201.33333333333331 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 264, + 201.33333333333331, + 460, + 214.99999999999997 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 264, + 214.99999999999997, + 460, + 228.66666666666663 + ], + "spans": [], + "index": 14 + }, + { + "bbox": [ + 264, + 228.66666666666663, + 460, + 242.3333333333333 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 264, + 242.3333333333333, + 460, + 255.99999999999994 + ], + "spans": [], + "index": 16 + } + ] + } + ], + "index": 9.0 + }, + { + "type": "image", + "bbox": [ + 151, + 207, + 228, + 280 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 151, + 207, + 228, + 280 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 151, + 207, + 228, + 280 + ], + "spans": [ + { + "bbox": [ + 151, + 207, + 228, + 280 + ], + "score": 0.706, + "type": "image", + "image_path": "52d2870f8dae13bbc0f1462a3db00c970f194d62fa9583c2df125b201342dd4a.jpg" + } + ] + } + ], + "index": 7.5, + "virtual_lines": [ + { + "bbox": [ + 151, + 207, + 228, + 243.5 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 151, + 243.5, + 228, + 280.0 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 148, + 285, + 229, + 303 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 148, + 285, + 231, + 295 + ], + "spans": [ + { + "bbox": [ + 148, + 285, + 231, + 295 + ], + "score": 1.0, + "content": "Figure 5: Kuka Block", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 148, + 294, + 194, + 303 + ], + "spans": [ + { + "bbox": [ + 148, + 294, + 194, + 303 + ], + "score": 1.0, + "content": "Stacking task.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "image_caption", + "bbox": [ + 252, + 262, + 460, + 303 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 252, + 262, + 461, + 271 + ], + "spans": [ + { + "bbox": [ + 252, + 262, + 461, + 271 + ], + "score": 1.0, + "content": "Table 3: Block Stacking through Constraint Minimization. Decision", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 251, + 270, + 461, + 280 + ], + "spans": [ + { + "bbox": [ + 251, + 270, + 461, + 280 + ], + "score": 1.0, + "content": "Diffuser (DD) improves over Diffuser in terms of the success rate of", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 251, + 279, + 461, + 287 + ], + "spans": [ + { + "bbox": [ + 251, + 279, + 461, + 287 + ], + "score": 1.0, + "content": "generating trajectories satisfying a set of block-stacking constraints. It", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 252, + 287, + 460, + 295 + ], + "spans": [ + { + "bbox": [ + 252, + 287, + 460, + 295 + ], + "score": 1.0, + "content": "can also flexibly combine multiple constraints during test time. We report", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 252, + 295, + 443, + 303 + ], + "spans": [ + { + "bbox": [ + 252, + 295, + 443, + 303 + ], + "score": 1.0, + "content": "the mean success rate and the standard error over 5 random seeds.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 19 + } + ], + "index": 9.5 + }, + { + "type": "title", + "bbox": [ + 107, + 311, + 222, + 322 + ], + "lines": [ + { + "bbox": [ + 105, + 309, + 223, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 309, + 223, + 324 + ], + "score": 1.0, + "content": "4.3 SKILL COMPOSITION", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 106, + 326, + 505, + 381 + ], + "lines": [ + { + "bbox": [ + 105, + 325, + 506, + 339 + ], + "spans": [ + { + "bbox": [ + 105, + 325, + 135, + 339 + ], + "score": 1.0, + "content": "Setup", + "type": "text" + }, + { + "bbox": [ + 142, + 326, + 506, + 339 + ], + "score": 1.0, + "content": "Finally, we look at how to compose different skills together. We consider the Unitree-go-", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 104, + 336, + 506, + 350 + ], + "spans": [ + { + "bbox": [ + 104, + 336, + 506, + 350 + ], + "score": 1.0, + "content": "running environment (Margolis & Agrawal, 2022), where a quadruped robot can be found running", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 349, + 505, + 361 + ], + "spans": [ + { + "bbox": [ + 106, + 349, + 505, + 361 + ], + "score": 1.0, + "content": "with various gaits, like bounding, pacing, and trotting. We explore if it is possible to generate", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 506, + 372 + ], + "score": 1.0, + "content": "trajectories that transition between these gaits after only training on individual gaits. For each gait,", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 370, + 432, + 382 + ], + "spans": [ + { + "bbox": [ + 106, + 370, + 432, + 382 + ], + "score": 1.0, + "content": "we collect a dataset of 2500 demonstrations on which we train Decision Diffuser.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 25, + "bbox_fs": [ + 104, + 325, + 506, + 382 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 384, + 505, + 505 + ], + "lines": [ + { + "bbox": [ + 105, + 383, + 505, + 397 + ], + "spans": [ + { + "bbox": [ + 105, + 383, + 505, + 397 + ], + "score": 1.0, + "content": "Results During testing, we use the noise model of our reverse diffusion process according to", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 395, + 506, + 408 + ], + "spans": [ + { + "bbox": [ + 105, + 395, + 506, + 408 + ], + "score": 1.0, + "content": "equation 9 to sample trajectories of the quadruped robot with entirely new running behavior. Figure 6", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 406, + 505, + 418 + ], + "spans": [ + { + "bbox": [ + 106, + 406, + 505, + 418 + ], + "score": 1.0, + "content": "shows a trajectory that begins with bounding but ends with pacing. Appendix I provides additional", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 418, + 505, + 429 + ], + "spans": [ + { + "bbox": [ + 106, + 418, + 505, + 429 + ], + "score": 1.0, + "content": "visualizations of running gaits being composed together. Although it visually appears that trajectories", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 427, + 505, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 427, + 505, + 441 + ], + "score": 1.0, + "content": "generated with the Decision Diffuser contain more than one gait, we would like to quantify exactly", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 104, + 438, + 506, + 452 + ], + "spans": [ + { + "bbox": [ + 104, + 438, + 506, + 452 + ], + "score": 1.0, + "content": "how well different gaits can be composed. To this end, we train a classifier to predict at every", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "spans": [ + { + "bbox": [ + 106, + 450, + 505, + 462 + ], + "score": 1.0, + "content": "time-step or frame in a trajectory the running gait of the quadruped (i.e. bound, pace, or trott). We", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 461, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 472 + ], + "score": 1.0, + "content": "reuse the demonstrations collected for training the Decision Diffuser to also train this classifier, where", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "our inputs are defined as robot joint states over a fixed period of time (i.e. state sub-sequences of", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 483, + 506, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 483, + 506, + 495 + ], + "score": 1.0, + "content": "length 10) and the label is the gait demonstrated in this sequence. The complete details of our gait", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 106, + 495, + 317, + 506 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 317, + 506 + ], + "score": 1.0, + "content": "classification procedure can be found in Appendix I.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 33, + "bbox_fs": [ + 104, + 383, + 506, + 506 + ] + }, + { + "type": "table", + "bbox": [ + 337, + 526, + 501, + 580 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 337, + 526, + 501, + 580 + ], + "group_id": 2, + "lines": [ + { + "bbox": [ + 337, + 526, + 501, + 580 + ], + "spans": [ + { + "bbox": [ + 337, + 526, + 501, + 580 + ], + "score": 0.956, + "html": "
ConditionTrottPaceBound
Only Bound0.81.098.2
Only Pace1.497.70.9
Bound + Pace1.438.560.1
", + "type": "table", + "image_path": "32c5686d1598622b99760125a23a8c4046cff1a0b7089a79a049bd8f9292a57e.jpg" + } + ] + } + ], + "index": 44.0, + "virtual_lines": [ + { + "bbox": [ + 337, + 526, + 501, + 539.5 + ], + "spans": [], + "index": 41 + }, + { + "bbox": [ + 337, + 539.5, + 501, + 553.0 + ], + "spans": [], + "index": 43 + }, + { + "bbox": [ + 337, + 553.0, + 501, + 566.5 + ], + "spans": [], + "index": 45 + }, + { + "bbox": [ + 337, + 566.5, + 501, + 580.0 + ], + "spans": [], + "index": 47 + } + ] + } + ], + "index": 44.0 + }, + { + "type": "image", + "bbox": [ + 106, + 516, + 329, + 592 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 516, + 329, + 592 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 516, + 329, + 592 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 329, + 592 + ], + "score": 0.907, + "type": "image", + "image_path": "8874639fd20e50d9304875bc9a9d40220fd33875e237e4b977f7e557544d6fc5.jpg" + } + ] + } + ], + "index": 43.0, + "virtual_lines": [ + { + "bbox": [ + 106, + 516, + 329, + 528.6666666666666 + ], + "spans": [], + "index": 39 + }, + { + "bbox": [ + 106, + 528.6666666666666, + 329, + 541.3333333333333 + ], + "spans": [], + "index": 40 + }, + { + "bbox": [ + 106, + 541.3333333333333, + 329, + 553.9999999999999 + ], + "spans": [], + "index": 42 + }, + { + "bbox": [ + 106, + 553.9999999999999, + 329, + 566.6666666666665 + ], + "spans": [], + "index": 44 + }, + { + "bbox": [ + 106, + 566.6666666666665, + 329, + 579.3333333333331 + ], + "spans": [], + "index": 46 + }, + { + "bbox": [ + 106, + 579.3333333333331, + 329, + 591.9999999999998 + ], + "spans": [], + "index": 48 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 599, + 506, + 640 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 105, + 598, + 506, + 613 + ], + "spans": [ + { + "bbox": [ + 105, + 598, + 506, + 613 + ], + "score": 1.0, + "content": "Figure 7: Classifying Running Gaits. A classifier predicts the running gait of the quadruped at every timestep.", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 106, + 609, + 505, + 621 + ], + "spans": [ + { + "bbox": [ + 106, + 609, + 505, + 621 + ], + "score": 1.0, + "content": "On trajectories generated by conditioning on a single skill, like only bounding or pacing, the classifier predicts", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 619, + 505, + 630 + ], + "spans": [ + { + "bbox": [ + 106, + 619, + 505, + 630 + ], + "score": 1.0, + "content": "the respective gait with largest probability. When conditioned on both skills, some timesteps are classified as", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 105, + 627, + 227, + 643 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 227, + 643 + ], + "score": 1.0, + "content": "bounding while others as pacing.", + "type": "text" + } + ], + "index": 52 + } + ], + "index": 50.5 + } + ], + "index": 46.75 + }, + { + "type": "text", + "bbox": [ + 106, + 644, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "score": 1.0, + "content": "We use our running gait classifier in two ways: to evaluate how the behavior of the quadruped changes", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 106, + 655, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 506, + 668 + ], + "score": 1.0, + "content": "over the course of a single, generated trajectory and to measure how often each gait emerges over", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 105, + 666, + 504, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 504, + 678 + ], + "score": 1.0, + "content": "several generated trajectories. In the former, we first sample three trajectories from the Decision", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 505, + 690 + ], + "score": 1.0, + "content": "Diffuser conditioned either on the bounding gait, the pacing gait, or both. For every trajectory, we", + "type": "text" + } + ], + "index": 56 + }, + { + "bbox": [ + 104, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 104, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "separately plot the classification probability of each gait over the length of the sequence. As shown in", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 105, + 698, + 505, + 713 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 713 + ], + "score": 1.0, + "content": "the plots of Figure 7, the classifier predicts bound and pace respectively to be the most likely running", + "type": "text" + } + ], + "index": 58 + }, + { + "bbox": [ + 105, + 710, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 710, + 505, + 723 + ], + "score": 1.0, + "content": "gait in trajectories sampled with this condition. When the trajectory is generated by conditioning", + "type": "text" + } + ], + "index": 59 + }, + { + "bbox": [ + 105, + 720, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 505, + 733 + ], + "score": 1.0, + "content": "on both gaits, the classifier transitions between predicting one gait with largest probability to the", + "type": "text" + } + ], + "index": 60 + }, + { + "bbox": [ + 106, + 173, + 505, + 186 + ], + "spans": [ + { + "bbox": [ + 106, + 173, + 505, + 186 + ], + "score": 1.0, + "content": "other. In fact, there are several instances where the behavior of the quadruped switches between", + "type": "text", + "cross_page": true + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 186, + 504, + 197 + ], + "spans": [ + { + "bbox": [ + 106, + 186, + 504, + 197 + ], + "score": 1.0, + "content": "bounding and pacing according to the classifier. This is consistent with the visualizations reported in", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 196, + 505, + 209 + ], + "spans": [ + { + "bbox": [ + 105, + 196, + 505, + 209 + ], + "score": 1.0, + "content": "Figure 6. In the table depicted in Figure 7, we consider 1000 trajectories generated with the Decision", + "type": "text", + "cross_page": true + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 207, + 505, + 219 + ], + "spans": [ + { + "bbox": [ + 105, + 207, + 505, + 219 + ], + "score": 1.0, + "content": "Diffuser when conditioned on one or both of the gaits as listed. We record the fraction of time that the", + "type": "text", + "cross_page": true + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 218, + 506, + 230 + ], + "spans": [ + { + "bbox": [ + 105, + 218, + 506, + 230 + ], + "score": 1.0, + "content": "quadruped’s running gait was classified as either trott, pace, or bound. It turns out that the classifier", + "type": "text", + "cross_page": true + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 229, + 505, + 241 + ], + "spans": [ + { + "bbox": [ + 105, + 229, + 266, + 241 + ], + "score": 1.0, + "content": "identifies the behavior as bounding for", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 266, + 229, + 293, + 240 + ], + "score": 0.88, + "content": "3 8 . 5 \\%", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 294, + 229, + 452, + 241 + ], + "score": 1.0, + "content": "of the time and as pacing for the other", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 452, + 229, + 479, + 240 + ], + "score": 0.88, + "content": "6 0 . 1 \\%", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 480, + 229, + 505, + 241 + ], + "score": 1.0, + "content": "when", + "type": "text", + "cross_page": true + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 240, + 506, + 253 + ], + "spans": [ + { + "bbox": [ + 106, + 240, + 506, + 253 + ], + "score": 1.0, + "content": "trajectories are sampled by composing both gaits. This corroborates the fact that the Decision Diffuser", + "type": "text", + "cross_page": true + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 252, + 448, + 263 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 448, + 263 + ], + "score": 1.0, + "content": "can indeed compose running behaviors despite only being trained on individual gaits.", + "type": "text", + "cross_page": true + } + ], + "index": 13 + } + ], + "index": 56.5, + "bbox_fs": [ + 104, + 643, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 205, + 80, + 406, + 128 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 205, + 80, + 406, + 128 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 205, + 80, + 406, + 128 + ], + "spans": [ + { + "bbox": [ + 205, + 80, + 406, + 128 + ], + "score": 0.961, + "type": "image", + "image_path": "ae5902e9ec2fa29c05349103c337c534e1a146b5725255f610070875b832cd24.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 205, + 80, + 406, + 96.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 205, + 96.0, + 406, + 112.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 205, + 112.0, + 406, + 128.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 108, + 137, + 505, + 168 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 136, + 506, + 149 + ], + "spans": [ + { + "bbox": [ + 106, + 136, + 506, + 149 + ], + "score": 1.0, + "content": "Figure 6: Composing Movement Skills. Decision Diffuser can imitate individual running gaits using expert", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 147, + 505, + 159 + ], + "spans": [ + { + "bbox": [ + 105, + 147, + 505, + 159 + ], + "score": 1.0, + "content": "demonstrations and compose multiple different skills together during test time. The results are best illustrated by", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 157, + 343, + 169 + ], + "spans": [ + { + "bbox": [ + 106, + 157, + 343, + 169 + ], + "score": 1.0, + "content": "videos viewable at https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 107, + 174, + 505, + 262 + ], + "lines": [ + { + "bbox": [ + 106, + 173, + 505, + 186 + ], + "spans": [ + { + "bbox": [ + 106, + 173, + 505, + 186 + ], + "score": 1.0, + "content": "other. In fact, there are several instances where the behavior of the quadruped switches between", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 186, + 504, + 197 + ], + "spans": [ + { + "bbox": [ + 106, + 186, + 504, + 197 + ], + "score": 1.0, + "content": "bounding and pacing according to the classifier. This is consistent with the visualizations reported in", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 196, + 505, + 209 + ], + "spans": [ + { + "bbox": [ + 105, + 196, + 505, + 209 + ], + "score": 1.0, + "content": "Figure 6. In the table depicted in Figure 7, we consider 1000 trajectories generated with the Decision", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 207, + 505, + 219 + ], + "spans": [ + { + "bbox": [ + 105, + 207, + 505, + 219 + ], + "score": 1.0, + "content": "Diffuser when conditioned on one or both of the gaits as listed. We record the fraction of time that the", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 218, + 506, + 230 + ], + "spans": [ + { + "bbox": [ + 105, + 218, + 506, + 230 + ], + "score": 1.0, + "content": "quadruped’s running gait was classified as either trott, pace, or bound. It turns out that the classifier", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 229, + 505, + 241 + ], + "spans": [ + { + "bbox": [ + 105, + 229, + 266, + 241 + ], + "score": 1.0, + "content": "identifies the behavior as bounding for", + "type": "text" + }, + { + "bbox": [ + 266, + 229, + 293, + 240 + ], + "score": 0.88, + "content": "3 8 . 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 294, + 229, + 452, + 241 + ], + "score": 1.0, + "content": "of the time and as pacing for the other", + "type": "text" + }, + { + "bbox": [ + 452, + 229, + 479, + 240 + ], + "score": 0.88, + "content": "6 0 . 1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 480, + 229, + 505, + 241 + ], + "score": 1.0, + "content": "when", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 240, + 506, + 253 + ], + "spans": [ + { + "bbox": [ + 106, + 240, + 506, + 253 + ], + "score": 1.0, + "content": "trajectories are sampled by composing both gaits. This corroborates the fact that the Decision Diffuser", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 252, + 448, + 263 + ], + "spans": [ + { + "bbox": [ + 105, + 252, + 448, + 263 + ], + "score": 1.0, + "content": "can indeed compose running behaviors despite only being trained on individual gaits.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 9.5 + }, + { + "type": "title", + "bbox": [ + 108, + 264, + 212, + 276 + ], + "lines": [ + { + "bbox": [ + 105, + 262, + 213, + 277 + ], + "spans": [ + { + "bbox": [ + 105, + 262, + 213, + 277 + ], + "score": 1.0, + "content": "5 RELATED WORK", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14 + }, + { + "type": "text", + "bbox": [ + 107, + 280, + 506, + 488 + ], + "lines": [ + { + "bbox": [ + 105, + 278, + 506, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 278, + 506, + 291 + ], + "score": 1.0, + "content": "Diffusion Models Diffusion Models is proficient in learning generative models of image and text", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 289, + 505, + 302 + ], + "spans": [ + { + "bbox": [ + 106, + 289, + 505, + 302 + ], + "score": 1.0, + "content": "data (Saharia et al., 2022; Nichol et al., 2021; Nichol & Dhariwal, 2021). It formulates the data", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 300, + 506, + 313 + ], + "spans": [ + { + "bbox": [ + 105, + 300, + 506, + 313 + ], + "score": 1.0, + "content": "sampling process as an iterative denoising procedure (Sohl-Dickstein et al., 2015; Ho et al., 2020).", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 311, + 506, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 311, + 506, + 324 + ], + "score": 1.0, + "content": "The denoising procedure can be alternatively interpreted as parameterizing the gradients of the data", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 323, + 506, + 335 + ], + "spans": [ + { + "bbox": [ + 106, + 323, + 506, + 335 + ], + "score": 1.0, + "content": "distribution (Song et al., 2021) optimizing the score matching objective (Hyvarinen ¨ , 2005) and thus", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 333, + 506, + 346 + ], + "spans": [ + { + "bbox": [ + 105, + 333, + 506, + 346 + ], + "score": 1.0, + "content": "as a Energy-Based Model (Du & Mordatch, 2019; Nijkamp et al., 2019; Grathwohl et al., 2020).", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 345, + 506, + 357 + ], + "spans": [ + { + "bbox": [ + 106, + 345, + 506, + 357 + ], + "score": 1.0, + "content": "To generate data samples (eg: images) conditioned on some additional information (eg:text), prior", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 354, + 507, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 354, + 507, + 369 + ], + "score": 1.0, + "content": "works (Nichol & Dhariwal, 2021) have learned a classifier to facilitate the conditional sampling.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 367, + 506, + 379 + ], + "spans": [ + { + "bbox": [ + 106, + 367, + 506, + 379 + ], + "score": 1.0, + "content": "More recent works (Ho & Salimans, 2022) have argued to leverage gradients of an implicit classifier,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 378, + 506, + 389 + ], + "spans": [ + { + "bbox": [ + 106, + 378, + 506, + 389 + ], + "score": 1.0, + "content": "formed by the difference in score functions of a conditional and an unconditional model, to facilitate", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 389, + 506, + 401 + ], + "spans": [ + { + "bbox": [ + 106, + 389, + 506, + 401 + ], + "score": 1.0, + "content": "conditional sampling. The resulting classifier-free guidance has been shown to generate better", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 399, + 506, + 411 + ], + "spans": [ + { + "bbox": [ + 105, + 399, + 506, + 411 + ], + "score": 1.0, + "content": "conditional samples than classifier-based guidance. Recent works have also used diffusion models to", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 411, + 505, + 422 + ], + "spans": [ + { + "bbox": [ + 106, + 411, + 505, + 422 + ], + "score": 1.0, + "content": "imitate human behavior (Pearce et al., 2023) and to parameterize policy in offline RL (Wang et al.,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 421, + 506, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 421, + 506, + 434 + ], + "score": 1.0, + "content": "2022). Janner et al. (2022) generate trajectories consisting of states and actions with an unconditional", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 432, + 506, + 445 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 506, + 445 + ], + "score": 1.0, + "content": "diffusion model, therefore requiring a trained reward function on noisy state-action pairs. At inference,", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 443, + 505, + 456 + ], + "spans": [ + { + "bbox": [ + 106, + 443, + 505, + 456 + ], + "score": 1.0, + "content": "the estimated reward function guides the reverse diffusion process towards samples of high-return", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 454, + 506, + 467 + ], + "spans": [ + { + "bbox": [ + 106, + 454, + 506, + 467 + ], + "score": 1.0, + "content": "trajectories. In contrast, we do not train reward functions or diffusion processes separately, but rather", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 466, + 505, + 477 + ], + "spans": [ + { + "bbox": [ + 106, + 466, + 505, + 477 + ], + "score": 1.0, + "content": "model the trajectories in our dataset with a single, conditional generative model. This ensures that the", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 475, + 504, + 490 + ], + "spans": [ + { + "bbox": [ + 105, + 475, + 504, + 490 + ], + "score": 1.0, + "content": "sampling procedure of the learned diffusion process is the same at inference as it is during training.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 107, + 491, + 505, + 568 + ], + "lines": [ + { + "bbox": [ + 106, + 489, + 506, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 506, + 502 + ], + "score": 1.0, + "content": "Reward Conditioned Policies Prior works (Kumar et al., 2019; Schmidhuber, 2019; Emmons", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 501, + 506, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 501, + 506, + 513 + ], + "score": 1.0, + "content": "et al., 2021; Chen et al., 2021) have studied learning of reward conditioned policies via reward", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "score": 1.0, + "content": "conditioned behavioral cloning. Chen et al. (2021) used a transformer (Vaswani et al., 2017) to model", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 522, + 506, + 536 + ], + "spans": [ + { + "bbox": [ + 105, + 522, + 506, + 536 + ], + "score": 1.0, + "content": "the reward conditioned policies and obtained a performance competitive with offline RL approaches.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 534, + 506, + 547 + ], + "spans": [ + { + "bbox": [ + 105, + 534, + 506, + 547 + ], + "score": 1.0, + "content": "Emmons et al. (2021) obtained similar performance as Chen et al. (2021) without using a transformer", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 545, + 505, + 558 + ], + "spans": [ + { + "bbox": [ + 106, + 545, + 505, + 558 + ], + "score": 1.0, + "content": "policy but relied on careful capacity tuning of MLP policy. In contrast, Decision Diffuser can also", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 555, + 347, + 569 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 347, + 569 + ], + "score": 1.0, + "content": "model constraints or skills and their resulting compositions.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 37 + }, + { + "type": "title", + "bbox": [ + 107, + 574, + 191, + 586 + ], + "lines": [ + { + "bbox": [ + 105, + 572, + 192, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 192, + 588 + ], + "score": 1.0, + "content": "6 DISCUSSION", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 41 + }, + { + "type": "text", + "bbox": [ + 106, + 589, + 506, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 589, + 505, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 589, + 505, + 601 + ], + "score": 1.0, + "content": "We propose Decision Diffuser, a conditional generative model for sequential decision making. It", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 600, + 505, + 612 + ], + "spans": [ + { + "bbox": [ + 106, + 600, + 505, + 612 + ], + "score": 1.0, + "content": "frames offline sequential decision making as conditional generative modeling and sidesteps the need", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "score": 1.0, + "content": "of reinforcement learning, thereby making the decision making pipeline simpler. By sampling for", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "score": 1.0, + "content": "high returns, it is able to capture the best behaviors in the dataset and outperforms existing offline", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 633, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 506, + 646 + ], + "score": 1.0, + "content": "RL approaches on standard D4RL benchmarks. In addition to returns, it can also be conditioned", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 104, + 643, + 506, + 658 + ], + "spans": [ + { + "bbox": [ + 104, + 643, + 506, + 658 + ], + "score": 1.0, + "content": "on constraints or skills and can generate novel behaviors by flexibly combining constraints or", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 654, + 506, + 670 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 506, + 670 + ], + "score": 1.0, + "content": "composing skills during test time. In this work, we focused on offline sequential decision making,", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "score": 1.0, + "content": "thus circumventing the need for exploration. Using ideas from Zheng et al. (2022), future works", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 106, + 677, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 506, + 690 + ], + "score": 1.0, + "content": "could look into online fine-tuning of Decision Diffuser by leveraging entropy of the state-sequence", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "model for exploration. While our work focused on state based environments, it can be extended", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "to image based environments by performing the diffusion in latent space, rather than observation", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 104, + 708, + 507, + 724 + ], + "spans": [ + { + "bbox": [ + 104, + 708, + 507, + 724 + ], + "score": 1.0, + "content": "space, as done in Rombach et al. (2022). For a detailed discussion on limitations of Decision Diffuser,", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 106, + 721, + 218, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 218, + 732 + ], + "score": 1.0, + "content": "please refer to Appendix L.", + "type": "text" + } + ], + "index": 54 + } + ], + "index": 48 + } + ], + "page_idx": 8, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "score": 1.0, + "content": "9", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 205, + 80, + 406, + 128 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 205, + 80, + 406, + 128 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 205, + 80, + 406, + 128 + ], + "spans": [ + { + "bbox": [ + 205, + 80, + 406, + 128 + ], + "score": 0.961, + "type": "image", + "image_path": "ae5902e9ec2fa29c05349103c337c534e1a146b5725255f610070875b832cd24.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 205, + 80, + 406, + 96.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 205, + 96.0, + 406, + 112.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 205, + 112.0, + 406, + 128.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 108, + 137, + 505, + 168 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 136, + 506, + 149 + ], + "spans": [ + { + "bbox": [ + 106, + 136, + 506, + 149 + ], + "score": 1.0, + "content": "Figure 6: Composing Movement Skills. Decision Diffuser can imitate individual running gaits using expert", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 147, + 505, + 159 + ], + "spans": [ + { + "bbox": [ + 105, + 147, + 505, + 159 + ], + "score": 1.0, + "content": "demonstrations and compose multiple different skills together during test time. The results are best illustrated by", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 157, + 343, + 169 + ], + "spans": [ + { + "bbox": [ + 106, + 157, + 343, + 169 + ], + "score": 1.0, + "content": "videos viewable at https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 107, + 174, + 505, + 262 + ], + "lines": [], + "index": 9.5, + "bbox_fs": [ + 105, + 173, + 506, + 263 + ], + "lines_deleted": true + }, + { + "type": "title", + "bbox": [ + 108, + 264, + 212, + 276 + ], + "lines": [ + { + "bbox": [ + 105, + 262, + 213, + 277 + ], + "spans": [ + { + "bbox": [ + 105, + 262, + 213, + 277 + ], + "score": 1.0, + "content": "5 RELATED WORK", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 14 + }, + { + "type": "text", + "bbox": [ + 107, + 280, + 506, + 488 + ], + "lines": [ + { + "bbox": [ + 105, + 278, + 506, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 278, + 506, + 291 + ], + "score": 1.0, + "content": "Diffusion Models Diffusion Models is proficient in learning generative models of image and text", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 289, + 505, + 302 + ], + "spans": [ + { + "bbox": [ + 106, + 289, + 505, + 302 + ], + "score": 1.0, + "content": "data (Saharia et al., 2022; Nichol et al., 2021; Nichol & Dhariwal, 2021). It formulates the data", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 300, + 506, + 313 + ], + "spans": [ + { + "bbox": [ + 105, + 300, + 506, + 313 + ], + "score": 1.0, + "content": "sampling process as an iterative denoising procedure (Sohl-Dickstein et al., 2015; Ho et al., 2020).", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 311, + 506, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 311, + 506, + 324 + ], + "score": 1.0, + "content": "The denoising procedure can be alternatively interpreted as parameterizing the gradients of the data", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 323, + 506, + 335 + ], + "spans": [ + { + "bbox": [ + 106, + 323, + 506, + 335 + ], + "score": 1.0, + "content": "distribution (Song et al., 2021) optimizing the score matching objective (Hyvarinen ¨ , 2005) and thus", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 333, + 506, + 346 + ], + "spans": [ + { + "bbox": [ + 105, + 333, + 506, + 346 + ], + "score": 1.0, + "content": "as a Energy-Based Model (Du & Mordatch, 2019; Nijkamp et al., 2019; Grathwohl et al., 2020).", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 345, + 506, + 357 + ], + "spans": [ + { + "bbox": [ + 106, + 345, + 506, + 357 + ], + "score": 1.0, + "content": "To generate data samples (eg: images) conditioned on some additional information (eg:text), prior", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 354, + 507, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 354, + 507, + 369 + ], + "score": 1.0, + "content": "works (Nichol & Dhariwal, 2021) have learned a classifier to facilitate the conditional sampling.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 367, + 506, + 379 + ], + "spans": [ + { + "bbox": [ + 106, + 367, + 506, + 379 + ], + "score": 1.0, + "content": "More recent works (Ho & Salimans, 2022) have argued to leverage gradients of an implicit classifier,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 378, + 506, + 389 + ], + "spans": [ + { + "bbox": [ + 106, + 378, + 506, + 389 + ], + "score": 1.0, + "content": "formed by the difference in score functions of a conditional and an unconditional model, to facilitate", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 389, + 506, + 401 + ], + "spans": [ + { + "bbox": [ + 106, + 389, + 506, + 401 + ], + "score": 1.0, + "content": "conditional sampling. The resulting classifier-free guidance has been shown to generate better", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 399, + 506, + 411 + ], + "spans": [ + { + "bbox": [ + 105, + 399, + 506, + 411 + ], + "score": 1.0, + "content": "conditional samples than classifier-based guidance. Recent works have also used diffusion models to", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 411, + 505, + 422 + ], + "spans": [ + { + "bbox": [ + 106, + 411, + 505, + 422 + ], + "score": 1.0, + "content": "imitate human behavior (Pearce et al., 2023) and to parameterize policy in offline RL (Wang et al.,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 421, + 506, + 434 + ], + "spans": [ + { + "bbox": [ + 106, + 421, + 506, + 434 + ], + "score": 1.0, + "content": "2022). Janner et al. (2022) generate trajectories consisting of states and actions with an unconditional", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 432, + 506, + 445 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 506, + 445 + ], + "score": 1.0, + "content": "diffusion model, therefore requiring a trained reward function on noisy state-action pairs. At inference,", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 443, + 505, + 456 + ], + "spans": [ + { + "bbox": [ + 106, + 443, + 505, + 456 + ], + "score": 1.0, + "content": "the estimated reward function guides the reverse diffusion process towards samples of high-return", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 454, + 506, + 467 + ], + "spans": [ + { + "bbox": [ + 106, + 454, + 506, + 467 + ], + "score": 1.0, + "content": "trajectories. In contrast, we do not train reward functions or diffusion processes separately, but rather", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 466, + 505, + 477 + ], + "spans": [ + { + "bbox": [ + 106, + 466, + 505, + 477 + ], + "score": 1.0, + "content": "model the trajectories in our dataset with a single, conditional generative model. This ensures that the", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 475, + 504, + 490 + ], + "spans": [ + { + "bbox": [ + 105, + 475, + 504, + 490 + ], + "score": 1.0, + "content": "sampling procedure of the learned diffusion process is the same at inference as it is during training.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 24, + "bbox_fs": [ + 105, + 278, + 507, + 490 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 491, + 505, + 568 + ], + "lines": [ + { + "bbox": [ + 106, + 489, + 506, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 506, + 502 + ], + "score": 1.0, + "content": "Reward Conditioned Policies Prior works (Kumar et al., 2019; Schmidhuber, 2019; Emmons", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 501, + 506, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 501, + 506, + 513 + ], + "score": 1.0, + "content": "et al., 2021; Chen et al., 2021) have studied learning of reward conditioned policies via reward", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "score": 1.0, + "content": "conditioned behavioral cloning. Chen et al. (2021) used a transformer (Vaswani et al., 2017) to model", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 522, + 506, + 536 + ], + "spans": [ + { + "bbox": [ + 105, + 522, + 506, + 536 + ], + "score": 1.0, + "content": "the reward conditioned policies and obtained a performance competitive with offline RL approaches.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 534, + 506, + 547 + ], + "spans": [ + { + "bbox": [ + 105, + 534, + 506, + 547 + ], + "score": 1.0, + "content": "Emmons et al. (2021) obtained similar performance as Chen et al. (2021) without using a transformer", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 545, + 505, + 558 + ], + "spans": [ + { + "bbox": [ + 106, + 545, + 505, + 558 + ], + "score": 1.0, + "content": "policy but relied on careful capacity tuning of MLP policy. In contrast, Decision Diffuser can also", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 555, + 347, + 569 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 347, + 569 + ], + "score": 1.0, + "content": "model constraints or skills and their resulting compositions.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 37, + "bbox_fs": [ + 105, + 489, + 506, + 569 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 574, + 191, + 586 + ], + "lines": [ + { + "bbox": [ + 105, + 572, + 192, + 588 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 192, + 588 + ], + "score": 1.0, + "content": "6 DISCUSSION", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 41 + }, + { + "type": "text", + "bbox": [ + 106, + 589, + 506, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 589, + 505, + 601 + ], + "spans": [ + { + "bbox": [ + 106, + 589, + 505, + 601 + ], + "score": 1.0, + "content": "We propose Decision Diffuser, a conditional generative model for sequential decision making. It", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 600, + 505, + 612 + ], + "spans": [ + { + "bbox": [ + 106, + 600, + 505, + 612 + ], + "score": 1.0, + "content": "frames offline sequential decision making as conditional generative modeling and sidesteps the need", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 506, + 624 + ], + "score": 1.0, + "content": "of reinforcement learning, thereby making the decision making pipeline simpler. By sampling for", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 505, + 635 + ], + "score": 1.0, + "content": "high returns, it is able to capture the best behaviors in the dataset and outperforms existing offline", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 633, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 506, + 646 + ], + "score": 1.0, + "content": "RL approaches on standard D4RL benchmarks. In addition to returns, it can also be conditioned", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 104, + 643, + 506, + 658 + ], + "spans": [ + { + "bbox": [ + 104, + 643, + 506, + 658 + ], + "score": 1.0, + "content": "on constraints or skills and can generate novel behaviors by flexibly combining constraints or", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 654, + 506, + 670 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 506, + 670 + ], + "score": 1.0, + "content": "composing skills during test time. In this work, we focused on offline sequential decision making,", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "score": 1.0, + "content": "thus circumventing the need for exploration. Using ideas from Zheng et al. (2022), future works", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 106, + 677, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 506, + 690 + ], + "score": 1.0, + "content": "could look into online fine-tuning of Decision Diffuser by leveraging entropy of the state-sequence", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "model for exploration. While our work focused on state based environments, it can be extended", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "to image based environments by performing the diffusion in latent space, rather than observation", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 104, + 708, + 507, + 724 + ], + "spans": [ + { + "bbox": [ + 104, + 708, + 507, + 724 + ], + "score": 1.0, + "content": "space, as done in Rombach et al. (2022). For a detailed discussion on limitations of Decision Diffuser,", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 106, + 721, + 218, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 218, + 732 + ], + "score": 1.0, + "content": "please refer to Appendix L.", + "type": "text" + } + ], + "index": 54 + } + ], + "index": 48, + "bbox_fs": [ + 104, + 589, + 507, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 82, + 225, + 93 + ], + "lines": [ + { + "bbox": [ + 107, + 81, + 226, + 96 + ], + "spans": [ + { + "bbox": [ + 107, + 81, + 226, + 96 + ], + "score": 1.0, + "content": "ACKNOWLEDGEMENTS", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 107, + 506, + 195 + ], + "lines": [ + { + "bbox": [ + 106, + 108, + 506, + 120 + ], + "spans": [ + { + "bbox": [ + 106, + 108, + 506, + 120 + ], + "score": 1.0, + "content": "The authors would like to thank Ofir Nachum, Anthony Simeonov and Richard Li for their helpful", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 118, + 507, + 131 + ], + "spans": [ + { + "bbox": [ + 105, + 118, + 507, + 131 + ], + "score": 1.0, + "content": "feedback on an earlier draft of the work; Jay Whang and Ge Yang for discussions on classifier-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 129, + 505, + 142 + ], + "spans": [ + { + "bbox": [ + 106, + 129, + 505, + 142 + ], + "score": 1.0, + "content": "free guidance; Gabe Margolis for helping with unitree experiments; Micheal Janner for providing", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 141, + 505, + 153 + ], + "spans": [ + { + "bbox": [ + 106, + 141, + 505, + 153 + ], + "score": 1.0, + "content": "visualization code for Kuka block stacking; and the members of Improbable AI Lab for discussions", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 151, + 506, + 164 + ], + "spans": [ + { + "bbox": [ + 105, + 151, + 506, + 164 + ], + "score": 1.0, + "content": "and helpful feedback. We thank MIT Supercloud and the Lincoln Laboratory Supercomputing Center", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 162, + 506, + 176 + ], + "spans": [ + { + "bbox": [ + 105, + 162, + 506, + 176 + ], + "score": 1.0, + "content": "for providing compute resources. This research was supported by an NSF graduate fellowship, a", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 173, + 507, + 186 + ], + "spans": [ + { + "bbox": [ + 105, + 173, + 507, + 186 + ], + "score": 1.0, + "content": "DARPA Machine Common Sense grant, a MURI grant, an MIT-IBM grant, and ARO W911NF-21-1-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 184, + 133, + 196 + ], + "spans": [ + { + "bbox": [ + 105, + 184, + 133, + 196 + ], + "score": 1.0, + "content": "0097.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 4.5 + }, + { + "type": "text", + "bbox": [ + 107, + 201, + 505, + 278 + ], + "lines": [ + { + "bbox": [ + 105, + 201, + 505, + 213 + ], + "spans": [ + { + "bbox": [ + 105, + 201, + 505, + 213 + ], + "score": 1.0, + "content": "This research was also partly sponsored by the United States Air Force Research Laboratory and the", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 211, + 505, + 225 + ], + "spans": [ + { + "bbox": [ + 106, + 211, + 505, + 225 + ], + "score": 1.0, + "content": "United States Air Force Artificial Intelligence Accelerator and was accomplished under Cooperative", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 223, + 506, + 235 + ], + "spans": [ + { + "bbox": [ + 105, + 223, + 506, + 235 + ], + "score": 1.0, + "content": "Agreement Number FA8750-19- 2-1000. The views and conclusions contained in this document", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 234, + 506, + 246 + ], + "spans": [ + { + "bbox": [ + 105, + 234, + 506, + 246 + ], + "score": 1.0, + "content": "are those of the authors and should not be interpreted as representing the official policies, either", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 244, + 506, + 258 + ], + "spans": [ + { + "bbox": [ + 105, + 244, + 506, + 258 + ], + "score": 1.0, + "content": "expressed or implied, of the United States Air Force or the U.S. Government. The U.S. Government", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 104, + 255, + 505, + 270 + ], + "spans": [ + { + "bbox": [ + 104, + 255, + 505, + 270 + ], + "score": 1.0, + "content": "is authorized to reproduce and distribute reprints for Government purposes, notwithstanding any", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 267, + 213, + 280 + ], + "spans": [ + { + "bbox": [ + 105, + 267, + 213, + 280 + ], + "score": 1.0, + "content": "copyright notation herein.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 12 + }, + { + "type": "title", + "bbox": [ + 108, + 295, + 242, + 307 + ], + "lines": [ + { + "bbox": [ + 106, + 294, + 243, + 310 + ], + "spans": [ + { + "bbox": [ + 106, + 294, + 243, + 310 + ], + "score": 1.0, + "content": "AUTHOR CONTRIBUTIONS", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 107, + 321, + 505, + 355 + ], + "lines": [ + { + "bbox": [ + 105, + 321, + 505, + 335 + ], + "spans": [ + { + "bbox": [ + 105, + 321, + 505, + 335 + ], + "score": 1.0, + "content": "Anurag Ajay conceived the framework of viewing decision-making as conditional diffusion", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 331, + 505, + 345 + ], + "spans": [ + { + "bbox": [ + 105, + 331, + 505, + 345 + ], + "score": 1.0, + "content": "generative modeling, implemented the Decision Diffuser algorithm, ran experiments on Offline", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 343, + 331, + 357 + ], + "spans": [ + { + "bbox": [ + 105, + 343, + 331, + 357 + ], + "score": 1.0, + "content": "RL and Skill Composition, and helped in paper writing.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 108, + 360, + 503, + 393 + ], + "lines": [ + { + "bbox": [ + 106, + 360, + 505, + 373 + ], + "spans": [ + { + "bbox": [ + 106, + 360, + 505, + 373 + ], + "score": 1.0, + "content": "Yilun Du helped in conceiving the framework of viewing decision-making as conditional diffusion", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 371, + 505, + 384 + ], + "spans": [ + { + "bbox": [ + 105, + 371, + 505, + 384 + ], + "score": 1.0, + "content": "generative modeling, ran experiments on Constraint Satisfaction, helped in paper writing and advised", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 381, + 142, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 381, + 142, + 396 + ], + "score": 1.0, + "content": "Anurag.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21 + }, + { + "type": "text", + "bbox": [ + 105, + 399, + 504, + 421 + ], + "lines": [ + { + "bbox": [ + 106, + 399, + 505, + 412 + ], + "spans": [ + { + "bbox": [ + 106, + 399, + 505, + 412 + ], + "score": 1.0, + "content": "Abhi Gupta helped in running experiments on Offline RL and Skill Composition, participated in", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 410, + 451, + 423 + ], + "spans": [ + { + "bbox": [ + 105, + 410, + 451, + 423 + ], + "score": 1.0, + "content": "research discussions, and played the leading role in paper writing and making figures.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 23.5 + }, + { + "type": "text", + "bbox": [ + 106, + 427, + 335, + 438 + ], + "lines": [ + { + "bbox": [ + 105, + 426, + 337, + 440 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 337, + 440 + ], + "score": 1.0, + "content": "Joshua Tenenbaum participated in research discussions.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 106, + 443, + 503, + 466 + ], + "lines": [ + { + "bbox": [ + 105, + 441, + 505, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 441, + 505, + 459 + ], + "score": 1.0, + "content": "Tommi Jaakkola participated in research discussions and suggested the experiment of classifying", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 455, + 164, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 164, + 468 + ], + "score": 1.0, + "content": "running gaits.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26.5 + }, + { + "type": "text", + "bbox": [ + 106, + 472, + 503, + 494 + ], + "lines": [ + { + "bbox": [ + 106, + 471, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 471, + 505, + 484 + ], + "score": 1.0, + "content": "Pulkit Agrawal was involved in research discussions, suggested experiments related to dynamic", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 104, + 482, + 473, + 497 + ], + "spans": [ + { + "bbox": [ + 104, + 482, + 473, + 497 + ], + "score": 1.0, + "content": "programming, provided feedback on writing, positioning of the work, and overall advising.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 28.5 + }, + { + "type": "title", + "bbox": [ + 108, + 511, + 175, + 523 + ], + "lines": [ + { + "bbox": [ + 106, + 511, + 176, + 524 + ], + "spans": [ + { + "bbox": [ + 106, + 511, + 176, + 524 + ], + "score": 1.0, + "content": "REFERENCES", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 106, + 537, + 504, + 570 + ], + "lines": [ + { + "bbox": [ + 105, + 536, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 536, + 505, + 550 + ], + "score": 1.0, + "content": "Pulkit Agrawal, Ashvin V Nair, Pieter Abbeel, Jitendra Malik, and Sergey Levine. Learning to poke", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 547, + 505, + 562 + ], + "spans": [ + { + "bbox": [ + 115, + 547, + 505, + 562 + ], + "score": 1.0, + "content": "by poking: Experiential learning of intuitive physics. Advances in neural information processing", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 114, + 559, + 192, + 571 + ], + "spans": [ + { + "bbox": [ + 114, + 559, + 192, + 571 + ], + "score": 1.0, + "content": "systems, 29, 2016.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32 + }, + { + "type": "text", + "bbox": [ + 105, + 580, + 504, + 603 + ], + "lines": [ + { + "bbox": [ + 106, + 580, + 505, + 593 + ], + "spans": [ + { + "bbox": [ + 106, + 580, + 505, + 593 + ], + "score": 1.0, + "content": "Anurag Ajay, Aviral Kumar, Pulkit Agrawal, Sergey Levine, and Ofir Nachum. Opal: Offline primitive", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 115, + 591, + 506, + 603 + ], + "spans": [ + { + "bbox": [ + 115, + 591, + 506, + 603 + ], + "score": 1.0, + "content": "discovery for accelerating offline reinforcement learning. arXiv preprint arXiv:2010.13611, 2020.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 34.5 + }, + { + "type": "text", + "bbox": [ + 108, + 611, + 505, + 689 + ], + "lines": [ + { + "bbox": [ + 106, + 612, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 106, + 612, + 506, + 624 + ], + "score": 1.0, + "content": "Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 116, + 623, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 116, + 623, + 505, + 635 + ], + "score": 1.0, + "content": "Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 633, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 115, + 633, + 506, + 646 + ], + "score": 1.0, + "content": "Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 114, + 643, + 507, + 658 + ], + "spans": [ + { + "bbox": [ + 114, + 643, + 507, + 658 + ], + "score": 1.0, + "content": "Jeffrey Wu, Clemens Winter, Chris Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 115, + 655, + 507, + 668 + ], + "spans": [ + { + "bbox": [ + 115, + 655, + 507, + 668 + ], + "score": 1.0, + "content": "Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever,", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 667, + 505, + 679 + ], + "spans": [ + { + "bbox": [ + 115, + 667, + 505, + 679 + ], + "score": 1.0, + "content": "and Dario Amodei. Language models are few-shot learners. In Advances in Neural Information", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 678, + 224, + 690 + ], + "spans": [ + { + "bbox": [ + 115, + 678, + 224, + 690 + ], + "score": 1.0, + "content": "Processing Systems, 2020.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 39 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "score": 1.0, + "content": "Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Misha Laskin, Pieter Abbeel,", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 115, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 115, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "Aravind Srinivas, and Igor Mordatch. Decision transformer: Reinforcement learning via sequence", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 720, + 411, + 733 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 411, + 733 + ], + "score": 1.0, + "content": "modeling. In Advances in Neural Information Processing Systems, 2021.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44 + } + ], + "page_idx": 9, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 752, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 82, + 225, + 93 + ], + "lines": [ + { + "bbox": [ + 107, + 81, + 226, + 96 + ], + "spans": [ + { + "bbox": [ + 107, + 81, + 226, + 96 + ], + "score": 1.0, + "content": "ACKNOWLEDGEMENTS", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 107, + 107, + 506, + 195 + ], + "lines": [ + { + "bbox": [ + 106, + 108, + 506, + 120 + ], + "spans": [ + { + "bbox": [ + 106, + 108, + 506, + 120 + ], + "score": 1.0, + "content": "The authors would like to thank Ofir Nachum, Anthony Simeonov and Richard Li for their helpful", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 118, + 507, + 131 + ], + "spans": [ + { + "bbox": [ + 105, + 118, + 507, + 131 + ], + "score": 1.0, + "content": "feedback on an earlier draft of the work; Jay Whang and Ge Yang for discussions on classifier-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 129, + 505, + 142 + ], + "spans": [ + { + "bbox": [ + 106, + 129, + 505, + 142 + ], + "score": 1.0, + "content": "free guidance; Gabe Margolis for helping with unitree experiments; Micheal Janner for providing", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 141, + 505, + 153 + ], + "spans": [ + { + "bbox": [ + 106, + 141, + 505, + 153 + ], + "score": 1.0, + "content": "visualization code for Kuka block stacking; and the members of Improbable AI Lab for discussions", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 151, + 506, + 164 + ], + "spans": [ + { + "bbox": [ + 105, + 151, + 506, + 164 + ], + "score": 1.0, + "content": "and helpful feedback. We thank MIT Supercloud and the Lincoln Laboratory Supercomputing Center", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 162, + 506, + 176 + ], + "spans": [ + { + "bbox": [ + 105, + 162, + 506, + 176 + ], + "score": 1.0, + "content": "for providing compute resources. This research was supported by an NSF graduate fellowship, a", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 173, + 507, + 186 + ], + "spans": [ + { + "bbox": [ + 105, + 173, + 507, + 186 + ], + "score": 1.0, + "content": "DARPA Machine Common Sense grant, a MURI grant, an MIT-IBM grant, and ARO W911NF-21-1-", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 184, + 133, + 196 + ], + "spans": [ + { + "bbox": [ + 105, + 184, + 133, + 196 + ], + "score": 1.0, + "content": "0097.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 4.5, + "bbox_fs": [ + 105, + 108, + 507, + 196 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 201, + 505, + 278 + ], + "lines": [ + { + "bbox": [ + 105, + 201, + 505, + 213 + ], + "spans": [ + { + "bbox": [ + 105, + 201, + 505, + 213 + ], + "score": 1.0, + "content": "This research was also partly sponsored by the United States Air Force Research Laboratory and the", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 211, + 505, + 225 + ], + "spans": [ + { + "bbox": [ + 106, + 211, + 505, + 225 + ], + "score": 1.0, + "content": "United States Air Force Artificial Intelligence Accelerator and was accomplished under Cooperative", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 223, + 506, + 235 + ], + "spans": [ + { + "bbox": [ + 105, + 223, + 506, + 235 + ], + "score": 1.0, + "content": "Agreement Number FA8750-19- 2-1000. The views and conclusions contained in this document", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 234, + 506, + 246 + ], + "spans": [ + { + "bbox": [ + 105, + 234, + 506, + 246 + ], + "score": 1.0, + "content": "are those of the authors and should not be interpreted as representing the official policies, either", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 244, + 506, + 258 + ], + "spans": [ + { + "bbox": [ + 105, + 244, + 506, + 258 + ], + "score": 1.0, + "content": "expressed or implied, of the United States Air Force or the U.S. Government. The U.S. Government", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 104, + 255, + 505, + 270 + ], + "spans": [ + { + "bbox": [ + 104, + 255, + 505, + 270 + ], + "score": 1.0, + "content": "is authorized to reproduce and distribute reprints for Government purposes, notwithstanding any", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 267, + 213, + 280 + ], + "spans": [ + { + "bbox": [ + 105, + 267, + 213, + 280 + ], + "score": 1.0, + "content": "copyright notation herein.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 12, + "bbox_fs": [ + 104, + 201, + 506, + 280 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 295, + 242, + 307 + ], + "lines": [ + { + "bbox": [ + 106, + 294, + 243, + 310 + ], + "spans": [ + { + "bbox": [ + 106, + 294, + 243, + 310 + ], + "score": 1.0, + "content": "AUTHOR CONTRIBUTIONS", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 107, + 321, + 505, + 355 + ], + "lines": [ + { + "bbox": [ + 105, + 321, + 505, + 335 + ], + "spans": [ + { + "bbox": [ + 105, + 321, + 505, + 335 + ], + "score": 1.0, + "content": "Anurag Ajay conceived the framework of viewing decision-making as conditional diffusion", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 331, + 505, + 345 + ], + "spans": [ + { + "bbox": [ + 105, + 331, + 505, + 345 + ], + "score": 1.0, + "content": "generative modeling, implemented the Decision Diffuser algorithm, ran experiments on Offline", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 343, + 331, + 357 + ], + "spans": [ + { + "bbox": [ + 105, + 343, + 331, + 357 + ], + "score": 1.0, + "content": "RL and Skill Composition, and helped in paper writing.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18, + "bbox_fs": [ + 105, + 321, + 505, + 357 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 360, + 503, + 393 + ], + "lines": [ + { + "bbox": [ + 106, + 360, + 505, + 373 + ], + "spans": [ + { + "bbox": [ + 106, + 360, + 505, + 373 + ], + "score": 1.0, + "content": "Yilun Du helped in conceiving the framework of viewing decision-making as conditional diffusion", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 371, + 505, + 384 + ], + "spans": [ + { + "bbox": [ + 105, + 371, + 505, + 384 + ], + "score": 1.0, + "content": "generative modeling, ran experiments on Constraint Satisfaction, helped in paper writing and advised", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 381, + 142, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 381, + 142, + 396 + ], + "score": 1.0, + "content": "Anurag.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21, + "bbox_fs": [ + 105, + 360, + 505, + 396 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 399, + 504, + 421 + ], + "lines": [ + { + "bbox": [ + 106, + 399, + 505, + 412 + ], + "spans": [ + { + "bbox": [ + 106, + 399, + 505, + 412 + ], + "score": 1.0, + "content": "Abhi Gupta helped in running experiments on Offline RL and Skill Composition, participated in", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 410, + 451, + 423 + ], + "spans": [ + { + "bbox": [ + 105, + 410, + 451, + 423 + ], + "score": 1.0, + "content": "research discussions, and played the leading role in paper writing and making figures.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 23.5, + "bbox_fs": [ + 105, + 399, + 505, + 423 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 427, + 335, + 438 + ], + "lines": [ + { + "bbox": [ + 105, + 426, + 337, + 440 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 337, + 440 + ], + "score": 1.0, + "content": "Joshua Tenenbaum participated in research discussions.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25, + "bbox_fs": [ + 105, + 426, + 337, + 440 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 443, + 503, + 466 + ], + "lines": [ + { + "bbox": [ + 105, + 441, + 505, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 441, + 505, + 459 + ], + "score": 1.0, + "content": "Tommi Jaakkola participated in research discussions and suggested the experiment of classifying", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 455, + 164, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 164, + 468 + ], + "score": 1.0, + "content": "running gaits.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26.5, + "bbox_fs": [ + 105, + 441, + 505, + 468 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 472, + 503, + 494 + ], + "lines": [ + { + "bbox": [ + 106, + 471, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 471, + 505, + 484 + ], + "score": 1.0, + "content": "Pulkit Agrawal was involved in research discussions, suggested experiments related to dynamic", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 104, + 482, + 473, + 497 + ], + "spans": [ + { + "bbox": [ + 104, + 482, + 473, + 497 + ], + "score": 1.0, + "content": "programming, provided feedback on writing, positioning of the work, and overall advising.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 28.5, + "bbox_fs": [ + 104, + 471, + 505, + 497 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 511, + 175, + 523 + ], + "lines": [ + { + "bbox": [ + 106, + 511, + 176, + 524 + ], + "spans": [ + { + "bbox": [ + 106, + 511, + 176, + 524 + ], + "score": 1.0, + "content": "REFERENCES", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 106, + 537, + 504, + 570 + ], + "lines": [ + { + "bbox": [ + 105, + 536, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 536, + 505, + 550 + ], + "score": 1.0, + "content": "Pulkit Agrawal, Ashvin V Nair, Pieter Abbeel, Jitendra Malik, and Sergey Levine. Learning to poke", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 547, + 505, + 562 + ], + "spans": [ + { + "bbox": [ + 115, + 547, + 505, + 562 + ], + "score": 1.0, + "content": "by poking: Experiential learning of intuitive physics. Advances in neural information processing", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 114, + 559, + 192, + 571 + ], + "spans": [ + { + "bbox": [ + 114, + 559, + 192, + 571 + ], + "score": 1.0, + "content": "systems, 29, 2016.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32, + "bbox_fs": [ + 105, + 536, + 505, + 571 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 580, + 504, + 603 + ], + "lines": [ + { + "bbox": [ + 106, + 580, + 505, + 593 + ], + "spans": [ + { + "bbox": [ + 106, + 580, + 505, + 593 + ], + "score": 1.0, + "content": "Anurag Ajay, Aviral Kumar, Pulkit Agrawal, Sergey Levine, and Ofir Nachum. Opal: Offline primitive", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 115, + 591, + 506, + 603 + ], + "spans": [ + { + "bbox": [ + 115, + 591, + 506, + 603 + ], + "score": 1.0, + "content": "discovery for accelerating offline reinforcement learning. arXiv preprint arXiv:2010.13611, 2020.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 34.5, + "bbox_fs": [ + 106, + 580, + 506, + 603 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 611, + 505, + 689 + ], + "lines": [ + { + "bbox": [ + 106, + 612, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 106, + 612, + 506, + 624 + ], + "score": 1.0, + "content": "Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 116, + 623, + 505, + 635 + ], + "spans": [ + { + "bbox": [ + 116, + 623, + 505, + 635 + ], + "score": 1.0, + "content": "Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 115, + 633, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 115, + 633, + 506, + 646 + ], + "score": 1.0, + "content": "Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 114, + 643, + 507, + 658 + ], + "spans": [ + { + "bbox": [ + 114, + 643, + 507, + 658 + ], + "score": 1.0, + "content": "Jeffrey Wu, Clemens Winter, Chris Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 115, + 655, + 507, + 668 + ], + "spans": [ + { + "bbox": [ + 115, + 655, + 507, + 668 + ], + "score": 1.0, + "content": "Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever,", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 667, + 505, + 679 + ], + "spans": [ + { + "bbox": [ + 115, + 667, + 505, + 679 + ], + "score": 1.0, + "content": "and Dario Amodei. Language models are few-shot learners. In Advances in Neural Information", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 678, + 224, + 690 + ], + "spans": [ + { + "bbox": [ + 115, + 678, + 224, + 690 + ], + "score": 1.0, + "content": "Processing Systems, 2020.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 39, + "bbox_fs": [ + 106, + 612, + 507, + 690 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "score": 1.0, + "content": "Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Misha Laskin, Pieter Abbeel,", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 115, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 115, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "Aravind Srinivas, and Igor Mordatch. Decision transformer: Reinforcement learning via sequence", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 720, + 411, + 733 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 411, + 733 + ], + "score": 1.0, + "content": "modeling. In Advances in Neural Information Processing Systems, 2021.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44, + "bbox_fs": [ + 105, + 699, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "Robert Dadashi, Shideh Rezaeifar, Nino Vieillard, Leonard Hussenot, Olivier Pietquin, and Matthieu ´", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 506, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 506, + 106 + ], + "score": 1.0, + "content": "Geist. Offline reinforcement learning with pseudometric learning. arXiv preprint arXiv:2103.01948,", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 104, + 142, + 116 + ], + "spans": [ + { + "bbox": [ + 115, + 104, + 142, + 116 + ], + "score": 1.0, + "content": "2021.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 105, + 123, + 505, + 146 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 505, + 136 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 505, + 136 + ], + "score": 1.0, + "content": "Yilun Du and Igor Mordatch. Implicit generation and generalization in energy-based models. In", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 114, + 134, + 355, + 146 + ], + "spans": [ + { + "bbox": [ + 114, + 134, + 355, + 146 + ], + "score": 1.0, + "content": "Advances in Neural Information Processing Systems, 2019.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 108, + 152, + 503, + 186 + ], + "lines": [ + { + "bbox": [ + 106, + 152, + 505, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 152, + 505, + 165 + ], + "score": 1.0, + "content": "Gabriel Dulac-Arnold, Nir Levine, Daniel J Mankowitz, Jerry Li, Cosmin Paduraru, Sven Gowal,", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 163, + 505, + 177 + ], + "spans": [ + { + "bbox": [ + 115, + 163, + 505, + 177 + ], + "score": 1.0, + "content": "and Todd Hester. Challenges of real-world reinforcement learning: definitions, benchmarks and", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 116, + 174, + 338, + 188 + ], + "spans": [ + { + "bbox": [ + 116, + 174, + 338, + 188 + ], + "score": 1.0, + "content": "analysis. Machine Learning, 110(9):2419–2468, 2021.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6 + }, + { + "type": "text", + "bbox": [ + 106, + 192, + 504, + 216 + ], + "lines": [ + { + "bbox": [ + 106, + 192, + 505, + 205 + ], + "spans": [ + { + "bbox": [ + 106, + 192, + 505, + 205 + ], + "score": 1.0, + "content": "Scott Emmons, Benjamin Eysenbach, Ilya Kostrikov, and Sergey Levine. Rvs: What is essential for", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 115, + 204, + 417, + 216 + ], + "spans": [ + { + "bbox": [ + 115, + 204, + 417, + 216 + ], + "score": 1.0, + "content": "offline rl via supervised learning? arXiv preprint arXiv:2112.10751, 2021.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 106, + 222, + 504, + 245 + ], + "lines": [ + { + "bbox": [ + 105, + 221, + 505, + 235 + ], + "spans": [ + { + "bbox": [ + 105, + 221, + 505, + 235 + ], + "score": 1.0, + "content": "Justin Fu, Aviral Kumar, Ofir Nachum, George Tucker, and Sergey Levine. D4RL: Datasets for deep", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 116, + 233, + 423, + 246 + ], + "spans": [ + { + "bbox": [ + 116, + 233, + 423, + 246 + ], + "score": 1.0, + "content": "data-driven reinforcement learning. arXiv preprint arXiv:2004.07219, 2020.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5 + }, + { + "type": "text", + "bbox": [ + 107, + 251, + 505, + 275 + ], + "lines": [ + { + "bbox": [ + 106, + 251, + 506, + 264 + ], + "spans": [ + { + "bbox": [ + 106, + 251, + 506, + 264 + ], + "score": 1.0, + "content": "Scott Fujimoto, Herke Hoof, and David Meger. Addressing function approximation error in actor-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 116, + 263, + 500, + 275 + ], + "spans": [ + { + "bbox": [ + 116, + 263, + 500, + 275 + ], + "score": 1.0, + "content": "critic methods. In International conference on machine learning, pp. 1587–1596. PMLR, 2018.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12.5 + }, + { + "type": "text", + "bbox": [ + 106, + 281, + 504, + 304 + ], + "lines": [ + { + "bbox": [ + 106, + 280, + 505, + 294 + ], + "spans": [ + { + "bbox": [ + 106, + 280, + 505, + 294 + ], + "score": 1.0, + "content": "Scott Fujimoto, David Meger, and Doina Precup. Off-policy deep reinforcement learning without", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 115, + 293, + 397, + 305 + ], + "spans": [ + { + "bbox": [ + 115, + 293, + 397, + 305 + ], + "score": 1.0, + "content": "exploration. In International Conference on Machine Learning, 2019.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5 + }, + { + "type": "text", + "bbox": [ + 106, + 310, + 505, + 334 + ], + "lines": [ + { + "bbox": [ + 105, + 309, + 506, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 309, + 506, + 324 + ], + "score": 1.0, + "content": "Dibya Ghosh, Anurag Ajay, Pulkit Agrawal, and Sergey Levine. Offline rl policies should be trained", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 322, + 505, + 334 + ], + "spans": [ + { + "bbox": [ + 115, + 322, + 505, + 334 + ], + "score": 1.0, + "content": "to be adaptive. In International Conference on Machine Learning, pp. 7513–7530. PMLR, 2022.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5 + }, + { + "type": "text", + "bbox": [ + 109, + 340, + 505, + 374 + ], + "lines": [ + { + "bbox": [ + 107, + 340, + 506, + 353 + ], + "spans": [ + { + "bbox": [ + 107, + 340, + 506, + 353 + ], + "score": 1.0, + "content": "Will Grathwohl, Kuan-Chieh Wang, Joern-Henrik Jacobsen, David Duvenaud, and Richard Zemel.", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 351, + 506, + 365 + ], + "spans": [ + { + "bbox": [ + 115, + 351, + 506, + 365 + ], + "score": 1.0, + "content": "Learning the stein discrepancy for training and evaluating energy-based models without sampling.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 116, + 363, + 346, + 375 + ], + "spans": [ + { + "bbox": [ + 116, + 363, + 346, + 375 + ], + "score": 1.0, + "content": "In International Conference on Machine Learning, 2020.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 108, + 380, + 504, + 415 + ], + "lines": [ + { + "bbox": [ + 105, + 380, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 380, + 506, + 394 + ], + "score": 1.0, + "content": "Abhishek Gupta, Russell Mendonca, YuXuan Liu, Pieter Abbeel, and Sergey Levine. Meta-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 116, + 393, + 505, + 404 + ], + "spans": [ + { + "bbox": [ + 116, + 393, + 505, + 404 + ], + "score": 1.0, + "content": "reinforcement learning of structured exploration strategies. Advances in neural information", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 405, + 237, + 415 + ], + "spans": [ + { + "bbox": [ + 115, + 405, + 237, + 415 + ], + "score": 1.0, + "content": "processing systems, 31, 2018.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 107, + 421, + 504, + 455 + ], + "lines": [ + { + "bbox": [ + 105, + 420, + 505, + 435 + ], + "spans": [ + { + "bbox": [ + 105, + 420, + 505, + 435 + ], + "score": 1.0, + "content": "Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, and Sergey Levine. Soft actor-critic: Off-policy", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 116, + 433, + 505, + 445 + ], + "spans": [ + { + "bbox": [ + 116, + 433, + 505, + 445 + ], + "score": 1.0, + "content": "maximum entropy deep reinforcement learning with a stochastic actor. In International Conference", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 116, + 444, + 232, + 456 + ], + "spans": [ + { + "bbox": [ + 116, + 444, + 232, + 456 + ], + "score": 1.0, + "content": "on Machine Learning, 2018.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 106, + 461, + 504, + 485 + ], + "lines": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "score": 1.0, + "content": "Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 114, + 472, + 143, + 486 + ], + "spans": [ + { + "bbox": [ + 114, + 472, + 143, + 486 + ], + "score": 1.0, + "content": "2022.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 107, + 491, + 504, + 514 + ], + "lines": [ + { + "bbox": [ + 105, + 491, + 505, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 491, + 505, + 504 + ], + "score": 1.0, + "content": "Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. In Advances in", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 115, + 502, + 303, + 515 + ], + "spans": [ + { + "bbox": [ + 115, + 502, + 303, + 515 + ], + "score": 1.0, + "content": "Neural Information Processing Systems, 2020.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5 + }, + { + "type": "text", + "bbox": [ + 106, + 520, + 504, + 544 + ], + "lines": [ + { + "bbox": [ + 106, + 520, + 506, + 535 + ], + "spans": [ + { + "bbox": [ + 106, + 520, + 506, + 535 + ], + "score": 1.0, + "content": "Aapo Hyvarinen. Estimation of non-normalized statistical models by score matching. ¨ Journal of", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 532, + 258, + 544 + ], + "spans": [ + { + "bbox": [ + 115, + 532, + 258, + 544 + ], + "score": 1.0, + "content": "Machine Learning Research, 2005.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 107, + 550, + 503, + 574 + ], + "lines": [ + { + "bbox": [ + 105, + 549, + 505, + 564 + ], + "spans": [ + { + "bbox": [ + 105, + 549, + 505, + 564 + ], + "score": 1.0, + "content": "Michael Janner, Qiyang Li, and Sergey Levine. Offline reinforcement learning as one big sequence", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 116, + 561, + 446, + 574 + ], + "spans": [ + { + "bbox": [ + 116, + 561, + 446, + 574 + ], + "score": 1.0, + "content": "modeling problem. In Advances in Neural Information Processing Systems, 2021.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5 + }, + { + "type": "text", + "bbox": [ + 107, + 579, + 504, + 603 + ], + "lines": [ + { + "bbox": [ + 105, + 580, + 506, + 593 + ], + "spans": [ + { + "bbox": [ + 105, + 580, + 506, + 593 + ], + "score": 1.0, + "content": "Michael Janner, Yilun Du, Joshua Tenenbaum, and Sergey Levine. Planning with diffusion for", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 116, + 591, + 458, + 604 + ], + "spans": [ + { + "bbox": [ + 116, + 591, + 458, + 604 + ], + "score": 1.0, + "content": "flexible behavior synthesis. In International Conference on Machine Learning, 2022.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5 + }, + { + "type": "text", + "bbox": [ + 106, + 609, + 506, + 643 + ], + "lines": [ + { + "bbox": [ + 106, + 609, + 507, + 622 + ], + "spans": [ + { + "bbox": [ + 106, + 609, + 507, + 622 + ], + "score": 1.0, + "content": "Rahul Kidambi, Aravind Rajeswaran, Praneeth Netrapalli, and Thorsten Joachims. MOReL: Model-", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 116, + 620, + 507, + 634 + ], + "spans": [ + { + "bbox": [ + 116, + 620, + 507, + 634 + ], + "score": 1.0, + "content": "based offline reinforcement learning. In Advances in Neural Information Processing Systems,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 631, + 142, + 643 + ], + "spans": [ + { + "bbox": [ + 115, + 631, + 142, + 643 + ], + "score": 1.0, + "content": "2020.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 106, + 650, + 504, + 673 + ], + "lines": [ + { + "bbox": [ + 105, + 649, + 506, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 506, + 663 + ], + "score": 1.0, + "content": "Diederik Kingma, Tim Salimans, Ben Poole, and Jonathan Ho. Variational diffusion models. Advances", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 662, + 383, + 673 + ], + "spans": [ + { + "bbox": [ + 115, + 662, + 383, + 673 + ], + "score": 1.0, + "content": "in neural information processing systems, 34:21696–21707, 2021.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 40.5 + }, + { + "type": "text", + "bbox": [ + 107, + 680, + 504, + 703 + ], + "lines": [ + { + "bbox": [ + 106, + 680, + 505, + 693 + ], + "spans": [ + { + "bbox": [ + 106, + 680, + 505, + 693 + ], + "score": 1.0, + "content": "Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In International", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 116, + 691, + 310, + 703 + ], + "spans": [ + { + "bbox": [ + 116, + 691, + 310, + 703 + ], + "score": 1.0, + "content": "Conference on Learning Representations, 2015.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42.5 + }, + { + "type": "text", + "bbox": [ + 107, + 709, + 503, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 708, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 505, + 723 + ], + "score": 1.0, + "content": "Ilya Kostrikov, Jonathan Tompson, Rob Fergus, and Ofir Nachum. Offline reinforcement learning", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 721, + 453, + 732 + ], + "spans": [ + { + "bbox": [ + 115, + 721, + 453, + 732 + ], + "score": 1.0, + "content": "with fisher divergence critic regularization. arXiv preprint arXiv:2103.08050, 2021.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44.5 + } + ], + "page_idx": 10, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 761 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 765 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 765 + ], + "score": 1.0, + "content": "11", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "Robert Dadashi, Shideh Rezaeifar, Nino Vieillard, Leonard Hussenot, Olivier Pietquin, and Matthieu ´", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 93, + 506, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 93, + 506, + 106 + ], + "score": 1.0, + "content": "Geist. Offline reinforcement learning with pseudometric learning. arXiv preprint arXiv:2103.01948,", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 104, + 142, + 116 + ], + "spans": [ + { + "bbox": [ + 115, + 104, + 142, + 116 + ], + "score": 1.0, + "content": "2021.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1, + "bbox_fs": [ + 105, + 82, + 506, + 116 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 123, + 505, + 146 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 505, + 136 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 505, + 136 + ], + "score": 1.0, + "content": "Yilun Du and Igor Mordatch. Implicit generation and generalization in energy-based models. In", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 114, + 134, + 355, + 146 + ], + "spans": [ + { + "bbox": [ + 114, + 134, + 355, + 146 + ], + "score": 1.0, + "content": "Advances in Neural Information Processing Systems, 2019.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5, + "bbox_fs": [ + 106, + 123, + 505, + 146 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 152, + 503, + 186 + ], + "lines": [ + { + "bbox": [ + 106, + 152, + 505, + 165 + ], + "spans": [ + { + "bbox": [ + 106, + 152, + 505, + 165 + ], + "score": 1.0, + "content": "Gabriel Dulac-Arnold, Nir Levine, Daniel J Mankowitz, Jerry Li, Cosmin Paduraru, Sven Gowal,", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 163, + 505, + 177 + ], + "spans": [ + { + "bbox": [ + 115, + 163, + 505, + 177 + ], + "score": 1.0, + "content": "and Todd Hester. Challenges of real-world reinforcement learning: definitions, benchmarks and", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 116, + 174, + 338, + 188 + ], + "spans": [ + { + "bbox": [ + 116, + 174, + 338, + 188 + ], + "score": 1.0, + "content": "analysis. Machine Learning, 110(9):2419–2468, 2021.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6, + "bbox_fs": [ + 106, + 152, + 505, + 188 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 192, + 504, + 216 + ], + "lines": [ + { + "bbox": [ + 106, + 192, + 505, + 205 + ], + "spans": [ + { + "bbox": [ + 106, + 192, + 505, + 205 + ], + "score": 1.0, + "content": "Scott Emmons, Benjamin Eysenbach, Ilya Kostrikov, and Sergey Levine. Rvs: What is essential for", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 115, + 204, + 417, + 216 + ], + "spans": [ + { + "bbox": [ + 115, + 204, + 417, + 216 + ], + "score": 1.0, + "content": "offline rl via supervised learning? arXiv preprint arXiv:2112.10751, 2021.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8.5, + "bbox_fs": [ + 106, + 192, + 505, + 216 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 222, + 504, + 245 + ], + "lines": [ + { + "bbox": [ + 105, + 221, + 505, + 235 + ], + "spans": [ + { + "bbox": [ + 105, + 221, + 505, + 235 + ], + "score": 1.0, + "content": "Justin Fu, Aviral Kumar, Ofir Nachum, George Tucker, and Sergey Levine. D4RL: Datasets for deep", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 116, + 233, + 423, + 246 + ], + "spans": [ + { + "bbox": [ + 116, + 233, + 423, + 246 + ], + "score": 1.0, + "content": "data-driven reinforcement learning. arXiv preprint arXiv:2004.07219, 2020.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10.5, + "bbox_fs": [ + 105, + 221, + 505, + 246 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 251, + 505, + 275 + ], + "lines": [ + { + "bbox": [ + 106, + 251, + 506, + 264 + ], + "spans": [ + { + "bbox": [ + 106, + 251, + 506, + 264 + ], + "score": 1.0, + "content": "Scott Fujimoto, Herke Hoof, and David Meger. Addressing function approximation error in actor-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 116, + 263, + 500, + 275 + ], + "spans": [ + { + "bbox": [ + 116, + 263, + 500, + 275 + ], + "score": 1.0, + "content": "critic methods. In International conference on machine learning, pp. 1587–1596. PMLR, 2018.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12.5, + "bbox_fs": [ + 106, + 251, + 506, + 275 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 281, + 504, + 304 + ], + "lines": [ + { + "bbox": [ + 106, + 280, + 505, + 294 + ], + "spans": [ + { + "bbox": [ + 106, + 280, + 505, + 294 + ], + "score": 1.0, + "content": "Scott Fujimoto, David Meger, and Doina Precup. Off-policy deep reinforcement learning without", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 115, + 293, + 397, + 305 + ], + "spans": [ + { + "bbox": [ + 115, + 293, + 397, + 305 + ], + "score": 1.0, + "content": "exploration. In International Conference on Machine Learning, 2019.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5, + "bbox_fs": [ + 106, + 280, + 505, + 305 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 310, + 505, + 334 + ], + "lines": [ + { + "bbox": [ + 105, + 309, + 506, + 324 + ], + "spans": [ + { + "bbox": [ + 105, + 309, + 506, + 324 + ], + "score": 1.0, + "content": "Dibya Ghosh, Anurag Ajay, Pulkit Agrawal, and Sergey Levine. Offline rl policies should be trained", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 322, + 505, + 334 + ], + "spans": [ + { + "bbox": [ + 115, + 322, + 505, + 334 + ], + "score": 1.0, + "content": "to be adaptive. In International Conference on Machine Learning, pp. 7513–7530. PMLR, 2022.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5, + "bbox_fs": [ + 105, + 309, + 506, + 334 + ] + }, + { + "type": "text", + "bbox": [ + 109, + 340, + 505, + 374 + ], + "lines": [ + { + "bbox": [ + 107, + 340, + 506, + 353 + ], + "spans": [ + { + "bbox": [ + 107, + 340, + 506, + 353 + ], + "score": 1.0, + "content": "Will Grathwohl, Kuan-Chieh Wang, Joern-Henrik Jacobsen, David Duvenaud, and Richard Zemel.", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 351, + 506, + 365 + ], + "spans": [ + { + "bbox": [ + 115, + 351, + 506, + 365 + ], + "score": 1.0, + "content": "Learning the stein discrepancy for training and evaluating energy-based models without sampling.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 116, + 363, + 346, + 375 + ], + "spans": [ + { + "bbox": [ + 116, + 363, + 346, + 375 + ], + "score": 1.0, + "content": "In International Conference on Machine Learning, 2020.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19, + "bbox_fs": [ + 107, + 340, + 506, + 375 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 380, + 504, + 415 + ], + "lines": [ + { + "bbox": [ + 105, + 380, + 506, + 394 + ], + "spans": [ + { + "bbox": [ + 105, + 380, + 506, + 394 + ], + "score": 1.0, + "content": "Abhishek Gupta, Russell Mendonca, YuXuan Liu, Pieter Abbeel, and Sergey Levine. Meta-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 116, + 393, + 505, + 404 + ], + "spans": [ + { + "bbox": [ + 116, + 393, + 505, + 404 + ], + "score": 1.0, + "content": "reinforcement learning of structured exploration strategies. Advances in neural information", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 405, + 237, + 415 + ], + "spans": [ + { + "bbox": [ + 115, + 405, + 237, + 415 + ], + "score": 1.0, + "content": "processing systems, 31, 2018.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22, + "bbox_fs": [ + 105, + 380, + 506, + 415 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 421, + 504, + 455 + ], + "lines": [ + { + "bbox": [ + 105, + 420, + 505, + 435 + ], + "spans": [ + { + "bbox": [ + 105, + 420, + 505, + 435 + ], + "score": 1.0, + "content": "Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, and Sergey Levine. Soft actor-critic: Off-policy", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 116, + 433, + 505, + 445 + ], + "spans": [ + { + "bbox": [ + 116, + 433, + 505, + 445 + ], + "score": 1.0, + "content": "maximum entropy deep reinforcement learning with a stochastic actor. In International Conference", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 116, + 444, + 232, + 456 + ], + "spans": [ + { + "bbox": [ + 116, + 444, + 232, + 456 + ], + "score": 1.0, + "content": "on Machine Learning, 2018.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25, + "bbox_fs": [ + 105, + 420, + 505, + 456 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 461, + 504, + 485 + ], + "lines": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "score": 1.0, + "content": "Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 114, + 472, + 143, + 486 + ], + "spans": [ + { + "bbox": [ + 114, + 472, + 143, + 486 + ], + "score": 1.0, + "content": "2022.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27.5, + "bbox_fs": [ + 105, + 461, + 505, + 486 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 491, + 504, + 514 + ], + "lines": [ + { + "bbox": [ + 105, + 491, + 505, + 504 + ], + "spans": [ + { + "bbox": [ + 105, + 491, + 505, + 504 + ], + "score": 1.0, + "content": "Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. In Advances in", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 115, + 502, + 303, + 515 + ], + "spans": [ + { + "bbox": [ + 115, + 502, + 303, + 515 + ], + "score": 1.0, + "content": "Neural Information Processing Systems, 2020.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29.5, + "bbox_fs": [ + 105, + 491, + 505, + 515 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 520, + 504, + 544 + ], + "lines": [ + { + "bbox": [ + 106, + 520, + 506, + 535 + ], + "spans": [ + { + "bbox": [ + 106, + 520, + 506, + 535 + ], + "score": 1.0, + "content": "Aapo Hyvarinen. Estimation of non-normalized statistical models by score matching. ¨ Journal of", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 532, + 258, + 544 + ], + "spans": [ + { + "bbox": [ + 115, + 532, + 258, + 544 + ], + "score": 1.0, + "content": "Machine Learning Research, 2005.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5, + "bbox_fs": [ + 106, + 520, + 506, + 544 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 550, + 503, + 574 + ], + "lines": [ + { + "bbox": [ + 105, + 549, + 505, + 564 + ], + "spans": [ + { + "bbox": [ + 105, + 549, + 505, + 564 + ], + "score": 1.0, + "content": "Michael Janner, Qiyang Li, and Sergey Levine. Offline reinforcement learning as one big sequence", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 116, + 561, + 446, + 574 + ], + "spans": [ + { + "bbox": [ + 116, + 561, + 446, + 574 + ], + "score": 1.0, + "content": "modeling problem. In Advances in Neural Information Processing Systems, 2021.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5, + "bbox_fs": [ + 105, + 549, + 505, + 574 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 579, + 504, + 603 + ], + "lines": [ + { + "bbox": [ + 105, + 580, + 506, + 593 + ], + "spans": [ + { + "bbox": [ + 105, + 580, + 506, + 593 + ], + "score": 1.0, + "content": "Michael Janner, Yilun Du, Joshua Tenenbaum, and Sergey Levine. Planning with diffusion for", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 116, + 591, + 458, + 604 + ], + "spans": [ + { + "bbox": [ + 116, + 591, + 458, + 604 + ], + "score": 1.0, + "content": "flexible behavior synthesis. In International Conference on Machine Learning, 2022.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5, + "bbox_fs": [ + 105, + 580, + 506, + 604 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 609, + 506, + 643 + ], + "lines": [ + { + "bbox": [ + 106, + 609, + 507, + 622 + ], + "spans": [ + { + "bbox": [ + 106, + 609, + 507, + 622 + ], + "score": 1.0, + "content": "Rahul Kidambi, Aravind Rajeswaran, Praneeth Netrapalli, and Thorsten Joachims. MOReL: Model-", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 116, + 620, + 507, + 634 + ], + "spans": [ + { + "bbox": [ + 116, + 620, + 507, + 634 + ], + "score": 1.0, + "content": "based offline reinforcement learning. In Advances in Neural Information Processing Systems,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 631, + 142, + 643 + ], + "spans": [ + { + "bbox": [ + 115, + 631, + 142, + 643 + ], + "score": 1.0, + "content": "2020.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38, + "bbox_fs": [ + 106, + 609, + 507, + 643 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 650, + 504, + 673 + ], + "lines": [ + { + "bbox": [ + 105, + 649, + 506, + 663 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 506, + 663 + ], + "score": 1.0, + "content": "Diederik Kingma, Tim Salimans, Ben Poole, and Jonathan Ho. Variational diffusion models. Advances", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 662, + 383, + 673 + ], + "spans": [ + { + "bbox": [ + 115, + 662, + 383, + 673 + ], + "score": 1.0, + "content": "in neural information processing systems, 34:21696–21707, 2021.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 40.5, + "bbox_fs": [ + 105, + 649, + 506, + 673 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 680, + 504, + 703 + ], + "lines": [ + { + "bbox": [ + 106, + 680, + 505, + 693 + ], + "spans": [ + { + "bbox": [ + 106, + 680, + 505, + 693 + ], + "score": 1.0, + "content": "Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In International", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 116, + 691, + 310, + 703 + ], + "spans": [ + { + "bbox": [ + 116, + 691, + 310, + 703 + ], + "score": 1.0, + "content": "Conference on Learning Representations, 2015.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42.5, + "bbox_fs": [ + 106, + 680, + 505, + 703 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 709, + 503, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 708, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 708, + 505, + 723 + ], + "score": 1.0, + "content": "Ilya Kostrikov, Jonathan Tompson, Rob Fergus, and Ofir Nachum. Offline reinforcement learning", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 721, + 453, + 732 + ], + "spans": [ + { + "bbox": [ + 115, + 721, + 453, + 732 + ], + "score": 1.0, + "content": "with fisher divergence critic regularization. arXiv preprint arXiv:2103.08050, 2021.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44.5, + "bbox_fs": [ + 105, + 708, + 505, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 82, + 504, + 106 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "Ilya Kostrikov, Ashvin Nair, and Sergey Levine. Offline reinforcement learning with implicit", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 116, + 93, + 424, + 106 + ], + "spans": [ + { + "bbox": [ + 116, + 93, + 424, + 106 + ], + "score": 1.0, + "content": "Q-learning. In International Conference on Learning Representations, 2022.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 106, + 111, + 504, + 135 + ], + "lines": [ + { + "bbox": [ + 105, + 110, + 505, + 126 + ], + "spans": [ + { + "bbox": [ + 105, + 110, + 505, + 126 + ], + "score": 1.0, + "content": "Aviral Kumar, Xue Bin Peng, and Sergey Levine. Reward-conditioned policies. arXiv preprint", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 115, + 123, + 219, + 135 + ], + "spans": [ + { + "bbox": [ + 115, + 123, + 219, + 135 + ], + "score": 1.0, + "content": "arXiv:1912.13465, 2019.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 106, + 142, + 504, + 165 + ], + "lines": [ + { + "bbox": [ + 107, + 143, + 505, + 155 + ], + "spans": [ + { + "bbox": [ + 107, + 143, + 505, + 155 + ], + "score": 1.0, + "content": "Aviral Kumar, Aurick Zhou, George Tucker, and Sergey Levine. Conservative Q-learning for offline", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 116, + 153, + 463, + 166 + ], + "spans": [ + { + "bbox": [ + 116, + 153, + 463, + 166 + ], + "score": 1.0, + "content": "reinforcement learning. In Advances in Neural Information Processing Systems, 2020.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4.5 + }, + { + "type": "text", + "bbox": [ + 105, + 172, + 504, + 195 + ], + "lines": [ + { + "bbox": [ + 106, + 172, + 505, + 185 + ], + "spans": [ + { + "bbox": [ + 106, + 172, + 505, + 185 + ], + "score": 1.0, + "content": "Aviral Kumar, Anikait Singh, Stephen Tian, Chelsea Finn, and Sergey Levine. A workflow for offline", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 115, + 182, + 453, + 196 + ], + "spans": [ + { + "bbox": [ + 115, + 182, + 453, + 196 + ], + "score": 1.0, + "content": "model-free robotic reinforcement learning. arXiv preprint arXiv:2109.10813, 2021.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 105, + 201, + 505, + 225 + ], + "lines": [ + { + "bbox": [ + 105, + 202, + 506, + 215 + ], + "spans": [ + { + "bbox": [ + 105, + 202, + 506, + 215 + ], + "score": 1.0, + "content": "Sergey Levine, Aviral Kumar, George Tucker, and Justin Fu. Offline reinforcement learning: Tutorial,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 115, + 213, + 455, + 226 + ], + "spans": [ + { + "bbox": [ + 115, + 213, + 455, + 226 + ], + "score": 1.0, + "content": "review, and perspectives on open problems. arXiv preprint arXiv:2005.01643, 2020.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8.5 + }, + { + "type": "text", + "bbox": [ + 107, + 231, + 505, + 266 + ], + "lines": [ + { + "bbox": [ + 105, + 230, + 505, + 246 + ], + "spans": [ + { + "bbox": [ + 105, + 230, + 505, + 246 + ], + "score": 1.0, + "content": "Aitor Lewkowycz, Anders Andreassen, David Dohan, Ethan Dyer, Henryk Michalewski, Vinay", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 115, + 242, + 505, + 257 + ], + "spans": [ + { + "bbox": [ + 115, + 242, + 505, + 257 + ], + "score": 1.0, + "content": "Ramasesh, Ambrose Slone, Cem Anil, Imanol Schlag, Theo Gutman-Solo, et al. Solving", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 116, + 255, + 503, + 267 + ], + "spans": [ + { + "bbox": [ + 116, + 255, + 503, + 267 + ], + "score": 1.0, + "content": "quantitative reasoning problems with language models. arXiv preprint arXiv:2206.14858, 2022.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 107, + 273, + 504, + 307 + ], + "lines": [ + { + "bbox": [ + 105, + 271, + 506, + 287 + ], + "spans": [ + { + "bbox": [ + 105, + 271, + 506, + 287 + ], + "score": 1.0, + "content": "Timothy P Lillicrap, Jonathan J Hunt, Alexander Pritzel, Nicolas Heess, Tom Erez, Yuval Tassa,", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 284, + 505, + 297 + ], + "spans": [ + { + "bbox": [ + 115, + 284, + 505, + 297 + ], + "score": 1.0, + "content": "David Silver, and Daan Wierstra. Continuous control with deep reinforcement learning. arXiv", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 114, + 296, + 254, + 307 + ], + "spans": [ + { + "bbox": [ + 114, + 296, + 254, + 307 + ], + "score": 1.0, + "content": "preprint arXiv:1509.02971, 2015.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14 + }, + { + "type": "text", + "bbox": [ + 106, + 313, + 504, + 337 + ], + "lines": [ + { + "bbox": [ + 106, + 315, + 505, + 326 + ], + "spans": [ + { + "bbox": [ + 106, + 315, + 505, + 326 + ], + "score": 1.0, + "content": "Nan Liu, Shuang Li, Yilun Du, Antonio Torralba, and Joshua B Tenenbaum. Compositional visual", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 325, + 467, + 338 + ], + "spans": [ + { + "bbox": [ + 115, + 325, + 467, + 338 + ], + "score": 1.0, + "content": "generation with composable diffusion models. arXiv preprint arXiv:2206.01714, 2022.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5 + }, + { + "type": "text", + "bbox": [ + 107, + 343, + 504, + 367 + ], + "lines": [ + { + "bbox": [ + 105, + 342, + 506, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 506, + 358 + ], + "score": 1.0, + "content": "Calvin Luo. Understanding diffusion models: A unified perspective. arXiv preprint arXiv:2208.11970,", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 354, + 142, + 367 + ], + "spans": [ + { + "bbox": [ + 115, + 354, + 142, + 367 + ], + "score": 1.0, + "content": "2022.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18.5 + }, + { + "type": "text", + "bbox": [ + 106, + 373, + 504, + 397 + ], + "lines": [ + { + "bbox": [ + 105, + 372, + 505, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 372, + 505, + 388 + ], + "score": 1.0, + "content": "Gabriel Margolis and Pulkit Agrawal. Walk these ways: Gait-conditioned policies yield diversified", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 115, + 385, + 363, + 397 + ], + "spans": [ + { + "bbox": [ + 115, + 385, + 363, + 397 + ], + "score": 1.0, + "content": "quadrupedal agility. In Conference on Robot Learning, 2022.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20.5 + }, + { + "type": "text", + "bbox": [ + 107, + 403, + 504, + 427 + ], + "lines": [ + { + "bbox": [ + 106, + 403, + 506, + 416 + ], + "spans": [ + { + "bbox": [ + 106, + 403, + 506, + 416 + ], + "score": 1.0, + "content": "Diganta Misra. Mish: A self regularized non-monotonic neural activation function. In British", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 415, + 256, + 426 + ], + "spans": [ + { + "bbox": [ + 115, + 415, + 256, + 426 + ], + "score": 1.0, + "content": "Machine Vision Conference, 2019.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22.5 + }, + { + "type": "text", + "bbox": [ + 106, + 433, + 506, + 468 + ], + "lines": [ + { + "bbox": [ + 105, + 432, + 506, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 506, + 447 + ], + "score": 1.0, + "content": "Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew,", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 116, + 445, + 505, + 457 + ], + "spans": [ + { + "bbox": [ + 116, + 445, + 505, + 457 + ], + "score": 1.0, + "content": "Ilya Sutskever, and Mark Chen. Glide: Towards photorealistic image generation and editing with", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 116, + 456, + 399, + 468 + ], + "spans": [ + { + "bbox": [ + 116, + 456, + 399, + 468 + ], + "score": 1.0, + "content": "text-guided diffusion models. arXiv preprint arXiv:2112.10741, 2021.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 106, + 474, + 505, + 498 + ], + "lines": [ + { + "bbox": [ + 106, + 474, + 506, + 487 + ], + "spans": [ + { + "bbox": [ + 106, + 474, + 506, + 487 + ], + "score": 1.0, + "content": "Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models.", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 486, + 347, + 498 + ], + "spans": [ + { + "bbox": [ + 115, + 486, + 347, + 498 + ], + "score": 1.0, + "content": "In International Conference on Machine Learning, 2021.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 106, + 504, + 505, + 538 + ], + "lines": [ + { + "bbox": [ + 105, + 503, + 506, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 503, + 506, + 519 + ], + "score": 1.0, + "content": "Erik Nijkamp, Mitch Hill, Song-Chun Zhu, and Ying Nian Wu. Learning non-convergent non-", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 115, + 516, + 505, + 529 + ], + "spans": [ + { + "bbox": [ + 115, + 516, + 505, + 529 + ], + "score": 1.0, + "content": "persistent short-run MCMC toward energy-based model. In Advances in Neural Information", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 116, + 527, + 224, + 539 + ], + "spans": [ + { + "bbox": [ + 116, + 527, + 224, + 539 + ], + "score": 1.0, + "content": "Processing Systems, 2019.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 107, + 545, + 505, + 580 + ], + "lines": [ + { + "bbox": [ + 105, + 544, + 507, + 559 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 507, + 559 + ], + "score": 1.0, + "content": "Pedro A Ortega, Markus Kunesch, Gregoire Del ´ etang, Tim Genewein, Jordi Grau-Moya, Joel Veness, ´", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 116, + 558, + 505, + 569 + ], + "spans": [ + { + "bbox": [ + 116, + 558, + 505, + 569 + ], + "score": 1.0, + "content": "Jonas Buchli, Jonas Degrave, Bilal Piot, Julien Perolat, et al. Shaking the foundations: delusions", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 568, + 469, + 580 + ], + "spans": [ + { + "bbox": [ + 115, + 568, + 469, + 580 + ], + "score": 1.0, + "content": "in sequence models for interaction and control. arXiv preprint arXiv:2110.10819, 2021.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33 + }, + { + "type": "text", + "bbox": [ + 104, + 586, + 504, + 610 + ], + "lines": [ + { + "bbox": [ + 105, + 585, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 505, + 600 + ], + "score": 1.0, + "content": "Keiran Paster, Sheila McIlraith, and Jimmy Ba. You can’t count on luck: Why decision transformers", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 116, + 597, + 406, + 610 + ], + "spans": [ + { + "bbox": [ + 116, + 597, + 406, + 610 + ], + "score": 1.0, + "content": "fail in stochastic environments. arXiv preprint arXiv:2205.15967, 2022.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5 + }, + { + "type": "text", + "bbox": [ + 106, + 616, + 505, + 661 + ], + "lines": [ + { + "bbox": [ + 105, + 615, + 505, + 630 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 505, + 630 + ], + "score": 1.0, + "content": "Deepak Pathak, Parsa Mahmoudieh, Guanghao Luo, Pulkit Agrawal, Dian Chen, Yide Shentu, Evan", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 116, + 628, + 505, + 640 + ], + "spans": [ + { + "bbox": [ + 116, + 628, + 505, + 640 + ], + "score": 1.0, + "content": "Shelhamer, Jitendra Malik, Alexei A Efros, and Trevor Darrell. Zero-shot visual imitation. In", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 114, + 637, + 507, + 653 + ], + "spans": [ + { + "bbox": [ + 114, + 637, + 507, + 653 + ], + "score": 1.0, + "content": "Proceedings of the IEEE conference on computer vision and pattern recognition workshops, pp.", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 116, + 650, + 192, + 660 + ], + "spans": [ + { + "bbox": [ + 116, + 650, + 192, + 660 + ], + "score": 1.0, + "content": "2050–2053, 2018.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 38.5 + }, + { + "type": "text", + "bbox": [ + 106, + 668, + 504, + 702 + ], + "lines": [ + { + "bbox": [ + 106, + 667, + 506, + 682 + ], + "spans": [ + { + "bbox": [ + 106, + 667, + 506, + 682 + ], + "score": 1.0, + "content": "Tim Pearce, Tabish Rashid, Anssi Kanervisto, Dave Bignell, Mingfei Sun, Raluca Georgescu,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 679, + 505, + 693 + ], + "spans": [ + { + "bbox": [ + 115, + 679, + 505, + 693 + ], + "score": 1.0, + "content": "Sergio Valcarcel Macua, Shan Zheng Tan, Ida Momennejad, Katja Hofmann, et al. Imitating", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 115, + 690, + 444, + 703 + ], + "spans": [ + { + "bbox": [ + 115, + 690, + 444, + 703 + ], + "score": 1.0, + "content": "human behaviour with diffusion models. arXiv preprint arXiv:2301.10677, 2023.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42 + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 503, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 720, + 466, + 732 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 466, + 732 + ], + "score": 1.0, + "content": "conditional image generation with clip latents. arXiv preprint arXiv:2204.06125, 2022.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44.5 + } + ], + "page_idx": 11, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "12", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 82, + 504, + 106 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "Ilya Kostrikov, Ashvin Nair, and Sergey Levine. Offline reinforcement learning with implicit", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 116, + 93, + 424, + 106 + ], + "spans": [ + { + "bbox": [ + 116, + 93, + 424, + 106 + ], + "score": 1.0, + "content": "Q-learning. In International Conference on Learning Representations, 2022.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5, + "bbox_fs": [ + 106, + 82, + 505, + 106 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 111, + 504, + 135 + ], + "lines": [ + { + "bbox": [ + 105, + 110, + 505, + 126 + ], + "spans": [ + { + "bbox": [ + 105, + 110, + 505, + 126 + ], + "score": 1.0, + "content": "Aviral Kumar, Xue Bin Peng, and Sergey Levine. Reward-conditioned policies. arXiv preprint", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 115, + 123, + 219, + 135 + ], + "spans": [ + { + "bbox": [ + 115, + 123, + 219, + 135 + ], + "score": 1.0, + "content": "arXiv:1912.13465, 2019.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5, + "bbox_fs": [ + 105, + 110, + 505, + 135 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 142, + 504, + 165 + ], + "lines": [ + { + "bbox": [ + 107, + 143, + 505, + 155 + ], + "spans": [ + { + "bbox": [ + 107, + 143, + 505, + 155 + ], + "score": 1.0, + "content": "Aviral Kumar, Aurick Zhou, George Tucker, and Sergey Levine. Conservative Q-learning for offline", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 116, + 153, + 463, + 166 + ], + "spans": [ + { + "bbox": [ + 116, + 153, + 463, + 166 + ], + "score": 1.0, + "content": "reinforcement learning. In Advances in Neural Information Processing Systems, 2020.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4.5, + "bbox_fs": [ + 107, + 143, + 505, + 166 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 172, + 504, + 195 + ], + "lines": [ + { + "bbox": [ + 106, + 172, + 505, + 185 + ], + "spans": [ + { + "bbox": [ + 106, + 172, + 505, + 185 + ], + "score": 1.0, + "content": "Aviral Kumar, Anikait Singh, Stephen Tian, Chelsea Finn, and Sergey Levine. A workflow for offline", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 115, + 182, + 453, + 196 + ], + "spans": [ + { + "bbox": [ + 115, + 182, + 453, + 196 + ], + "score": 1.0, + "content": "model-free robotic reinforcement learning. arXiv preprint arXiv:2109.10813, 2021.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5, + "bbox_fs": [ + 106, + 172, + 505, + 196 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 201, + 505, + 225 + ], + "lines": [ + { + "bbox": [ + 105, + 202, + 506, + 215 + ], + "spans": [ + { + "bbox": [ + 105, + 202, + 506, + 215 + ], + "score": 1.0, + "content": "Sergey Levine, Aviral Kumar, George Tucker, and Justin Fu. Offline reinforcement learning: Tutorial,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 115, + 213, + 455, + 226 + ], + "spans": [ + { + "bbox": [ + 115, + 213, + 455, + 226 + ], + "score": 1.0, + "content": "review, and perspectives on open problems. arXiv preprint arXiv:2005.01643, 2020.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8.5, + "bbox_fs": [ + 105, + 202, + 506, + 226 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 231, + 505, + 266 + ], + "lines": [ + { + "bbox": [ + 105, + 230, + 505, + 246 + ], + "spans": [ + { + "bbox": [ + 105, + 230, + 505, + 246 + ], + "score": 1.0, + "content": "Aitor Lewkowycz, Anders Andreassen, David Dohan, Ethan Dyer, Henryk Michalewski, Vinay", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 115, + 242, + 505, + 257 + ], + "spans": [ + { + "bbox": [ + 115, + 242, + 505, + 257 + ], + "score": 1.0, + "content": "Ramasesh, Ambrose Slone, Cem Anil, Imanol Schlag, Theo Gutman-Solo, et al. Solving", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 116, + 255, + 503, + 267 + ], + "spans": [ + { + "bbox": [ + 116, + 255, + 503, + 267 + ], + "score": 1.0, + "content": "quantitative reasoning problems with language models. arXiv preprint arXiv:2206.14858, 2022.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 230, + 505, + 267 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 273, + 504, + 307 + ], + "lines": [ + { + "bbox": [ + 105, + 271, + 506, + 287 + ], + "spans": [ + { + "bbox": [ + 105, + 271, + 506, + 287 + ], + "score": 1.0, + "content": "Timothy P Lillicrap, Jonathan J Hunt, Alexander Pritzel, Nicolas Heess, Tom Erez, Yuval Tassa,", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 284, + 505, + 297 + ], + "spans": [ + { + "bbox": [ + 115, + 284, + 505, + 297 + ], + "score": 1.0, + "content": "David Silver, and Daan Wierstra. Continuous control with deep reinforcement learning. arXiv", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 114, + 296, + 254, + 307 + ], + "spans": [ + { + "bbox": [ + 114, + 296, + 254, + 307 + ], + "score": 1.0, + "content": "preprint arXiv:1509.02971, 2015.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14, + "bbox_fs": [ + 105, + 271, + 506, + 307 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 313, + 504, + 337 + ], + "lines": [ + { + "bbox": [ + 106, + 315, + 505, + 326 + ], + "spans": [ + { + "bbox": [ + 106, + 315, + 505, + 326 + ], + "score": 1.0, + "content": "Nan Liu, Shuang Li, Yilun Du, Antonio Torralba, and Joshua B Tenenbaum. Compositional visual", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 325, + 467, + 338 + ], + "spans": [ + { + "bbox": [ + 115, + 325, + 467, + 338 + ], + "score": 1.0, + "content": "generation with composable diffusion models. arXiv preprint arXiv:2206.01714, 2022.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5, + "bbox_fs": [ + 106, + 315, + 505, + 338 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 343, + 504, + 367 + ], + "lines": [ + { + "bbox": [ + 105, + 342, + 506, + 358 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 506, + 358 + ], + "score": 1.0, + "content": "Calvin Luo. Understanding diffusion models: A unified perspective. arXiv preprint arXiv:2208.11970,", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 354, + 142, + 367 + ], + "spans": [ + { + "bbox": [ + 115, + 354, + 142, + 367 + ], + "score": 1.0, + "content": "2022.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18.5, + "bbox_fs": [ + 105, + 342, + 506, + 367 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 373, + 504, + 397 + ], + "lines": [ + { + "bbox": [ + 105, + 372, + 505, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 372, + 505, + 388 + ], + "score": 1.0, + "content": "Gabriel Margolis and Pulkit Agrawal. Walk these ways: Gait-conditioned policies yield diversified", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 115, + 385, + 363, + 397 + ], + "spans": [ + { + "bbox": [ + 115, + 385, + 363, + 397 + ], + "score": 1.0, + "content": "quadrupedal agility. In Conference on Robot Learning, 2022.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20.5, + "bbox_fs": [ + 105, + 372, + 505, + 397 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 403, + 504, + 427 + ], + "lines": [ + { + "bbox": [ + 106, + 403, + 506, + 416 + ], + "spans": [ + { + "bbox": [ + 106, + 403, + 506, + 416 + ], + "score": 1.0, + "content": "Diganta Misra. Mish: A self regularized non-monotonic neural activation function. In British", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 415, + 256, + 426 + ], + "spans": [ + { + "bbox": [ + 115, + 415, + 256, + 426 + ], + "score": 1.0, + "content": "Machine Vision Conference, 2019.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22.5, + "bbox_fs": [ + 106, + 403, + 506, + 426 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 433, + 506, + 468 + ], + "lines": [ + { + "bbox": [ + 105, + 432, + 506, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 506, + 447 + ], + "score": 1.0, + "content": "Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew,", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 116, + 445, + 505, + 457 + ], + "spans": [ + { + "bbox": [ + 116, + 445, + 505, + 457 + ], + "score": 1.0, + "content": "Ilya Sutskever, and Mark Chen. Glide: Towards photorealistic image generation and editing with", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 116, + 456, + 399, + 468 + ], + "spans": [ + { + "bbox": [ + 116, + 456, + 399, + 468 + ], + "score": 1.0, + "content": "text-guided diffusion models. arXiv preprint arXiv:2112.10741, 2021.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25, + "bbox_fs": [ + 105, + 432, + 506, + 468 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 474, + 505, + 498 + ], + "lines": [ + { + "bbox": [ + 106, + 474, + 506, + 487 + ], + "spans": [ + { + "bbox": [ + 106, + 474, + 506, + 487 + ], + "score": 1.0, + "content": "Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models.", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 486, + 347, + 498 + ], + "spans": [ + { + "bbox": [ + 115, + 486, + 347, + 498 + ], + "score": 1.0, + "content": "In International Conference on Machine Learning, 2021.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27.5, + "bbox_fs": [ + 106, + 474, + 506, + 498 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 504, + 505, + 538 + ], + "lines": [ + { + "bbox": [ + 105, + 503, + 506, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 503, + 506, + 519 + ], + "score": 1.0, + "content": "Erik Nijkamp, Mitch Hill, Song-Chun Zhu, and Ying Nian Wu. Learning non-convergent non-", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 115, + 516, + 505, + 529 + ], + "spans": [ + { + "bbox": [ + 115, + 516, + 505, + 529 + ], + "score": 1.0, + "content": "persistent short-run MCMC toward energy-based model. In Advances in Neural Information", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 116, + 527, + 224, + 539 + ], + "spans": [ + { + "bbox": [ + 116, + 527, + 224, + 539 + ], + "score": 1.0, + "content": "Processing Systems, 2019.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30, + "bbox_fs": [ + 105, + 503, + 506, + 539 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 545, + 505, + 580 + ], + "lines": [ + { + "bbox": [ + 105, + 544, + 507, + 559 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 507, + 559 + ], + "score": 1.0, + "content": "Pedro A Ortega, Markus Kunesch, Gregoire Del ´ etang, Tim Genewein, Jordi Grau-Moya, Joel Veness, ´", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 116, + 558, + 505, + 569 + ], + "spans": [ + { + "bbox": [ + 116, + 558, + 505, + 569 + ], + "score": 1.0, + "content": "Jonas Buchli, Jonas Degrave, Bilal Piot, Julien Perolat, et al. Shaking the foundations: delusions", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 568, + 469, + 580 + ], + "spans": [ + { + "bbox": [ + 115, + 568, + 469, + 580 + ], + "score": 1.0, + "content": "in sequence models for interaction and control. arXiv preprint arXiv:2110.10819, 2021.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33, + "bbox_fs": [ + 105, + 544, + 507, + 580 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 586, + 504, + 610 + ], + "lines": [ + { + "bbox": [ + 105, + 585, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 505, + 600 + ], + "score": 1.0, + "content": "Keiran Paster, Sheila McIlraith, and Jimmy Ba. You can’t count on luck: Why decision transformers", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 116, + 597, + 406, + 610 + ], + "spans": [ + { + "bbox": [ + 116, + 597, + 406, + 610 + ], + "score": 1.0, + "content": "fail in stochastic environments. arXiv preprint arXiv:2205.15967, 2022.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5, + "bbox_fs": [ + 105, + 585, + 505, + 610 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 616, + 505, + 661 + ], + "lines": [ + { + "bbox": [ + 105, + 615, + 505, + 630 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 505, + 630 + ], + "score": 1.0, + "content": "Deepak Pathak, Parsa Mahmoudieh, Guanghao Luo, Pulkit Agrawal, Dian Chen, Yide Shentu, Evan", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 116, + 628, + 505, + 640 + ], + "spans": [ + { + "bbox": [ + 116, + 628, + 505, + 640 + ], + "score": 1.0, + "content": "Shelhamer, Jitendra Malik, Alexei A Efros, and Trevor Darrell. Zero-shot visual imitation. In", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 114, + 637, + 507, + 653 + ], + "spans": [ + { + "bbox": [ + 114, + 637, + 507, + 653 + ], + "score": 1.0, + "content": "Proceedings of the IEEE conference on computer vision and pattern recognition workshops, pp.", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 116, + 650, + 192, + 660 + ], + "spans": [ + { + "bbox": [ + 116, + 650, + 192, + 660 + ], + "score": 1.0, + "content": "2050–2053, 2018.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 38.5, + "bbox_fs": [ + 105, + 615, + 507, + 660 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 668, + 504, + 702 + ], + "lines": [ + { + "bbox": [ + 106, + 667, + 506, + 682 + ], + "spans": [ + { + "bbox": [ + 106, + 667, + 506, + 682 + ], + "score": 1.0, + "content": "Tim Pearce, Tabish Rashid, Anssi Kanervisto, Dave Bignell, Mingfei Sun, Raluca Georgescu,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 115, + 679, + 505, + 693 + ], + "spans": [ + { + "bbox": [ + 115, + 679, + 505, + 693 + ], + "score": 1.0, + "content": "Sergio Valcarcel Macua, Shan Zheng Tan, Ida Momennejad, Katja Hofmann, et al. Imitating", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 115, + 690, + 444, + 703 + ], + "spans": [ + { + "bbox": [ + 115, + 690, + 444, + 703 + ], + "score": 1.0, + "content": "human behaviour with diffusion models. arXiv preprint arXiv:2301.10677, 2023.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42, + "bbox_fs": [ + 106, + 667, + 506, + 703 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 709, + 503, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 115, + 720, + 466, + 732 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 466, + 732 + ], + "score": 1.0, + "content": "conditional image generation with clip latents. arXiv preprint arXiv:2204.06125, 2022.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 44.5, + "bbox_fs": [ + 106, + 709, + 505, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 109, + 82, + 505, + 116 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 506, + 95 + ], + "score": 1.0, + "content": "Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bjorn Ommer. High- ¨", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 94, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 94, + 505, + 106 + ], + "score": 1.0, + "content": "resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 104, + 448, + 118 + ], + "spans": [ + { + "bbox": [ + 115, + 104, + 448, + 118 + ], + "score": 1.0, + "content": "Conference on Computer Vision and Pattern Recognition, pp. 10684–10695, 2022.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 107, + 123, + 506, + 168 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "score": 1.0, + "content": "Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily Denton, Seyed", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 135, + 507, + 146 + ], + "spans": [ + { + "bbox": [ + 115, + 135, + 507, + 146 + ], + "score": 1.0, + "content": "Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, S Sara Mahdavi, Rapha Gontijo Lopes, et al.", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 145, + 505, + 158 + ], + "spans": [ + { + "bbox": [ + 115, + 145, + 505, + 158 + ], + "score": 1.0, + "content": "Photorealistic text-to-image diffusion models with deep language understanding. arXiv preprint", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 156, + 219, + 168 + ], + "spans": [ + { + "bbox": [ + 115, + 156, + 219, + 168 + ], + "score": 1.0, + "content": "arXiv:2205.11487, 2022.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + }, + { + "type": "text", + "bbox": [ + 105, + 174, + 503, + 198 + ], + "lines": [ + { + "bbox": [ + 105, + 174, + 505, + 189 + ], + "spans": [ + { + "bbox": [ + 105, + 174, + 505, + 189 + ], + "score": 1.0, + "content": "Juergen Schmidhuber. Reinforcement learning upside down: Don’t predict rewards–just map them to", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 115, + 186, + 313, + 198 + ], + "spans": [ + { + "bbox": [ + 115, + 186, + 313, + 198 + ], + "score": 1.0, + "content": "actions. arXiv preprint arXiv:1912.02875, 2019.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 106, + 204, + 506, + 238 + ], + "lines": [ + { + "bbox": [ + 105, + 204, + 505, + 218 + ], + "spans": [ + { + "bbox": [ + 105, + 204, + 505, + 218 + ], + "score": 1.0, + "content": "Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 115, + 215, + 507, + 230 + ], + "spans": [ + { + "bbox": [ + 115, + 215, + 507, + 230 + ], + "score": 1.0, + "content": "learning using nonequilibrium thermodynamics. In International Conference on Machine Learning,", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 115, + 225, + 143, + 239 + ], + "spans": [ + { + "bbox": [ + 115, + 225, + 143, + 239 + ], + "score": 1.0, + "content": "2015.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + }, + { + "type": "text", + "bbox": [ + 107, + 245, + 504, + 268 + ], + "lines": [ + { + "bbox": [ + 105, + 245, + 506, + 259 + ], + "spans": [ + { + "bbox": [ + 105, + 245, + 506, + 259 + ], + "score": 1.0, + "content": "Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. In", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 115, + 256, + 365, + 269 + ], + "spans": [ + { + "bbox": [ + 115, + 256, + 365, + 269 + ], + "score": 1.0, + "content": "International Conference on Learning Representations, 2021.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12.5 + }, + { + "type": "text", + "bbox": [ + 105, + 275, + 504, + 298 + ], + "lines": [ + { + "bbox": [ + 105, + 273, + 507, + 289 + ], + "spans": [ + { + "bbox": [ + 105, + 273, + 507, + 289 + ], + "score": 1.0, + "content": "Richard S. Sutton. Learning to predict by the methods of temporal differences. Machine Learning,", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 114, + 284, + 142, + 300 + ], + "spans": [ + { + "bbox": [ + 114, + 284, + 142, + 300 + ], + "score": 1.0, + "content": "1988.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5 + }, + { + "type": "text", + "bbox": [ + 105, + 305, + 504, + 317 + ], + "lines": [ + { + "bbox": [ + 105, + 304, + 506, + 319 + ], + "spans": [ + { + "bbox": [ + 105, + 304, + 506, + 319 + ], + "score": 1.0, + "content": "Richard S Sutton and Andrew G Barto. Reinforcement learning: An introduction. MIT press, 2018.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 106, + 324, + 491, + 336 + ], + "lines": [ + { + "bbox": [ + 106, + 324, + 490, + 337 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 490, + 337 + ], + "score": 1.0, + "content": "Russ Tedrake. Underactuated Robotics. 2022. URL http://underactuated.mit.edu.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 17 + }, + { + "type": "text", + "bbox": [ + 107, + 343, + 506, + 377 + ], + "lines": [ + { + "bbox": [ + 105, + 342, + 505, + 356 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 505, + 356 + ], + "score": 1.0, + "content": "Hado Van Hasselt, Yotam Doron, Florian Strub, Matteo Hessel, Nicolas Sonnerat, and Joseph", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 353, + 506, + 367 + ], + "spans": [ + { + "bbox": [ + 115, + 353, + 506, + 367 + ], + "score": 1.0, + "content": "Modayil. Deep reinforcement learning and the deadly triad. arXiv preprint arXiv:1812.02648,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 363, + 144, + 378 + ], + "spans": [ + { + "bbox": [ + 115, + 363, + 144, + 378 + ], + "score": 1.0, + "content": "2018.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 107, + 384, + 506, + 417 + ], + "lines": [ + { + "bbox": [ + 105, + 384, + 505, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 384, + 505, + 396 + ], + "score": 1.0, + "content": "Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 395, + 505, + 408 + ], + "spans": [ + { + "bbox": [ + 115, + 395, + 505, + 408 + ], + "score": 1.0, + "content": "Kaiser, and Illia Polosukhin. Attention is all you need. In Advances in Neural Information", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 406, + 224, + 419 + ], + "spans": [ + { + "bbox": [ + 115, + 406, + 224, + 419 + ], + "score": 1.0, + "content": "Processing Systems, 2017.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 107, + 424, + 505, + 459 + ], + "lines": [ + { + "bbox": [ + 105, + 424, + 505, + 438 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 505, + 438 + ], + "score": 1.0, + "content": "Adam R Villaflor, Zhe Huang, Swapnil Pande, John M Dolan, and Jeff Schneider. Addressing", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 436, + 506, + 449 + ], + "spans": [ + { + "bbox": [ + 115, + 436, + 506, + 449 + ], + "score": 1.0, + "content": "optimism bias in sequence modeling for reinforcement learning. In International Conference on", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 115, + 447, + 326, + 459 + ], + "spans": [ + { + "bbox": [ + 115, + 447, + 326, + 459 + ], + "score": 1.0, + "content": "Machine Learning, pp. 22270–22283. PMLR, 2022.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 107, + 466, + 505, + 499 + ], + "lines": [ + { + "bbox": [ + 105, + 464, + 506, + 479 + ], + "spans": [ + { + "bbox": [ + 105, + 464, + 506, + 479 + ], + "score": 1.0, + "content": "Homer Walke, Jonathan Yang, Albert Yu, Aviral Kumar, Jedrzej Orbik, Avi Singh, and Sergey Levine.", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 116, + 477, + 505, + 489 + ], + "spans": [ + { + "bbox": [ + 116, + 477, + 505, + 489 + ], + "score": 1.0, + "content": "Don’t start from scratch: Leveraging prior data to automate robotic reinforcement learning. arXiv", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 115, + 489, + 254, + 499 + ], + "spans": [ + { + "bbox": [ + 115, + 489, + 254, + 499 + ], + "score": 1.0, + "content": "preprint arXiv:2207.04703, 2022.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 104, + 506, + 504, + 529 + ], + "lines": [ + { + "bbox": [ + 105, + 505, + 505, + 521 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 505, + 521 + ], + "score": 1.0, + "content": "Zhendong Wang, Jonathan J Hunt, and Mingyuan Zhou. Diffusion policies as an expressive policy", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 116, + 517, + 440, + 530 + ], + "spans": [ + { + "bbox": [ + 116, + 517, + 440, + 530 + ], + "score": 1.0, + "content": "class for offline reinforcement learning. arXiv preprint arXiv:2208.06193, 2022.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30.5 + }, + { + "type": "text", + "bbox": [ + 105, + 536, + 505, + 559 + ], + "lines": [ + { + "bbox": [ + 105, + 534, + 507, + 551 + ], + "spans": [ + { + "bbox": [ + 105, + 534, + 507, + 551 + ], + "score": 1.0, + "content": "Yifan Wu, George Tucker, and Ofir Nachum. Behavior regularized offline reinforcement learning.", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 115, + 547, + 279, + 560 + ], + "spans": [ + { + "bbox": [ + 115, + 547, + 279, + 560 + ], + "score": 1.0, + "content": "arXiv preprint arXiv:1911.11361, 2019.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32.5 + }, + { + "type": "text", + "bbox": [ + 105, + 566, + 505, + 589 + ], + "lines": [ + { + "bbox": [ + 105, + 565, + 507, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 507, + 581 + ], + "score": 1.0, + "content": "Yuxin Wu and Kaiming He. Group normalization. In European Conference on Computer Vision,", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 115, + 575, + 144, + 591 + ], + "spans": [ + { + "bbox": [ + 115, + 575, + 144, + 591 + ], + "score": 1.0, + "content": "2018.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 34.5 + }, + { + "type": "text", + "bbox": [ + 105, + 596, + 504, + 619 + ], + "lines": [ + { + "bbox": [ + 106, + 596, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 106, + 596, + 506, + 609 + ], + "score": 1.0, + "content": "Mengjiao Yang, Dale Schuurmans, Pieter Abbeel, and Ofir Nachum. Dichotomy of control:", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 116, + 607, + 504, + 620 + ], + "spans": [ + { + "bbox": [ + 116, + 607, + 504, + 620 + ], + "score": 1.0, + "content": "Separating what you can control from what you cannot. arXiv preprint arXiv:2210.13435, 2022.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5 + }, + { + "type": "text", + "bbox": [ + 104, + 626, + 504, + 650 + ], + "lines": [ + { + "bbox": [ + 105, + 626, + 505, + 639 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 505, + 639 + ], + "score": 1.0, + "content": "Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Beyond empirical", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 637, + 356, + 649 + ], + "spans": [ + { + "bbox": [ + 115, + 637, + 356, + 649 + ], + "score": 1.0, + "content": "risk minimization. arXiv preprint arXiv:1710.09412, 2017.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5 + }, + { + "type": "text", + "bbox": [ + 106, + 656, + 504, + 679 + ], + "lines": [ + { + "bbox": [ + 106, + 655, + 506, + 669 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 506, + 669 + ], + "score": 1.0, + "content": "Qinqing Zheng, Amy Zhang, and Aditya Grover. Online decision transformer. arXiv preprint", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 667, + 220, + 680 + ], + "spans": [ + { + "bbox": [ + 115, + 667, + 220, + 680 + ], + "score": 1.0, + "content": "arXiv:2202.05607, 2022.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 40.5 + } + ], + "page_idx": 12, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "13", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 109, + 82, + 505, + 116 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 506, + 95 + ], + "score": 1.0, + "content": "Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bjorn Ommer. High- ¨", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 115, + 94, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 115, + 94, + 505, + 106 + ], + "score": 1.0, + "content": "resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 104, + 448, + 118 + ], + "spans": [ + { + "bbox": [ + 115, + 104, + 448, + 118 + ], + "score": 1.0, + "content": "Conference on Computer Vision and Pattern Recognition, pp. 10684–10695, 2022.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1, + "bbox_fs": [ + 106, + 82, + 506, + 118 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 123, + 506, + 168 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "score": 1.0, + "content": "Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily Denton, Seyed", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 135, + 507, + 146 + ], + "spans": [ + { + "bbox": [ + 115, + 135, + 507, + 146 + ], + "score": 1.0, + "content": "Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, S Sara Mahdavi, Rapha Gontijo Lopes, et al.", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 145, + 505, + 158 + ], + "spans": [ + { + "bbox": [ + 115, + 145, + 505, + 158 + ], + "score": 1.0, + "content": "Photorealistic text-to-image diffusion models with deep language understanding. arXiv preprint", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 156, + 219, + 168 + ], + "spans": [ + { + "bbox": [ + 115, + 156, + 219, + 168 + ], + "score": 1.0, + "content": "arXiv:2205.11487, 2022.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5, + "bbox_fs": [ + 106, + 123, + 507, + 168 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 174, + 503, + 198 + ], + "lines": [ + { + "bbox": [ + 105, + 174, + 505, + 189 + ], + "spans": [ + { + "bbox": [ + 105, + 174, + 505, + 189 + ], + "score": 1.0, + "content": "Juergen Schmidhuber. Reinforcement learning upside down: Don’t predict rewards–just map them to", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 115, + 186, + 313, + 198 + ], + "spans": [ + { + "bbox": [ + 115, + 186, + 313, + 198 + ], + "score": 1.0, + "content": "actions. arXiv preprint arXiv:1912.02875, 2019.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5, + "bbox_fs": [ + 105, + 174, + 505, + 198 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 204, + 506, + 238 + ], + "lines": [ + { + "bbox": [ + 105, + 204, + 505, + 218 + ], + "spans": [ + { + "bbox": [ + 105, + 204, + 505, + 218 + ], + "score": 1.0, + "content": "Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 115, + 215, + 507, + 230 + ], + "spans": [ + { + "bbox": [ + 115, + 215, + 507, + 230 + ], + "score": 1.0, + "content": "learning using nonequilibrium thermodynamics. In International Conference on Machine Learning,", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 115, + 225, + 143, + 239 + ], + "spans": [ + { + "bbox": [ + 115, + 225, + 143, + 239 + ], + "score": 1.0, + "content": "2015.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10, + "bbox_fs": [ + 105, + 204, + 507, + 239 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 245, + 504, + 268 + ], + "lines": [ + { + "bbox": [ + 105, + 245, + 506, + 259 + ], + "spans": [ + { + "bbox": [ + 105, + 245, + 506, + 259 + ], + "score": 1.0, + "content": "Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. In", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 115, + 256, + 365, + 269 + ], + "spans": [ + { + "bbox": [ + 115, + 256, + 365, + 269 + ], + "score": 1.0, + "content": "International Conference on Learning Representations, 2021.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 12.5, + "bbox_fs": [ + 105, + 245, + 506, + 269 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 275, + 504, + 298 + ], + "lines": [ + { + "bbox": [ + 105, + 273, + 507, + 289 + ], + "spans": [ + { + "bbox": [ + 105, + 273, + 507, + 289 + ], + "score": 1.0, + "content": "Richard S. Sutton. Learning to predict by the methods of temporal differences. Machine Learning,", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 114, + 284, + 142, + 300 + ], + "spans": [ + { + "bbox": [ + 114, + 284, + 142, + 300 + ], + "score": 1.0, + "content": "1988.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14.5, + "bbox_fs": [ + 105, + 273, + 507, + 300 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 305, + 504, + 317 + ], + "lines": [ + { + "bbox": [ + 105, + 304, + 506, + 319 + ], + "spans": [ + { + "bbox": [ + 105, + 304, + 506, + 319 + ], + "score": 1.0, + "content": "Richard S Sutton and Andrew G Barto. Reinforcement learning: An introduction. MIT press, 2018.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 16, + "bbox_fs": [ + 105, + 304, + 506, + 319 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 324, + 491, + 336 + ], + "lines": [ + { + "bbox": [ + 106, + 324, + 490, + 337 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 490, + 337 + ], + "score": 1.0, + "content": "Russ Tedrake. Underactuated Robotics. 2022. URL http://underactuated.mit.edu.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 17, + "bbox_fs": [ + 106, + 324, + 490, + 337 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 343, + 506, + 377 + ], + "lines": [ + { + "bbox": [ + 105, + 342, + 505, + 356 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 505, + 356 + ], + "score": 1.0, + "content": "Hado Van Hasselt, Yotam Doron, Florian Strub, Matteo Hessel, Nicolas Sonnerat, and Joseph", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 353, + 506, + 367 + ], + "spans": [ + { + "bbox": [ + 115, + 353, + 506, + 367 + ], + "score": 1.0, + "content": "Modayil. Deep reinforcement learning and the deadly triad. arXiv preprint arXiv:1812.02648,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 363, + 144, + 378 + ], + "spans": [ + { + "bbox": [ + 115, + 363, + 144, + 378 + ], + "score": 1.0, + "content": "2018.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19, + "bbox_fs": [ + 105, + 342, + 506, + 378 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 384, + 506, + 417 + ], + "lines": [ + { + "bbox": [ + 105, + 384, + 505, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 384, + 505, + 396 + ], + "score": 1.0, + "content": "Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 395, + 505, + 408 + ], + "spans": [ + { + "bbox": [ + 115, + 395, + 505, + 408 + ], + "score": 1.0, + "content": "Kaiser, and Illia Polosukhin. Attention is all you need. In Advances in Neural Information", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 115, + 406, + 224, + 419 + ], + "spans": [ + { + "bbox": [ + 115, + 406, + 224, + 419 + ], + "score": 1.0, + "content": "Processing Systems, 2017.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22, + "bbox_fs": [ + 105, + 384, + 505, + 419 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 424, + 505, + 459 + ], + "lines": [ + { + "bbox": [ + 105, + 424, + 505, + 438 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 505, + 438 + ], + "score": 1.0, + "content": "Adam R Villaflor, Zhe Huang, Swapnil Pande, John M Dolan, and Jeff Schneider. Addressing", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 436, + 506, + 449 + ], + "spans": [ + { + "bbox": [ + 115, + 436, + 506, + 449 + ], + "score": 1.0, + "content": "optimism bias in sequence modeling for reinforcement learning. In International Conference on", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 115, + 447, + 326, + 459 + ], + "spans": [ + { + "bbox": [ + 115, + 447, + 326, + 459 + ], + "score": 1.0, + "content": "Machine Learning, pp. 22270–22283. PMLR, 2022.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25, + "bbox_fs": [ + 105, + 424, + 506, + 459 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 466, + 505, + 499 + ], + "lines": [ + { + "bbox": [ + 105, + 464, + 506, + 479 + ], + "spans": [ + { + "bbox": [ + 105, + 464, + 506, + 479 + ], + "score": 1.0, + "content": "Homer Walke, Jonathan Yang, Albert Yu, Aviral Kumar, Jedrzej Orbik, Avi Singh, and Sergey Levine.", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 116, + 477, + 505, + 489 + ], + "spans": [ + { + "bbox": [ + 116, + 477, + 505, + 489 + ], + "score": 1.0, + "content": "Don’t start from scratch: Leveraging prior data to automate robotic reinforcement learning. arXiv", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 115, + 489, + 254, + 499 + ], + "spans": [ + { + "bbox": [ + 115, + 489, + 254, + 499 + ], + "score": 1.0, + "content": "preprint arXiv:2207.04703, 2022.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 28, + "bbox_fs": [ + 105, + 464, + 506, + 499 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 506, + 504, + 529 + ], + "lines": [ + { + "bbox": [ + 105, + 505, + 505, + 521 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 505, + 521 + ], + "score": 1.0, + "content": "Zhendong Wang, Jonathan J Hunt, and Mingyuan Zhou. Diffusion policies as an expressive policy", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 116, + 517, + 440, + 530 + ], + "spans": [ + { + "bbox": [ + 116, + 517, + 440, + 530 + ], + "score": 1.0, + "content": "class for offline reinforcement learning. arXiv preprint arXiv:2208.06193, 2022.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30.5, + "bbox_fs": [ + 105, + 505, + 505, + 530 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 536, + 505, + 559 + ], + "lines": [ + { + "bbox": [ + 105, + 534, + 507, + 551 + ], + "spans": [ + { + "bbox": [ + 105, + 534, + 507, + 551 + ], + "score": 1.0, + "content": "Yifan Wu, George Tucker, and Ofir Nachum. Behavior regularized offline reinforcement learning.", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 115, + 547, + 279, + 560 + ], + "spans": [ + { + "bbox": [ + 115, + 547, + 279, + 560 + ], + "score": 1.0, + "content": "arXiv preprint arXiv:1911.11361, 2019.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 32.5, + "bbox_fs": [ + 105, + 534, + 507, + 560 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 566, + 505, + 589 + ], + "lines": [ + { + "bbox": [ + 105, + 565, + 507, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 507, + 581 + ], + "score": 1.0, + "content": "Yuxin Wu and Kaiming He. Group normalization. In European Conference on Computer Vision,", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 115, + 575, + 144, + 591 + ], + "spans": [ + { + "bbox": [ + 115, + 575, + 144, + 591 + ], + "score": 1.0, + "content": "2018.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 34.5, + "bbox_fs": [ + 105, + 565, + 507, + 591 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 596, + 504, + 619 + ], + "lines": [ + { + "bbox": [ + 106, + 596, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 106, + 596, + 506, + 609 + ], + "score": 1.0, + "content": "Mengjiao Yang, Dale Schuurmans, Pieter Abbeel, and Ofir Nachum. Dichotomy of control:", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 116, + 607, + 504, + 620 + ], + "spans": [ + { + "bbox": [ + 116, + 607, + 504, + 620 + ], + "score": 1.0, + "content": "Separating what you can control from what you cannot. arXiv preprint arXiv:2210.13435, 2022.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5, + "bbox_fs": [ + 106, + 596, + 506, + 620 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 626, + 504, + 650 + ], + "lines": [ + { + "bbox": [ + 105, + 626, + 505, + 639 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 505, + 639 + ], + "score": 1.0, + "content": "Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Beyond empirical", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 637, + 356, + 649 + ], + "spans": [ + { + "bbox": [ + 115, + 637, + 356, + 649 + ], + "score": 1.0, + "content": "risk minimization. arXiv preprint arXiv:1710.09412, 2017.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5, + "bbox_fs": [ + 105, + 626, + 505, + 649 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 656, + 504, + 679 + ], + "lines": [ + { + "bbox": [ + 106, + 655, + 506, + 669 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 506, + 669 + ], + "score": 1.0, + "content": "Qinqing Zheng, Amy Zhang, and Aditya Grover. Online decision transformer. arXiv preprint", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 115, + 667, + 220, + 680 + ], + "spans": [ + { + "bbox": [ + 115, + 667, + 220, + 680 + ], + "score": 1.0, + "content": "arXiv:2202.05607, 2022.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 40.5, + "bbox_fs": [ + 106, + 655, + 506, + 680 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 79, + 167, + 95 + ], + "lines": [ + { + "bbox": [ + 104, + 78, + 170, + 97 + ], + "spans": [ + { + "bbox": [ + 104, + 78, + 170, + 97 + ], + "score": 1.0, + "content": "Appendix", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 106, + 104, + 505, + 181 + ], + "lines": [ + { + "bbox": [ + 106, + 104, + 505, + 117 + ], + "spans": [ + { + "bbox": [ + 106, + 104, + 505, + 117 + ], + "score": 1.0, + "content": "In this appendix, we discuss details of the illustrative examples in Section A. Next, we discuss", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 115, + 505, + 128 + ], + "spans": [ + { + "bbox": [ + 106, + 115, + 505, + 128 + ], + "score": 1.0, + "content": "hyperparameters and architectural details in Section B. We analyze the importance of low temperature", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 127, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 106, + 127, + 505, + 138 + ], + "score": 1.0, + "content": "sampling in Section C, further explain composition of conditioning variable in Section D, discuss the", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 138, + 505, + 149 + ], + "spans": [ + { + "bbox": [ + 106, + 138, + 505, + 149 + ], + "score": 1.0, + "content": "run-time characteristics of decision diffuser in Section E, discuss when to use inverse dynamics in", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 147, + 506, + 161 + ], + "spans": [ + { + "bbox": [ + 105, + 147, + 506, + 161 + ], + "score": 1.0, + "content": "Section F and analyze robustness of Decision Diffuser to stochastic dynamics in Section G. Finally,", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 159, + 506, + 171 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 506, + 171 + ], + "score": 1.0, + "content": "we provide details of the Kuka Block Stacking environment in Section H and the Unitree environment", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 170, + 158, + 182 + ], + "spans": [ + { + "bbox": [ + 106, + 170, + 158, + 182 + ], + "score": 1.0, + "content": "in Section I.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 4 + }, + { + "type": "title", + "bbox": [ + 108, + 196, + 263, + 209 + ], + "lines": [ + { + "bbox": [ + 105, + 195, + 265, + 211 + ], + "spans": [ + { + "bbox": [ + 105, + 195, + 265, + 211 + ], + "score": 1.0, + "content": "A ILLUSTRATIVE EXAMPLES", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "title", + "bbox": [ + 107, + 222, + 287, + 234 + ], + "lines": [ + { + "bbox": [ + 106, + 222, + 288, + 235 + ], + "spans": [ + { + "bbox": [ + 106, + 222, + 288, + 235 + ], + "score": 1.0, + "content": "A.1 IMPLICIT DYNAMIC PROGRAMMING", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "image", + "bbox": [ + 207, + 240, + 404, + 313 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 207, + 240, + 404, + 313 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 207, + 240, + 404, + 313 + ], + "spans": [ + { + "bbox": [ + 207, + 240, + 404, + 313 + ], + "score": 0.965, + "type": "image", + "image_path": "06c49faf819f369ba151ad9d3ff002a905c14d4f9d446bec58e52f70a39c1d4b.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 207, + 240, + 404, + 254.6 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 207, + 254.6, + 404, + 269.2 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 207, + 269.2, + 404, + 283.8 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 207, + 283.8, + 404, + 298.40000000000003 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 207, + 298.40000000000003, + 404, + 313.00000000000006 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 323, + 504, + 354 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 323, + 505, + 335 + ], + "spans": [ + { + "bbox": [ + 106, + 323, + 505, + 335 + ], + "score": 1.0, + "content": "Figure A1: Illustrative example. We demonstrate the ability of Decision Diffuser to stitch together sub-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 333, + 506, + 347 + ], + "spans": [ + { + "bbox": [ + 105, + 333, + 506, + 347 + ], + "score": 1.0, + "content": "optimal trajectories in training dataset to obtain (near) optimal trajectories, thereby implicitly performing", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 344, + 380, + 355 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 380, + 355 + ], + "score": 1.0, + "content": "dynamic programming in Maze2D-open environment from Fu et al. (2020).", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16 + } + ], + "index": 14.0 + }, + { + "type": "text", + "bbox": [ + 107, + 364, + 505, + 442 + ], + "lines": [ + { + "bbox": [ + 104, + 361, + 506, + 379 + ], + "spans": [ + { + "bbox": [ + 104, + 361, + 506, + 379 + ], + "score": 1.0, + "content": "We empirically demonstrate the ability of Decision Diffuser to perform implicit dynamic programming", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 375, + 505, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 505, + 388 + ], + "score": 1.0, + "content": "in Maze2D-open environment from Fu et al. (2020). The task in Maze2D-open environment is to", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 386, + 505, + 398 + ], + "spans": [ + { + "bbox": [ + 106, + 386, + 505, + 398 + ], + "score": 1.0, + "content": "reach point C and the reward is negative distance from point C. The training dataset consists of 500", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 397, + 505, + 410 + ], + "spans": [ + { + "bbox": [ + 106, + 397, + 505, + 410 + ], + "score": 1.0, + "content": "trajectories from point A to point B and 500 trajectories from point B to point C. The maximum", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 408, + 506, + 420 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 506, + 420 + ], + "score": 1.0, + "content": "trajectory length is 50. During test time, the agent starts from point A and needs to reach point C", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 418, + 506, + 433 + ], + "spans": [ + { + "bbox": [ + 105, + 418, + 506, + 433 + ], + "score": 1.0, + "content": "as quickly as possible. As shown in Figure A1, Decision Diffuser can stitch trajectories in training", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 429, + 443, + 442 + ], + "spans": [ + { + "bbox": [ + 105, + 429, + 443, + 442 + ], + "score": 1.0, + "content": "dataset to form trajectories that goes from point A to point B in (near) straight lines.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21 + }, + { + "type": "title", + "bbox": [ + 108, + 453, + 255, + 465 + ], + "lines": [ + { + "bbox": [ + 106, + 452, + 257, + 467 + ], + "spans": [ + { + "bbox": [ + 106, + 452, + 257, + 467 + ], + "score": 1.0, + "content": "A.2 CONSTRAINT COMBINATION", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 106, + 471, + 506, + 548 + ], + "lines": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "Setup In linear system robot navigation, Decision Diffuser is trained on 1000 expert trajectories", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 482, + 507, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 234, + 495 + ], + "score": 1.0, + "content": "either satisfying the constraint", + "type": "text" + }, + { + "bbox": [ + 235, + 482, + 280, + 494 + ], + "score": 0.86, + "content": "\\lVert s _ { T } \\rVert \\leq R", + "type": "inline_equation" + }, + { + "bbox": [ + 285, + 482, + 315, + 493 + ], + "score": 0.76, + "content": "R = 1", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 482, + 390, + 495 + ], + "score": 1.0, + "content": ") or the constraint", + "type": "text" + }, + { + "bbox": [ + 390, + 482, + 432, + 494 + ], + "score": 0.87, + "content": "\\| s _ { T } \\| \\geq r", + "type": "inline_equation" + }, + { + "bbox": [ + 437, + 483, + 473, + 493 + ], + "score": 0.67, + "content": "\\mathit { r } = 0 . 7", + "type": "inline_equation" + }, + { + "bbox": [ + 473, + 482, + 507, + 495 + ], + "score": 1.0, + "content": "). Here,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 107, + 493, + 506, + 506 + ], + "spans": [ + { + "bbox": [ + 107, + 493, + 167, + 505 + ], + "score": 0.92, + "content": "s _ { T } = [ x _ { T } , y _ { T } ]", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 493, + 506, + 506 + ], + "score": 1.0, + "content": "represents the final robot state in a trajectory, specifying its final 2d position. The", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 504, + 505, + 516 + ], + "spans": [ + { + "bbox": [ + 105, + 504, + 505, + 516 + ], + "score": 1.0, + "content": "maximum trajectory length is 50. During test time, Decision Diffuser is asked to generate trajectories", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 515, + 507, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 149, + 528 + ], + "score": 1.0, + "content": "satisfying", + "type": "text" + }, + { + "bbox": [ + 150, + 515, + 195, + 527 + ], + "score": 0.92, + "content": "\\| s _ { T } \\| \\le R", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 515, + 215, + 528 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 215, + 515, + 258, + 527 + ], + "score": 0.93, + "content": "\\| s _ { T } \\| \\geq r", + "type": "inline_equation" + }, + { + "bbox": [ + 258, + 515, + 507, + 528 + ], + "score": 1.0, + "content": "to test its ability to satisfy single constraints. Furthermore,", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 526, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 366, + 540 + ], + "score": 1.0, + "content": "Decision Diffuser is also asked to generate trajectories satisfying", + "type": "text" + }, + { + "bbox": [ + 366, + 526, + 428, + 538 + ], + "score": 0.94, + "content": "r \\leq \\left\\lceil | s _ { T } | | \\leq R \\right\\rceil", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 526, + 506, + 540 + ], + "score": 1.0, + "content": "to test its ability to", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 537, + 225, + 550 + ], + "spans": [ + { + "bbox": [ + 106, + 537, + 225, + 550 + ], + "score": 1.0, + "content": "satisfy combined constraints.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 551, + 505, + 595 + ], + "lines": [ + { + "bbox": [ + 105, + 551, + 505, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 551, + 505, + 563 + ], + "score": 1.0, + "content": "Results Figure 2 shows that Decision Diffuser learns to generate trajectories perfectly (i.e. with", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 561, + 506, + 575 + ], + "spans": [ + { + "bbox": [ + 106, + 562, + 131, + 573 + ], + "score": 0.88, + "content": "1 0 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 561, + 506, + 575 + ], + "score": 1.0, + "content": "success rate) satisfying single constraints in linear system robot navigation. Furthermore, it", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 573, + 505, + 585 + ], + "spans": [ + { + "bbox": [ + 105, + 573, + 505, + 585 + ], + "score": 1.0, + "content": "learns to generate trajectories satisfying the composed constraint in linear system robot navigation", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 583, + 468, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 584, + 126, + 596 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 127, + 583, + 190, + 596 + ], + "score": 0.95, + "content": "9 1 . 3 \\hat { \\% } ( \\pm 2 . 6 \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 584, + 468, + 596 + ], + "score": 1.0, + "content": "accuracy where the standard error is calculated over 5 random seeds.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34.5 + }, + { + "type": "title", + "bbox": [ + 108, + 611, + 390, + 623 + ], + "lines": [ + { + "bbox": [ + 105, + 609, + 392, + 625 + ], + "spans": [ + { + "bbox": [ + 105, + 609, + 392, + 625 + ], + "score": 1.0, + "content": "B HYPERPARAMETER AND ARCHITECTURAL DETAILS", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 106, + 635, + 412, + 647 + ], + "lines": [ + { + "bbox": [ + 105, + 634, + 414, + 649 + ], + "spans": [ + { + "bbox": [ + 105, + 634, + 414, + 649 + ], + "score": 1.0, + "content": "In this section, we describe various architectural and hyperparameter details:", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 106, + 654, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 653, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 232, + 668 + ], + "score": 1.0, + "content": "• We represent the noise model", + "type": "text" + }, + { + "bbox": [ + 233, + 657, + 243, + 666 + ], + "score": 0.84, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 243, + 653, + 506, + 668 + ], + "score": 1.0, + "content": "with a temporal U-Net (Janner et al., 2022), consisting of a U-Net", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 114, + 666, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 114, + 666, + 505, + 678 + ], + "score": 1.0, + "content": "structure with 6 repeated residual blocks. Each block consisted of two temporal convolutions, each", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 113, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 113, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "followed by group norm (Wu & He, 2018), and a final Mish nonlinearity (Misra, 2019). Timestep", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 114, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 114, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "and condition embeddings, both 128-dimensional vectors, are produced by separate 2-layered MLP", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 114, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 114, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "(with 256 hidden units and Mish nonlinearity) and are concatenated together before getting added", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 114, + 710, + 505, + 721 + ], + "spans": [ + { + "bbox": [ + 114, + 710, + 505, + 721 + ], + "score": 1.0, + "content": "to the activations of the first temporal convolution within each block. We borrow the code for", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 114, + 721, + 349, + 733 + ], + "spans": [ + { + "bbox": [ + 114, + 721, + 349, + 733 + ], + "score": 1.0, + "content": "temporal U-Net from https://github.com/jannerm/diffuser.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 42 + } + ], + "page_idx": 13, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 752, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 107, + 79, + 167, + 95 + ], + "lines": [ + { + "bbox": [ + 104, + 78, + 170, + 97 + ], + "spans": [ + { + "bbox": [ + 104, + 78, + 170, + 97 + ], + "score": 1.0, + "content": "Appendix", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "text", + "bbox": [ + 106, + 104, + 505, + 181 + ], + "lines": [ + { + "bbox": [ + 106, + 104, + 505, + 117 + ], + "spans": [ + { + "bbox": [ + 106, + 104, + 505, + 117 + ], + "score": 1.0, + "content": "In this appendix, we discuss details of the illustrative examples in Section A. Next, we discuss", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 106, + 115, + 505, + 128 + ], + "spans": [ + { + "bbox": [ + 106, + 115, + 505, + 128 + ], + "score": 1.0, + "content": "hyperparameters and architectural details in Section B. We analyze the importance of low temperature", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 127, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 106, + 127, + 505, + 138 + ], + "score": 1.0, + "content": "sampling in Section C, further explain composition of conditioning variable in Section D, discuss the", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 138, + 505, + 149 + ], + "spans": [ + { + "bbox": [ + 106, + 138, + 505, + 149 + ], + "score": 1.0, + "content": "run-time characteristics of decision diffuser in Section E, discuss when to use inverse dynamics in", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 147, + 506, + 161 + ], + "spans": [ + { + "bbox": [ + 105, + 147, + 506, + 161 + ], + "score": 1.0, + "content": "Section F and analyze robustness of Decision Diffuser to stochastic dynamics in Section G. Finally,", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 159, + 506, + 171 + ], + "spans": [ + { + "bbox": [ + 105, + 159, + 506, + 171 + ], + "score": 1.0, + "content": "we provide details of the Kuka Block Stacking environment in Section H and the Unitree environment", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 170, + 158, + 182 + ], + "spans": [ + { + "bbox": [ + 106, + 170, + 158, + 182 + ], + "score": 1.0, + "content": "in Section I.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 4, + "bbox_fs": [ + 105, + 104, + 506, + 182 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 196, + 263, + 209 + ], + "lines": [ + { + "bbox": [ + 105, + 195, + 265, + 211 + ], + "spans": [ + { + "bbox": [ + 105, + 195, + 265, + 211 + ], + "score": 1.0, + "content": "A ILLUSTRATIVE EXAMPLES", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "title", + "bbox": [ + 107, + 222, + 287, + 234 + ], + "lines": [ + { + "bbox": [ + 106, + 222, + 288, + 235 + ], + "spans": [ + { + "bbox": [ + 106, + 222, + 288, + 235 + ], + "score": 1.0, + "content": "A.1 IMPLICIT DYNAMIC PROGRAMMING", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "image", + "bbox": [ + 207, + 240, + 404, + 313 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 207, + 240, + 404, + 313 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 207, + 240, + 404, + 313 + ], + "spans": [ + { + "bbox": [ + 207, + 240, + 404, + 313 + ], + "score": 0.965, + "type": "image", + "image_path": "06c49faf819f369ba151ad9d3ff002a905c14d4f9d446bec58e52f70a39c1d4b.jpg" + } + ] + } + ], + "index": 12, + "virtual_lines": [ + { + "bbox": [ + 207, + 240, + 404, + 254.6 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 207, + 254.6, + 404, + 269.2 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 207, + 269.2, + 404, + 283.8 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 207, + 283.8, + 404, + 298.40000000000003 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 207, + 298.40000000000003, + 404, + 313.00000000000006 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 323, + 504, + 354 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 323, + 505, + 335 + ], + "spans": [ + { + "bbox": [ + 106, + 323, + 505, + 335 + ], + "score": 1.0, + "content": "Figure A1: Illustrative example. We demonstrate the ability of Decision Diffuser to stitch together sub-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 333, + 506, + 347 + ], + "spans": [ + { + "bbox": [ + 105, + 333, + 506, + 347 + ], + "score": 1.0, + "content": "optimal trajectories in training dataset to obtain (near) optimal trajectories, thereby implicitly performing", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 344, + 380, + 355 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 380, + 355 + ], + "score": 1.0, + "content": "dynamic programming in Maze2D-open environment from Fu et al. (2020).", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16 + } + ], + "index": 14.0 + }, + { + "type": "text", + "bbox": [ + 107, + 364, + 505, + 442 + ], + "lines": [ + { + "bbox": [ + 104, + 361, + 506, + 379 + ], + "spans": [ + { + "bbox": [ + 104, + 361, + 506, + 379 + ], + "score": 1.0, + "content": "We empirically demonstrate the ability of Decision Diffuser to perform implicit dynamic programming", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 375, + 505, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 505, + 388 + ], + "score": 1.0, + "content": "in Maze2D-open environment from Fu et al. (2020). The task in Maze2D-open environment is to", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 386, + 505, + 398 + ], + "spans": [ + { + "bbox": [ + 106, + 386, + 505, + 398 + ], + "score": 1.0, + "content": "reach point C and the reward is negative distance from point C. The training dataset consists of 500", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 397, + 505, + 410 + ], + "spans": [ + { + "bbox": [ + 106, + 397, + 505, + 410 + ], + "score": 1.0, + "content": "trajectories from point A to point B and 500 trajectories from point B to point C. The maximum", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 408, + 506, + 420 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 506, + 420 + ], + "score": 1.0, + "content": "trajectory length is 50. During test time, the agent starts from point A and needs to reach point C", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 418, + 506, + 433 + ], + "spans": [ + { + "bbox": [ + 105, + 418, + 506, + 433 + ], + "score": 1.0, + "content": "as quickly as possible. As shown in Figure A1, Decision Diffuser can stitch trajectories in training", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 429, + 443, + 442 + ], + "spans": [ + { + "bbox": [ + 105, + 429, + 443, + 442 + ], + "score": 1.0, + "content": "dataset to form trajectories that goes from point A to point B in (near) straight lines.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 21, + "bbox_fs": [ + 104, + 361, + 506, + 442 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 453, + 255, + 465 + ], + "lines": [ + { + "bbox": [ + 106, + 452, + 257, + 467 + ], + "spans": [ + { + "bbox": [ + 106, + 452, + 257, + 467 + ], + "score": 1.0, + "content": "A.2 CONSTRAINT COMBINATION", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 106, + 471, + 506, + 548 + ], + "lines": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 506, + 484 + ], + "score": 1.0, + "content": "Setup In linear system robot navigation, Decision Diffuser is trained on 1000 expert trajectories", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 482, + 507, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 482, + 234, + 495 + ], + "score": 1.0, + "content": "either satisfying the constraint", + "type": "text" + }, + { + "bbox": [ + 235, + 482, + 280, + 494 + ], + "score": 0.86, + "content": "\\lVert s _ { T } \\rVert \\leq R", + "type": "inline_equation" + }, + { + "bbox": [ + 285, + 482, + 315, + 493 + ], + "score": 0.76, + "content": "R = 1", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 482, + 390, + 495 + ], + "score": 1.0, + "content": ") or the constraint", + "type": "text" + }, + { + "bbox": [ + 390, + 482, + 432, + 494 + ], + "score": 0.87, + "content": "\\| s _ { T } \\| \\geq r", + "type": "inline_equation" + }, + { + "bbox": [ + 437, + 483, + 473, + 493 + ], + "score": 0.67, + "content": "\\mathit { r } = 0 . 7", + "type": "inline_equation" + }, + { + "bbox": [ + 473, + 482, + 507, + 495 + ], + "score": 1.0, + "content": "). Here,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 107, + 493, + 506, + 506 + ], + "spans": [ + { + "bbox": [ + 107, + 493, + 167, + 505 + ], + "score": 0.92, + "content": "s _ { T } = [ x _ { T } , y _ { T } ]", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 493, + 506, + 506 + ], + "score": 1.0, + "content": "represents the final robot state in a trajectory, specifying its final 2d position. The", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 504, + 505, + 516 + ], + "spans": [ + { + "bbox": [ + 105, + 504, + 505, + 516 + ], + "score": 1.0, + "content": "maximum trajectory length is 50. During test time, Decision Diffuser is asked to generate trajectories", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 515, + 507, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 149, + 528 + ], + "score": 1.0, + "content": "satisfying", + "type": "text" + }, + { + "bbox": [ + 150, + 515, + 195, + 527 + ], + "score": 0.92, + "content": "\\| s _ { T } \\| \\le R", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 515, + 215, + 528 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 215, + 515, + 258, + 527 + ], + "score": 0.93, + "content": "\\| s _ { T } \\| \\geq r", + "type": "inline_equation" + }, + { + "bbox": [ + 258, + 515, + 507, + 528 + ], + "score": 1.0, + "content": "to test its ability to satisfy single constraints. Furthermore,", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 526, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 366, + 540 + ], + "score": 1.0, + "content": "Decision Diffuser is also asked to generate trajectories satisfying", + "type": "text" + }, + { + "bbox": [ + 366, + 526, + 428, + 538 + ], + "score": 0.94, + "content": "r \\leq \\left\\lceil | s _ { T } | | \\leq R \\right\\rceil", + "type": "inline_equation" + }, + { + "bbox": [ + 428, + 526, + 506, + 540 + ], + "score": 1.0, + "content": "to test its ability to", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 537, + 225, + 550 + ], + "spans": [ + { + "bbox": [ + 106, + 537, + 225, + 550 + ], + "score": 1.0, + "content": "satisfy combined constraints.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 29, + "bbox_fs": [ + 105, + 471, + 507, + 550 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 551, + 505, + 595 + ], + "lines": [ + { + "bbox": [ + 105, + 551, + 505, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 551, + 505, + 563 + ], + "score": 1.0, + "content": "Results Figure 2 shows that Decision Diffuser learns to generate trajectories perfectly (i.e. with", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 561, + 506, + 575 + ], + "spans": [ + { + "bbox": [ + 106, + 562, + 131, + 573 + ], + "score": 0.88, + "content": "1 0 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 131, + 561, + 506, + 575 + ], + "score": 1.0, + "content": "success rate) satisfying single constraints in linear system robot navigation. Furthermore, it", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 573, + 505, + 585 + ], + "spans": [ + { + "bbox": [ + 105, + 573, + 505, + 585 + ], + "score": 1.0, + "content": "learns to generate trajectories satisfying the composed constraint in linear system robot navigation", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 583, + 468, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 584, + 126, + 596 + ], + "score": 1.0, + "content": "with", + "type": "text" + }, + { + "bbox": [ + 127, + 583, + 190, + 596 + ], + "score": 0.95, + "content": "9 1 . 3 \\hat { \\% } ( \\pm 2 . 6 \\% )", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 584, + 468, + 596 + ], + "score": 1.0, + "content": "accuracy where the standard error is calculated over 5 random seeds.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 34.5, + "bbox_fs": [ + 105, + 551, + 506, + 596 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 611, + 390, + 623 + ], + "lines": [ + { + "bbox": [ + 105, + 609, + 392, + 625 + ], + "spans": [ + { + "bbox": [ + 105, + 609, + 392, + 625 + ], + "score": 1.0, + "content": "B HYPERPARAMETER AND ARCHITECTURAL DETAILS", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 37 + }, + { + "type": "text", + "bbox": [ + 106, + 635, + 412, + 647 + ], + "lines": [ + { + "bbox": [ + 105, + 634, + 414, + 649 + ], + "spans": [ + { + "bbox": [ + 105, + 634, + 414, + 649 + ], + "score": 1.0, + "content": "In this section, we describe various architectural and hyperparameter details:", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38, + "bbox_fs": [ + 105, + 634, + 414, + 649 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 654, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 653, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 653, + 232, + 668 + ], + "score": 1.0, + "content": "• We represent the noise model", + "type": "text" + }, + { + "bbox": [ + 233, + 657, + 243, + 666 + ], + "score": 0.84, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 243, + 653, + 506, + 668 + ], + "score": 1.0, + "content": "with a temporal U-Net (Janner et al., 2022), consisting of a U-Net", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 114, + 666, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 114, + 666, + 505, + 678 + ], + "score": 1.0, + "content": "structure with 6 repeated residual blocks. Each block consisted of two temporal convolutions, each", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 113, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 113, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "followed by group norm (Wu & He, 2018), and a final Mish nonlinearity (Misra, 2019). Timestep", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 114, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 114, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "and condition embeddings, both 128-dimensional vectors, are produced by separate 2-layered MLP", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 114, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 114, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "(with 256 hidden units and Mish nonlinearity) and are concatenated together before getting added", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 114, + 710, + 505, + 721 + ], + "spans": [ + { + "bbox": [ + 114, + 710, + 505, + 721 + ], + "score": 1.0, + "content": "to the activations of the first temporal convolution within each block. We borrow the code for", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 114, + 721, + 349, + 733 + ], + "spans": [ + { + "bbox": [ + 114, + 721, + 349, + 733 + ], + "score": 1.0, + "content": "temporal U-Net from https://github.com/jannerm/diffuser.", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 42, + "bbox_fs": [ + 105, + 653, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 82, + 504, + 105 + ], + "lines": [ + { + "bbox": [ + 104, + 80, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 104, + 80, + 260, + 95 + ], + "score": 1.0, + "content": "• We represent the inverse dynamics", + "type": "text" + }, + { + "bbox": [ + 261, + 83, + 272, + 95 + ], + "score": 0.88, + "content": "f _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 80, + 505, + 95 + ], + "score": 1.0, + "content": "with a 2-layered MLP with 512 hidden units and ReLU", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 114, + 94, + 163, + 106 + ], + "spans": [ + { + "bbox": [ + 114, + 94, + 163, + 106 + ], + "score": 1.0, + "content": "activations.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 106, + 114, + 504, + 126 + ], + "lines": [ + { + "bbox": [ + 104, + 112, + 506, + 128 + ], + "spans": [ + { + "bbox": [ + 104, + 112, + 506, + 128 + ], + "score": 1.0, + "content": "• We represent the gait classifier with a 3-layered MLP with 1024 hidden units and ReLU activations.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 2 + }, + { + "type": "text", + "bbox": [ + 107, + 135, + 504, + 158 + ], + "lines": [ + { + "bbox": [ + 105, + 133, + 505, + 148 + ], + "spans": [ + { + "bbox": [ + 105, + 133, + 150, + 148 + ], + "score": 1.0, + "content": "• We train", + "type": "text" + }, + { + "bbox": [ + 150, + 137, + 160, + 146 + ], + "score": 0.81, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 160, + 133, + 178, + 148 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 179, + 136, + 190, + 147 + ], + "score": 0.88, + "content": "f _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 133, + 476, + 148 + ], + "score": 1.0, + "content": "using the Adam optimizer (Kingma & Ba, 2015) with a learning rate of", + "type": "text" + }, + { + "bbox": [ + 477, + 136, + 505, + 146 + ], + "score": 0.81, + "content": "2 e - 4", + "type": "inline_equation" + } + ], + "index": 3 + }, + { + "bbox": [ + 114, + 145, + 274, + 159 + ], + "spans": [ + { + "bbox": [ + 114, + 145, + 274, + 159 + ], + "score": 1.0, + "content": "and batch size of 32 for 2e6 train steps.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 106, + 167, + 505, + 190 + ], + "lines": [ + { + "bbox": [ + 105, + 165, + 505, + 180 + ], + "spans": [ + { + "bbox": [ + 105, + 165, + 417, + 180 + ], + "score": 1.0, + "content": "• We train the gait classifier using the Adam optimizer with a learning rate of", + "type": "text" + }, + { + "bbox": [ + 417, + 168, + 445, + 178 + ], + "score": 0.7, + "content": "2 e - 4", + "type": "inline_equation" + }, + { + "bbox": [ + 446, + 165, + 505, + 180 + ], + "score": 1.0, + "content": "and batch size", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 114, + 177, + 215, + 191 + ], + "spans": [ + { + "bbox": [ + 114, + 177, + 215, + 191 + ], + "score": 1.0, + "content": "of 64 for 1e6 train steps.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + }, + { + "type": "text", + "bbox": [ + 108, + 199, + 441, + 211 + ], + "lines": [ + { + "bbox": [ + 105, + 198, + 442, + 212 + ], + "spans": [ + { + "bbox": [ + 105, + 198, + 221, + 212 + ], + "score": 1.0, + "content": "• We choose the probability", + "type": "text" + }, + { + "bbox": [ + 221, + 201, + 227, + 210 + ], + "score": 0.79, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 198, + 442, + 212 + ], + "score": 1.0, + "content": "of removing the conditioning information to be 0.25.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 107, + 220, + 248, + 232 + ], + "lines": [ + { + "bbox": [ + 105, + 218, + 249, + 234 + ], + "spans": [ + { + "bbox": [ + 105, + 218, + 145, + 234 + ], + "score": 1.0, + "content": "• We use", + "type": "text" + }, + { + "bbox": [ + 146, + 221, + 185, + 231 + ], + "score": 0.89, + "content": "K = 1 0 0", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 218, + 249, + 234 + ], + "score": 1.0, + "content": "diffusion steps.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 108, + 241, + 506, + 274 + ], + "lines": [ + { + "bbox": [ + 105, + 240, + 506, + 253 + ], + "spans": [ + { + "bbox": [ + 105, + 240, + 223, + 253 + ], + "score": 1.0, + "content": "• We use a planning horizon", + "type": "text" + }, + { + "bbox": [ + 223, + 242, + 233, + 251 + ], + "score": 0.81, + "content": "H", + "type": "inline_equation" + }, + { + "bbox": [ + 234, + 240, + 506, + 253 + ], + "score": 1.0, + "content": "of 100 in all the D4RL locomotion tasks, 56 in D4RL kitchen tasks,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 114, + 252, + 506, + 264 + ], + "spans": [ + { + "bbox": [ + 114, + 252, + 506, + 264 + ], + "score": 1.0, + "content": "128 in Kuka block stacking, 56 in unitree-go-running tasks, 50 in the illustrative example and 60 in", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 114, + 262, + 188, + 276 + ], + "spans": [ + { + "bbox": [ + 114, + 262, + 188, + 276 + ], + "score": 1.0, + "content": "Block push tasks.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + }, + { + "type": "text", + "bbox": [ + 108, + 283, + 447, + 296 + ], + "lines": [ + { + "bbox": [ + 105, + 283, + 448, + 297 + ], + "spans": [ + { + "bbox": [ + 105, + 283, + 213, + 297 + ], + "score": 1.0, + "content": "• We use a guidance scale", + "type": "text" + }, + { + "bbox": [ + 214, + 284, + 306, + 296 + ], + "score": 0.91, + "content": "s \\in \\{ 1 . 2 , 1 . 4 , 1 . 6 , 1 . 8 \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 306, + 283, + 448, + 297 + ], + "score": 1.0, + "content": "but the exact choice varies by task.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 12 + }, + { + "type": "text", + "bbox": [ + 106, + 304, + 317, + 317 + ], + "lines": [ + { + "bbox": [ + 105, + 302, + 318, + 320 + ], + "spans": [ + { + "bbox": [ + 105, + 302, + 160, + 320 + ], + "score": 1.0, + "content": "• We choose", + "type": "text" + }, + { + "bbox": [ + 160, + 306, + 194, + 315 + ], + "score": 0.88, + "content": "\\alpha = 0 . 5", + "type": "inline_equation" + }, + { + "bbox": [ + 194, + 302, + 318, + 320 + ], + "score": 1.0, + "content": "for low temperature sampling.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 107, + 326, + 254, + 337 + ], + "lines": [ + { + "bbox": [ + 106, + 324, + 252, + 339 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 219, + 339 + ], + "score": 1.0, + "content": "• We choose context length", + "type": "text" + }, + { + "bbox": [ + 219, + 326, + 252, + 336 + ], + "score": 0.89, + "content": "C = 2 0", + "type": "inline_equation" + } + ], + "index": 14 + } + ], + "index": 14 + }, + { + "type": "title", + "bbox": [ + 106, + 360, + 374, + 372 + ], + "lines": [ + { + "bbox": [ + 106, + 358, + 375, + 374 + ], + "spans": [ + { + "bbox": [ + 106, + 358, + 375, + 374 + ], + "score": 1.0, + "content": "C IMPORTANCE OF LOW TEMPERATURE SAMPLING", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 106, + 389, + 506, + 467 + ], + "lines": [ + { + "bbox": [ + 105, + 389, + 505, + 402 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 219, + 402 + ], + "score": 1.0, + "content": "In Algorithm 1, we compute", + "type": "text" + }, + { + "bbox": [ + 220, + 391, + 242, + 401 + ], + "score": 0.85, + "content": "\\mu _ { k - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 242, + 389, + 259, + 402 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 259, + 390, + 282, + 400 + ], + "score": 0.89, + "content": "\\Sigma _ { k - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 283, + 389, + 505, + 402 + ], + "score": 1.0, + "content": "from a noisy sequence of states and predicted noise. We", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 104, + 399, + 506, + 414 + ], + "spans": [ + { + "bbox": [ + 104, + 399, + 182, + 414 + ], + "score": 1.0, + "content": "find that sampling", + "type": "text" + }, + { + "bbox": [ + 182, + 401, + 289, + 412 + ], + "score": 0.91, + "content": "x _ { k - 1 } \\sim { \\mathcal { N } } ( \\mu _ { k - 1 } , \\alpha \\Sigma _ { k - 1 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 399, + 321, + 414 + ], + "score": 1.0, + "content": "(where", + "type": "text" + }, + { + "bbox": [ + 321, + 400, + 362, + 412 + ], + "score": 0.91, + "content": "\\alpha \\in [ 0 , 1 )", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 399, + 506, + 414 + ], + "score": 1.0, + "content": ") with a reduced variance produces", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 411, + 505, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 505, + 424 + ], + "score": 1.0, + "content": "high-likelihood state sequences. We refer to this as low-temperature sampling. To empirically show", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 421, + 506, + 435 + ], + "spans": [ + { + "bbox": [ + 105, + 421, + 450, + 435 + ], + "score": 1.0, + "content": "its importance, we compare performances of Decision Diffuser with different values of", + "type": "text" + }, + { + "bbox": [ + 451, + 424, + 459, + 432 + ], + "score": 0.75, + "content": "\\alpha", + "type": "inline_equation" + }, + { + "bbox": [ + 459, + 421, + 506, + 435 + ], + "score": 1.0, + "content": "(Table A1).", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 432, + 506, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 269, + 447 + ], + "score": 1.0, + "content": "We show that low temperature sampling (", + "type": "text" + }, + { + "bbox": [ + 269, + 433, + 304, + 444 + ], + "score": 0.84, + "content": "\\alpha = 0 . 5", + "type": "inline_equation" + }, + { + "bbox": [ + 304, + 432, + 506, + 447 + ], + "score": 1.0, + "content": ") gives the best average returns. However, reducing", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 443, + 504, + 457 + ], + "spans": [ + { + "bbox": [ + 105, + 443, + 120, + 457 + ], + "score": 1.0, + "content": "the", + "type": "text" + }, + { + "bbox": [ + 121, + 445, + 129, + 454 + ], + "score": 0.8, + "content": "\\alpha", + "type": "inline_equation" + }, + { + "bbox": [ + 129, + 443, + 470, + 457 + ], + "score": 1.0, + "content": "to 0 eliminates the entropy in sampling and leads to lower returns. On the other hand,", + "type": "text" + }, + { + "bbox": [ + 470, + 444, + 504, + 455 + ], + "score": 0.87, + "content": "\\alpha = 1 . 0", + "type": "inline_equation" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 455, + 361, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 361, + 468 + ], + "score": 1.0, + "content": "leads to a higher variance in terms of returns of the trajectories.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 19 + }, + { + "type": "table", + "bbox": [ + 212, + 483, + 397, + 537 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 212, + 483, + 397, + 537 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 212, + 483, + 397, + 537 + ], + "spans": [ + { + "bbox": [ + 212, + 483, + 397, + 537 + ], + "score": 0.974, + "html": "
Decision DiffuserHopper-Medium-Expert
α=0104.3 ± 0.7
α= 0.5111.8 ±1.6
α= 1.0107.1 ± 3.5
", + "type": "table", + "image_path": "e59f79eab126e624fbc836f6cebf586440a9d6ef00746eabcad19e1c336e7e3d.jpg" + } + ] + } + ], + "index": 24, + "virtual_lines": [ + { + "bbox": [ + 212, + 483, + 397, + 501.0 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 212, + 501.0, + 397, + 519.0 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 212, + 519.0, + 397, + 537.0 + ], + "spans": [], + "index": 25 + } + ] + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 106, + 545, + 506, + 578 + ], + "lines": [ + { + "bbox": [ + 104, + 543, + 506, + 558 + ], + "spans": [ + { + "bbox": [ + 104, + 543, + 260, + 558 + ], + "score": 1.0, + "content": "Table A1: Low-temperature sampling", + "type": "text" + }, + { + "bbox": [ + 260, + 545, + 296, + 556 + ], + "score": 0.87, + "content": "( \\alpha = 0 . 5 )", + "type": "inline_equation" + }, + { + "bbox": [ + 296, + 543, + 506, + 558 + ], + "score": 1.0, + "content": ") allows us to get high return trajectories consistently.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 555, + 505, + 568 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 132, + 568 + ], + "score": 1.0, + "content": "While", + "type": "text" + }, + { + "bbox": [ + 132, + 556, + 167, + 566 + ], + "score": 0.89, + "content": "\\alpha = 1 . 0", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 555, + 384, + 568 + ], + "score": 1.0, + "content": "leads to a higher variance in returns of the trajectories,", + "type": "text" + }, + { + "bbox": [ + 384, + 556, + 419, + 566 + ], + "score": 0.9, + "content": "\\alpha = 0 . 0", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 555, + 505, + 568 + ], + "score": 1.0, + "content": "eliminates entropy in", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 567, + 267, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 567, + 267, + 578 + ], + "score": 1.0, + "content": "the sampling and leads to lower returns.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27 + }, + { + "type": "title", + "bbox": [ + 107, + 606, + 334, + 619 + ], + "lines": [ + { + "bbox": [ + 105, + 605, + 334, + 622 + ], + "spans": [ + { + "bbox": [ + 105, + 605, + 334, + 622 + ], + "score": 1.0, + "content": "D COMPOSING CONDITIONING VARIABLES", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 106, + 635, + 505, + 704 + ], + "lines": [ + { + "bbox": [ + 106, + 635, + 505, + 648 + ], + "spans": [ + { + "bbox": [ + 106, + 635, + 505, + 648 + ], + "score": 1.0, + "content": "In this section, we detail how Decision Diffuser trained with different conditioning variables", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 107, + 645, + 505, + 662 + ], + "spans": [ + { + "bbox": [ + 107, + 646, + 154, + 659 + ], + "score": 0.9, + "content": "\\{ \\pmb { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 154, + 645, + 505, + 662 + ], + "score": 1.0, + "content": "composes these conditioning variables together. It learns the denoising model", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 107, + 657, + 506, + 670 + ], + "spans": [ + { + "bbox": [ + 107, + 658, + 186, + 670 + ], + "score": 0.88, + "content": "\\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , { \\pmb y } ^ { i } ( \\tau ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 186, + 657, + 338, + 670 + ], + "score": 1.0, + "content": "for a given conditioning variable", + "type": "text" + }, + { + "bbox": [ + 338, + 658, + 362, + 670 + ], + "score": 0.92, + "content": "\\mathbf { \\dot { \\mathbf { y } } } ^ { i } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 657, + 506, + 670 + ], + "score": 1.0, + "content": ". From the derivations outlined", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 667, + 505, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 667, + 379, + 684 + ], + "score": 1.0, + "content": "in prior works (Luo, 2022; Song et al., 2021), we know that", + "type": "text" + }, + { + "bbox": [ + 380, + 668, + 505, + 681 + ], + "score": 0.89, + "content": "\\nabla _ { \\pmb { x } _ { k } ( \\tau ) } \\log q ( \\pmb { x } _ { k } ( \\tau ) | \\pmb { y } ^ { i } ( \\tau ) ) \\propto", + "type": "inline_equation" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 679, + 505, + 697 + ], + "spans": [ + { + "bbox": [ + 106, + 681, + 194, + 694 + ], + "score": 0.91, + "content": "- \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , { \\pmb y } ^ { i } ( \\tau ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 194, + 679, + 400, + 697 + ], + "score": 1.0, + "content": ". Therefore, each conditional trajectory distribution", + "type": "text" + }, + { + "bbox": [ + 401, + 681, + 487, + 694 + ], + "score": 0.9, + "content": "\\{ q ( \\pmb { x } _ { k } ( \\tau ) | \\pmb { y } ^ { i } ( \\tau ) ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 488, + 679, + 505, + 697 + ], + "score": 1.0, + "content": "can", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 691, + 485, + 705 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 280, + 705 + ], + "score": 1.0, + "content": "be modelled with a single denoising model", + "type": "text" + }, + { + "bbox": [ + 280, + 694, + 290, + 704 + ], + "score": 0.84, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 691, + 456, + 705 + ], + "score": 1.0, + "content": "that conditions on the respective variable", + "type": "text" + }, + { + "bbox": [ + 457, + 694, + 480, + 705 + ], + "score": 0.9, + "content": "y ^ { i } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 480, + 691, + 485, + 705 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 32.5 + }, + { + "type": "text", + "bbox": [ + 105, + 709, + 504, + 733 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 192, + 722 + ], + "score": 1.0, + "content": "In order to compose", + "type": "text" + }, + { + "bbox": [ + 193, + 712, + 200, + 720 + ], + "score": 0.73, + "content": "n", + "type": "inline_equation" + }, + { + "bbox": [ + 201, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "different conditioning variables (i.e. skills or constraints), we would like", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 719, + 506, + 735 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 146, + 735 + ], + "score": 1.0, + "content": "to model", + "type": "text" + }, + { + "bbox": [ + 146, + 720, + 233, + 733 + ], + "score": 0.93, + "content": "q ( \\pmb { x } _ { k } ( \\hat { \\tau } ) | \\{ \\pmb { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } )", + "type": "inline_equation" + }, + { + "bbox": [ + 233, + 719, + 307, + 735 + ], + "score": 1.0, + "content": ". We assume that", + "type": "text" + }, + { + "bbox": [ + 307, + 720, + 354, + 733 + ], + "score": 0.94, + "content": "\\{ y ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 354, + 719, + 506, + 735 + ], + "score": 1.0, + "content": "are conditionally independent given", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5 + } + ], + "page_idx": 14, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 82, + 504, + 105 + ], + "lines": [ + { + "bbox": [ + 104, + 80, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 104, + 80, + 260, + 95 + ], + "score": 1.0, + "content": "• We represent the inverse dynamics", + "type": "text" + }, + { + "bbox": [ + 261, + 83, + 272, + 95 + ], + "score": 0.88, + "content": "f _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 272, + 80, + 505, + 95 + ], + "score": 1.0, + "content": "with a 2-layered MLP with 512 hidden units and ReLU", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 114, + 94, + 163, + 106 + ], + "spans": [ + { + "bbox": [ + 114, + 94, + 163, + 106 + ], + "score": 1.0, + "content": "activations.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5, + "bbox_fs": [ + 104, + 80, + 505, + 106 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 114, + 504, + 126 + ], + "lines": [ + { + "bbox": [ + 104, + 112, + 506, + 128 + ], + "spans": [ + { + "bbox": [ + 104, + 112, + 506, + 128 + ], + "score": 1.0, + "content": "• We represent the gait classifier with a 3-layered MLP with 1024 hidden units and ReLU activations.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 2, + "bbox_fs": [ + 104, + 112, + 506, + 128 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 135, + 504, + 158 + ], + "lines": [ + { + "bbox": [ + 105, + 133, + 505, + 148 + ], + "spans": [ + { + "bbox": [ + 105, + 133, + 150, + 148 + ], + "score": 1.0, + "content": "• We train", + "type": "text" + }, + { + "bbox": [ + 150, + 137, + 160, + 146 + ], + "score": 0.81, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 160, + 133, + 178, + 148 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 179, + 136, + 190, + 147 + ], + "score": 0.88, + "content": "f _ { \\phi }", + "type": "inline_equation" + }, + { + "bbox": [ + 190, + 133, + 476, + 148 + ], + "score": 1.0, + "content": "using the Adam optimizer (Kingma & Ba, 2015) with a learning rate of", + "type": "text" + }, + { + "bbox": [ + 477, + 136, + 505, + 146 + ], + "score": 0.81, + "content": "2 e - 4", + "type": "inline_equation" + } + ], + "index": 3 + }, + { + "bbox": [ + 114, + 145, + 274, + 159 + ], + "spans": [ + { + "bbox": [ + 114, + 145, + 274, + 159 + ], + "score": 1.0, + "content": "and batch size of 32 for 2e6 train steps.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5, + "bbox_fs": [ + 105, + 133, + 505, + 159 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 167, + 505, + 190 + ], + "lines": [ + { + "bbox": [ + 105, + 165, + 505, + 180 + ], + "spans": [ + { + "bbox": [ + 105, + 165, + 417, + 180 + ], + "score": 1.0, + "content": "• We train the gait classifier using the Adam optimizer with a learning rate of", + "type": "text" + }, + { + "bbox": [ + 417, + 168, + 445, + 178 + ], + "score": 0.7, + "content": "2 e - 4", + "type": "inline_equation" + }, + { + "bbox": [ + 446, + 165, + 505, + 180 + ], + "score": 1.0, + "content": "and batch size", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 114, + 177, + 215, + 191 + ], + "spans": [ + { + "bbox": [ + 114, + 177, + 215, + 191 + ], + "score": 1.0, + "content": "of 64 for 1e6 train steps.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5, + "bbox_fs": [ + 105, + 165, + 505, + 191 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 199, + 441, + 211 + ], + "lines": [ + { + "bbox": [ + 105, + 198, + 442, + 212 + ], + "spans": [ + { + "bbox": [ + 105, + 198, + 221, + 212 + ], + "score": 1.0, + "content": "• We choose the probability", + "type": "text" + }, + { + "bbox": [ + 221, + 201, + 227, + 210 + ], + "score": 0.79, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 228, + 198, + 442, + 212 + ], + "score": 1.0, + "content": "of removing the conditioning information to be 0.25.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 7, + "bbox_fs": [ + 105, + 198, + 442, + 212 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 220, + 248, + 232 + ], + "lines": [ + { + "bbox": [ + 105, + 218, + 249, + 234 + ], + "spans": [ + { + "bbox": [ + 105, + 218, + 145, + 234 + ], + "score": 1.0, + "content": "• We use", + "type": "text" + }, + { + "bbox": [ + 146, + 221, + 185, + 231 + ], + "score": 0.89, + "content": "K = 1 0 0", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 218, + 249, + 234 + ], + "score": 1.0, + "content": "diffusion steps.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8, + "bbox_fs": [ + 105, + 218, + 249, + 234 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 241, + 506, + 274 + ], + "lines": [ + { + "bbox": [ + 105, + 240, + 506, + 253 + ], + "spans": [ + { + "bbox": [ + 105, + 240, + 223, + 253 + ], + "score": 1.0, + "content": "• We use a planning horizon", + "type": "text" + }, + { + "bbox": [ + 223, + 242, + 233, + 251 + ], + "score": 0.81, + "content": "H", + "type": "inline_equation" + }, + { + "bbox": [ + 234, + 240, + 506, + 253 + ], + "score": 1.0, + "content": "of 100 in all the D4RL locomotion tasks, 56 in D4RL kitchen tasks,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 114, + 252, + 506, + 264 + ], + "spans": [ + { + "bbox": [ + 114, + 252, + 506, + 264 + ], + "score": 1.0, + "content": "128 in Kuka block stacking, 56 in unitree-go-running tasks, 50 in the illustrative example and 60 in", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 114, + 262, + 188, + 276 + ], + "spans": [ + { + "bbox": [ + 114, + 262, + 188, + 276 + ], + "score": 1.0, + "content": "Block push tasks.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10, + "bbox_fs": [ + 105, + 240, + 506, + 276 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 283, + 447, + 296 + ], + "lines": [ + { + "bbox": [ + 105, + 283, + 448, + 297 + ], + "spans": [ + { + "bbox": [ + 105, + 283, + 213, + 297 + ], + "score": 1.0, + "content": "• We use a guidance scale", + "type": "text" + }, + { + "bbox": [ + 214, + 284, + 306, + 296 + ], + "score": 0.91, + "content": "s \\in \\{ 1 . 2 , 1 . 4 , 1 . 6 , 1 . 8 \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 306, + 283, + 448, + 297 + ], + "score": 1.0, + "content": "but the exact choice varies by task.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 12, + "bbox_fs": [ + 105, + 283, + 448, + 297 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 304, + 317, + 317 + ], + "lines": [ + { + "bbox": [ + 105, + 302, + 318, + 320 + ], + "spans": [ + { + "bbox": [ + 105, + 302, + 160, + 320 + ], + "score": 1.0, + "content": "• We choose", + "type": "text" + }, + { + "bbox": [ + 160, + 306, + 194, + 315 + ], + "score": 0.88, + "content": "\\alpha = 0 . 5", + "type": "inline_equation" + }, + { + "bbox": [ + 194, + 302, + 318, + 320 + ], + "score": 1.0, + "content": "for low temperature sampling.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 13, + "bbox_fs": [ + 105, + 302, + 318, + 320 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 326, + 254, + 337 + ], + "lines": [ + { + "bbox": [ + 106, + 324, + 252, + 339 + ], + "spans": [ + { + "bbox": [ + 106, + 324, + 219, + 339 + ], + "score": 1.0, + "content": "• We choose context length", + "type": "text" + }, + { + "bbox": [ + 219, + 326, + 252, + 336 + ], + "score": 0.89, + "content": "C = 2 0", + "type": "inline_equation" + } + ], + "index": 14 + } + ], + "index": 14, + "bbox_fs": [ + 106, + 324, + 252, + 339 + ] + }, + { + "type": "title", + "bbox": [ + 106, + 360, + 374, + 372 + ], + "lines": [ + { + "bbox": [ + 106, + 358, + 375, + 374 + ], + "spans": [ + { + "bbox": [ + 106, + 358, + 375, + 374 + ], + "score": 1.0, + "content": "C IMPORTANCE OF LOW TEMPERATURE SAMPLING", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 106, + 389, + 506, + 467 + ], + "lines": [ + { + "bbox": [ + 105, + 389, + 505, + 402 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 219, + 402 + ], + "score": 1.0, + "content": "In Algorithm 1, we compute", + "type": "text" + }, + { + "bbox": [ + 220, + 391, + 242, + 401 + ], + "score": 0.85, + "content": "\\mu _ { k - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 242, + 389, + 259, + 402 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 259, + 390, + 282, + 400 + ], + "score": 0.89, + "content": "\\Sigma _ { k - 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 283, + 389, + 505, + 402 + ], + "score": 1.0, + "content": "from a noisy sequence of states and predicted noise. We", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 104, + 399, + 506, + 414 + ], + "spans": [ + { + "bbox": [ + 104, + 399, + 182, + 414 + ], + "score": 1.0, + "content": "find that sampling", + "type": "text" + }, + { + "bbox": [ + 182, + 401, + 289, + 412 + ], + "score": 0.91, + "content": "x _ { k - 1 } \\sim { \\mathcal { N } } ( \\mu _ { k - 1 } , \\alpha \\Sigma _ { k - 1 } )", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 399, + 321, + 414 + ], + "score": 1.0, + "content": "(where", + "type": "text" + }, + { + "bbox": [ + 321, + 400, + 362, + 412 + ], + "score": 0.91, + "content": "\\alpha \\in [ 0 , 1 )", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 399, + 506, + 414 + ], + "score": 1.0, + "content": ") with a reduced variance produces", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 411, + 505, + 424 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 505, + 424 + ], + "score": 1.0, + "content": "high-likelihood state sequences. We refer to this as low-temperature sampling. To empirically show", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 421, + 506, + 435 + ], + "spans": [ + { + "bbox": [ + 105, + 421, + 450, + 435 + ], + "score": 1.0, + "content": "its importance, we compare performances of Decision Diffuser with different values of", + "type": "text" + }, + { + "bbox": [ + 451, + 424, + 459, + 432 + ], + "score": 0.75, + "content": "\\alpha", + "type": "inline_equation" + }, + { + "bbox": [ + 459, + 421, + 506, + 435 + ], + "score": 1.0, + "content": "(Table A1).", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 432, + 506, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 432, + 269, + 447 + ], + "score": 1.0, + "content": "We show that low temperature sampling (", + "type": "text" + }, + { + "bbox": [ + 269, + 433, + 304, + 444 + ], + "score": 0.84, + "content": "\\alpha = 0 . 5", + "type": "inline_equation" + }, + { + "bbox": [ + 304, + 432, + 506, + 447 + ], + "score": 1.0, + "content": ") gives the best average returns. However, reducing", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 443, + 504, + 457 + ], + "spans": [ + { + "bbox": [ + 105, + 443, + 120, + 457 + ], + "score": 1.0, + "content": "the", + "type": "text" + }, + { + "bbox": [ + 121, + 445, + 129, + 454 + ], + "score": 0.8, + "content": "\\alpha", + "type": "inline_equation" + }, + { + "bbox": [ + 129, + 443, + 470, + 457 + ], + "score": 1.0, + "content": "to 0 eliminates the entropy in sampling and leads to lower returns. On the other hand,", + "type": "text" + }, + { + "bbox": [ + 470, + 444, + 504, + 455 + ], + "score": 0.87, + "content": "\\alpha = 1 . 0", + "type": "inline_equation" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 455, + 361, + 468 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 361, + 468 + ], + "score": 1.0, + "content": "leads to a higher variance in terms of returns of the trajectories.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 19, + "bbox_fs": [ + 104, + 389, + 506, + 468 + ] + }, + { + "type": "table", + "bbox": [ + 212, + 483, + 397, + 537 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 212, + 483, + 397, + 537 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 212, + 483, + 397, + 537 + ], + "spans": [ + { + "bbox": [ + 212, + 483, + 397, + 537 + ], + "score": 0.974, + "html": "
Decision DiffuserHopper-Medium-Expert
α=0104.3 ± 0.7
α= 0.5111.8 ±1.6
α= 1.0107.1 ± 3.5
", + "type": "table", + "image_path": "e59f79eab126e624fbc836f6cebf586440a9d6ef00746eabcad19e1c336e7e3d.jpg" + } + ] + } + ], + "index": 24, + "virtual_lines": [ + { + "bbox": [ + 212, + 483, + 397, + 501.0 + ], + "spans": [], + "index": 23 + }, + { + "bbox": [ + 212, + 501.0, + 397, + 519.0 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 212, + 519.0, + 397, + 537.0 + ], + "spans": [], + "index": 25 + } + ] + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 106, + 545, + 506, + 578 + ], + "lines": [ + { + "bbox": [ + 104, + 543, + 506, + 558 + ], + "spans": [ + { + "bbox": [ + 104, + 543, + 260, + 558 + ], + "score": 1.0, + "content": "Table A1: Low-temperature sampling", + "type": "text" + }, + { + "bbox": [ + 260, + 545, + 296, + 556 + ], + "score": 0.87, + "content": "( \\alpha = 0 . 5 )", + "type": "inline_equation" + }, + { + "bbox": [ + 296, + 543, + 506, + 558 + ], + "score": 1.0, + "content": ") allows us to get high return trajectories consistently.", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 555, + 505, + 568 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 132, + 568 + ], + "score": 1.0, + "content": "While", + "type": "text" + }, + { + "bbox": [ + 132, + 556, + 167, + 566 + ], + "score": 0.89, + "content": "\\alpha = 1 . 0", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 555, + 384, + 568 + ], + "score": 1.0, + "content": "leads to a higher variance in returns of the trajectories,", + "type": "text" + }, + { + "bbox": [ + 384, + 556, + 419, + 566 + ], + "score": 0.9, + "content": "\\alpha = 0 . 0", + "type": "inline_equation" + }, + { + "bbox": [ + 419, + 555, + 505, + 568 + ], + "score": 1.0, + "content": "eliminates entropy in", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 567, + 267, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 567, + 267, + 578 + ], + "score": 1.0, + "content": "the sampling and leads to lower returns.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27, + "bbox_fs": [ + 104, + 543, + 506, + 578 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 606, + 334, + 619 + ], + "lines": [ + { + "bbox": [ + 105, + 605, + 334, + 622 + ], + "spans": [ + { + "bbox": [ + 105, + 605, + 334, + 622 + ], + "score": 1.0, + "content": "D COMPOSING CONDITIONING VARIABLES", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 106, + 635, + 505, + 704 + ], + "lines": [ + { + "bbox": [ + 106, + 635, + 505, + 648 + ], + "spans": [ + { + "bbox": [ + 106, + 635, + 505, + 648 + ], + "score": 1.0, + "content": "In this section, we detail how Decision Diffuser trained with different conditioning variables", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 107, + 645, + 505, + 662 + ], + "spans": [ + { + "bbox": [ + 107, + 646, + 154, + 659 + ], + "score": 0.9, + "content": "\\{ \\pmb { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 154, + 645, + 505, + 662 + ], + "score": 1.0, + "content": "composes these conditioning variables together. It learns the denoising model", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 107, + 657, + 506, + 670 + ], + "spans": [ + { + "bbox": [ + 107, + 658, + 186, + 670 + ], + "score": 0.88, + "content": "\\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , { \\pmb y } ^ { i } ( \\tau ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 186, + 657, + 338, + 670 + ], + "score": 1.0, + "content": "for a given conditioning variable", + "type": "text" + }, + { + "bbox": [ + 338, + 658, + 362, + 670 + ], + "score": 0.92, + "content": "\\mathbf { \\dot { \\mathbf { y } } } ^ { i } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 657, + 506, + 670 + ], + "score": 1.0, + "content": ". From the derivations outlined", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 667, + 505, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 667, + 379, + 684 + ], + "score": 1.0, + "content": "in prior works (Luo, 2022; Song et al., 2021), we know that", + "type": "text" + }, + { + "bbox": [ + 380, + 668, + 505, + 681 + ], + "score": 0.89, + "content": "\\nabla _ { \\pmb { x } _ { k } ( \\tau ) } \\log q ( \\pmb { x } _ { k } ( \\tau ) | \\pmb { y } ^ { i } ( \\tau ) ) \\propto", + "type": "inline_equation" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 679, + 505, + 697 + ], + "spans": [ + { + "bbox": [ + 106, + 681, + 194, + 694 + ], + "score": 0.91, + "content": "- \\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , { \\pmb y } ^ { i } ( \\tau ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 194, + 679, + 400, + 697 + ], + "score": 1.0, + "content": ". Therefore, each conditional trajectory distribution", + "type": "text" + }, + { + "bbox": [ + 401, + 681, + 487, + 694 + ], + "score": 0.9, + "content": "\\{ q ( \\pmb { x } _ { k } ( \\tau ) | \\pmb { y } ^ { i } ( \\tau ) ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 488, + 679, + 505, + 697 + ], + "score": 1.0, + "content": "can", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 691, + 485, + 705 + ], + "spans": [ + { + "bbox": [ + 105, + 691, + 280, + 705 + ], + "score": 1.0, + "content": "be modelled with a single denoising model", + "type": "text" + }, + { + "bbox": [ + 280, + 694, + 290, + 704 + ], + "score": 0.84, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 691, + 456, + 705 + ], + "score": 1.0, + "content": "that conditions on the respective variable", + "type": "text" + }, + { + "bbox": [ + 457, + 694, + 480, + 705 + ], + "score": 0.9, + "content": "y ^ { i } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 480, + 691, + 485, + 705 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 32.5, + "bbox_fs": [ + 105, + 635, + 506, + 705 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 709, + 504, + 733 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 192, + 722 + ], + "score": 1.0, + "content": "In order to compose", + "type": "text" + }, + { + "bbox": [ + 193, + 712, + 200, + 720 + ], + "score": 0.73, + "content": "n", + "type": "inline_equation" + }, + { + "bbox": [ + 201, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "different conditioning variables (i.e. skills or constraints), we would like", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 719, + 506, + 735 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 146, + 735 + ], + "score": 1.0, + "content": "to model", + "type": "text" + }, + { + "bbox": [ + 146, + 720, + 233, + 733 + ], + "score": 0.93, + "content": "q ( \\pmb { x } _ { k } ( \\hat { \\tau } ) | \\{ \\pmb { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } )", + "type": "inline_equation" + }, + { + "bbox": [ + 233, + 719, + 307, + 735 + ], + "score": 1.0, + "content": ". We assume that", + "type": "text" + }, + { + "bbox": [ + 307, + 720, + 354, + 733 + ], + "score": 0.94, + "content": "\\{ y ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 354, + 719, + 506, + 735 + ], + "score": 1.0, + "content": "are conditionally independent given", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36.5, + "bbox_fs": [ + 105, + 709, + 506, + 735 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 274, + 94 + ], + "lines": [ + { + "bbox": [ + 107, + 81, + 276, + 95 + ], + "spans": [ + { + "bbox": [ + 107, + 82, + 132, + 95 + ], + "score": 0.91, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 132, + 81, + 276, + 95 + ], + "score": 1.0, + "content": ". Thus, we can factorize as follows:", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "interline_equation", + "bbox": [ + 111, + 106, + 514, + 255 + ], + "lines": [ + { + "bbox": [ + 111, + 106, + 514, + 255 + ], + "spans": [ + { + "bbox": [ + 111, + 106, + 514, + 255 + ], + "score": 0.91, + "content": "\\begin{array} { r l } { \\displaystyle q ( \\mathbf { x } _ { k } ( \\tau ) | \\{ \\boldsymbol { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } ) \\propto q ( \\mathbf { x } _ { k } ( \\tau ) ) \\displaystyle \\prod _ { i = 1 } ^ { n } \\frac { q ( \\alpha _ { k } ( \\tau ) | \\boldsymbol { y } ^ { i } ( \\tau ) ) } { q ( \\alpha _ { k } ( \\tau ) ) } } & { ( \\mathrm { B a y e s ~ R u l e } ) } \\\\ { \\displaystyle \\Rightarrow \\log q ( \\alpha _ { k } ( \\tau ) | \\{ \\boldsymbol { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } ) \\propto \\log q ( \\alpha _ { k } ( \\tau ) ) + \\displaystyle \\sum _ { i = 1 } ^ { n } ( \\log q ( \\alpha _ { k } ( \\tau ) | \\boldsymbol { y } ^ { i } ( \\tau ) ) - \\log q ( \\alpha _ { k } ( \\tau ) ) ) } \\\\ { \\displaystyle \\Rightarrow \\nabla _ { \\mathbf { x } _ { k } ( \\tau ) } \\log q ( x _ { k } ( \\tau ) | \\{ \\boldsymbol { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } ) = \\nabla _ { \\mathbf { x } _ { k } ( \\tau ) } \\log q ( \\alpha _ { k } ( \\tau ) ) } \\\\ & { \\displaystyle + \\sum _ { i = 1 } ^ { n } ( \\nabla _ { \\mathbf { x } _ { k } ( \\tau ) } \\log q ( \\alpha _ { k } ( \\tau ) ) y ^ { i } ( \\tau ) ) - \\nabla _ { \\mathbf { x } _ { k } ( \\tau ) } \\log q ( \\alpha _ { k } ( \\tau ) ) ) } \\\\ { \\displaystyle \\Rightarrow \\epsilon _ { \\theta } ( \\alpha _ { k } ( \\tau ) , \\{ \\boldsymbol { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } , k ) = \\epsilon _ { \\theta } ( \\alpha _ { k } ( \\tau ) , \\boldsymbol { \\mathcal { O } } , k ) + \\displaystyle \\sum _ { i = 1 } ^ { n } ( \\epsilon _ { \\theta } ( \\alpha _ { k } ( \\tau ) , \\boldsymbol { y } ^ { i } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( x _ { k } ( \\tau ) , \\boldsymbol { \\mathcal { O } } , k ) ) } \\end{array}", + "type": "interline_equation", + "image_path": "9538bdc8a2e5be9ba6de812f3c83201e00416af038f79d46a42db9796578e664.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 111, + 106, + 514, + 155.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 111, + 155.66666666666666, + 514, + 205.33333333333331 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 111, + 205.33333333333331, + 514, + 254.99999999999997 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 265, + 503, + 289 + ], + "lines": [ + { + "bbox": [ + 106, + 264, + 505, + 279 + ], + "spans": [ + { + "bbox": [ + 106, + 264, + 301, + 279 + ], + "score": 1.0, + "content": "Using the above equations, we can sample from", + "type": "text" + }, + { + "bbox": [ + 302, + 265, + 388, + 279 + ], + "score": 0.92, + "content": "q ( \\pmb { x } _ { 0 } ( \\tau ) | \\{ \\pmb { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } )", + "type": "inline_equation" + }, + { + "bbox": [ + 388, + 264, + 505, + 279 + ], + "score": 1.0, + "content": "with classifier free guidance", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 277, + 212, + 290 + ], + "spans": [ + { + "bbox": [ + 106, + 277, + 212, + 290 + ], + "score": 1.0, + "content": "using the perturbed noise:", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4.5 + }, + { + "type": "interline_equation", + "bbox": [ + 160, + 297, + 450, + 348 + ], + "lines": [ + { + "bbox": [ + 160, + 297, + 450, + 348 + ], + "spans": [ + { + "bbox": [ + 160, + 297, + 450, + 348 + ], + "score": 0.93, + "content": "\\begin{array} { r l } & { \\hat { \\boldsymbol { \\epsilon } } : = \\boldsymbol { \\epsilon } _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\mathcal { O } , \\boldsymbol { k } ) + \\omega ( \\epsilon _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\{ \\boldsymbol { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } , \\boldsymbol { k } ) - \\epsilon _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\mathcal { O } , \\boldsymbol { k } ) ) } \\\\ & { \\quad = \\epsilon _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\mathcal { O } , \\boldsymbol { k } ) + \\omega \\displaystyle \\sum _ { i = 1 } ^ { n } ( \\epsilon _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\boldsymbol { y } ^ { i } ( \\tau ) , \\boldsymbol { k } ) - \\epsilon _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\mathcal { O } , \\boldsymbol { k } ) ) } \\end{array}", + "type": "interline_equation", + "image_path": "a809c3861b59cffd688cf17ec377461efa3936753cc310b6ef48ca1eb981f9d5.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 160, + 297, + 450, + 314.0 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 160, + 314.0, + 450, + 331.0 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 160, + 331.0, + 450, + 348.0 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 363, + 503, + 386 + ], + "lines": [ + { + "bbox": [ + 105, + 363, + 505, + 376 + ], + "spans": [ + { + "bbox": [ + 105, + 363, + 505, + 376 + ], + "score": 1.0, + "content": "We use the perturbed noise to compose skills or combine constraints at test time. This derivation was", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 374, + 395, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 374, + 395, + 387 + ], + "score": 1.0, + "content": "borrowed from Liu et al. (2022) and is presented here for completeness.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "text", + "bbox": [ + 106, + 391, + 505, + 458 + ], + "lines": [ + { + "bbox": [ + 104, + 390, + 506, + 406 + ], + "spans": [ + { + "bbox": [ + 104, + 390, + 313, + 406 + ], + "score": 1.0, + "content": "While the composition of conditioning variables", + "type": "text" + }, + { + "bbox": [ + 313, + 391, + 361, + 404 + ], + "score": 0.93, + "content": "\\{ \\pmb { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 361, + 390, + 506, + 406 + ], + "score": 1.0, + "content": "requires them to be conditionally", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 402, + 506, + 415 + ], + "spans": [ + { + "bbox": [ + 106, + 402, + 256, + 415 + ], + "score": 1.0, + "content": "independent given the state trajectory", + "type": "text" + }, + { + "bbox": [ + 256, + 403, + 281, + 415 + ], + "score": 0.92, + "content": "\\pmb { x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 282, + 402, + 506, + 415 + ], + "score": 1.0, + "content": ", we empirically observe that this condition doesn’t have", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 104, + 412, + 507, + 426 + ], + "spans": [ + { + "bbox": [ + 104, + 412, + 507, + 426 + ], + "score": 1.0, + "content": "to be strictly satisfied. However, we require composition of conditioning variables to be feasible (i.e.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 423, + 505, + 437 + ], + "spans": [ + { + "bbox": [ + 106, + 424, + 140, + 437 + ], + "score": 0.82, + "content": "\\exists x _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 423, + 505, + 437 + ], + "score": 1.0, + "content": "that satisfies all the conditioning variables). When the composition is infeasible, Decision", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "score": 1.0, + "content": "Diffuser produces trajectories with incoherent behavior, as expected. This is best illustrated by videos", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 446, + 340, + 458 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 340, + 458 + ], + "score": 1.0, + "content": "viewable at https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 13.5 + }, + { + "type": "text", + "bbox": [ + 106, + 460, + 505, + 537 + ], + "lines": [ + { + "bbox": [ + 106, + 460, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 106, + 460, + 505, + 472 + ], + "score": 1.0, + "content": "Requirements on the dataset First, the dataset should have a diverse set of demonstrations", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 471, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 393, + 484 + ], + "score": 1.0, + "content": "that shows different ways of satisfying each conditioning variable", + "type": "text" + }, + { + "bbox": [ + 393, + 471, + 417, + 483 + ], + "score": 0.93, + "content": "y ^ { i } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 418, + 471, + 505, + 484 + ], + "score": 1.0, + "content": ". This would allow", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 481, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 481, + 447, + 495 + ], + "score": 1.0, + "content": "Decision Diffuser to learn diverse ways of satisfying each conditioning variable", + "type": "text" + }, + { + "bbox": [ + 447, + 482, + 471, + 494 + ], + "score": 0.92, + "content": "y ^ { i } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 471, + 481, + 505, + 495 + ], + "score": 1.0, + "content": ". Since", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 493, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 106, + 493, + 428, + 506 + ], + "score": 1.0, + "content": "we use inverse dynamics to extract actions from the predicted state trajectory", + "type": "text" + }, + { + "bbox": [ + 428, + 493, + 453, + 505 + ], + "score": 0.92, + "content": "\\pmb { x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 453, + 493, + 505, + 506 + ], + "score": 1.0, + "content": ", we assume", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 504, + 505, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 504, + 204, + 516 + ], + "score": 1.0, + "content": "that the state trajectory", + "type": "text" + }, + { + "bbox": [ + 204, + 504, + 230, + 516 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 230, + 504, + 505, + 516 + ], + "score": 1.0, + "content": "resulting from the composition of different conditioning variables", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 515, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 237, + 528 + ], + "score": 1.0, + "content": "contains consecutive state pairs", + "type": "text" + }, + { + "bbox": [ + 237, + 515, + 276, + 527 + ], + "score": 0.92, + "content": "\\left( { { s _ { t } } , { s _ { t + 1 } } } \\right)", + "type": "inline_equation" + }, + { + "bbox": [ + 276, + 515, + 505, + 528 + ], + "score": 1.0, + "content": "that come from the same distribution that generated the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 525, + 443, + 538 + ], + "spans": [ + { + "bbox": [ + 105, + 525, + 443, + 538 + ], + "score": 1.0, + "content": "demonstration dataset. Otherwise, inverse dynamics can give erroneous predictions.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 20 + }, + { + "type": "title", + "bbox": [ + 107, + 559, + 393, + 573 + ], + "lines": [ + { + "bbox": [ + 105, + 559, + 394, + 574 + ], + "spans": [ + { + "bbox": [ + 105, + 559, + 394, + 574 + ], + "score": 1.0, + "content": "E RUNTIME CHARACTERISTIC OF DECISION DIFFUSER", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 106, + 588, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 588, + 505, + 602 + ], + "spans": [ + { + "bbox": [ + 105, + 588, + 505, + 602 + ], + "score": 1.0, + "content": "We analyze the runtime characteristics of Decision Diffuser in this section. After training the Decision", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 598, + 506, + 614 + ], + "spans": [ + { + "bbox": [ + 105, + 598, + 506, + 614 + ], + "score": 1.0, + "content": "Diffuser on trajectories from the D4RL Hopper-Medium-Expert dataset, we plan in the corresponding", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 611, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 611, + 506, + 624 + ], + "score": 1.0, + "content": "environment according to Algorithm 1. Every action taken in the environment requires running 100", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 622, + 506, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 506, + 634 + ], + "score": 1.0, + "content": "reverse diffusion steps to generate a state sequence taking on average 1.26s in wall-clock time. We", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 633, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 633, + 506, + 646 + ], + "score": 1.0, + "content": "can improve the run-time of planning by warm-starting the state diffusion as suggested in Janner et al.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "score": 1.0, + "content": "(2022). Here, we start with a generated state sequence (from the previous environment step), run", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "score": 1.0, + "content": "forward diffusion for a fixed number of steps, and finally run the same number of reverse diffusion", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 666, + 506, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 506, + 679 + ], + "score": 1.0, + "content": "steps from the partially noised state sequence to generate another state sequence. Warm-starting in", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "this way allows us to decrease the number of denoising steps to 40 (0.48s on average) without any", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "loss in performance, to 20 (0.21s on average) with minimal loss in performance, and to 5 with less", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 125, + 712 + ], + "score": 1.0, + "content": "than", + "type": "text" + }, + { + "bbox": [ + 125, + 699, + 145, + 709 + ], + "score": 0.88, + "content": "2 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "loss in performance (0.06s on average). We demonstrate the trade-off between performance,", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "measured by normalized average return achieved in the environment, and planning time, measured in", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 720, + 424, + 734 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 424, + 734 + ], + "score": 1.0, + "content": "wall-clock time after warm-starting the reverse diffusion process, in Figure A2.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 31 + } + ], + "page_idx": 15, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "16", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 274, + 94 + ], + "lines": [ + { + "bbox": [ + 107, + 81, + 276, + 95 + ], + "spans": [ + { + "bbox": [ + 107, + 82, + 132, + 95 + ], + "score": 0.91, + "content": "{ \\pmb x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 132, + 81, + 276, + 95 + ], + "score": 1.0, + "content": ". Thus, we can factorize as follows:", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0, + "bbox_fs": [ + 107, + 81, + 276, + 95 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 111, + 106, + 514, + 255 + ], + "lines": [ + { + "bbox": [ + 111, + 106, + 514, + 255 + ], + "spans": [ + { + "bbox": [ + 111, + 106, + 514, + 255 + ], + "score": 0.91, + "content": "\\begin{array} { r l } { \\displaystyle q ( \\mathbf { x } _ { k } ( \\tau ) | \\{ \\boldsymbol { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } ) \\propto q ( \\mathbf { x } _ { k } ( \\tau ) ) \\displaystyle \\prod _ { i = 1 } ^ { n } \\frac { q ( \\alpha _ { k } ( \\tau ) | \\boldsymbol { y } ^ { i } ( \\tau ) ) } { q ( \\alpha _ { k } ( \\tau ) ) } } & { ( \\mathrm { B a y e s ~ R u l e } ) } \\\\ { \\displaystyle \\Rightarrow \\log q ( \\alpha _ { k } ( \\tau ) | \\{ \\boldsymbol { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } ) \\propto \\log q ( \\alpha _ { k } ( \\tau ) ) + \\displaystyle \\sum _ { i = 1 } ^ { n } ( \\log q ( \\alpha _ { k } ( \\tau ) | \\boldsymbol { y } ^ { i } ( \\tau ) ) - \\log q ( \\alpha _ { k } ( \\tau ) ) ) } \\\\ { \\displaystyle \\Rightarrow \\nabla _ { \\mathbf { x } _ { k } ( \\tau ) } \\log q ( x _ { k } ( \\tau ) | \\{ \\boldsymbol { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } ) = \\nabla _ { \\mathbf { x } _ { k } ( \\tau ) } \\log q ( \\alpha _ { k } ( \\tau ) ) } \\\\ & { \\displaystyle + \\sum _ { i = 1 } ^ { n } ( \\nabla _ { \\mathbf { x } _ { k } ( \\tau ) } \\log q ( \\alpha _ { k } ( \\tau ) ) y ^ { i } ( \\tau ) ) - \\nabla _ { \\mathbf { x } _ { k } ( \\tau ) } \\log q ( \\alpha _ { k } ( \\tau ) ) ) } \\\\ { \\displaystyle \\Rightarrow \\epsilon _ { \\theta } ( \\alpha _ { k } ( \\tau ) , \\{ \\boldsymbol { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } , k ) = \\epsilon _ { \\theta } ( \\alpha _ { k } ( \\tau ) , \\boldsymbol { \\mathcal { O } } , k ) + \\displaystyle \\sum _ { i = 1 } ^ { n } ( \\epsilon _ { \\theta } ( \\alpha _ { k } ( \\tau ) , \\boldsymbol { y } ^ { i } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( x _ { k } ( \\tau ) , \\boldsymbol { \\mathcal { O } } , k ) ) } \\end{array}", + "type": "interline_equation", + "image_path": "9538bdc8a2e5be9ba6de812f3c83201e00416af038f79d46a42db9796578e664.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 111, + 106, + 514, + 155.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 111, + 155.66666666666666, + 514, + 205.33333333333331 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 111, + 205.33333333333331, + 514, + 254.99999999999997 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 265, + 503, + 289 + ], + "lines": [ + { + "bbox": [ + 106, + 264, + 505, + 279 + ], + "spans": [ + { + "bbox": [ + 106, + 264, + 301, + 279 + ], + "score": 1.0, + "content": "Using the above equations, we can sample from", + "type": "text" + }, + { + "bbox": [ + 302, + 265, + 388, + 279 + ], + "score": 0.92, + "content": "q ( \\pmb { x } _ { 0 } ( \\tau ) | \\{ \\pmb { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } )", + "type": "inline_equation" + }, + { + "bbox": [ + 388, + 264, + 505, + 279 + ], + "score": 1.0, + "content": "with classifier free guidance", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 277, + 212, + 290 + ], + "spans": [ + { + "bbox": [ + 106, + 277, + 212, + 290 + ], + "score": 1.0, + "content": "using the perturbed noise:", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4.5, + "bbox_fs": [ + 106, + 264, + 505, + 290 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 160, + 297, + 450, + 348 + ], + "lines": [ + { + "bbox": [ + 160, + 297, + 450, + 348 + ], + "spans": [ + { + "bbox": [ + 160, + 297, + 450, + 348 + ], + "score": 0.93, + "content": "\\begin{array} { r l } & { \\hat { \\boldsymbol { \\epsilon } } : = \\boldsymbol { \\epsilon } _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\mathcal { O } , \\boldsymbol { k } ) + \\omega ( \\epsilon _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\{ \\boldsymbol { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n } , \\boldsymbol { k } ) - \\epsilon _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\mathcal { O } , \\boldsymbol { k } ) ) } \\\\ & { \\quad = \\epsilon _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\mathcal { O } , \\boldsymbol { k } ) + \\omega \\displaystyle \\sum _ { i = 1 } ^ { n } ( \\epsilon _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\boldsymbol { y } ^ { i } ( \\tau ) , \\boldsymbol { k } ) - \\epsilon _ { \\boldsymbol { \\theta } } ( \\mathbf { x } _ { k } ( \\tau ) , \\mathcal { O } , \\boldsymbol { k } ) ) } \\end{array}", + "type": "interline_equation", + "image_path": "a809c3861b59cffd688cf17ec377461efa3936753cc310b6ef48ca1eb981f9d5.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 160, + 297, + 450, + 314.0 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 160, + 314.0, + 450, + 331.0 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 160, + 331.0, + 450, + 348.0 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 363, + 503, + 386 + ], + "lines": [ + { + "bbox": [ + 105, + 363, + 505, + 376 + ], + "spans": [ + { + "bbox": [ + 105, + 363, + 505, + 376 + ], + "score": 1.0, + "content": "We use the perturbed noise to compose skills or combine constraints at test time. This derivation was", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 374, + 395, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 374, + 395, + 387 + ], + "score": 1.0, + "content": "borrowed from Liu et al. (2022) and is presented here for completeness.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5, + "bbox_fs": [ + 105, + 363, + 505, + 387 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 391, + 505, + 458 + ], + "lines": [ + { + "bbox": [ + 104, + 390, + 506, + 406 + ], + "spans": [ + { + "bbox": [ + 104, + 390, + 313, + 406 + ], + "score": 1.0, + "content": "While the composition of conditioning variables", + "type": "text" + }, + { + "bbox": [ + 313, + 391, + 361, + 404 + ], + "score": 0.93, + "content": "\\{ \\pmb { y } ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 361, + 390, + 506, + 406 + ], + "score": 1.0, + "content": "requires them to be conditionally", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 402, + 506, + 415 + ], + "spans": [ + { + "bbox": [ + 106, + 402, + 256, + 415 + ], + "score": 1.0, + "content": "independent given the state trajectory", + "type": "text" + }, + { + "bbox": [ + 256, + 403, + 281, + 415 + ], + "score": 0.92, + "content": "\\pmb { x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 282, + 402, + 506, + 415 + ], + "score": 1.0, + "content": ", we empirically observe that this condition doesn’t have", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 104, + 412, + 507, + 426 + ], + "spans": [ + { + "bbox": [ + 104, + 412, + 507, + 426 + ], + "score": 1.0, + "content": "to be strictly satisfied. However, we require composition of conditioning variables to be feasible (i.e.", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 423, + 505, + 437 + ], + "spans": [ + { + "bbox": [ + 106, + 424, + 140, + 437 + ], + "score": 0.82, + "content": "\\exists x _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 141, + 423, + 505, + 437 + ], + "score": 1.0, + "content": "that satisfies all the conditioning variables). When the composition is infeasible, Decision", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "score": 1.0, + "content": "Diffuser produces trajectories with incoherent behavior, as expected. This is best illustrated by videos", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 446, + 340, + 458 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 340, + 458 + ], + "score": 1.0, + "content": "viewable at https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 13.5, + "bbox_fs": [ + 104, + 390, + 507, + 458 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 460, + 505, + 537 + ], + "lines": [ + { + "bbox": [ + 106, + 460, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 106, + 460, + 505, + 472 + ], + "score": 1.0, + "content": "Requirements on the dataset First, the dataset should have a diverse set of demonstrations", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 471, + 505, + 484 + ], + "spans": [ + { + "bbox": [ + 105, + 471, + 393, + 484 + ], + "score": 1.0, + "content": "that shows different ways of satisfying each conditioning variable", + "type": "text" + }, + { + "bbox": [ + 393, + 471, + 417, + 483 + ], + "score": 0.93, + "content": "y ^ { i } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 418, + 471, + 505, + 484 + ], + "score": 1.0, + "content": ". This would allow", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 481, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 105, + 481, + 447, + 495 + ], + "score": 1.0, + "content": "Decision Diffuser to learn diverse ways of satisfying each conditioning variable", + "type": "text" + }, + { + "bbox": [ + 447, + 482, + 471, + 494 + ], + "score": 0.92, + "content": "y ^ { i } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 471, + 481, + 505, + 495 + ], + "score": 1.0, + "content": ". Since", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 493, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 106, + 493, + 428, + 506 + ], + "score": 1.0, + "content": "we use inverse dynamics to extract actions from the predicted state trajectory", + "type": "text" + }, + { + "bbox": [ + 428, + 493, + 453, + 505 + ], + "score": 0.92, + "content": "\\pmb { x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 453, + 493, + 505, + 506 + ], + "score": 1.0, + "content": ", we assume", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 504, + 505, + 516 + ], + "spans": [ + { + "bbox": [ + 106, + 504, + 204, + 516 + ], + "score": 1.0, + "content": "that the state trajectory", + "type": "text" + }, + { + "bbox": [ + 204, + 504, + 230, + 516 + ], + "score": 0.92, + "content": "{ \\pmb x } _ { 0 } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 230, + 504, + 505, + 516 + ], + "score": 1.0, + "content": "resulting from the composition of different conditioning variables", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 515, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 515, + 237, + 528 + ], + "score": 1.0, + "content": "contains consecutive state pairs", + "type": "text" + }, + { + "bbox": [ + 237, + 515, + 276, + 527 + ], + "score": 0.92, + "content": "\\left( { { s _ { t } } , { s _ { t + 1 } } } \\right)", + "type": "inline_equation" + }, + { + "bbox": [ + 276, + 515, + 505, + 528 + ], + "score": 1.0, + "content": "that come from the same distribution that generated the", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 525, + 443, + 538 + ], + "spans": [ + { + "bbox": [ + 105, + 525, + 443, + 538 + ], + "score": 1.0, + "content": "demonstration dataset. Otherwise, inverse dynamics can give erroneous predictions.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 20, + "bbox_fs": [ + 105, + 460, + 505, + 538 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 559, + 393, + 573 + ], + "lines": [ + { + "bbox": [ + 105, + 559, + 394, + 574 + ], + "spans": [ + { + "bbox": [ + 105, + 559, + 394, + 574 + ], + "score": 1.0, + "content": "E RUNTIME CHARACTERISTIC OF DECISION DIFFUSER", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 106, + 588, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 588, + 505, + 602 + ], + "spans": [ + { + "bbox": [ + 105, + 588, + 505, + 602 + ], + "score": 1.0, + "content": "We analyze the runtime characteristics of Decision Diffuser in this section. After training the Decision", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 598, + 506, + 614 + ], + "spans": [ + { + "bbox": [ + 105, + 598, + 506, + 614 + ], + "score": 1.0, + "content": "Diffuser on trajectories from the D4RL Hopper-Medium-Expert dataset, we plan in the corresponding", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 611, + 506, + 624 + ], + "spans": [ + { + "bbox": [ + 105, + 611, + 506, + 624 + ], + "score": 1.0, + "content": "environment according to Algorithm 1. Every action taken in the environment requires running 100", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 622, + 506, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 506, + 634 + ], + "score": 1.0, + "content": "reverse diffusion steps to generate a state sequence taking on average 1.26s in wall-clock time. We", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 633, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 633, + 506, + 646 + ], + "score": 1.0, + "content": "can improve the run-time of planning by warm-starting the state diffusion as suggested in Janner et al.", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 505, + 657 + ], + "score": 1.0, + "content": "(2022). Here, we start with a generated state sequence (from the previous environment step), run", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "spans": [ + { + "bbox": [ + 106, + 654, + 505, + 666 + ], + "score": 1.0, + "content": "forward diffusion for a fixed number of steps, and finally run the same number of reverse diffusion", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 666, + 506, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 506, + 679 + ], + "score": 1.0, + "content": "steps from the partially noised state sequence to generate another state sequence. Warm-starting in", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "this way allows us to decrease the number of denoising steps to 40 (0.48s on average) without any", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "loss in performance, to 20 (0.21s on average) with minimal loss in performance, and to 5 with less", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 125, + 712 + ], + "score": 1.0, + "content": "than", + "type": "text" + }, + { + "bbox": [ + 125, + 699, + 145, + 709 + ], + "score": 0.88, + "content": "2 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 145, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "loss in performance (0.06s on average). We demonstrate the trade-off between performance,", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "measured by normalized average return achieved in the environment, and planning time, measured in", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 720, + 424, + 734 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 424, + 734 + ], + "score": 1.0, + "content": "wall-clock time after warm-starting the reverse diffusion process, in Figure A2.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 31, + "bbox_fs": [ + 105, + 588, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 205, + 79, + 406, + 250 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 205, + 79, + 406, + 250 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 205, + 79, + 406, + 250 + ], + "spans": [ + { + "bbox": [ + 205, + 79, + 406, + 250 + ], + "score": 0.925, + "type": "image", + "image_path": "582a316558824d8218903d31c84d2ea265682dcdd6126b4d53b0d942b675b615.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 205, + 79, + 406, + 92.15384615384616 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 205, + 92.15384615384616, + 406, + 105.30769230769232 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 205, + 105.30769230769232, + 406, + 118.46153846153848 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 205, + 118.46153846153848, + 406, + 131.61538461538464 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 205, + 131.61538461538464, + 406, + 144.7692307692308 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 205, + 144.7692307692308, + 406, + 157.92307692307696 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 205, + 157.92307692307696, + 406, + 171.07692307692312 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 205, + 171.07692307692312, + 406, + 184.23076923076928 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 205, + 184.23076923076928, + 406, + 197.38461538461544 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 205, + 197.38461538461544, + 406, + 210.5384615384616 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 205, + 210.5384615384616, + 406, + 223.69230769230776 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 205, + 223.69230769230776, + 406, + 236.84615384615392 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 205, + 236.84615384615392, + 406, + 250.00000000000009 + ], + "spans": [], + "index": 12 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 259, + 504, + 290 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 259, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 106, + 259, + 505, + 271 + ], + "score": 1.0, + "content": "Figure A2: Performance vs planning time. We visualize the trade-off between performance, measured by", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 270, + 506, + 281 + ], + "spans": [ + { + "bbox": [ + 106, + 270, + 506, + 281 + ], + "score": 1.0, + "content": "normalized average return achieved in the environment, and planning time, measured in wall-clock time after", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 280, + 266, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 266, + 291 + ], + "score": 1.0, + "content": "warm-starting the reverse diffusion process.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14 + } + ], + "index": 10.0 + }, + { + "type": "table", + "bbox": [ + 228, + 295, + 490, + 338 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 228, + 295, + 490, + 338 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 228, + 295, + 490, + 338 + ], + "spans": [ + { + "bbox": [ + 228, + 295, + 490, + 338 + ], + "score": 0.966, + "html": "
EnvironmentBCCondDiffuserDecision Diffuser
Position Control57.3 ±1.287.3 ±3.187.8 ±2.8
Torque Control55.2 ±1.571.8 ±3.484.7 ±2.2
", + "type": "table", + "image_path": "1299108cf864d28cd5fd5354a22cc2b848692c2a21bde63eba1dbe87607eda12.jpg" + } + ] + } + ], + "index": 17, + "virtual_lines": [ + { + "bbox": [ + 228, + 295, + 490, + 309.3333333333333 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 228, + 309.3333333333333, + 490, + 323.66666666666663 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 228, + 323.66666666666663, + 490, + 337.99999999999994 + ], + "spans": [], + "index": 18 + } + ] + } + ], + "index": 17 + }, + { + "type": "image", + "bbox": [ + 107, + 306, + 207, + 388 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 306, + 207, + 388 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 107, + 306, + 207, + 388 + ], + "spans": [ + { + "bbox": [ + 107, + 306, + 207, + 388 + ], + "score": 0.963, + "type": "image", + "image_path": "b7e5f34bba8b05d1a3659b787a4b3d2616878cdfbc9e2f790df9d0e668b1b9ef.jpg" + } + ] + } + ], + "index": 20.5, + "virtual_lines": [ + { + "bbox": [ + 107, + 306, + 207, + 347.0 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 107, + 347.0, + 207, + 388.0 + ], + "spans": [], + "index": 19 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 396, + 207, + 417 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 396, + 208, + 407 + ], + "spans": [ + { + "bbox": [ + 106, + 396, + 208, + 407 + ], + "score": 1.0, + "content": "Figure A3: Block push", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 407, + 159, + 417 + ], + "spans": [ + { + "bbox": [ + 106, + 407, + 159, + 417 + ], + "score": 1.0, + "content": "environment.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20.5 + } + ], + "index": 20.5 + }, + { + "type": "text", + "bbox": [ + 228, + 347, + 488, + 417 + ], + "lines": [ + { + "bbox": [ + 227, + 347, + 488, + 358 + ], + "spans": [ + { + "bbox": [ + 227, + 347, + 488, + 358 + ], + "score": 1.0, + "content": "Table A2: Block pushing with different controls. Decision Diffuser", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 227, + 357, + 489, + 369 + ], + "spans": [ + { + "bbox": [ + 227, + 357, + 489, + 369 + ], + "score": 1.0, + "content": "and CondDiffuser perform similarly when the agent uses position control.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 227, + 367, + 488, + 379 + ], + "spans": [ + { + "bbox": [ + 227, + 367, + 488, + 379 + ], + "score": 1.0, + "content": "However, when the agent uses torque control, CondDiffuser performs", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 227, + 377, + 488, + 388 + ], + "spans": [ + { + "bbox": [ + 227, + 377, + 488, + 388 + ], + "score": 1.0, + "content": "worse than Decision Diffuser given it’s harder to diffuse over non-smooth", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 227, + 387, + 488, + 398 + ], + "spans": [ + { + "bbox": [ + 227, + 387, + 488, + 398 + ], + "score": 1.0, + "content": "action trajectories. We use the success rate of the red cube reaching the", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 227, + 397, + 488, + 408 + ], + "spans": [ + { + "bbox": [ + 227, + 397, + 488, + 408 + ], + "score": 1.0, + "content": "green circle as the performance metric. We report the mean success rate", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 227, + 407, + 387, + 417 + ], + "spans": [ + { + "bbox": [ + 227, + 407, + 387, + 417 + ], + "score": 1.0, + "content": "and the standard error over 5 random seeds.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 26 + }, + { + "type": "title", + "bbox": [ + 107, + 430, + 312, + 443 + ], + "lines": [ + { + "bbox": [ + 105, + 430, + 313, + 445 + ], + "spans": [ + { + "bbox": [ + 105, + 430, + 313, + 445 + ], + "score": 1.0, + "content": "F WHEN TO USE INVERSE DYNAMICS?", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 107, + 461, + 505, + 528 + ], + "lines": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "score": 1.0, + "content": "In this section, we try to analyze further when using inverse dynamics is better than diffusing over", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 472, + 506, + 485 + ], + "spans": [ + { + "bbox": [ + 105, + 472, + 506, + 485 + ], + "score": 1.0, + "content": "actions. Table 2 showed that Decision Diffuser outperformed CondDiffuser on 3 hopper environment,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 484, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 106, + 484, + 505, + 496 + ], + "score": 1.0, + "content": "thereby suggesting that inverse dynamics is a better alternative to diffusing over actions. Our intuition", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 495, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 505, + 506 + ], + "score": 1.0, + "content": "was that sequences over actions, represented as joint torques in our environments, tend to be more", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 505, + 506, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 506, + 518 + ], + "score": 1.0, + "content": "high-frequency and less smooth, thus making it harder for the diffusion model to predict (Kingma", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 516, + 347, + 529 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 347, + 529 + ], + "score": 1.0, + "content": "et al., 2021). We now try to verify this intuition empirically.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 33.5 + }, + { + "type": "text", + "bbox": [ + 107, + 533, + 505, + 610 + ], + "lines": [ + { + "bbox": [ + 105, + 532, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 105, + 532, + 505, + 546 + ], + "score": 1.0, + "content": "Setup We choose Block Push environment adapted from Gupta et al. (2018) where the goal is to", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 544, + 507, + 558 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 507, + 558 + ], + "score": 1.0, + "content": "push the red cube to the green circle. When the red cube reaches the green circle, the agent gets a", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 555, + 506, + 568 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 147, + 568 + ], + "score": 1.0, + "content": "reward of", + "type": "text" + }, + { + "bbox": [ + 147, + 556, + 160, + 565 + ], + "score": 0.84, + "content": "+ 1", + "type": "inline_equation" + }, + { + "bbox": [ + 160, + 555, + 506, + 568 + ], + "score": 1.0, + "content": ". The state space is 10-dimensional consisting of joint angles (3) and velocities (3) of", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 565, + 506, + 579 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 506, + 579 + ], + "score": 1.0, + "content": "the gripper, COM of the gripper (2) and position of the red cube (2). The green circle’s position is", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 577, + 505, + 590 + ], + "spans": [ + { + "bbox": [ + 105, + 577, + 505, + 590 + ], + "score": 1.0, + "content": "fixed and at an initial distance of 0.5 from COM of the gripper. The red cube (of size 0.03) is initially", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 588, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 588, + 353, + 600 + ], + "score": 1.0, + "content": "at a distance of 0.1 from COM of the gripper and at an angle", + "type": "text" + }, + { + "bbox": [ + 353, + 588, + 360, + 598 + ], + "score": 0.78, + "content": "\\theta", + "type": "inline_equation" + }, + { + "bbox": [ + 360, + 588, + 419, + 600 + ], + "score": 1.0, + "content": "sampled from", + "type": "text" + }, + { + "bbox": [ + 419, + 588, + 479, + 600 + ], + "score": 0.92, + "content": "\\mathcal { U } ( - \\pi / 4 , \\pi / 4 )", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 588, + 505, + 600 + ], + "score": 1.0, + "content": "at the", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 599, + 330, + 612 + ], + "spans": [ + { + "bbox": [ + 105, + 599, + 330, + 612 + ], + "score": 1.0, + "content": "start of every episode. The task horizon is 60 timesteps.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 40 + }, + { + "type": "text", + "bbox": [ + 107, + 616, + 505, + 671 + ], + "lines": [ + { + "bbox": [ + 105, + 615, + 505, + 629 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 505, + 629 + ], + "score": 1.0, + "content": "There are 2 control types: (i) torque control, where the agent needs to specify joint torques (3", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 626, + 505, + 640 + ], + "spans": [ + { + "bbox": [ + 106, + 626, + 505, + 640 + ], + "score": 1.0, + "content": "dimensional) and (ii) position control where the agent needs to specify the position change of COM", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 637, + 506, + 652 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 350, + 652 + ], + "score": 1.0, + "content": "of the gripper and the angular change in gripper’s orientation", + "type": "text" + }, + { + "bbox": [ + 350, + 638, + 409, + 650 + ], + "score": 0.92, + "content": "( \\Delta x , \\Delta y , \\Delta \\phi )", + "type": "inline_equation" + }, + { + "bbox": [ + 409, + 637, + 506, + 652 + ], + "score": 1.0, + "content": "(3 dimensional). While", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 648, + 505, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 505, + 662 + ], + "score": 1.0, + "content": "action trajectories from position control are smooth, the action trajectories from torque control have", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 660, + 230, + 673 + ], + "spans": [ + { + "bbox": [ + 105, + 660, + 230, + 673 + ], + "score": 1.0, + "content": "higher frequency components.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 46 + }, + { + "type": "text", + "bbox": [ + 107, + 677, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 676, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 506, + 689 + ], + "score": 1.0, + "content": "Offline dataset collection To collect the offline data, we use Soft Actor-Critic (SAC) (Haarnoja", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "et al., 2018) first to train an expert policy for 1 million environment steps. We then use 1 million", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 698, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 711 + ], + "score": 1.0, + "content": "environment transitions as our offline dataset, which contains expert trajectories collected towards", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "the end of the training and random action trajectories collected at the beginning of the training. We", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 105, + 719, + 284, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 284, + 734 + ], + "score": 1.0, + "content": "collect 2 datasets, one for each control type.", + "type": "text" + } + ], + "index": 53 + } + ], + "index": 51 + } + ], + "page_idx": 16, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 205, + 79, + 406, + 250 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 205, + 79, + 406, + 250 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 205, + 79, + 406, + 250 + ], + "spans": [ + { + "bbox": [ + 205, + 79, + 406, + 250 + ], + "score": 0.925, + "type": "image", + "image_path": "582a316558824d8218903d31c84d2ea265682dcdd6126b4d53b0d942b675b615.jpg" + } + ] + } + ], + "index": 6, + "virtual_lines": [ + { + "bbox": [ + 205, + 79, + 406, + 92.15384615384616 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 205, + 92.15384615384616, + 406, + 105.30769230769232 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 205, + 105.30769230769232, + 406, + 118.46153846153848 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 205, + 118.46153846153848, + 406, + 131.61538461538464 + ], + "spans": [], + "index": 3 + }, + { + "bbox": [ + 205, + 131.61538461538464, + 406, + 144.7692307692308 + ], + "spans": [], + "index": 4 + }, + { + "bbox": [ + 205, + 144.7692307692308, + 406, + 157.92307692307696 + ], + "spans": [], + "index": 5 + }, + { + "bbox": [ + 205, + 157.92307692307696, + 406, + 171.07692307692312 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 205, + 171.07692307692312, + 406, + 184.23076923076928 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 205, + 184.23076923076928, + 406, + 197.38461538461544 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 205, + 197.38461538461544, + 406, + 210.5384615384616 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 205, + 210.5384615384616, + 406, + 223.69230769230776 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 205, + 223.69230769230776, + 406, + 236.84615384615392 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 205, + 236.84615384615392, + 406, + 250.00000000000009 + ], + "spans": [], + "index": 12 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 259, + 504, + 290 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 259, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 106, + 259, + 505, + 271 + ], + "score": 1.0, + "content": "Figure A2: Performance vs planning time. We visualize the trade-off between performance, measured by", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 270, + 506, + 281 + ], + "spans": [ + { + "bbox": [ + 106, + 270, + 506, + 281 + ], + "score": 1.0, + "content": "normalized average return achieved in the environment, and planning time, measured in wall-clock time after", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 280, + 266, + 291 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 266, + 291 + ], + "score": 1.0, + "content": "warm-starting the reverse diffusion process.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14 + } + ], + "index": 10.0 + }, + { + "type": "table", + "bbox": [ + 228, + 295, + 490, + 338 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 228, + 295, + 490, + 338 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 228, + 295, + 490, + 338 + ], + "spans": [ + { + "bbox": [ + 228, + 295, + 490, + 338 + ], + "score": 0.966, + "html": "
EnvironmentBCCondDiffuserDecision Diffuser
Position Control57.3 ±1.287.3 ±3.187.8 ±2.8
Torque Control55.2 ±1.571.8 ±3.484.7 ±2.2
", + "type": "table", + "image_path": "1299108cf864d28cd5fd5354a22cc2b848692c2a21bde63eba1dbe87607eda12.jpg" + } + ] + } + ], + "index": 17, + "virtual_lines": [ + { + "bbox": [ + 228, + 295, + 490, + 309.3333333333333 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 228, + 309.3333333333333, + 490, + 323.66666666666663 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 228, + 323.66666666666663, + 490, + 337.99999999999994 + ], + "spans": [], + "index": 18 + } + ] + } + ], + "index": 17 + }, + { + "type": "image", + "bbox": [ + 107, + 306, + 207, + 388 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 306, + 207, + 388 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 107, + 306, + 207, + 388 + ], + "spans": [ + { + "bbox": [ + 107, + 306, + 207, + 388 + ], + "score": 0.963, + "type": "image", + "image_path": "b7e5f34bba8b05d1a3659b787a4b3d2616878cdfbc9e2f790df9d0e668b1b9ef.jpg" + } + ] + } + ], + "index": 20.5, + "virtual_lines": [ + { + "bbox": [ + 107, + 306, + 207, + 347.0 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 107, + 347.0, + 207, + 388.0 + ], + "spans": [], + "index": 19 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 396, + 207, + 417 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 396, + 208, + 407 + ], + "spans": [ + { + "bbox": [ + 106, + 396, + 208, + 407 + ], + "score": 1.0, + "content": "Figure A3: Block push", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 407, + 159, + 417 + ], + "spans": [ + { + "bbox": [ + 106, + 407, + 159, + 417 + ], + "score": 1.0, + "content": "environment.", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 20.5 + } + ], + "index": 20.5 + }, + { + "type": "text", + "bbox": [ + 228, + 347, + 488, + 417 + ], + "lines": [ + { + "bbox": [ + 227, + 347, + 488, + 358 + ], + "spans": [ + { + "bbox": [ + 227, + 347, + 488, + 358 + ], + "score": 1.0, + "content": "Table A2: Block pushing with different controls. Decision Diffuser", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 227, + 357, + 489, + 369 + ], + "spans": [ + { + "bbox": [ + 227, + 357, + 489, + 369 + ], + "score": 1.0, + "content": "and CondDiffuser perform similarly when the agent uses position control.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 227, + 367, + 488, + 379 + ], + "spans": [ + { + "bbox": [ + 227, + 367, + 488, + 379 + ], + "score": 1.0, + "content": "However, when the agent uses torque control, CondDiffuser performs", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 227, + 377, + 488, + 388 + ], + "spans": [ + { + "bbox": [ + 227, + 377, + 488, + 388 + ], + "score": 1.0, + "content": "worse than Decision Diffuser given it’s harder to diffuse over non-smooth", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 227, + 387, + 488, + 398 + ], + "spans": [ + { + "bbox": [ + 227, + 387, + 488, + 398 + ], + "score": 1.0, + "content": "action trajectories. We use the success rate of the red cube reaching the", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 227, + 397, + 488, + 408 + ], + "spans": [ + { + "bbox": [ + 227, + 397, + 488, + 408 + ], + "score": 1.0, + "content": "green circle as the performance metric. We report the mean success rate", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 227, + 407, + 387, + 417 + ], + "spans": [ + { + "bbox": [ + 227, + 407, + 387, + 417 + ], + "score": 1.0, + "content": "and the standard error over 5 random seeds.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 26, + "bbox_fs": [ + 227, + 347, + 489, + 417 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 430, + 312, + 443 + ], + "lines": [ + { + "bbox": [ + 105, + 430, + 313, + 445 + ], + "spans": [ + { + "bbox": [ + 105, + 430, + 313, + 445 + ], + "score": 1.0, + "content": "F WHEN TO USE INVERSE DYNAMICS?", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 107, + 461, + 505, + 528 + ], + "lines": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "score": 1.0, + "content": "In this section, we try to analyze further when using inverse dynamics is better than diffusing over", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 472, + 506, + 485 + ], + "spans": [ + { + "bbox": [ + 105, + 472, + 506, + 485 + ], + "score": 1.0, + "content": "actions. Table 2 showed that Decision Diffuser outperformed CondDiffuser on 3 hopper environment,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 484, + 505, + 496 + ], + "spans": [ + { + "bbox": [ + 106, + 484, + 505, + 496 + ], + "score": 1.0, + "content": "thereby suggesting that inverse dynamics is a better alternative to diffusing over actions. Our intuition", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 495, + 505, + 506 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 505, + 506 + ], + "score": 1.0, + "content": "was that sequences over actions, represented as joint torques in our environments, tend to be more", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 505, + 506, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 506, + 518 + ], + "score": 1.0, + "content": "high-frequency and less smooth, thus making it harder for the diffusion model to predict (Kingma", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 516, + 347, + 529 + ], + "spans": [ + { + "bbox": [ + 106, + 516, + 347, + 529 + ], + "score": 1.0, + "content": "et al., 2021). We now try to verify this intuition empirically.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 33.5, + "bbox_fs": [ + 105, + 461, + 506, + 529 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 533, + 505, + 610 + ], + "lines": [ + { + "bbox": [ + 105, + 532, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 105, + 532, + 505, + 546 + ], + "score": 1.0, + "content": "Setup We choose Block Push environment adapted from Gupta et al. (2018) where the goal is to", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 544, + 507, + 558 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 507, + 558 + ], + "score": 1.0, + "content": "push the red cube to the green circle. When the red cube reaches the green circle, the agent gets a", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 555, + 506, + 568 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 147, + 568 + ], + "score": 1.0, + "content": "reward of", + "type": "text" + }, + { + "bbox": [ + 147, + 556, + 160, + 565 + ], + "score": 0.84, + "content": "+ 1", + "type": "inline_equation" + }, + { + "bbox": [ + 160, + 555, + 506, + 568 + ], + "score": 1.0, + "content": ". The state space is 10-dimensional consisting of joint angles (3) and velocities (3) of", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 565, + 506, + 579 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 506, + 579 + ], + "score": 1.0, + "content": "the gripper, COM of the gripper (2) and position of the red cube (2). The green circle’s position is", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 577, + 505, + 590 + ], + "spans": [ + { + "bbox": [ + 105, + 577, + 505, + 590 + ], + "score": 1.0, + "content": "fixed and at an initial distance of 0.5 from COM of the gripper. The red cube (of size 0.03) is initially", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 588, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 588, + 353, + 600 + ], + "score": 1.0, + "content": "at a distance of 0.1 from COM of the gripper and at an angle", + "type": "text" + }, + { + "bbox": [ + 353, + 588, + 360, + 598 + ], + "score": 0.78, + "content": "\\theta", + "type": "inline_equation" + }, + { + "bbox": [ + 360, + 588, + 419, + 600 + ], + "score": 1.0, + "content": "sampled from", + "type": "text" + }, + { + "bbox": [ + 419, + 588, + 479, + 600 + ], + "score": 0.92, + "content": "\\mathcal { U } ( - \\pi / 4 , \\pi / 4 )", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 588, + 505, + 600 + ], + "score": 1.0, + "content": "at the", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 599, + 330, + 612 + ], + "spans": [ + { + "bbox": [ + 105, + 599, + 330, + 612 + ], + "score": 1.0, + "content": "start of every episode. The task horizon is 60 timesteps.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 40, + "bbox_fs": [ + 105, + 532, + 507, + 612 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 616, + 505, + 671 + ], + "lines": [ + { + "bbox": [ + 105, + 615, + 505, + 629 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 505, + 629 + ], + "score": 1.0, + "content": "There are 2 control types: (i) torque control, where the agent needs to specify joint torques (3", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 106, + 626, + 505, + 640 + ], + "spans": [ + { + "bbox": [ + 106, + 626, + 505, + 640 + ], + "score": 1.0, + "content": "dimensional) and (ii) position control where the agent needs to specify the position change of COM", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 637, + 506, + 652 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 350, + 652 + ], + "score": 1.0, + "content": "of the gripper and the angular change in gripper’s orientation", + "type": "text" + }, + { + "bbox": [ + 350, + 638, + 409, + 650 + ], + "score": 0.92, + "content": "( \\Delta x , \\Delta y , \\Delta \\phi )", + "type": "inline_equation" + }, + { + "bbox": [ + 409, + 637, + 506, + 652 + ], + "score": 1.0, + "content": "(3 dimensional). While", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 648, + 505, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 505, + 662 + ], + "score": 1.0, + "content": "action trajectories from position control are smooth, the action trajectories from torque control have", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 660, + 230, + 673 + ], + "spans": [ + { + "bbox": [ + 105, + 660, + 230, + 673 + ], + "score": 1.0, + "content": "higher frequency components.", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 46, + "bbox_fs": [ + 105, + 615, + 506, + 673 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 677, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 676, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 506, + 689 + ], + "score": 1.0, + "content": "Offline dataset collection To collect the offline data, we use Soft Actor-Critic (SAC) (Haarnoja", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "et al., 2018) first to train an expert policy for 1 million environment steps. We then use 1 million", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 698, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 711 + ], + "score": 1.0, + "content": "environment transitions as our offline dataset, which contains expert trajectories collected towards", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "the end of the training and random action trajectories collected at the beginning of the training. We", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 105, + 719, + 284, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 284, + 734 + ], + "score": 1.0, + "content": "collect 2 datasets, one for each control type.", + "type": "text" + } + ], + "index": 53 + } + ], + "index": 51, + "bbox_fs": [ + 105, + 676, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 505, + 138 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 504, + 94 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 504, + 94 + ], + "score": 1.0, + "content": "Results Table A2 shows that Decision Diffuser and CondDiffuser perform similarly when the", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 93, + 506, + 106 + ], + "spans": [ + { + "bbox": [ + 105, + 93, + 506, + 106 + ], + "score": 1.0, + "content": "agent uses position control. This is because action trajectories resulting from position control are", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 506, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 506, + 117 + ], + "score": 1.0, + "content": "smoother and hence easier to model with diffusion. However, when the agent uses torque control,", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 114, + 505, + 128 + ], + "spans": [ + { + "bbox": [ + 106, + 114, + 505, + 128 + ], + "score": 1.0, + "content": "CondDiffuser performs worse than Decision Diffuser, given the action trajectories have higher", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 126, + 383, + 139 + ], + "spans": [ + { + "bbox": [ + 106, + 126, + 383, + 139 + ], + "score": 1.0, + "content": "frequency components and hence are harder to model with diffusion.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 2 + }, + { + "type": "title", + "bbox": [ + 107, + 158, + 340, + 171 + ], + "lines": [ + { + "bbox": [ + 105, + 156, + 341, + 173 + ], + "spans": [ + { + "bbox": [ + 105, + 156, + 341, + 173 + ], + "score": 1.0, + "content": "G ROBUSTNESS TO STOCHASTIC DYNAMICS", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "table", + "bbox": [ + 183, + 196, + 426, + 259 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 183, + 196, + 426, + 259 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 183, + 196, + 426, + 259 + ], + "spans": [ + { + "bbox": [ + 183, + 196, + 426, + 259 + ], + "score": 0.975, + "html": "
pBCDecision DiffuserDiffuserCQL
0.0055.2±1.584.7±2.272.4±1.473.2±2.3
0.0549.3±3.677.3±3.163.2±2.961.8±3.7
0.1025.8±3.853.2±4.152.3±4.651.2±4.3
0.1515.1±4.341.3±4.941.6±5.142.2±5.5
", + "type": "table", + "image_path": "9a10b8466b781048f4175dc18a2454a22f02774d489b7c634b57535433222a26.jpg" + } + ] + } + ], + "index": 7.5, + "virtual_lines": [ + { + "bbox": [ + 183, + 196, + 426, + 211.75 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 183, + 211.75, + 426, + 227.5 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 183, + 227.5, + 426, + 243.25 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 183, + 243.25, + 426, + 259.0 + ], + "spans": [], + "index": 9 + } + ] + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 106, + 268, + 506, + 318 + ], + "lines": [ + { + "bbox": [ + 105, + 267, + 505, + 279 + ], + "spans": [ + { + "bbox": [ + 105, + 267, + 505, + 279 + ], + "score": 1.0, + "content": "Table A3: Robustness to stochastic dynamics. Decision Diffuser’s performance suffers when stochasticity is", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 277, + 505, + 289 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 408, + 289 + ], + "score": 1.0, + "content": "introduced in dynamics function. While it still outperforms Diffuser and CQL when", + "type": "text" + }, + { + "bbox": [ + 409, + 278, + 443, + 288 + ], + "score": 0.9, + "content": "p = 0 . 0 5", + "type": "inline_equation" + }, + { + "bbox": [ + 444, + 277, + 505, + 289 + ], + "score": 1.0, + "content": ", its performance", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 286, + 505, + 300 + ], + "spans": [ + { + "bbox": [ + 105, + 286, + 300, + 300 + ], + "score": 1.0, + "content": "becomes similar to that of Diffuser and CQL for higher", + "type": "text" + }, + { + "bbox": [ + 301, + 289, + 307, + 298 + ], + "score": 0.8, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 308, + 286, + 505, + 300 + ], + "score": 1.0, + "content": "values. We use the success rate of the red cube reaching", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 299, + 505, + 308 + ], + "spans": [ + { + "bbox": [ + 106, + 299, + 505, + 308 + ], + "score": 1.0, + "content": "the green circle as the performance metric. We report the mean success rate and the standard error over 5 random", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 104, + 307, + 130, + 318 + ], + "spans": [ + { + "bbox": [ + 104, + 307, + 130, + 318 + ], + "score": 1.0, + "content": "seeds.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 12 + }, + { + "type": "text", + "bbox": [ + 107, + 327, + 481, + 339 + ], + "lines": [ + { + "bbox": [ + 104, + 326, + 483, + 342 + ], + "spans": [ + { + "bbox": [ + 104, + 326, + 483, + 342 + ], + "score": 1.0, + "content": "We empirically analyze robustness of Decision Diffuser to stochasticity in dynamics function.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 106, + 341, + 505, + 386 + ], + "lines": [ + { + "bbox": [ + 105, + 341, + 506, + 354 + ], + "spans": [ + { + "bbox": [ + 105, + 341, + 506, + 354 + ], + "score": 1.0, + "content": "Setup We use Block Push environment, described in Appendix F, with torque control. However,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 353, + 505, + 365 + ], + "spans": [ + { + "bbox": [ + 106, + 353, + 505, + 365 + ], + "score": 1.0, + "content": "we inject stochasticity into the environment dynamics. For every environment step, we either sample", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 363, + 505, + 376 + ], + "spans": [ + { + "bbox": [ + 105, + 363, + 194, + 376 + ], + "score": 1.0, + "content": "a random action from", + "type": "text" + }, + { + "bbox": [ + 194, + 363, + 295, + 375 + ], + "score": 0.74, + "content": "\\mathcal { U } ( [ - 1 , - 1 , - 1 ] , [ 1 , 1 , \\dot { 1 } ] )", + "type": "inline_equation" + }, + { + "bbox": [ + 295, + 363, + 363, + 376 + ], + "score": 1.0, + "content": "with probability", + "type": "text" + }, + { + "bbox": [ + 363, + 365, + 370, + 375 + ], + "score": 0.79, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 370, + 363, + 505, + 376 + ], + "score": 1.0, + "content": "or execute the action given by the", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 375, + 442, + 388 + ], + "spans": [ + { + "bbox": [ + 106, + 375, + 200, + 388 + ], + "score": 1.0, + "content": "policy with probability", + "type": "text" + }, + { + "bbox": [ + 201, + 375, + 231, + 387 + ], + "score": 0.91, + "content": "( 1 - p )", + "type": "inline_equation" + }, + { + "bbox": [ + 231, + 375, + 267, + 388 + ], + "score": 1.0, + "content": ". We use", + "type": "text" + }, + { + "bbox": [ + 267, + 375, + 362, + 387 + ], + "score": 0.9, + "content": "p \\in \\{ 0 , 0 . 0 5 , 0 . 1 , 0 . 1 5 \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 375, + 442, + 388 + ], + "score": 1.0, + "content": "in our experiments.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 17.5 + }, + { + "type": "text", + "bbox": [ + 108, + 388, + 504, + 421 + ], + "lines": [ + { + "bbox": [ + 106, + 388, + 504, + 400 + ], + "spans": [ + { + "bbox": [ + 106, + 388, + 504, + 400 + ], + "score": 1.0, + "content": "Offline dataset collection We collect separate offline datasets for different block push", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 400, + 505, + 411 + ], + "spans": [ + { + "bbox": [ + 106, + 400, + 348, + 411 + ], + "score": 1.0, + "content": "environments, each characterized by a different value of", + "type": "text" + }, + { + "bbox": [ + 348, + 401, + 354, + 411 + ], + "score": 0.78, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 355, + 400, + 505, + 411 + ], + "score": 1.0, + "content": ". Each offline dataset consists of 1", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 410, + 448, + 423 + ], + "spans": [ + { + "bbox": [ + 105, + 410, + 448, + 423 + ], + "score": 1.0, + "content": "million environment transitions collected using the method described in Appendix F.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21 + }, + { + "type": "text", + "bbox": [ + 107, + 424, + 505, + 468 + ], + "lines": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "score": 1.0, + "content": "Results Table A3 characterizes how the performance of BC, Decision Diffuser, Diffuser, and CQL", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 435, + 505, + 447 + ], + "spans": [ + { + "bbox": [ + 106, + 435, + 505, + 447 + ], + "score": 1.0, + "content": "changes with increasing stochasticity in the environment dynamics. We observe that the Decision", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 445, + 505, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 445, + 284, + 459 + ], + "score": 1.0, + "content": "Diffuser outperforms Diffuser and CQL for", + "type": "text" + }, + { + "bbox": [ + 284, + 447, + 321, + 457 + ], + "score": 0.9, + "content": "p = 0 . 0 5", + "type": "inline_equation" + }, + { + "bbox": [ + 322, + 445, + 505, + 459 + ], + "score": 1.0, + "content": ", however all methods including the Decision", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 456, + 354, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 343, + 470 + ], + "score": 1.0, + "content": "Diffuser settle to a similar performance for larger values of", + "type": "text" + }, + { + "bbox": [ + 344, + 459, + 350, + 469 + ], + "score": 0.78, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 351, + 456, + 354, + 470 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 24.5 + }, + { + "type": "text", + "bbox": [ + 107, + 473, + 505, + 551 + ], + "lines": [ + { + "bbox": [ + 106, + 473, + 506, + 486 + ], + "spans": [ + { + "bbox": [ + 106, + 473, + 506, + 486 + ], + "score": 1.0, + "content": "Several works (Paster et al., 2022; Yang et al., 2022) have shown that the performance of return-", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 485, + 506, + 497 + ], + "spans": [ + { + "bbox": [ + 105, + 485, + 506, + 497 + ], + "score": 1.0, + "content": "conditioned policies suffers as the stochasticity in environment dynamics increases. This is because", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 496, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 505, + 508 + ], + "score": 1.0, + "content": "the return-conditioned policies aren’t able to distinguish between high returns from good actions", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 508, + 504, + 519 + ], + "spans": [ + { + "bbox": [ + 106, + 508, + 504, + 519 + ], + "score": 1.0, + "content": "and high returns from environment stochasticity. Hence, these return-conditioned policies can learn", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 518, + 506, + 531 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 506, + 531 + ], + "score": 1.0, + "content": "sub-optimal actions that got associated with high-return trajectories in the dataset due to environment", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 529, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 529, + 505, + 541 + ], + "score": 1.0, + "content": "stochasticity. Given Decision diffuser uses return conditioning to generate actions in offline RL, its", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 540, + 429, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 540, + 429, + 552 + ], + "score": 1.0, + "content": "performance also suffers when stochasticity in environment dynamics increases.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 108, + 556, + 505, + 601 + ], + "lines": [ + { + "bbox": [ + 105, + 555, + 506, + 569 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 506, + 569 + ], + "score": 1.0, + "content": "Some recent works (Yang et al., 2022; Villaflor et al., 2022) address the above issue by learning a", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 567, + 506, + 580 + ], + "spans": [ + { + "bbox": [ + 105, + 567, + 506, + 580 + ], + "score": 1.0, + "content": "latent model for future states and then conditioning the policy on predicted latent future states rather", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 578, + 505, + 590 + ], + "spans": [ + { + "bbox": [ + 106, + 578, + 505, + 590 + ], + "score": 1.0, + "content": "than returns. Conditioning Decision Diffuser on future state information, rather than returns, would", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 590, + 491, + 602 + ], + "spans": [ + { + "bbox": [ + 106, + 590, + 491, + 602 + ], + "score": 1.0, + "content": "make it more robust to stochastic dynamics and could be an interesting avenue for future works.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 35.5 + }, + { + "type": "title", + "bbox": [ + 108, + 620, + 257, + 633 + ], + "lines": [ + { + "bbox": [ + 105, + 619, + 258, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 619, + 258, + 635 + ], + "score": 1.0, + "content": "H KUKA BLOCK STACKING", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 107, + 648, + 505, + 682 + ], + "lines": [ + { + "bbox": [ + 105, + 647, + 505, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 647, + 505, + 662 + ], + "score": 1.0, + "content": "In the Kuka blocking stacking environment, the underlying goal is to stack a set of blocks on top", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "score": 1.0, + "content": "of each other. Models have trained on a set of demonstration data, where a set of 4 blocks are", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 671, + 363, + 683 + ], + "spans": [ + { + "bbox": [ + 105, + 671, + 363, + 683 + ], + "score": 1.0, + "content": "sequentially stacked on top of each other to form a block tower.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 40 + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "score": 1.0, + "content": "We construct state-space plans of length 128. Following (Janner et al., 2022), we utilize a close-loop", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "score": 1.0, + "content": "controller to generate actions for each state in our state-space plan (controlling the 7 degrees of", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "freedom in joints). The total maximum trajectory length plan in Kuka block stacking is 384. We", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 720, + 450, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 450, + 733 + ], + "score": 1.0, + "content": "detail differences between the two consider conditional stacking environments below:", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 43.5 + } + ], + "page_idx": 17, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 763 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 763 + ], + "score": 1.0, + "content": "18", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 505, + 138 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 504, + 94 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 504, + 94 + ], + "score": 1.0, + "content": "Results Table A2 shows that Decision Diffuser and CondDiffuser perform similarly when the", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 93, + 506, + 106 + ], + "spans": [ + { + "bbox": [ + 105, + 93, + 506, + 106 + ], + "score": 1.0, + "content": "agent uses position control. This is because action trajectories resulting from position control are", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 506, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 506, + 117 + ], + "score": 1.0, + "content": "smoother and hence easier to model with diffusion. However, when the agent uses torque control,", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 114, + 505, + 128 + ], + "spans": [ + { + "bbox": [ + 106, + 114, + 505, + 128 + ], + "score": 1.0, + "content": "CondDiffuser performs worse than Decision Diffuser, given the action trajectories have higher", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 126, + 383, + 139 + ], + "spans": [ + { + "bbox": [ + 106, + 126, + 383, + 139 + ], + "score": 1.0, + "content": "frequency components and hence are harder to model with diffusion.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 2, + "bbox_fs": [ + 105, + 82, + 506, + 139 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 158, + 340, + 171 + ], + "lines": [ + { + "bbox": [ + 105, + 156, + 341, + 173 + ], + "spans": [ + { + "bbox": [ + 105, + 156, + 341, + 173 + ], + "score": 1.0, + "content": "G ROBUSTNESS TO STOCHASTIC DYNAMICS", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 5 + }, + { + "type": "table", + "bbox": [ + 183, + 196, + 426, + 259 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 183, + 196, + 426, + 259 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 183, + 196, + 426, + 259 + ], + "spans": [ + { + "bbox": [ + 183, + 196, + 426, + 259 + ], + "score": 0.975, + "html": "
pBCDecision DiffuserDiffuserCQL
0.0055.2±1.584.7±2.272.4±1.473.2±2.3
0.0549.3±3.677.3±3.163.2±2.961.8±3.7
0.1025.8±3.853.2±4.152.3±4.651.2±4.3
0.1515.1±4.341.3±4.941.6±5.142.2±5.5
", + "type": "table", + "image_path": "9a10b8466b781048f4175dc18a2454a22f02774d489b7c634b57535433222a26.jpg" + } + ] + } + ], + "index": 7.5, + "virtual_lines": [ + { + "bbox": [ + 183, + 196, + 426, + 211.75 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 183, + 211.75, + 426, + 227.5 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 183, + 227.5, + 426, + 243.25 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 183, + 243.25, + 426, + 259.0 + ], + "spans": [], + "index": 9 + } + ] + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 106, + 268, + 506, + 318 + ], + "lines": [ + { + "bbox": [ + 105, + 267, + 505, + 279 + ], + "spans": [ + { + "bbox": [ + 105, + 267, + 505, + 279 + ], + "score": 1.0, + "content": "Table A3: Robustness to stochastic dynamics. Decision Diffuser’s performance suffers when stochasticity is", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 277, + 505, + 289 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 408, + 289 + ], + "score": 1.0, + "content": "introduced in dynamics function. While it still outperforms Diffuser and CQL when", + "type": "text" + }, + { + "bbox": [ + 409, + 278, + 443, + 288 + ], + "score": 0.9, + "content": "p = 0 . 0 5", + "type": "inline_equation" + }, + { + "bbox": [ + 444, + 277, + 505, + 289 + ], + "score": 1.0, + "content": ", its performance", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 286, + 505, + 300 + ], + "spans": [ + { + "bbox": [ + 105, + 286, + 300, + 300 + ], + "score": 1.0, + "content": "becomes similar to that of Diffuser and CQL for higher", + "type": "text" + }, + { + "bbox": [ + 301, + 289, + 307, + 298 + ], + "score": 0.8, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 308, + 286, + 505, + 300 + ], + "score": 1.0, + "content": "values. We use the success rate of the red cube reaching", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 299, + 505, + 308 + ], + "spans": [ + { + "bbox": [ + 106, + 299, + 505, + 308 + ], + "score": 1.0, + "content": "the green circle as the performance metric. We report the mean success rate and the standard error over 5 random", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 104, + 307, + 130, + 318 + ], + "spans": [ + { + "bbox": [ + 104, + 307, + 130, + 318 + ], + "score": 1.0, + "content": "seeds.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 12, + "bbox_fs": [ + 104, + 267, + 505, + 318 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 327, + 481, + 339 + ], + "lines": [ + { + "bbox": [ + 104, + 326, + 483, + 342 + ], + "spans": [ + { + "bbox": [ + 104, + 326, + 483, + 342 + ], + "score": 1.0, + "content": "We empirically analyze robustness of Decision Diffuser to stochasticity in dynamics function.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 15, + "bbox_fs": [ + 104, + 326, + 483, + 342 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 341, + 505, + 386 + ], + "lines": [ + { + "bbox": [ + 105, + 341, + 506, + 354 + ], + "spans": [ + { + "bbox": [ + 105, + 341, + 506, + 354 + ], + "score": 1.0, + "content": "Setup We use Block Push environment, described in Appendix F, with torque control. However,", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 353, + 505, + 365 + ], + "spans": [ + { + "bbox": [ + 106, + 353, + 505, + 365 + ], + "score": 1.0, + "content": "we inject stochasticity into the environment dynamics. For every environment step, we either sample", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 363, + 505, + 376 + ], + "spans": [ + { + "bbox": [ + 105, + 363, + 194, + 376 + ], + "score": 1.0, + "content": "a random action from", + "type": "text" + }, + { + "bbox": [ + 194, + 363, + 295, + 375 + ], + "score": 0.74, + "content": "\\mathcal { U } ( [ - 1 , - 1 , - 1 ] , [ 1 , 1 , \\dot { 1 } ] )", + "type": "inline_equation" + }, + { + "bbox": [ + 295, + 363, + 363, + 376 + ], + "score": 1.0, + "content": "with probability", + "type": "text" + }, + { + "bbox": [ + 363, + 365, + 370, + 375 + ], + "score": 0.79, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 370, + 363, + 505, + 376 + ], + "score": 1.0, + "content": "or execute the action given by the", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 375, + 442, + 388 + ], + "spans": [ + { + "bbox": [ + 106, + 375, + 200, + 388 + ], + "score": 1.0, + "content": "policy with probability", + "type": "text" + }, + { + "bbox": [ + 201, + 375, + 231, + 387 + ], + "score": 0.91, + "content": "( 1 - p )", + "type": "inline_equation" + }, + { + "bbox": [ + 231, + 375, + 267, + 388 + ], + "score": 1.0, + "content": ". We use", + "type": "text" + }, + { + "bbox": [ + 267, + 375, + 362, + 387 + ], + "score": 0.9, + "content": "p \\in \\{ 0 , 0 . 0 5 , 0 . 1 , 0 . 1 5 \\}", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 375, + 442, + 388 + ], + "score": 1.0, + "content": "in our experiments.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 17.5, + "bbox_fs": [ + 105, + 341, + 506, + 388 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 388, + 504, + 421 + ], + "lines": [ + { + "bbox": [ + 106, + 388, + 504, + 400 + ], + "spans": [ + { + "bbox": [ + 106, + 388, + 504, + 400 + ], + "score": 1.0, + "content": "Offline dataset collection We collect separate offline datasets for different block push", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 400, + 505, + 411 + ], + "spans": [ + { + "bbox": [ + 106, + 400, + 348, + 411 + ], + "score": 1.0, + "content": "environments, each characterized by a different value of", + "type": "text" + }, + { + "bbox": [ + 348, + 401, + 354, + 411 + ], + "score": 0.78, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 355, + 400, + 505, + 411 + ], + "score": 1.0, + "content": ". Each offline dataset consists of 1", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 410, + 448, + 423 + ], + "spans": [ + { + "bbox": [ + 105, + 410, + 448, + 423 + ], + "score": 1.0, + "content": "million environment transitions collected using the method described in Appendix F.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21, + "bbox_fs": [ + 105, + 388, + 505, + 423 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 424, + 505, + 468 + ], + "lines": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "score": 1.0, + "content": "Results Table A3 characterizes how the performance of BC, Decision Diffuser, Diffuser, and CQL", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 435, + 505, + 447 + ], + "spans": [ + { + "bbox": [ + 106, + 435, + 505, + 447 + ], + "score": 1.0, + "content": "changes with increasing stochasticity in the environment dynamics. We observe that the Decision", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 445, + 505, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 445, + 284, + 459 + ], + "score": 1.0, + "content": "Diffuser outperforms Diffuser and CQL for", + "type": "text" + }, + { + "bbox": [ + 284, + 447, + 321, + 457 + ], + "score": 0.9, + "content": "p = 0 . 0 5", + "type": "inline_equation" + }, + { + "bbox": [ + 322, + 445, + 505, + 459 + ], + "score": 1.0, + "content": ", however all methods including the Decision", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 456, + 354, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 343, + 470 + ], + "score": 1.0, + "content": "Diffuser settle to a similar performance for larger values of", + "type": "text" + }, + { + "bbox": [ + 344, + 459, + 350, + 469 + ], + "score": 0.78, + "content": "p", + "type": "inline_equation" + }, + { + "bbox": [ + 351, + 456, + 354, + 470 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 24.5, + "bbox_fs": [ + 105, + 424, + 505, + 470 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 473, + 505, + 551 + ], + "lines": [ + { + "bbox": [ + 106, + 473, + 506, + 486 + ], + "spans": [ + { + "bbox": [ + 106, + 473, + 506, + 486 + ], + "score": 1.0, + "content": "Several works (Paster et al., 2022; Yang et al., 2022) have shown that the performance of return-", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 485, + 506, + 497 + ], + "spans": [ + { + "bbox": [ + 105, + 485, + 506, + 497 + ], + "score": 1.0, + "content": "conditioned policies suffers as the stochasticity in environment dynamics increases. This is because", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 496, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 105, + 496, + 505, + 508 + ], + "score": 1.0, + "content": "the return-conditioned policies aren’t able to distinguish between high returns from good actions", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 508, + 504, + 519 + ], + "spans": [ + { + "bbox": [ + 106, + 508, + 504, + 519 + ], + "score": 1.0, + "content": "and high returns from environment stochasticity. Hence, these return-conditioned policies can learn", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 518, + 506, + 531 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 506, + 531 + ], + "score": 1.0, + "content": "sub-optimal actions that got associated with high-return trajectories in the dataset due to environment", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 529, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 529, + 505, + 541 + ], + "score": 1.0, + "content": "stochasticity. Given Decision diffuser uses return conditioning to generate actions in offline RL, its", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 540, + 429, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 540, + 429, + 552 + ], + "score": 1.0, + "content": "performance also suffers when stochasticity in environment dynamics increases.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 30, + "bbox_fs": [ + 105, + 473, + 506, + 552 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 556, + 505, + 601 + ], + "lines": [ + { + "bbox": [ + 105, + 555, + 506, + 569 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 506, + 569 + ], + "score": 1.0, + "content": "Some recent works (Yang et al., 2022; Villaflor et al., 2022) address the above issue by learning a", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 567, + 506, + 580 + ], + "spans": [ + { + "bbox": [ + 105, + 567, + 506, + 580 + ], + "score": 1.0, + "content": "latent model for future states and then conditioning the policy on predicted latent future states rather", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 578, + 505, + 590 + ], + "spans": [ + { + "bbox": [ + 106, + 578, + 505, + 590 + ], + "score": 1.0, + "content": "than returns. Conditioning Decision Diffuser on future state information, rather than returns, would", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 590, + 491, + 602 + ], + "spans": [ + { + "bbox": [ + 106, + 590, + 491, + 602 + ], + "score": 1.0, + "content": "make it more robust to stochastic dynamics and could be an interesting avenue for future works.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 35.5, + "bbox_fs": [ + 105, + 555, + 506, + 602 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 620, + 257, + 633 + ], + "lines": [ + { + "bbox": [ + 105, + 619, + 258, + 635 + ], + "spans": [ + { + "bbox": [ + 105, + 619, + 258, + 635 + ], + "score": 1.0, + "content": "H KUKA BLOCK STACKING", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 107, + 648, + 505, + 682 + ], + "lines": [ + { + "bbox": [ + 105, + 647, + 505, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 647, + 505, + 662 + ], + "score": 1.0, + "content": "In the Kuka blocking stacking environment, the underlying goal is to stack a set of blocks on top", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 505, + 672 + ], + "score": 1.0, + "content": "of each other. Models have trained on a set of demonstration data, where a set of 4 blocks are", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 671, + 363, + 683 + ], + "spans": [ + { + "bbox": [ + 105, + 671, + 363, + 683 + ], + "score": 1.0, + "content": "sequentially stacked on top of each other to form a block tower.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 40, + "bbox_fs": [ + 105, + 647, + 505, + 683 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "score": 1.0, + "content": "We construct state-space plans of length 128. Following (Janner et al., 2022), we utilize a close-loop", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "score": 1.0, + "content": "controller to generate actions for each state in our state-space plan (controlling the 7 degrees of", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "freedom in joints). The total maximum trajectory length plan in Kuka block stacking is 384. We", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 720, + 450, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 450, + 733 + ], + "score": 1.0, + "content": "detail differences between the two consider conditional stacking environments below:", + "type": "text" + } + ], + "index": 45 + } + ], + "index": 43.5, + "bbox_fs": [ + 105, + 687, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 504, + 105 + ], + "lines": [ + { + "bbox": [ + 104, + 81, + 506, + 96 + ], + "spans": [ + { + "bbox": [ + 104, + 81, + 506, + 96 + ], + "score": 1.0, + "content": "• Stacking In the stacking environment, at test time we wish to again construct a tower of four", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 113, + 91, + 147, + 107 + ], + "spans": [ + { + "bbox": [ + 113, + 91, + 147, + 107 + ], + "score": 1.0, + "content": "blocks.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 107, + 110, + 505, + 154 + ], + "lines": [ + { + "bbox": [ + 106, + 109, + 506, + 122 + ], + "spans": [ + { + "bbox": [ + 106, + 109, + 506, + 122 + ], + "score": 1.0, + "content": "• Rearrangement In the rearrangement environment, at test time wish to stack blocks in a", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 115, + 122, + 505, + 132 + ], + "spans": [ + { + "bbox": [ + 115, + 122, + 505, + 132 + ], + "score": 1.0, + "content": "configuration where a set of blocks are above a second set. This set of stack-place relations", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 114, + 132, + 506, + 144 + ], + "spans": [ + { + "bbox": [ + 114, + 132, + 506, + 144 + ], + "score": 1.0, + "content": "may not precisely correspond to a single block tower (can instead construct two block towers),", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 114, + 142, + 348, + 156 + ], + "spans": [ + { + "bbox": [ + 114, + 142, + 348, + 156 + ], + "score": 1.0, + "content": "making this environment an out-of-distribution challenge.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 107, + 164, + 505, + 198 + ], + "lines": [ + { + "bbox": [ + 105, + 164, + 505, + 177 + ], + "spans": [ + { + "bbox": [ + 105, + 164, + 505, + 177 + ], + "score": 1.0, + "content": "In addition to Diffuser (Janner et al., 2022), we used goal-conditioned variants of CQL (Kumar et al.,", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 175, + 505, + 188 + ], + "spans": [ + { + "bbox": [ + 105, + 175, + 505, + 188 + ], + "score": 1.0, + "content": "2020) and BCQ (Fujimoto et al., 2019) as baselines for the block stacking and rearrangement with", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 186, + 338, + 199 + ], + "spans": [ + { + "bbox": [ + 105, + 186, + 338, + 199 + ], + "score": 1.0, + "content": "single constraint. However, they get a success rate of 0.0.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "title", + "bbox": [ + 108, + 215, + 244, + 228 + ], + "lines": [ + { + "bbox": [ + 104, + 214, + 244, + 231 + ], + "spans": [ + { + "bbox": [ + 104, + 214, + 244, + 231 + ], + "score": 1.0, + "content": "I UNITREE GO RUNNING", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 107, + 242, + 504, + 276 + ], + "lines": [ + { + "bbox": [ + 106, + 243, + 505, + 254 + ], + "spans": [ + { + "bbox": [ + 106, + 243, + 505, + 254 + ], + "score": 1.0, + "content": "We consider Unitree-go-running environment (Margolis & Agrawal, 2022) where a quadruped robot", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 254, + 505, + 266 + ], + "spans": [ + { + "bbox": [ + 106, + 254, + 505, + 266 + ], + "score": 1.0, + "content": "runs in 3 different gaits: bounding, pacing, and trotting. The state space is 56 dimensional, the action", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 264, + 379, + 277 + ], + "spans": [ + { + "bbox": [ + 105, + 264, + 379, + 277 + ], + "score": 1.0, + "content": "space is 12 dimensional, and the maximum trajectory length is 250.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 107, + 281, + 505, + 336 + ], + "lines": [ + { + "bbox": [ + 105, + 280, + 505, + 294 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 505, + 294 + ], + "score": 1.0, + "content": "As described in Section 4.3, we train Decision Diffuser on expert trajectories demonstrating individual", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 292, + 505, + 305 + ], + "spans": [ + { + "bbox": [ + 105, + 292, + 505, + 305 + ], + "score": 1.0, + "content": "gaits. During testing, we compose the noise model of our reverse diffusion process according to", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 104, + 302, + 505, + 317 + ], + "spans": [ + { + "bbox": [ + 104, + 302, + 505, + 317 + ], + "score": 1.0, + "content": "equation 9. This allows us to sample trajectories of the quadruped robot with entirely new running", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 313, + 505, + 328 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 505, + 328 + ], + "score": 1.0, + "content": "behavior. Figures A4,A5,A6 shows the ability of Decision Diffuser to imitate bounding, trotting and", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 325, + 232, + 338 + ], + "spans": [ + { + "bbox": [ + 105, + 325, + 232, + 338 + ], + "score": 1.0, + "content": "pacing and their combinations.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 15 + }, + { + "type": "title", + "bbox": [ + 106, + 352, + 334, + 363 + ], + "lines": [ + { + "bbox": [ + 105, + 351, + 335, + 365 + ], + "spans": [ + { + "bbox": [ + 105, + 351, + 335, + 365 + ], + "score": 1.0, + "content": "I.1 QUANTITATIVE VERIFICATION OF COMPOSITION", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 107, + 372, + 504, + 395 + ], + "lines": [ + { + "bbox": [ + 105, + 371, + 505, + 386 + ], + "spans": [ + { + "bbox": [ + 105, + 371, + 505, + 386 + ], + "score": 1.0, + "content": "We now try to quantitatively verify whether the trajectories resulting from composition of 2 gaits", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 384, + 263, + 397 + ], + "spans": [ + { + "bbox": [ + 106, + 384, + 263, + 397 + ], + "score": 1.0, + "content": "does indeed contain only those 2 gaits.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5 + }, + { + "type": "text", + "bbox": [ + 106, + 397, + 506, + 528 + ], + "lines": [ + { + "bbox": [ + 106, + 398, + 505, + 410 + ], + "spans": [ + { + "bbox": [ + 106, + 398, + 134, + 409 + ], + "score": 1.0, + "content": "Setup", + "type": "text" + }, + { + "bbox": [ + 142, + 398, + 505, + 410 + ], + "score": 1.0, + "content": "We learn a gait classifier that takes in a sub-sequence of states (of length 10) and predicts", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 408, + 506, + 421 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 506, + 421 + ], + "score": 1.0, + "content": "the gait-ID. It is represented by a 3-layered MLP with 1024 hidden units and ReLU activations that", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 420, + 505, + 432 + ], + "spans": [ + { + "bbox": [ + 106, + 420, + 505, + 432 + ], + "score": 1.0, + "content": "concatenates the sub-sequence of states (of length 10) into a single vector of dimension 560 before", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 431, + 505, + 443 + ], + "spans": [ + { + "bbox": [ + 106, + 431, + 505, + 443 + ], + "score": 1.0, + "content": "taking it in as an input. We train the gait classifier on the demonstration dataset. To ensure that", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 441, + 506, + 455 + ], + "spans": [ + { + "bbox": [ + 106, + 441, + 506, + 455 + ], + "score": 1.0, + "content": "the learned classifier can predict gait-ID on trajectories generated by the composition of skills, we", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 452, + 506, + 465 + ], + "spans": [ + { + "bbox": [ + 105, + 452, + 506, + 465 + ], + "score": 1.0, + "content": "use MixUp-style (Zhang et al., 2017) data augmentation during training. We create a synthetic sub-", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 464, + 506, + 476 + ], + "spans": [ + { + "bbox": [ + 105, + 464, + 506, + 476 + ], + "score": 1.0, + "content": "sequence of length 10 by concatenating two sampled sub-sequence (from the demonstration dataset)", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 103, + 470, + 506, + 494 + ], + "spans": [ + { + "bbox": [ + 103, + 470, + 143, + 494 + ], + "score": 1.0, + "content": "of length", + "type": "text" + }, + { + "bbox": [ + 143, + 476, + 151, + 487 + ], + "score": 0.85, + "content": "l _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 151, + 470, + 168, + 494 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 168, + 476, + 177, + 488 + ], + "score": 0.86, + "content": "l _ { j }", + "type": "inline_equation" + }, + { + "bbox": [ + 177, + 470, + 207, + 494 + ], + "score": 1.0, + "content": "(where", + "type": "text" + }, + { + "bbox": [ + 207, + 475, + 256, + 488 + ], + "score": 0.91, + "content": "l _ { i } + l _ { j } = 1 0", + "type": "inline_equation" + }, + { + "bbox": [ + 257, + 470, + 321, + 494 + ], + "score": 1.0, + "content": ") from gaits with", + "type": "text" + }, + { + "bbox": [ + 321, + 475, + 339, + 486 + ], + "score": 0.27, + "content": "\\operatorname { I D } i", + "type": "inline_equation" + }, + { + "bbox": [ + 340, + 470, + 356, + 494 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 356, + 476, + 362, + 487 + ], + "score": 0.82, + "content": "j", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 470, + 433, + 494 + ], + "score": 1.0, + "content": "and give it a label", + "type": "text" + }, + { + "bbox": [ + 434, + 474, + 454, + 491 + ], + "score": 0.91, + "content": "\\frac { l _ { i } } { l _ { i } + l _ { j } }", + "type": "inline_equation" + }, + { + "bbox": [ + 455, + 470, + 484, + 494 + ], + "score": 1.0, + "content": "one-ho", + "type": "text" + }, + { + "bbox": [ + 485, + 475, + 506, + 488 + ], + "score": 0.74, + "content": "( i ) +", + "type": "inline_equation" + } + ], + "index": 28 + }, + { + "bbox": [ + 107, + 489, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 107, + 489, + 127, + 506 + ], + "score": 0.83, + "content": "\\frac { l _ { j } } { l _ { i } + l _ { j } }", + "type": "inline_equation" + }, + { + "bbox": [ + 128, + 489, + 158, + 507 + ], + "score": 1.0, + "content": "one-hot", + "type": "text" + }, + { + "bbox": [ + 158, + 491, + 171, + 504 + ], + "score": 0.37, + "content": "( j )", + "type": "inline_equation" + }, + { + "bbox": [ + 171, + 489, + 506, + 507 + ], + "score": 1.0, + "content": ". During training, we sample a sub-sequence from the demonstration dataset with", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 504, + 505, + 519 + ], + "spans": [ + { + "bbox": [ + 106, + 506, + 126, + 516 + ], + "score": 0.86, + "content": "7 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 504, + 316, + 519 + ], + "score": 1.0, + "content": "probability and a sythenthic sub-sequence with", + "type": "text" + }, + { + "bbox": [ + 316, + 505, + 336, + 516 + ], + "score": 0.88, + "content": "3 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 336, + 504, + 488, + 519 + ], + "score": 1.0, + "content": "probability. We train the classifier for", + "type": "text" + }, + { + "bbox": [ + 488, + 506, + 505, + 516 + ], + "score": 0.36, + "content": "2 e 6", + "type": "inline_equation" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 517, + 361, + 529 + ], + "spans": [ + { + "bbox": [ + 106, + 517, + 240, + 529 + ], + "score": 1.0, + "content": "train steps with a learning rate of", + "type": "text" + }, + { + "bbox": [ + 240, + 517, + 268, + 527 + ], + "score": 0.83, + "content": "2 e - 4", + "type": "inline_equation" + }, + { + "bbox": [ + 269, + 517, + 361, + 529 + ], + "score": 1.0, + "content": "and a batch size of 64.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 26 + }, + { + "type": "text", + "bbox": [ + 106, + 531, + 505, + 619 + ], + "lines": [ + { + "bbox": [ + 106, + 531, + 506, + 543 + ], + "spans": [ + { + "bbox": [ + 106, + 531, + 506, + 543 + ], + "score": 1.0, + "content": "Results Figures A4,A5,A6 show that the classifier’s prediction is consistent with the visualized", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 542, + 505, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 505, + 554 + ], + "score": 1.0, + "content": "composed trajectories. Furthermore, we use Decision diffuser to act in the environment and generate", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 552, + 505, + 565 + ], + "spans": [ + { + "bbox": [ + 105, + 552, + 505, + 565 + ], + "score": 1.0, + "content": "1000 trott trajectories, 1000 pace trajectories, 1000 bound trajectories, and 1000 composed trajectories", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 563, + 505, + 576 + ], + "spans": [ + { + "bbox": [ + 105, + 563, + 505, + 576 + ], + "score": 1.0, + "content": "for each possible pair of individual gaits. We then evaluate the learned gait classifier on these", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 574, + 506, + 587 + ], + "spans": [ + { + "bbox": [ + 105, + 574, + 506, + 587 + ], + "score": 1.0, + "content": "trajectories and compute the percentage of timesteps a particular gait has the highest probability.", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 585, + 505, + 598 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 505, + 598 + ], + "score": 1.0, + "content": "From Figures A4,A5,A6, we can see that if trajectories are generated by the composition of two", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 596, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 596, + 506, + 609 + ], + "score": 1.0, + "content": "gaits, then those two gaits will have the two highest probabilities across different timesteps in those", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 608, + 156, + 620 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 156, + 620 + ], + "score": 1.0, + "content": "trajectories.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 35.5 + }, + { + "type": "title", + "bbox": [ + 108, + 634, + 298, + 644 + ], + "lines": [ + { + "bbox": [ + 106, + 633, + 299, + 646 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 299, + 646 + ], + "score": 1.0, + "content": "I.2 A SIMPLE BASELINE FOR COMPOSITION", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 40 + }, + { + "type": "text", + "bbox": [ + 107, + 654, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 654, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 154, + 668 + ], + "score": 1.0, + "content": "Let one-hot", + "type": "text" + }, + { + "bbox": [ + 154, + 655, + 165, + 667 + ], + "score": 0.56, + "content": "( i )", + "type": "inline_equation" + }, + { + "bbox": [ + 165, + 654, + 200, + 668 + ], + "score": 1.0, + "content": "and one-", + "type": "text" + }, + { + "bbox": [ + 200, + 655, + 227, + 667 + ], + "score": 0.48, + "content": "\\cdot \\mathrm { h o t } ( j )", + "type": "inline_equation" + }, + { + "bbox": [ + 227, + 654, + 506, + 668 + ], + "score": 1.0, + "content": "represent two different gaits that can be generated using noise models", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 107, + 665, + 506, + 679 + ], + "spans": [ + { + "bbox": [ + 107, + 666, + 204, + 678 + ], + "score": 0.67, + "content": "\\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , \\mathrm { o n e - h o t } ( i ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 205, + 665, + 226, + 679 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 226, + 666, + 325, + 678 + ], + "score": 0.81, + "content": "\\begin{array} { r } { \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\mathrm { o n e - h o t } ( j ) , k ) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 325, + 665, + 506, + 679 + ], + "score": 1.0, + "content": "respectively. To compose these gaits, we", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "compose the above-mentioned noise models using equation 9. As an alternative, we see if the noise", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 135, + 700 + ], + "score": 1.0, + "content": "model", + "type": "text" + }, + { + "bbox": [ + 136, + 687, + 290, + 699 + ], + "score": 0.79, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\mathrm { o n e - h o t } ( i ) + \\mathrm { o n e - h o t } ( j ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "can lead to composed gaits. However, we observe", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 125, + 712 + ], + "score": 1.0, + "content": "that", + "type": "text" + }, + { + "bbox": [ + 126, + 699, + 164, + 711 + ], + "score": 0.89, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 164, + 698, + 184, + 712 + ], + "score": 1.0, + "content": ", one-", + "type": "text" + }, + { + "bbox": [ + 184, + 699, + 280, + 711 + ], + "score": 0.82, + "content": "\\mathbf { \\cdot h o t } ( i ) + \\mathbf { o n e - h o t } ( j ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 281, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "catastrophically fail to generate any gait (see videos", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "score": 1.0, + "content": "at https://anuragajay.github.io/decision-diffuser/). This happens because the condition variable", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 720, + 421, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 128, + 734 + ], + "score": 1.0, + "content": "one-", + "type": "text" + }, + { + "bbox": [ + 128, + 721, + 204, + 732 + ], + "score": 0.79, + "content": "\\mathsf { 1 o t } ( i ) + \\mathsf { o n e { \\mathrm { - } } h o t } ( j )", + "type": "inline_equation" + }, + { + "bbox": [ + 205, + 720, + 345, + 734 + ], + "score": 1.0, + "content": "was never seen by the noise model", + "type": "text" + }, + { + "bbox": [ + 346, + 722, + 356, + 732 + ], + "score": 0.85, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 356, + 720, + 421, + 734 + ], + "score": 1.0, + "content": "during training.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 44 + } + ], + "page_idx": 18, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 106, + 82, + 504, + 105 + ], + "lines": [ + { + "bbox": [ + 104, + 81, + 506, + 96 + ], + "spans": [ + { + "bbox": [ + 104, + 81, + 506, + 96 + ], + "score": 1.0, + "content": "• Stacking In the stacking environment, at test time we wish to again construct a tower of four", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 113, + 91, + 147, + 107 + ], + "spans": [ + { + "bbox": [ + 113, + 91, + 147, + 107 + ], + "score": 1.0, + "content": "blocks.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5, + "bbox_fs": [ + 104, + 81, + 506, + 107 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 110, + 505, + 154 + ], + "lines": [ + { + "bbox": [ + 106, + 109, + 506, + 122 + ], + "spans": [ + { + "bbox": [ + 106, + 109, + 506, + 122 + ], + "score": 1.0, + "content": "• Rearrangement In the rearrangement environment, at test time wish to stack blocks in a", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 115, + 122, + 505, + 132 + ], + "spans": [ + { + "bbox": [ + 115, + 122, + 505, + 132 + ], + "score": 1.0, + "content": "configuration where a set of blocks are above a second set. This set of stack-place relations", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 114, + 132, + 506, + 144 + ], + "spans": [ + { + "bbox": [ + 114, + 132, + 506, + 144 + ], + "score": 1.0, + "content": "may not precisely correspond to a single block tower (can instead construct two block towers),", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 114, + 142, + 348, + 156 + ], + "spans": [ + { + "bbox": [ + 114, + 142, + 348, + 156 + ], + "score": 1.0, + "content": "making this environment an out-of-distribution challenge.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 3.5, + "bbox_fs": [ + 106, + 109, + 506, + 156 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 164, + 505, + 198 + ], + "lines": [ + { + "bbox": [ + 105, + 164, + 505, + 177 + ], + "spans": [ + { + "bbox": [ + 105, + 164, + 505, + 177 + ], + "score": 1.0, + "content": "In addition to Diffuser (Janner et al., 2022), we used goal-conditioned variants of CQL (Kumar et al.,", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 175, + 505, + 188 + ], + "spans": [ + { + "bbox": [ + 105, + 175, + 505, + 188 + ], + "score": 1.0, + "content": "2020) and BCQ (Fujimoto et al., 2019) as baselines for the block stacking and rearrangement with", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 186, + 338, + 199 + ], + "spans": [ + { + "bbox": [ + 105, + 186, + 338, + 199 + ], + "score": 1.0, + "content": "single constraint. However, they get a success rate of 0.0.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7, + "bbox_fs": [ + 105, + 164, + 505, + 199 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 215, + 244, + 228 + ], + "lines": [ + { + "bbox": [ + 104, + 214, + 244, + 231 + ], + "spans": [ + { + "bbox": [ + 104, + 214, + 244, + 231 + ], + "score": 1.0, + "content": "I UNITREE GO RUNNING", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 107, + 242, + 504, + 276 + ], + "lines": [ + { + "bbox": [ + 106, + 243, + 505, + 254 + ], + "spans": [ + { + "bbox": [ + 106, + 243, + 505, + 254 + ], + "score": 1.0, + "content": "We consider Unitree-go-running environment (Margolis & Agrawal, 2022) where a quadruped robot", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 254, + 505, + 266 + ], + "spans": [ + { + "bbox": [ + 106, + 254, + 505, + 266 + ], + "score": 1.0, + "content": "runs in 3 different gaits: bounding, pacing, and trotting. The state space is 56 dimensional, the action", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 264, + 379, + 277 + ], + "spans": [ + { + "bbox": [ + 105, + 264, + 379, + 277 + ], + "score": 1.0, + "content": "space is 12 dimensional, and the maximum trajectory length is 250.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11, + "bbox_fs": [ + 105, + 243, + 505, + 277 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 281, + 505, + 336 + ], + "lines": [ + { + "bbox": [ + 105, + 280, + 505, + 294 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 505, + 294 + ], + "score": 1.0, + "content": "As described in Section 4.3, we train Decision Diffuser on expert trajectories demonstrating individual", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 292, + 505, + 305 + ], + "spans": [ + { + "bbox": [ + 105, + 292, + 505, + 305 + ], + "score": 1.0, + "content": "gaits. During testing, we compose the noise model of our reverse diffusion process according to", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 104, + 302, + 505, + 317 + ], + "spans": [ + { + "bbox": [ + 104, + 302, + 505, + 317 + ], + "score": 1.0, + "content": "equation 9. This allows us to sample trajectories of the quadruped robot with entirely new running", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 313, + 505, + 328 + ], + "spans": [ + { + "bbox": [ + 105, + 313, + 505, + 328 + ], + "score": 1.0, + "content": "behavior. Figures A4,A5,A6 shows the ability of Decision Diffuser to imitate bounding, trotting and", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 325, + 232, + 338 + ], + "spans": [ + { + "bbox": [ + 105, + 325, + 232, + 338 + ], + "score": 1.0, + "content": "pacing and their combinations.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 15, + "bbox_fs": [ + 104, + 280, + 505, + 338 + ] + }, + { + "type": "title", + "bbox": [ + 106, + 352, + 334, + 363 + ], + "lines": [ + { + "bbox": [ + 105, + 351, + 335, + 365 + ], + "spans": [ + { + "bbox": [ + 105, + 351, + 335, + 365 + ], + "score": 1.0, + "content": "I.1 QUANTITATIVE VERIFICATION OF COMPOSITION", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 107, + 372, + 504, + 395 + ], + "lines": [ + { + "bbox": [ + 105, + 371, + 505, + 386 + ], + "spans": [ + { + "bbox": [ + 105, + 371, + 505, + 386 + ], + "score": 1.0, + "content": "We now try to quantitatively verify whether the trajectories resulting from composition of 2 gaits", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 384, + 263, + 397 + ], + "spans": [ + { + "bbox": [ + 106, + 384, + 263, + 397 + ], + "score": 1.0, + "content": "does indeed contain only those 2 gaits.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5, + "bbox_fs": [ + 105, + 371, + 505, + 397 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 397, + 506, + 528 + ], + "lines": [ + { + "bbox": [ + 106, + 398, + 505, + 410 + ], + "spans": [ + { + "bbox": [ + 106, + 398, + 134, + 409 + ], + "score": 1.0, + "content": "Setup", + "type": "text" + }, + { + "bbox": [ + 142, + 398, + 505, + 410 + ], + "score": 1.0, + "content": "We learn a gait classifier that takes in a sub-sequence of states (of length 10) and predicts", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 408, + 506, + 421 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 506, + 421 + ], + "score": 1.0, + "content": "the gait-ID. It is represented by a 3-layered MLP with 1024 hidden units and ReLU activations that", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 420, + 505, + 432 + ], + "spans": [ + { + "bbox": [ + 106, + 420, + 505, + 432 + ], + "score": 1.0, + "content": "concatenates the sub-sequence of states (of length 10) into a single vector of dimension 560 before", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 431, + 505, + 443 + ], + "spans": [ + { + "bbox": [ + 106, + 431, + 505, + 443 + ], + "score": 1.0, + "content": "taking it in as an input. We train the gait classifier on the demonstration dataset. To ensure that", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 441, + 506, + 455 + ], + "spans": [ + { + "bbox": [ + 106, + 441, + 506, + 455 + ], + "score": 1.0, + "content": "the learned classifier can predict gait-ID on trajectories generated by the composition of skills, we", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 452, + 506, + 465 + ], + "spans": [ + { + "bbox": [ + 105, + 452, + 506, + 465 + ], + "score": 1.0, + "content": "use MixUp-style (Zhang et al., 2017) data augmentation during training. We create a synthetic sub-", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 464, + 506, + 476 + ], + "spans": [ + { + "bbox": [ + 105, + 464, + 506, + 476 + ], + "score": 1.0, + "content": "sequence of length 10 by concatenating two sampled sub-sequence (from the demonstration dataset)", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 103, + 470, + 506, + 494 + ], + "spans": [ + { + "bbox": [ + 103, + 470, + 143, + 494 + ], + "score": 1.0, + "content": "of length", + "type": "text" + }, + { + "bbox": [ + 143, + 476, + 151, + 487 + ], + "score": 0.85, + "content": "l _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 151, + 470, + 168, + 494 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 168, + 476, + 177, + 488 + ], + "score": 0.86, + "content": "l _ { j }", + "type": "inline_equation" + }, + { + "bbox": [ + 177, + 470, + 207, + 494 + ], + "score": 1.0, + "content": "(where", + "type": "text" + }, + { + "bbox": [ + 207, + 475, + 256, + 488 + ], + "score": 0.91, + "content": "l _ { i } + l _ { j } = 1 0", + "type": "inline_equation" + }, + { + "bbox": [ + 257, + 470, + 321, + 494 + ], + "score": 1.0, + "content": ") from gaits with", + "type": "text" + }, + { + "bbox": [ + 321, + 475, + 339, + 486 + ], + "score": 0.27, + "content": "\\operatorname { I D } i", + "type": "inline_equation" + }, + { + "bbox": [ + 340, + 470, + 356, + 494 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 356, + 476, + 362, + 487 + ], + "score": 0.82, + "content": "j", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 470, + 433, + 494 + ], + "score": 1.0, + "content": "and give it a label", + "type": "text" + }, + { + "bbox": [ + 434, + 474, + 454, + 491 + ], + "score": 0.91, + "content": "\\frac { l _ { i } } { l _ { i } + l _ { j } }", + "type": "inline_equation" + }, + { + "bbox": [ + 455, + 470, + 484, + 494 + ], + "score": 1.0, + "content": "one-ho", + "type": "text" + }, + { + "bbox": [ + 485, + 475, + 506, + 488 + ], + "score": 0.74, + "content": "( i ) +", + "type": "inline_equation" + } + ], + "index": 28 + }, + { + "bbox": [ + 107, + 489, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 107, + 489, + 127, + 506 + ], + "score": 0.83, + "content": "\\frac { l _ { j } } { l _ { i } + l _ { j } }", + "type": "inline_equation" + }, + { + "bbox": [ + 128, + 489, + 158, + 507 + ], + "score": 1.0, + "content": "one-hot", + "type": "text" + }, + { + "bbox": [ + 158, + 491, + 171, + 504 + ], + "score": 0.37, + "content": "( j )", + "type": "inline_equation" + }, + { + "bbox": [ + 171, + 489, + 506, + 507 + ], + "score": 1.0, + "content": ". During training, we sample a sub-sequence from the demonstration dataset with", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 504, + 505, + 519 + ], + "spans": [ + { + "bbox": [ + 106, + 506, + 126, + 516 + ], + "score": 0.86, + "content": "7 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 126, + 504, + 316, + 519 + ], + "score": 1.0, + "content": "probability and a sythenthic sub-sequence with", + "type": "text" + }, + { + "bbox": [ + 316, + 505, + 336, + 516 + ], + "score": 0.88, + "content": "3 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 336, + 504, + 488, + 519 + ], + "score": 1.0, + "content": "probability. We train the classifier for", + "type": "text" + }, + { + "bbox": [ + 488, + 506, + 505, + 516 + ], + "score": 0.36, + "content": "2 e 6", + "type": "inline_equation" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 517, + 361, + 529 + ], + "spans": [ + { + "bbox": [ + 106, + 517, + 240, + 529 + ], + "score": 1.0, + "content": "train steps with a learning rate of", + "type": "text" + }, + { + "bbox": [ + 240, + 517, + 268, + 527 + ], + "score": 0.83, + "content": "2 e - 4", + "type": "inline_equation" + }, + { + "bbox": [ + 269, + 517, + 361, + 529 + ], + "score": 1.0, + "content": "and a batch size of 64.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 26, + "bbox_fs": [ + 103, + 398, + 506, + 529 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 531, + 505, + 619 + ], + "lines": [ + { + "bbox": [ + 106, + 531, + 506, + 543 + ], + "spans": [ + { + "bbox": [ + 106, + 531, + 506, + 543 + ], + "score": 1.0, + "content": "Results Figures A4,A5,A6 show that the classifier’s prediction is consistent with the visualized", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 542, + 505, + 554 + ], + "spans": [ + { + "bbox": [ + 105, + 542, + 505, + 554 + ], + "score": 1.0, + "content": "composed trajectories. Furthermore, we use Decision diffuser to act in the environment and generate", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 552, + 505, + 565 + ], + "spans": [ + { + "bbox": [ + 105, + 552, + 505, + 565 + ], + "score": 1.0, + "content": "1000 trott trajectories, 1000 pace trajectories, 1000 bound trajectories, and 1000 composed trajectories", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 563, + 505, + 576 + ], + "spans": [ + { + "bbox": [ + 105, + 563, + 505, + 576 + ], + "score": 1.0, + "content": "for each possible pair of individual gaits. We then evaluate the learned gait classifier on these", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 574, + 506, + 587 + ], + "spans": [ + { + "bbox": [ + 105, + 574, + 506, + 587 + ], + "score": 1.0, + "content": "trajectories and compute the percentage of timesteps a particular gait has the highest probability.", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 585, + 505, + 598 + ], + "spans": [ + { + "bbox": [ + 105, + 585, + 505, + 598 + ], + "score": 1.0, + "content": "From Figures A4,A5,A6, we can see that if trajectories are generated by the composition of two", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 596, + 506, + 609 + ], + "spans": [ + { + "bbox": [ + 105, + 596, + 506, + 609 + ], + "score": 1.0, + "content": "gaits, then those two gaits will have the two highest probabilities across different timesteps in those", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 608, + 156, + 620 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 156, + 620 + ], + "score": 1.0, + "content": "trajectories.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 35.5, + "bbox_fs": [ + 105, + 531, + 506, + 620 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 634, + 298, + 644 + ], + "lines": [ + { + "bbox": [ + 106, + 633, + 299, + 646 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 299, + 646 + ], + "score": 1.0, + "content": "I.2 A SIMPLE BASELINE FOR COMPOSITION", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 40 + }, + { + "type": "text", + "bbox": [ + 107, + 654, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 654, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 154, + 668 + ], + "score": 1.0, + "content": "Let one-hot", + "type": "text" + }, + { + "bbox": [ + 154, + 655, + 165, + 667 + ], + "score": 0.56, + "content": "( i )", + "type": "inline_equation" + }, + { + "bbox": [ + 165, + 654, + 200, + 668 + ], + "score": 1.0, + "content": "and one-", + "type": "text" + }, + { + "bbox": [ + 200, + 655, + 227, + 667 + ], + "score": 0.48, + "content": "\\cdot \\mathrm { h o t } ( j )", + "type": "inline_equation" + }, + { + "bbox": [ + 227, + 654, + 506, + 668 + ], + "score": 1.0, + "content": "represent two different gaits that can be generated using noise models", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 107, + 665, + 506, + 679 + ], + "spans": [ + { + "bbox": [ + 107, + 666, + 204, + 678 + ], + "score": 0.67, + "content": "\\epsilon _ { \\theta } ( { \\pmb x } _ { k } ( \\tau ) , \\mathrm { o n e - h o t } ( i ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 205, + 665, + 226, + 679 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 226, + 666, + 325, + 678 + ], + "score": 0.81, + "content": "\\begin{array} { r } { \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\mathrm { o n e - h o t } ( j ) , k ) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 325, + 665, + 506, + 679 + ], + "score": 1.0, + "content": "respectively. To compose these gaits, we", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "compose the above-mentioned noise models using equation 9. As an alternative, we see if the noise", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 135, + 700 + ], + "score": 1.0, + "content": "model", + "type": "text" + }, + { + "bbox": [ + 136, + 687, + 290, + 699 + ], + "score": 0.79, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\mathrm { o n e - h o t } ( i ) + \\mathrm { o n e - h o t } ( j ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 290, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "can lead to composed gaits. However, we observe", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 125, + 712 + ], + "score": 1.0, + "content": "that", + "type": "text" + }, + { + "bbox": [ + 126, + 699, + 164, + 711 + ], + "score": 0.89, + "content": "\\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau )", + "type": "inline_equation" + }, + { + "bbox": [ + 164, + 698, + 184, + 712 + ], + "score": 1.0, + "content": ", one-", + "type": "text" + }, + { + "bbox": [ + 184, + 699, + 280, + 711 + ], + "score": 0.82, + "content": "\\mathbf { \\cdot h o t } ( i ) + \\mathbf { o n e - h o t } ( j ) , k )", + "type": "inline_equation" + }, + { + "bbox": [ + 281, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "catastrophically fail to generate any gait (see videos", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "score": 1.0, + "content": "at https://anuragajay.github.io/decision-diffuser/). This happens because the condition variable", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 720, + 421, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 128, + 734 + ], + "score": 1.0, + "content": "one-", + "type": "text" + }, + { + "bbox": [ + 128, + 721, + 204, + 732 + ], + "score": 0.79, + "content": "\\mathsf { 1 o t } ( i ) + \\mathsf { o n e { \\mathrm { - } } h o t } ( j )", + "type": "inline_equation" + }, + { + "bbox": [ + 205, + 720, + 345, + 734 + ], + "score": 1.0, + "content": "was never seen by the noise model", + "type": "text" + }, + { + "bbox": [ + 346, + 722, + 356, + 732 + ], + "score": 0.85, + "content": "\\epsilon _ { \\theta }", + "type": "inline_equation" + }, + { + "bbox": [ + 356, + 720, + 421, + 734 + ], + "score": 1.0, + "content": "during training.", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 44, + "bbox_fs": [ + 105, + 654, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 80, + 506, + 255 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 80, + 506, + 255 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 80, + 505, + 255 + ], + "spans": [ + { + "bbox": [ + 106, + 80, + 505, + 255 + ], + "score": 0.957, + "type": "image", + "image_path": "c6eca9326f1240e24c5bc046689af995227c5ce856ee5e3438368121d3c1e16c.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 80, + 506, + 138.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 138.33333333333334, + 506, + 196.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 196.66666666666669, + 506, + 255.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 266, + 506, + 298 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 265, + 506, + 278 + ], + "spans": [ + { + "bbox": [ + 106, + 265, + 506, + 278 + ], + "score": 1.0, + "content": "Figure A4: Composing Trott and Pace. Decision Diffuser can imitate individual running gaits using expert", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 275, + 505, + 288 + ], + "spans": [ + { + "bbox": [ + 105, + 275, + 505, + 288 + ], + "score": 1.0, + "content": "demonstrations and compose multiple different skills together during test time. The results are best illustrated by", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 286, + 343, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 286, + 343, + 298 + ], + "score": 1.0, + "content": "videos viewable at https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 106, + 309, + 513, + 484 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 309, + 513, + 484 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 309, + 507, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 309, + 507, + 484 + ], + "score": 0.966, + "type": "image", + "image_path": "718ee11d8a554aad4ad0463dc91824091eb322e60b29c2deeafd23463aa4b13e.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 106, + 309, + 513, + 367.3333333333333 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 106, + 367.3333333333333, + 513, + 425.66666666666663 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 106, + 425.66666666666663, + 513, + 483.99999999999994 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 496, + 507, + 527 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 495, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 506, + 507 + ], + "score": 1.0, + "content": "Figure A5: Composing Trott and Bound. Decision Diffuser can imitate individual running gaits using expert", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 505, + 505, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 505, + 518 + ], + "score": 1.0, + "content": "demonstrations and compose multiple different skills together during test time. The results are best illustrated by", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 516, + 343, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 516, + 343, + 528 + ], + "score": 1.0, + "content": "videos viewable at https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + } + ], + "index": 8.5 + }, + { + "type": "title", + "bbox": [ + 106, + 547, + 370, + 561 + ], + "lines": [ + { + "bbox": [ + 105, + 547, + 371, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 371, + 563 + ], + "score": 1.0, + "content": "J NOT COMPOSITIONS WITH DECISION DIFFUSER", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 12 + }, + { + "type": "text", + "bbox": [ + 107, + 574, + 505, + 619 + ], + "lines": [ + { + "bbox": [ + 105, + 573, + 506, + 587 + ], + "spans": [ + { + "bbox": [ + 105, + 573, + 506, + 587 + ], + "score": 1.0, + "content": "Decision diffuser can also support ”NOT” composition. Suppose we wanted to sample from", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 107, + 584, + 507, + 600 + ], + "spans": [ + { + "bbox": [ + 107, + 585, + 194, + 597 + ], + "score": 0.9, + "content": "q ( \\pmb { x } _ { 0 } ( \\tau ) | \\mathrm { N O T } \\ y ^ { j } ( \\tau ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 194, + 584, + 218, + 600 + ], + "score": 1.0, + "content": ". Let", + "type": "text" + }, + { + "bbox": [ + 218, + 585, + 265, + 598 + ], + "score": 0.92, + "content": "\\{ y ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 266, + 584, + 507, + 600 + ], + "score": 1.0, + "content": "be the set of all conditioning variables. Then, following", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 596, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 105, + 596, + 281, + 610 + ], + "score": 1.0, + "content": "derivations from Liu et al. (2022) and using", + "type": "text" + }, + { + "bbox": [ + 281, + 597, + 307, + 608 + ], + "score": 0.9, + "content": "\\beta = 1", + "type": "inline_equation" + }, + { + "bbox": [ + 307, + 596, + 393, + 610 + ], + "score": 1.0, + "content": ", we can sample from", + "type": "text" + }, + { + "bbox": [ + 393, + 596, + 479, + 609 + ], + "score": 0.92, + "content": "q ( \\pmb { x } _ { 0 } ( \\tau ) | \\mathrm { N O T } \\ y ^ { j } ( \\tau ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 480, + 596, + 505, + 610 + ], + "score": 1.0, + "content": "using", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 608, + 188, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 188, + 619 + ], + "score": 1.0, + "content": "the perturbed noise:", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 14.5 + }, + { + "type": "interline_equation", + "bbox": [ + 163, + 620, + 448, + 667 + ], + "lines": [ + { + "bbox": [ + 163, + 620, + 448, + 667 + ], + "spans": [ + { + "bbox": [ + 163, + 620, + 448, + 667 + ], + "score": 0.91, + "content": "\\begin{array} { l } { \\hat { \\epsilon } : = \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\emptyset , k ) + \\omega ( \\displaystyle \\sum _ { i \\neq j } ( \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ^ { i } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\emptyset , k ) ) } \\\\ { - \\left( \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ^ { j } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\emptyset , k ) ) \\right) } \\end{array}", + "type": "interline_equation", + "image_path": "f64ef7f63313279d0fe7890652b4ffe3953e1b4e7a222366a273bc1f6af03c7b.jpg" + } + ] + } + ], + "index": 18, + "virtual_lines": [ + { + "bbox": [ + 163, + 620, + 448, + 635.6666666666666 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 163, + 635.6666666666666, + 448, + 651.3333333333333 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 163, + 651.3333333333333, + 448, + 666.9999999999999 + ], + "spans": [], + "index": 19 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 676, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "We demonstrate the ability of Decision Diffuser to support ”NOT” composition by using it to satisfy", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 687, + 504, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 250, + 700 + ], + "score": 1.0, + "content": "constraint of type BlockHeight", + "type": "text" + }, + { + "bbox": [ + 251, + 687, + 276, + 700 + ], + "score": 0.58, + "content": "( i ) >", + "type": "inline_equation" + }, + { + "bbox": [ + 277, + 687, + 346, + 700 + ], + "score": 1.0, + "content": "BlockHeight", + "type": "text" + }, + { + "bbox": [ + 347, + 688, + 359, + 700 + ], + "score": 0.57, + "content": "( j )", + "type": "inline_equation" + }, + { + "bbox": [ + 360, + 687, + 476, + 700 + ], + "score": 1.0, + "content": "AND (NOT BlockHeight", + "type": "text" + }, + { + "bbox": [ + 477, + 687, + 504, + 700 + ], + "score": 0.53, + "content": ". ( j ) ~ >", + "type": "inline_equation" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 698, + 507, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 172, + 712 + ], + "score": 1.0, + "content": "BlockHeight", + "type": "text" + }, + { + "bbox": [ + 173, + 699, + 185, + 711 + ], + "score": 0.47, + "content": "( i )", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 698, + 507, + 712 + ], + "score": 1.0, + "content": ") in Kuka block stacking task, as visualized in videos at", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "https://anuragajay.github.io/decision-diffuser/. As the Decision Diffuser does not provide", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 720, + 442, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 442, + 733 + ], + "score": 1.0, + "content": "an explicit density estimate for each skill, it can’t natively support OR composition.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 22 + } + ], + "page_idx": 19, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "spans": [ + { + "bbox": [ + 298, + 749, + 313, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 15 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 80, + 506, + 255 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 80, + 506, + 255 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 80, + 505, + 255 + ], + "spans": [ + { + "bbox": [ + 106, + 80, + 505, + 255 + ], + "score": 0.957, + "type": "image", + "image_path": "c6eca9326f1240e24c5bc046689af995227c5ce856ee5e3438368121d3c1e16c.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 80, + 506, + 138.33333333333334 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 138.33333333333334, + 506, + 196.66666666666669 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 196.66666666666669, + 506, + 255.00000000000003 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 266, + 506, + 298 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 265, + 506, + 278 + ], + "spans": [ + { + "bbox": [ + 106, + 265, + 506, + 278 + ], + "score": 1.0, + "content": "Figure A4: Composing Trott and Pace. Decision Diffuser can imitate individual running gaits using expert", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 275, + 505, + 288 + ], + "spans": [ + { + "bbox": [ + 105, + 275, + 505, + 288 + ], + "score": 1.0, + "content": "demonstrations and compose multiple different skills together during test time. The results are best illustrated by", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 286, + 343, + 298 + ], + "spans": [ + { + "bbox": [ + 106, + 286, + 343, + 298 + ], + "score": 1.0, + "content": "videos viewable at https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "image", + "bbox": [ + 106, + 309, + 513, + 484 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 309, + 513, + 484 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 309, + 507, + 484 + ], + "spans": [ + { + "bbox": [ + 106, + 309, + 507, + 484 + ], + "score": 0.966, + "type": "image", + "image_path": "718ee11d8a554aad4ad0463dc91824091eb322e60b29c2deeafd23463aa4b13e.jpg" + } + ] + } + ], + "index": 7, + "virtual_lines": [ + { + "bbox": [ + 106, + 309, + 513, + 367.3333333333333 + ], + "spans": [], + "index": 6 + }, + { + "bbox": [ + 106, + 367.3333333333333, + 513, + 425.66666666666663 + ], + "spans": [], + "index": 7 + }, + { + "bbox": [ + 106, + 425.66666666666663, + 513, + 483.99999999999994 + ], + "spans": [], + "index": 8 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 496, + 507, + 527 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 106, + 495, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 106, + 495, + 506, + 507 + ], + "score": 1.0, + "content": "Figure A5: Composing Trott and Bound. Decision Diffuser can imitate individual running gaits using expert", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 505, + 505, + 518 + ], + "spans": [ + { + "bbox": [ + 105, + 505, + 505, + 518 + ], + "score": 1.0, + "content": "demonstrations and compose multiple different skills together during test time. The results are best illustrated by", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 516, + 343, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 516, + 343, + 528 + ], + "score": 1.0, + "content": "videos viewable at https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 10 + } + ], + "index": 8.5 + }, + { + "type": "title", + "bbox": [ + 106, + 547, + 370, + 561 + ], + "lines": [ + { + "bbox": [ + 105, + 547, + 371, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 547, + 371, + 563 + ], + "score": 1.0, + "content": "J NOT COMPOSITIONS WITH DECISION DIFFUSER", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 12 + }, + { + "type": "text", + "bbox": [ + 107, + 574, + 505, + 619 + ], + "lines": [ + { + "bbox": [ + 105, + 573, + 506, + 587 + ], + "spans": [ + { + "bbox": [ + 105, + 573, + 506, + 587 + ], + "score": 1.0, + "content": "Decision diffuser can also support ”NOT” composition. Suppose we wanted to sample from", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 107, + 584, + 507, + 600 + ], + "spans": [ + { + "bbox": [ + 107, + 585, + 194, + 597 + ], + "score": 0.9, + "content": "q ( \\pmb { x } _ { 0 } ( \\tau ) | \\mathrm { N O T } \\ y ^ { j } ( \\tau ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 194, + 584, + 218, + 600 + ], + "score": 1.0, + "content": ". Let", + "type": "text" + }, + { + "bbox": [ + 218, + 585, + 265, + 598 + ], + "score": 0.92, + "content": "\\{ y ^ { i } ( \\tau ) \\} _ { i = 1 } ^ { n }", + "type": "inline_equation" + }, + { + "bbox": [ + 266, + 584, + 507, + 600 + ], + "score": 1.0, + "content": "be the set of all conditioning variables. Then, following", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 596, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 105, + 596, + 281, + 610 + ], + "score": 1.0, + "content": "derivations from Liu et al. (2022) and using", + "type": "text" + }, + { + "bbox": [ + 281, + 597, + 307, + 608 + ], + "score": 0.9, + "content": "\\beta = 1", + "type": "inline_equation" + }, + { + "bbox": [ + 307, + 596, + 393, + 610 + ], + "score": 1.0, + "content": ", we can sample from", + "type": "text" + }, + { + "bbox": [ + 393, + 596, + 479, + 609 + ], + "score": 0.92, + "content": "q ( \\pmb { x } _ { 0 } ( \\tau ) | \\mathrm { N O T } \\ y ^ { j } ( \\tau ) )", + "type": "inline_equation" + }, + { + "bbox": [ + 480, + 596, + 505, + 610 + ], + "score": 1.0, + "content": "using", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 608, + 188, + 619 + ], + "spans": [ + { + "bbox": [ + 106, + 608, + 188, + 619 + ], + "score": 1.0, + "content": "the perturbed noise:", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 14.5, + "bbox_fs": [ + 105, + 573, + 507, + 619 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 163, + 620, + 448, + 667 + ], + "lines": [ + { + "bbox": [ + 163, + 620, + 448, + 667 + ], + "spans": [ + { + "bbox": [ + 163, + 620, + 448, + 667 + ], + "score": 0.91, + "content": "\\begin{array} { l } { \\hat { \\epsilon } : = \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\emptyset , k ) + \\omega ( \\displaystyle \\sum _ { i \\neq j } ( \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ^ { i } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\emptyset , k ) ) } \\\\ { - \\left( \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\pmb { y } ^ { j } ( \\tau ) , k ) - \\epsilon _ { \\theta } ( \\pmb { x } _ { k } ( \\tau ) , \\emptyset , k ) ) \\right) } \\end{array}", + "type": "interline_equation", + "image_path": "f64ef7f63313279d0fe7890652b4ffe3953e1b4e7a222366a273bc1f6af03c7b.jpg" + } + ] + } + ], + "index": 18, + "virtual_lines": [ + { + "bbox": [ + 163, + 620, + 448, + 635.6666666666666 + ], + "spans": [], + "index": 17 + }, + { + "bbox": [ + 163, + 635.6666666666666, + 448, + 651.3333333333333 + ], + "spans": [], + "index": 18 + }, + { + "bbox": [ + 163, + 651.3333333333333, + 448, + 666.9999999999999 + ], + "spans": [], + "index": 19 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 676, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 676, + 505, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 505, + 689 + ], + "score": 1.0, + "content": "We demonstrate the ability of Decision Diffuser to support ”NOT” composition by using it to satisfy", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 687, + 504, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 250, + 700 + ], + "score": 1.0, + "content": "constraint of type BlockHeight", + "type": "text" + }, + { + "bbox": [ + 251, + 687, + 276, + 700 + ], + "score": 0.58, + "content": "( i ) >", + "type": "inline_equation" + }, + { + "bbox": [ + 277, + 687, + 346, + 700 + ], + "score": 1.0, + "content": "BlockHeight", + "type": "text" + }, + { + "bbox": [ + 347, + 688, + 359, + 700 + ], + "score": 0.57, + "content": "( j )", + "type": "inline_equation" + }, + { + "bbox": [ + 360, + 687, + 476, + 700 + ], + "score": 1.0, + "content": "AND (NOT BlockHeight", + "type": "text" + }, + { + "bbox": [ + 477, + 687, + 504, + 700 + ], + "score": 0.53, + "content": ". ( j ) ~ >", + "type": "inline_equation" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 698, + 507, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 172, + 712 + ], + "score": 1.0, + "content": "BlockHeight", + "type": "text" + }, + { + "bbox": [ + 173, + 699, + 185, + 711 + ], + "score": 0.47, + "content": "( i )", + "type": "inline_equation" + }, + { + "bbox": [ + 185, + 698, + 507, + 712 + ], + "score": 1.0, + "content": ") in Kuka block stacking task, as visualized in videos at", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "https://anuragajay.github.io/decision-diffuser/. As the Decision Diffuser does not provide", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 720, + 442, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 442, + 733 + ], + "score": 1.0, + "content": "an explicit density estimate for each skill, it can’t natively support OR composition.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 22, + "bbox_fs": [ + 105, + 676, + 507, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 79, + 512, + 255 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 79, + 512, + 255 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 79, + 507, + 255 + ], + "spans": [ + { + "bbox": [ + 106, + 79, + 507, + 255 + ], + "score": 0.948, + "type": "image", + "image_path": "79557db21666a3824496a144965288975b56e26dcafa64a8c6e3a89d15f4a7a3.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 79, + 512, + 137.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 137.66666666666666, + 512, + 196.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 196.33333333333331, + 512, + 254.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 266, + 506, + 297 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 265, + 506, + 278 + ], + "spans": [ + { + "bbox": [ + 105, + 265, + 506, + 278 + ], + "score": 1.0, + "content": "Figure A6: Composing Bound and Pace. Decision Diffuser can imitate individual running gaits using expert", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 275, + 505, + 288 + ], + "spans": [ + { + "bbox": [ + 105, + 275, + 505, + 288 + ], + "score": 1.0, + "content": "demonstrations and compose multiple different skills together during test time. The results are best illustrated by", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 286, + 343, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 286, + 343, + 298 + ], + "score": 1.0, + "content": "videos viewable at https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "title", + "bbox": [ + 108, + 320, + 483, + 346 + ], + "lines": [ + { + "bbox": [ + 105, + 317, + 484, + 334 + ], + "spans": [ + { + "bbox": [ + 105, + 317, + 484, + 334 + ], + "score": 1.0, + "content": "K COMPARING Q-FUNCTION GUIDED DIFFUSION AND CLASSIFIER-FREE", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 128, + 335, + 228, + 347 + ], + "spans": [ + { + "bbox": [ + 128, + 335, + 228, + 347 + ], + "score": 1.0, + "content": "GUIDED DIFFUSION", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 108, + 361, + 505, + 405 + ], + "lines": [ + { + "bbox": [ + 105, + 359, + 506, + 374 + ], + "spans": [ + { + "bbox": [ + 105, + 359, + 506, + 374 + ], + "score": 1.0, + "content": "Classifier-free guided diffusion and Q-value guided diffusion are theoretically equivalent. However, as", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 372, + 505, + 384 + ], + "spans": [ + { + "bbox": [ + 105, + 372, + 505, + 384 + ], + "score": 1.0, + "content": "noted in several works (Nichol et al., 2021; Ho & Salimans, 2022; Saharia et al., 2022), classifier-free", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 383, + 507, + 395 + ], + "spans": [ + { + "bbox": [ + 105, + 383, + 507, + 395 + ], + "score": 1.0, + "content": "guidance performs better than classifier guidance (i.e. Q function guidance in our case) in practice.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 393, + 239, + 406 + ], + "spans": [ + { + "bbox": [ + 105, + 393, + 239, + 406 + ], + "score": 1.0, + "content": "This is due to following reasons:", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 9.5 + }, + { + "type": "text", + "bbox": [ + 133, + 416, + 505, + 619 + ], + "lines": [ + { + "bbox": [ + 135, + 416, + 505, + 428 + ], + "spans": [ + { + "bbox": [ + 135, + 416, + 505, + 428 + ], + "score": 1.0, + "content": "• Classifier-guided diffusion models learns an unconditional diffusion model along with", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 427, + 505, + 439 + ], + "spans": [ + { + "bbox": [ + 141, + 427, + 505, + 439 + ], + "score": 1.0, + "content": "a classifier (Q-function in our case) and uses gradients from the classifier to perform", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 438, + 505, + 450 + ], + "spans": [ + { + "bbox": [ + 141, + 438, + 505, + 450 + ], + "score": 1.0, + "content": "conditional sampling. However, the unconditional diffusion model doesn’t need to focus on", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 448, + 505, + 462 + ], + "spans": [ + { + "bbox": [ + 141, + 448, + 505, + 462 + ], + "score": 1.0, + "content": "conditional modeling during training and only cares about conditional generation during", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 142, + 460, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 142, + 460, + 505, + 472 + ], + "score": 1.0, + "content": "testing after it has been trained. In contrast, classifier-free guidance relies on conditional", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 471, + 505, + 483 + ], + "spans": [ + { + "bbox": [ + 141, + 471, + 505, + 483 + ], + "score": 1.0, + "content": "diffusion model to estimate gradients of the implicit classifier. Since the conditional diffusion", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 481, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 141, + 481, + 505, + 495 + ], + "score": 1.0, + "content": "model, learned when using classifier-free guidance, focuses on conditional modeling during", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 493, + 427, + 505 + ], + "spans": [ + { + "bbox": [ + 141, + 493, + 427, + 505 + ], + "score": 1.0, + "content": "train time, it performs better in conditional generation during test time.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 138, + 509, + 506, + 522 + ], + "spans": [ + { + "bbox": [ + 138, + 509, + 506, + 522 + ], + "score": 1.0, + "content": "Q function trained on an offline dataset can erroneously predict high Q values for out-of-", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 141, + 520, + 506, + 533 + ], + "spans": [ + { + "bbox": [ + 141, + 520, + 506, + 533 + ], + "score": 1.0, + "content": "distribution actions given any state. This problem has been extensively studied in offline", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 141, + 532, + 505, + 544 + ], + "spans": [ + { + "bbox": [ + 141, + 532, + 505, + 544 + ], + "score": 1.0, + "content": "RL literature (Kumar et al., 2020; Fujimoto et al., 2019; Levine et al., 2020). In online RL,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 141, + 542, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 141, + 542, + 505, + 555 + ], + "score": 1.0, + "content": "this issue is automatically corrected when the policy acts in the environment, thinking an", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 141, + 554, + 505, + 565 + ], + "spans": [ + { + "bbox": [ + 141, + 554, + 505, + 565 + ], + "score": 1.0, + "content": "action to be good but then receives a low reward for it. In offline RL, this issue can’t be", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 141, + 565, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 141, + 565, + 506, + 577 + ], + "score": 1.0, + "content": "corrected easily; hence, the learned Q-function can often guide the diffusion model towards", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 141, + 576, + 506, + 588 + ], + "spans": [ + { + "bbox": [ + 141, + 576, + 506, + 588 + ], + "score": 1.0, + "content": "out-of-distribution actions that might be sub-optimal. In contrast, classifier-free guidance", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 142, + 587, + 505, + 598 + ], + "spans": [ + { + "bbox": [ + 142, + 587, + 505, + 598 + ], + "score": 1.0, + "content": "circumvents the issue of learning a Q-function and directly conditions the diffusion model on", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 141, + 598, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 141, + 598, + 505, + 610 + ], + "score": 1.0, + "content": "returns. Hence, classifier-free guidance doesn’t suffer due to errors in learned Q-functions", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 141, + 608, + 385, + 622 + ], + "spans": [ + { + "bbox": [ + 141, + 608, + 385, + 622 + ], + "score": 1.0, + "content": "and hence performs better than Q-function guided diffusion.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 20.5 + }, + { + "type": "title", + "bbox": [ + 107, + 638, + 320, + 651 + ], + "lines": [ + { + "bbox": [ + 105, + 637, + 321, + 653 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 321, + 653 + ], + "score": 1.0, + "content": "L LIMITATIONS OF DECISION DIFFUSER", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 107, + 665, + 313, + 677 + ], + "lines": [ + { + "bbox": [ + 105, + 664, + 315, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 315, + 679 + ], + "score": 1.0, + "content": "We summarize the limitations of Decision Diffuser:", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31 + }, + { + "type": "text", + "bbox": [ + 135, + 687, + 505, + 731 + ], + "lines": [ + { + "bbox": [ + 133, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 133, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "• No partial observability Decision Diffuser works with fully observable MDPs. Naive", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 141, + 698, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 141, + 698, + 506, + 711 + ], + "score": 1.0, + "content": "extensions to partially observed MDPs (POMDPs) may cause self-delusions (Ortega et al.,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 141, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 141, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "2021) in Decision Diffuser. Hence, extending Decision Diffuser to POMDPs could be an", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 141, + 721, + 273, + 732 + ], + "spans": [ + { + "bbox": [ + 141, + 721, + 273, + 732 + ], + "score": 1.0, + "content": "exciting avenue for future work.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 33.5 + } + ], + "page_idx": 20, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 298, + 750, + 312, + 765 + ], + "spans": [ + { + "bbox": [ + 298, + 750, + 312, + 765 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 14 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 79, + 512, + 255 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 79, + 512, + 255 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 79, + 507, + 255 + ], + "spans": [ + { + "bbox": [ + 106, + 79, + 507, + 255 + ], + "score": 0.948, + "type": "image", + "image_path": "79557db21666a3824496a144965288975b56e26dcafa64a8c6e3a89d15f4a7a3.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 79, + 512, + 137.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 137.66666666666666, + 512, + 196.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 196.33333333333331, + 512, + 254.99999999999997 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 107, + 266, + 506, + 297 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 265, + 506, + 278 + ], + "spans": [ + { + "bbox": [ + 105, + 265, + 506, + 278 + ], + "score": 1.0, + "content": "Figure A6: Composing Bound and Pace. Decision Diffuser can imitate individual running gaits using expert", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 275, + 505, + 288 + ], + "spans": [ + { + "bbox": [ + 105, + 275, + 505, + 288 + ], + "score": 1.0, + "content": "demonstrations and compose multiple different skills together during test time. The results are best illustrated by", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 286, + 343, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 286, + 343, + 298 + ], + "score": 1.0, + "content": "videos viewable at https://anuragajay.github.io/decision-diffuser/.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "title", + "bbox": [ + 108, + 320, + 483, + 346 + ], + "lines": [ + { + "bbox": [ + 105, + 317, + 484, + 334 + ], + "spans": [ + { + "bbox": [ + 105, + 317, + 484, + 334 + ], + "score": 1.0, + "content": "K COMPARING Q-FUNCTION GUIDED DIFFUSION AND CLASSIFIER-FREE", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 128, + 335, + 228, + 347 + ], + "spans": [ + { + "bbox": [ + 128, + 335, + 228, + 347 + ], + "score": 1.0, + "content": "GUIDED DIFFUSION", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 6.5 + }, + { + "type": "text", + "bbox": [ + 108, + 361, + 505, + 405 + ], + "lines": [ + { + "bbox": [ + 105, + 359, + 506, + 374 + ], + "spans": [ + { + "bbox": [ + 105, + 359, + 506, + 374 + ], + "score": 1.0, + "content": "Classifier-free guided diffusion and Q-value guided diffusion are theoretically equivalent. However, as", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 372, + 505, + 384 + ], + "spans": [ + { + "bbox": [ + 105, + 372, + 505, + 384 + ], + "score": 1.0, + "content": "noted in several works (Nichol et al., 2021; Ho & Salimans, 2022; Saharia et al., 2022), classifier-free", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 383, + 507, + 395 + ], + "spans": [ + { + "bbox": [ + 105, + 383, + 507, + 395 + ], + "score": 1.0, + "content": "guidance performs better than classifier guidance (i.e. Q function guidance in our case) in practice.", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 393, + 239, + 406 + ], + "spans": [ + { + "bbox": [ + 105, + 393, + 239, + 406 + ], + "score": 1.0, + "content": "This is due to following reasons:", + "type": "text" + } + ], + "index": 11 + } + ], + "index": 9.5, + "bbox_fs": [ + 105, + 359, + 507, + 406 + ] + }, + { + "type": "text", + "bbox": [ + 133, + 416, + 505, + 619 + ], + "lines": [ + { + "bbox": [ + 135, + 416, + 505, + 428 + ], + "spans": [ + { + "bbox": [ + 135, + 416, + 505, + 428 + ], + "score": 1.0, + "content": "• Classifier-guided diffusion models learns an unconditional diffusion model along with", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 427, + 505, + 439 + ], + "spans": [ + { + "bbox": [ + 141, + 427, + 505, + 439 + ], + "score": 1.0, + "content": "a classifier (Q-function in our case) and uses gradients from the classifier to perform", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 438, + 505, + 450 + ], + "spans": [ + { + "bbox": [ + 141, + 438, + 505, + 450 + ], + "score": 1.0, + "content": "conditional sampling. However, the unconditional diffusion model doesn’t need to focus on", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 448, + 505, + 462 + ], + "spans": [ + { + "bbox": [ + 141, + 448, + 505, + 462 + ], + "score": 1.0, + "content": "conditional modeling during training and only cares about conditional generation during", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 142, + 460, + 505, + 472 + ], + "spans": [ + { + "bbox": [ + 142, + 460, + 505, + 472 + ], + "score": 1.0, + "content": "testing after it has been trained. In contrast, classifier-free guidance relies on conditional", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 471, + 505, + 483 + ], + "spans": [ + { + "bbox": [ + 141, + 471, + 505, + 483 + ], + "score": 1.0, + "content": "diffusion model to estimate gradients of the implicit classifier. Since the conditional diffusion", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 481, + 505, + 495 + ], + "spans": [ + { + "bbox": [ + 141, + 481, + 505, + 495 + ], + "score": 1.0, + "content": "model, learned when using classifier-free guidance, focuses on conditional modeling during", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 493, + 427, + 505 + ], + "spans": [ + { + "bbox": [ + 141, + 493, + 427, + 505 + ], + "score": 1.0, + "content": "train time, it performs better in conditional generation during test time.", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 138, + 509, + 506, + 522 + ], + "spans": [ + { + "bbox": [ + 138, + 509, + 506, + 522 + ], + "score": 1.0, + "content": "Q function trained on an offline dataset can erroneously predict high Q values for out-of-", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 141, + 520, + 506, + 533 + ], + "spans": [ + { + "bbox": [ + 141, + 520, + 506, + 533 + ], + "score": 1.0, + "content": "distribution actions given any state. This problem has been extensively studied in offline", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 141, + 532, + 505, + 544 + ], + "spans": [ + { + "bbox": [ + 141, + 532, + 505, + 544 + ], + "score": 1.0, + "content": "RL literature (Kumar et al., 2020; Fujimoto et al., 2019; Levine et al., 2020). In online RL,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 141, + 542, + 505, + 555 + ], + "spans": [ + { + "bbox": [ + 141, + 542, + 505, + 555 + ], + "score": 1.0, + "content": "this issue is automatically corrected when the policy acts in the environment, thinking an", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 141, + 554, + 505, + 565 + ], + "spans": [ + { + "bbox": [ + 141, + 554, + 505, + 565 + ], + "score": 1.0, + "content": "action to be good but then receives a low reward for it. In offline RL, this issue can’t be", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 141, + 565, + 506, + 577 + ], + "spans": [ + { + "bbox": [ + 141, + 565, + 506, + 577 + ], + "score": 1.0, + "content": "corrected easily; hence, the learned Q-function can often guide the diffusion model towards", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 141, + 576, + 506, + 588 + ], + "spans": [ + { + "bbox": [ + 141, + 576, + 506, + 588 + ], + "score": 1.0, + "content": "out-of-distribution actions that might be sub-optimal. In contrast, classifier-free guidance", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 142, + 587, + 505, + 598 + ], + "spans": [ + { + "bbox": [ + 142, + 587, + 505, + 598 + ], + "score": 1.0, + "content": "circumvents the issue of learning a Q-function and directly conditions the diffusion model on", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 141, + 598, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 141, + 598, + 505, + 610 + ], + "score": 1.0, + "content": "returns. Hence, classifier-free guidance doesn’t suffer due to errors in learned Q-functions", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 141, + 608, + 385, + 622 + ], + "spans": [ + { + "bbox": [ + 141, + 608, + 385, + 622 + ], + "score": 1.0, + "content": "and hence performs better than Q-function guided diffusion.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 20.5, + "bbox_fs": [ + 135, + 416, + 506, + 622 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 638, + 320, + 651 + ], + "lines": [ + { + "bbox": [ + 105, + 637, + 321, + 653 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 321, + 653 + ], + "score": 1.0, + "content": "L LIMITATIONS OF DECISION DIFFUSER", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 30 + }, + { + "type": "text", + "bbox": [ + 107, + 665, + 313, + 677 + ], + "lines": [ + { + "bbox": [ + 105, + 664, + 315, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 315, + 679 + ], + "score": 1.0, + "content": "We summarize the limitations of Decision Diffuser:", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 31, + "bbox_fs": [ + 105, + 664, + 315, + 679 + ] + }, + { + "type": "text", + "bbox": [ + 135, + 687, + 505, + 731 + ], + "lines": [ + { + "bbox": [ + 133, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 133, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "• No partial observability Decision Diffuser works with fully observable MDPs. Naive", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 141, + 698, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 141, + 698, + 506, + 711 + ], + "score": 1.0, + "content": "extensions to partially observed MDPs (POMDPs) may cause self-delusions (Ortega et al.,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 141, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 141, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "2021) in Decision Diffuser. Hence, extending Decision Diffuser to POMDPs could be an", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 141, + 721, + 273, + 732 + ], + "spans": [ + { + "bbox": [ + 141, + 721, + 273, + 732 + ], + "score": 1.0, + "content": "exciting avenue for future work.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 33.5, + "bbox_fs": [ + 133, + 687, + 506, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 133, + 82, + 505, + 127 + ], + "lines": [ + { + "bbox": [ + 133, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 133, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "• Inability to explore the environment and update itself in online setting In this work, we", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 141, + 93, + 507, + 106 + ], + "spans": [ + { + "bbox": [ + 141, + 93, + 507, + 106 + ], + "score": 1.0, + "content": "focused on offline sequential decision making, thus circumventing the need for exploration.", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 141, + 104, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 141, + 104, + 505, + 118 + ], + "score": 1.0, + "content": "Using ideas from Zheng et al. (2022), future works could look into online fine-tuning of", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 141, + 115, + 483, + 128 + ], + "spans": [ + { + "bbox": [ + 141, + 115, + 483, + 128 + ], + "score": 1.0, + "content": "Decision Diffuser by leveraging entropy of the state-sequence model for exploration.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 1.5 + }, + { + "type": "text", + "bbox": [ + 134, + 131, + 504, + 164 + ], + "lines": [ + { + "bbox": [ + 133, + 130, + 505, + 143 + ], + "spans": [ + { + "bbox": [ + 133, + 130, + 505, + 143 + ], + "score": 1.0, + "content": "• Experiments on only state-based environments While our work focused on state based", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 141, + 141, + 505, + 154 + ], + "spans": [ + { + "bbox": [ + 141, + 141, + 505, + 154 + ], + "score": 1.0, + "content": "environments, it can be extended to image based environments by performing the diffusion", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 141, + 152, + 463, + 165 + ], + "spans": [ + { + "bbox": [ + 141, + 152, + 463, + 165 + ], + "score": 1.0, + "content": "in latent space, rather than observation space, as done in Rombach et al. (2022).", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 134, + 167, + 504, + 200 + ], + "lines": [ + { + "bbox": [ + 133, + 166, + 505, + 180 + ], + "spans": [ + { + "bbox": [ + 133, + 166, + 505, + 180 + ], + "score": 1.0, + "content": "• Only AND and NOT compositions are supported Since Decision Diffuser does not", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 141, + 178, + 505, + 191 + ], + "spans": [ + { + "bbox": [ + 141, + 178, + 505, + 191 + ], + "score": 1.0, + "content": "provide an explicit density estimate for each condition variable, it can’t natively support OR", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 141, + 189, + 197, + 203 + ], + "spans": [ + { + "bbox": [ + 141, + 189, + 197, + 203 + ], + "score": 1.0, + "content": "composition.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 133, + 204, + 505, + 281 + ], + "lines": [ + { + "bbox": [ + 133, + 204, + 505, + 216 + ], + "spans": [ + { + "bbox": [ + 133, + 204, + 505, + 216 + ], + "score": 1.0, + "content": "• Performance degradation in environments with stochastic dynamics In environments", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 214, + 505, + 228 + ], + "spans": [ + { + "bbox": [ + 141, + 214, + 505, + 228 + ], + "score": 1.0, + "content": "with highly stochastic dynamics, Decision Diffuser loses its advantage and performs similarly", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 225, + 506, + 240 + ], + "spans": [ + { + "bbox": [ + 141, + 225, + 506, + 240 + ], + "score": 1.0, + "content": "to Diffuser and CQL. To tackle environments with stochastic dynamics, recent works (Yang", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 236, + 506, + 250 + ], + "spans": [ + { + "bbox": [ + 141, + 236, + 506, + 250 + ], + "score": 1.0, + "content": "et al., 2022; Villaflor et al., 2022) propose learning a latent model for future states and then", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 248, + 505, + 261 + ], + "spans": [ + { + "bbox": [ + 141, + 248, + 505, + 261 + ], + "score": 1.0, + "content": "conditioning the policy on predicted latent future states rather than returns. Conditioning", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 142, + 259, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 142, + 259, + 505, + 271 + ], + "score": 1.0, + "content": "Decision Diffuser on future state information, rather than returns, would make it more robust", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 270, + 445, + 282 + ], + "spans": [ + { + "bbox": [ + 141, + 270, + 445, + 282 + ], + "score": 1.0, + "content": "to stochastic dynamics and could be an interesting avenue for future works.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 135, + 285, + 505, + 307 + ], + "lines": [ + { + "bbox": [ + 133, + 284, + 505, + 297 + ], + "spans": [ + { + "bbox": [ + 133, + 284, + 505, + 297 + ], + "score": 1.0, + "content": "• Performance in limited data regime Since diffusion models are prone to overfitting in case", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 295, + 478, + 308 + ], + "spans": [ + { + "bbox": [ + 141, + 295, + 478, + 308 + ], + "score": 1.0, + "content": "of limited data, Decision Diffuser is also prone to overfitting in limited data regime.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5 + } + ], + "page_idx": 21, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "22", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 133, + 82, + 505, + 127 + ], + "lines": [ + { + "bbox": [ + 133, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 133, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "• Inability to explore the environment and update itself in online setting In this work, we", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 141, + 93, + 507, + 106 + ], + "spans": [ + { + "bbox": [ + 141, + 93, + 507, + 106 + ], + "score": 1.0, + "content": "focused on offline sequential decision making, thus circumventing the need for exploration.", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 141, + 104, + 505, + 118 + ], + "spans": [ + { + "bbox": [ + 141, + 104, + 505, + 118 + ], + "score": 1.0, + "content": "Using ideas from Zheng et al. (2022), future works could look into online fine-tuning of", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 141, + 115, + 483, + 128 + ], + "spans": [ + { + "bbox": [ + 141, + 115, + 483, + 128 + ], + "score": 1.0, + "content": "Decision Diffuser by leveraging entropy of the state-sequence model for exploration.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 1.5, + "bbox_fs": [ + 133, + 82, + 507, + 128 + ] + }, + { + "type": "text", + "bbox": [ + 134, + 131, + 504, + 164 + ], + "lines": [ + { + "bbox": [ + 133, + 130, + 505, + 143 + ], + "spans": [ + { + "bbox": [ + 133, + 130, + 505, + 143 + ], + "score": 1.0, + "content": "• Experiments on only state-based environments While our work focused on state based", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 141, + 141, + 505, + 154 + ], + "spans": [ + { + "bbox": [ + 141, + 141, + 505, + 154 + ], + "score": 1.0, + "content": "environments, it can be extended to image based environments by performing the diffusion", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 141, + 152, + 463, + 165 + ], + "spans": [ + { + "bbox": [ + 141, + 152, + 463, + 165 + ], + "score": 1.0, + "content": "in latent space, rather than observation space, as done in Rombach et al. (2022).", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5, + "bbox_fs": [ + 133, + 130, + 505, + 165 + ] + }, + { + "type": "text", + "bbox": [ + 134, + 167, + 504, + 200 + ], + "lines": [ + { + "bbox": [ + 133, + 166, + 505, + 180 + ], + "spans": [ + { + "bbox": [ + 133, + 166, + 505, + 180 + ], + "score": 1.0, + "content": "• Only AND and NOT compositions are supported Since Decision Diffuser does not", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 141, + 178, + 505, + 191 + ], + "spans": [ + { + "bbox": [ + 141, + 178, + 505, + 191 + ], + "score": 1.0, + "content": "provide an explicit density estimate for each condition variable, it can’t natively support OR", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 141, + 189, + 197, + 203 + ], + "spans": [ + { + "bbox": [ + 141, + 189, + 197, + 203 + ], + "score": 1.0, + "content": "composition.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8, + "bbox_fs": [ + 133, + 166, + 505, + 203 + ] + }, + { + "type": "text", + "bbox": [ + 133, + 204, + 505, + 281 + ], + "lines": [ + { + "bbox": [ + 133, + 204, + 505, + 216 + ], + "spans": [ + { + "bbox": [ + 133, + 204, + 505, + 216 + ], + "score": 1.0, + "content": "• Performance degradation in environments with stochastic dynamics In environments", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 214, + 505, + 228 + ], + "spans": [ + { + "bbox": [ + 141, + 214, + 505, + 228 + ], + "score": 1.0, + "content": "with highly stochastic dynamics, Decision Diffuser loses its advantage and performs similarly", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 225, + 506, + 240 + ], + "spans": [ + { + "bbox": [ + 141, + 225, + 506, + 240 + ], + "score": 1.0, + "content": "to Diffuser and CQL. To tackle environments with stochastic dynamics, recent works (Yang", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 236, + 506, + 250 + ], + "spans": [ + { + "bbox": [ + 141, + 236, + 506, + 250 + ], + "score": 1.0, + "content": "et al., 2022; Villaflor et al., 2022) propose learning a latent model for future states and then", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 248, + 505, + 261 + ], + "spans": [ + { + "bbox": [ + 141, + 248, + 505, + 261 + ], + "score": 1.0, + "content": "conditioning the policy on predicted latent future states rather than returns. Conditioning", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 142, + 259, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 142, + 259, + 505, + 271 + ], + "score": 1.0, + "content": "Decision Diffuser on future state information, rather than returns, would make it more robust", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 141, + 270, + 445, + 282 + ], + "spans": [ + { + "bbox": [ + 141, + 270, + 445, + 282 + ], + "score": 1.0, + "content": "to stochastic dynamics and could be an interesting avenue for future works.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 13, + "bbox_fs": [ + 133, + 204, + 506, + 282 + ] + }, + { + "type": "text", + "bbox": [ + 135, + 285, + 505, + 307 + ], + "lines": [ + { + "bbox": [ + 133, + 284, + 505, + 297 + ], + "spans": [ + { + "bbox": [ + 133, + 284, + 505, + 297 + ], + "score": 1.0, + "content": "• Performance in limited data regime Since diffusion models are prone to overfitting in case", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 295, + 478, + 308 + ], + "spans": [ + { + "bbox": [ + 141, + 295, + 478, + 308 + ], + "score": 1.0, + "content": "of limited data, Decision Diffuser is also prone to overfitting in limited data regime.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 17.5, + "bbox_fs": [ + 133, + 284, + 505, + 308 + ] + } + ] + } + ], + "_backend": "pipeline", + "_version_name": "2.2.2" +} \ No newline at end of file diff --git a/parse/dev/sde_7ZzGXOE/images/17f4c675f601e8649c69da852afbb57c9bc53518793db9912e5cb86a945d71b9.jpg b/parse/dev/sde_7ZzGXOE/images/17f4c675f601e8649c69da852afbb57c9bc53518793db9912e5cb86a945d71b9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b9a84b82783b7ed804cd8f12632fe0760c4daca3 --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/17f4c675f601e8649c69da852afbb57c9bc53518793db9912e5cb86a945d71b9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ab7eff7aa215c3a1a1e8c21f388319f69ffb9e2f5542d078108e722e020e0d7 +size 7322 diff --git a/parse/dev/sde_7ZzGXOE/images/1d304630e386653569fabb5853f7b4ea4b50ba295f72f2eff117a658f689ea9a.jpg b/parse/dev/sde_7ZzGXOE/images/1d304630e386653569fabb5853f7b4ea4b50ba295f72f2eff117a658f689ea9a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..13b0e68e4402241efe88b31f5172316e80a2bdd6 --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/1d304630e386653569fabb5853f7b4ea4b50ba295f72f2eff117a658f689ea9a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e8af9915e0446dff7f96d84c48a4cb68dbf539aab7f75ed914e9bbb8870a0cd +size 8905 diff --git a/parse/dev/sde_7ZzGXOE/images/2973cdde62a109363d6db7e26e0a860ac60a5781d325af728997ac9821c70b2c.jpg b/parse/dev/sde_7ZzGXOE/images/2973cdde62a109363d6db7e26e0a860ac60a5781d325af728997ac9821c70b2c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dcbee2fda027c91da3159289233d2c01b3e23a9f --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/2973cdde62a109363d6db7e26e0a860ac60a5781d325af728997ac9821c70b2c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b6a60d21053471bcc1d1fc729013e91817cc1210074ea0213837f76503a1934 +size 9594 diff --git a/parse/dev/sde_7ZzGXOE/images/445d2a6f3b3a7adf12fefeb870eca3a119428b19452345a11e6d9d2fe1f6461f.jpg b/parse/dev/sde_7ZzGXOE/images/445d2a6f3b3a7adf12fefeb870eca3a119428b19452345a11e6d9d2fe1f6461f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d251609521e14f8b7b79176b592b4e2f2178807a --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/445d2a6f3b3a7adf12fefeb870eca3a119428b19452345a11e6d9d2fe1f6461f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e54062baa4ebb6908e165b73bbfceed9a4ef9c8e0a5aa174f866f0ef6869285 +size 11630 diff --git a/parse/dev/sde_7ZzGXOE/images/476a7c2d2fbeed17edbe07ce195b725a07ef1115434afa0b3ba660962d5e7a30.jpg b/parse/dev/sde_7ZzGXOE/images/476a7c2d2fbeed17edbe07ce195b725a07ef1115434afa0b3ba660962d5e7a30.jpg new file mode 100644 index 0000000000000000000000000000000000000000..39b293ff9349cb35907e3b4acd2f1ad8c0443416 --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/476a7c2d2fbeed17edbe07ce195b725a07ef1115434afa0b3ba660962d5e7a30.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:90a01aee09f707fc31bf86be5129f40f12bab48ba22072734a36331c355f74d3 +size 7801 diff --git a/parse/dev/sde_7ZzGXOE/images/493f0aa925cf2c2f999f1ada18a2bdc2c88187706115cd3ef76fb1f76abe140d.jpg b/parse/dev/sde_7ZzGXOE/images/493f0aa925cf2c2f999f1ada18a2bdc2c88187706115cd3ef76fb1f76abe140d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9461a61fb3fbdaddaf3b4e880e5acb4006728e72 --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/493f0aa925cf2c2f999f1ada18a2bdc2c88187706115cd3ef76fb1f76abe140d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2018fcf70c9cdd9ee6532ab4eab19162a457e5403386bd3b3d5c6c03013d3a3b +size 12974 diff --git a/parse/dev/sde_7ZzGXOE/images/5da25bf1e49ec1befc5b0fbc1af6cf3a3a4fda3beb0c466a4e155fb4f8504660.jpg b/parse/dev/sde_7ZzGXOE/images/5da25bf1e49ec1befc5b0fbc1af6cf3a3a4fda3beb0c466a4e155fb4f8504660.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fc090189c8160e627ac0a3800d2168533b8a538d --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/5da25bf1e49ec1befc5b0fbc1af6cf3a3a4fda3beb0c466a4e155fb4f8504660.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2f307140e06f38be24877d4650995976edd1f8d8ec84217d4cd2ae8237418b1d +size 4838 diff --git a/parse/dev/sde_7ZzGXOE/images/771f388a0acf02cf575f5b34878919d4b98678252e8fd0813e3635f5a0a537a3.jpg b/parse/dev/sde_7ZzGXOE/images/771f388a0acf02cf575f5b34878919d4b98678252e8fd0813e3635f5a0a537a3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a456caf89314f4513340474dfe9ffa0d4ba3e652 --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/771f388a0acf02cf575f5b34878919d4b98678252e8fd0813e3635f5a0a537a3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f37c7571c90e6c6841f6b731343d2836eb226d170974d5cf8e6d46e220e0f0b4 +size 4137 diff --git a/parse/dev/sde_7ZzGXOE/images/87c1dd47d9a72a457873bc6e445f9bd010d3c6fa439b0425bbc3529a4ed0569e.jpg b/parse/dev/sde_7ZzGXOE/images/87c1dd47d9a72a457873bc6e445f9bd010d3c6fa439b0425bbc3529a4ed0569e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..eb5bd33145655806acc5a48c4619015d893bf67e --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/87c1dd47d9a72a457873bc6e445f9bd010d3c6fa439b0425bbc3529a4ed0569e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:15110a673f26d0e2bebeb179e7a079c753ea2762ab9e338f032c728bb22f851e +size 4464 diff --git a/parse/dev/sde_7ZzGXOE/images/a22f1f6610c06ecbc9d0558b64009beff6ce860ffb7b48b8f68079715e548d88.jpg b/parse/dev/sde_7ZzGXOE/images/a22f1f6610c06ecbc9d0558b64009beff6ce860ffb7b48b8f68079715e548d88.jpg new file mode 100644 index 0000000000000000000000000000000000000000..83fab013f724a4eedf6fcc46e507a16e88b90c4b --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/a22f1f6610c06ecbc9d0558b64009beff6ce860ffb7b48b8f68079715e548d88.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4870912e8412a143ce692561be152767b77e4c03d0685bf1964d6537d3c0da54 +size 8917 diff --git a/parse/dev/sde_7ZzGXOE/images/abc25bd72f555ee3a0591da4949c90acf37aea34739a0852c279ca21e93f140a.jpg b/parse/dev/sde_7ZzGXOE/images/abc25bd72f555ee3a0591da4949c90acf37aea34739a0852c279ca21e93f140a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ac8b18d98ae234ffe51371b5281f326a95316776 --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/abc25bd72f555ee3a0591da4949c90acf37aea34739a0852c279ca21e93f140a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4e9665b1412a2ee3c6ae1f0284316cafb2dbc1524f2f11ff686bfc28a0d6d80f +size 8327 diff --git a/parse/dev/sde_7ZzGXOE/images/b10b5aab01bae97bef324b4ed22518baa1c8212cd751ea5090655952d04c1e09.jpg b/parse/dev/sde_7ZzGXOE/images/b10b5aab01bae97bef324b4ed22518baa1c8212cd751ea5090655952d04c1e09.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1797aa41cecb3d0902fa4f6b1cbff7f5d5f470ac --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/b10b5aab01bae97bef324b4ed22518baa1c8212cd751ea5090655952d04c1e09.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1267d2b59af88d1cc29b795090dc69ec30a860a84287a01787fa0713ccd0bcf3 +size 4625 diff --git a/parse/dev/sde_7ZzGXOE/images/bf2f97232b528a8f6d9004e3691335ab6f1119a2f123eb5f4370344b8df0e7ee.jpg b/parse/dev/sde_7ZzGXOE/images/bf2f97232b528a8f6d9004e3691335ab6f1119a2f123eb5f4370344b8df0e7ee.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b60715fefb086add63fd91604179e9f040f99063 --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/bf2f97232b528a8f6d9004e3691335ab6f1119a2f123eb5f4370344b8df0e7ee.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13b1af1d17f9f7bffedd9093c3c38248dabc305f6a206e212294c3d24c919d17 +size 7041 diff --git a/parse/dev/sde_7ZzGXOE/images/e20b2e4e75f563987df590c49ac04b35a09d1d933373a5743f807974753e556a.jpg b/parse/dev/sde_7ZzGXOE/images/e20b2e4e75f563987df590c49ac04b35a09d1d933373a5743f807974753e556a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8e348e68d1ccde5568aae7a1ae6c4b1e862b7695 --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/e20b2e4e75f563987df590c49ac04b35a09d1d933373a5743f807974753e556a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9ace4fec601a5f685cc5e29f8111e6e7d712c41be600612e55b8cb0bb6828b9a +size 43766 diff --git a/parse/dev/sde_7ZzGXOE/images/e2d9c371f3be8f2517595f0ad9b84c1cd24328aab6852a65d8bdfa54bd71a924.jpg b/parse/dev/sde_7ZzGXOE/images/e2d9c371f3be8f2517595f0ad9b84c1cd24328aab6852a65d8bdfa54bd71a924.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7928dbfe6aba787825d2be7a8906219343f440f7 --- /dev/null +++ b/parse/dev/sde_7ZzGXOE/images/e2d9c371f3be8f2517595f0ad9b84c1cd24328aab6852a65d8bdfa54bd71a924.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c4484fad458704fbb9f0ad9b3e74625d041722cedebf07e218dec9b76d99fda +size 8295 diff --git a/parse/dev/uKiE0VIluA-/images/050ad3c56fb65192044ff4f90cd1d727474e9d47f611a76cd90679ca2e80bdfd.jpg b/parse/dev/uKiE0VIluA-/images/050ad3c56fb65192044ff4f90cd1d727474e9d47f611a76cd90679ca2e80bdfd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3e0248c8404e6c29eabd90435bcea190f3c571d5 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/050ad3c56fb65192044ff4f90cd1d727474e9d47f611a76cd90679ca2e80bdfd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62d33fcfc350b7fc7e29f60e62c1d95a2e46c05817b4ed5b8779947ff52b043b +size 15874 diff --git a/parse/dev/uKiE0VIluA-/images/06df1d28e896c52e2c2be29a592a85ec5522a5be3242071f76840602944f2744.jpg b/parse/dev/uKiE0VIluA-/images/06df1d28e896c52e2c2be29a592a85ec5522a5be3242071f76840602944f2744.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0aaf2a7f1a1127ef07c41a5c6f102710268a686a --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/06df1d28e896c52e2c2be29a592a85ec5522a5be3242071f76840602944f2744.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9530c4b1ef43ba3fff729279bac06c0be3e43b3efddbe9157bc6d089c72d7b7 +size 12906 diff --git a/parse/dev/uKiE0VIluA-/images/071353b312af1a8ba2b23b8ffb446b50c99600e47b3e1c0e79f1e5dc04fac28d.jpg b/parse/dev/uKiE0VIluA-/images/071353b312af1a8ba2b23b8ffb446b50c99600e47b3e1c0e79f1e5dc04fac28d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2fcea1a1ee10ab8c27767a4ba83f9f0893b28c55 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/071353b312af1a8ba2b23b8ffb446b50c99600e47b3e1c0e79f1e5dc04fac28d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89cf9becf20943b35060a0d52e84af3d1abe283df4dcc8688698e2fe39af1f5f +size 27758 diff --git a/parse/dev/uKiE0VIluA-/images/10a3fba96f9bc80055d306e21f18e65c5da87283527dae0426e74206aaf0cacc.jpg b/parse/dev/uKiE0VIluA-/images/10a3fba96f9bc80055d306e21f18e65c5da87283527dae0426e74206aaf0cacc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..84c63385663614c9213d713e2c7374fe051d4f9b --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/10a3fba96f9bc80055d306e21f18e65c5da87283527dae0426e74206aaf0cacc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:64602f393b45b89b650c565151b6cebc10dba6c72d7c7cfb7e13635d282d8880 +size 29236 diff --git a/parse/dev/uKiE0VIluA-/images/11a6b0fc6172afbfd3e3708bd8b67d632b7f1d49093d4dd038a7155f8b5ef5e3.jpg b/parse/dev/uKiE0VIluA-/images/11a6b0fc6172afbfd3e3708bd8b67d632b7f1d49093d4dd038a7155f8b5ef5e3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..815aed770a81130a864ae5597d38aa68b786662a --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/11a6b0fc6172afbfd3e3708bd8b67d632b7f1d49093d4dd038a7155f8b5ef5e3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5468fff9ba4bf3f6371158b3f0138e1b2748f69b0f832a9cc1e946d17097bf8d +size 9740 diff --git a/parse/dev/uKiE0VIluA-/images/1452e0c91ffb48d8c69c950a13dd2e6d70e8415c077692e13aa2d08eb8ec41de.jpg b/parse/dev/uKiE0VIluA-/images/1452e0c91ffb48d8c69c950a13dd2e6d70e8415c077692e13aa2d08eb8ec41de.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d131ca17e40562fe2056a1e63204690b07b40d80 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/1452e0c91ffb48d8c69c950a13dd2e6d70e8415c077692e13aa2d08eb8ec41de.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c026fc1f1df1448ccd7fed1b6372d2eb05fc6b18d504435644444a128808bc67 +size 68538 diff --git a/parse/dev/uKiE0VIluA-/images/15c9a02ba082d20a9e60253556f92d0a6016b55cef19a9b49624003be0a80b6b.jpg b/parse/dev/uKiE0VIluA-/images/15c9a02ba082d20a9e60253556f92d0a6016b55cef19a9b49624003be0a80b6b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b34f9c858c5f426e56243d2d1cbb74c0bd8b05c7 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/15c9a02ba082d20a9e60253556f92d0a6016b55cef19a9b49624003be0a80b6b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:48673d6673ebb2ba810a49af5c7c2ddc92fdf77aa99553f340a05d52dd5c122d +size 49980 diff --git a/parse/dev/uKiE0VIluA-/images/17c9842392d31a939ce412d846d6b1a1cd3fb3a611e6b9a832c2e5b3792eae51.jpg b/parse/dev/uKiE0VIluA-/images/17c9842392d31a939ce412d846d6b1a1cd3fb3a611e6b9a832c2e5b3792eae51.jpg new file mode 100644 index 0000000000000000000000000000000000000000..087284ff9ddfffb3c2f9d2765082db4c06b897e5 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/17c9842392d31a939ce412d846d6b1a1cd3fb3a611e6b9a832c2e5b3792eae51.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a0d3569bee357616bead1dccc8dc857174903ad1a86b7fd8c75a56611b0a54c +size 21544 diff --git a/parse/dev/uKiE0VIluA-/images/18d572fc180e042bb360a3a9313ddd3c194ada5258ef89fbb3ceae2608dbf0bc.jpg b/parse/dev/uKiE0VIluA-/images/18d572fc180e042bb360a3a9313ddd3c194ada5258ef89fbb3ceae2608dbf0bc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2f2539d7a3dbf43be476cd11cc0ea8049709617f --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/18d572fc180e042bb360a3a9313ddd3c194ada5258ef89fbb3ceae2608dbf0bc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c36669a129f650153228529ad4c7c27c608cfa2c4ab208c08e9b568d25f58ef7 +size 6123 diff --git a/parse/dev/uKiE0VIluA-/images/1d393bed800a50a3c2e3af0d4a7fea2c914bcd39ac3e9c4b05bddd5e3be3d99b.jpg b/parse/dev/uKiE0VIluA-/images/1d393bed800a50a3c2e3af0d4a7fea2c914bcd39ac3e9c4b05bddd5e3be3d99b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1cfe17f8a5a67c9e1509d1e7fbaccdc86b0a9ca7 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/1d393bed800a50a3c2e3af0d4a7fea2c914bcd39ac3e9c4b05bddd5e3be3d99b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c04867db9ac8fe0a56acea55560f714537be8b2edba9f8730edd48915b779ad +size 5153 diff --git a/parse/dev/uKiE0VIluA-/images/22695240a5423e46d6f20b16bc0734d445454f3654be146e18e54df972d1c69e.jpg b/parse/dev/uKiE0VIluA-/images/22695240a5423e46d6f20b16bc0734d445454f3654be146e18e54df972d1c69e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..00fcd5de03a76c5b122d338b29bddf99f509cf9f --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/22695240a5423e46d6f20b16bc0734d445454f3654be146e18e54df972d1c69e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e2b5877adc9b057a208a65b0e26ccf10c2b5630bdd48db56f02637f7ce6c344 +size 13572 diff --git a/parse/dev/uKiE0VIluA-/images/235f321ca103acb2a1cc4079299b5aef0bdc169ce5db31a3e362d6ae2d1b6678.jpg b/parse/dev/uKiE0VIluA-/images/235f321ca103acb2a1cc4079299b5aef0bdc169ce5db31a3e362d6ae2d1b6678.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fadb19f31313a68aa035339ad705163851629f37 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/235f321ca103acb2a1cc4079299b5aef0bdc169ce5db31a3e362d6ae2d1b6678.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:28f29f574d0714899f0744ccf42db38ed2a1694e0587311e24d244f0893ba604 +size 8643 diff --git a/parse/dev/uKiE0VIluA-/images/2a191dd236647a96c8800f8f836ac76dc4fdfdd47d4e482992c3f69e5cf0012c.jpg b/parse/dev/uKiE0VIluA-/images/2a191dd236647a96c8800f8f836ac76dc4fdfdd47d4e482992c3f69e5cf0012c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9d120abb5c29daa7e7e2e03a1002e63e1c7dd4c8 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/2a191dd236647a96c8800f8f836ac76dc4fdfdd47d4e482992c3f69e5cf0012c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e41ba1c0a55914f11cc95a61f50469f43558ee62ebde9da0b684efb5b865697 +size 3555 diff --git a/parse/dev/uKiE0VIluA-/images/2e194fafb01eaf8f542ca19586543c18e55ed01adf5c21ffda92bccfc8f70e5e.jpg b/parse/dev/uKiE0VIluA-/images/2e194fafb01eaf8f542ca19586543c18e55ed01adf5c21ffda92bccfc8f70e5e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4fd98a10fef14a25507229370bc951bd53500702 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/2e194fafb01eaf8f542ca19586543c18e55ed01adf5c21ffda92bccfc8f70e5e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14baf75f6797a55ef05917b12e2a4f39d78ec6e8593e6607cd90aabcf47cdf0f +size 21510 diff --git a/parse/dev/uKiE0VIluA-/images/3297acfa0052326f3a64ba5ab217be6274d0e368b4d8aeef9d391b5fb344ea43.jpg b/parse/dev/uKiE0VIluA-/images/3297acfa0052326f3a64ba5ab217be6274d0e368b4d8aeef9d391b5fb344ea43.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6fca600c07b2e9328d4d15ec216889a323a3d76e --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/3297acfa0052326f3a64ba5ab217be6274d0e368b4d8aeef9d391b5fb344ea43.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24f4ced29a8ab75dd109f4dbf3ab0c746b167486365b14628b58f05bead5cd7a +size 39638 diff --git a/parse/dev/uKiE0VIluA-/images/3daef70170f60f8d02bc727e408d4f971b4d6b8415772862bbdc6b2f50bc1b7e.jpg b/parse/dev/uKiE0VIluA-/images/3daef70170f60f8d02bc727e408d4f971b4d6b8415772862bbdc6b2f50bc1b7e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..aa38e4b9d981ea85567d2dae29a754601e578a61 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/3daef70170f60f8d02bc727e408d4f971b4d6b8415772862bbdc6b2f50bc1b7e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fd71b73384b744b33e6eefd18b4fcdeb0d9b3905fc79da93fb89887d326c585a +size 7243 diff --git a/parse/dev/uKiE0VIluA-/images/4004829abfdd8793b862f9f64e933039aa1da7bcfd0fc169d68ad19441427d65.jpg b/parse/dev/uKiE0VIluA-/images/4004829abfdd8793b862f9f64e933039aa1da7bcfd0fc169d68ad19441427d65.jpg new file mode 100644 index 0000000000000000000000000000000000000000..32953a934414fb181a4ab5f0e0b4072949792f9b --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/4004829abfdd8793b862f9f64e933039aa1da7bcfd0fc169d68ad19441427d65.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9200f3ce6f8d5bc1cf8ccedc224be8f3ec747e8074c4de69f46cfb6fee821213 +size 13391 diff --git a/parse/dev/uKiE0VIluA-/images/40559583cd69984d8020e878a9f5939abc1c20fe4352361e0470b4726c6f60a2.jpg b/parse/dev/uKiE0VIluA-/images/40559583cd69984d8020e878a9f5939abc1c20fe4352361e0470b4726c6f60a2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2becf249bb68801c97864e8ab86041ed83ca037c --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/40559583cd69984d8020e878a9f5939abc1c20fe4352361e0470b4726c6f60a2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5a454236e9220e9156ef5d5e9c72087d3959c41448427bc0dc92e73803030ed +size 12178 diff --git a/parse/dev/uKiE0VIluA-/images/438c5ccd20e31bd30576f642f76b01ae5f5a23315aefbad2c545b6355eba1c6b.jpg b/parse/dev/uKiE0VIluA-/images/438c5ccd20e31bd30576f642f76b01ae5f5a23315aefbad2c545b6355eba1c6b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ea2f7742137efc09cee05df1928ce35a2bdf02c0 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/438c5ccd20e31bd30576f642f76b01ae5f5a23315aefbad2c545b6355eba1c6b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07c5852b5f29445f31b8478ac4daa5e9d9afd69912fa01053264dd87a8a0b261 +size 4080 diff --git a/parse/dev/uKiE0VIluA-/images/46b02f8100fca73efdd7437649208bc5cdb78801bf5aa89432920d041f393050.jpg b/parse/dev/uKiE0VIluA-/images/46b02f8100fca73efdd7437649208bc5cdb78801bf5aa89432920d041f393050.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5145af62c65ecad7080dba55ffb0d562c82fcbb7 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/46b02f8100fca73efdd7437649208bc5cdb78801bf5aa89432920d041f393050.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a269f040f195e0fdbc95fc2c65119e65e02d27cf886ba13638ef6468656d992 +size 13473 diff --git a/parse/dev/uKiE0VIluA-/images/4ad4c97b85d236bb562674c2f22b8d431da33db42292cc1252fa4ba51665a475.jpg b/parse/dev/uKiE0VIluA-/images/4ad4c97b85d236bb562674c2f22b8d431da33db42292cc1252fa4ba51665a475.jpg new file mode 100644 index 0000000000000000000000000000000000000000..61cd3297fea154e329566141a0d002a0b92e951c --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/4ad4c97b85d236bb562674c2f22b8d431da33db42292cc1252fa4ba51665a475.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ad396fcb40f199ca61a990357bd679cc3bd67f39475d7849be41c6cdecf2684 +size 56344 diff --git a/parse/dev/uKiE0VIluA-/images/4db37699081dc07a244a83fc948efcd7aab2b0e6e69593d0850195661fb61a46.jpg b/parse/dev/uKiE0VIluA-/images/4db37699081dc07a244a83fc948efcd7aab2b0e6e69593d0850195661fb61a46.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fd2bae2b9aa19ba89c08753296ab262336b0d240 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/4db37699081dc07a244a83fc948efcd7aab2b0e6e69593d0850195661fb61a46.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13f393f79b1ce582f2288571a861497665585cc08138ec85d2164ad04e05d7f6 +size 12978 diff --git a/parse/dev/uKiE0VIluA-/images/5dbb21f1762af92ef31f5cfe4b8a11681e2156ddb5c3fbbb9b1dc2a089e912d9.jpg b/parse/dev/uKiE0VIluA-/images/5dbb21f1762af92ef31f5cfe4b8a11681e2156ddb5c3fbbb9b1dc2a089e912d9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e40aad8dea2c00ad620f0555c9f82f321df68c8a --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/5dbb21f1762af92ef31f5cfe4b8a11681e2156ddb5c3fbbb9b1dc2a089e912d9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:03b00e2de0aff680722e1415fd123a41f96e87ff98d70f493f1d5fba94cfea17 +size 7601 diff --git a/parse/dev/uKiE0VIluA-/images/60d38a1a70c21289ccec8eaaefea4aa28bdbb045d384a5bcfff0cbc7432d61ad.jpg b/parse/dev/uKiE0VIluA-/images/60d38a1a70c21289ccec8eaaefea4aa28bdbb045d384a5bcfff0cbc7432d61ad.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ca4e8ec9bf2778063d95bde544a5b664ad3a6414 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/60d38a1a70c21289ccec8eaaefea4aa28bdbb045d384a5bcfff0cbc7432d61ad.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:868740eb0ab34ed36034613d334a890b8e553210d52caec58a9382a1ed9fb4cc +size 8700 diff --git a/parse/dev/uKiE0VIluA-/images/61cd98e377972ec309ba0da19367131f4587e9a6373bb97c8912b48f47d0ae9b.jpg b/parse/dev/uKiE0VIluA-/images/61cd98e377972ec309ba0da19367131f4587e9a6373bb97c8912b48f47d0ae9b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..779a31e06c874d1d59935acb817414d0e7749fd6 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/61cd98e377972ec309ba0da19367131f4587e9a6373bb97c8912b48f47d0ae9b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9660199ea01a1cc5820cff5f10fa7b0abf8cccf5b4b9057c4756f99bd9362b57 +size 10859 diff --git a/parse/dev/uKiE0VIluA-/images/63519ae2d5008b3f18d7ddfe5ad6477c2279f125edf2fa198691f60c91cc42ff.jpg b/parse/dev/uKiE0VIluA-/images/63519ae2d5008b3f18d7ddfe5ad6477c2279f125edf2fa198691f60c91cc42ff.jpg new file mode 100644 index 0000000000000000000000000000000000000000..72996bb83ec99b1d14c11301a5918cb722379ad6 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/63519ae2d5008b3f18d7ddfe5ad6477c2279f125edf2fa198691f60c91cc42ff.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:11ff1720ae533a92574ce5ec6b29bc5b70bc1dcc654aca2fe9d33e11c8f39112 +size 9778 diff --git a/parse/dev/uKiE0VIluA-/images/65136910729af2c3b21887723d3b15e4d573cfcd3151c77bd4fc734d06ef3609.jpg b/parse/dev/uKiE0VIluA-/images/65136910729af2c3b21887723d3b15e4d573cfcd3151c77bd4fc734d06ef3609.jpg new file mode 100644 index 0000000000000000000000000000000000000000..33c6cf31bc8389599888e044922e7818b8975d71 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/65136910729af2c3b21887723d3b15e4d573cfcd3151c77bd4fc734d06ef3609.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e0a9262444854b0e43adb11f231d03bd25e533fc4983836ea61b14324228b97 +size 5417 diff --git a/parse/dev/uKiE0VIluA-/images/655579d8918eedd5740dde28380a52f290ea593857826f0ac468a86003c72bf2.jpg b/parse/dev/uKiE0VIluA-/images/655579d8918eedd5740dde28380a52f290ea593857826f0ac468a86003c72bf2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4f5b71529814c15582fadf82c860bf31dc64dac8 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/655579d8918eedd5740dde28380a52f290ea593857826f0ac468a86003c72bf2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c9a4a7323f2a48d5e6d956afd3b4ff284ca5868edf00df8e44f625804a33b94 +size 59820 diff --git a/parse/dev/uKiE0VIluA-/images/81451cd141501135c342a853fb5be92df45dfe3b2ccb1bd58acab05104a938e9.jpg b/parse/dev/uKiE0VIluA-/images/81451cd141501135c342a853fb5be92df45dfe3b2ccb1bd58acab05104a938e9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fb70354d2769182c82b446d4023869f545c704aa --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/81451cd141501135c342a853fb5be92df45dfe3b2ccb1bd58acab05104a938e9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c95a777ee919f9e1ed59b03cb48bdf1a0f3dd90949ca55cc3151808d7494e345 +size 17898 diff --git a/parse/dev/uKiE0VIluA-/images/860b8c91154f6c199c75bcff86df7eb9c23f1b08a560ba7c4171d88aeb6b3267.jpg b/parse/dev/uKiE0VIluA-/images/860b8c91154f6c199c75bcff86df7eb9c23f1b08a560ba7c4171d88aeb6b3267.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1ffa6fd3173f38a4b787f8fa13fa612fecfd500a --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/860b8c91154f6c199c75bcff86df7eb9c23f1b08a560ba7c4171d88aeb6b3267.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1986eed1a6e874a8dba3874cdff1c857222b142fc755d2b5015b6b24f70da86a +size 9154 diff --git a/parse/dev/uKiE0VIluA-/images/862e3ec3edf6f97cec9fb3c1bce2ab14f12cb517ae03f65e8c5e5a7c1b0128c3.jpg b/parse/dev/uKiE0VIluA-/images/862e3ec3edf6f97cec9fb3c1bce2ab14f12cb517ae03f65e8c5e5a7c1b0128c3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2b4e91bbcfa4061e59fdd3b0b14c089e4b3d071d --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/862e3ec3edf6f97cec9fb3c1bce2ab14f12cb517ae03f65e8c5e5a7c1b0128c3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7cb3eb48d141461087bad48162de9c78cf35936509f32bc1c5966f8e9dcc4062 +size 44528 diff --git a/parse/dev/uKiE0VIluA-/images/8635afab64698827a375347d41edaf19fec9830dd4c3370e611eae68eb2eec2d.jpg b/parse/dev/uKiE0VIluA-/images/8635afab64698827a375347d41edaf19fec9830dd4c3370e611eae68eb2eec2d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0cab3b9d7e864f787060e1a7c638c74708acbfc2 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/8635afab64698827a375347d41edaf19fec9830dd4c3370e611eae68eb2eec2d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b4581c91d74e7cd6b2a90c994ee9ba096037422185c71187d3552119db01d01 +size 4382 diff --git a/parse/dev/uKiE0VIluA-/images/90ec9ca92ef688e0299b0a61faf7dae77cf91b2517de17c8b0ddbaecc7b7e221.jpg b/parse/dev/uKiE0VIluA-/images/90ec9ca92ef688e0299b0a61faf7dae77cf91b2517de17c8b0ddbaecc7b7e221.jpg new file mode 100644 index 0000000000000000000000000000000000000000..189aade3fb6a7da1c86eba91869a4ae243b45adb --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/90ec9ca92ef688e0299b0a61faf7dae77cf91b2517de17c8b0ddbaecc7b7e221.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d62131ca1a7d8eac36cb2af5aedd3dac3b9c4850d1853b989214a19a97b9014 +size 3597 diff --git a/parse/dev/uKiE0VIluA-/images/9181374b21d9c1c59996b4bf381d3c99906da516aa9a04e9499658b9242eef57.jpg b/parse/dev/uKiE0VIluA-/images/9181374b21d9c1c59996b4bf381d3c99906da516aa9a04e9499658b9242eef57.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7c9eec8d3de66c0fdb8ebcde4525d94e1ec19e45 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/9181374b21d9c1c59996b4bf381d3c99906da516aa9a04e9499658b9242eef57.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:534a5f97fc473de05fa7e708ba1163484c03605ae085127979dcdf5f9552f33b +size 11546 diff --git a/parse/dev/uKiE0VIluA-/images/a0560fcfc93d28d385577be35b17936d964b15936df134b9157fc246495f9df9.jpg b/parse/dev/uKiE0VIluA-/images/a0560fcfc93d28d385577be35b17936d964b15936df134b9157fc246495f9df9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..037e3f317ec83a11946ec1f7374a6768ca31e431 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/a0560fcfc93d28d385577be35b17936d964b15936df134b9157fc246495f9df9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6abd5c84aa40aa0eda04380475ca4f0e27c4b329b2af53cd91a2a55ecfb509ab +size 3945 diff --git a/parse/dev/uKiE0VIluA-/images/a0df8a933cb13800488f84a51e2064d531323074a78c2c1978a5b8e93f4d5d2d.jpg b/parse/dev/uKiE0VIluA-/images/a0df8a933cb13800488f84a51e2064d531323074a78c2c1978a5b8e93f4d5d2d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..afaa054380072ca92731ed82bfc1d000892672f6 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/a0df8a933cb13800488f84a51e2064d531323074a78c2c1978a5b8e93f4d5d2d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:691e5178278c9b8cf696bd43b6e4fee072bb6804075be4e9309a59f054122c3e +size 7104 diff --git a/parse/dev/uKiE0VIluA-/images/a2854e0109b4abb7b98ee3f8bf08f076ec74452aaa065bc618d1908e4ab5cf3b.jpg b/parse/dev/uKiE0VIluA-/images/a2854e0109b4abb7b98ee3f8bf08f076ec74452aaa065bc618d1908e4ab5cf3b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..28881fc32b59a54cbd13f53d6b428d987b067f52 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/a2854e0109b4abb7b98ee3f8bf08f076ec74452aaa065bc618d1908e4ab5cf3b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f025cc68144ddf515a107aebfe6ac193520642401591d8c512a5d84c85bc49ea +size 57518 diff --git a/parse/dev/uKiE0VIluA-/images/abcd2a6a4d5ad44659342f34f14582507eb9cdcf9b14f316067a2c3f6bafb4f8.jpg b/parse/dev/uKiE0VIluA-/images/abcd2a6a4d5ad44659342f34f14582507eb9cdcf9b14f316067a2c3f6bafb4f8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..145e5a53f38456e643caf3cfb63d209e806dcf2c --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/abcd2a6a4d5ad44659342f34f14582507eb9cdcf9b14f316067a2c3f6bafb4f8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44139fb8d78185d79a72b072bb892ea4d5d2ab3db90f132b6d76472d537bfed5 +size 8307 diff --git a/parse/dev/uKiE0VIluA-/images/abe7676f25966510f9ae1c31c17c6b982a506034b914b91bf8fd5982856a0f50.jpg b/parse/dev/uKiE0VIluA-/images/abe7676f25966510f9ae1c31c17c6b982a506034b914b91bf8fd5982856a0f50.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5b683ae769019b1b37081b206008778be6103ecd --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/abe7676f25966510f9ae1c31c17c6b982a506034b914b91bf8fd5982856a0f50.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e65ce4ff4cc1b5e75585f36426b022988acf3238339bd4302e20249dcfc19036 +size 78888 diff --git a/parse/dev/uKiE0VIluA-/images/adca1ae1d8aab4e8478b70f53410ac27a5fe57c198eab572493f1b5b7fb78652.jpg b/parse/dev/uKiE0VIluA-/images/adca1ae1d8aab4e8478b70f53410ac27a5fe57c198eab572493f1b5b7fb78652.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ed2d3ffcd5ec92689fc829c02cac21c75e6aec9a --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/adca1ae1d8aab4e8478b70f53410ac27a5fe57c198eab572493f1b5b7fb78652.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:341a0d444263741375c04dde2199edfa3c530d0dab6e2cb018bfa291be2b4e5f +size 8528 diff --git a/parse/dev/uKiE0VIluA-/images/b1f5874fdc173511c7c0ecd6dc917e77b83c43f2ffb05173196382a662c81bfd.jpg b/parse/dev/uKiE0VIluA-/images/b1f5874fdc173511c7c0ecd6dc917e77b83c43f2ffb05173196382a662c81bfd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cba0ad59c9b02d00563c104ebbe3a54485a9411e --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/b1f5874fdc173511c7c0ecd6dc917e77b83c43f2ffb05173196382a662c81bfd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8734ab7dcaad6e8a443b19d509b8e155015c7c01c548a35b331fdd53277d5d2d +size 10408 diff --git a/parse/dev/uKiE0VIluA-/images/b8460a204b5d2e8b5d0b92d127966e09dc7212c3dfac7c343aa3507771e7c765.jpg b/parse/dev/uKiE0VIluA-/images/b8460a204b5d2e8b5d0b92d127966e09dc7212c3dfac7c343aa3507771e7c765.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dabed9f4d89c1c6076eb7b6ffff88041bd2e5430 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/b8460a204b5d2e8b5d0b92d127966e09dc7212c3dfac7c343aa3507771e7c765.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20f0588482d68c1953a6bbf46dd2164b233359720773e48ffe61d4cafa1f1668 +size 6435 diff --git a/parse/dev/uKiE0VIluA-/images/bb1f4eb8452d13b0e61b73d6acf1f9e790634ea9e60508148710a86b85fd98c2.jpg b/parse/dev/uKiE0VIluA-/images/bb1f4eb8452d13b0e61b73d6acf1f9e790634ea9e60508148710a86b85fd98c2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f5c1dee5228d945dce593c2540abf6b40e041bd1 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/bb1f4eb8452d13b0e61b73d6acf1f9e790634ea9e60508148710a86b85fd98c2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14b944a311a53dc3aa3fe67a36dd216df5c829d263fd1e36b99e181acba05ab1 +size 5979 diff --git a/parse/dev/uKiE0VIluA-/images/be436fcf0bafc0f907ee7a6c7d7e0536f891d28b81f369293938758f8c0ab65d.jpg b/parse/dev/uKiE0VIluA-/images/be436fcf0bafc0f907ee7a6c7d7e0536f891d28b81f369293938758f8c0ab65d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..40cc02fcf82b138d22cc7cf8048e9d31751c7a36 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/be436fcf0bafc0f907ee7a6c7d7e0536f891d28b81f369293938758f8c0ab65d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26cd5939b9cf7c17021d9a38094c7d13ce6f637eccfae90a7057b472f2e382ba +size 5336 diff --git a/parse/dev/uKiE0VIluA-/images/c4a3682e5956cace17fe93f6004fa05e961d98ff23bf3e056505db2ff8ac4e7d.jpg b/parse/dev/uKiE0VIluA-/images/c4a3682e5956cace17fe93f6004fa05e961d98ff23bf3e056505db2ff8ac4e7d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..222baef2673fde70be3a48b004914d3769923c71 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/c4a3682e5956cace17fe93f6004fa05e961d98ff23bf3e056505db2ff8ac4e7d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79c84f3b76b315daa471fd84763863581164a5a45cb3cc57011333330810037a +size 8317 diff --git a/parse/dev/uKiE0VIluA-/images/cb05020a6bcb8aa060d3a58bdb891c42ab9078eb32ece7d52ec44ccdc99b38ec.jpg b/parse/dev/uKiE0VIluA-/images/cb05020a6bcb8aa060d3a58bdb891c42ab9078eb32ece7d52ec44ccdc99b38ec.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5e647b0f5b6a842074bc962c84049ebd4fd7bc5d --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/cb05020a6bcb8aa060d3a58bdb891c42ab9078eb32ece7d52ec44ccdc99b38ec.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fa50209ff5b98b9a64a69228c69b721ca5a698626b175600971041b0dc3c1fde +size 11195 diff --git a/parse/dev/uKiE0VIluA-/images/cbe24c27e05abbeba1ead3a84d2b76c8c8a74b7f66ffec42da0d279996e6e840.jpg b/parse/dev/uKiE0VIluA-/images/cbe24c27e05abbeba1ead3a84d2b76c8c8a74b7f66ffec42da0d279996e6e840.jpg new file mode 100644 index 0000000000000000000000000000000000000000..799a1439454a8741284344f2bec2d428c37e55a2 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/cbe24c27e05abbeba1ead3a84d2b76c8c8a74b7f66ffec42da0d279996e6e840.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e7527beb7e7e0ec558c655cd33e49444cf8546d89a49adfa6b3457c3e7b7af1b +size 11960 diff --git a/parse/dev/uKiE0VIluA-/images/cfe557448da84f09bbadb359b75eedb28f00f014aa4c99dcca35fe416b5d5b3b.jpg b/parse/dev/uKiE0VIluA-/images/cfe557448da84f09bbadb359b75eedb28f00f014aa4c99dcca35fe416b5d5b3b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a607be90ec740a9edbb96615caf5b42ac7c9ad9d --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/cfe557448da84f09bbadb359b75eedb28f00f014aa4c99dcca35fe416b5d5b3b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6c274eaad8205f6d7be8b8fc78e93830de3926a222a41d2356c9e791af28e98 +size 6149 diff --git a/parse/dev/uKiE0VIluA-/images/d4937db94ffa95314385e9594028a2c749d990a3337f16352d0dfebb7a3f16a3.jpg b/parse/dev/uKiE0VIluA-/images/d4937db94ffa95314385e9594028a2c749d990a3337f16352d0dfebb7a3f16a3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c043a1664b36f8afb567ca4c580cefe2611e85ea --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/d4937db94ffa95314385e9594028a2c749d990a3337f16352d0dfebb7a3f16a3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d00ecb77ef585df0d988626b663cfd01f2197daff2280ad43473a12885b6d299 +size 19037 diff --git a/parse/dev/uKiE0VIluA-/images/d7fb78da63b80a51cd9fc87f35d7928a183727c2b4018bb55f9d89dd68da5ede.jpg b/parse/dev/uKiE0VIluA-/images/d7fb78da63b80a51cd9fc87f35d7928a183727c2b4018bb55f9d89dd68da5ede.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a419c2a6e1aee0554d0f7a24c4f535393bbd4e30 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/d7fb78da63b80a51cd9fc87f35d7928a183727c2b4018bb55f9d89dd68da5ede.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e09e55c348f40c000a82540c32bd1269d60890adb8939c012618cdc14c3aef11 +size 93341 diff --git a/parse/dev/uKiE0VIluA-/images/dbf308461216717c536237b145b80e4fe33da73608dfc6ff50d5dcbd2e71b481.jpg b/parse/dev/uKiE0VIluA-/images/dbf308461216717c536237b145b80e4fe33da73608dfc6ff50d5dcbd2e71b481.jpg new file mode 100644 index 0000000000000000000000000000000000000000..84d0bdf68f73ff62e0e99a72928b4dce56c78983 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/dbf308461216717c536237b145b80e4fe33da73608dfc6ff50d5dcbd2e71b481.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca9d70f480b6221339051244c944daf95e66eeb723b1878ea560e38a9769eceb +size 8393 diff --git a/parse/dev/uKiE0VIluA-/images/e0522814d19514cc5c270f107d8b02f1f07a9cb874e40fb53ca22dae952cfd9e.jpg b/parse/dev/uKiE0VIluA-/images/e0522814d19514cc5c270f107d8b02f1f07a9cb874e40fb53ca22dae952cfd9e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..22eae73702fc3ba12b1509fc44b9255a4f93a269 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/e0522814d19514cc5c270f107d8b02f1f07a9cb874e40fb53ca22dae952cfd9e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0bcf216304417da56b3fa09ef24c0c0b12145830d7cd6ad4707ed4f538753d39 +size 13130 diff --git a/parse/dev/uKiE0VIluA-/images/e3ec0992f98540f483d4a0e9f25349e01e88a2320d97af9e79546fbd01b72b58.jpg b/parse/dev/uKiE0VIluA-/images/e3ec0992f98540f483d4a0e9f25349e01e88a2320d97af9e79546fbd01b72b58.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5bdd5b1da8766508443b7aa7b40cae77249666b1 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/e3ec0992f98540f483d4a0e9f25349e01e88a2320d97af9e79546fbd01b72b58.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:912b12ac09c8be4d64cc8ebaebde7d7cc7496dcb0fade58fac3261181846298f +size 40450 diff --git a/parse/dev/uKiE0VIluA-/images/eb9a96d049e36e29731e8e8da239aa4525e7568f6c9b8873877942c15e65a43b.jpg b/parse/dev/uKiE0VIluA-/images/eb9a96d049e36e29731e8e8da239aa4525e7568f6c9b8873877942c15e65a43b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f0c8c03fbd1080689f3a3e361b5382af233bb21c --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/eb9a96d049e36e29731e8e8da239aa4525e7568f6c9b8873877942c15e65a43b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:18f3c2aa3e410eb206fc71cc287f43c7259112747402c511be711f07fcbe0930 +size 10558 diff --git a/parse/dev/uKiE0VIluA-/images/f343c0bad91cc2a7e2b6a155fdd54a25018c26c8bef57be83633e247701befec.jpg b/parse/dev/uKiE0VIluA-/images/f343c0bad91cc2a7e2b6a155fdd54a25018c26c8bef57be83633e247701befec.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7b1a8841dc3644f8cb6de71fb45308680ff46157 --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/f343c0bad91cc2a7e2b6a155fdd54a25018c26c8bef57be83633e247701befec.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:048e450f41c0076547c06a5b4ea160a715a5187019681b5022e428912b102a59 +size 41940 diff --git a/parse/dev/uKiE0VIluA-/images/fbe2cc2cd585fb1d9e9ab8f6d7d54b6c4ae9962a7acb1a075e85ad1bf04c71bc.jpg b/parse/dev/uKiE0VIluA-/images/fbe2cc2cd585fb1d9e9ab8f6d7d54b6c4ae9962a7acb1a075e85ad1bf04c71bc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f303797cea81f5400c3f92f32a1e28f6d4e3911a --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/fbe2cc2cd585fb1d9e9ab8f6d7d54b6c4ae9962a7acb1a075e85ad1bf04c71bc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:108cfbd048606dec6ff2cd937f0120696d6ac800927e53ab3082e08b4e1af817 +size 3654 diff --git a/parse/dev/uKiE0VIluA-/images/fddbab7ca60418fca249ea92f139d852298a4ac20ee3022b062f161f12f221b9.jpg b/parse/dev/uKiE0VIluA-/images/fddbab7ca60418fca249ea92f139d852298a4ac20ee3022b062f161f12f221b9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2a20399bc2557becea38a6b710d0f85e2013c8fb --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/fddbab7ca60418fca249ea92f139d852298a4ac20ee3022b062f161f12f221b9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bc5d273dc41f0f3d9032b3f4efb5f5949c89002c44b66eeb64a1d37238d04282 +size 5924 diff --git a/parse/dev/uKiE0VIluA-/images/ff84a241b204a9faf7774602b42e409bd7c2e0dc1aec414adc96862201ad6e06.jpg b/parse/dev/uKiE0VIluA-/images/ff84a241b204a9faf7774602b42e409bd7c2e0dc1aec414adc96862201ad6e06.jpg new file mode 100644 index 0000000000000000000000000000000000000000..930977f170c5f0d7d992853b6e8218b171fc908a --- /dev/null +++ b/parse/dev/uKiE0VIluA-/images/ff84a241b204a9faf7774602b42e409bd7c2e0dc1aec414adc96862201ad6e06.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82d5ada201d577d607d43cc120124bd5c618e8c1d8e1ce0df4bd824f9554cdb5 +size 5903 diff --git a/parse/dev/um2BxfgkT2_/um2BxfgkT2__model.json b/parse/dev/um2BxfgkT2_/um2BxfgkT2__model.json new file mode 100644 index 0000000000000000000000000000000000000000..77d89a54d91603777a9443044eeb89f9c5afab65 --- /dev/null +++ b/parse/dev/um2BxfgkT2_/um2BxfgkT2__model.json @@ -0,0 +1,24417 @@ +[ + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1362, + 1405, + 1362, + 1405, + 1697, + 298, + 1697 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 299, + 1133, + 1404, + 1133, + 1404, + 1347, + 299, + 1347 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 299, + 1710, + 1403, + 1710, + 1403, + 1894, + 299, + 1894 + ], + "score": 0.971 + }, + { + "category_id": 1, + "poly": [ + 397, + 636, + 1304, + 636, + 1304, + 1002, + 397, + 1002 + ], + "score": 0.964 + }, + { + "category_id": 1, + "poly": [ + 434, + 421, + 1263, + 421, + 1263, + 526, + 434, + 526 + ], + "score": 0.963 + }, + { + "category_id": 2, + "poly": [ + 331, + 1919, + 861, + 1919, + 861, + 1979, + 331, + 1979 + ], + "score": 0.935 + }, + { + "category_id": 0, + "poly": [ + 340, + 271, + 1357, + 271, + 1357, + 327, + 340, + 327 + ], + "score": 0.914 + }, + { + "category_id": 0, + "poly": [ + 299, + 1062, + 530, + 1062, + 530, + 1101, + 299, + 1101 + ], + "score": 0.91 + }, + { + "category_id": 2, + "poly": [ + 297, + 2033, + 1071, + 2033, + 1071, + 2062, + 297, + 2062 + ], + "score": 0.895 + }, + { + "category_id": 0, + "poly": [ + 787, + 580, + 913, + 580, + 913, + 617, + 787, + 617 + ], + "score": 0.887 + }, + { + "category_id": 13, + "poly": [ + 525, + 424, + 608, + 424, + 608, + 456, + 525, + 456 + ], + "score": 0.61, + "latex": "{ \\bf K i m ^ { 1 * } }" + }, + { + "category_id": 13, + "poly": [ + 625, + 492, + 681, + 492, + 681, + 524, + 625, + 524 + ], + "score": 0.48, + "latex": "^ 2 \\mathrm { L G }" + }, + { + "category_id": 13, + "poly": [ + 994, + 423, + 1061, + 423, + 1061, + 456, + 994, + 456 + ], + "score": 0.26, + "latex": "\\mathbf { M } \\mathbf { i n } ^ { 2 }" + }, + { + "category_id": 15, + "poly": [ + 328.0, + 1913.0, + 864.0, + 1913.0, + 864.0, + 1953.0, + 328.0, + 1953.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 331.0, + 1944.0, + 643.0, + 1944.0, + 643.0, + 1986.0, + 331.0, + 1986.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 337.0, + 265.0, + 1363.0, + 265.0, + 1363.0, + 338.0, + 337.0, + 338.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1059.0, + 536.0, + 1059.0, + 536.0, + 1107.0, + 292.0, + 1107.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 2030.0, + 1071.0, + 2030.0, + 1071.0, + 2066.0, + 295.0, + 2066.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 783.0, + 580.0, + 918.0, + 580.0, + 918.0, + 619.0, + 783.0, + 619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1360.0, + 1408.0, + 1360.0, + 1408.0, + 1397.0, + 292.0, + 1397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1391.0, + 1405.0, + 1391.0, + 1405.0, + 1428.0, + 293.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1421.0, + 1405.0, + 1421.0, + 1405.0, + 1458.0, + 293.0, + 1458.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1453.0, + 1405.0, + 1453.0, + 1405.0, + 1487.0, + 295.0, + 1487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1479.0, + 1406.0, + 1479.0, + 1406.0, + 1522.0, + 292.0, + 1522.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1512.0, + 1408.0, + 1512.0, + 1408.0, + 1551.0, + 293.0, + 1551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1543.0, + 1407.0, + 1543.0, + 1407.0, + 1578.0, + 295.0, + 1578.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1575.0, + 1405.0, + 1575.0, + 1405.0, + 1610.0, + 295.0, + 1610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1600.0, + 1406.0, + 1600.0, + 1406.0, + 1642.0, + 292.0, + 1642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1633.0, + 1405.0, + 1633.0, + 1405.0, + 1672.0, + 293.0, + 1672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1667.0, + 1388.0, + 1667.0, + 1388.0, + 1701.0, + 296.0, + 1701.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1135.0, + 1404.0, + 1135.0, + 1404.0, + 1169.0, + 295.0, + 1169.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1166.0, + 1408.0, + 1166.0, + 1408.0, + 1201.0, + 295.0, + 1201.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1196.0, + 1405.0, + 1196.0, + 1405.0, + 1230.0, + 295.0, + 1230.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1224.0, + 1406.0, + 1224.0, + 1406.0, + 1263.0, + 293.0, + 1263.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1256.0, + 1405.0, + 1256.0, + 1405.0, + 1288.0, + 293.0, + 1288.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1285.0, + 1405.0, + 1285.0, + 1405.0, + 1323.0, + 293.0, + 1323.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1316.0, + 1394.0, + 1316.0, + 1394.0, + 1352.0, + 293.0, + 1352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1708.0, + 1407.0, + 1708.0, + 1407.0, + 1745.0, + 294.0, + 1745.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1739.0, + 1405.0, + 1739.0, + 1405.0, + 1775.0, + 293.0, + 1775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1769.0, + 1405.0, + 1769.0, + 1405.0, + 1807.0, + 293.0, + 1807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1801.0, + 1405.0, + 1801.0, + 1405.0, + 1835.0, + 293.0, + 1835.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1831.0, + 1404.0, + 1831.0, + 1404.0, + 1865.0, + 293.0, + 1865.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1862.0, + 1069.0, + 1862.0, + 1069.0, + 1900.0, + 293.0, + 1900.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 396.0, + 638.0, + 1305.0, + 638.0, + 1305.0, + 670.0, + 396.0, + 670.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 668.0, + 1305.0, + 668.0, + 1305.0, + 704.0, + 393.0, + 704.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 699.0, + 1304.0, + 699.0, + 1304.0, + 733.0, + 392.0, + 733.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 394.0, + 729.0, + 1305.0, + 729.0, + 1305.0, + 764.0, + 394.0, + 764.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 758.0, + 1306.0, + 758.0, + 1306.0, + 795.0, + 393.0, + 795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 789.0, + 1307.0, + 789.0, + 1307.0, + 824.0, + 393.0, + 824.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 819.0, + 1309.0, + 819.0, + 1309.0, + 854.0, + 392.0, + 854.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 849.0, + 1307.0, + 849.0, + 1307.0, + 885.0, + 393.0, + 885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 879.0, + 1304.0, + 879.0, + 1304.0, + 914.0, + 393.0, + 914.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 913.0, + 1305.0, + 913.0, + 1305.0, + 945.0, + 395.0, + 945.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 395.0, + 940.0, + 1304.0, + 940.0, + 1304.0, + 975.0, + 395.0, + 975.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 393.0, + 970.0, + 1023.0, + 970.0, + 1023.0, + 1007.0, + 393.0, + 1007.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 431.0, + 420.0, + 524.0, + 420.0, + 524.0, + 461.0, + 431.0, + 461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 609.0, + 420.0, + 993.0, + 420.0, + 993.0, + 461.0, + 609.0, + 461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1062.0, + 420.0, + 1266.0, + 420.0, + 1266.0, + 461.0, + 1062.0, + 461.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 513.0, + 456.0, + 712.0, + 456.0, + 712.0, + 487.0, + 513.0, + 487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 727.0, + 451.0, + 1190.0, + 451.0, + 1190.0, + 493.0, + 727.0, + 493.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 521.0, + 485.0, + 624.0, + 485.0, + 624.0, + 533.0, + 521.0, + 533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 682.0, + 485.0, + 1175.0, + 485.0, + 1175.0, + 533.0, + 682.0, + 533.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 0, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1205, + 1405, + 1205, + 1405, + 1481, + 297, + 1481 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 856, + 1405, + 856, + 1405, + 1192, + 298, + 1192 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 1727, + 1404, + 1727, + 1404, + 2008, + 297, + 2008 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 297, + 1591, + 1404, + 1591, + 1404, + 1714, + 297, + 1714 + ], + "score": 0.975 + }, + { + "category_id": 4, + "poly": [ + 297, + 678, + 1406, + 678, + 1406, + 803, + 297, + 803 + ], + "score": 0.961 + }, + { + "category_id": 0, + "poly": [ + 296, + 1522, + 954, + 1522, + 954, + 1562, + 296, + 1562 + ], + "score": 0.938 + }, + { + "category_id": 2, + "poly": [ + 629, + 200, + 1075, + 200, + 1075, + 223, + 629, + 223 + ], + "score": 0.832 + }, + { + "category_id": 3, + "poly": [ + 312, + 238, + 1378, + 238, + 1378, + 656, + 312, + 656 + ], + "score": 0.803 + }, + { + "category_id": 3, + "poly": [ + 312, + 236, + 1381, + 236, + 1381, + 657, + 312, + 657 + ], + "score": 0.801 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 858, + 2061, + 858, + 2085, + 841, + 2085 + ], + "score": 0.716 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.169 + }, + { + "category_id": 13, + "poly": [ + 296, + 1819, + 631, + 1819, + 631, + 1857, + 296, + 1857 + ], + "score": 0.92, + "latex": "\\mathbf { X } = [ \\bar { \\mathbf { X } } ^ { \\nu } ; \\mathbf { X } ^ { \\varepsilon } ] \\in \\mathbb { R } ^ { ( n + m ) \\times C }" + }, + { + "category_id": 13, + "poly": [ + 344, + 1729, + 477, + 1729, + 477, + 1762, + 344, + 1762 + ], + "score": 0.92, + "latex": "\\mathcal { G } = ( \\nu , \\mathcal { E } )" + }, + { + "category_id": 13, + "poly": [ + 631, + 1758, + 786, + 1758, + 786, + 1788, + 631, + 1788 + ], + "score": 0.92, + "latex": "\\mathbf { X } ^ { \\nu } \\in \\mathbb { R } ^ { n \\times C }" + }, + { + "category_id": 13, + "poly": [ + 762, + 1790, + 862, + 1790, + 862, + 1823, + 762, + 1823 + ], + "score": 0.92, + "latex": "( n + m )" + }, + { + "category_id": 13, + "poly": [ + 1171, + 1729, + 1405, + 1729, + 1405, + 1763, + 1171, + 1763 + ], + "score": 0.91, + "latex": "\\mathcal { E } = \\{ e _ { 1 } , . . . , e _ { m } \\} \\subseteq" + }, + { + "category_id": 13, + "poly": [ + 297, + 1758, + 331, + 1758, + 331, + 1788, + 297, + 1788 + ], + "score": 0.85, + "latex": "\\mathcal { V } ^ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 844, + 1759, + 1005, + 1759, + 1005, + 1788, + 844, + 1788 + ], + "score": 0.82, + "latex": "\\mathbf { X } ^ { \\varepsilon } \\in \\mathbb { R } ^ { m \\times C }" + }, + { + "category_id": 13, + "poly": [ + 564, + 1162, + 582, + 1162, + 582, + 1188, + 564, + 1188 + ], + "score": 0.81, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 420, + 1132, + 438, + 1132, + 438, + 1158, + 420, + 1158 + ], + "score": 0.8, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1015, + 1132, + 1034, + 1132, + 1034, + 1158, + 1015, + 1158 + ], + "score": 0.8, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 816, + 1729, + 1010, + 1729, + 1010, + 1760, + 816, + 1760 + ], + "score": 0.8, + "latex": "\\mathcal { V } = \\{ v _ { 1 } , . . . , v _ { n } \\}" + }, + { + "category_id": 13, + "poly": [ + 716, + 1735, + 737, + 1735, + 737, + 1757, + 716, + 1757 + ], + "score": 0.76, + "latex": "n" + }, + { + "category_id": 13, + "poly": [ + 852, + 1162, + 870, + 1162, + 870, + 1188, + 852, + 1188 + ], + "score": 0.7, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1065, + 1734, + 1094, + 1734, + 1094, + 1758, + 1065, + 1758 + ], + "score": 0.68, + "latex": "m" + }, + { + "category_id": 13, + "poly": [ + 847, + 1886, + 874, + 1886, + 874, + 1912, + 847, + 1912 + ], + "score": 0.63, + "latex": "\\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 1374, + 1824, + 1403, + 1824, + 1403, + 1852, + 1374, + 1852 + ], + "score": 0.41, + "latex": "\\mathbf { X }" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 680.0, + 1404.0, + 680.0, + 1404.0, + 712.0, + 296.0, + 712.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 709.0, + 1406.0, + 709.0, + 1406.0, + 744.0, + 292.0, + 744.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 739.0, + 1407.0, + 739.0, + 1407.0, + 778.0, + 291.0, + 778.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 769.0, + 1188.0, + 769.0, + 1188.0, + 806.0, + 292.0, + 806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1519.0, + 955.0, + 1519.0, + 955.0, + 1569.0, + 290.0, + 1569.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 625.0, + 192.0, + 1078.0, + 192.0, + 1078.0, + 229.0, + 625.0, + 229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1268.0, + 229.0, + 1380.0, + 229.0, + 1380.0, + 268.0, + 1268.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1025.0, + 331.0, + 1228.0, + 331.0, + 1228.0, + 366.0, + 1025.0, + 366.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 623.0, + 437.0, + 683.0, + 437.0, + 683.0, + 474.0, + 623.0, + 474.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 733.0, + 435.0, + 765.0, + 435.0, + 765.0, + 463.0, + 733.0, + 463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 781.0, + 435.0, + 812.0, + 435.0, + 812.0, + 463.0, + 781.0, + 463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 914.0, + 437.0, + 943.0, + 437.0, + 943.0, + 462.0, + 914.0, + 462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 960.0, + 436.0, + 994.0, + 436.0, + 994.0, + 464.0, + 960.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1014.0, + 438.0, + 1042.0, + 438.0, + 1042.0, + 462.0, + 1014.0, + 462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1060.0, + 436.0, + 1094.0, + 436.0, + 1094.0, + 464.0, + 1060.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1110.0, + 436.0, + 1141.0, + 436.0, + 1141.0, + 464.0, + 1110.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1160.0, + 436.0, + 1192.0, + 436.0, + 1192.0, + 464.0, + 1160.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1212.0, + 437.0, + 1240.0, + 437.0, + 1240.0, + 463.0, + 1212.0, + 463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1259.0, + 435.0, + 1291.0, + 435.0, + 1291.0, + 464.0, + 1259.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 600.0, + 459.0, + 704.0, + 459.0, + 704.0, + 491.0, + 600.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 720.0, + 461.0, + 824.0, + 461.0, + 824.0, + 493.0, + 720.0, + 493.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 390.0, + 507.0, + 424.0, + 507.0, + 424.0, + 557.0, + 390.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 530.0, + 507.0, + 564.0, + 507.0, + 564.0, + 558.0, + 530.0, + 558.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 599.0, + 498.0, + 841.0, + 498.0, + 841.0, + 563.0, + 599.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 922.0, + 519.0, + 934.0, + 519.0, + 934.0, + 530.0, + 922.0, + 530.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1017.0, + 491.0, + 1038.0, + 491.0, + 1038.0, + 535.0, + 1017.0, + 535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1067.0, + 491.0, + 1088.0, + 491.0, + 1088.0, + 537.0, + 1067.0, + 537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1115.0, + 488.0, + 1139.0, + 488.0, + 1139.0, + 539.0, + 1115.0, + 539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1164.0, + 489.0, + 1190.0, + 489.0, + 1190.0, + 535.0, + 1164.0, + 535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1212.0, + 487.0, + 1241.0, + 487.0, + 1241.0, + 542.0, + 1212.0, + 542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1262.0, + 488.0, + 1288.0, + 488.0, + 1288.0, + 540.0, + 1262.0, + 540.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 319.0, + 581.0, + 345.0, + 581.0, + 345.0, + 601.0, + 319.0, + 601.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 459.0, + 580.0, + 490.0, + 580.0, + 490.0, + 603.0, + 459.0, + 603.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1163.0, + 585.0, + 1173.0, + 585.0, + 1173.0, + 594.0, + 1163.0, + 594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 326.0, + 608.0, + 342.0, + 608.0, + 342.0, + 625.0, + 326.0, + 625.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 991.0, + 599.0, + 1217.0, + 599.0, + 1217.0, + 634.0, + 991.0, + 634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1287.0, + 600.0, + 1362.0, + 600.0, + 1362.0, + 639.0, + 1287.0, + 639.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 375.0, + 625.0, + 502.0, + 625.0, + 502.0, + 661.0, + 375.0, + 661.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 969.0, + 622.0, + 1239.0, + 622.0, + 1239.0, + 663.0, + 969.0, + 663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1170.75, + 567.5, + 1194.75, + 567.5, + 1194.75, + 577.5, + 1170.75, + 577.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1269.0, + 232.0, + 1379.0, + 232.0, + 1379.0, + 265.0, + 1269.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1026.0, + 333.0, + 1228.0, + 333.0, + 1228.0, + 365.0, + 1026.0, + 365.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 621.0, + 440.0, + 682.0, + 440.0, + 682.0, + 470.0, + 621.0, + 470.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 737.0, + 439.0, + 759.0, + 439.0, + 759.0, + 458.0, + 737.0, + 458.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 782.0, + 436.0, + 810.0, + 436.0, + 810.0, + 463.0, + 782.0, + 463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 912.0, + 435.0, + 945.0, + 435.0, + 945.0, + 464.0, + 912.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 960.0, + 435.0, + 995.0, + 435.0, + 995.0, + 464.0, + 960.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1010.0, + 436.0, + 1044.0, + 436.0, + 1044.0, + 464.0, + 1010.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1058.0, + 435.0, + 1094.0, + 435.0, + 1094.0, + 464.0, + 1058.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1110.0, + 435.0, + 1142.0, + 435.0, + 1142.0, + 464.0, + 1110.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1159.0, + 435.0, + 1194.0, + 435.0, + 1194.0, + 464.0, + 1159.0, + 464.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1217.0, + 439.0, + 1236.0, + 439.0, + 1236.0, + 459.0, + 1217.0, + 459.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1262.0, + 436.0, + 1290.0, + 436.0, + 1290.0, + 463.0, + 1262.0, + 463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 599.0, + 460.0, + 704.0, + 460.0, + 704.0, + 491.0, + 599.0, + 491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 718.0, + 459.0, + 825.0, + 459.0, + 825.0, + 494.0, + 718.0, + 494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 391.0, + 508.0, + 422.0, + 508.0, + 422.0, + 557.0, + 391.0, + 557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 531.0, + 508.0, + 564.0, + 508.0, + 564.0, + 555.0, + 531.0, + 555.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 599.0, + 499.0, + 839.0, + 499.0, + 839.0, + 563.0, + 599.0, + 563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 920.0, + 493.0, + 939.0, + 493.0, + 939.0, + 535.0, + 920.0, + 535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 969.0, + 518.0, + 987.0, + 518.0, + 987.0, + 534.0, + 969.0, + 534.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1017.0, + 492.0, + 1039.0, + 492.0, + 1039.0, + 537.0, + 1017.0, + 537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1064.0, + 490.0, + 1090.0, + 490.0, + 1090.0, + 539.0, + 1064.0, + 539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1113.0, + 489.0, + 1140.0, + 489.0, + 1140.0, + 539.0, + 1113.0, + 539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1164.0, + 489.0, + 1189.0, + 489.0, + 1189.0, + 537.0, + 1164.0, + 537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1213.0, + 489.0, + 1237.0, + 489.0, + 1237.0, + 539.0, + 1213.0, + 539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1265.0, + 492.0, + 1285.0, + 492.0, + 1285.0, + 537.0, + 1265.0, + 537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1063.0, + 565.0, + 1073.0, + 565.0, + 1073.0, + 578.0, + 1063.0, + 578.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1180.0, + 567.0, + 1190.0, + 567.0, + 1190.0, + 578.0, + 1180.0, + 578.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 318.0, + 579.0, + 346.0, + 579.0, + 346.0, + 603.0, + 318.0, + 603.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 458.0, + 581.0, + 491.0, + 581.0, + 491.0, + 603.0, + 458.0, + 603.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 878.0, + 577.0, + 893.0, + 577.0, + 893.0, + 587.0, + 878.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 324.0, + 604.0, + 345.0, + 604.0, + 345.0, + 627.0, + 324.0, + 627.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 992.0, + 597.0, + 1217.0, + 597.0, + 1217.0, + 634.0, + 992.0, + 634.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1287.0, + 599.0, + 1364.0, + 599.0, + 1364.0, + 638.0, + 1287.0, + 638.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 375.0, + 624.0, + 503.0, + 624.0, + 503.0, + 661.0, + 375.0, + 661.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 967.0, + 623.0, + 1240.0, + 623.0, + 1240.0, + 662.0, + 967.0, + 662.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2093.0, + 839.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2093.0, + 838.0, + 2093.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1205.0, + 1405.0, + 1205.0, + 1405.0, + 1240.0, + 294.0, + 1240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1236.0, + 1407.0, + 1236.0, + 1407.0, + 1273.0, + 295.0, + 1273.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1265.0, + 1405.0, + 1265.0, + 1405.0, + 1303.0, + 294.0, + 1303.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1296.0, + 1406.0, + 1296.0, + 1406.0, + 1333.0, + 294.0, + 1333.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1327.0, + 1408.0, + 1327.0, + 1408.0, + 1364.0, + 294.0, + 1364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1355.0, + 1407.0, + 1355.0, + 1407.0, + 1393.0, + 294.0, + 1393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1386.0, + 1406.0, + 1386.0, + 1406.0, + 1424.0, + 292.0, + 1424.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1417.0, + 1406.0, + 1417.0, + 1406.0, + 1453.0, + 294.0, + 1453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1449.0, + 1133.0, + 1449.0, + 1133.0, + 1485.0, + 295.0, + 1485.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 856.0, + 1406.0, + 856.0, + 1406.0, + 893.0, + 293.0, + 893.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 887.0, + 1408.0, + 887.0, + 1408.0, + 925.0, + 293.0, + 925.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 919.0, + 1406.0, + 919.0, + 1406.0, + 953.0, + 295.0, + 953.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 950.0, + 1403.0, + 950.0, + 1403.0, + 984.0, + 295.0, + 984.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 977.0, + 1407.0, + 977.0, + 1407.0, + 1014.0, + 293.0, + 1014.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1010.0, + 1406.0, + 1010.0, + 1406.0, + 1045.0, + 295.0, + 1045.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1039.0, + 1407.0, + 1039.0, + 1407.0, + 1076.0, + 292.0, + 1076.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1070.0, + 1406.0, + 1070.0, + 1406.0, + 1105.0, + 295.0, + 1105.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1101.0, + 1405.0, + 1101.0, + 1405.0, + 1135.0, + 295.0, + 1135.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1129.0, + 419.0, + 1129.0, + 419.0, + 1166.0, + 293.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 439.0, + 1129.0, + 1014.0, + 1129.0, + 1014.0, + 1166.0, + 439.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1035.0, + 1129.0, + 1405.0, + 1129.0, + 1405.0, + 1166.0, + 1035.0, + 1166.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1160.0, + 563.0, + 1160.0, + 563.0, + 1195.0, + 295.0, + 1195.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 583.0, + 1160.0, + 851.0, + 1160.0, + 851.0, + 1195.0, + 583.0, + 1195.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 871.0, + 1160.0, + 976.0, + 1160.0, + 976.0, + 1195.0, + 871.0, + 1195.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1725.0, + 343.0, + 1725.0, + 343.0, + 1770.0, + 291.0, + 1770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 478.0, + 1725.0, + 715.0, + 1725.0, + 715.0, + 1770.0, + 478.0, + 1770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 738.0, + 1725.0, + 815.0, + 1725.0, + 815.0, + 1770.0, + 738.0, + 1770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1011.0, + 1725.0, + 1064.0, + 1725.0, + 1064.0, + 1770.0, + 1011.0, + 1770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1095.0, + 1725.0, + 1170.0, + 1725.0, + 1170.0, + 1770.0, + 1095.0, + 1770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1751.0, + 296.0, + 1751.0, + 296.0, + 1795.0, + 291.0, + 1795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 332.0, + 1751.0, + 630.0, + 1751.0, + 630.0, + 1795.0, + 332.0, + 1795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 1751.0, + 843.0, + 1751.0, + 843.0, + 1795.0, + 787.0, + 1795.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1010.0, + 1760.0, + 1404.0, + 1760.0, + 1404.0, + 1791.0, + 1010.0, + 1791.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1788.0, + 761.0, + 1788.0, + 761.0, + 1825.0, + 295.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 863.0, + 1788.0, + 1405.0, + 1788.0, + 1405.0, + 1825.0, + 863.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1811.0, + 295.0, + 1811.0, + 295.0, + 1866.0, + 289.0, + 1866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 632.0, + 1811.0, + 1373.0, + 1811.0, + 1373.0, + 1866.0, + 632.0, + 1866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1404.0, + 1811.0, + 1410.0, + 1811.0, + 1410.0, + 1866.0, + 1404.0, + 1866.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1853.0, + 1405.0, + 1853.0, + 1405.0, + 1890.0, + 294.0, + 1890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1883.0, + 846.0, + 1883.0, + 846.0, + 1920.0, + 292.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 875.0, + 1883.0, + 1404.0, + 1883.0, + 1404.0, + 1920.0, + 875.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1911.0, + 1405.0, + 1911.0, + 1405.0, + 1951.0, + 294.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1946.0, + 1406.0, + 1946.0, + 1406.0, + 1979.0, + 296.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1972.0, + 1223.0, + 1972.0, + 1223.0, + 2012.0, + 294.0, + 2012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 884.75, + 1753.0, + 1009.75, + 1753.0, + 1009.75, + 1791.5, + 884.75, + 1791.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1593.0, + 1407.0, + 1593.0, + 1407.0, + 1626.0, + 295.0, + 1626.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1623.0, + 1405.0, + 1623.0, + 1405.0, + 1656.0, + 295.0, + 1656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1654.0, + 1404.0, + 1654.0, + 1404.0, + 1690.0, + 294.0, + 1690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1684.0, + 824.0, + 1684.0, + 824.0, + 1716.0, + 295.0, + 1716.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 1, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1725, + 1405, + 1725, + 1405, + 2008, + 297, + 2008 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 297, + 1150, + 1405, + 1150, + 1405, + 1365, + 297, + 1365 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 952, + 1404, + 952, + 1404, + 1137, + 298, + 1137 + ], + "score": 0.974 + }, + { + "category_id": 1, + "poly": [ + 298, + 453, + 1406, + 453, + 1406, + 641, + 298, + 641 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 300, + 1394, + 1404, + 1394, + 1404, + 1522, + 300, + 1522 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 301, + 654, + 1404, + 654, + 1404, + 747, + 301, + 747 + ], + "score": 0.967 + }, + { + "category_id": 1, + "poly": [ + 297, + 202, + 1402, + 202, + 1402, + 266, + 297, + 266 + ], + "score": 0.949 + }, + { + "category_id": 1, + "poly": [ + 299, + 1633, + 1404, + 1633, + 1404, + 1697, + 299, + 1697 + ], + "score": 0.946 + }, + { + "category_id": 1, + "poly": [ + 370, + 1539, + 1405, + 1539, + 1405, + 1616, + 370, + 1616 + ], + "score": 0.911 + }, + { + "category_id": 1, + "poly": [ + 367, + 765, + 1405, + 765, + 1405, + 934, + 367, + 934 + ], + "score": 0.91 + }, + { + "category_id": 1, + "poly": [ + 371, + 361, + 1249, + 361, + 1249, + 436, + 371, + 436 + ], + "score": 0.907 + }, + { + "category_id": 1, + "poly": [ + 299, + 278, + 1406, + 278, + 1406, + 341, + 299, + 341 + ], + "score": 0.855 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.627 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.473 + }, + { + "category_id": 13, + "poly": [ + 560, + 1580, + 676, + 1580, + 676, + 1615, + 560, + 1615 + ], + "score": 0.94, + "latex": "( u , v ) \\in \\mathcal { E }" + }, + { + "category_id": 13, + "poly": [ + 1069, + 1425, + 1196, + 1425, + 1196, + 1459, + 1069, + 1459 + ], + "score": 0.93, + "latex": "\\mathcal { G } = ( \\nu , \\mathcal { E } )" + }, + { + "category_id": 13, + "poly": [ + 1194, + 547, + 1324, + 547, + 1324, + 582, + 1194, + 582 + ], + "score": 0.93, + "latex": "k \\in \\mathsf { \\Gamma } ( u , v )" + }, + { + "category_id": 13, + "poly": [ + 604, + 1819, + 786, + 1819, + 786, + 1855, + 604, + 1855 + ], + "score": 0.93, + "latex": "\\mathbf { X } _ { [ \\mathrm { g r a p h } ] } \\mathbf { \\bar { \\Pi } } \\in \\mathbf { \\bar { \\Pi } } \\mathbb { R } ^ { d }" + }, + { + "category_id": 13, + "poly": [ + 560, + 400, + 676, + 400, + 676, + 435, + 560, + 435 + ], + "score": 0.93, + "latex": "( u , v ) \\in \\mathcal { E }" + }, + { + "category_id": 13, + "poly": [ + 596, + 1456, + 861, + 1456, + 861, + 1492, + 596, + 1492 + ], + "score": 0.93, + "latex": "{ \\bf E } = [ { \\bf E } ^ { \\nu } ; { \\bf E } ^ { \\varepsilon } ] \\in \\mathbb { R } ^ { 2 \\times d _ { e } }" + }, + { + "category_id": 13, + "poly": [ + 564, + 279, + 691, + 279, + 691, + 312, + 564, + 312 + ], + "score": 0.92, + "latex": "\\mathcal { G } = ( \\nu , \\mathcal { E } )" + }, + { + "category_id": 13, + "poly": [ + 571, + 867, + 975, + 867, + 975, + 900, + 571, + 900 + ], + "score": 0.92, + "latex": " { \\Delta } = { \\mathbf { I } } - { \\mathbf { D } } ^ { - 1 / 2 } { \\mathbf { A } } { \\mathbf { D } } ^ { - 1 / 2 } = { \\mathbf { U } } ^ { \\top } { \\mathbf { A } } \\bar { { \\mathbf { U } } }" + }, + { + "category_id": 13, + "poly": [ + 969, + 515, + 1088, + 515, + 1088, + 548, + 969, + 548 + ], + "score": 0.92, + "latex": "\\boldsymbol { e } = \\left( u , v \\right)" + }, + { + "category_id": 13, + "poly": [ + 967, + 361, + 1116, + 361, + 1116, + 396, + 967, + 396 + ], + "score": 0.92, + "latex": "\\big [ \\mathbf { X } _ { v } , \\mathbf { P } _ { v } , \\mathbf { P } _ { v } \\big ]" + }, + { + "category_id": 13, + "poly": [ + 939, + 1580, + 1126, + 1580, + 1126, + 1618, + 939, + 1618 + ], + "score": 0.92, + "latex": "[ \\mathbf { X } _ { ( u , v ) } , \\mathbf { P } _ { u } , \\mathbf { P } _ { v } ]" + }, + { + "category_id": 13, + "poly": [ + 1077, + 1538, + 1277, + 1538, + 1277, + 1575, + 1077, + 1575 + ], + "score": 0.92, + "latex": "[ \\mathbf { X } _ { v } , \\mathbf { P } _ { v } , \\mathbf { P } _ { v } , \\mathbf { E } ^ { \\nu } ]" + }, + { + "category_id": 13, + "poly": [ + 737, + 545, + 1019, + 545, + 1019, + 582, + 737, + 582 + ], + "score": 0.92, + "latex": "[ { \\bf P } _ { u } , { \\bf P } _ { v } ] [ { \\bf P } _ { k } , { \\bf P } _ { k } ] ^ { \\top } = 1" + }, + { + "category_id": 13, + "poly": [ + 1161, + 1579, + 1397, + 1579, + 1397, + 1618, + 1161, + 1618 + ], + "score": 0.92, + "latex": "[ \\mathbf { X } _ { ( u , v ) } , \\mathbf { P } _ { u } , \\mathbf { P } _ { v } , \\mathbf { E } ^ { \\mathcal { E } } ]" + }, + { + "category_id": 13, + "poly": [ + 1267, + 276, + 1399, + 276, + 1399, + 308, + 1267, + 308 + ], + "score": 0.92, + "latex": "\\mathbf { P } \\in \\mathbb { R } ^ { n \\times d _ { p } }" + }, + { + "category_id": 13, + "poly": [ + 1057, + 870, + 1184, + 870, + 1184, + 899, + 1057, + 899 + ], + "score": 0.91, + "latex": "\\mathbf { A } \\in \\mathbb { R } ^ { n \\times n }" + }, + { + "category_id": 13, + "poly": [ + 893, + 1540, + 1042, + 1540, + 1042, + 1575, + 893, + 1575 + ], + "score": 0.91, + "latex": "[ \\mathbf { X } _ { v } , \\mathbf { P } _ { v } , \\mathbf { P } _ { v } ]" + }, + { + "category_id": 13, + "poly": [ + 938, + 401, + 1014, + 401, + 1014, + 438, + 938, + 438 + ], + "score": 0.91, + "latex": "\\mathbf { X } _ { ( u , v ) }" + }, + { + "category_id": 13, + "poly": [ + 1273, + 766, + 1400, + 766, + 1400, + 799, + 1273, + 799 + ], + "score": 0.91, + "latex": "\\mathbf { Q } \\in \\mathbb { R } ^ { n \\times n }" + }, + { + "category_id": 13, + "poly": [ + 297, + 1911, + 782, + 1911, + 782, + 1949, + 297, + 1949 + ], + "score": 0.91, + "latex": "{ \\bf Z } ^ { ( 0 ) } = [ { \\bf X } _ { [ \\mathrm { g r a p h } ] } ; { \\bf X } ^ { i n } w ^ { i n } ] \\bar { \\in } \\mathbb { R } ^ { ( 1 + n \\bar { + } m ) \\times d }" + }, + { + "category_id": 13, + "poly": [ + 1080, + 798, + 1207, + 798, + 1207, + 828, + 1080, + 828 + ], + "score": 0.9, + "latex": "\\mathbf { G } \\in \\mathbb { R } ^ { n \\times n }" + }, + { + "category_id": 13, + "poly": [ + 1051, + 400, + 1237, + 400, + 1237, + 438, + 1051, + 438 + ], + "score": 0.9, + "latex": "[ \\mathbf { X } _ { ( u , v ) } , \\mathbf { P } _ { u } , \\mathbf { P } _ { v } ]" + }, + { + "category_id": 13, + "poly": [ + 1146, + 1756, + 1400, + 1756, + 1400, + 1789, + 1146, + 1789 + ], + "score": 0.9, + "latex": "\\overline { { w } } ^ { i n } \\in \\mathbb { R } ^ { ( C + 2 d _ { p } + d _ { e } ) \\times d }" + }, + { + "category_id": 13, + "poly": [ + 296, + 1757, + 610, + 1757, + 610, + 1789, + 296, + 1789 + ], + "score": 0.89, + "latex": "\\mathbf { X } ^ { i n } \\in \\mathbb { R } ^ { ( n + m ) \\times ( C + 2 d _ { p } + d _ { e } ) }" + }, + { + "category_id": 13, + "poly": [ + 560, + 364, + 630, + 364, + 630, + 392, + 560, + 392 + ], + "score": 0.89, + "latex": "v \\in \\mathcal V" + }, + { + "category_id": 13, + "poly": [ + 1232, + 1456, + 1273, + 1456, + 1273, + 1487, + 1232, + 1487 + ], + "score": 0.89, + "latex": "\\mathbf { E } ^ { \\nu }" + }, + { + "category_id": 13, + "poly": [ + 1323, + 1456, + 1362, + 1456, + 1362, + 1487, + 1323, + 1487 + ], + "score": 0.89, + "latex": "\\mathbf { E } ^ { \\mathcal { E } }" + }, + { + "category_id": 13, + "poly": [ + 560, + 1543, + 630, + 1543, + 630, + 1571, + 560, + 1571 + ], + "score": 0.89, + "latex": "v \\in \\mathcal V" + }, + { + "category_id": 13, + "poly": [ + 892, + 363, + 932, + 363, + 932, + 394, + 892, + 394 + ], + "score": 0.88, + "latex": "\\mathbf { X } _ { v }" + }, + { + "category_id": 13, + "poly": [ + 1382, + 516, + 1401, + 516, + 1401, + 543, + 1382, + 543 + ], + "score": 0.78, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 889, + 286, + 910, + 286, + 910, + 307, + 889, + 307 + ], + "score": 0.74, + "latex": "n" + }, + { + "category_id": 13, + "poly": [ + 657, + 655, + 684, + 655, + 684, + 683, + 657, + 683 + ], + "score": 0.68, + "latex": "\\mathbf { P }" + }, + { + "category_id": 13, + "poly": [ + 1019, + 310, + 1046, + 310, + 1046, + 337, + 1019, + 337 + ], + "score": 0.56, + "latex": "\\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 396, + 902, + 425, + 902, + 425, + 930, + 396, + 930 + ], + "score": 0.49, + "latex": "\\mathbf { D }" + }, + { + "category_id": 13, + "poly": [ + 665, + 903, + 689, + 903, + 689, + 929, + 665, + 929 + ], + "score": 0.44, + "latex": "\\pmb { \\Lambda }" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1724.0, + 1405.0, + 1724.0, + 1405.0, + 1762.0, + 292.0, + 1762.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 287.0, + 1748.0, + 295.0, + 1748.0, + 295.0, + 1799.0, + 287.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 611.0, + 1748.0, + 1145.0, + 1748.0, + 1145.0, + 1799.0, + 611.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1401.0, + 1748.0, + 1411.0, + 1748.0, + 1411.0, + 1799.0, + 1401.0, + 1799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1791.0, + 1401.0, + 1791.0, + 1401.0, + 1825.0, + 295.0, + 1825.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1816.0, + 603.0, + 1816.0, + 603.0, + 1862.0, + 290.0, + 1862.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 1816.0, + 1408.0, + 1816.0, + 1408.0, + 1862.0, + 787.0, + 1862.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1852.0, + 1406.0, + 1852.0, + 1406.0, + 1886.0, + 295.0, + 1886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1878.0, + 1406.0, + 1878.0, + 1406.0, + 1917.0, + 292.0, + 1917.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1905.0, + 296.0, + 1905.0, + 296.0, + 1957.0, + 289.0, + 1957.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 783.0, + 1905.0, + 1413.0, + 1905.0, + 1413.0, + 1957.0, + 783.0, + 1957.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1945.0, + 1405.0, + 1945.0, + 1405.0, + 1979.0, + 295.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1974.0, + 1304.0, + 1974.0, + 1304.0, + 2010.0, + 294.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1152.0, + 1402.0, + 1152.0, + 1402.0, + 1186.0, + 295.0, + 1186.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1180.0, + 1407.0, + 1180.0, + 1407.0, + 1216.0, + 294.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1212.0, + 1406.0, + 1212.0, + 1406.0, + 1246.0, + 294.0, + 1246.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1242.0, + 1406.0, + 1242.0, + 1406.0, + 1276.0, + 294.0, + 1276.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1272.0, + 1405.0, + 1272.0, + 1405.0, + 1307.0, + 295.0, + 1307.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1303.0, + 1405.0, + 1303.0, + 1405.0, + 1337.0, + 294.0, + 1337.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1331.0, + 1408.0, + 1331.0, + 1408.0, + 1368.0, + 294.0, + 1368.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 954.0, + 1404.0, + 954.0, + 1404.0, + 986.0, + 296.0, + 986.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 985.0, + 1405.0, + 985.0, + 1405.0, + 1020.0, + 294.0, + 1020.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1014.0, + 1405.0, + 1014.0, + 1405.0, + 1050.0, + 294.0, + 1050.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1044.0, + 1405.0, + 1044.0, + 1405.0, + 1080.0, + 294.0, + 1080.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1078.0, + 1405.0, + 1078.0, + 1405.0, + 1109.0, + 296.0, + 1109.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1106.0, + 1330.0, + 1106.0, + 1330.0, + 1140.0, + 293.0, + 1140.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 452.0, + 1404.0, + 452.0, + 1404.0, + 490.0, + 293.0, + 490.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 485.0, + 1404.0, + 485.0, + 1404.0, + 517.0, + 296.0, + 517.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 515.0, + 968.0, + 515.0, + 968.0, + 551.0, + 295.0, + 551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1089.0, + 515.0, + 1381.0, + 515.0, + 1381.0, + 551.0, + 1089.0, + 551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 515.0, + 1406.0, + 515.0, + 1406.0, + 551.0, + 1402.0, + 551.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 544.0, + 736.0, + 544.0, + 736.0, + 587.0, + 291.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1020.0, + 544.0, + 1193.0, + 544.0, + 1193.0, + 587.0, + 1020.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1325.0, + 544.0, + 1408.0, + 544.0, + 1408.0, + 587.0, + 1325.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 577.0, + 1408.0, + 577.0, + 1408.0, + 614.0, + 293.0, + 614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 607.0, + 1328.0, + 607.0, + 1328.0, + 643.0, + 293.0, + 643.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1395.0, + 1405.0, + 1395.0, + 1405.0, + 1428.0, + 296.0, + 1428.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1425.0, + 1068.0, + 1425.0, + 1068.0, + 1462.0, + 295.0, + 1462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1197.0, + 1425.0, + 1408.0, + 1425.0, + 1408.0, + 1462.0, + 1197.0, + 1462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1453.0, + 595.0, + 1453.0, + 595.0, + 1494.0, + 293.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 862.0, + 1453.0, + 1231.0, + 1453.0, + 1231.0, + 1494.0, + 862.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1274.0, + 1453.0, + 1322.0, + 1453.0, + 1322.0, + 1494.0, + 1274.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1363.0, + 1453.0, + 1406.0, + 1453.0, + 1406.0, + 1494.0, + 1363.0, + 1494.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1486.0, + 1401.0, + 1486.0, + 1401.0, + 1526.0, + 294.0, + 1526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 652.0, + 656.0, + 652.0, + 656.0, + 690.0, + 295.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 685.0, + 652.0, + 1406.0, + 652.0, + 1406.0, + 690.0, + 685.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 685.0, + 1405.0, + 685.0, + 1405.0, + 719.0, + 295.0, + 719.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 715.0, + 928.0, + 715.0, + 928.0, + 749.0, + 296.0, + 749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 202.0, + 1406.0, + 202.0, + 1406.0, + 238.0, + 295.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 235.0, + 1128.0, + 235.0, + 1128.0, + 267.0, + 296.0, + 267.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1633.0, + 1405.0, + 1633.0, + 1405.0, + 1669.0, + 295.0, + 1669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1662.0, + 1409.0, + 1662.0, + 1409.0, + 1700.0, + 294.0, + 1700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 364.0, + 1534.0, + 559.0, + 1534.0, + 559.0, + 1579.0, + 364.0, + 1579.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 631.0, + 1534.0, + 892.0, + 1534.0, + 892.0, + 1579.0, + 631.0, + 1579.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1043.0, + 1534.0, + 1076.0, + 1534.0, + 1076.0, + 1579.0, + 1043.0, + 1579.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1278.0, + 1534.0, + 1288.0, + 1534.0, + 1288.0, + 1579.0, + 1278.0, + 1579.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 1574.0, + 559.0, + 1574.0, + 559.0, + 1620.0, + 367.0, + 1620.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 677.0, + 1574.0, + 938.0, + 1574.0, + 938.0, + 1620.0, + 677.0, + 1620.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1127.0, + 1574.0, + 1160.0, + 1574.0, + 1160.0, + 1620.0, + 1127.0, + 1620.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1398.0, + 1574.0, + 1408.0, + 1574.0, + 1408.0, + 1620.0, + 1398.0, + 1620.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 371.0, + 762.0, + 1272.0, + 762.0, + 1272.0, + 803.0, + 371.0, + 803.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1401.0, + 762.0, + 1406.0, + 762.0, + 1406.0, + 803.0, + 1401.0, + 803.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 392.0, + 792.0, + 1079.0, + 792.0, + 1079.0, + 836.0, + 392.0, + 836.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1208.0, + 792.0, + 1315.0, + 792.0, + 1315.0, + 836.0, + 1208.0, + 836.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 379.0, + 835.0, + 1409.0, + 835.0, + 1409.0, + 875.0, + 379.0, + 875.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 389.0, + 861.0, + 570.0, + 861.0, + 570.0, + 911.0, + 389.0, + 911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 861.0, + 1056.0, + 861.0, + 1056.0, + 911.0, + 976.0, + 911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1185.0, + 861.0, + 1411.0, + 861.0, + 1411.0, + 911.0, + 1185.0, + 911.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 426.0, + 899.0, + 664.0, + 899.0, + 664.0, + 938.0, + 426.0, + 938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 690.0, + 899.0, + 1409.0, + 899.0, + 1409.0, + 938.0, + 690.0, + 938.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 357.0, + 559.0, + 357.0, + 559.0, + 400.0, + 369.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 631.0, + 357.0, + 891.0, + 357.0, + 891.0, + 400.0, + 631.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 933.0, + 357.0, + 966.0, + 357.0, + 966.0, + 400.0, + 933.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1117.0, + 357.0, + 1126.0, + 357.0, + 1126.0, + 400.0, + 1117.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 394.0, + 559.0, + 394.0, + 559.0, + 441.0, + 369.0, + 441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 677.0, + 394.0, + 937.0, + 394.0, + 937.0, + 441.0, + 677.0, + 441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1015.0, + 394.0, + 1050.0, + 394.0, + 1050.0, + 441.0, + 1015.0, + 441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1238.0, + 394.0, + 1248.0, + 394.0, + 1248.0, + 441.0, + 1238.0, + 441.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 275.0, + 563.0, + 275.0, + 563.0, + 313.0, + 292.0, + 313.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 692.0, + 275.0, + 888.0, + 275.0, + 888.0, + 313.0, + 692.0, + 313.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 911.0, + 275.0, + 1266.0, + 275.0, + 1266.0, + 313.0, + 911.0, + 313.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1400.0, + 275.0, + 1404.0, + 275.0, + 1404.0, + 313.0, + 1400.0, + 313.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 308.0, + 1018.0, + 308.0, + 1018.0, + 344.0, + 296.0, + 344.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1047.0, + 308.0, + 1408.0, + 308.0, + 1408.0, + 344.0, + 1047.0, + 344.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 2, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 201, + 1407, + 201, + 1407, + 507, + 297, + 507 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 297, + 613, + 1405, + 613, + 1405, + 829, + 297, + 829 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 918, + 1406, + 918, + 1406, + 1074, + 298, + 1074 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 296, + 1094, + 1405, + 1094, + 1405, + 1324, + 296, + 1324 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 299, + 1853, + 1404, + 1853, + 1404, + 2008, + 299, + 2008 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 298, + 1340, + 1404, + 1340, + 1404, + 1496, + 298, + 1496 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 297, + 1525, + 1401, + 1525, + 1401, + 1619, + 297, + 1619 + ], + "score": 0.966 + }, + { + "category_id": 8, + "poly": [ + 572, + 1700, + 1125, + 1700, + 1125, + 1748, + 572, + 1748 + ], + "score": 0.948 + }, + { + "category_id": 1, + "poly": [ + 298, + 1755, + 1405, + 1755, + 1405, + 1834, + 298, + 1834 + ], + "score": 0.945 + }, + { + "category_id": 1, + "poly": [ + 297, + 1629, + 1400, + 1629, + 1400, + 1694, + 297, + 1694 + ], + "score": 0.935 + }, + { + "category_id": 0, + "poly": [ + 298, + 863, + 1179, + 863, + 1179, + 897, + 298, + 897 + ], + "score": 0.927 + }, + { + "category_id": 0, + "poly": [ + 299, + 548, + 642, + 548, + 642, + 586, + 299, + 586 + ], + "score": 0.919 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1712, + 1400, + 1712, + 1400, + 1741, + 1366, + 1741 + ], + "score": 0.857 + }, + { + "category_id": 2, + "poly": [ + 841, + 2062, + 858, + 2062, + 858, + 2084, + 841, + 2084 + ], + "score": 0.778 + }, + { + "category_id": 13, + "poly": [ + 427, + 1754, + 496, + 1754, + 496, + 1798, + 427, + 1798 + ], + "score": 0.93, + "latex": "L _ { k k } ^ { ( t ) }" + }, + { + "category_id": 13, + "poly": [ + 897, + 1251, + 1106, + 1251, + 1106, + 1285, + 897, + 1285 + ], + "score": 0.93, + "latex": "{ \\bf { \\dot { F } } } ( \\pi \\cdot { \\bf { X } } ) = F ( { \\bf { X } } )" + }, + { + "category_id": 13, + "poly": [ + 1127, + 1283, + 1263, + 1283, + 1263, + 1318, + 1127, + 1318 + ], + "score": 0.93, + "latex": "\\mathbf { X } \\in \\mathbb { R } ^ { n ^ { k } \\times d }" + }, + { + "category_id": 13, + "poly": [ + 459, + 1288, + 704, + 1288, + 704, + 1323, + 459, + 1323 + ], + "score": 0.93, + "latex": "F ( { \\boldsymbol \\pi } \\cdot \\mathbf { X } ) = { \\boldsymbol \\pi } \\cdot F ( \\mathbf { X } )" + }, + { + "category_id": 13, + "poly": [ + 681, + 948, + 784, + 948, + 784, + 981, + 681, + 981 + ], + "score": 0.93, + "latex": "\\mathbf { X } _ { i } \\in \\mathbb { R } ^ { d }" + }, + { + "category_id": 13, + "poly": [ + 296, + 1126, + 466, + 1126, + 466, + 1161, + 296, + 1161 + ], + "score": 0.93, + "latex": "\\mathbf { X } _ { i _ { 1 } , \\dots , i _ { k } } \\in \\mathbb { R } ^ { d }" + }, + { + "category_id": 13, + "poly": [ + 1117, + 1336, + 1254, + 1336, + 1254, + 1371, + 1117, + 1371 + ], + "score": 0.93, + "latex": "\\mathbf { X } \\in \\mathbb { R } ^ { n ^ { k } \\times d }" + }, + { + "category_id": 13, + "poly": [ + 633, + 1186, + 1000, + 1186, + 1000, + 1221, + 633, + 1221 + ], + "score": 0.93, + "latex": "\\pi ^ { - 1 } ( \\mathbf { i } ) = ( \\pi ^ { - 1 } ( i _ { 1 } ) , . . . , \\pi ^ { - 1 } ( i _ { k } ) )" + }, + { + "category_id": 13, + "poly": [ + 1130, + 1221, + 1258, + 1221, + 1258, + 1254, + 1130, + 1254 + ], + "score": 0.93, + "latex": "F ( \\mathbf { X } ) \\approx T" + }, + { + "category_id": 13, + "poly": [ + 1110, + 1624, + 1316, + 1624, + 1316, + 1662, + 1110, + 1662 + ], + "score": 0.92, + "latex": "F _ { k } : \\mathbb { R } ^ { n ^ { k } \\times d _ { 0 } } \\mathbb { R }" + }, + { + "category_id": 13, + "poly": [ + 656, + 1010, + 832, + 1010, + 832, + 1043, + 656, + 1043 + ], + "score": 0.92, + "latex": "[ n ] = \\{ 1 , . . . , n \\}" + }, + { + "category_id": 13, + "poly": [ + 1054, + 1090, + 1191, + 1090, + 1191, + 1125, + 1054, + 1125 + ], + "score": 0.92, + "latex": "\\mathbf { X } \\in \\mathbb { R } ^ { n ^ { k } \\times d }" + }, + { + "category_id": 13, + "poly": [ + 478, + 948, + 603, + 948, + 603, + 979, + 478, + 979 + ], + "score": 0.92, + "latex": "\\mathbf { X } \\in \\mathbb { R } ^ { n \\times d }" + }, + { + "category_id": 13, + "poly": [ + 334, + 1187, + 553, + 1187, + 553, + 1224, + 334, + 1224 + ], + "score": 0.91, + "latex": "( { \\boldsymbol { \\pi } } \\cdot \\mathbf { X } ) _ { \\mathbf { i } } = \\mathbf { X } _ { \\pi ^ { - 1 } ( \\mathbf { i } ) }" + }, + { + "category_id": 13, + "poly": [ + 1155, + 1042, + 1239, + 1042, + 1239, + 1072, + 1155, + 1072 + ], + "score": 0.91, + "latex": "\\pi \\in S _ { n }" + }, + { + "category_id": 13, + "poly": [ + 1335, + 1097, + 1406, + 1097, + 1406, + 1128, + 1335, + 1128 + ], + "score": 0.91, + "latex": "\\mathbf { X _ { i } } =" + }, + { + "category_id": 13, + "poly": [ + 346, + 1040, + 568, + 1040, + 568, + 1077, + 346, + 1077 + ], + "score": 0.91, + "latex": "( \\pi \\cdot \\mathbf { \\bar { X } } ) _ { i } = \\mathbf { X } _ { \\pi ^ { - 1 } ( i ) } ." + }, + { + "category_id": 13, + "poly": [ + 347, + 1803, + 414, + 1803, + 414, + 1832, + 347, + 1832 + ], + "score": 0.9, + "latex": "L _ { k 0 }" + }, + { + "category_id": 13, + "poly": [ + 1313, + 1288, + 1397, + 1288, + 1397, + 1320, + 1313, + 1320 + ], + "score": 0.9, + "latex": "\\pi \\in S _ { n }" + }, + { + "category_id": 13, + "poly": [ + 1143, + 1161, + 1231, + 1161, + 1231, + 1188, + 1143, + 1188 + ], + "score": 0.9, + "latex": "\\pi \\in S _ { n }" + }, + { + "category_id": 13, + "poly": [ + 1052, + 1127, + 1298, + 1127, + 1298, + 1160, + 1052, + 1160 + ], + "score": 0.9, + "latex": "\\mathbf { i } = ( i _ { 1 } , . . . , i _ { k } ) \\in [ n ] ^ { k }" + }, + { + "category_id": 14, + "poly": [ + 573, + 1699, + 1124, + 1699, + 1124, + 1746, + 573, + 1746 + ], + "score": 0.9, + "latex": "F _ { k } = \\mathbf { M } \\mathbf { L } \\mathbf { P } \\circ L _ { k 0 } \\circ L _ { k k } ^ { ( T ) } \\circ \\sigma \\circ \\dots \\circ \\sigma \\circ L _ { k k } ^ { ( 1 ) } ," + }, + { + "category_id": 13, + "poly": [ + 1106, + 982, + 1140, + 982, + 1140, + 1011, + 1106, + 1011 + ], + "score": 0.89, + "latex": "S _ { n }" + }, + { + "category_id": 13, + "poly": [ + 1184, + 1190, + 1249, + 1190, + 1249, + 1216, + 1184, + 1216 + ], + "score": 0.87, + "latex": "\\pi \\cdot { \\bf X }" + }, + { + "category_id": 13, + "poly": [ + 884, + 1011, + 950, + 1011, + 950, + 1039, + 884, + 1039 + ], + "score": 0.86, + "latex": "\\pi \\cdot \\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 935, + 1374, + 1010, + 1374, + 1010, + 1402, + 935, + 1402 + ], + "score": 0.85, + "latex": "k = 2" + }, + { + "category_id": 13, + "poly": [ + 676, + 1374, + 752, + 1374, + 752, + 1402, + 676, + 1402 + ], + "score": 0.85, + "latex": " { \\left( k = 1 \\right. }" + }, + { + "category_id": 13, + "poly": [ + 724, + 1043, + 788, + 1043, + 788, + 1069, + 724, + 1069 + ], + "score": 0.85, + "latex": "\\pi \\cdot \\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 1287, + 1374, + 1364, + 1374, + 1364, + 1403, + 1287, + 1403 + ], + "score": 0.84, + "latex": "( k > 2 )" + }, + { + "category_id": 13, + "poly": [ + 709, + 1886, + 727, + 1886, + 727, + 1912, + 709, + 1912 + ], + "score": 0.83, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 618, + 1916, + 636, + 1916, + 636, + 1942, + 618, + 1942 + ], + "score": 0.83, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 958, + 1098, + 976, + 1098, + 976, + 1124, + 958, + 1124 + ], + "score": 0.82, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1100, + 1856, + 1118, + 1856, + 1118, + 1882, + 1100, + 1882 + ], + "score": 0.82, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 412, + 1254, + 435, + 1254, + 435, + 1279, + 412, + 1279 + ], + "score": 0.81, + "latex": "T" + }, + { + "category_id": 13, + "poly": [ + 793, + 1344, + 811, + 1344, + 811, + 1370, + 793, + 1370 + ], + "score": 0.8, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1177, + 1916, + 1195, + 1916, + 1195, + 1942, + 1177, + 1942 + ], + "score": 0.79, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 574, + 799, + 592, + 799, + 592, + 825, + 574, + 825 + ], + "score": 0.78, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1115, + 799, + 1134, + 799, + 1134, + 825, + 1115, + 825 + ], + "score": 0.78, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 625, + 1227, + 644, + 1227, + 644, + 1249, + 625, + 1249 + ], + "score": 0.76, + "latex": "\\pi" + }, + { + "category_id": 13, + "poly": [ + 1377, + 1016, + 1397, + 1016, + 1397, + 1038, + 1377, + 1038 + ], + "score": 0.76, + "latex": "\\pi" + }, + { + "category_id": 13, + "poly": [ + 596, + 1015, + 617, + 1015, + 617, + 1038, + 596, + 1038 + ], + "score": 0.76, + "latex": "\\pi" + }, + { + "category_id": 13, + "poly": [ + 767, + 1794, + 916, + 1794, + 916, + 1830, + 767, + 1830 + ], + "score": 0.75, + "latex": "\\mathbb { R } ^ { n ^ { k } \\times d _ { T } } t o \\mathbb { R }" + }, + { + "category_id": 13, + "poly": [ + 516, + 1435, + 537, + 1435, + 537, + 1461, + 516, + 1461 + ], + "score": 0.74, + "latex": "\\mathcal { E }" + }, + { + "category_id": 13, + "poly": [ + 1092, + 926, + 1112, + 926, + 1112, + 947, + 1092, + 947 + ], + "score": 0.73, + "latex": "n" + }, + { + "category_id": 13, + "poly": [ + 1203, + 799, + 1220, + 799, + 1220, + 826, + 1203, + 826 + ], + "score": 0.73, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1290, + 1012, + 1316, + 1012, + 1316, + 1038, + 1290, + 1038 + ], + "score": 0.72, + "latex": "\\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 562, + 1633, + 579, + 1633, + 579, + 1659, + 562, + 1659 + ], + "score": 0.71, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1105, + 1189, + 1133, + 1189, + 1133, + 1216, + 1105, + 1216 + ], + "score": 0.7, + "latex": "\\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 1007, + 953, + 1020, + 953, + 1020, + 978, + 1007, + 978 + ], + "score": 0.68, + "latex": "i" + }, + { + "category_id": 13, + "poly": [ + 961, + 1887, + 978, + 1887, + 978, + 1912, + 961, + 1912 + ], + "score": 0.68, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 646, + 1043, + 674, + 1043, + 674, + 1068, + 646, + 1068 + ], + "score": 0.67, + "latex": "\\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 1045, + 1755, + 1131, + 1755, + 1131, + 1790, + 1045, + 1790 + ], + "score": 0.65, + "latex": "\\mathbb { R } ^ { n ^ { k } \\times d _ { t } }" + }, + { + "category_id": 13, + "poly": [ + 1143, + 1768, + 1164, + 1768, + 1164, + 1789, + 1143, + 1789 + ], + "score": 0.65, + "latex": "\\sigma" + }, + { + "category_id": 13, + "poly": [ + 902, + 1755, + 1012, + 1755, + 1012, + 1789, + 902, + 1789 + ], + "score": 0.63, + "latex": "\\mathbb { R } ^ { n ^ { k } \\times d _ { t - 1 } }" + }, + { + "category_id": 13, + "poly": [ + 1375, + 1158, + 1402, + 1158, + 1402, + 1186, + 1375, + 1186 + ], + "score": 0.62, + "latex": "\\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 879, + 1633, + 896, + 1633, + 896, + 1659, + 879, + 1659 + ], + "score": 0.62, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1375, + 1221, + 1403, + 1221, + 1403, + 1250, + 1375, + 1250 + ], + "score": 0.56, + "latex": "\\mathbf { X }" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 858.0, + 1182.0, + 858.0, + 1182.0, + 906.0, + 291.0, + 906.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 543.0, + 645.0, + 543.0, + 645.0, + 593.0, + 290.0, + 593.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2059.0, + 862.0, + 2059.0, + 862.0, + 2091.0, + 838.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 200.0, + 1409.0, + 200.0, + 1409.0, + 239.0, + 294.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 231.0, + 1407.0, + 231.0, + 1407.0, + 270.0, + 294.0, + 270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 261.0, + 1408.0, + 261.0, + 1408.0, + 299.0, + 292.0, + 299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 294.0, + 1409.0, + 294.0, + 1409.0, + 330.0, + 294.0, + 330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 320.0, + 1408.0, + 320.0, + 1408.0, + 363.0, + 291.0, + 363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 355.0, + 1405.0, + 355.0, + 1405.0, + 387.0, + 296.0, + 387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 383.0, + 1405.0, + 383.0, + 1405.0, + 420.0, + 294.0, + 420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 411.0, + 1405.0, + 411.0, + 1405.0, + 453.0, + 292.0, + 453.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 445.0, + 1405.0, + 445.0, + 1405.0, + 481.0, + 295.0, + 481.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 476.0, + 1158.0, + 476.0, + 1158.0, + 510.0, + 294.0, + 510.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 615.0, + 1406.0, + 615.0, + 1406.0, + 648.0, + 294.0, + 648.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 648.0, + 1406.0, + 648.0, + 1406.0, + 679.0, + 294.0, + 679.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 677.0, + 1405.0, + 677.0, + 1405.0, + 711.0, + 295.0, + 711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 707.0, + 1405.0, + 707.0, + 1405.0, + 739.0, + 294.0, + 739.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 736.0, + 1405.0, + 736.0, + 1405.0, + 773.0, + 292.0, + 773.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 767.0, + 1407.0, + 767.0, + 1407.0, + 805.0, + 292.0, + 805.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 798.0, + 573.0, + 798.0, + 573.0, + 832.0, + 295.0, + 832.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 593.0, + 798.0, + 1114.0, + 798.0, + 1114.0, + 832.0, + 593.0, + 832.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1135.0, + 798.0, + 1202.0, + 798.0, + 1202.0, + 832.0, + 1135.0, + 832.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1221.0, + 798.0, + 1351.0, + 798.0, + 1351.0, + 832.0, + 1221.0, + 832.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 917.0, + 1091.0, + 917.0, + 1091.0, + 955.0, + 295.0, + 955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1113.0, + 917.0, + 1406.0, + 917.0, + 1406.0, + 955.0, + 1113.0, + 955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 941.0, + 477.0, + 941.0, + 477.0, + 987.0, + 292.0, + 987.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 604.0, + 941.0, + 680.0, + 941.0, + 680.0, + 987.0, + 604.0, + 987.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 785.0, + 941.0, + 1006.0, + 941.0, + 1006.0, + 987.0, + 785.0, + 987.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1021.0, + 941.0, + 1409.0, + 941.0, + 1409.0, + 987.0, + 1021.0, + 987.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 976.0, + 1105.0, + 976.0, + 1105.0, + 1017.0, + 293.0, + 1017.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1141.0, + 976.0, + 1407.0, + 976.0, + 1407.0, + 1017.0, + 1141.0, + 1017.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1009.0, + 595.0, + 1009.0, + 595.0, + 1047.0, + 295.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 618.0, + 1009.0, + 655.0, + 1009.0, + 655.0, + 1047.0, + 618.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 833.0, + 1009.0, + 883.0, + 1009.0, + 883.0, + 1047.0, + 833.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 951.0, + 1009.0, + 1289.0, + 1009.0, + 1289.0, + 1047.0, + 951.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1317.0, + 1009.0, + 1376.0, + 1009.0, + 1376.0, + 1047.0, + 1317.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1398.0, + 1009.0, + 1409.0, + 1009.0, + 1409.0, + 1047.0, + 1398.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1036.0, + 345.0, + 1036.0, + 345.0, + 1079.0, + 292.0, + 1079.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 569.0, + 1036.0, + 645.0, + 1036.0, + 645.0, + 1079.0, + 569.0, + 1079.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 675.0, + 1036.0, + 723.0, + 1036.0, + 723.0, + 1079.0, + 675.0, + 1079.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 1036.0, + 1154.0, + 1036.0, + 1154.0, + 1079.0, + 789.0, + 1079.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1240.0, + 1036.0, + 1253.0, + 1036.0, + 1253.0, + 1079.0, + 1240.0, + 1079.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1085.0, + 957.0, + 1085.0, + 957.0, + 1136.0, + 290.0, + 1136.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 977.0, + 1085.0, + 1053.0, + 1085.0, + 1053.0, + 1136.0, + 977.0, + 1136.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1192.0, + 1085.0, + 1334.0, + 1085.0, + 1334.0, + 1136.0, + 1192.0, + 1136.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1123.0, + 295.0, + 1123.0, + 295.0, + 1164.0, + 289.0, + 1164.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 467.0, + 1123.0, + 1051.0, + 1123.0, + 1051.0, + 1164.0, + 467.0, + 1164.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1299.0, + 1123.0, + 1410.0, + 1123.0, + 1410.0, + 1164.0, + 1299.0, + 1164.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1155.0, + 1142.0, + 1155.0, + 1142.0, + 1195.0, + 293.0, + 1195.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1232.0, + 1155.0, + 1374.0, + 1155.0, + 1374.0, + 1195.0, + 1232.0, + 1195.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1403.0, + 1155.0, + 1407.0, + 1155.0, + 1407.0, + 1195.0, + 1403.0, + 1195.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1185.0, + 333.0, + 1185.0, + 333.0, + 1223.0, + 291.0, + 1223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 554.0, + 1185.0, + 632.0, + 1185.0, + 632.0, + 1223.0, + 554.0, + 1223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1001.0, + 1185.0, + 1104.0, + 1185.0, + 1104.0, + 1223.0, + 1001.0, + 1223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1134.0, + 1185.0, + 1183.0, + 1185.0, + 1183.0, + 1223.0, + 1134.0, + 1223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1250.0, + 1185.0, + 1406.0, + 1185.0, + 1406.0, + 1223.0, + 1250.0, + 1223.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1217.0, + 624.0, + 1217.0, + 624.0, + 1258.0, + 291.0, + 1258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 1217.0, + 1129.0, + 1217.0, + 1129.0, + 1258.0, + 645.0, + 1258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1259.0, + 1217.0, + 1374.0, + 1217.0, + 1374.0, + 1258.0, + 1259.0, + 1258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1404.0, + 1217.0, + 1407.0, + 1217.0, + 1407.0, + 1258.0, + 1404.0, + 1258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1250.0, + 411.0, + 1250.0, + 411.0, + 1288.0, + 295.0, + 1288.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 436.0, + 1250.0, + 896.0, + 1250.0, + 896.0, + 1288.0, + 436.0, + 1288.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1107.0, + 1250.0, + 1407.0, + 1250.0, + 1407.0, + 1288.0, + 1107.0, + 1288.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1279.0, + 458.0, + 1279.0, + 458.0, + 1330.0, + 289.0, + 1330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 705.0, + 1279.0, + 1126.0, + 1279.0, + 1126.0, + 1330.0, + 705.0, + 1330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1264.0, + 1279.0, + 1312.0, + 1279.0, + 1312.0, + 1330.0, + 1264.0, + 1330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1398.0, + 1279.0, + 1413.0, + 1279.0, + 1413.0, + 1330.0, + 1398.0, + 1330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1855.0, + 1099.0, + 1855.0, + 1099.0, + 1888.0, + 294.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1119.0, + 1855.0, + 1404.0, + 1855.0, + 1404.0, + 1888.0, + 1119.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1883.0, + 708.0, + 1883.0, + 708.0, + 1920.0, + 293.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 728.0, + 1883.0, + 960.0, + 1883.0, + 960.0, + 1920.0, + 728.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 1883.0, + 1408.0, + 1883.0, + 1408.0, + 1920.0, + 979.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1913.0, + 617.0, + 1913.0, + 617.0, + 1950.0, + 294.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 637.0, + 1913.0, + 1176.0, + 1913.0, + 1176.0, + 1950.0, + 637.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1196.0, + 1913.0, + 1406.0, + 1913.0, + 1406.0, + 1950.0, + 1196.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1942.0, + 1406.0, + 1942.0, + 1406.0, + 1982.0, + 293.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1972.0, + 783.0, + 1972.0, + 783.0, + 2012.0, + 293.0, + 2012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1334.0, + 792.0, + 1334.0, + 792.0, + 1379.0, + 292.0, + 1379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 812.0, + 1334.0, + 1116.0, + 1334.0, + 1116.0, + 1379.0, + 812.0, + 1379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1255.0, + 1334.0, + 1409.0, + 1334.0, + 1409.0, + 1379.0, + 1255.0, + 1379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1370.0, + 675.0, + 1370.0, + 675.0, + 1409.0, + 293.0, + 1409.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 753.0, + 1370.0, + 934.0, + 1370.0, + 934.0, + 1409.0, + 753.0, + 1409.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1011.0, + 1370.0, + 1286.0, + 1370.0, + 1286.0, + 1409.0, + 1011.0, + 1409.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1365.0, + 1370.0, + 1406.0, + 1370.0, + 1406.0, + 1409.0, + 1365.0, + 1409.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1402.0, + 1406.0, + 1402.0, + 1406.0, + 1439.0, + 292.0, + 1439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1429.0, + 515.0, + 1429.0, + 515.0, + 1474.0, + 292.0, + 1474.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 538.0, + 1429.0, + 1406.0, + 1429.0, + 1406.0, + 1474.0, + 538.0, + 1474.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1465.0, + 1041.0, + 1465.0, + 1041.0, + 1498.0, + 296.0, + 1498.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1525.0, + 1405.0, + 1525.0, + 1405.0, + 1562.0, + 293.0, + 1562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1557.0, + 1405.0, + 1557.0, + 1405.0, + 1591.0, + 295.0, + 1591.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1588.0, + 1376.0, + 1588.0, + 1376.0, + 1623.0, + 292.0, + 1623.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 420.0, + 1747.0, + 426.0, + 1747.0, + 426.0, + 1806.0, + 420.0, + 1806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 497.0, + 1747.0, + 901.0, + 1747.0, + 901.0, + 1806.0, + 497.0, + 1806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1013.0, + 1747.0, + 1044.0, + 1747.0, + 1044.0, + 1806.0, + 1013.0, + 1806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1132.0, + 1747.0, + 1142.0, + 1747.0, + 1142.0, + 1806.0, + 1132.0, + 1806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1165.0, + 1747.0, + 1416.0, + 1747.0, + 1416.0, + 1806.0, + 1165.0, + 1806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1790.0, + 346.0, + 1790.0, + 346.0, + 1839.0, + 289.0, + 1839.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 415.0, + 1790.0, + 766.0, + 1790.0, + 766.0, + 1839.0, + 415.0, + 1839.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 917.0, + 1790.0, + 929.0, + 1790.0, + 929.0, + 1839.0, + 917.0, + 1839.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1749.5, + 488.0, + 1749.5, + 488.0, + 1798.5, + 291.0, + 1798.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1622.0, + 561.0, + 1622.0, + 561.0, + 1669.0, + 291.0, + 1669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 580.0, + 1622.0, + 878.0, + 1622.0, + 878.0, + 1669.0, + 580.0, + 1669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 897.0, + 1622.0, + 1109.0, + 1622.0, + 1109.0, + 1669.0, + 897.0, + 1669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1317.0, + 1622.0, + 1407.0, + 1622.0, + 1407.0, + 1669.0, + 1317.0, + 1669.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1656.0, + 492.0, + 1656.0, + 492.0, + 1702.0, + 291.0, + 1702.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 3, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1181, + 1404, + 1181, + 1404, + 1312, + 297, + 1312 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 295, + 1533, + 1406, + 1533, + 1406, + 1757, + 295, + 1757 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 297, + 1003, + 1404, + 1003, + 1404, + 1171, + 297, + 1171 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 297, + 560, + 1406, + 560, + 1406, + 689, + 297, + 689 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 297, + 1366, + 1405, + 1366, + 1405, + 1523, + 297, + 1523 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 296, + 1804, + 1405, + 1804, + 1405, + 1931, + 296, + 1931 + ], + "score": 0.975 + }, + { + "category_id": 1, + "poly": [ + 298, + 435, + 1406, + 435, + 1406, + 543, + 298, + 543 + ], + "score": 0.974 + }, + { + "category_id": 8, + "poly": [ + 548, + 1939, + 1151, + 1939, + 1151, + 2003, + 548, + 2003 + ], + "score": 0.954 + }, + { + "category_id": 8, + "poly": [ + 593, + 350, + 1104, + 350, + 1104, + 425, + 593, + 425 + ], + "score": 0.952 + }, + { + "category_id": 1, + "poly": [ + 292, + 775, + 1404, + 775, + 1404, + 839, + 292, + 839 + ], + "score": 0.949 + }, + { + "category_id": 1, + "poly": [ + 294, + 201, + 1404, + 201, + 1404, + 266, + 294, + 266 + ], + "score": 0.949 + }, + { + "category_id": 8, + "poly": [ + 591, + 1319, + 1105, + 1319, + 1105, + 1359, + 591, + 1359 + ], + "score": 0.947 + }, + { + "category_id": 8, + "poly": [ + 405, + 847, + 1289, + 847, + 1289, + 940, + 405, + 940 + ], + "score": 0.941 + }, + { + "category_id": 8, + "poly": [ + 650, + 1763, + 1046, + 1763, + 1046, + 1798, + 650, + 1798 + ], + "score": 0.94 + }, + { + "category_id": 1, + "poly": [ + 294, + 272, + 1401, + 272, + 1401, + 342, + 294, + 342 + ], + "score": 0.932 + }, + { + "category_id": 0, + "poly": [ + 296, + 720, + 981, + 720, + 981, + 755, + 296, + 755 + ], + "score": 0.93 + }, + { + "category_id": 1, + "poly": [ + 297, + 949, + 1390, + 949, + 1390, + 992, + 297, + 992 + ], + "score": 0.924 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1766, + 1400, + 1766, + 1400, + 1795, + 1366, + 1795 + ], + "score": 0.874 + }, + { + "category_id": 9, + "poly": [ + 1366, + 1326, + 1400, + 1326, + 1400, + 1355, + 1366, + 1355 + ], + "score": 0.871 + }, + { + "category_id": 9, + "poly": [ + 1365, + 362, + 1400, + 362, + 1400, + 392, + 1365, + 392 + ], + "score": 0.866 + }, + { + "category_id": 9, + "poly": [ + 1366, + 879, + 1400, + 879, + 1400, + 909, + 1366, + 909 + ], + "score": 0.864 + }, + { + "category_id": 9, + "poly": [ + 1365, + 1956, + 1401, + 1956, + 1401, + 1985, + 1365, + 1985 + ], + "score": 0.864 + }, + { + "category_id": 2, + "poly": [ + 841, + 2061, + 859, + 2061, + 859, + 2085, + 841, + 2085 + ], + "score": 0.748 + }, + { + "category_id": 14, + "poly": [ + 592, + 350, + 1104, + 350, + 1104, + 427, + 592, + 427 + ], + "score": 0.95, + "latex": "L _ { k \\to l } ( { \\bf X } ) _ { \\bf i } = \\sum _ { \\mu } \\sum _ { { \\bf j } } { \\bf B } _ { { \\bf i } , { \\bf j } } ^ { \\mu } { \\bf X } _ { { \\bf j } } w _ { \\mu } + \\sum _ { \\lambda } { \\bf C } _ { { \\bf i } } ^ { \\lambda } b _ { \\lambda } ," + }, + { + "category_id": 13, + "poly": [ + 1156, + 1628, + 1323, + 1628, + 1323, + 1664, + 1156, + 1664 + ], + "score": 0.93, + "latex": "\\mathbf { S } _ { i j } = a \\mathbf { P } _ { i } ^ { \\top } \\mathbf { P } _ { j }" + }, + { + "category_id": 13, + "poly": [ + 759, + 1660, + 888, + 1660, + 888, + 1698, + 759, + 1698 + ], + "score": 0.93, + "latex": "{ \\bf P } _ { i } ^ { \\top } { \\bf P } _ { j } = 1" + }, + { + "category_id": 13, + "poly": [ + 911, + 269, + 1194, + 269, + 1194, + 309, + 911, + 309 + ], + "score": 0.93, + "latex": "L _ { k \\to l } : \\mathbb { R } ^ { n ^ { k } \\times d } \\mathbb { R } ^ { n ^ { l } \\times d ^ { \\prime } }" + }, + { + "category_id": 13, + "poly": [ + 493, + 307, + 628, + 307, + 628, + 342, + 493, + 342 + ], + "score": 0.93, + "latex": "\\bar { \\mathbf { X } } \\in \\mathbb { R } ^ { n ^ { k } \\times d }" + }, + { + "category_id": 13, + "poly": [ + 770, + 1459, + 857, + 1459, + 857, + 1488, + 770, + 1488 + ], + "score": 0.93, + "latex": "{ \\pmb { \\alpha } } ^ { 1 } \\approx { \\bf I }" + }, + { + "category_id": 13, + "poly": [ + 1015, + 1835, + 1200, + 1835, + 1200, + 1868, + 1015, + 1868 + ], + "score": 0.92, + "latex": "\\mathbf { B } ^ { \\mu _ { 2 } } = \\mathbf { \\bar { 1 1 } } ^ { \\top } - \\mathbf { I }" + }, + { + "category_id": 13, + "poly": [ + 751, + 1369, + 853, + 1369, + 853, + 1397, + 751, + 1397 + ], + "score": 0.92, + "latex": "\\mathbf { B } ^ { \\mu _ { 1 } } = \\mathbf { I }" + }, + { + "category_id": 13, + "poly": [ + 400, + 1279, + 524, + 1279, + 524, + 1309, + 400, + 1309 + ], + "score": 0.92, + "latex": "\\mathbf { X } \\in \\mathbb { R } ^ { n \\times d }" + }, + { + "category_id": 14, + "poly": [ + 405, + 845, + 1291, + 845, + 1291, + 943, + 405, + 943 + ], + "score": 0.91, + "latex": "\\mathbf { M S A } ( \\mathbf { X } ) _ { i } = \\sum _ { h = 1 } ^ { H } \\sum _ { j } \\alpha _ { i j } ^ { h } \\mathbf { X } _ { j } w _ { h } ^ { V } w _ { h } ^ { O } \\mathrm { ~ w h e r e ~ } \\alpha ^ { h } = \\mathrm { s o f t m a x } \\left( \\frac { \\mathbf { X } w _ { h } ^ { Q } ( \\mathbf { X } w _ { h } ^ { K } ) ^ { \\top } } { \\sqrt { d _ { H } } } \\right) ," + }, + { + "category_id": 13, + "poly": [ + 374, + 437, + 563, + 437, + 563, + 473, + 374, + 473 + ], + "score": 0.91, + "latex": "\\mathbf { i } \\in [ n ] ^ { l } , \\mathbf { j } \\in [ n ] ^ { k }" + }, + { + "category_id": 13, + "poly": [ + 788, + 1215, + 854, + 1215, + 854, + 1245, + 788, + 1245 + ], + "score": 0.91, + "latex": "L _ { 1 1 }" + }, + { + "category_id": 13, + "poly": [ + 1112, + 1532, + 1245, + 1532, + 1245, + 1564, + 1112, + 1564 + ], + "score": 0.91, + "latex": "\\mathbf { P } \\in \\mathbb { R } ^ { n \\times d _ { p } }" + }, + { + "category_id": 14, + "poly": [ + 548, + 1935, + 1152, + 1935, + 1152, + 2004, + 548, + 2004 + ], + "score": 0.91, + "latex": "\\alpha ^ { 2 } = \\operatorname { s o f t m a x } \\left( - a \\mathbf { I } \\right) \\to { \\frac { 1 } { n - 1 } } ( \\mathbf { 1 1 } ^ { \\top } - \\mathbf { I } ) { \\mathrm { ~ a s ~ } } a \\to \\infty ." + }, + { + "category_id": 13, + "poly": [ + 540, + 472, + 654, + 472, + 654, + 507, + 540, + 507 + ], + "score": 0.91, + "latex": "\\mathbf { C } ^ { \\lambda } \\in \\mathbb { R } ^ { n ^ { l } }" + }, + { + "category_id": 13, + "poly": [ + 488, + 1138, + 644, + 1138, + 644, + 1171, + 488, + 1171 + ], + "score": 0.91, + "latex": "H = { \\mathsf { b e l l } } ( 2 k )" + }, + { + "category_id": 14, + "poly": [ + 649, + 1760, + 1050, + 1760, + 1050, + 1799, + 649, + 1799 + ], + "score": 0.91, + "latex": "\\alpha ^ { 1 } = \\operatorname { s o f t m a x } ( a \\mathbf { I } ) \\to \\mathbf { I } \\operatorname { a s } a \\to \\infty ." + }, + { + "category_id": 13, + "poly": [ + 1189, + 477, + 1268, + 477, + 1268, + 511, + 1189, + 511 + ], + "score": 0.91, + "latex": "( l + k )" + }, + { + "category_id": 13, + "poly": [ + 973, + 1665, + 1037, + 1665, + 1037, + 1696, + 973, + 1696 + ], + "score": 0.91, + "latex": "i = j" + }, + { + "category_id": 13, + "poly": [ + 345, + 474, + 486, + 474, + 486, + 507, + 345, + 507 + ], + "score": 0.91, + "latex": "\\mathbf { B } ^ { \\mu } \\in \\mathbb { R } ^ { n ^ { l + k } }" + }, + { + "category_id": 14, + "poly": [ + 590, + 1318, + 1107, + 1318, + 1107, + 1359, + 590, + 1359 + ], + "score": 0.91, + "latex": "L _ { 1 \\to 1 } ( \\mathbf { X } ) = \\mathbf { I } \\mathbf { X } w _ { 1 } + ( \\mathbf { 1 1 } ^ { \\top } - \\mathbf { I } ) \\mathbf { X } w _ { 2 } + \\mathbf { 1 } b ^ { \\top } ." + }, + { + "category_id": 13, + "poly": [ + 346, + 1244, + 535, + 1244, + 535, + 1277, + 346, + 1277 + ], + "score": 0.91, + "latex": "\\mathbf { B } ^ { \\mu _ { 2 } } = \\mathbf { 1 1 } ^ { \\top } - \\mathbf { I }" + }, + { + "category_id": 13, + "poly": [ + 773, + 435, + 914, + 435, + 914, + 475, + 773, + 475 + ], + "score": 0.9, + "latex": "w _ { \\boldsymbol { \\mu } } \\in \\mathbb { R } ^ { d \\times d ^ { \\prime } }" + }, + { + "category_id": 13, + "poly": [ + 1039, + 1246, + 1145, + 1246, + 1145, + 1275, + 1039, + 1275 + ], + "score": 0.9, + "latex": "\\mathbf { C } ^ { \\lambda _ { 1 } } = \\mathbf { 1 }" + }, + { + "category_id": 13, + "poly": [ + 824, + 1066, + 1112, + 1066, + 1112, + 1103, + 824, + 1103 + ], + "score": 0.9, + "latex": "L _ { k \\to k } : \\mathbb { R } ^ { n ^ { k } \\times d } \\to \\mathbb { R } ^ { n ^ { k } \\times d }" + }, + { + "category_id": 13, + "poly": [ + 1109, + 1003, + 1147, + 1003, + 1147, + 1034, + 1109, + 1034 + ], + "score": 0.9, + "latex": "\\alpha ^ { h }" + }, + { + "category_id": 13, + "poly": [ + 354, + 1631, + 426, + 1631, + 426, + 1659, + 354, + 1659 + ], + "score": 0.9, + "latex": "a > 0" + }, + { + "category_id": 13, + "poly": [ + 298, + 1139, + 367, + 1139, + 367, + 1169, + 298, + 1169 + ], + "score": 0.89, + "latex": "L _ { k k }" + }, + { + "category_id": 13, + "poly": [ + 627, + 956, + 665, + 956, + 665, + 986, + 627, + 986 + ], + "score": 0.89, + "latex": "d _ { H }" + }, + { + "category_id": 13, + "poly": [ + 1018, + 1105, + 1113, + 1105, + 1113, + 1135, + 1018, + 1135 + ], + "score": 0.89, + "latex": "N = n ^ { k }" + }, + { + "category_id": 13, + "poly": [ + 766, + 1868, + 1000, + 1868, + 1000, + 1901, + 766, + 1901 + ], + "score": 0.89, + "latex": "\\pmb { \\alpha } ^ { 2 } = \\mathrm { s o f t m a x } ( - a \\mathbf { I } )" + }, + { + "category_id": 13, + "poly": [ + 328, + 1837, + 364, + 1837, + 364, + 1867, + 328, + 1867 + ], + "score": 0.89, + "latex": "\\alpha ^ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 297, + 1694, + 386, + 1694, + 386, + 1723, + 297, + 1723 + ], + "score": 0.89, + "latex": "\\mathbf { S } = a \\mathbf { I }" + }, + { + "category_id": 13, + "poly": [ + 1214, + 512, + 1278, + 512, + 1278, + 541, + 1214, + 541 + ], + "score": 0.89, + "latex": "L _ { k l }" + }, + { + "category_id": 13, + "poly": [ + 1336, + 510, + 1398, + 510, + 1398, + 537, + 1336, + 537 + ], + "score": 0.89, + "latex": "l = 0" + }, + { + "category_id": 13, + "poly": [ + 1168, + 653, + 1208, + 653, + 1208, + 683, + 1168, + 683 + ], + "score": 0.89, + "latex": "\\mathbf { C } ^ { \\lambda }" + }, + { + "category_id": 13, + "poly": [ + 1301, + 1214, + 1401, + 1214, + 1401, + 1241, + 1301, + 1241 + ], + "score": 0.88, + "latex": "\\mathbf { B } ^ { \\mu _ { 1 } } = \\mathbf { I }" + }, + { + "category_id": 13, + "poly": [ + 1083, + 1565, + 1399, + 1565, + 1399, + 1602, + 1083, + 1602 + ], + "score": 0.88, + "latex": "\\mathbf { X } ^ { i n } = [ \\mathbf { X } , \\mathbf { P } ] \\in \\mathbb { R } ^ { n \\times ( d + d _ { p } ) }" + }, + { + "category_id": 13, + "poly": [ + 1119, + 1367, + 1154, + 1367, + 1154, + 1397, + 1119, + 1397 + ], + "score": 0.88, + "latex": "\\alpha ^ { 1 }" + }, + { + "category_id": 13, + "poly": [ + 464, + 1698, + 557, + 1698, + 557, + 1723, + 464, + 1723 + ], + "score": 0.88, + "latex": "a \\infty" + }, + { + "category_id": 13, + "poly": [ + 1361, + 1599, + 1401, + 1599, + 1401, + 1632, + 1361, + 1632 + ], + "score": 0.88, + "latex": "\\sqrt { a }" + }, + { + "category_id": 13, + "poly": [ + 926, + 434, + 1030, + 434, + 1030, + 472, + 926, + 472 + ], + "score": 0.86, + "latex": "b _ { \\lambda } \\in \\mathbb { R } ^ { d ^ { \\prime } }" + }, + { + "category_id": 13, + "poly": [ + 711, + 1036, + 750, + 1036, + 750, + 1064, + 711, + 1064 + ], + "score": 0.86, + "latex": "\\mathbf { B } ^ { \\mu }" + }, + { + "category_id": 13, + "poly": [ + 298, + 1904, + 386, + 1904, + 386, + 1926, + 298, + 1926 + ], + "score": 0.85, + "latex": "a \\to \\infty" + }, + { + "category_id": 13, + "poly": [ + 567, + 656, + 607, + 656, + 607, + 683, + 567, + 683 + ], + "score": 0.85, + "latex": "\\mathbf { B } ^ { \\mu }" + }, + { + "category_id": 13, + "poly": [ + 371, + 957, + 399, + 957, + 399, + 983, + 371, + 983 + ], + "score": 0.83, + "latex": "H" + }, + { + "category_id": 13, + "poly": [ + 874, + 1277, + 1048, + 1277, + 1048, + 1313, + 874, + 1313 + ], + "score": 0.8, + "latex": "w _ { 1 } , w _ { 2 } \\in \\mathbb { R } ^ { d \\times d }" + }, + { + "category_id": 13, + "poly": [ + 796, + 1248, + 894, + 1248, + 894, + 1281, + 796, + 1281 + ], + "score": 0.8, + "latex": "1 ( 1 ) = 1" + }, + { + "category_id": 13, + "poly": [ + 1218, + 1139, + 1258, + 1139, + 1258, + 1165, + 1218, + 1165 + ], + "score": 0.8, + "latex": "\\mathbf { B } ^ { \\mu }" + }, + { + "category_id": 13, + "poly": [ + 904, + 1806, + 930, + 1806, + 930, + 1833, + 904, + 1833 + ], + "score": 0.79, + "latex": "\\mathbf { P }" + }, + { + "category_id": 13, + "poly": [ + 407, + 315, + 425, + 315, + 425, + 341, + 407, + 341 + ], + "score": 0.79, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 837, + 1903, + 849, + 1903, + 849, + 1926, + 837, + 1926 + ], + "score": 0.78, + "latex": "i" + }, + { + "category_id": 13, + "poly": [ + 822, + 1402, + 834, + 1402, + 834, + 1427, + 822, + 1427 + ], + "score": 0.77, + "latex": "i" + }, + { + "category_id": 13, + "poly": [ + 1222, + 1602, + 1247, + 1602, + 1247, + 1627, + 1222, + 1627 + ], + "score": 0.76, + "latex": "\\mathbf { P }" + }, + { + "category_id": 13, + "poly": [ + 748, + 1543, + 768, + 1543, + 768, + 1563, + 748, + 1563 + ], + "score": 0.76, + "latex": "n" + }, + { + "category_id": 13, + "poly": [ + 589, + 512, + 607, + 512, + 607, + 537, + 589, + 537 + ], + "score": 0.75, + "latex": "\\lambda" + }, + { + "category_id": 13, + "poly": [ + 652, + 1901, + 665, + 1901, + 665, + 1926, + 652, + 1926 + ], + "score": 0.75, + "latex": "i" + }, + { + "category_id": 13, + "poly": [ + 873, + 1629, + 1077, + 1629, + 1077, + 1660, + 873, + 1660 + ], + "score": 0.74, + "latex": "\\bar { \\mathbf { \\alpha } } ^ { 1 } = \\bar { \\mathrm { s o f t m a x } } ( \\mathbf { S } )" + }, + { + "category_id": 13, + "poly": [ + 460, + 1402, + 472, + 1402, + 472, + 1427, + 460, + 1427 + ], + "score": 0.74, + "latex": "i" + }, + { + "category_id": 13, + "poly": [ + 678, + 1104, + 958, + 1104, + 958, + 1136, + 678, + 1136 + ], + "score": 0.72, + "latex": "\\mathbf { M S A } : \\mathbb { R } ^ { N \\times d ^ { \\prime } } \\mathbb { R } ^ { N \\times d ^ { \\prime } }" + }, + { + "category_id": 13, + "poly": [ + 1059, + 1278, + 1141, + 1278, + 1141, + 1310, + 1059, + 1310 + ], + "score": 0.71, + "latex": "b \\in \\mathbb { R } ^ { d }" + }, + { + "category_id": 13, + "poly": [ + 1019, + 1214, + 1150, + 1214, + 1150, + 1246, + 1019, + 1246 + ], + "score": 0.68, + "latex": "{ \\mathsf { b e l l } } ( 2 ) = 2" + }, + { + "category_id": 13, + "poly": [ + 782, + 1431, + 808, + 1431, + 808, + 1457, + 782, + 1457 + ], + "score": 0.66, + "latex": "\\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 1376, + 1839, + 1401, + 1839, + 1401, + 1866, + 1376, + 1866 + ], + "score": 0.65, + "latex": "\\mathbf { P }" + }, + { + "category_id": 13, + "poly": [ + 858, + 950, + 1064, + 950, + 1064, + 991, + 858, + 991 + ], + "score": 0.65, + "latex": "w _ { h } ^ { Q } , w _ { h } ^ { K } \\in \\mathbb { R } ^ { d \\times d _ { H } }" + }, + { + "category_id": 13, + "poly": [ + 1279, + 622, + 1404, + 622, + 1404, + 656, + 1279, + 656 + ], + "score": 0.62, + "latex": "\\mathrm { b e l l } ( k + l )" + }, + { + "category_id": 13, + "poly": [ + 518, + 516, + 538, + 516, + 538, + 542, + 518, + 542 + ], + "score": 0.62, + "latex": "\\mu" + }, + { + "category_id": 13, + "poly": [ + 1025, + 1601, + 1052, + 1601, + 1052, + 1627, + 1025, + 1627 + ], + "score": 0.62, + "latex": "\\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 854, + 950, + 1385, + 950, + 1385, + 991, + 854, + 991 + ], + "score": 0.58, + "latex": "w _ { h } ^ { Q } , w _ { h } ^ { K } \\in \\mathbb R ^ { d \\times d _ { H } } , w _ { h } ^ { V } \\in \\mathbb R ^ { d \\times d _ { v } } w _ { h } ^ { O } \\in \\mathbb R ^ { d _ { v } \\times d } ." + }, + { + "category_id": 13, + "poly": [ + 372, + 1491, + 398, + 1491, + 398, + 1518, + 372, + 1518 + ], + "score": 0.55, + "latex": "\\mathbf { X }" + }, + { + "category_id": 13, + "poly": [ + 623, + 1664, + 648, + 1664, + 648, + 1692, + 623, + 1692 + ], + "score": 0.54, + "latex": "\\mathbf { P }" + }, + { + "category_id": 13, + "poly": [ + 1390, + 480, + 1401, + 480, + 1401, + 506, + 1390, + 506 + ], + "score": 0.52, + "latex": "l" + }, + { + "category_id": 13, + "poly": [ + 864, + 656, + 894, + 656, + 894, + 689, + 864, + 689 + ], + "score": 0.47, + "latex": "( l )" + }, + { + "category_id": 13, + "poly": [ + 820, + 655, + 895, + 655, + 895, + 689, + 820, + 689 + ], + "score": 0.43, + "latex": "\\mathsf { b e l l } ( l )" + }, + { + "category_id": 13, + "poly": [ + 1169, + 1491, + 1196, + 1491, + 1196, + 1518, + 1169, + 1518 + ], + "score": 0.42, + "latex": "\\mathbf { X }" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 720.0, + 983.0, + 720.0, + 983.0, + 760.0, + 293.0, + 760.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2092.0, + 838.0, + 2092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1179.0, + 1406.0, + 1179.0, + 1406.0, + 1220.0, + 295.0, + 1220.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1212.0, + 787.0, + 1212.0, + 787.0, + 1249.0, + 295.0, + 1249.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 855.0, + 1212.0, + 1018.0, + 1212.0, + 1018.0, + 1249.0, + 855.0, + 1249.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1151.0, + 1212.0, + 1300.0, + 1212.0, + 1300.0, + 1249.0, + 1151.0, + 1249.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 1212.0, + 1407.0, + 1212.0, + 1407.0, + 1249.0, + 1402.0, + 1249.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1243.0, + 345.0, + 1243.0, + 345.0, + 1283.0, + 295.0, + 1283.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 536.0, + 1243.0, + 795.0, + 1243.0, + 795.0, + 1283.0, + 536.0, + 1283.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 895.0, + 1243.0, + 1038.0, + 1243.0, + 1038.0, + 1283.0, + 895.0, + 1283.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1146.0, + 1243.0, + 1405.0, + 1243.0, + 1405.0, + 1283.0, + 1146.0, + 1283.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1277.0, + 399.0, + 1277.0, + 399.0, + 1315.0, + 293.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 525.0, + 1277.0, + 873.0, + 1277.0, + 873.0, + 1315.0, + 525.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1049.0, + 1277.0, + 1058.0, + 1277.0, + 1058.0, + 1315.0, + 1049.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1142.0, + 1277.0, + 1155.0, + 1277.0, + 1155.0, + 1315.0, + 1142.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1531.0, + 747.0, + 1531.0, + 747.0, + 1570.0, + 292.0, + 1570.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 769.0, + 1531.0, + 1111.0, + 1531.0, + 1111.0, + 1570.0, + 769.0, + 1570.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1246.0, + 1531.0, + 1407.0, + 1531.0, + 1407.0, + 1570.0, + 1246.0, + 1570.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1565.0, + 1082.0, + 1565.0, + 1082.0, + 1603.0, + 289.0, + 1603.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1400.0, + 1565.0, + 1409.0, + 1565.0, + 1409.0, + 1603.0, + 1400.0, + 1603.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1597.0, + 1024.0, + 1597.0, + 1024.0, + 1636.0, + 292.0, + 1636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1053.0, + 1597.0, + 1221.0, + 1597.0, + 1221.0, + 1636.0, + 1053.0, + 1636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1248.0, + 1597.0, + 1360.0, + 1597.0, + 1360.0, + 1636.0, + 1248.0, + 1636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 1597.0, + 1407.0, + 1597.0, + 1407.0, + 1636.0, + 1402.0, + 1636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1627.0, + 353.0, + 1627.0, + 353.0, + 1667.0, + 292.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 427.0, + 1627.0, + 872.0, + 1627.0, + 872.0, + 1667.0, + 427.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1078.0, + 1627.0, + 1155.0, + 1627.0, + 1155.0, + 1667.0, + 1078.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1324.0, + 1627.0, + 1411.0, + 1627.0, + 1411.0, + 1667.0, + 1324.0, + 1667.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1660.0, + 622.0, + 1660.0, + 622.0, + 1699.0, + 294.0, + 1699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 649.0, + 1660.0, + 758.0, + 1660.0, + 758.0, + 1699.0, + 649.0, + 1699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 889.0, + 1660.0, + 972.0, + 1660.0, + 972.0, + 1699.0, + 889.0, + 1699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1038.0, + 1660.0, + 1407.0, + 1660.0, + 1407.0, + 1699.0, + 1038.0, + 1699.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1691.0, + 296.0, + 1691.0, + 296.0, + 1732.0, + 292.0, + 1732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 1691.0, + 463.0, + 1691.0, + 463.0, + 1732.0, + 387.0, + 1732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 558.0, + 1691.0, + 1407.0, + 1691.0, + 1407.0, + 1732.0, + 558.0, + 1732.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1720.0, + 1081.0, + 1720.0, + 1081.0, + 1763.0, + 293.0, + 1763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1000.0, + 1108.0, + 1000.0, + 1108.0, + 1040.0, + 294.0, + 1040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1148.0, + 1000.0, + 1405.0, + 1000.0, + 1405.0, + 1040.0, + 1148.0, + 1040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1036.0, + 710.0, + 1036.0, + 710.0, + 1069.0, + 294.0, + 1069.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 751.0, + 1036.0, + 1404.0, + 1036.0, + 1404.0, + 1069.0, + 751.0, + 1069.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 1063.0, + 823.0, + 1063.0, + 823.0, + 1110.0, + 288.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1113.0, + 1063.0, + 1411.0, + 1063.0, + 1411.0, + 1110.0, + 1113.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1098.0, + 677.0, + 1098.0, + 677.0, + 1143.0, + 291.0, + 1143.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 959.0, + 1098.0, + 1017.0, + 1098.0, + 1017.0, + 1143.0, + 959.0, + 1143.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1114.0, + 1098.0, + 1406.0, + 1098.0, + 1406.0, + 1143.0, + 1114.0, + 1143.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1135.0, + 297.0, + 1135.0, + 297.0, + 1172.0, + 294.0, + 1172.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 1135.0, + 487.0, + 1135.0, + 487.0, + 1172.0, + 368.0, + 1172.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 645.0, + 1135.0, + 1217.0, + 1135.0, + 1217.0, + 1172.0, + 645.0, + 1172.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1259.0, + 1135.0, + 1277.0, + 1135.0, + 1277.0, + 1172.0, + 1259.0, + 1172.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 562.0, + 1407.0, + 562.0, + 1407.0, + 596.0, + 295.0, + 596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 591.0, + 1406.0, + 591.0, + 1406.0, + 627.0, + 292.0, + 627.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 621.0, + 1278.0, + 621.0, + 1278.0, + 658.0, + 294.0, + 658.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1405.0, + 621.0, + 1408.0, + 621.0, + 1408.0, + 658.0, + 1405.0, + 658.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 653.0, + 566.0, + 653.0, + 566.0, + 690.0, + 292.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 608.0, + 653.0, + 819.0, + 653.0, + 819.0, + 690.0, + 608.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 896.0, + 653.0, + 1167.0, + 653.0, + 1167.0, + 690.0, + 896.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1209.0, + 653.0, + 1350.0, + 653.0, + 1350.0, + 690.0, + 1209.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1367.0, + 750.0, + 1367.0, + 750.0, + 1400.0, + 296.0, + 1400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 854.0, + 1367.0, + 1118.0, + 1367.0, + 1118.0, + 1400.0, + 854.0, + 1400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1155.0, + 1367.0, + 1402.0, + 1367.0, + 1402.0, + 1400.0, + 1155.0, + 1400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1399.0, + 459.0, + 1399.0, + 459.0, + 1434.0, + 292.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 473.0, + 1399.0, + 821.0, + 1399.0, + 821.0, + 1434.0, + 473.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 835.0, + 1399.0, + 1405.0, + 1399.0, + 1405.0, + 1434.0, + 835.0, + 1434.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1428.0, + 781.0, + 1428.0, + 781.0, + 1465.0, + 292.0, + 1465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 809.0, + 1428.0, + 1406.0, + 1428.0, + 1406.0, + 1465.0, + 809.0, + 1465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1458.0, + 769.0, + 1458.0, + 769.0, + 1495.0, + 294.0, + 1495.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 858.0, + 1458.0, + 1405.0, + 1458.0, + 1405.0, + 1495.0, + 858.0, + 1495.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1490.0, + 371.0, + 1490.0, + 371.0, + 1527.0, + 295.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 399.0, + 1490.0, + 1168.0, + 1490.0, + 1168.0, + 1527.0, + 399.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1197.0, + 1490.0, + 1410.0, + 1490.0, + 1410.0, + 1527.0, + 1197.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1802.0, + 903.0, + 1802.0, + 903.0, + 1842.0, + 291.0, + 1842.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 931.0, + 1802.0, + 1405.0, + 1802.0, + 1405.0, + 1842.0, + 931.0, + 1842.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1836.0, + 327.0, + 1836.0, + 327.0, + 1872.0, + 293.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 365.0, + 1836.0, + 1014.0, + 1836.0, + 1014.0, + 1872.0, + 365.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1201.0, + 1836.0, + 1375.0, + 1836.0, + 1375.0, + 1872.0, + 1201.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 1836.0, + 1406.0, + 1836.0, + 1406.0, + 1872.0, + 1402.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1867.0, + 765.0, + 1867.0, + 765.0, + 1905.0, + 293.0, + 1905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1001.0, + 1867.0, + 1406.0, + 1867.0, + 1406.0, + 1905.0, + 1001.0, + 1905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1897.0, + 297.0, + 1897.0, + 297.0, + 1935.0, + 290.0, + 1935.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 387.0, + 1897.0, + 651.0, + 1897.0, + 651.0, + 1935.0, + 387.0, + 1935.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 666.0, + 1897.0, + 836.0, + 1897.0, + 836.0, + 1935.0, + 666.0, + 1935.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 850.0, + 1897.0, + 1408.0, + 1897.0, + 1408.0, + 1935.0, + 850.0, + 1935.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 428.0, + 373.0, + 428.0, + 373.0, + 483.0, + 290.0, + 483.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 564.0, + 428.0, + 772.0, + 428.0, + 772.0, + 483.0, + 564.0, + 483.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 915.0, + 428.0, + 925.0, + 428.0, + 925.0, + 483.0, + 915.0, + 483.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1031.0, + 428.0, + 1412.0, + 428.0, + 1412.0, + 483.0, + 1031.0, + 483.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 487.0, + 469.0, + 539.0, + 469.0, + 539.0, + 518.0, + 487.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 655.0, + 469.0, + 1188.0, + 469.0, + 1188.0, + 518.0, + 655.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1269.0, + 469.0, + 1389.0, + 469.0, + 1389.0, + 518.0, + 1269.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1402.0, + 469.0, + 1408.0, + 469.0, + 1408.0, + 518.0, + 1402.0, + 518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 509.0, + 517.0, + 509.0, + 517.0, + 545.0, + 293.0, + 545.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 539.0, + 509.0, + 588.0, + 509.0, + 588.0, + 545.0, + 539.0, + 545.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 608.0, + 509.0, + 1213.0, + 509.0, + 1213.0, + 545.0, + 608.0, + 545.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1279.0, + 509.0, + 1335.0, + 509.0, + 1335.0, + 545.0, + 1279.0, + 545.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1399.0, + 509.0, + 1409.0, + 509.0, + 1409.0, + 545.0, + 1399.0, + 545.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 467.5, + 497.0, + 467.5, + 497.0, + 514.5, + 290.0, + 514.5 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 778.0, + 1406.0, + 778.0, + 1406.0, + 810.0, + 296.0, + 810.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 807.0, + 1152.0, + 807.0, + 1152.0, + 839.0, + 296.0, + 839.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 203.0, + 1407.0, + 203.0, + 1407.0, + 239.0, + 296.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 234.0, + 1290.0, + 234.0, + 1290.0, + 268.0, + 293.0, + 268.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 259.0, + 910.0, + 259.0, + 910.0, + 320.0, + 288.0, + 320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 259.0, + 1410.0, + 259.0, + 1410.0, + 320.0, + 1195.0, + 320.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 306.0, + 643.0, + 306.0, + 643.0, + 349.0, + 289.0, + 349.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 947.0, + 370.0, + 947.0, + 370.0, + 995.0, + 289.0, + 995.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 400.0, + 947.0, + 626.0, + 947.0, + 626.0, + 995.0, + 400.0, + 995.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 666.0, + 947.0, + 853.0, + 947.0, + 853.0, + 995.0, + 666.0, + 995.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1386.0, + 947.0, + 1395.0, + 947.0, + 1395.0, + 995.0, + 1386.0, + 995.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 4, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 311, + 1405, + 311, + 1405, + 527, + 297, + 527 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 298, + 726, + 1405, + 726, + 1405, + 887, + 298, + 887 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 298, + 1148, + 1405, + 1148, + 1405, + 1252, + 298, + 1252 + ], + "score": 0.972 + }, + { + "category_id": 1, + "poly": [ + 299, + 897, + 1404, + 897, + 1404, + 1024, + 299, + 1024 + ], + "score": 0.971 + }, + { + "category_id": 1, + "poly": [ + 298, + 1041, + 1406, + 1041, + 1406, + 1140, + 298, + 1140 + ], + "score": 0.97 + }, + { + "category_id": 1, + "poly": [ + 298, + 1461, + 1405, + 1461, + 1405, + 1554, + 298, + 1554 + ], + "score": 0.967 + }, + { + "category_id": 1, + "poly": [ + 298, + 1343, + 1405, + 1343, + 1405, + 1440, + 298, + 1440 + ], + "score": 0.966 + }, + { + "category_id": 1, + "poly": [ + 299, + 202, + 1402, + 202, + 1402, + 298, + 299, + 298 + ], + "score": 0.965 + }, + { + "category_id": 1, + "poly": [ + 298, + 1607, + 1406, + 1607, + 1406, + 1705, + 298, + 1705 + ], + "score": 0.964 + }, + { + "category_id": 1, + "poly": [ + 298, + 1913, + 1404, + 1913, + 1404, + 2008, + 298, + 2008 + ], + "score": 0.963 + }, + { + "category_id": 1, + "poly": [ + 300, + 614, + 1403, + 614, + 1403, + 709, + 300, + 709 + ], + "score": 0.961 + }, + { + "category_id": 1, + "poly": [ + 294, + 1271, + 1403, + 1271, + 1403, + 1335, + 294, + 1335 + ], + "score": 0.942 + }, + { + "category_id": 1, + "poly": [ + 296, + 1845, + 1402, + 1845, + 1402, + 1908, + 296, + 1908 + ], + "score": 0.934 + }, + { + "category_id": 0, + "poly": [ + 296, + 558, + 949, + 558, + 949, + 593, + 296, + 593 + ], + "score": 0.931 + }, + { + "category_id": 1, + "poly": [ + 298, + 1566, + 1066, + 1566, + 1066, + 1599, + 298, + 1599 + ], + "score": 0.92 + }, + { + "category_id": 1, + "poly": [ + 299, + 1763, + 1404, + 1763, + 1404, + 1826, + 299, + 1826 + ], + "score": 0.919 + }, + { + "category_id": 1, + "poly": [ + 300, + 1724, + 802, + 1724, + 802, + 1757, + 300, + 1757 + ], + "score": 0.918 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 859, + 2062, + 859, + 2085, + 840, + 2085 + ], + "score": 0.787 + }, + { + "category_id": 13, + "poly": [ + 512, + 1145, + 648, + 1145, + 648, + 1181, + 512, + 1181 + ], + "score": 0.94, + "latex": "\\mathbf { X } \\in \\mathbb { R } ^ { n ^ { k } \\times d }" + }, + { + "category_id": 13, + "poly": [ + 329, + 852, + 635, + 852, + 635, + 887, + 329, + 887 + ], + "score": 0.93, + "latex": "[ \\mathbf { X } _ { i _ { 1 } , . . . , i _ { k } } , \\mathbf { \\bar { P } } _ { i _ { 1 } } , . . . , \\mathbf { P } _ { i _ { k } } , \\mathbf { E } ^ { \\gamma } ]" + }, + { + "category_id": 13, + "poly": [ + 1262, + 722, + 1398, + 722, + 1398, + 757, + 1262, + 757 + ], + "score": 0.93, + "latex": "\\mathbf { X } \\in \\mathbb { R } ^ { n ^ { k } \\times d }" + }, + { + "category_id": 13, + "poly": [ + 1030, + 898, + 1180, + 898, + 1180, + 931, + 1030, + 931 + ], + "score": 0.92, + "latex": "\\left[ { \\bf X } _ { i } , { \\bf P } _ { i } , { \\bf E } ^ { \\gamma _ { 1 } } \\right]" + }, + { + "category_id": 13, + "poly": [ + 715, + 853, + 880, + 853, + 880, + 886, + 715, + 886 + ], + "score": 0.92, + "latex": "( i _ { 1 } , . . . , i _ { k } ) \\in \\gamma" + }, + { + "category_id": 13, + "poly": [ + 943, + 1183, + 1071, + 1183, + 1071, + 1218, + 943, + 1218 + ], + "score": 0.92, + "latex": "\\mathbf { B } ^ { \\mu } \\in \\mathbb { R } ^ { n ^ { 2 k } }" + }, + { + "category_id": 13, + "poly": [ + 625, + 233, + 810, + 233, + 810, + 266, + 625, + 266 + ], + "score": 0.92, + "latex": "\\mathbf { B } ^ { \\mu _ { 2 } } = \\mathbf { 1 1 } ^ { \\top } - \\bar { \\mathbf { I } }" + }, + { + "category_id": 13, + "poly": [ + 824, + 959, + 1030, + 959, + 1030, + 994, + 824, + 994 + ], + "score": 0.92, + "latex": "[ { \\bf X } _ { i j } , { \\bf P } _ { i } , { \\bf P } _ { j } , { \\bf E } ^ { \\gamma _ { 2 } } ]" + }, + { + "category_id": 13, + "poly": [ + 1039, + 1037, + 1306, + 1037, + 1306, + 1073, + 1039, + 1073 + ], + "score": 0.92, + "latex": "\\mathbf { X } ^ { i n } \\in \\mathbb { R } ^ { n ^ { k } \\times ( d + k d _ { p } + d _ { e } ) }" + }, + { + "category_id": 13, + "poly": [ + 535, + 1340, + 675, + 1340, + 675, + 1376, + 535, + 1376 + ], + "score": 0.92, + "latex": "\\mathbf { X } \\in \\mathbb { R } ^ { n ^ { k } \\times d }" + }, + { + "category_id": 13, + "poly": [ + 1127, + 792, + 1301, + 792, + 1301, + 824, + 1127, + 824 + ], + "score": 0.92, + "latex": "\\mathbf { E } ^ { \\gamma _ { 1 } } , . . . , \\mathbf { E } ^ { \\gamma _ { \\mathrm { b e l l } } ( k ) }" + }, + { + "category_id": 13, + "poly": [ + 402, + 959, + 459, + 959, + 459, + 992, + 402, + 992 + ], + "score": 0.92, + "latex": "( i , j )" + }, + { + "category_id": 13, + "poly": [ + 526, + 1604, + 663, + 1604, + 663, + 1640, + 526, + 1640 + ], + "score": 0.92, + "latex": "\\mathbf { X } \\in \\mathbb { R } ^ { n ^ { k } \\times d }" + }, + { + "category_id": 13, + "poly": [ + 535, + 1075, + 800, + 1075, + 800, + 1107, + 535, + 1107 + ], + "score": 0.91, + "latex": "w ^ { i n } \\in \\mathbb { R } ^ { ( d + k d _ { p } + d _ { e } ) \\times d \\tau }" + }, + { + "category_id": 13, + "poly": [ + 645, + 789, + 822, + 789, + 822, + 821, + 645, + 821 + ], + "score": 0.91, + "latex": "\\mathbf { E } \\in \\mathbb { R } ^ { \\mathrm { b e l l } ( k ) \\times d _ { e } }" + }, + { + "category_id": 13, + "poly": [ + 712, + 1376, + 805, + 1376, + 805, + 1406, + 712, + 1406 + ], + "score": 0.91, + "latex": "\\mathbf { X } ^ { i n } w ^ { i n }" + }, + { + "category_id": 13, + "poly": [ + 1324, + 1642, + 1403, + 1642, + 1403, + 1675, + 1324, + 1675 + ], + "score": 0.91, + "latex": "F _ { k } ( \\mathbf { X } )" + }, + { + "category_id": 13, + "poly": [ + 298, + 1408, + 412, + 1408, + 412, + 1441, + 298, + 1441 + ], + "score": 0.91, + "latex": "L _ { k \\to k } ( \\mathbf X )" + }, + { + "category_id": 13, + "poly": [ + 1198, + 929, + 1402, + 929, + 1402, + 961, + 1198, + 961 + ], + "score": 0.91, + "latex": "\\left[ { \\bf X } _ { i i } , { \\bf P } _ { i } , { \\bf P } _ { i } , { \\bf E } ^ { \\gamma _ { 1 } } \\right]" + }, + { + "category_id": 13, + "poly": [ + 298, + 405, + 367, + 405, + 367, + 435, + 298, + 435 + ], + "score": 0.91, + "latex": "L _ { k k }" + }, + { + "category_id": 13, + "poly": [ + 523, + 344, + 590, + 344, + 590, + 374, + 523, + 374 + ], + "score": 0.91, + "latex": "L _ { 1 1 }" + }, + { + "category_id": 13, + "poly": [ + 759, + 991, + 879, + 991, + 879, + 1020, + 759, + 1020 + ], + "score": 0.91, + "latex": "\\mathbf { E } ^ { \\nu } = \\mathbf { E } ^ { \\tilde { \\gamma } _ { 1 } }" + }, + { + "category_id": 13, + "poly": [ + 1058, + 990, + 1177, + 990, + 1177, + 1020, + 1058, + 1020 + ], + "score": 0.91, + "latex": "\\mathbf { E } ^ { \\mathcal { E } } = \\mathbf { E } ^ { \\gamma _ { 2 } }" + }, + { + "category_id": 13, + "poly": [ + 361, + 1221, + 430, + 1221, + 430, + 1251, + 361, + 1251 + ], + "score": 0.91, + "latex": "L _ { k k }" + }, + { + "category_id": 13, + "poly": [ + 864, + 929, + 918, + 929, + 918, + 961, + 864, + 961 + ], + "score": 0.91, + "latex": "( i , i )" + }, + { + "category_id": 13, + "poly": [ + 643, + 757, + 786, + 757, + 786, + 787, + 643, + 787 + ], + "score": 0.9, + "latex": "\\mathbf { P } \\in \\mathbb { R } ^ { n \\times d _ { p } }" + }, + { + "category_id": 13, + "poly": [ + 466, + 1187, + 559, + 1187, + 559, + 1218, + 466, + 1218 + ], + "score": 0.9, + "latex": "\\mathbf { X } ^ { i n } w ^ { i n }" + }, + { + "category_id": 13, + "poly": [ + 493, + 1641, + 587, + 1641, + 587, + 1671, + 493, + 1671 + ], + "score": 0.9, + "latex": "\\mathbf { X } ^ { i n } w ^ { i n }" + }, + { + "category_id": 13, + "poly": [ + 298, + 495, + 367, + 495, + 367, + 526, + 298, + 526 + ], + "score": 0.9, + "latex": "L _ { k k }" + }, + { + "category_id": 13, + "poly": [ + 538, + 899, + 608, + 899, + 608, + 927, + 538, + 927 + ], + "score": 0.89, + "latex": "k = 1" + }, + { + "category_id": 13, + "poly": [ + 1184, + 647, + 1253, + 647, + 1253, + 675, + 1184, + 675 + ], + "score": 0.89, + "latex": "k > 2" + }, + { + "category_id": 13, + "poly": [ + 404, + 647, + 474, + 647, + 474, + 675, + 404, + 675 + ], + "score": 0.89, + "latex": "k = 2" + }, + { + "category_id": 13, + "poly": [ + 628, + 929, + 698, + 929, + 698, + 957, + 628, + 957 + ], + "score": 0.89, + "latex": "k = 2" + }, + { + "category_id": 13, + "poly": [ + 1224, + 202, + 1260, + 202, + 1260, + 232, + 1224, + 232 + ], + "score": 0.89, + "latex": "\\alpha ^ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 1175, + 1079, + 1212, + 1079, + 1212, + 1110, + 1175, + 1110 + ], + "score": 0.88, + "latex": "d \\tau" + }, + { + "category_id": 13, + "poly": [ + 918, + 1151, + 966, + 1151, + 966, + 1180, + 918, + 1180 + ], + "score": 0.88, + "latex": "\\mathbf { X } ^ { i n }" + }, + { + "category_id": 13, + "poly": [ + 1262, + 1150, + 1300, + 1150, + 1300, + 1182, + 1262, + 1182 + ], + "score": 0.88, + "latex": "\\pmb { \\alpha } ^ { h }" + }, + { + "category_id": 13, + "poly": [ + 932, + 1610, + 981, + 1610, + 981, + 1640, + 932, + 1640 + ], + "score": 0.87, + "latex": "\\mathbf { X } ^ { i n }" + }, + { + "category_id": 13, + "poly": [ + 954, + 1345, + 1002, + 1345, + 1002, + 1376, + 954, + 1376 + ], + "score": 0.87, + "latex": "\\mathbf { X } ^ { i n }" + }, + { + "category_id": 13, + "poly": [ + 559, + 959, + 633, + 959, + 633, + 991, + 559, + 991 + ], + "score": 0.86, + "latex": "( i \\neq j )" + }, + { + "category_id": 13, + "poly": [ + 694, + 1109, + 786, + 1109, + 786, + 1137, + 694, + 1137 + ], + "score": 0.86, + "latex": "\\mathbf { X } ^ { i n } w ^ { i n }" + }, + { + "category_id": 13, + "poly": [ + 1155, + 730, + 1172, + 730, + 1172, + 756, + 1155, + 756 + ], + "score": 0.83, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 658, + 824, + 677, + 824, + 677, + 850, + 658, + 850 + ], + "score": 0.83, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 943, + 1046, + 961, + 1046, + 961, + 1072, + 943, + 1072 + ], + "score": 0.82, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 825, + 1304, + 843, + 1304, + 843, + 1330, + 825, + 1330 + ], + "score": 0.81, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1380, + 617, + 1398, + 617, + 1398, + 643, + 1380, + 643 + ], + "score": 0.81, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1191, + 1848, + 1209, + 1848, + 1209, + 1874, + 1191, + 1874 + ], + "score": 0.81, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 876, + 828, + 895, + 828, + 895, + 854, + 876, + 854 + ], + "score": 0.79, + "latex": "\\gamma" + }, + { + "category_id": 13, + "poly": [ + 1241, + 1643, + 1260, + 1643, + 1260, + 1669, + 1241, + 1669 + ], + "score": 0.77, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1012, + 404, + 1065, + 404, + 1065, + 432, + 1012, + 432 + ], + "score": 0.77, + "latex": "\\mathbf { B } ^ { \\mu } ?" + }, + { + "category_id": 13, + "poly": [ + 740, + 902, + 752, + 902, + 752, + 926, + 740, + 926 + ], + "score": 0.77, + "latex": "i" + }, + { + "category_id": 13, + "poly": [ + 1156, + 765, + 1176, + 765, + 1176, + 786, + 1156, + 786 + ], + "score": 0.77, + "latex": "n" + }, + { + "category_id": 13, + "poly": [ + 1076, + 435, + 1095, + 435, + 1095, + 461, + 1076, + 461 + ], + "score": 0.77, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1176, + 1190, + 1194, + 1190, + 1194, + 1217, + 1176, + 1217 + ], + "score": 0.76, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 297, + 1947, + 316, + 1947, + 316, + 1974, + 297, + 1974 + ], + "score": 0.75, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 909, + 373, + 959, + 373, + 959, + 405, + 909, + 405 + ], + "score": 0.73, + "latex": "( 2 k )" + }, + { + "category_id": 13, + "poly": [ + 297, + 1795, + 316, + 1795, + 316, + 1821, + 297, + 1821 + ], + "score": 0.72, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1109, + 1379, + 1127, + 1379, + 1127, + 1405, + 1109, + 1405 + ], + "score": 0.68, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1304, + 1613, + 1327, + 1613, + 1327, + 1639, + 1304, + 1639 + ], + "score": 0.68, + "latex": "T" + }, + { + "category_id": 13, + "poly": [ + 1042, + 1273, + 1091, + 1273, + 1091, + 1304, + 1042, + 1304 + ], + "score": 0.68, + "latex": "( 2 k )" + }, + { + "category_id": 13, + "poly": [ + 1353, + 1347, + 1403, + 1347, + 1403, + 1380, + 1353, + 1380 + ], + "score": 0.66, + "latex": "( 2 k )" + }, + { + "category_id": 13, + "poly": [ + 790, + 313, + 816, + 313, + 816, + 340, + 790, + 340 + ], + "score": 0.6, + "latex": "\\mathbf { P }" + }, + { + "category_id": 13, + "poly": [ + 979, + 792, + 1060, + 792, + 1060, + 826, + 979, + 826 + ], + "score": 0.56, + "latex": "\\mathsf { b e l l } ( k )" + }, + { + "category_id": 13, + "poly": [ + 592, + 1796, + 610, + 1796, + 610, + 1821, + 592, + 1821 + ], + "score": 0.54, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1023, + 792, + 1060, + 792, + 1060, + 825, + 1023, + 825 + ], + "score": 0.52, + "latex": "( k )" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 551.0, + 951.0, + 551.0, + 951.0, + 603.0, + 292.0, + 603.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2060.0, + 862.0, + 2060.0, + 862.0, + 2091.0, + 840.0, + 2091.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 309.0, + 789.0, + 309.0, + 789.0, + 349.0, + 294.0, + 349.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 817.0, + 309.0, + 1406.0, + 309.0, + 1406.0, + 349.0, + 817.0, + 349.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 344.0, + 522.0, + 344.0, + 522.0, + 378.0, + 295.0, + 378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 591.0, + 344.0, + 1403.0, + 344.0, + 1403.0, + 378.0, + 591.0, + 378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 372.0, + 908.0, + 372.0, + 908.0, + 407.0, + 295.0, + 407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 960.0, + 372.0, + 1405.0, + 372.0, + 1405.0, + 407.0, + 960.0, + 407.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 403.0, + 297.0, + 403.0, + 297.0, + 439.0, + 294.0, + 439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 403.0, + 1011.0, + 403.0, + 1011.0, + 439.0, + 368.0, + 439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1066.0, + 403.0, + 1407.0, + 403.0, + 1407.0, + 439.0, + 1066.0, + 439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 431.0, + 1075.0, + 431.0, + 1075.0, + 471.0, + 292.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1096.0, + 431.0, + 1405.0, + 431.0, + 1405.0, + 471.0, + 1096.0, + 471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 461.0, + 1406.0, + 461.0, + 1406.0, + 500.0, + 292.0, + 500.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 368.0, + 494.0, + 1153.0, + 494.0, + 1153.0, + 530.0, + 368.0, + 530.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 722.0, + 1154.0, + 722.0, + 1154.0, + 763.0, + 291.0, + 763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1173.0, + 722.0, + 1261.0, + 722.0, + 1261.0, + 763.0, + 1173.0, + 763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1399.0, + 722.0, + 1410.0, + 722.0, + 1410.0, + 763.0, + 1399.0, + 763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 749.0, + 642.0, + 749.0, + 642.0, + 797.0, + 292.0, + 797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 749.0, + 1155.0, + 749.0, + 1155.0, + 797.0, + 787.0, + 797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1177.0, + 749.0, + 1407.0, + 749.0, + 1407.0, + 797.0, + 1177.0, + 797.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 783.0, + 644.0, + 783.0, + 644.0, + 830.0, + 292.0, + 830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 823.0, + 783.0, + 978.0, + 783.0, + 978.0, + 830.0, + 823.0, + 830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1061.0, + 783.0, + 1126.0, + 783.0, + 1126.0, + 830.0, + 1061.0, + 830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1302.0, + 783.0, + 1410.0, + 783.0, + 1410.0, + 830.0, + 1302.0, + 830.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 819.0, + 657.0, + 819.0, + 657.0, + 860.0, + 293.0, + 860.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 678.0, + 819.0, + 875.0, + 819.0, + 875.0, + 860.0, + 678.0, + 860.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 896.0, + 819.0, + 1406.0, + 819.0, + 1406.0, + 860.0, + 896.0, + 860.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 847.0, + 328.0, + 847.0, + 328.0, + 894.0, + 292.0, + 894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 636.0, + 847.0, + 714.0, + 847.0, + 714.0, + 894.0, + 636.0, + 894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 881.0, + 847.0, + 894.0, + 847.0, + 894.0, + 894.0, + 881.0, + 894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1141.0, + 511.0, + 1141.0, + 511.0, + 1191.0, + 291.0, + 1191.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 649.0, + 1141.0, + 917.0, + 1141.0, + 917.0, + 1191.0, + 649.0, + 1191.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 967.0, + 1141.0, + 1261.0, + 1141.0, + 1261.0, + 1191.0, + 967.0, + 1191.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1301.0, + 1141.0, + 1408.0, + 1141.0, + 1408.0, + 1191.0, + 1301.0, + 1191.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1183.0, + 465.0, + 1183.0, + 465.0, + 1224.0, + 292.0, + 1224.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 560.0, + 1183.0, + 942.0, + 1183.0, + 942.0, + 1224.0, + 560.0, + 1224.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1074.0, + 1186.0, + 1175.0, + 1186.0, + 1175.0, + 1226.0, + 1074.0, + 1226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1195.0, + 1186.0, + 1405.0, + 1186.0, + 1405.0, + 1226.0, + 1195.0, + 1226.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1219.0, + 360.0, + 1219.0, + 360.0, + 1255.0, + 296.0, + 1255.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 431.0, + 1219.0, + 1068.0, + 1219.0, + 1068.0, + 1255.0, + 431.0, + 1255.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 898.0, + 537.0, + 898.0, + 537.0, + 931.0, + 294.0, + 931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 609.0, + 898.0, + 739.0, + 898.0, + 739.0, + 931.0, + 609.0, + 931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 753.0, + 898.0, + 1029.0, + 898.0, + 1029.0, + 931.0, + 753.0, + 931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1181.0, + 898.0, + 1405.0, + 898.0, + 1405.0, + 931.0, + 1181.0, + 931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 926.0, + 627.0, + 926.0, + 627.0, + 962.0, + 293.0, + 962.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 699.0, + 926.0, + 863.0, + 926.0, + 863.0, + 962.0, + 699.0, + 962.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 919.0, + 926.0, + 1197.0, + 926.0, + 1197.0, + 962.0, + 919.0, + 962.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 957.0, + 401.0, + 957.0, + 401.0, + 992.0, + 293.0, + 992.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 460.0, + 957.0, + 558.0, + 957.0, + 558.0, + 992.0, + 460.0, + 992.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 634.0, + 957.0, + 823.0, + 957.0, + 823.0, + 992.0, + 634.0, + 992.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1031.0, + 957.0, + 1407.0, + 957.0, + 1407.0, + 992.0, + 1031.0, + 992.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 990.0, + 758.0, + 990.0, + 758.0, + 1026.0, + 293.0, + 1026.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 880.0, + 990.0, + 1057.0, + 990.0, + 1057.0, + 1026.0, + 880.0, + 1026.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1178.0, + 990.0, + 1190.0, + 990.0, + 1190.0, + 1026.0, + 1178.0, + 1026.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1035.0, + 942.0, + 1035.0, + 942.0, + 1082.0, + 291.0, + 1082.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 962.0, + 1035.0, + 1038.0, + 1035.0, + 1038.0, + 1082.0, + 962.0, + 1082.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1307.0, + 1035.0, + 1408.0, + 1035.0, + 1408.0, + 1082.0, + 1307.0, + 1082.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 1066.0, + 534.0, + 1066.0, + 534.0, + 1118.0, + 288.0, + 1118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 801.0, + 1066.0, + 1174.0, + 1066.0, + 1174.0, + 1118.0, + 801.0, + 1118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1213.0, + 1066.0, + 1413.0, + 1066.0, + 1413.0, + 1118.0, + 1213.0, + 1118.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1104.0, + 693.0, + 1104.0, + 693.0, + 1144.0, + 292.0, + 1144.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 787.0, + 1104.0, + 1304.0, + 1104.0, + 1304.0, + 1144.0, + 787.0, + 1144.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1460.0, + 1405.0, + 1460.0, + 1405.0, + 1498.0, + 293.0, + 1498.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1491.0, + 1405.0, + 1491.0, + 1405.0, + 1527.0, + 293.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1521.0, + 1231.0, + 1521.0, + 1231.0, + 1557.0, + 295.0, + 1557.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1339.0, + 534.0, + 1339.0, + 534.0, + 1384.0, + 291.0, + 1384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 676.0, + 1339.0, + 953.0, + 1339.0, + 953.0, + 1384.0, + 676.0, + 1384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1003.0, + 1339.0, + 1352.0, + 1339.0, + 1352.0, + 1384.0, + 1003.0, + 1384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1404.0, + 1339.0, + 1407.0, + 1339.0, + 1407.0, + 1384.0, + 1404.0, + 1384.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1372.0, + 711.0, + 1372.0, + 711.0, + 1415.0, + 291.0, + 1415.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 806.0, + 1372.0, + 1108.0, + 1372.0, + 1108.0, + 1415.0, + 806.0, + 1415.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1128.0, + 1372.0, + 1407.0, + 1372.0, + 1407.0, + 1415.0, + 1128.0, + 1415.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1406.0, + 297.0, + 1406.0, + 297.0, + 1443.0, + 294.0, + 1443.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 413.0, + 1406.0, + 825.0, + 1406.0, + 825.0, + 1443.0, + 413.0, + 1443.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 202.0, + 1223.0, + 202.0, + 1223.0, + 240.0, + 294.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1261.0, + 202.0, + 1406.0, + 202.0, + 1406.0, + 240.0, + 1261.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 234.0, + 624.0, + 234.0, + 624.0, + 272.0, + 293.0, + 272.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 811.0, + 234.0, + 1406.0, + 234.0, + 1406.0, + 272.0, + 811.0, + 272.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 266.0, + 1383.0, + 266.0, + 1383.0, + 303.0, + 293.0, + 303.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1603.0, + 525.0, + 1603.0, + 525.0, + 1649.0, + 292.0, + 1649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 664.0, + 1603.0, + 931.0, + 1603.0, + 931.0, + 1649.0, + 664.0, + 1649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 982.0, + 1603.0, + 1303.0, + 1603.0, + 1303.0, + 1649.0, + 982.0, + 1649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1328.0, + 1603.0, + 1407.0, + 1603.0, + 1407.0, + 1649.0, + 1328.0, + 1649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1635.0, + 492.0, + 1635.0, + 492.0, + 1680.0, + 291.0, + 1680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 588.0, + 1635.0, + 1240.0, + 1635.0, + 1240.0, + 1680.0, + 588.0, + 1680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1261.0, + 1635.0, + 1323.0, + 1635.0, + 1323.0, + 1680.0, + 1261.0, + 1680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1404.0, + 1635.0, + 1407.0, + 1635.0, + 1407.0, + 1680.0, + 1404.0, + 1680.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1673.0, + 704.0, + 1673.0, + 704.0, + 1708.0, + 295.0, + 1708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1911.0, + 1405.0, + 1911.0, + 1405.0, + 1951.0, + 295.0, + 1951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1942.0, + 296.0, + 1942.0, + 296.0, + 1982.0, + 292.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 317.0, + 1942.0, + 1404.0, + 1942.0, + 1404.0, + 1982.0, + 317.0, + 1982.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1976.0, + 742.0, + 1976.0, + 742.0, + 2010.0, + 294.0, + 2010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 615.0, + 1379.0, + 615.0, + 1379.0, + 649.0, + 296.0, + 649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1399.0, + 615.0, + 1408.0, + 615.0, + 1408.0, + 649.0, + 1399.0, + 649.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 647.0, + 403.0, + 647.0, + 403.0, + 681.0, + 295.0, + 681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 475.0, + 647.0, + 1183.0, + 647.0, + 1183.0, + 681.0, + 475.0, + 681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1254.0, + 647.0, + 1403.0, + 647.0, + 1403.0, + 681.0, + 1254.0, + 681.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 677.0, + 1140.0, + 677.0, + 1140.0, + 711.0, + 295.0, + 711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1271.0, + 1041.0, + 1271.0, + 1041.0, + 1307.0, + 296.0, + 1307.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1092.0, + 1271.0, + 1408.0, + 1271.0, + 1408.0, + 1307.0, + 1092.0, + 1307.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1299.0, + 824.0, + 1299.0, + 824.0, + 1339.0, + 293.0, + 1339.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 844.0, + 1299.0, + 1356.0, + 1299.0, + 1356.0, + 1339.0, + 844.0, + 1339.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 1847.0, + 1190.0, + 1847.0, + 1190.0, + 1879.0, + 298.0, + 1879.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1210.0, + 1847.0, + 1403.0, + 1847.0, + 1403.0, + 1879.0, + 1210.0, + 1879.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1878.0, + 1238.0, + 1878.0, + 1238.0, + 1910.0, + 296.0, + 1910.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1563.0, + 1070.0, + 1563.0, + 1070.0, + 1608.0, + 293.0, + 1608.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1760.0, + 1404.0, + 1760.0, + 1404.0, + 1800.0, + 295.0, + 1800.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 317.0, + 1793.0, + 591.0, + 1793.0, + 591.0, + 1828.0, + 317.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 611.0, + 1793.0, + 899.0, + 1793.0, + 899.0, + 1828.0, + 611.0, + 1828.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1717.0, + 806.0, + 1717.0, + 806.0, + 1766.0, + 294.0, + 1766.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 5, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1131, + 1405, + 1131, + 1405, + 1801, + 297, + 1801 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 349, + 1404, + 349, + 1404, + 684, + 298, + 684 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 710, + 1406, + 710, + 1406, + 1105, + 298, + 1105 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 300, + 1887, + 1405, + 1887, + 1405, + 2009, + 300, + 2009 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 261, + 1401, + 261, + 1401, + 322, + 298, + 322 + ], + "score": 0.954 + }, + { + "category_id": 0, + "poly": [ + 299, + 198, + 548, + 198, + 548, + 236, + 299, + 236 + ], + "score": 0.912 + }, + { + "category_id": 0, + "poly": [ + 298, + 1824, + 532, + 1824, + 532, + 1862, + 298, + 1862 + ], + "score": 0.906 + }, + { + "category_id": 2, + "poly": [ + 842, + 2061, + 858, + 2061, + 858, + 2084, + 842, + 2084 + ], + "score": 0.693 + }, + { + "category_id": 2, + "poly": [ + 842, + 2061, + 859, + 2061, + 859, + 2084, + 842, + 2084 + ], + "score": 0.139 + }, + { + "category_id": 13, + "poly": [ + 1313, + 1074, + 1382, + 1074, + 1382, + 1104, + 1313, + 1104 + ], + "score": 0.9, + "latex": "k \\geq 2" + }, + { + "category_id": 13, + "poly": [ + 975, + 1075, + 994, + 1075, + 994, + 1101, + 975, + 1101 + ], + "score": 0.81, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 611, + 623, + 629, + 623, + 629, + 649, + 611, + 649 + ], + "score": 0.8, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 1179, + 1618, + 1198, + 1618, + 1198, + 1644, + 1179, + 1644 + ], + "score": 0.78, + "latex": "k" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 195.0, + 552.0, + 195.0, + 552.0, + 241.0, + 293.0, + 241.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 1818.0, + 538.0, + 1818.0, + 538.0, + 1872.0, + 288.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2059.0, + 860.0, + 2059.0, + 860.0, + 2092.0, + 840.0, + 2092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 840.0, + 2060.0, + 861.0, + 2060.0, + 861.0, + 2092.0, + 840.0, + 2092.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1132.0, + 1405.0, + 1132.0, + 1405.0, + 1167.0, + 295.0, + 1167.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1162.0, + 1405.0, + 1162.0, + 1405.0, + 1198.0, + 292.0, + 1198.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1191.0, + 1405.0, + 1191.0, + 1405.0, + 1229.0, + 294.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1220.0, + 1408.0, + 1220.0, + 1408.0, + 1261.0, + 292.0, + 1261.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1252.0, + 1410.0, + 1252.0, + 1410.0, + 1290.0, + 291.0, + 1290.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1284.0, + 1407.0, + 1284.0, + 1407.0, + 1319.0, + 294.0, + 1319.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1310.0, + 1407.0, + 1310.0, + 1407.0, + 1351.0, + 291.0, + 1351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1343.0, + 1406.0, + 1343.0, + 1406.0, + 1381.0, + 291.0, + 1381.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1375.0, + 1405.0, + 1375.0, + 1405.0, + 1410.0, + 295.0, + 1410.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1406.0, + 1405.0, + 1406.0, + 1405.0, + 1438.0, + 296.0, + 1438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1434.0, + 1405.0, + 1434.0, + 1405.0, + 1471.0, + 294.0, + 1471.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1466.0, + 1407.0, + 1466.0, + 1407.0, + 1501.0, + 295.0, + 1501.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1493.0, + 1407.0, + 1493.0, + 1407.0, + 1532.0, + 294.0, + 1532.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1520.0, + 1410.0, + 1520.0, + 1410.0, + 1565.0, + 291.0, + 1565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1553.0, + 1408.0, + 1553.0, + 1408.0, + 1594.0, + 291.0, + 1594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1586.0, + 1407.0, + 1586.0, + 1407.0, + 1622.0, + 294.0, + 1622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1617.0, + 1178.0, + 1617.0, + 1178.0, + 1652.0, + 295.0, + 1652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1199.0, + 1617.0, + 1407.0, + 1617.0, + 1407.0, + 1652.0, + 1199.0, + 1652.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1647.0, + 1407.0, + 1647.0, + 1407.0, + 1683.0, + 294.0, + 1683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1677.0, + 1407.0, + 1677.0, + 1407.0, + 1713.0, + 294.0, + 1713.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1705.0, + 1407.0, + 1705.0, + 1407.0, + 1742.0, + 294.0, + 1742.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1737.0, + 1406.0, + 1737.0, + 1406.0, + 1775.0, + 291.0, + 1775.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1766.0, + 1373.0, + 1766.0, + 1373.0, + 1804.0, + 292.0, + 1804.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 347.0, + 1405.0, + 347.0, + 1405.0, + 386.0, + 294.0, + 386.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 379.0, + 1406.0, + 379.0, + 1406.0, + 416.0, + 292.0, + 416.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 410.0, + 1406.0, + 410.0, + 1406.0, + 445.0, + 294.0, + 445.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 441.0, + 1405.0, + 441.0, + 1405.0, + 476.0, + 294.0, + 476.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 466.0, + 1405.0, + 466.0, + 1405.0, + 509.0, + 292.0, + 509.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 500.0, + 1402.0, + 500.0, + 1402.0, + 535.0, + 294.0, + 535.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 531.0, + 1405.0, + 531.0, + 1405.0, + 566.0, + 294.0, + 566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 558.0, + 1405.0, + 558.0, + 1405.0, + 599.0, + 292.0, + 599.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 592.0, + 1405.0, + 592.0, + 1405.0, + 626.0, + 294.0, + 626.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 622.0, + 610.0, + 622.0, + 610.0, + 656.0, + 293.0, + 656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 630.0, + 622.0, + 1404.0, + 622.0, + 1404.0, + 656.0, + 630.0, + 656.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 654.0, + 1356.0, + 654.0, + 1356.0, + 688.0, + 294.0, + 688.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 705.0, + 1407.0, + 705.0, + 1407.0, + 749.0, + 292.0, + 749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 739.0, + 1404.0, + 739.0, + 1404.0, + 776.0, + 295.0, + 776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 770.0, + 1406.0, + 770.0, + 1406.0, + 806.0, + 295.0, + 806.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 801.0, + 1404.0, + 801.0, + 1404.0, + 837.0, + 295.0, + 837.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 831.0, + 1403.0, + 831.0, + 1403.0, + 864.0, + 296.0, + 864.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 860.0, + 1407.0, + 860.0, + 1407.0, + 896.0, + 295.0, + 896.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 893.0, + 1404.0, + 893.0, + 1404.0, + 926.0, + 295.0, + 926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 923.0, + 1403.0, + 923.0, + 1403.0, + 955.0, + 295.0, + 955.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 950.0, + 1408.0, + 950.0, + 1408.0, + 990.0, + 292.0, + 990.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 980.0, + 1406.0, + 980.0, + 1406.0, + 1018.0, + 293.0, + 1018.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1014.0, + 1407.0, + 1014.0, + 1407.0, + 1047.0, + 296.0, + 1047.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1042.0, + 1406.0, + 1042.0, + 1406.0, + 1079.0, + 293.0, + 1079.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1073.0, + 974.0, + 1073.0, + 974.0, + 1110.0, + 292.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 995.0, + 1073.0, + 1312.0, + 1073.0, + 1312.0, + 1110.0, + 995.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1383.0, + 1073.0, + 1396.0, + 1073.0, + 1396.0, + 1110.0, + 1383.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1886.0, + 1409.0, + 1886.0, + 1409.0, + 1921.0, + 295.0, + 1921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1918.0, + 1406.0, + 1918.0, + 1406.0, + 1950.0, + 296.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1948.0, + 1405.0, + 1948.0, + 1405.0, + 1984.0, + 295.0, + 1984.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1978.0, + 1396.0, + 1978.0, + 1396.0, + 2013.0, + 295.0, + 2013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 258.0, + 1406.0, + 258.0, + 1406.0, + 295.0, + 294.0, + 295.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 288.0, + 1306.0, + 288.0, + 1306.0, + 325.0, + 294.0, + 325.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 6, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 1562, + 1404, + 1562, + 1404, + 1841, + 298, + 1841 + ], + "score": 0.98 + }, + { + "category_id": 1, + "poly": [ + 299, + 1854, + 1405, + 1854, + 1405, + 2007, + 299, + 2007 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 298, + 1396, + 1405, + 1396, + 1405, + 1549, + 298, + 1549 + ], + "score": 0.975 + }, + { + "category_id": 5, + "poly": [ + 349, + 332, + 1349, + 332, + 1349, + 666, + 349, + 666 + ], + "score": 0.972, + "html": "
node id.type id.dense inputsparse input
train L2↓test L2↓train L2↓test L2↓
××47.95± 0.60053.93 ± 1.42629.88±0.45034.70±1.167
×O32.38± 0.44840.06±1.20215.92 ± 0.27520.39±0.765
Random (first-order)032.19 ± 0.47632.49 ± 3.68715.87 ± 0.24716.56 ± 0.904
ORF (first-order)032.35 ± 0.36939.87 ± 1.26315.87 ± 0.24716.56 ± 0.908
Random×5.909 ±0.0195.548 ± 0.0908.152 ± 0.0428.270± 0.285
ORF×5.472 ± 0.0355.143 ± 0.0787.167 ± 0.0257.190 ± 0.217
Laplacian eigenvector×1.899 ± 3.0501.702 ± 2.9120.288 ± 0.0190.064 ± 0.010
Random00.375±0.0090.234 ± 0.0110.990±0.1080.875 ± 0.042
ORF00.080 ± 0.0010.009 ± 5e-50.129 ± 0.0020.011 ± 0.002
Laplacian eigenvector○0.053 ± 1.5e-50.005 ± 1e-40.101 ± 0.0030.019 ± 0.007
" + }, + { + "category_id": 3, + "poly": [ + 307, + 692, + 1390, + 692, + 1390, + 1178, + 307, + 1178 + ], + "score": 0.966 + }, + { + "category_id": 6, + "poly": [ + 298, + 216, + 1404, + 216, + 1404, + 308, + 298, + 308 + ], + "score": 0.946 + }, + { + "category_id": 4, + "poly": [ + 296, + 1188, + 1406, + 1188, + 1406, + 1312, + 296, + 1312 + ], + "score": 0.941 + }, + { + "category_id": 0, + "poly": [ + 298, + 1337, + 936, + 1337, + 936, + 1371, + 298, + 1371 + ], + "score": 0.929 + }, + { + "category_id": 2, + "poly": [ + 840, + 2062, + 858, + 2062, + 858, + 2085, + 840, + 2085 + ], + "score": 0.8 + }, + { + "category_id": 13, + "poly": [ + 576, + 1564, + 650, + 1564, + 650, + 1592, + 576, + 1592 + ], + "score": 0.91, + "latex": "k = 2" + }, + { + "category_id": 13, + "poly": [ + 1162, + 1428, + 1231, + 1428, + 1231, + 1459, + 1162, + 1459 + ], + "score": 0.91, + "latex": "L _ { k k }" + }, + { + "category_id": 13, + "poly": [ + 1132, + 1594, + 1282, + 1594, + 1282, + 1627, + 1132, + 1627 + ], + "score": 0.91, + "latex": "( 2 + 2 ) = 1 5" + }, + { + "category_id": 13, + "poly": [ + 298, + 1486, + 424, + 1486, + 424, + 1519, + 298, + 1519 + ], + "score": 0.9, + "latex": "\\hat { \\pmb { \\alpha } ^ { 1 } } , . . . , \\pmb { \\alpha } ^ { H }" + }, + { + "category_id": 13, + "poly": [ + 1056, + 1488, + 1245, + 1488, + 1245, + 1519, + 1056, + 1519 + ], + "score": 0.9, + "latex": "\\mathbf { B } ^ { \\mu _ { 1 } } , . . . , \\mathbf { B } ^ { \\mu _ { \\mathrm { b e l l } } ( 2 k ) }" + }, + { + "category_id": 13, + "poly": [ + 835, + 1624, + 873, + 1624, + 873, + 1655, + 835, + 1655 + ], + "score": 0.89, + "latex": "\\pmb { \\alpha } ^ { h }" + }, + { + "category_id": 13, + "poly": [ + 443, + 1658, + 495, + 1658, + 495, + 1686, + 443, + 1686 + ], + "score": 0.88, + "latex": "\\mathbf { B } ^ { \\mu _ { h } }" + }, + { + "category_id": 13, + "poly": [ + 1314, + 1747, + 1346, + 1747, + 1346, + 1777, + 1314, + 1777 + ], + "score": 0.86, + "latex": "n ^ { 2 }" + }, + { + "category_id": 13, + "poly": [ + 706, + 1429, + 725, + 1429, + 725, + 1455, + 706, + 1455 + ], + "score": 0.8, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 768, + 1916, + 800, + 1916, + 800, + 1946, + 768, + 1946 + ], + "score": 0.76, + "latex": "\\bigcirc" + }, + { + "category_id": 13, + "poly": [ + 1273, + 1783, + 1301, + 1783, + 1301, + 1806, + 1273, + 1806 + ], + "score": 0.74, + "latex": "m" + }, + { + "category_id": 13, + "poly": [ + 730, + 1753, + 751, + 1753, + 751, + 1776, + 730, + 1776 + ], + "score": 0.73, + "latex": "n" + }, + { + "category_id": 13, + "poly": [ + 520, + 1946, + 554, + 1946, + 554, + 1977, + 520, + 1977 + ], + "score": 0.7, + "latex": "\\bigcirc" + }, + { + "category_id": 13, + "poly": [ + 874, + 1753, + 902, + 1753, + 902, + 1777, + 874, + 1777 + ], + "score": 0.59, + "latex": "m" + }, + { + "category_id": 13, + "poly": [ + 626, + 395, + 648, + 395, + 648, + 417, + 626, + 417 + ], + "score": 0.4, + "latex": "\\times" + }, + { + "category_id": 13, + "poly": [ + 805, + 1156, + 829, + 1156, + 829, + 1178, + 805, + 1178 + ], + "score": 0.32, + "latex": "\\mathsf { \\Omega } \\propto ^ { h }" + }, + { + "category_id": 13, + "poly": [ + 625, + 507, + 647, + 507, + 647, + 528, + 625, + 528 + ], + "score": 0.27, + "latex": "\\times" + }, + { + "category_id": 15, + "poly": [ + 339.0, + 686.0, + 484.0, + 686.0, + 484.0, + 725.0, + 339.0, + 725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 550.0, + 688.0, + 682.0, + 688.0, + 682.0, + 724.0, + 550.0, + 724.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 756.0, + 690.0, + 888.0, + 690.0, + 888.0, + 721.0, + 756.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 933.0, + 688.0, + 1120.0, + 688.0, + 1120.0, + 723.0, + 933.0, + 723.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1148.0, + 691.0, + 1317.0, + 691.0, + 1317.0, + 721.0, + 1148.0, + 721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1350.0, + 709.0, + 1388.0, + 709.0, + 1388.0, + 730.0, + 1350.0, + 730.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1350.0, + 748.0, + 1389.0, + 748.0, + 1389.0, + 769.0, + 1350.0, + 769.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1350.0, + 788.0, + 1389.0, + 788.0, + 1389.0, + 809.0, + 1350.0, + 809.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1350.0, + 829.0, + 1389.0, + 829.0, + 1389.0, + 850.0, + 1350.0, + 850.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1350.0, + 868.0, + 1389.0, + 868.0, + 1389.0, + 889.0, + 1350.0, + 889.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1353.0, + 909.0, + 1388.0, + 909.0, + 1388.0, + 926.0, + 1353.0, + 926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1350.0, + 929.0, + 1388.0, + 929.0, + 1388.0, + 951.0, + 1350.0, + 951.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1350.0, + 968.0, + 1389.0, + 968.0, + 1389.0, + 989.0, + 1350.0, + 989.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1349.0, + 1008.0, + 1389.0, + 1008.0, + 1389.0, + 1029.0, + 1349.0, + 1029.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1350.0, + 1049.0, + 1389.0, + 1049.0, + 1389.0, + 1069.0, + 1350.0, + 1069.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1350.0, + 1088.0, + 1389.0, + 1088.0, + 1389.0, + 1108.0, + 1350.0, + 1108.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1353.0, + 1129.0, + 1388.0, + 1129.0, + 1388.0, + 1146.0, + 1353.0, + 1146.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 609.0, + 1151.0, + 804.0, + 1151.0, + 804.0, + 1183.0, + 609.0, + 1183.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 830.0, + 1151.0, + 835.0, + 1151.0, + 835.0, + 1183.0, + 830.0, + 1183.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1158.0, + 1154.0, + 1307.0, + 1154.0, + 1307.0, + 1182.0, + 1158.0, + 1182.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 215.0, + 1404.0, + 215.0, + 1404.0, + 252.0, + 293.0, + 252.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 245.0, + 1402.0, + 245.0, + 1402.0, + 279.0, + 294.0, + 279.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 278.0, + 773.0, + 278.0, + 773.0, + 311.0, + 296.0, + 311.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1188.0, + 1407.0, + 1188.0, + 1407.0, + 1224.0, + 294.0, + 1224.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1219.0, + 1408.0, + 1219.0, + 1408.0, + 1253.0, + 294.0, + 1253.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1248.0, + 1406.0, + 1248.0, + 1406.0, + 1286.0, + 293.0, + 1286.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1280.0, + 1344.0, + 1280.0, + 1344.0, + 1315.0, + 293.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1335.0, + 941.0, + 1335.0, + 941.0, + 1378.0, + 294.0, + 1378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 838.0, + 2058.0, + 862.0, + 2058.0, + 862.0, + 2090.0, + 838.0, + 2090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1561.0, + 575.0, + 1561.0, + 575.0, + 1599.0, + 295.0, + 1599.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 651.0, + 1561.0, + 1405.0, + 1561.0, + 1405.0, + 1599.0, + 651.0, + 1599.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1592.0, + 1131.0, + 1592.0, + 1131.0, + 1628.0, + 292.0, + 1628.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1283.0, + 1592.0, + 1404.0, + 1592.0, + 1404.0, + 1628.0, + 1283.0, + 1628.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1625.0, + 834.0, + 1625.0, + 834.0, + 1662.0, + 293.0, + 1662.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 874.0, + 1625.0, + 1406.0, + 1625.0, + 1406.0, + 1662.0, + 874.0, + 1662.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1654.0, + 442.0, + 1654.0, + 442.0, + 1694.0, + 292.0, + 1694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 496.0, + 1654.0, + 1406.0, + 1654.0, + 1406.0, + 1694.0, + 496.0, + 1694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1688.0, + 1404.0, + 1688.0, + 1404.0, + 1721.0, + 296.0, + 1721.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1716.0, + 1407.0, + 1716.0, + 1407.0, + 1755.0, + 292.0, + 1755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1748.0, + 729.0, + 1748.0, + 729.0, + 1786.0, + 292.0, + 1786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 752.0, + 1748.0, + 873.0, + 1748.0, + 873.0, + 1786.0, + 752.0, + 1786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 903.0, + 1748.0, + 1313.0, + 1748.0, + 1313.0, + 1786.0, + 903.0, + 1786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1347.0, + 1748.0, + 1409.0, + 1748.0, + 1409.0, + 1786.0, + 1347.0, + 1786.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1776.0, + 1272.0, + 1776.0, + 1272.0, + 1815.0, + 293.0, + 1815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1302.0, + 1776.0, + 1406.0, + 1776.0, + 1406.0, + 1815.0, + 1302.0, + 1815.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1807.0, + 982.0, + 1807.0, + 982.0, + 1844.0, + 294.0, + 1844.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1853.0, + 1407.0, + 1853.0, + 1407.0, + 1888.0, + 296.0, + 1888.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1880.0, + 1407.0, + 1880.0, + 1407.0, + 1923.0, + 292.0, + 1923.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1912.0, + 767.0, + 1912.0, + 767.0, + 1952.0, + 293.0, + 1952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 801.0, + 1912.0, + 1410.0, + 1912.0, + 1410.0, + 1952.0, + 801.0, + 1952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1942.0, + 519.0, + 1942.0, + 519.0, + 1981.0, + 293.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 555.0, + 1942.0, + 1405.0, + 1942.0, + 1405.0, + 1981.0, + 555.0, + 1981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1974.0, + 1407.0, + 1974.0, + 1407.0, + 2012.0, + 293.0, + 2012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1395.0, + 1406.0, + 1395.0, + 1406.0, + 1432.0, + 293.0, + 1432.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1426.0, + 705.0, + 1426.0, + 705.0, + 1463.0, + 295.0, + 1463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 726.0, + 1426.0, + 1161.0, + 1426.0, + 1161.0, + 1463.0, + 726.0, + 1463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1232.0, + 1426.0, + 1407.0, + 1426.0, + 1407.0, + 1463.0, + 1232.0, + 1463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1458.0, + 1406.0, + 1458.0, + 1406.0, + 1491.0, + 295.0, + 1491.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1482.0, + 297.0, + 1482.0, + 297.0, + 1527.0, + 291.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 425.0, + 1482.0, + 1055.0, + 1482.0, + 1055.0, + 1527.0, + 425.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1246.0, + 1482.0, + 1410.0, + 1482.0, + 1410.0, + 1527.0, + 1246.0, + 1527.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1519.0, + 527.0, + 1519.0, + 527.0, + 1550.0, + 294.0, + 1550.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 7, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 297, + 1701, + 1405, + 1701, + 1405, + 2007, + 297, + 2007 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 299, + 1566, + 1404, + 1566, + 1404, + 1688, + 299, + 1688 + ], + "score": 0.977 + }, + { + "category_id": 1, + "poly": [ + 299, + 1263, + 1404, + 1263, + 1404, + 1418, + 299, + 1418 + ], + "score": 0.972 + }, + { + "category_id": 4, + "poly": [ + 302, + 1148, + 1403, + 1148, + 1403, + 1240, + 302, + 1240 + ], + "score": 0.956 + }, + { + "category_id": 1, + "poly": [ + 298, + 1431, + 1398, + 1431, + 1398, + 1493, + 298, + 1493 + ], + "score": 0.952 + }, + { + "category_id": 0, + "poly": [ + 299, + 1520, + 707, + 1520, + 707, + 1554, + 299, + 1554 + ], + "score": 0.923 + }, + { + "category_id": 6, + "poly": [ + 297, + 200, + 1400, + 200, + 1400, + 262, + 297, + 262 + ], + "score": 0.92 + }, + { + "category_id": 2, + "poly": [ + 840, + 2061, + 859, + 2061, + 859, + 2084, + 840, + 2084 + ], + "score": 0.796 + }, + { + "category_id": 5, + "poly": [ + 428, + 273, + 1262, + 273, + 1262, + 1137, + 428, + 1137 + ], + "score": 0.77, + "html": "
method# parametersvalid MAE↓test-dev MAE↓asymptotics
Message-passingGNNs
GCN [27] 2.0M0.13790.1398O(n+m)
GIN [27] GAT3.8M 6.7M0.1195 0.13020.1218 N/AO(n+m) O(n+m)
GCN-VN[27]4.9M0.11530.1152O(n+m)
GIN-VN [27]6.7M0.10830.1084O(n+m)
GAT-VN0.1192N/AO(n+m)
GAT-VN (large)6.7M0.1361N/AO(n+m)
55.2M
Transformerswith strong graph-specific modifications
Graphormer[63]48.3M0.0864N/AO(n²)
EGT[29]89.3M0.08690.0872O(n2)
GRPE[54]46.2M0.08900.0898O(n2)
Pure Transformers
Transformer48.5M0.2340N/AO((n+m)²)
TokenGT (ORF)48.6M0.0962N/AO((n+ m)²)
TokenGT (Lap)48.5M0.09100.0919O((n+m)²)
TokenGT(Lap)+Performer48.5M0.0935N/AO(n+m)
TokenGT (ORF) 4TokenGT (Lap) . · .
(irr grretesertrttteertrs 3 OD 8 2 · . : : 1 + 0 .O Head 1 . Head 32 . 5 10 Network depth (layer)4 (irr erretsrt rtrtttineerls 3 + 2 . + 0 . 0. + O I O COn CD C . .00 CICI 0 . . . . : . . + ·· . Head 32 . :. Head 1
" + }, + { + "category_id": 3, + "poly": [ + 428, + 273, + 1262, + 273, + 1262, + 1137, + 428, + 1137 + ], + "score": 0.133 + }, + { + "category_id": 13, + "poly": [ + 474, + 1356, + 574, + 1356, + 574, + 1388, + 474, + 1388 + ], + "score": 0.92, + "latex": "[ \\mathbf { P } _ { u } , \\mathbf { P } _ { v } ]" + }, + { + "category_id": 13, + "poly": [ + 685, + 1356, + 749, + 1356, + 749, + 1389, + 685, + 1389 + ], + "score": 0.91, + "latex": "( u , v )" + }, + { + "category_id": 13, + "poly": [ + 1079, + 1884, + 1353, + 1884, + 1353, + 1917, + 1079, + 1917 + ], + "score": 0.9, + "latex": "( \\beta _ { 1 } , \\beta _ { 2 } ) = \\mathsf { \\bar { ( 0 . 9 9 , 0 . 9 9 9 ) } }" + }, + { + "category_id": 13, + "poly": [ + 441, + 1326, + 476, + 1326, + 476, + 1356, + 441, + 1356 + ], + "score": 0.88, + "latex": "\\mathbf { P } _ { t }" + }, + { + "category_id": 13, + "poly": [ + 1131, + 516, + 1187, + 516, + 1187, + 539, + 1131, + 539 + ], + "score": 0.79, + "latex": "\\scriptstyle { \\mathcal { O } } ( n ^ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 1209, + 1265, + 1245, + 1265, + 1245, + 1297, + 1209, + 1297 + ], + "score": 0.78, + "latex": "\\textcircled{)}" + }, + { + "category_id": 13, + "poly": [ + 1031, + 1326, + 1065, + 1326, + 1065, + 1358, + 1031, + 1358 + ], + "score": 0.78, + "latex": "\\bigcirc" + }, + { + "category_id": 13, + "poly": [ + 549, + 1887, + 572, + 1887, + 572, + 1912, + 549, + 1912 + ], + "score": 0.77, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 1130, + 564, + 1187, + 564, + 1187, + 587, + 1130, + 587 + ], + "score": 0.68, + "latex": "\\mathcal { O } ( n ^ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 1130, + 540, + 1187, + 540, + 1187, + 564, + 1130, + 564 + ], + "score": 0.66, + "latex": "O ( n ^ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 1128, + 665, + 1245, + 665, + 1245, + 688, + 1128, + 688 + ], + "score": 0.61, + "latex": "O ( ( n + m ) ^ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 804, + 1328, + 818, + 1328, + 818, + 1353, + 804, + 1353 + ], + "score": 0.61, + "latex": "t" + }, + { + "category_id": 13, + "poly": [ + 1130, + 641, + 1246, + 641, + 1246, + 664, + 1130, + 664 + ], + "score": 0.56, + "latex": "O ( ( n + m ) ^ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 1130, + 617, + 1243, + 617, + 1243, + 640, + 1130, + 640 + ], + "score": 0.48, + "latex": "O ( ( n + m ) ^ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 1129, + 689, + 1221, + 689, + 1221, + 711, + 1129, + 711 + ], + "score": 0.48, + "latex": "{ \\mathcal { O } } ( n + m )" + }, + { + "category_id": 13, + "poly": [ + 1119, + 1658, + 1184, + 1658, + 1184, + 1688, + 1119, + 1688 + ], + "score": 0.42, + "latex": "3 . 7 \\mathbf { M }" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1148.0, + 1404.0, + 1148.0, + 1404.0, + 1182.0, + 297.0, + 1182.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1178.0, + 1409.0, + 1178.0, + 1409.0, + 1216.0, + 295.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1207.0, + 821.0, + 1207.0, + 821.0, + 1244.0, + 296.0, + 1244.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1514.0, + 710.0, + 1514.0, + 710.0, + 1563.0, + 293.0, + 1563.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 199.0, + 1404.0, + 199.0, + 1404.0, + 235.0, + 295.0, + 235.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 230.0, + 1400.0, + 230.0, + 1400.0, + 265.0, + 295.0, + 265.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 839.0, + 2060.0, + 861.0, + 2060.0, + 861.0, + 2090.0, + 839.0, + 2090.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 439.0, + 271.0, + 513.0, + 271.0, + 513.0, + 296.0, + 439.0, + 296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 702.0, + 271.0, + 820.0, + 271.0, + 820.0, + 297.0, + 702.0, + 297.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 833.0, + 267.0, + 956.0, + 267.0, + 956.0, + 299.0, + 833.0, + 299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 966.0, + 267.0, + 1115.0, + 267.0, + 1115.0, + 298.0, + 966.0, + 298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1125.0, + 272.0, + 1235.0, + 272.0, + 1235.0, + 296.0, + 1125.0, + 296.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 439.0, + 296.0, + 652.0, + 296.0, + 652.0, + 325.0, + 439.0, + 325.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 439.0, + 320.0, + 536.0, + 320.0, + 536.0, + 352.0, + 439.0, + 352.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 733.0, + 322.0, + 791.0, + 322.0, + 791.0, + 350.0, + 733.0, + 350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 860.0, + 322.0, + 930.0, + 322.0, + 930.0, + 350.0, + 860.0, + 350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 322.0, + 1074.0, + 322.0, + 1074.0, + 350.0, + 1006.0, + 350.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1125.0, + 320.0, + 1224.0, + 320.0, + 1224.0, + 353.0, + 1125.0, + 353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 342.0, + 531.0, + 342.0, + 531.0, + 375.0, + 440.0, + 375.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 733.0, + 345.0, + 790.0, + 345.0, + 790.0, + 372.0, + 733.0, + 372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 860.0, + 345.0, + 929.0, + 345.0, + 929.0, + 372.0, + 860.0, + 372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 346.0, + 1074.0, + 346.0, + 1074.0, + 372.0, + 1006.0, + 372.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1126.0, + 342.0, + 1224.0, + 342.0, + 1224.0, + 377.0, + 1126.0, + 377.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 368.0, + 493.0, + 368.0, + 493.0, + 395.0, + 440.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 733.0, + 368.0, + 790.0, + 368.0, + 790.0, + 395.0, + 733.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 859.0, + 368.0, + 929.0, + 368.0, + 929.0, + 395.0, + 859.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1012.0, + 368.0, + 1067.0, + 368.0, + 1067.0, + 395.0, + 1012.0, + 395.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1126.0, + 365.0, + 1224.0, + 365.0, + 1224.0, + 400.0, + 1126.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 441.0, + 392.0, + 572.0, + 392.0, + 572.0, + 418.0, + 441.0, + 418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 734.0, + 392.0, + 790.0, + 392.0, + 790.0, + 416.0, + 734.0, + 416.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 858.0, + 391.0, + 928.0, + 391.0, + 928.0, + 415.0, + 858.0, + 415.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 391.0, + 1074.0, + 391.0, + 1074.0, + 417.0, + 1006.0, + 417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1126.0, + 387.0, + 1223.0, + 387.0, + 1223.0, + 422.0, + 1126.0, + 422.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 412.0, + 567.0, + 412.0, + 567.0, + 442.0, + 440.0, + 442.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 733.0, + 415.0, + 788.0, + 415.0, + 788.0, + 439.0, + 733.0, + 439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 860.0, + 413.0, + 929.0, + 413.0, + 929.0, + 439.0, + 860.0, + 439.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1007.0, + 415.0, + 1073.0, + 415.0, + 1073.0, + 438.0, + 1007.0, + 438.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1126.0, + 409.0, + 1224.0, + 409.0, + 1224.0, + 444.0, + 1126.0, + 444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 437.0, + 527.0, + 437.0, + 527.0, + 463.0, + 440.0, + 463.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 734.0, + 438.0, + 790.0, + 438.0, + 790.0, + 462.0, + 734.0, + 462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 859.0, + 436.0, + 930.0, + 436.0, + 930.0, + 462.0, + 859.0, + 462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1015.0, + 437.0, + 1063.0, + 437.0, + 1063.0, + 462.0, + 1015.0, + 462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1126.0, + 433.0, + 1224.0, + 433.0, + 1224.0, + 466.0, + 1126.0, + 466.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 457.0, + 587.0, + 457.0, + 587.0, + 488.0, + 440.0, + 488.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 729.0, + 459.0, + 795.0, + 459.0, + 795.0, + 486.0, + 729.0, + 486.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 860.0, + 459.0, + 928.0, + 459.0, + 928.0, + 486.0, + 860.0, + 486.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1015.0, + 459.0, + 1063.0, + 459.0, + 1063.0, + 487.0, + 1015.0, + 487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1126.0, + 457.0, + 1223.0, + 457.0, + 1223.0, + 489.0, + 1126.0, + 489.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 441.0, + 486.0, + 904.0, + 486.0, + 904.0, + 514.0, + 441.0, + 514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 511.0, + 594.0, + 511.0, + 594.0, + 542.0, + 440.0, + 542.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 729.0, + 512.0, + 796.0, + 512.0, + 796.0, + 539.0, + 729.0, + 539.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 859.0, + 513.0, + 928.0, + 513.0, + 928.0, + 537.0, + 859.0, + 537.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1015.0, + 513.0, + 1063.0, + 513.0, + 1063.0, + 538.0, + 1015.0, + 538.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1125.0, + 510.0, + 1130.0, + 510.0, + 1130.0, + 541.0, + 1125.0, + 541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1188.0, + 510.0, + 1191.0, + 510.0, + 1191.0, + 541.0, + 1188.0, + 541.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 534.0, + 532.0, + 534.0, + 532.0, + 566.0, + 440.0, + 566.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 730.0, + 537.0, + 796.0, + 537.0, + 796.0, + 562.0, + 730.0, + 562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 860.0, + 535.0, + 929.0, + 535.0, + 929.0, + 561.0, + 860.0, + 561.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 536.0, + 1074.0, + 536.0, + 1074.0, + 562.0, + 1006.0, + 562.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1124.0, + 531.0, + 1129.0, + 531.0, + 1129.0, + 565.0, + 1124.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1188.0, + 531.0, + 1191.0, + 531.0, + 1191.0, + 565.0, + 1188.0, + 565.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 439.0, + 559.0, + 543.0, + 559.0, + 543.0, + 586.0, + 439.0, + 586.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 729.0, + 559.0, + 796.0, + 559.0, + 796.0, + 587.0, + 729.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 860.0, + 559.0, + 929.0, + 559.0, + 929.0, + 587.0, + 860.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1006.0, + 559.0, + 1074.0, + 559.0, + 1074.0, + 587.0, + 1006.0, + 587.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1125.0, + 556.0, + 1129.0, + 556.0, + 1129.0, + 588.0, + 1125.0, + 588.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1188.0, + 556.0, + 1191.0, + 556.0, + 1191.0, + 588.0, + 1188.0, + 588.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 588.0, + 607.0, + 588.0, + 607.0, + 614.0, + 440.0, + 614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 611.0, + 555.0, + 611.0, + 555.0, + 642.0, + 440.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 730.0, + 615.0, + 796.0, + 615.0, + 796.0, + 640.0, + 730.0, + 640.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 860.0, + 614.0, + 929.0, + 614.0, + 929.0, + 640.0, + 860.0, + 640.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1016.0, + 610.0, + 1064.0, + 610.0, + 1064.0, + 641.0, + 1016.0, + 641.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1126.0, + 612.0, + 1129.0, + 612.0, + 1129.0, + 642.0, + 1126.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1244.0, + 612.0, + 1252.0, + 612.0, + 1252.0, + 642.0, + 1244.0, + 642.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 441.0, + 636.0, + 588.0, + 636.0, + 588.0, + 664.0, + 441.0, + 664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 729.0, + 636.0, + 796.0, + 636.0, + 796.0, + 663.0, + 729.0, + 663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 860.0, + 637.0, + 929.0, + 637.0, + 929.0, + 663.0, + 860.0, + 663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1016.0, + 637.0, + 1062.0, + 637.0, + 1062.0, + 662.0, + 1016.0, + 662.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1126.0, + 633.0, + 1129.0, + 633.0, + 1129.0, + 665.0, + 1126.0, + 665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1247.0, + 633.0, + 1250.0, + 633.0, + 1250.0, + 665.0, + 1247.0, + 665.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 658.0, + 583.0, + 658.0, + 583.0, + 691.0, + 440.0, + 691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 730.0, + 662.0, + 796.0, + 662.0, + 796.0, + 687.0, + 730.0, + 687.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 861.0, + 662.0, + 929.0, + 662.0, + 929.0, + 685.0, + 861.0, + 685.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1007.0, + 662.0, + 1073.0, + 662.0, + 1073.0, + 685.0, + 1007.0, + 685.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1246.0, + 658.0, + 1252.0, + 658.0, + 1252.0, + 690.0, + 1246.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 440.0, + 683.0, + 690.0, + 683.0, + 690.0, + 711.0, + 440.0, + 711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 730.0, + 685.0, + 796.0, + 685.0, + 796.0, + 710.0, + 730.0, + 710.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 861.0, + 683.0, + 929.0, + 683.0, + 929.0, + 710.0, + 861.0, + 710.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1015.0, + 683.0, + 1063.0, + 683.0, + 1063.0, + 711.0, + 1015.0, + 711.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1222.0, + 681.0, + 1225.0, + 681.0, + 1225.0, + 713.0, + 1222.0, + 713.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 579.0, + 725.0, + 759.0, + 725.0, + 759.0, + 760.0, + 579.0, + 760.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1003.0, + 722.0, + 1172.0, + 722.0, + 1172.0, + 763.0, + 1003.0, + 763.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 495.0, + 756.0, + 520.0, + 756.0, + 520.0, + 781.0, + 495.0, + 781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 913.0, + 756.0, + 938.0, + 756.0, + 938.0, + 781.0, + 913.0, + 781.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 458.0, + 765.0, + 493.0, + 765.0, + 493.0, + 1073.0, + 458.0, + 1073.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 877.0, + 767.0, + 910.0, + 767.0, + 910.0, + 1072.0, + 877.0, + 1072.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 999.0, + 769.0, + 1013.0, + 769.0, + 1013.0, + 782.0, + 999.0, + 782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1023.0, + 779.0, + 1036.0, + 779.0, + 1036.0, + 792.0, + 1023.0, + 792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1048.0, + 783.0, + 1057.0, + 783.0, + 1057.0, + 792.0, + 1048.0, + 792.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 495.0, + 829.0, + 519.0, + 829.0, + 519.0, + 855.0, + 495.0, + 855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 913.0, + 829.0, + 938.0, + 829.0, + 938.0, + 855.0, + 913.0, + 855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1002.0, + 840.0, + 1010.0, + 840.0, + 1010.0, + 849.0, + 1002.0, + 849.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 535.0, + 846.0, + 548.0, + 846.0, + 548.0, + 893.0, + 535.0, + 893.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 557.0, + 855.0, + 570.0, + 855.0, + 570.0, + 894.0, + 557.0, + 894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 581.0, + 852.0, + 594.0, + 852.0, + 594.0, + 897.0, + 581.0, + 897.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 606.0, + 859.0, + 617.0, + 859.0, + 617.0, + 918.0, + 606.0, + 918.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 976.0, + 855.0, + 989.0, + 855.0, + 989.0, + 886.0, + 976.0, + 886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1001.0, + 859.0, + 1011.0, + 859.0, + 1011.0, + 879.0, + 1001.0, + 879.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1026.0, + 860.0, + 1033.0, + 860.0, + 1033.0, + 871.0, + 1026.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 632.0, + 868.0, + 639.0, + 868.0, + 639.0, + 884.0, + 632.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 655.0, + 870.0, + 662.0, + 870.0, + 662.0, + 880.0, + 655.0, + 880.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 795.0, + 873.0, + 803.0, + 873.0, + 803.0, + 885.0, + 795.0, + 885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 954.0, + 865.0, + 965.0, + 865.0, + 965.0, + 881.0, + 954.0, + 881.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 632.0, + 882.0, + 639.0, + 882.0, + 639.0, + 905.0, + 632.0, + 905.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 882.0, + 987.0, + 882.0, + 987.0, + 892.0, + 979.0, + 892.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1002.0, + 884.0, + 1010.0, + 884.0, + 1010.0, + 891.0, + 1002.0, + 891.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1072.0, + 887.0, + 1080.0, + 887.0, + 1080.0, + 904.0, + 1072.0, + 904.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1023.0, + 888.0, + 1036.0, + 888.0, + 1036.0, + 934.0, + 1023.0, + 934.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1048.0, + 889.0, + 1058.0, + 889.0, + 1058.0, + 949.0, + 1048.0, + 949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1096.0, + 893.0, + 1104.0, + 893.0, + 1104.0, + 904.0, + 1096.0, + 904.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 497.0, + 904.0, + 516.0, + 904.0, + 516.0, + 928.0, + 497.0, + 928.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 536.0, + 903.0, + 546.0, + 903.0, + 546.0, + 921.0, + 536.0, + 921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 583.0, + 904.0, + 592.0, + 904.0, + 592.0, + 914.0, + 583.0, + 914.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 914.0, + 904.0, + 935.0, + 904.0, + 935.0, + 928.0, + 914.0, + 928.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 909.0, + 986.0, + 909.0, + 986.0, + 919.0, + 979.0, + 919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1002.0, + 904.0, + 1010.0, + 904.0, + 1010.0, + 913.0, + 1002.0, + 913.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1071.0, + 905.0, + 1082.0, + 905.0, + 1082.0, + 927.0, + 1071.0, + 927.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 539.0, + 926.0, + 544.0, + 926.0, + 544.0, + 931.0, + 539.0, + 931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 559.0, + 919.0, + 569.0, + 919.0, + 569.0, + 937.0, + 559.0, + 937.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 583.0, + 922.0, + 592.0, + 922.0, + 592.0, + 932.0, + 583.0, + 932.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 607.0, + 921.0, + 615.0, + 921.0, + 615.0, + 929.0, + 607.0, + 929.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 956.0, + 924.0, + 964.0, + 924.0, + 964.0, + 931.0, + 956.0, + 931.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1002.0, + 924.0, + 1010.0, + 924.0, + 1010.0, + 935.0, + 1002.0, + 935.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1024.0, + 934.0, + 1035.0, + 934.0, + 1035.0, + 956.0, + 1024.0, + 956.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1001.0, + 940.0, + 1011.0, + 940.0, + 1011.0, + 957.0, + 1001.0, + 957.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 559.0, + 950.0, + 570.0, + 950.0, + 570.0, + 973.0, + 559.0, + 973.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 681.0, + 959.0, + 776.0, + 959.0, + 776.0, + 986.0, + 681.0, + 986.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 956.0, + 964.0, + 964.0, + 964.0, + 964.0, + 972.0, + 956.0, + 972.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 978.0, + 972.0, + 987.0, + 972.0, + 987.0, + 979.0, + 978.0, + 979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1002.0, + 954.0, + 1010.0, + 954.0, + 1010.0, + 980.0, + 1002.0, + 980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1025.0, + 972.0, + 1033.0, + 972.0, + 1033.0, + 980.0, + 1025.0, + 980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1099.0, + 959.0, + 1193.0, + 959.0, + 1193.0, + 986.0, + 1099.0, + 986.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 917.0, + 980.0, + 933.0, + 980.0, + 933.0, + 999.0, + 917.0, + 999.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 700.0, + 995.0, + 724.0, + 995.0, + 724.0, + 1013.0, + 700.0, + 1013.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1120.0, + 996.0, + 1141.0, + 996.0, + 1141.0, + 1012.0, + 1120.0, + 1012.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 681.0, + 1028.0, + 693.0, + 1028.0, + 693.0, + 1040.0, + 681.0, + 1040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 698.0, + 1018.0, + 789.0, + 1018.0, + 789.0, + 1048.0, + 698.0, + 1048.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 956.0, + 1012.0, + 964.0, + 1012.0, + 964.0, + 1021.0, + 956.0, + 1021.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1000.0, + 1014.0, + 1012.0, + 1014.0, + 1012.0, + 1027.0, + 1000.0, + 1027.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1100.0, + 1028.0, + 1112.0, + 1028.0, + 1112.0, + 1040.0, + 1100.0, + 1040.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1116.0, + 1018.0, + 1206.0, + 1018.0, + 1206.0, + 1048.0, + 1116.0, + 1048.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 497.0, + 1053.0, + 517.0, + 1053.0, + 517.0, + 1076.0, + 497.0, + 1076.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 535.0, + 1052.0, + 548.0, + 1052.0, + 548.0, + 1067.0, + 535.0, + 1067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 914.0, + 1053.0, + 936.0, + 1053.0, + 936.0, + 1076.0, + 914.0, + 1076.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 953.0, + 1043.0, + 965.0, + 1043.0, + 965.0, + 1069.0, + 953.0, + 1069.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 975.0, + 1045.0, + 990.0, + 1045.0, + 990.0, + 1067.0, + 975.0, + 1067.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 531.0, + 1081.0, + 551.0, + 1081.0, + 551.0, + 1104.0, + 531.0, + 1104.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 647.0, + 1078.0, + 670.0, + 1078.0, + 670.0, + 1106.0, + 647.0, + 1106.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 758.0, + 1078.0, + 792.0, + 1078.0, + 792.0, + 1106.0, + 758.0, + 1106.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 950.0, + 1081.0, + 969.0, + 1081.0, + 969.0, + 1104.0, + 950.0, + 1104.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1065.0, + 1078.0, + 1088.0, + 1078.0, + 1088.0, + 1106.0, + 1065.0, + 1106.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1177.0, + 1078.0, + 1209.0, + 1078.0, + 1209.0, + 1106.0, + 1177.0, + 1106.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 560.0, + 1104.0, + 778.0, + 1104.0, + 778.0, + 1140.0, + 560.0, + 1140.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 979.0, + 1103.0, + 1195.0, + 1103.0, + 1195.0, + 1141.0, + 979.0, + 1141.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1699.0, + 1407.0, + 1699.0, + 1407.0, + 1739.0, + 292.0, + 1739.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1735.0, + 1407.0, + 1735.0, + 1407.0, + 1766.0, + 296.0, + 1766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1760.0, + 1410.0, + 1760.0, + 1410.0, + 1801.0, + 292.0, + 1801.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1793.0, + 1406.0, + 1793.0, + 1406.0, + 1829.0, + 295.0, + 1829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1825.0, + 1407.0, + 1825.0, + 1407.0, + 1857.0, + 295.0, + 1857.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1853.0, + 1407.0, + 1853.0, + 1407.0, + 1891.0, + 294.0, + 1891.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1884.0, + 548.0, + 1884.0, + 548.0, + 1920.0, + 295.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 573.0, + 1884.0, + 1078.0, + 1884.0, + 1078.0, + 1920.0, + 573.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1354.0, + 1884.0, + 1406.0, + 1884.0, + 1406.0, + 1920.0, + 1354.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1915.0, + 1405.0, + 1915.0, + 1405.0, + 1950.0, + 295.0, + 1950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1943.0, + 1408.0, + 1943.0, + 1408.0, + 1983.0, + 291.0, + 1983.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1976.0, + 1408.0, + 1976.0, + 1408.0, + 2011.0, + 295.0, + 2011.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1567.0, + 1406.0, + 1567.0, + 1406.0, + 1599.0, + 295.0, + 1599.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1598.0, + 1404.0, + 1598.0, + 1404.0, + 1630.0, + 294.0, + 1630.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1626.0, + 1405.0, + 1626.0, + 1405.0, + 1662.0, + 293.0, + 1662.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1658.0, + 1118.0, + 1658.0, + 1118.0, + 1693.0, + 293.0, + 1693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1185.0, + 1658.0, + 1392.0, + 1658.0, + 1392.0, + 1693.0, + 1185.0, + 1693.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1262.0, + 1208.0, + 1262.0, + 1208.0, + 1302.0, + 294.0, + 1302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1246.0, + 1262.0, + 1405.0, + 1262.0, + 1405.0, + 1302.0, + 1246.0, + 1302.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1297.0, + 1404.0, + 1297.0, + 1404.0, + 1330.0, + 297.0, + 1330.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1323.0, + 440.0, + 1323.0, + 440.0, + 1363.0, + 292.0, + 1363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 477.0, + 1323.0, + 803.0, + 1323.0, + 803.0, + 1363.0, + 477.0, + 1363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 819.0, + 1323.0, + 1030.0, + 1323.0, + 1030.0, + 1363.0, + 819.0, + 1363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1066.0, + 1323.0, + 1406.0, + 1323.0, + 1406.0, + 1363.0, + 1066.0, + 1363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1351.0, + 473.0, + 1351.0, + 473.0, + 1394.0, + 293.0, + 1394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 575.0, + 1351.0, + 684.0, + 1351.0, + 684.0, + 1394.0, + 575.0, + 1394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 750.0, + 1351.0, + 1406.0, + 1351.0, + 1406.0, + 1394.0, + 750.0, + 1394.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1387.0, + 1254.0, + 1387.0, + 1254.0, + 1420.0, + 297.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1427.0, + 1404.0, + 1427.0, + 1404.0, + 1470.0, + 292.0, + 1470.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1461.0, + 1078.0, + 1461.0, + 1078.0, + 1496.0, + 295.0, + 1496.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 8, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 298, + 612, + 1404, + 612, + 1404, + 1068, + 298, + 1068 + ], + "score": 0.984 + }, + { + "category_id": 1, + "poly": [ + 298, + 1421, + 1405, + 1421, + 1405, + 1695, + 298, + 1695 + ], + "score": 0.983 + }, + { + "category_id": 1, + "poly": [ + 298, + 1162, + 1404, + 1162, + 1404, + 1407, + 298, + 1407 + ], + "score": 0.982 + }, + { + "category_id": 1, + "poly": [ + 297, + 202, + 1405, + 202, + 1405, + 597, + 297, + 597 + ], + "score": 0.981 + }, + { + "category_id": 1, + "poly": [ + 298, + 1709, + 1403, + 1709, + 1403, + 1862, + 298, + 1862 + ], + "score": 0.978 + }, + { + "category_id": 1, + "poly": [ + 300, + 1884, + 1403, + 1884, + 1403, + 2007, + 300, + 2007 + ], + "score": 0.976 + }, + { + "category_id": 0, + "poly": [ + 298, + 1104, + 509, + 1104, + 509, + 1142, + 298, + 1142 + ], + "score": 0.904 + }, + { + "category_id": 2, + "poly": [ + 837, + 2062, + 865, + 2062, + 865, + 2085, + 837, + 2085 + ], + "score": 0.842 + }, + { + "category_id": 13, + "poly": [ + 1022, + 1450, + 1175, + 1450, + 1175, + 1485, + 1022, + 1485 + ], + "score": 0.93, + "latex": "O ( ( n + m ) ^ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 511, + 1740, + 609, + 1740, + 609, + 1773, + 511, + 1773 + ], + "score": 0.92, + "latex": "( n + m )" + }, + { + "category_id": 13, + "poly": [ + 391, + 1513, + 513, + 1513, + 513, + 1546, + 391, + 1546 + ], + "score": 0.92, + "latex": "O ( n + m )" + }, + { + "category_id": 13, + "poly": [ + 1013, + 537, + 1134, + 537, + 1134, + 570, + 1013, + 570 + ], + "score": 0.92, + "latex": "{ \\mathcal { O } } ( n + m )" + }, + { + "category_id": 13, + "poly": [ + 382, + 505, + 458, + 505, + 458, + 539, + 382, + 539 + ], + "score": 0.92, + "latex": "\\mathcal { O } ( n ^ { 2 } )" + }, + { + "category_id": 13, + "poly": [ + 769, + 1226, + 788, + 1226, + 788, + 1252, + 769, + 1252 + ], + "score": 0.82, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 895, + 1226, + 914, + 1226, + 914, + 1253, + 895, + 1253 + ], + "score": 0.81, + "latex": "k" + }, + { + "category_id": 13, + "poly": [ + 479, + 540, + 503, + 540, + 503, + 565, + 479, + 565 + ], + "score": 0.69, + "latex": "^ +" + }, + { + "category_id": 13, + "poly": [ + 905, + 415, + 1057, + 415, + 1057, + 446, + 905, + 446 + ], + "score": 0.39, + "latex": "0 . 0 9 1 0 ~ \\mathrm { M A E }" + }, + { + "category_id": 13, + "poly": [ + 297, + 294, + 442, + 294, + 442, + 324, + 297, + 324 + ], + "score": 0.31, + "latex": "0 . 0 9 6 2 \\mathrm { M A E }" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1100.0, + 514.0, + 1100.0, + 514.0, + 1149.0, + 292.0, + 1149.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 608.0, + 1405.0, + 608.0, + 1405.0, + 650.0, + 292.0, + 650.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 641.0, + 1405.0, + 641.0, + 1405.0, + 677.0, + 293.0, + 677.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 670.0, + 1408.0, + 670.0, + 1408.0, + 708.0, + 293.0, + 708.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 702.0, + 1406.0, + 702.0, + 1406.0, + 741.0, + 293.0, + 741.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 732.0, + 1406.0, + 732.0, + 1406.0, + 770.0, + 293.0, + 770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 763.0, + 1408.0, + 763.0, + 1408.0, + 802.0, + 293.0, + 802.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 794.0, + 1406.0, + 794.0, + 1406.0, + 829.0, + 293.0, + 829.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 825.0, + 1406.0, + 825.0, + 1406.0, + 860.0, + 294.0, + 860.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 855.0, + 1406.0, + 855.0, + 1406.0, + 890.0, + 294.0, + 890.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 886.0, + 1405.0, + 886.0, + 1405.0, + 919.0, + 293.0, + 919.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 916.0, + 1406.0, + 916.0, + 1406.0, + 950.0, + 294.0, + 950.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 945.0, + 1405.0, + 945.0, + 1405.0, + 980.0, + 294.0, + 980.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 975.0, + 1404.0, + 975.0, + 1404.0, + 1010.0, + 293.0, + 1010.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1008.0, + 1405.0, + 1008.0, + 1405.0, + 1042.0, + 294.0, + 1042.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1037.0, + 1387.0, + 1037.0, + 1387.0, + 1072.0, + 294.0, + 1072.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1420.0, + 1405.0, + 1420.0, + 1405.0, + 1457.0, + 295.0, + 1457.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1450.0, + 1021.0, + 1450.0, + 1021.0, + 1487.0, + 295.0, + 1487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1176.0, + 1450.0, + 1405.0, + 1450.0, + 1405.0, + 1487.0, + 1176.0, + 1487.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1481.0, + 1406.0, + 1481.0, + 1406.0, + 1518.0, + 295.0, + 1518.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1510.0, + 390.0, + 1510.0, + 390.0, + 1550.0, + 293.0, + 1550.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 514.0, + 1510.0, + 1407.0, + 1510.0, + 1407.0, + 1550.0, + 514.0, + 1550.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1541.0, + 1407.0, + 1541.0, + 1407.0, + 1577.0, + 295.0, + 1577.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1568.0, + 1407.0, + 1568.0, + 1407.0, + 1610.0, + 293.0, + 1610.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1603.0, + 1407.0, + 1603.0, + 1407.0, + 1640.0, + 295.0, + 1640.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1632.0, + 1403.0, + 1632.0, + 1403.0, + 1672.0, + 293.0, + 1672.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1663.0, + 1401.0, + 1663.0, + 1401.0, + 1700.0, + 295.0, + 1700.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1161.0, + 1406.0, + 1161.0, + 1406.0, + 1199.0, + 293.0, + 1199.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1193.0, + 1404.0, + 1193.0, + 1404.0, + 1229.0, + 295.0, + 1229.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1226.0, + 768.0, + 1226.0, + 768.0, + 1259.0, + 296.0, + 1259.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 789.0, + 1226.0, + 894.0, + 1226.0, + 894.0, + 1259.0, + 789.0, + 1259.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 915.0, + 1226.0, + 1405.0, + 1226.0, + 1405.0, + 1259.0, + 915.0, + 1259.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1255.0, + 1405.0, + 1255.0, + 1405.0, + 1289.0, + 292.0, + 1289.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1286.0, + 1406.0, + 1286.0, + 1406.0, + 1319.0, + 294.0, + 1319.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1314.0, + 1406.0, + 1314.0, + 1406.0, + 1349.0, + 293.0, + 1349.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1344.0, + 1405.0, + 1344.0, + 1405.0, + 1382.0, + 293.0, + 1382.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1373.0, + 1259.0, + 1373.0, + 1259.0, + 1411.0, + 293.0, + 1411.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 202.0, + 1406.0, + 202.0, + 1406.0, + 238.0, + 295.0, + 238.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 232.0, + 1406.0, + 232.0, + 1406.0, + 272.0, + 292.0, + 272.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 265.0, + 1403.0, + 265.0, + 1403.0, + 298.0, + 294.0, + 298.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 443.0, + 294.0, + 1403.0, + 294.0, + 1403.0, + 329.0, + 443.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 321.0, + 1405.0, + 321.0, + 1405.0, + 363.0, + 292.0, + 363.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 354.0, + 1405.0, + 354.0, + 1405.0, + 390.0, + 294.0, + 390.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 385.0, + 1402.0, + 385.0, + 1402.0, + 418.0, + 296.0, + 418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 414.0, + 904.0, + 414.0, + 904.0, + 452.0, + 291.0, + 452.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1058.0, + 414.0, + 1407.0, + 414.0, + 1407.0, + 452.0, + 1058.0, + 452.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 446.0, + 1407.0, + 446.0, + 1407.0, + 482.0, + 294.0, + 482.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 472.0, + 1406.0, + 472.0, + 1406.0, + 513.0, + 291.0, + 513.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 504.0, + 381.0, + 504.0, + 381.0, + 544.0, + 292.0, + 544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 459.0, + 504.0, + 1408.0, + 504.0, + 1408.0, + 544.0, + 459.0, + 544.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 534.0, + 478.0, + 534.0, + 478.0, + 574.0, + 292.0, + 574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 504.0, + 534.0, + 1012.0, + 534.0, + 1012.0, + 574.0, + 504.0, + 574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 1135.0, + 534.0, + 1410.0, + 534.0, + 1410.0, + 574.0, + 1135.0, + 574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 566.0, + 1332.0, + 566.0, + 1332.0, + 602.0, + 294.0, + 602.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1708.0, + 1404.0, + 1708.0, + 1404.0, + 1743.0, + 296.0, + 1743.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1737.0, + 510.0, + 1737.0, + 510.0, + 1776.0, + 293.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 610.0, + 1737.0, + 1407.0, + 1737.0, + 1407.0, + 1776.0, + 610.0, + 1776.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1768.0, + 1405.0, + 1768.0, + 1405.0, + 1808.0, + 293.0, + 1808.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1799.0, + 1404.0, + 1799.0, + 1404.0, + 1834.0, + 294.0, + 1834.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 1832.0, + 1345.0, + 1832.0, + 1345.0, + 1865.0, + 296.0, + 1865.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1884.0, + 1407.0, + 1884.0, + 1407.0, + 1920.0, + 295.0, + 1920.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1913.0, + 1407.0, + 1913.0, + 1407.0, + 1949.0, + 293.0, + 1949.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1947.0, + 1405.0, + 1947.0, + 1405.0, + 1979.0, + 295.0, + 1979.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1976.0, + 840.0, + 1976.0, + 840.0, + 2012.0, + 294.0, + 2012.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 9, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 293, + 288, + 1410, + 288, + 1410, + 2008, + 293, + 2008 + ], + "score": 0.963 + }, + { + "category_id": 0, + "poly": [ + 298, + 199, + 456, + 199, + 456, + 235, + 298, + 235 + ], + "score": 0.887 + }, + { + "category_id": 2, + "poly": [ + 835, + 2061, + 864, + 2061, + 864, + 2087, + 835, + 2087 + ], + "score": 0.835 + }, + { + "category_id": 15, + "poly": [ + 295.0, + 197.0, + 460.0, + 197.0, + 460.0, + 240.0, + 295.0, + 240.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 831.0, + 2058.0, + 869.0, + 2058.0, + 869.0, + 2097.0, + 831.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 293.0, + 1365.0, + 293.0, + 1365.0, + 329.0, + 306.0, + 329.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 333.0, + 1407.0, + 333.0, + 1407.0, + 369.0, + 306.0, + 369.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 357.0, + 893.0, + 357.0, + 893.0, + 393.0, + 348.0, + 393.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 395.0, + 1403.0, + 395.0, + 1403.0, + 431.0, + 306.0, + 431.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 426.0, + 829.0, + 426.0, + 829.0, + 462.0, + 348.0, + 462.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 464.0, + 1403.0, + 464.0, + 1403.0, + 500.0, + 306.0, + 500.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 494.0, + 395.0, + 494.0, + 395.0, + 528.0, + 348.0, + 528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 526.0, + 1409.0, + 526.0, + 1409.0, + 570.0, + 302.0, + 570.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 558.0, + 1068.0, + 558.0, + 1068.0, + 594.0, + 349.0, + 594.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 302.0, + 592.0, + 1411.0, + 592.0, + 1411.0, + 636.0, + 302.0, + 636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 625.0, + 1409.0, + 625.0, + 1409.0, + 661.0, + 348.0, + 661.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 342.0, + 647.0, + 1411.0, + 647.0, + 1411.0, + 691.0, + 342.0, + 691.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 344.0, + 676.0, + 1409.0, + 676.0, + 1409.0, + 720.0, + 344.0, + 720.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 706.0, + 707.0, + 706.0, + 707.0, + 742.0, + 348.0, + 742.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 746.0, + 1393.0, + 746.0, + 1393.0, + 782.0, + 306.0, + 782.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 306.0, + 784.0, + 1409.0, + 784.0, + 1409.0, + 820.0, + 306.0, + 820.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 812.0, + 1409.0, + 812.0, + 1409.0, + 848.0, + 349.0, + 848.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 850.0, + 1407.0, + 850.0, + 1407.0, + 886.0, + 294.0, + 886.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 879.0, + 553.0, + 879.0, + 553.0, + 915.0, + 348.0, + 915.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 917.0, + 1407.0, + 917.0, + 1407.0, + 953.0, + 292.0, + 953.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 945.0, + 1409.0, + 945.0, + 1409.0, + 981.0, + 348.0, + 981.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 970.0, + 743.0, + 970.0, + 743.0, + 1008.0, + 346.0, + 1008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1012.0, + 1405.0, + 1012.0, + 1405.0, + 1048.0, + 294.0, + 1048.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1042.0, + 956.0, + 1042.0, + 956.0, + 1072.0, + 351.0, + 1072.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1078.0, + 1348.0, + 1078.0, + 1348.0, + 1114.0, + 294.0, + 1114.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1116.0, + 1095.0, + 1116.0, + 1095.0, + 1152.0, + 292.0, + 1152.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1154.0, + 1405.0, + 1154.0, + 1405.0, + 1190.0, + 292.0, + 1190.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1182.0, + 718.0, + 1182.0, + 718.0, + 1218.0, + 349.0, + 1218.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1222.0, + 1407.0, + 1222.0, + 1407.0, + 1258.0, + 292.0, + 1258.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1246.0, + 658.0, + 1246.0, + 658.0, + 1285.0, + 346.0, + 1285.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1286.0, + 1407.0, + 1286.0, + 1407.0, + 1322.0, + 292.0, + 1322.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1315.0, + 1104.0, + 1315.0, + 1104.0, + 1351.0, + 349.0, + 1351.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1353.0, + 1407.0, + 1353.0, + 1407.0, + 1387.0, + 292.0, + 1387.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1381.0, + 1407.0, + 1381.0, + 1407.0, + 1417.0, + 349.0, + 1417.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1408.0, + 1114.0, + 1408.0, + 1114.0, + 1444.0, + 349.0, + 1444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1448.0, + 1407.0, + 1448.0, + 1407.0, + 1484.0, + 294.0, + 1484.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 346.0, + 1474.0, + 410.0, + 1474.0, + 410.0, + 1514.0, + 346.0, + 1514.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1512.0, + 1407.0, + 1512.0, + 1407.0, + 1548.0, + 292.0, + 1548.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1544.0, + 713.0, + 1544.0, + 713.0, + 1575.0, + 349.0, + 1575.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1576.0, + 1409.0, + 1576.0, + 1409.0, + 1618.0, + 292.0, + 1618.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1607.0, + 810.0, + 1607.0, + 810.0, + 1643.0, + 349.0, + 1643.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1647.0, + 1407.0, + 1647.0, + 1407.0, + 1683.0, + 294.0, + 1683.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1673.0, + 728.0, + 1673.0, + 728.0, + 1709.0, + 348.0, + 1709.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1713.0, + 1409.0, + 1713.0, + 1409.0, + 1749.0, + 294.0, + 1749.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1743.0, + 507.0, + 1743.0, + 507.0, + 1774.0, + 349.0, + 1774.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1777.0, + 1405.0, + 1777.0, + 1405.0, + 1813.0, + 294.0, + 1813.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1808.0, + 720.0, + 1808.0, + 720.0, + 1838.0, + 351.0, + 1838.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1844.0, + 1289.0, + 1844.0, + 1289.0, + 1880.0, + 294.0, + 1880.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1880.0, + 1411.0, + 1880.0, + 1411.0, + 1922.0, + 292.0, + 1922.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1914.0, + 726.0, + 1914.0, + 726.0, + 1944.0, + 348.0, + 1944.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 1950.0, + 1407.0, + 1950.0, + 1407.0, + 1986.0, + 294.0, + 1986.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1978.0, + 994.0, + 1978.0, + 994.0, + 2014.0, + 348.0, + 2014.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 10, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 293, + 189, + 1411, + 189, + 1411, + 2011, + 293, + 2011 + ], + "score": 0.958 + }, + { + "category_id": 2, + "poly": [ + 835, + 2060, + 866, + 2060, + 866, + 2087, + 835, + 2087 + ], + "score": 0.838 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2097.0, + 832.0, + 2097.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 203.0, + 1409.0, + 203.0, + 1409.0, + 241.0, + 295.0, + 241.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 227.0, + 562.0, + 227.0, + 562.0, + 269.0, + 348.0, + 269.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 271.0, + 1405.0, + 271.0, + 1405.0, + 303.0, + 297.0, + 303.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 297.0, + 946.0, + 297.0, + 946.0, + 335.0, + 349.0, + 335.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 333.0, + 1407.0, + 333.0, + 1407.0, + 371.0, + 293.0, + 371.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 365.0, + 1405.0, + 365.0, + 1405.0, + 397.0, + 351.0, + 397.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 389.0, + 1190.0, + 389.0, + 1190.0, + 427.0, + 345.0, + 427.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 427.0, + 1405.0, + 427.0, + 1405.0, + 465.0, + 293.0, + 465.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 453.0, + 936.0, + 453.0, + 936.0, + 493.0, + 347.0, + 493.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 493.0, + 1407.0, + 493.0, + 1407.0, + 531.0, + 293.0, + 531.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 519.0, + 994.0, + 519.0, + 994.0, + 559.0, + 345.0, + 559.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 559.0, + 1409.0, + 559.0, + 1409.0, + 597.0, + 293.0, + 597.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 589.0, + 493.0, + 589.0, + 493.0, + 622.0, + 351.0, + 622.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 621.0, + 1407.0, + 621.0, + 1407.0, + 664.0, + 291.0, + 664.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 654.0, + 946.0, + 654.0, + 946.0, + 690.0, + 347.0, + 690.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 692.0, + 1363.0, + 692.0, + 1363.0, + 730.0, + 295.0, + 730.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 724.0, + 1411.0, + 724.0, + 1411.0, + 770.0, + 295.0, + 770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 758.0, + 562.0, + 758.0, + 562.0, + 790.0, + 351.0, + 790.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 792.0, + 1409.0, + 792.0, + 1409.0, + 832.0, + 295.0, + 832.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 818.0, + 1405.0, + 818.0, + 1405.0, + 864.0, + 345.0, + 864.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 852.0, + 794.0, + 852.0, + 794.0, + 884.0, + 351.0, + 884.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 888.0, + 1405.0, + 888.0, + 1405.0, + 926.0, + 293.0, + 926.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 916.0, + 932.0, + 916.0, + 932.0, + 952.0, + 347.0, + 952.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 952.0, + 1407.0, + 952.0, + 1407.0, + 990.0, + 295.0, + 990.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 984.0, + 1225.0, + 984.0, + 1225.0, + 1016.0, + 349.0, + 1016.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1020.0, + 1409.0, + 1020.0, + 1409.0, + 1058.0, + 293.0, + 1058.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1050.0, + 666.0, + 1050.0, + 666.0, + 1082.0, + 349.0, + 1082.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1080.0, + 1411.0, + 1080.0, + 1411.0, + 1126.0, + 289.0, + 1126.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1110.0, + 616.0, + 1110.0, + 616.0, + 1150.0, + 347.0, + 1150.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1148.0, + 1407.0, + 1148.0, + 1407.0, + 1188.0, + 291.0, + 1188.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1178.0, + 1211.0, + 1178.0, + 1211.0, + 1216.0, + 349.0, + 1216.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1218.0, + 1152.0, + 1218.0, + 1152.0, + 1250.0, + 295.0, + 1250.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1256.0, + 1339.0, + 1256.0, + 1339.0, + 1288.0, + 295.0, + 1288.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1294.0, + 1405.0, + 1294.0, + 1405.0, + 1326.0, + 295.0, + 1326.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1322.0, + 748.0, + 1322.0, + 748.0, + 1354.0, + 349.0, + 1354.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1358.0, + 1409.0, + 1358.0, + 1409.0, + 1396.0, + 293.0, + 1396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1388.0, + 658.0, + 1388.0, + 658.0, + 1420.0, + 349.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1422.0, + 1409.0, + 1422.0, + 1409.0, + 1460.0, + 293.0, + 1460.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1454.0, + 708.0, + 1454.0, + 708.0, + 1486.0, + 351.0, + 1486.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1488.0, + 1409.0, + 1488.0, + 1409.0, + 1526.0, + 293.0, + 1526.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1520.0, + 519.0, + 1520.0, + 519.0, + 1552.0, + 349.0, + 1552.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 289.0, + 1548.0, + 1411.0, + 1548.0, + 1411.0, + 1595.0, + 289.0, + 1595.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1583.0, + 483.0, + 1583.0, + 483.0, + 1621.0, + 347.0, + 1621.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 291.0, + 1614.0, + 1409.0, + 1614.0, + 1409.0, + 1663.0, + 291.0, + 1663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1649.0, + 1329.0, + 1649.0, + 1329.0, + 1687.0, + 347.0, + 1687.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 1687.0, + 1407.0, + 1687.0, + 1407.0, + 1725.0, + 297.0, + 1725.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 348.0, + 1704.0, + 648.0, + 1704.0, + 648.0, + 1755.0, + 348.0, + 1755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1751.0, + 1409.0, + 1751.0, + 1409.0, + 1789.0, + 293.0, + 1789.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1779.0, + 668.0, + 1779.0, + 668.0, + 1817.0, + 349.0, + 1817.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1817.0, + 1409.0, + 1817.0, + 1409.0, + 1855.0, + 293.0, + 1855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1847.0, + 483.0, + 1847.0, + 483.0, + 1885.0, + 349.0, + 1885.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1883.0, + 1403.0, + 1883.0, + 1403.0, + 1921.0, + 295.0, + 1921.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1909.0, + 858.0, + 1909.0, + 858.0, + 1947.0, + 349.0, + 1947.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1947.0, + 1405.0, + 1947.0, + 1405.0, + 1985.0, + 295.0, + 1985.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1979.0, + 614.0, + 1979.0, + 614.0, + 2011.0, + 349.0, + 2011.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 11, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 292, + 250, + 1411, + 250, + 1411, + 1995, + 292, + 1995 + ], + "score": 0.962 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2086, + 836, + 2086 + ], + "score": 0.834 + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2058.0, + 870.0, + 2058.0, + 870.0, + 2096.0, + 832.0, + 2096.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 234.0, + 482.0, + 234.0, + 482.0, + 270.0, + 347.0, + 270.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 269.0, + 1408.0, + 269.0, + 1408.0, + 306.0, + 293.0, + 306.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 298.0, + 1408.0, + 298.0, + 1408.0, + 335.0, + 347.0, + 335.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 325.0, + 1244.0, + 325.0, + 1244.0, + 361.0, + 349.0, + 361.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 363.0, + 1406.0, + 363.0, + 1406.0, + 400.0, + 293.0, + 400.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 392.0, + 975.0, + 392.0, + 975.0, + 429.0, + 349.0, + 429.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 425.0, + 1408.0, + 425.0, + 1408.0, + 467.0, + 292.0, + 467.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 456.0, + 560.0, + 456.0, + 560.0, + 492.0, + 351.0, + 492.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 488.0, + 1408.0, + 488.0, + 1408.0, + 533.0, + 292.0, + 533.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 519.0, + 891.0, + 519.0, + 891.0, + 559.0, + 345.0, + 559.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 559.0, + 1410.0, + 559.0, + 1410.0, + 596.0, + 293.0, + 596.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 588.0, + 627.0, + 588.0, + 627.0, + 619.0, + 351.0, + 619.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 619.0, + 1408.0, + 619.0, + 1408.0, + 663.0, + 290.0, + 663.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 352.0, + 654.0, + 621.0, + 654.0, + 621.0, + 684.0, + 352.0, + 684.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 684.0, + 1408.0, + 684.0, + 1408.0, + 729.0, + 290.0, + 729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 715.0, + 1185.0, + 715.0, + 1185.0, + 755.0, + 347.0, + 755.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 754.0, + 1408.0, + 754.0, + 1408.0, + 790.0, + 293.0, + 790.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 780.0, + 1192.0, + 780.0, + 1192.0, + 817.0, + 349.0, + 817.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 819.0, + 1408.0, + 819.0, + 1408.0, + 855.0, + 293.0, + 855.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 857.0, + 1408.0, + 857.0, + 1408.0, + 894.0, + 293.0, + 894.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 886.0, + 1360.0, + 886.0, + 1360.0, + 923.0, + 349.0, + 923.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 923.0, + 1406.0, + 923.0, + 1406.0, + 959.0, + 293.0, + 959.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 948.0, + 768.0, + 948.0, + 768.0, + 986.0, + 347.0, + 986.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 990.0, + 1408.0, + 990.0, + 1408.0, + 1021.0, + 295.0, + 1021.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1013.0, + 983.0, + 1013.0, + 983.0, + 1051.0, + 347.0, + 1051.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1055.0, + 1404.0, + 1055.0, + 1404.0, + 1086.0, + 295.0, + 1086.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1080.0, + 712.0, + 1080.0, + 712.0, + 1117.0, + 349.0, + 1117.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1119.0, + 1408.0, + 1119.0, + 1408.0, + 1155.0, + 293.0, + 1155.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1146.0, + 627.0, + 1146.0, + 627.0, + 1182.0, + 347.0, + 1182.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 290.0, + 1178.0, + 1410.0, + 1178.0, + 1410.0, + 1222.0, + 290.0, + 1222.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 345.0, + 1209.0, + 615.0, + 1209.0, + 615.0, + 1248.0, + 345.0, + 1248.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1245.0, + 1408.0, + 1245.0, + 1408.0, + 1282.0, + 293.0, + 1282.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1274.0, + 642.0, + 1274.0, + 642.0, + 1313.0, + 347.0, + 1313.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1311.0, + 1408.0, + 1311.0, + 1408.0, + 1353.0, + 292.0, + 1353.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1342.0, + 1071.0, + 1342.0, + 1071.0, + 1378.0, + 349.0, + 1378.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 288.0, + 1374.0, + 1406.0, + 1374.0, + 1406.0, + 1418.0, + 288.0, + 1418.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1407.0, + 941.0, + 1407.0, + 941.0, + 1443.0, + 349.0, + 1443.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1443.0, + 1406.0, + 1443.0, + 1406.0, + 1480.0, + 293.0, + 1480.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1472.0, + 1122.0, + 1472.0, + 1122.0, + 1509.0, + 349.0, + 1509.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1507.0, + 1404.0, + 1507.0, + 1404.0, + 1543.0, + 292.0, + 1543.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1538.0, + 1078.0, + 1538.0, + 1078.0, + 1574.0, + 351.0, + 1574.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1574.0, + 1408.0, + 1574.0, + 1408.0, + 1611.0, + 293.0, + 1611.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1605.0, + 520.0, + 1605.0, + 520.0, + 1636.0, + 351.0, + 1636.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 292.0, + 1639.0, + 1406.0, + 1639.0, + 1406.0, + 1676.0, + 292.0, + 1676.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1668.0, + 1118.0, + 1668.0, + 1118.0, + 1705.0, + 349.0, + 1705.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1705.0, + 1406.0, + 1705.0, + 1406.0, + 1741.0, + 293.0, + 1741.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 351.0, + 1736.0, + 802.0, + 1736.0, + 802.0, + 1766.0, + 351.0, + 1766.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1770.0, + 1406.0, + 1770.0, + 1406.0, + 1807.0, + 293.0, + 1807.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1797.0, + 796.0, + 1797.0, + 796.0, + 1834.0, + 349.0, + 1834.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1836.0, + 1406.0, + 1836.0, + 1406.0, + 1872.0, + 293.0, + 1872.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 347.0, + 1862.0, + 1034.0, + 1862.0, + 1034.0, + 1899.0, + 347.0, + 1899.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 1901.0, + 1406.0, + 1901.0, + 1406.0, + 1937.0, + 293.0, + 1937.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 349.0, + 1932.0, + 644.0, + 1932.0, + 644.0, + 1962.0, + 349.0, + 1962.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 295.0, + 1966.0, + 1164.0, + 1966.0, + 1164.0, + 2003.0, + 295.0, + 2003.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 12, + "width": 1700, + "height": 2200 + } + }, + { + "layout_dets": [ + { + "category_id": 1, + "poly": [ + 407, + 1157, + 1404, + 1157, + 1404, + 1444, + 407, + 1444 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 407, + 1493, + 1404, + 1493, + 1404, + 1755, + 407, + 1755 + ], + "score": 0.976 + }, + { + "category_id": 1, + "poly": [ + 406, + 735, + 1403, + 735, + 1403, + 963, + 406, + 963 + ], + "score": 0.973 + }, + { + "category_id": 1, + "poly": [ + 408, + 1010, + 1404, + 1010, + 1404, + 1108, + 408, + 1108 + ], + "score": 0.965 + }, + { + "category_id": 1, + "poly": [ + 408, + 1804, + 1404, + 1804, + 1404, + 1998, + 408, + 1998 + ], + "score": 0.963 + }, + { + "category_id": 1, + "poly": [ + 300, + 266, + 1405, + 266, + 1405, + 390, + 300, + 390 + ], + "score": 0.962 + }, + { + "category_id": 1, + "poly": [ + 304, + 577, + 1406, + 577, + 1406, + 670, + 304, + 670 + ], + "score": 0.96 + }, + { + "category_id": 1, + "poly": [ + 371, + 411, + 1403, + 411, + 1403, + 555, + 371, + 555 + ], + "score": 0.936 + }, + { + "category_id": 1, + "poly": [ + 364, + 971, + 840, + 971, + 840, + 1003, + 364, + 1003 + ], + "score": 0.913 + }, + { + "category_id": 1, + "poly": [ + 363, + 1117, + 671, + 1117, + 671, + 1148, + 363, + 1148 + ], + "score": 0.908 + }, + { + "category_id": 0, + "poly": [ + 298, + 198, + 435, + 198, + 435, + 235, + 298, + 235 + ], + "score": 0.899 + }, + { + "category_id": 1, + "poly": [ + 365, + 695, + 580, + 695, + 580, + 725, + 365, + 725 + ], + "score": 0.881 + }, + { + "category_id": 1, + "poly": [ + 365, + 1764, + 1200, + 1764, + 1200, + 1795, + 365, + 1795 + ], + "score": 0.877 + }, + { + "category_id": 1, + "poly": [ + 367, + 1453, + 1402, + 1453, + 1402, + 1485, + 367, + 1485 + ], + "score": 0.872 + }, + { + "category_id": 2, + "poly": [ + 836, + 2061, + 865, + 2061, + 865, + 2085, + 836, + 2085 + ], + "score": 0.852 + }, + { + "category_id": 15, + "poly": [ + 296.0, + 197.0, + 440.0, + 197.0, + 440.0, + 239.0, + 296.0, + 239.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 832.0, + 2057.0, + 871.0, + 2057.0, + 871.0, + 2095.0, + 832.0, + 2095.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 1155.0, + 1408.0, + 1155.0, + 1408.0, + 1192.0, + 404.0, + 1192.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 1187.0, + 1404.0, + 1187.0, + 1404.0, + 1218.0, + 450.0, + 1218.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 1218.0, + 803.0, + 1218.0, + 803.0, + 1252.0, + 450.0, + 1252.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 1250.0, + 1404.0, + 1250.0, + 1404.0, + 1287.0, + 405.0, + 1287.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 448.0, + 1284.0, + 848.0, + 1284.0, + 848.0, + 1315.0, + 448.0, + 1315.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 407.0, + 1315.0, + 1409.0, + 1315.0, + 1409.0, + 1355.0, + 407.0, + 1355.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 1349.0, + 944.0, + 1349.0, + 944.0, + 1379.0, + 449.0, + 1379.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 407.0, + 1379.0, + 1405.0, + 1379.0, + 1405.0, + 1420.0, + 407.0, + 1420.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 452.0, + 1414.0, + 1192.0, + 1414.0, + 1192.0, + 1445.0, + 452.0, + 1445.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 1493.0, + 1363.0, + 1493.0, + 1363.0, + 1528.0, + 404.0, + 1528.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 403.0, + 1526.0, + 1179.0, + 1526.0, + 1179.0, + 1564.0, + 403.0, + 1564.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 403.0, + 1561.0, + 1405.0, + 1561.0, + 1405.0, + 1599.0, + 403.0, + 1599.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 1591.0, + 933.0, + 1591.0, + 933.0, + 1626.0, + 450.0, + 1626.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 406.0, + 1626.0, + 1403.0, + 1626.0, + 1403.0, + 1661.0, + 406.0, + 1661.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 1658.0, + 706.0, + 1658.0, + 706.0, + 1694.0, + 450.0, + 1694.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 408.0, + 1691.0, + 1404.0, + 1691.0, + 1404.0, + 1729.0, + 408.0, + 1729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 450.0, + 1724.0, + 905.0, + 1724.0, + 905.0, + 1759.0, + 450.0, + 1759.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 732.0, + 1405.0, + 732.0, + 1405.0, + 770.0, + 405.0, + 770.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 764.0, + 1124.0, + 764.0, + 1124.0, + 799.0, + 449.0, + 799.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 799.0, + 1211.0, + 799.0, + 1211.0, + 834.0, + 404.0, + 834.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 834.0, + 1405.0, + 834.0, + 1405.0, + 871.0, + 404.0, + 871.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 863.0, + 572.0, + 863.0, + 572.0, + 899.0, + 449.0, + 899.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 896.0, + 1405.0, + 896.0, + 1405.0, + 937.0, + 405.0, + 937.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 448.0, + 926.0, + 596.0, + 926.0, + 596.0, + 968.0, + 448.0, + 968.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 1008.0, + 1408.0, + 1008.0, + 1408.0, + 1044.0, + 405.0, + 1044.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 402.0, + 1044.0, + 1405.0, + 1044.0, + 1405.0, + 1080.0, + 402.0, + 1080.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 452.0, + 1075.0, + 712.0, + 1075.0, + 712.0, + 1110.0, + 452.0, + 1110.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 404.0, + 1801.0, + 1408.0, + 1801.0, + 1408.0, + 1839.0, + 404.0, + 1839.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 1835.0, + 660.0, + 1835.0, + 660.0, + 1869.0, + 449.0, + 1869.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 405.0, + 1868.0, + 1405.0, + 1868.0, + 1405.0, + 1903.0, + 405.0, + 1903.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 1899.0, + 953.0, + 1899.0, + 953.0, + 1937.0, + 449.0, + 1937.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 408.0, + 1937.0, + 446.0, + 1937.0, + 446.0, + 1966.0, + 408.0, + 1966.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 448.0, + 1933.0, + 1405.0, + 1933.0, + 1405.0, + 1969.0, + 448.0, + 1969.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 449.0, + 1967.0, + 926.0, + 1967.0, + 926.0, + 2001.0, + 449.0, + 2001.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 296.0, + 266.0, + 1405.0, + 266.0, + 1405.0, + 299.0, + 296.0, + 299.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 295.0, + 1407.0, + 295.0, + 1407.0, + 331.0, + 294.0, + 331.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 294.0, + 326.0, + 1405.0, + 326.0, + 1405.0, + 364.0, + 294.0, + 364.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 293.0, + 356.0, + 1298.0, + 356.0, + 1298.0, + 396.0, + 293.0, + 396.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 297.0, + 576.0, + 1406.0, + 576.0, + 1406.0, + 614.0, + 297.0, + 614.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 607.0, + 1406.0, + 607.0, + 1406.0, + 644.0, + 298.0, + 644.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 298.0, + 637.0, + 1383.0, + 637.0, + 1383.0, + 673.0, + 298.0, + 673.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 369.0, + 408.0, + 1218.0, + 408.0, + 1218.0, + 444.0, + 369.0, + 444.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 367.0, + 446.0, + 1406.0, + 446.0, + 1406.0, + 489.0, + 367.0, + 489.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 391.0, + 484.0, + 532.0, + 484.0, + 532.0, + 521.0, + 391.0, + 521.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 371.0, + 519.0, + 1065.0, + 519.0, + 1065.0, + 558.0, + 371.0, + 558.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 359.0, + 968.0, + 846.0, + 968.0, + 846.0, + 1008.0, + 359.0, + 1008.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 357.0, + 1111.0, + 675.0, + 1111.0, + 675.0, + 1155.0, + 357.0, + 1155.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 361.0, + 690.0, + 586.0, + 690.0, + 586.0, + 729.0, + 361.0, + 729.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 357.0, + 1759.0, + 1205.0, + 1759.0, + 1205.0, + 1802.0, + 357.0, + 1802.0 + ], + "score": 1.0, + "text": "" + }, + { + "category_id": 15, + "poly": [ + 360.0, + 1448.0, + 1409.0, + 1448.0, + 1409.0, + 1491.0, + 360.0, + 1491.0 + ], + "score": 1.0, + "text": "" + } + ], + "page_info": { + "page_no": 13, + "width": 1700, + "height": 2200 + } + } +] \ No newline at end of file diff --git a/parse/dev/vKBdabh_WV/images/004e1408ffac58b0f7f37ea0b36650cec00e17e6541f60cd8a7c8bcc380ac26b.jpg b/parse/dev/vKBdabh_WV/images/004e1408ffac58b0f7f37ea0b36650cec00e17e6541f60cd8a7c8bcc380ac26b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..597e04451e3274bf7d4a402b950913f1ce62360b --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/004e1408ffac58b0f7f37ea0b36650cec00e17e6541f60cd8a7c8bcc380ac26b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02c996ce15d925f499bbf92d273930532ddc8b73c28468777f6a5be2d093d67a +size 69483 diff --git a/parse/dev/vKBdabh_WV/images/06d1cf6a3bd1083a28d821dd1b5eeb2c6e64466395962b7687226b59e85a25a8.jpg b/parse/dev/vKBdabh_WV/images/06d1cf6a3bd1083a28d821dd1b5eeb2c6e64466395962b7687226b59e85a25a8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a7c77f5621fe61c58e3a667f12235ea0946b8d56 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/06d1cf6a3bd1083a28d821dd1b5eeb2c6e64466395962b7687226b59e85a25a8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:144e33ce2a5921be48e3841bcdf2d07404f249ddf0692a8d6836ae5181b22ec5 +size 29278 diff --git a/parse/dev/vKBdabh_WV/images/09d3dfa82fa398eb9d40ec3acb54491a4d36751e7aae18c43668615713bf8f86.jpg b/parse/dev/vKBdabh_WV/images/09d3dfa82fa398eb9d40ec3acb54491a4d36751e7aae18c43668615713bf8f86.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d972d6c57c5f1f282e054df6621449fb3efbe1df --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/09d3dfa82fa398eb9d40ec3acb54491a4d36751e7aae18c43668615713bf8f86.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1fbd97e6a85a225225ed3fd4d7070c45e25457fb6d2d3f89db05e4ee7954a7a1 +size 453682 diff --git a/parse/dev/vKBdabh_WV/images/13b70170c613a799408f9646f8968f0334867f496d6726f314786247b8bf270b.jpg b/parse/dev/vKBdabh_WV/images/13b70170c613a799408f9646f8968f0334867f496d6726f314786247b8bf270b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..edef91d035c56e201dae48f46ca016d37256530b --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/13b70170c613a799408f9646f8968f0334867f496d6726f314786247b8bf270b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:84bc40b13e9a5d40c58b4b93f77d1af16df4d1d1cacd007c11db2ec7662bdc33 +size 8585 diff --git a/parse/dev/vKBdabh_WV/images/171f70ead8ef060a2cfbc4abfd78f00baa417cb963825edba7b7b6062e5f1aa1.jpg b/parse/dev/vKBdabh_WV/images/171f70ead8ef060a2cfbc4abfd78f00baa417cb963825edba7b7b6062e5f1aa1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b39075e2e0eccb78d90b4b4e18f6806091820bc1 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/171f70ead8ef060a2cfbc4abfd78f00baa417cb963825edba7b7b6062e5f1aa1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f2bb08168c41f4a85741e5f70048d02b62a39dd069fdd34501ff5219ae8e855e +size 41175 diff --git a/parse/dev/vKBdabh_WV/images/212fa793b7d765338284ea2ba0f91b5f07b3ad99a494d09a9c633eca7b27d5d8.jpg b/parse/dev/vKBdabh_WV/images/212fa793b7d765338284ea2ba0f91b5f07b3ad99a494d09a9c633eca7b27d5d8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..07e5c44da203ab65a57ce4951265bbf68eb7ad0b --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/212fa793b7d765338284ea2ba0f91b5f07b3ad99a494d09a9c633eca7b27d5d8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b742fff5794e2dcec9ba436965918a88ac4739fddd5bc675381cf1ef6ddce81 +size 21495 diff --git a/parse/dev/vKBdabh_WV/images/23aacb3e79153ea71efcbeaceba9f227c9e98f5b10823f4bca8f23d92bb22f7e.jpg b/parse/dev/vKBdabh_WV/images/23aacb3e79153ea71efcbeaceba9f227c9e98f5b10823f4bca8f23d92bb22f7e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..db546fde0a46f4c702a53fa0642893173ab2f13a --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/23aacb3e79153ea71efcbeaceba9f227c9e98f5b10823f4bca8f23d92bb22f7e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd43e82fabc0f82ee45af2c0e8107b44b7a5db5113929a645c963a9fcda9b269 +size 5685 diff --git a/parse/dev/vKBdabh_WV/images/276bdcc166d343a82e2fe0d299041d6a498983e773f8441fcc2d6150ccaa3910.jpg b/parse/dev/vKBdabh_WV/images/276bdcc166d343a82e2fe0d299041d6a498983e773f8441fcc2d6150ccaa3910.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c03cdde9c3cf743da748514814960cf31f4b0707 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/276bdcc166d343a82e2fe0d299041d6a498983e773f8441fcc2d6150ccaa3910.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6a960a19f57101a23f1f60514ab7b46ad847733187274439b24d8481cea5e132 +size 25666 diff --git a/parse/dev/vKBdabh_WV/images/27fa4a59a578adc236321b15d5bf0ad37405c73accca0679f9791677662c3d91.jpg b/parse/dev/vKBdabh_WV/images/27fa4a59a578adc236321b15d5bf0ad37405c73accca0679f9791677662c3d91.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b4f096cb954746a156a3543aed7e13301def16ae --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/27fa4a59a578adc236321b15d5bf0ad37405c73accca0679f9791677662c3d91.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6ad8916d1e0241dbd82e483bd1dc80208c1196734e0c0fea407e3c13065c2e6 +size 12262 diff --git a/parse/dev/vKBdabh_WV/images/2969166062520b5fc2c010e34f35b5fc2fc286c9c1fea1d27a9ea4c4fa921b41.jpg b/parse/dev/vKBdabh_WV/images/2969166062520b5fc2c010e34f35b5fc2fc286c9c1fea1d27a9ea4c4fa921b41.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b2f09bef2d0d970317998ff2f54ec66682e2f63c --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/2969166062520b5fc2c010e34f35b5fc2fc286c9c1fea1d27a9ea4c4fa921b41.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76fff0f708e0bed1c04ca9463ba89df4d46553da1e1802c5847b0f8bd810cd49 +size 20737 diff --git a/parse/dev/vKBdabh_WV/images/2a7561c5a223c69270bed3de19d1f17c34eec95efd9e19ffe0c1e790ce82cdc0.jpg b/parse/dev/vKBdabh_WV/images/2a7561c5a223c69270bed3de19d1f17c34eec95efd9e19ffe0c1e790ce82cdc0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4f2a0a7df6bd4d312598397ea960b3b654909f0a --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/2a7561c5a223c69270bed3de19d1f17c34eec95efd9e19ffe0c1e790ce82cdc0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87c8be84758d5e411454320eddaa7de5a1a3fd8d63cc13bb1d8b07ff625a3669 +size 22985 diff --git a/parse/dev/vKBdabh_WV/images/3f20dd6b171e97fbcafe42fda12f6bbf12e0527507667de291ea1dcf9db9c3c6.jpg b/parse/dev/vKBdabh_WV/images/3f20dd6b171e97fbcafe42fda12f6bbf12e0527507667de291ea1dcf9db9c3c6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9b5b1f3ab619ab03da27d3e48f2ed7fa501c3a6e --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/3f20dd6b171e97fbcafe42fda12f6bbf12e0527507667de291ea1dcf9db9c3c6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:684a2ffe56762885367b0b6c8c85ea2753422b7ecb578ed922e4b0a661c18317 +size 155342 diff --git a/parse/dev/vKBdabh_WV/images/43926468ccefafd6bd75543dc8e531e2988a002ca5de2d33378b929385d38043.jpg b/parse/dev/vKBdabh_WV/images/43926468ccefafd6bd75543dc8e531e2988a002ca5de2d33378b929385d38043.jpg new file mode 100644 index 0000000000000000000000000000000000000000..06e628a5cb2885577dfe22fcb861f56dca7722c8 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/43926468ccefafd6bd75543dc8e531e2988a002ca5de2d33378b929385d38043.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ede0e8bba1f58bf60128c737d6a204c2141ad2acafabb699c6a3a2d15284c1ff +size 4525 diff --git a/parse/dev/vKBdabh_WV/images/48226ef1e175617f0165acf17ede7ac4bacdd98bc79a27ad8f47d234fe87bd67.jpg b/parse/dev/vKBdabh_WV/images/48226ef1e175617f0165acf17ede7ac4bacdd98bc79a27ad8f47d234fe87bd67.jpg new file mode 100644 index 0000000000000000000000000000000000000000..769841f8c7793bc9d309d01a5fa6ff85c692e9e0 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/48226ef1e175617f0165acf17ede7ac4bacdd98bc79a27ad8f47d234fe87bd67.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f0fc7238b7eb475c2a7378b53a26f3fb88127b38ec96b679cd95959de29eb98 +size 8408 diff --git a/parse/dev/vKBdabh_WV/images/4dba508f5c7da7efe6502c507c2200385a4027a752792df4c28c05a2e65c1c25.jpg b/parse/dev/vKBdabh_WV/images/4dba508f5c7da7efe6502c507c2200385a4027a752792df4c28c05a2e65c1c25.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8ea2d1f0605bad528af76fa2d502698e37857624 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/4dba508f5c7da7efe6502c507c2200385a4027a752792df4c28c05a2e65c1c25.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0dd7a81a8aa6fd0fd804441da726d9e99a672cb0e10066df03e272470dad671 +size 11119 diff --git a/parse/dev/vKBdabh_WV/images/535c42f81677a785c24479eac345b13f21ff954dc392f93f7fe587e75bc49998.jpg b/parse/dev/vKBdabh_WV/images/535c42f81677a785c24479eac345b13f21ff954dc392f93f7fe587e75bc49998.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b34291d60fecf197652ed55e6314c29a0552c0ae --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/535c42f81677a785c24479eac345b13f21ff954dc392f93f7fe587e75bc49998.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:739c821cdaf06df0496548d114f8d181901f545a2579ac88dab7a932a525f52c +size 5139 diff --git a/parse/dev/vKBdabh_WV/images/5758a67dd58d89db3a283fb03efae06e8ceae580f7cb8686e24050e0a96f62f9.jpg b/parse/dev/vKBdabh_WV/images/5758a67dd58d89db3a283fb03efae06e8ceae580f7cb8686e24050e0a96f62f9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9bc47c6170206c231dec7841edc74c94b8556926 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/5758a67dd58d89db3a283fb03efae06e8ceae580f7cb8686e24050e0a96f62f9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cdb202f167ef241cbd37f3dd21070f49c961c73e689479ad52a6b7917ee9ab44 +size 30799 diff --git a/parse/dev/vKBdabh_WV/images/5ba930c598e50dda0eab7737ce6e68f43f25082f6852d01333ecfbde4e9e26a7.jpg b/parse/dev/vKBdabh_WV/images/5ba930c598e50dda0eab7737ce6e68f43f25082f6852d01333ecfbde4e9e26a7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2a6177d4de909c8858702681fe0c454b623c41c4 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/5ba930c598e50dda0eab7737ce6e68f43f25082f6852d01333ecfbde4e9e26a7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc4f002a4fee84e89c0a0d941ec517bf00b3f07e820af7e5290b08e9d6b279b2 +size 3850 diff --git a/parse/dev/vKBdabh_WV/images/5fb76eecaa0a33cd03ef513471751d35391927d1a78276b562f376a3d961a18a.jpg b/parse/dev/vKBdabh_WV/images/5fb76eecaa0a33cd03ef513471751d35391927d1a78276b562f376a3d961a18a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b00d7cd428f20b7eef85bdcafc5dd3239938ffc4 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/5fb76eecaa0a33cd03ef513471751d35391927d1a78276b562f376a3d961a18a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9c37f6280b3d73b3324beecfbf4085968e8bce1be87e6c92c6804e16727cd4a +size 4675 diff --git a/parse/dev/vKBdabh_WV/images/631085016e5fbf7f63dae62f3bc0911dadc36cbcffde1225e4ee7f107c08ebd9.jpg b/parse/dev/vKBdabh_WV/images/631085016e5fbf7f63dae62f3bc0911dadc36cbcffde1225e4ee7f107c08ebd9.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6ced1a6c015e13c24c59f2a9caacea8d4c569878 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/631085016e5fbf7f63dae62f3bc0911dadc36cbcffde1225e4ee7f107c08ebd9.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef741b9944d10b62d6c80780708dcd0676aa82a582cdbcc49236d19b95cff686 +size 26065 diff --git a/parse/dev/vKBdabh_WV/images/66b53dde9e387914f911c46162b2ee1500e1db7f81732c8a89d889919359d4da.jpg b/parse/dev/vKBdabh_WV/images/66b53dde9e387914f911c46162b2ee1500e1db7f81732c8a89d889919359d4da.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6bd9d9e9a0d349245352eb06ec193edc60f9f5c1 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/66b53dde9e387914f911c46162b2ee1500e1db7f81732c8a89d889919359d4da.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:717fe0ec3fbb848e424f0caccea7e69ed512dc832fe6f133be60311d09448228 +size 83333 diff --git a/parse/dev/vKBdabh_WV/images/6cc461505612aaa96f12cb9218bf3e55dbb5b35c739b02afaa9565d7bef24dba.jpg b/parse/dev/vKBdabh_WV/images/6cc461505612aaa96f12cb9218bf3e55dbb5b35c739b02afaa9565d7bef24dba.jpg new file mode 100644 index 0000000000000000000000000000000000000000..881154d6187b99b2e59628fc9b8448e611383e0a --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/6cc461505612aaa96f12cb9218bf3e55dbb5b35c739b02afaa9565d7bef24dba.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:299652aeb8409a3562ff8feae8db005d7a5d2cdf415b6de9ee82b85720b72649 +size 453494 diff --git a/parse/dev/vKBdabh_WV/images/76fb1ff43fe83d341b98323589d65260720c16d6caae0f2c92cf0d6edc85bb5a.jpg b/parse/dev/vKBdabh_WV/images/76fb1ff43fe83d341b98323589d65260720c16d6caae0f2c92cf0d6edc85bb5a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..113fedc411f85e30d97a5c9ebd2b6f0fdcf5ff2b --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/76fb1ff43fe83d341b98323589d65260720c16d6caae0f2c92cf0d6edc85bb5a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:622dbe9aa154242e1463b769ce4a16efd5db4c3215385da86549564a98970ce8 +size 15321 diff --git a/parse/dev/vKBdabh_WV/images/7a3bdfe695e37513957941a7260fb0d02052baa42c291e9d7d7c183015706ee3.jpg b/parse/dev/vKBdabh_WV/images/7a3bdfe695e37513957941a7260fb0d02052baa42c291e9d7d7c183015706ee3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a7d6dc12bdaec18671d6224fa1379465c59272c7 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/7a3bdfe695e37513957941a7260fb0d02052baa42c291e9d7d7c183015706ee3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f08395c519e176493cb733b9f70f96d5c207af2b72b1daa28b3d3e17d7ca112b +size 43198 diff --git a/parse/dev/vKBdabh_WV/images/7af1ac4f0437daa502ac9da63b7ed2232221c26b74d5115b26c01519c59a0591.jpg b/parse/dev/vKBdabh_WV/images/7af1ac4f0437daa502ac9da63b7ed2232221c26b74d5115b26c01519c59a0591.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b02c1d70b0c6c7dc9a3746efb75ee4ed6d097d62 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/7af1ac4f0437daa502ac9da63b7ed2232221c26b74d5115b26c01519c59a0591.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb80bfbed70b914385b033a42b620c35b47070591311ecab88c1aabee446c7a1 +size 146868 diff --git a/parse/dev/vKBdabh_WV/images/7ea21caf63ada6ad0b51e0d82cdc61ee856aec1238a842e4d09a8e8117ddcc9d.jpg b/parse/dev/vKBdabh_WV/images/7ea21caf63ada6ad0b51e0d82cdc61ee856aec1238a842e4d09a8e8117ddcc9d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..09d18e6c08edd45473edfa828384c6206c2f62c3 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/7ea21caf63ada6ad0b51e0d82cdc61ee856aec1238a842e4d09a8e8117ddcc9d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8a7627440f3e70a89871b36c431eab5d3374d2c7d84fae4645e59eceee1934c4 +size 5992 diff --git a/parse/dev/vKBdabh_WV/images/8616f618b471ecd63b92b6a20c8447fc70dc69c27908f8e572ff97dbc0c4f3ff.jpg b/parse/dev/vKBdabh_WV/images/8616f618b471ecd63b92b6a20c8447fc70dc69c27908f8e572ff97dbc0c4f3ff.jpg new file mode 100644 index 0000000000000000000000000000000000000000..52292db81ffacf89f5c80b058503bd0f6efecf9d --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/8616f618b471ecd63b92b6a20c8447fc70dc69c27908f8e572ff97dbc0c4f3ff.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65ab72da6c865ffbdd0123e3c2678973580534cb7e8bc9a7e60ff250f8c43be7 +size 6734 diff --git a/parse/dev/vKBdabh_WV/images/95ae2364d302a149a728b11b5cd0623c379ff2f52461bc44c169eb33670a5c8b.jpg b/parse/dev/vKBdabh_WV/images/95ae2364d302a149a728b11b5cd0623c379ff2f52461bc44c169eb33670a5c8b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..22c202b5c50997de4250ebd5a0e45f11d1ca5f4d --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/95ae2364d302a149a728b11b5cd0623c379ff2f52461bc44c169eb33670a5c8b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df365f1f7a5b6ae0e1bd5c039c6485cb78cb222bc994ce62f00ecaf33dd9fa96 +size 28416 diff --git a/parse/dev/vKBdabh_WV/images/9a09d1236ab28dbb00606bd9bc6f797a967152dafce85430f4477a1435b58647.jpg b/parse/dev/vKBdabh_WV/images/9a09d1236ab28dbb00606bd9bc6f797a967152dafce85430f4477a1435b58647.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4eeab328f6c5cafdd4a4734f736e4f034539d20a --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/9a09d1236ab28dbb00606bd9bc6f797a967152dafce85430f4477a1435b58647.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9503b716e9fc6959df49b3e28bcc6b88fd77aab437707a40ee62aa86e385009c +size 3566 diff --git a/parse/dev/vKBdabh_WV/images/a65dabc1328759986cd9453b0ec2f8ff4816cbe1138e8a0dfd80e63e988cdf67.jpg b/parse/dev/vKBdabh_WV/images/a65dabc1328759986cd9453b0ec2f8ff4816cbe1138e8a0dfd80e63e988cdf67.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e263c88fa3f13c997f95b4b6fa571c90b645c839 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/a65dabc1328759986cd9453b0ec2f8ff4816cbe1138e8a0dfd80e63e988cdf67.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4cab2f7e2670668f2771c60fd1124a89fef810e33e36bbcaa79a2bf4eaa5d521 +size 175250 diff --git a/parse/dev/vKBdabh_WV/images/af8ae56bdc99160a056fbfb731e519000f1dc5da50e7146c194806677a95f1c8.jpg b/parse/dev/vKBdabh_WV/images/af8ae56bdc99160a056fbfb731e519000f1dc5da50e7146c194806677a95f1c8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d7d95b232c11dfd41833378960f72a31d3b8059d --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/af8ae56bdc99160a056fbfb731e519000f1dc5da50e7146c194806677a95f1c8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b7873bcb67745bc51a7e4bdaf93bb7d534356aed9cf0006651f72b750e159ea +size 5201 diff --git a/parse/dev/vKBdabh_WV/images/b047e4b4917bf548ef253d17d45801d1af71bb57627ba75831b7997778ad4bc2.jpg b/parse/dev/vKBdabh_WV/images/b047e4b4917bf548ef253d17d45801d1af71bb57627ba75831b7997778ad4bc2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d2e53732bb9988de48d153513bde9d5011b7b28a --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/b047e4b4917bf548ef253d17d45801d1af71bb57627ba75831b7997778ad4bc2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:288129b0dc6ff8b3ba4ae261884eb6fa8210058e9f1a95f5ed88f3a464bea0b8 +size 35336 diff --git a/parse/dev/vKBdabh_WV/images/b0f8ace7c05a07b0cb2ca3c7caf42a97a1fb8797fac2d3ea98936132a49cf965.jpg b/parse/dev/vKBdabh_WV/images/b0f8ace7c05a07b0cb2ca3c7caf42a97a1fb8797fac2d3ea98936132a49cf965.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b8edb818e21be529eec53f265ecf09e21eb4d9a6 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/b0f8ace7c05a07b0cb2ca3c7caf42a97a1fb8797fac2d3ea98936132a49cf965.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:efc6d19770d14a54a6b9c931984052acf28f347f86c50ed98b469ee0ac797ea5 +size 17425 diff --git a/parse/dev/vKBdabh_WV/images/b46cb90a5498469cea62cbbfd652a7a59a6d0f29c551951f28956819f8061d83.jpg b/parse/dev/vKBdabh_WV/images/b46cb90a5498469cea62cbbfd652a7a59a6d0f29c551951f28956819f8061d83.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8d32ce395456a58ef9c3d91fb7888b493d3f9c31 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/b46cb90a5498469cea62cbbfd652a7a59a6d0f29c551951f28956819f8061d83.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7290a5c34ba8d423fe854eb5e78657ad58b7d380c79fbb3bfb3c9b750c27e643 +size 51004 diff --git a/parse/dev/vKBdabh_WV/images/b4875f3868a4d4a94d5d4942798a95328942a2086ebc5c3ebabf55c61381105d.jpg b/parse/dev/vKBdabh_WV/images/b4875f3868a4d4a94d5d4942798a95328942a2086ebc5c3ebabf55c61381105d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f0744908662a67f4cfc9b7e325116ba50dc96e43 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/b4875f3868a4d4a94d5d4942798a95328942a2086ebc5c3ebabf55c61381105d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4d967f7e4e4332fbfd96393d3b3d6caef58dc2849bf1b10243fcc8d1e0e1ede +size 4609 diff --git a/parse/dev/vKBdabh_WV/images/c4807975bc736e8bedf554a75b321eca90de0361c71d4219df36c06d482e14bf.jpg b/parse/dev/vKBdabh_WV/images/c4807975bc736e8bedf554a75b321eca90de0361c71d4219df36c06d482e14bf.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d68a509831a05335b259c75644c35542260b01c9 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/c4807975bc736e8bedf554a75b321eca90de0361c71d4219df36c06d482e14bf.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a007f51c9970eb8d3ad440b186667a60932457e9d9e904c45aecf4e5c5e0cd2 +size 452765 diff --git a/parse/dev/vKBdabh_WV/images/c65d9b80bc771a7f24a4c9491464f0d264572127c957ca71436a6b0b2057fd7e.jpg b/parse/dev/vKBdabh_WV/images/c65d9b80bc771a7f24a4c9491464f0d264572127c957ca71436a6b0b2057fd7e.jpg new file mode 100644 index 0000000000000000000000000000000000000000..26f20804f5059346bf43412838a41dac207dfe5f --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/c65d9b80bc771a7f24a4c9491464f0d264572127c957ca71436a6b0b2057fd7e.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87578bc2f83dccf08aaf2e11c5beef51e97e7484c6c7ee0bc96b95249edbccb1 +size 101555 diff --git a/parse/dev/vKBdabh_WV/images/cda5ef6a0b48f9ae80509ef87ba413a6390f0555e52fe0a5fec253b18b72c316.jpg b/parse/dev/vKBdabh_WV/images/cda5ef6a0b48f9ae80509ef87ba413a6390f0555e52fe0a5fec253b18b72c316.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ddbd2e72d68052fe5b00918ee9db2cf255d944b3 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/cda5ef6a0b48f9ae80509ef87ba413a6390f0555e52fe0a5fec253b18b72c316.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c4b08647264928b0728f5c3fbf2a0949451ce976e53fd7c5cf410463483d439f +size 10359 diff --git a/parse/dev/vKBdabh_WV/images/d0a47347790edfbe68c6b659cc9e6a37308741ce026c66e6c94b8805bf2a3e91.jpg b/parse/dev/vKBdabh_WV/images/d0a47347790edfbe68c6b659cc9e6a37308741ce026c66e6c94b8805bf2a3e91.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0483e1d8493bf08a214e9d984e6e943f920b90e6 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/d0a47347790edfbe68c6b659cc9e6a37308741ce026c66e6c94b8805bf2a3e91.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c9336554f3131bbce2e361a19f798acbbcf023c569bbf0844b1c07f29417315 +size 5767 diff --git a/parse/dev/vKBdabh_WV/images/e63c8c8932e900e8f3403624ddeac57138060ddc5f58c9f0d4a25a74b770be8c.jpg b/parse/dev/vKBdabh_WV/images/e63c8c8932e900e8f3403624ddeac57138060ddc5f58c9f0d4a25a74b770be8c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0300b65b2a0cbb69bd8acd2cb6d126ec6a3f3c81 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/e63c8c8932e900e8f3403624ddeac57138060ddc5f58c9f0d4a25a74b770be8c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1cceca996178402a3189ea491f5d81d46f3dd3019ee455ee7ef53f8ede81e02f +size 9702 diff --git a/parse/dev/vKBdabh_WV/images/f7bbb5d79b7cda97e9e1d6c48ae3af59f3c7f64ec4cc5c54faf887beff7cfc31.jpg b/parse/dev/vKBdabh_WV/images/f7bbb5d79b7cda97e9e1d6c48ae3af59f3c7f64ec4cc5c54faf887beff7cfc31.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c424ba024b4c62ad53effa27830bcae9e1c588a2 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/f7bbb5d79b7cda97e9e1d6c48ae3af59f3c7f64ec4cc5c54faf887beff7cfc31.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49f194d79b5056007da54aa23970c52926b9426594992c348fa0388bb09673c3 +size 5193 diff --git a/parse/dev/vKBdabh_WV/images/fa260f974a9b2b80c3a56d5a6e468a5b7aea9da171c2c0112b86ee16a281daff.jpg b/parse/dev/vKBdabh_WV/images/fa260f974a9b2b80c3a56d5a6e468a5b7aea9da171c2c0112b86ee16a281daff.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f6ec334dd50eb2c96716b4d71fe3edafcc5d2441 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/fa260f974a9b2b80c3a56d5a6e468a5b7aea9da171c2c0112b86ee16a281daff.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:00d97883d236a45d732c3325c43a5d595ecfae107ecbf52b0afe2e7202661e67 +size 2668 diff --git a/parse/dev/vKBdabh_WV/images/fc0214da94ee7b1a8c3a8f6552324165f4c51da45aeeb6290ff4de45df5f06f3.jpg b/parse/dev/vKBdabh_WV/images/fc0214da94ee7b1a8c3a8f6552324165f4c51da45aeeb6290ff4de45df5f06f3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..499ba6f69d8f5a584a20e6bd1345d882bebcefdc --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/fc0214da94ee7b1a8c3a8f6552324165f4c51da45aeeb6290ff4de45df5f06f3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9cdfe87617bf6380b866059723b43333ff67285393a055576b7218c5ba52e87a +size 8384 diff --git a/parse/dev/vKBdabh_WV/images/fd0d7e0fc8e6d3e6f612fa9f919707ac0cca436a75c45e1c5ffa751c053281af.jpg b/parse/dev/vKBdabh_WV/images/fd0d7e0fc8e6d3e6f612fa9f919707ac0cca436a75c45e1c5ffa751c053281af.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7463e3a23156b95680b808279c0136216283fcb4 --- /dev/null +++ b/parse/dev/vKBdabh_WV/images/fd0d7e0fc8e6d3e6f612fa9f919707ac0cca436a75c45e1c5ffa751c053281af.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca9d4c99fb8aef3b4d3bed82bfb9422478724ccf73aac428b5a1e83bfae8d247 +size 30218 diff --git a/parse/dev/vruwp11pWnO/vruwp11pWnO.md b/parse/dev/vruwp11pWnO/vruwp11pWnO.md new file mode 100644 index 0000000000000000000000000000000000000000..987aa392ec64708e552a105373cb151c2ba86532 --- /dev/null +++ b/parse/dev/vruwp11pWnO/vruwp11pWnO.md @@ -0,0 +1,285 @@ +# IMPROVING AND ASSESSING ANOMALY DETECTORS FOR LARGE-SCALE SETTINGS + +Anonymous authors Paper under double-blind review + +# ABSTRACT + +Detecting out-of-distribution examples is important for safety-critical machine learning applications such as detecting novel biological phenomena and self-driving cars. However, existing research mainly focuses on simple small-scale settings. To set the stage for more realistic out-of-distribution detection, we depart from small-scale settings and explore large-scale multiclass and multi-label settings with high-resolution images and thousands of classes. To make future work in real-world settings possible, we create new benchmarks for three large-scale settings. To test ImageNet multiclass anomaly detectors, we introduce a new dataset of anomalous species. We leverage ImageNet-21K to evaluate PASCAL VOC and COCO multilabel anomaly detectors. Third, we introduce a new benchmark for anomaly segmentation by introducing a segmentation benchmark with road anomalies. We conduct extensive experiments in these more realistic settings for out-of-distribution detection and find that a surprisingly simple detector based on the maximum logit outperforms prior methods in all the large-scale multi-class, multi-label, and segmentation tasks, establishing a simple new baseline for future work. + +# 1 INTRODUCTION + +Out-of-distribution (OOD) detection is a valuable tool for developing safe and reliable machine learning (ML) systems. Detecting anomalous inputs allows systems to initiate a conservative fallback policy or defer to human judgment. As an important component of ML Safety (Hendrycks et al., 2021), OOD detection is important for safety-critical applications such as self-driving cars and detecting novel microorganisms. Accordingly, research on out-of-distribution detection has a rich history spanning several decades (Schölkopf et al., 1999; Breunig et al., 2000; Emmott et al., 2015). Recent work leverages deep neural representations for out-of-distribution detection in complex domains, such as image data (Hendrycks & Gimpel, 2017; Lee et al., 2018a; Mohseni et al., 2020; Hendrycks et al., 2019b). However, these works still primarily use small-scale datasets with low-resolution images and few classes. As the community moves towards more realistic, large-scale settings, strong baselines and high-quality benchmarks are imperative for future progress. + +Large-scale datasets such as ImageNet (Deng et al., 2009) and Places365 (Zhou et al., 2017) present unique challenges not seen in small-scale settings, such as a plethora of fine-grained object classes. We demonstrate that the maximum softmax probability (MSP) detector, a state-of-the-art method for small-scale problems, does not scale well to these challenging conditions. Through extensive experiments, we identify a detector based on the maximum logit (MaxLogit) that greatly outperforms the MSP and other strong baselines in large-scale multi-class anomaly segmentation. To facilitate further research in this setting, we also collect a new out-of-distribution test dataset suitable for models trained on highly diverse datasets. Shown in Figure 2, our Species dataset contains diverse, anomalous species that do not overlap ImageNet-21K which has approximately twenty two thousand classes. Species avoids data leakage and enables a stricter evaluation methodology for ImageNet-21K models. Using Species to conduct more controlled experiments without train-test overlap, we find that contrary to prior claims (Fort et al., 2021; Koner et al., 2021), Vision Transformers (Dosovitskiy et al., 2021a) pre-trained on ImageNet-21K are not substantially better at out-of-distribution detection. + +Moreover, in the common real-world case of multi-label data, the MSP detector cannot naturally be applied in the first place, as it requires softmax probabilities. To enable research into the multi-label setting for anomaly detection, we contribute a multi-label experimental setup and explore various methods on large-scale multi-label datasets. We find that the MaxLogit detector from our investigation into the large-scale multi-class setting generalizes well to multi-label data and again outperforms all other baselines. + +![](images/8595abd6b64885b25129dadf2053a7309f448d326b5e56de08cc3f4b6a1ec9f1.jpg) +Figure 1: We scale up out-of-distribution detection to large-scale multi-class datasets with thousands of classes, multi-label datasets with complex scenes, and anomaly segmentation in driving environments. We introduce new benchmarks for all three settings. In all of these settings, we find that an OOD detector based on the maximum logit outperforms previous methods, establishing a strong and versatile baseline for future work on large-scale OOD detection. The bottom-right shows a scene from our new anomaly segmentation benchmark and the predicted anomaly using a state-of-the-art detector. + +In addition to focusing on small-scale datasets, most existing benchmarks for anomaly detection treat entire images as anomalies. In practice, an image could be anomalous in localized regions while being in-distribution elsewhere. Knowing which regions of an image are anomalous could allow for safer handling of unfamiliar objects in the case of self-driving cars. Creating a benchmark for this task is difficult, though, as simply cutting and pasting anomalous objects into images introduces various unnatural giveaway cues such as edge effects, mismatched orientation, and lighting, all of which trivialize the task of anomaly segmentation (Blum et al., 2019). + +To overcome these issues, we utilize a simulated driving environment to create the novel StreetHazards dataset for anomaly segmentation. Using the Unreal Engine and the open-source CARLA simulation environment (Dosovitskiy et al., 2017), we insert a diverse array of foreign objects into driving scenes and re-render the scenes with these novel objects. This enables integration of the foreign objects into their surrounding context with correct lighting and orientation, sidestepping giveaway cues. + +To complement the StreetHazards dataset, we convert the BDD100K semantic segmentation dataset (Yu et al., 2018) into an anomaly segmentation dataset, which we call BDD-Anomaly. By leveraging the large scale of BDD100K, we reserve infrequent object classes to be anomalies. We combine this dataset with StreetHazards to form the Combined Anomalous Object Segmentation (CAOS) benchmark. The CAOS benchmark improves over previous evaluations for anomaly segmentation in driving scenes by evaluating detectors on realistic and diverse anomalies. We evaluate several baselines on the CAOS benchmark and discuss problems with porting existing approaches from earlier formulations of out-of-distribution detection. + +Despite its simplicity, we find that the MaxLogit detector outperforms all baselines on Species, our multi-class benchmark, and CAOS. In each of these three settings, we discuss why MaxLogit provides superior performance, and we show that these gains are hidden if one looks at small-scale problems alone. The code for our experiments and the Species and CAOS datasets are available at [anonymized]. Our new baseline combined with Species and CAOS benchmarks pave the way for future research on large-scale out-of-distribution detection. + +# Anomalous Species Dataset + +![](images/654db4cbaa3ed751b6c6dce51acbc564fb80912d3bccb104063743fd809b6f2d.jpg) +Figure 2: The Species out-of-distribution dataset is designed for large-scale anomaly detectors pretrained on datasets as diverse as ImageNet-21K. When models are pretrained on ImageNet-21K, many previous OOD detection datasets may overlap with the pretraining set, resulting in erroneous evaluations. To rectify this, Species is comprised of hundreds of anomalous species that are disjoint from ImageNet-21K classes and enables the evaluation of cutting-edge models. + +# 2 RELATED WORK + +Multi-Class Out-of-Distribution Detection. A recent line of work leverages deep neural representations from multi-class classifiers to perform out-of-distribution (OOD) detection on highdimensional data, including images, text, and speech data. Hendrycks & Gimpel (2017) formulate the task and propose the simple baseline of using the maximum softmax probability of the classifier on an input to gauge whether the input is out-of-distribution. In particular, they formulate the task as distinguishing between examples from an in-distribution dataset and various OOD datasets. Importantly, entire images are treated as out-of-distribution. + +Continuing this line of work, Lee et al. (2018a) propose to improve the neural representation of the classifier to better separate OOD examples. They use generative adversarial networks to produce neardistribution examples and induce uniform posteriors on these synthetic OOD examples. Hendrycks et al. (2019b) observe that outliers are often easy to obtain in large quantity from diverse, realistic datasets and demonstrate that OOD detectors trained on these outliers generalize to unseen classes of anomalies. Other work investigates improving the anomaly detectors themselves given a fixed classifier (DeVries & Taylor, 2018; Liang et al., 2018). However, as Hendrycks et al. (2019b) observe, many of these works tune hyperparameters on a particular type of anomaly that is also seen at test time, so their evaluation setting is more lenient. In this paper, all anomalies seen at test time come from entirely unseen categories and are not tuned on in any way. Hence, we do not compare to techniques such as ODIN (Liang et al., 2018). Additionally, in a point of departure from prior work, we focus primarily on large-scale images and datasets with many classes. + +Recent work has suggested that stronger representations from Vision Transformers pre-trained on ImageNet-21K can make out-of-distribution detection trivial (Fort et al., 2021; Koner et al., 2021). They evaluate models on detecting CIFAR-10 when fine-tuned on CIFAR-100 or vice versa, using models pretrained on ImageNet-21K. However, over 1,000 classes in ImageNet-21K overlap with CIFAR-10, so it is still unclear how Vision Transformers perform at detecting entirely unseen OOD categories. We create a new OOD test dataset of anomalous species to investigate how well Vision Transformers perform in controlled OOD detection settings without data leakage and overlap. We find that Vision Transformers pre-trained on ImageNet-21K are far from solving OOD detection in large-scale settings. + +
FPR95↓AUROC↑AUPR↑
DinMSPDeVriesMaxLogitMSPDeVries MaxLogitMSPDeVriesMaxLogit
ImageNet44.246.035.884.676.987.238.230.545.8
Places36552.685.836.676.031.185.88.22.019.2
+ +Table 1: Multi-class out-of-distribution detection results using the maximum softmax probability (MSP) baseline (Hendrycks & Gimpel, 2017), the confidence branch detector of DeVries & Taylor (2018), and our maximum logit baseline. All values are percentages and average across five out-ofdistribution test datasets. Full results on individual OOD test datasets are in the Appendix. + +Anomaly Segmentation. Several prior works explore segmenting anomalous image regions. One line of work uses the WildDash dataset (Zendel et al., 2018), which contains numerous annotated driving scenes in conditions such as snow, fog, and rain. The WildDash test set contains fifteen “negative images” from different domains for which the goal is to mark the entire image as out-ofdistribution. Thus, while the task is segmentation, the anomalies do not exist as objects within an otherwise in-distribution scene. This setting is similar to that explored by Hendrycks & Gimpel (2017), in which whole images from other datasets serve as out-of-distribution examples. + +To approach anomaly segmentation on WildDash, Krešo et al. (2018) train on multiple semantic segmentation domains and treat regions of images from the WildDash driving dataset as out-ofdistribution if they are segmented as regions from different domains, i.e. indoor classes. Bevandic´ et al. (2018) use ILSVRC 2012 images and train their network to segment the entirety of these images as out-of-distribution. + +In medical anomaly segmentation and product fault detection, anomalies are regions of otherwise in-distribution images. Baur et al. (2019) segment anomalous regions in brain MRIs using pixelwise reconstruction loss. Similarly, Haselmann et al. (2018) perform product fault detection using pixel-wise reconstruction loss and introduce an expansive dataset for segmentation of product faults. In these relatively simple domains, reconstruction-based approaches work well. In contrast to medical anomaly segmentation and fault detection, we consider complex images from street scenes. These images have high variability in scene layout and lighting, and hence are less amenable to reconstruction-based techniques. + +The two works closest to our own are the Lost and Found (Pinggera et al., 2016) and Fishyscapes (Blum et al., 2019) datasets. The Lost and Found dataset consists of real images in a driving environment with small road hazards. The images were collected to mirror the Cityscapes dataset (Cordts et al., 2016) but are only collected from one city and so have less diversity. The dataset contains 35 unique anomalous objects, and methods are allowed to train on many of these. For Lost and Found, only nine unique objects are truly unseen at test time. Crucially, this is a different evaluation setting from our own, where anomalous objects are not revealed at training time, so their dataset is not directly comparable. Nevertheless, the BDD-Anomaly dataset fills several gaps in Lost and Found. First, the images are more diverse, because they are sourced from a more recent and comprehensive semantic segmentation dataset. Second, the anomalies are not restricted to small, sparse road hazards. Concretely, anomalous regions in Lost and Found take up $0 . 1 1 \%$ of the image on average, whereas anomalous regions in the BDD-Anomaly dataset are larger and fill $0 . 8 3 \%$ of the image on average. Finally, although the BDD-Anomaly dataset treats three categories as anomalous, compared to Lost and Found it has far more unique anomalous objects. + +The Fishyscapes benchmark for anomaly segmentation consists of cut-and-paste anomalies from out-of-distribution domains. This is problematic, because the anomalies stand out as clearly unnatural in context. For instance, the orientation of anomalous objects is unnatural, and the lighting of the cut-and-paste patch differs from the lighting in the original image, providing an unnatural cue to anomaly detectors that would not exist for real anomalies. Figure 7 shows an example of these inconsistencies. Techniques for detecting image manipulation (Zhou et al., 2018; Johnson & Farid, 2005) are competent at detecting artificial image elements of this kind. Our StreetHazards dataset overcomes these issues by leveraging a simulated driving environment to naturally insert anomalous 3D models into a scene rather than overlaying 2D images. These anomalies are integrated into the scene with proper lighting and orientation, mimicking real-world anomalies and making them significantly more difficult to detect. + +![](images/c3e38c3fa532816e71e5ccfdee9b5bd80f517c0c2915dee5bd921821eb1372be.jpg) +Figure 3: Small-scale datasets such as CIFAR-10 have relatively disjoint classes, but larger-scale datasets including ImageNet-1K have several classes with high visual similarity to other classes. This implies that large-scale classifiers disperse probability mass among several classes. If the prediction confidence is used for out-of-distribution detection, then images which have similarities to other classes will often wrongly be deemed out-of-distribution due to low and dispersed confidence. This motivates our MaxLogit out-of-distribution detector. + +# 3 MULTI-CLASS PREDICTION FOR OOD DETECTION + +Problem with existing baselines. Existing baselines for anomaly detection can work well in small-scale settings. However, in more realistic settings image classification networks are often tasked with distinguishing hundreds or thousands of classes, possibly with subtle differences. This is problematic for the maximum softmax probability (MSP) baseline (Hendrycks & Gimpel, 2017), which uses the negative maximum softmax probability as the anomaly score, or $\begin{array} { r } { \dot { { \bf \varphi } } - \operatorname* { m a x } _ { k } \exp f ( x ) _ { k } / \sum _ { i } \exp f ( \bar { x _ { i } } ) _ { i } = - \operatorname* { m a x } _ { k } \hat { p } ( y = k \mid x ) } \end{array}$ , where $f ( x )$ is the unnormalized logits of classifier $f$ on input $x$ . Classifiers tend to have higher confidence on in-distribution examples than out-of-distribution examples, enabling OOD detection. Assuming single-model evaluation and no access to other anomalies or test-time adaptation, the MSP attains state-of-the-art anomaly detection performance in small-scale settings. However, we show that the MSP is problematic for realistic in-distribution datasets with many classes, such as ImageNet and Places365 (Zhou et al., 2017). Probability mass can be dispersed among visually similar classes, as shown in Figure 3. Consequently, a classifier may produce a low confidence prediction for an in-distribution image, not because the image is unfamiliar, but because the object’s exact class is difficult to determine. To circumvent this problem, we propose using the negative of the maximum unnormalized logit for an anomaly score $- \operatorname* { m a x } _ { k } f ( x ) _ { k }$ , which we call MaxLogit. Since the logits are unnormalized, they are not affected by the number of classes and can serve as a better baseline for large-scale out-of-distribution detection. + +The Species Out-Of-Distribution Dataset. To enable controlled experiments and high-quality evaluations of anomaly detectors in large-scale settings, we create the Species dataset, a new outof-distribution test dataset that has no overlapping classes with ImageNet-21K. The Species dataset is comprised of images scraped from the iNaturalist website and contains hundreds of anomalous species grouped into seven high-level categories: Plants, Microorganisms, Amphibians, Protozoa, Fungi, Arachnids, and Insects. Example images from the Species dataset are in Figure 2. + +Setup. To evaluate the MSP baseline out-of-distribution detector and the MaxLogit detector, we use ImageNet-21K as the in-distribution dataset $\mathcal { D } _ { \mathrm { i n } }$ . To obtain representations for anomaly detection, we use models trained on ImageNet-21K-P, a cleaned version of ImageNet-21K with a train/val split (Ridnik et al., 2021a). We evaluate a TResNet-M, ViT-B-16, and Mixer-B-16 (Ridnik et al., 2021b; Dosovitskiy et al., 2021b; Tolstikhin et al., 2021), and the validation split is used for obtaining in-distribution scores. For out-of-distribution test datasets $\mathcal { D } _ { \mathrm { o u t } }$ , we use categories from the Species dataset, all of which are unseen during training. Results for these experiments are in Table 2. We also use ImageNet-1K and Places365 as in-distribution datasets $\mathcal { D } _ { \mathrm { i n } }$ , for which we use pretrained ResNet-50 models and use several out-of-distribution test datasets $\mathcal { D } _ { \mathrm { o u t } }$ . Full results with ImageNet and Places365 as in-distribution are in the Appendix. + +Table 2: Results on Species. Models and the processed version of ImageNet-21K (ImageNet-21K-P) are from Ridnik et al. (2021a). All values are percent AUROC. Species enables evaluating anomaly detectors trained on ImageNet-21K and evades class overlap issues present in prior work. Using Species to conduct more controlled experiments without class overlap issues, we find that contrary to recent claims (Fort et al., 2021), simply scaling up Vision Transformers does not make OOD detection trivial. + +
ResNetViTMLPMixer
DinDtest outMSPMaxLogitMSPMaxLogitMSPMaxLogit
Plants80.387.878.284.880.385.0
Microorganisms77.483.471.182.474.486.0
Amphibians41.848.641.948.844.451.7
Protozoa70.780.469.380.968.077.7
Fungi66.477.464.776.164.176.9
hetArachnids46.956.746.656.848.958.8
Insects47.656.448.054.648.653.8
Mean61.670.160.069.261.270.0
+ +Metrics. To evaluate out-of-distribution detectors in large-scale settings, we use three standard metrics of detection performance: area under the ROC curve (AUROC), false positive rate at $9 5 \%$ recall (FPR95), and area under the precision-recall curve (AUPR). The AUROC and AUPR are important metrics, because they give a holistic measure of performance when the cutoff for detecting anomalies is not a priori obvious or when we want to represent the performance of a detection method across several different cutoffs. + +The AUROC can be thought of as the probability that an anomalous example is given a higher score than an ordinary example. Thus, a higher score is better, and an uninformative detector has a AUROC of $50 \%$ . AUPR provides a metric more attuned to class imbalances, which is relevant in anomaly and failure detection, when the number of anomalies or failures may be relatively small. Last, the FPR95 metric consists of measuring the false positive rate at $9 5 \%$ . Since these measures are correlated, we occasionally solely present the AUROC for brevity and to preserve space. + +Results. Results on Species are shown in Table 2. Results with ImageNet-1K and Places365 as in-distribution datasets are in Table 1. We find that the proposed MaxLogit method outperforms the maximum softmax probability baseline on all out-of-distribution test datasets $\mathcal { D } _ { \mathrm { o u t } }$ . This holds true for all three models trained on ImageNet-21K. The MSP baseline is not much better than random and is has similar performance for all three model classes. This suggests that contrary to recent claims, (Fort et al., 2021) simply scaling up Vision Transformers does not make OOD detection trivial. + +# 4 MULTI-LABEL PREDICTION FOR OOD DETECTION + +Current work on out-of-distribution detection primarily considers multi-class or unsupervised settings. Yet as classifiers become more useful in realistic settings, the multi-label formulation becomes increasingly natural. To investigate out-of-distribution detection in multi-label settings, we provide a baseline and evaluation setup. + +Setup. For multi-label classification we use PASCAL VOC (Everingham et al., 2009) and MSCOCO (Lin et al., 2014) as in-distribution data. To evaluate anomaly detectors for these in-distribution datasets, we use 20 out-of-distribution classes from ImageNet-21K. These classes have no overlap with ImageNet-1K, PASCAL VOC, or MS-COCO. The 20 classes are chosen not to overlap with ImageNet-1K since the multi-label classifiers models are pre-trained on ImageNet-1K. We list the class WordNet IDs in the Appendix. + +Methods. For our experiments, we use a ResNet-101 backbone architecture pre-trained on ImageNet-1K. We replace the final layer with 2 fully connected layers and apply the logistic sigmoid function for multi-label prediction. During training we freeze the batch normalization parameters due to an insufficient number of images for proper mean and variance estimation. We train each model for 50 epochs using the Adam optimizer (Kingma & Ba, 2014) with hyperparameter values $1 0 ^ { - 4 }$ and + +Table 3: Multi-label out-of-distribution detection comparison of the Isolation Forest (iForest), Local Outlier Factor (LOF), Dropout, logit average, maximum softmax probability, and maximum logit anomaly detectors on PASCAL VOC and MS-COCO. The same network architecture is used for all three detectors. All results shown are percentages. + +
iForestLOFDropoutLogitAvgMSPMaxLogit
PASCAL VOCFPR9598.684.097.298.282.335.6
AUROC46.368.449.247.974.290.9
AUPR→↑↑37.158.445.341.365.581.2
COCOFPR95√95.678.493.394.581.840.4
AUROC个41.470.258.055.570.790.3
AUPR个63.782.076.374.082.994.0
+ +$1 0 ^ { - 5 }$ for $\beta _ { 1 }$ and $\beta _ { 2 }$ respectively. For data augmentation we use standard resizing, random crops, and random flips to obtain images of size $2 5 6 \times 2 5 6 \times 3$ . As a result of this training procedure, the mAP of the ResNet-101 on PASCAL VOC is $8 9 . 1 1 \%$ and $7 2 . 0 \%$ for MS-COCO. + +As there has been little work on out-of-distribution detection in multilabel settings, we include comparisons to classic anomaly detectors for general settings. Isolation Forest, denoted by iForest, works by randomly partitioning the space into half spaces to form a decision tree. The score is determined by how close a point is to the root of the tree. The local outlier factor (LOF) (Breunig et al., 2000) computes a local density ratio between every element and its neighbors. We set the number of neighbors as 20. iForest and LOF are both computed on features from the penultimate layer of the networks. MSP denotes a natural extension of the maximum softmax probability detector in the multi-label setting, obtained by taking the sigmoid of each output score $f ( { \boldsymbol { x } } ) _ { i }$ and computing $- \operatorname* { m a x } _ { i } \sigma ( f ( x ) _ { i } )$ . Alternatively, one can average the logit values, denoted by LogitAvg. These serve as our baseline detectors for multi-label OOD detection. We compare these baselines to the MaxLogit detector that we introduce in Section 3. As in the multi-class case, the MaxLogit anomaly score for multi-label classification is $- \operatorname* { m a x } _ { i } f ( x ) _ { i }$ . + +Results. Results are shown in Table 3. We find that MaxLogit obtains the highest performance in all cases. MaxLogit bears similarity to the MSP baseline (Hendrycks & Gimpel, 2017) but is naturally applicable to multi-label problems. These results establish the MaxLogit as an effective and natural baseline for large-scale multi-label problems. Further, the evaluation setup enables future work in out-of-distribution detection with multi-label datasets. + +# 5 THE CAOS BENCHMARK + +The Combined Anomalous Object Segmentation (CAOS) benchmark is comprised of two complementary datasets for evaluating anomaly segmentation systems on diverse, realistic anomalies. First is the StreetHazards dataset, which leverages simulation to provide a large variety of anomalous objects realistically inserted into driving scenes. Second is the BDD-Anomaly dataset, which consists of real images taken from the BDD100K dataset (Yu et al., 2018). StreetHazards contains a highly diverse array of anomalies; BDD-Anomaly contains anomalies in real-world images. Together, these datasets allow researchers to judge techniques on their ability to segment diverse anomalies as well as anomalies in real images. All images have $7 2 0 \times 1 2 8 0$ resolution. + +The StreetHazards Dataset. StreetHazards is an anomaly segmentation dataset that leverages simulation to provide diverse, realistically-inserted anomalous objects. To create the StreetHazards dataset, we use the Unreal Engine along with the CARLA simulation environment (Dosovitskiy et al., 2017). From several months of development and testing including customization of the Unreal Engine and CARLA, we can insert foreign entities into a scene while having them be properly integrated. Unlike previous work, this avoids the issues of inconsistent chromatic aberration, inconsistent lighting, edge effects, and other simple cues that an object is anomalous. Additionally, using a simulated environment allows us to dynamically insert diverse anomalous objects in any location and have them render properly with changes to lighting and weather including time of day, cloudy skies, and rain. + +We use 3 towns from CARLA for training, from which we collect RGB images and their respective semantic segmentation maps to serve as training data for semantic segmentation models. We generate a validation set from the fourth town. Finally, we reserve the fifth and sixth town as our test set. We insert anomalies taken from the Digimation Model Bank Library and semantic ShapeNet (ShapeNetSem) (Savva et al., 2015) into the test set in order to evaluate methods for out-of-distribution detection. In total, we use 250 unique anomaly models of diverse types. There are 12 classes used for training: background, road, street lines, traffic signs, sidewalk, pedestrian, vehicle, building, wall, pole, fence, and vegetation. The thirteenth class is the anomaly class that is only used at test time. We collect 5,125 image and semantic segmentation ground truth pairs for training, 1,031 pairs without anomalies for validation, and 1,500 test pairs with anomalies. + +![](images/9f6a01add5da434aa775b1cb401e714b45aeff0bd523f7abf8d24e9f6ac738fc.jpg) +Figure 4: A sample of anomalous scenes from the CAOS benchmark with model predictions and anomaly scores. The anomaly scores are thresholded to the top $10 \%$ of values for visualization. GT is ground truth, the autoencoder model is based on the spatial autoencoder used in Baur et al. (2019), MSP is the maximum softmax probability baseline (Hendrycks & Gimpel, 2017), and MaxLogit is the method we propose as a new baseline for large-scale settings. Compared to baselines, the MaxLogit detector places lower scores on in-distribution image regions, including object outlines, while also doing a better job of highlighting anomalous objects. + +The BDD-Anomaly Dataset. BDD-Anomaly is an anomaly segmentation dataset with real images in diverse conditions. We source BDD-Anomaly from BDD100K (Yu et al., 2018), a large-scale semantic segmentation dataset with diverse driving conditions. The original data consists in 7,000 images for training and 1,000 for validation. There are 18 original classes. We choose motorcycle, train, and bicycle as the anomalous object classes and remove all images with these objects from the training and validation sets. This yields 6,280 training pairs, 910 validation pairs without anomalies, and 810 testing pairs with anomalous objects. + +# 5.1 EXPERIMENTS + +Evaluation. In anomaly segmentation experiments, each pixel is treated as a prediction, resulting in many predictions to evaluate. To fit these in memory, we compute the metrics on each image and average over the images to obtain final values. + +Methods. Our first baseline is pixel-wise Maximum Softmax Probability (MSP). Introduced by Hendrycks & Gimpel (2017) for multi-class out-of-distribution detection, we directly port this baseline to anomaly segmentation. Alternatively, the background class might serve as an anomaly detector, because it contains everything not in the other classes. To test this hypothesis, “Background” uses the posterior probability of the background class as the anomaly score. The Dropout method leverages MC Dropout (Gal & Ghahramani, 2016) to obtain an epistemic uncertainty estimate. Following Kendall et al. (2015), we compute the pixel-wise posterior variance over multiple dropout masks and average across all classes, which serves as the anomaly score. We also experiment with an autoencoder baseline similar to Baur et al. (2019); Haselmann et al. (2018) where pixel-wise reconstruction loss is used as the anomaly score. This method is called AE. The “Branch” method is a direct port of the confidence branch detector from DeVries & Taylor (2018) to pixel-wise prediction. Finally, we use the MaxLogit method described in earlier sections independently on each pixel. + +Table 4: Results on the CAOS benchmark. AUPR is low across the board due to the large class imbalance, but all methods perform substantially better than chance. MaxLogit obtains the best performance. All results are percentages. + +
MSPBranchBackgroundDropoutAEMaxLogit
StreetHazardsFPR9533.768.469.079.491.726.5
AUROC87.765.758.669.966.189.3
↓↑ AUPR 个6.61.54.57.52.210.6
BDD-Anomaly AUROCFPR9524.525.640.116.674.114.0
87.785.669.790.864.092.6
AUPR3.73.91.14.30.75.4
+ +For all of the baselines except the autoencoder, we train a PSPNet (Zhao et al., 2017) decoder with a ResNet-101 encoder (He et al., 2015) for 20 epochs. We train both the encoder and decoder using SGD with momentum of 0.9, a learning rate of $2 \times 1 0 ^ { - 2 }$ , and learning rate decay of $1 0 ^ { - 4 }$ . For AE, we use a 4-layer U-Net (Ronneberger et al., 2015) with a spatial latent code as in Baur et al. (2019). The U-Net also uses batch norm and is trained for 10 epochs. Results are in Table 4. + +Results and Analysis. MaxLogit outperforms all other methods across the board by a substantial margin. The intuitive baseline of using the posterior for the background class to detect anomalies performs poorly, which suggests that the background class may not align with rare visual features. Even though reconstruction-based scores succeed in product fault segmentation, we find that the AE method performs poorly on the CAOS benchmark, which may be due to the more complex domain. AUPR for all methods is low, indicating that the large class imbalance presents a serious challenge. However, the substantial improvements with the MaxLogit method suggest that progress on this task is possible and there is much room for improvement. A comparison with other datasets is in Figure 5 (Pinggera et al., 2016; Blum et al., 2019; Jung et al., 2021). + +Figure 5: Auxiliary analysis of the MSP and the MaxLogit AUROCs using prior less comprehensive anomaly segmentation datasets. + +
MethodMSPMaxLogit
FS Lost and Found87.0%92.0%
Road Anomaly73.8%78.0%
+ +In Figure 4, we see that both MaxLogit and MSP have many false positives, as they assign high anomaly scores to semantic boundaries, a problem also observed in the recent works of (Blum et al., 2019; Angus, 2019). However, the problem is less severe with MaxLogit. A potential explanation for this is that even when the prediction confidence dips at semantic boundaries, the maximum logit can remain the same in a ‘hand-off’ procedure between the classes. Thus, MaxLogit provides a natural mechanism to combat semantic boundary artifacts that could be further explored in future work. + +# 6 CONCLUSION + +We scaled out-of-distribution detection to settings with thousands of classes and high-resolution images. We identified an issue faced by existing baselines when scaling to these settings and proposed the maximum logit detector as a natural solution. We introduced the Species dataset to enable more controlled experiments without class overlap and also investigated using multi-label classifiers for OOD detection, establishing an experimental setup for this previously unexplored setting. Finally, we introduced the CAOS benchmark for anomaly segmentation, consisting of diverse, naturally-integrated anomalous objects in driving scenes. Baseline methods on the CAOS benchmark substantially improve on random guessing but are still lacking, indicating potential for future work. Interestingly, the MaxLogit detector also provides consistent and significant gains in the multi-label and anomaly segmentation settings, thereby establishing it as a new baseline in place of the maximum softmax probability baseline on large-scale OOD detection problems. In all, we we hope that our contributions will enable further research on out-of-distribution detection for real-world safety-critical environments. + +# REFERENCES + +Matt Angus. Towards pixel-level ood detection for semantic segmentation, 2019. + +Christoph Baur, Benedikt Wiestler, Shadi Albarqouni, and Nassir Navab. Deep autoencoding models for unsupervised anomaly segmentation in brain mr images. Lecture Notes in Computer Science, pp. 161–169, 2019. ISSN 1611-3349. doi: 10.1007/978-3-030-11723-8_16. + +Petra Bevandic, Ivan Krešo, Marin Orši ´ c, and Siniša Šegvi ´ c. Discriminative out-of-distribution ´ detection for semantic segmentation, 2018. + +Hermann Blum, Paul-Edouard Sarlin, Juan Nieto, Roland Siegwart, and Cesar Cadena. The fishyscapes benchmark: Measuring blind spots in semantic segmentation, 2019. + +Markus M Breunig, Hans-Peter Kriegel, Raymond T $\mathrm { N g }$ , and Jörg Sander. Lof: identifying densitybased local outliers. In ACM sigmod record, volume 29, pp. 93–104. ACM, 2000. + +M. Cimpoi, S. Maji, I. Kokkinos, S. Mohamed, , and A. Vedaldi. Describing textures in the wild. In Computer Vision and Pattern Recognition, 2014. + +Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. The cityscapes dataset for semantic urban scene understanding. In Proc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016. + +J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei. ImageNet: A Large-Scale Hierarchical Image Database. In CVPR09, 2009. + +Terrance DeVries and Graham W Taylor. Learning confidence for out-of-distribution detection in neural networks. arXiv preprint arXiv:1802.04865, 2018. + +Alexey Dosovitskiy, German Ros, Felipe Codevilla, Antonio Lopez, and Vladlen Koltun. CARLA: An open urban driving simulator. In Proceedings of the 1st Annual Conference on Robot Learning, pp. 1–16, 2017. + +Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. ArXiv, abs/2010.11929, 2021a. + +Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. In International Conference on Learning Representations, 2021b. URL https://openreview. net/forum?id ${ . } = { }$ YicbFdNTTy. + +Andrew Emmott, Shubhomoy Das, Thomas Dietterich, Alan Fern, and Weng-Keen Wong. A meta-analysis of the anomaly detection problem, 2015. + +Mark Everingham, Luc Van Gool, Christopher K. I. Williams, John M. Winn, and Andrew Zisserman. The pascal visual object classes (voc) challenge. International Journal of Computer Vision, 88: 303–338, 2009. + +Stanislav Fort, Jie Ren, and Balaji Lakshminarayanan. Exploring the limits of out-of-distribution detection. arXiv preprint arXiv:2106.03004, 2021. + +Yarin Gal and Zoubin Ghahramani. Dropout as a bayesian approximation: Representing model uncertainty in deep learning. International Conference on Machine Learning, 2016. + +Matthias Haselmann, Dieter P Gruber, and Paul Tabatabai. Anomaly detection using deep learning based image completion. In 2018 17th IEEE International Conference on Machine Learning and Applications (ICMLA), pp. 1237–1242. IEEE, 2018. + +Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. CVPR, 2015. + +Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution examples in neural networks. ICLR, 2017. + +Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier exposure. Proceedings of the International Conference on Learning Representations, 2019a. + +Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier exposure. In International Conference on Learning Representations, 2019b. + +Dan Hendrycks, Mantas Mazeika, Saurav Kadavath, and Dawn Song. Using self-supervised learning can improve model robustness and uncertainty. In NeurIPS, 2019c. + +Dan Hendrycks, Nicholas Carlini, John Schulman, and Jacob Steinhardt. Unsolved problems in ml safety. ArXiv, abs/2109.13916, 2021. + +Micah K Johnson and Hany Farid. Exposing digital forgeries by detecting inconsistencies in lighting. In Proceedings of the 7th workshop on Multimedia and security, pp. 1–10, 2005. + +Sanghun Jung, Jungsoo Lee, Daehoon Gwak, Sungha Choi, and Jaegul Choo. Standardized max logits: A simple yet effective approach for identifying unexpected road obstacles in urban-scene segmentation. ArXiv, abs/2107.11264, 2021. + +Alex Kendall, Vijay Badrinarayanan, and Roberto Cipolla. Bayesian segnet: Model uncertainty in deep convolutional encoder-decoder architectures for scene understanding. ArXiv, abs/1511.02680, 2015. + +Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. ICLR, 2014. + +Rajat Koner, Poulami Sinhamahapatra, Karsten Roscher, Stephan Günnemann, and Volker Tresp. Oodformer: Out-of-distribution detection transformer. ArXiv, 2021. + +Ivan Krešo, Marin Oršic, Petra Bevandi ´ c, and Siniša Šegvi ´ c. Robust semantic segmentation with ´ ladder-densenet models, 2018. + +Kimin Lee, Honglak Lee, Kibok Lee, and Jinwoo Shin. Training confidence-calibrated classifiers for detecting out-of-distribution samples. International Conference on Learning Representations, 2018a. + +Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. A simple unified framework for detecting out-of-distribution samples and adversarial attacks. ArXiv, abs/1807.03888, 2018b. + +Shiyu Liang, Yixuan Li, and R. Srikant. Enhancing the reliability of out-of-distribution image detection in neural networks. International Conference on Learning Representations, 2018. + +Tsung-Yi Lin, Michael Maire, Serge Belongie, Lubomir Bourdev, Ross Girshick, James Hays, Pietro Perona, Deva Ramanan, C. Lawrence Zitnick, and Piotr Dollar. Microsoft COCO: Common objects in context. ECCV, 2014. + +Sina Mohseni, Mandar Pitale, Jbs Yadawa, and Zhangyang Wang. Self-supervised learning for generalizable out-of-distribution detection. In AAAI, 2020. + +Peter Pinggera, Sebastian Ramos, Stefan K. Gehrig, Uwe Franke, Carsten Rother, and Rudolf Mester. Lost and found: Detecting small road hazards for self-driving vehicles. 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 1099–1106, 2016. + +T. Ridnik, Emanuel Ben-Baruch, Asaf Noy, and Lihi Zelnik-Manor. Imagenet-21k pretraining for the masses. ArXiv, abs/2104.10972, 2021a. + +Tal Ridnik, Hussam Lawen, Asaf Noy, Emanuel Ben Baruch, Gilad Sharir, and Itamar Friedman. Tresnet: High performance gpu-dedicated architecture. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 1400–1409, 2021b. + +Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015, pp. 234–241, 2015. ISSN 1611-3349. doi: 10.1007/978-3-319-24574-4_28. + +Manolis Savva, Angel X. Chang, and Pat Hanrahan. Semantically-Enriched 3D Models for Commonsense Knowledge. CVPR 2015 Workshop on Functionality, Physics, Intentionality and Causality, 2015. + +Bernhard Schölkopf, Robert Williamson, Alex Smola, John Shawe-Taylor, and John Platt. Support vector method for novelty detection. In Proceedings of the 12th International Conference on Neural Information Processing Systems, NIPS’99, pp. 582–588, Cambridge, MA, USA, 1999. MIT Press. + +Ilya Tolstikhin, Neil Houlsby, Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Thomas Unterthiner, Jessica Yung, Daniel Keysers, Jakob Uszkoreit, Mario Lucic, et al. Mlp-mixer: An all-mlp architecture for vision. arXiv preprint arXiv:2105.01601, 2021. + +Fisher Yu, Yinda Zhang, Shuran Song, Ari Seff, and Jianxiong Xiao. LSUN: construction of a large-scale image dataset using deep learning with humans in the loop. CoRR, 2015. + +Fisher Yu, Wenqi Xian, Yingying Chen, Fangchen Liu, Mike Liao, Vashisht Madhavan, and Trevor Darrell. BDD100K: A diverse driving video database with scalable annotation tooling. CoRR, abs/1805.04687, 2018. + +Oliver Zendel, Katrin Honauer, Markus Murschitz, Daniel Steininger, and Gustavo Fernandez Dominguez. Wilddash-creating hazard-aware benchmarks. In Proceedings of the European Conference on Computer Vision (ECCV), pp. 402–416, 2018. + +Hengshuang Zhao, Jianping Shi, Xiaojuan Qi, Xiaogang Wang, and Jiaya Jia. Pyramid scene parsing network. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 6230–6239, 2017. + +Bolei Zhou, Agata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba. Places: A 10 million image database for scene recognition. PAMI, 2017. + +Peng Zhou, Xintong Han, Vlad I Morariu, and Larry S Davis. Learning rich features for image manipulation detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1053–1061, 2018. + +# A APPENDIX + +Table 5: B is for the maximum softmax probability baseline, M is for maximum logit, D is for the method in DeVries & Taylor (2018), and K is our own KL method described below. Both M and K are ours. Results are on ImageNet and Places365. All values are percentages and are rounded so that 99.95 rounds to 100. + +
DestFPR95↓AUROC ↑AUPR↑
DinBM DKBMDKBMDK
Gaussian2054100100 9798329252521985579
1eeeeeeRademacher2144158998 9893706254
Blobs26327288079 379917793
Textures685674598087 7685361648
LSUN6663596075 777679221938
Places3656459637279 83797927322446
Mean41.335.84636.185.287.276.988.7 3745.830.559.7
PPssse536Gaussian10671129396 35931624216
Rademacher20109118993 1010011 515.91.688
Blobs59688277298 159341231
Textures86728774657943 7911112
Places6988899291616452 65636
Mean5336.685.840.97685.831.185.819.2230.5
+ +# B FULL MULTICLASS OOD DETECTION RESULTS + +Datasets. To evaluate the MSP baseline out-of-distribution detector and the MaxLogit detector, we use the ImageNet-1K object recognition dataset and Places365 scene recognition dataset as in-distribution datasets $\mathcal { D } _ { \mathrm { i n } }$ . We use several out-of-distribution test datasets $\mathcal { D } _ { \mathrm { o u t } }$ , all of which are unseen during training. The first out-of-distribution dataset is Gaussian noise, where each example’s pixels are i.i.d. sampled from $\mathcal { N } ( 0 , 0 . 5 )$ and clipped to be contained within $[ - 1 , 1 ]$ . Another type of test-time noise is Rademacher noise, in which each pixel is i.i.d. sampled from 2 · Bernoulli $( 0 . 5 ) - 1$ , i.e. each pixel is 1 or $- 1$ with equal probability. Blob examples are more structured than noise; they are algorithmically generated blob images. Meanwhile, Textures is a dataset consisting in images of describable textures (Cimpoi et al., 2014). When evaluating the ImageNet-1K detector, we use LSUN images, a scene recognition dataset (Yu et al., 2015). Our final $\mathcal { D } _ { \mathrm { o u t } }$ is Places69, a scene classification dataset that does not share classes with Places365. In all, we evaluate against out-of-distribution examples spanning synthetic and realistic images. + +KL Matching Method. To verify our intuitions that led us to develop the MaxLogit detector, we developed a less convenient but similarly powerful technique applicable for the multiclass setting. Recall that some classes tend to be predicted with low confidence and others high confidence. The shape of predicted posterior distributions is often class dependent. + +We capture the typical shape of each class’s posterior distribution and form posterior distribution templates for each class. During test time, the network’s softmax posterior distribution is compared to these templates and an anomaly score is generated. More concretely, we compute $k$ different distributions $d _ { k }$ , one for each class. We write $d _ { k } = \mathbb { E } _ { x ^ { \prime } \sim \mathcal { X } _ { \mathrm { v a l } } } [ p ( y | x ^ { \prime } ) ]$ where $k = \mathrm { a r g m a x } _ { k } p ( y =$ $k \mid x ^ { \prime } )$ . Then for a new test input $x$ , we calculate the anomaly score $\mathrm { { \ddot { m i n } } } _ { k } ~ \mathrm { K L } [ p ( y \mid x ) \mid \mid d _ { k } ]$ rather than the MSP baseline $- \operatorname* { m a x } _ { k } p ( y = k \mid x ) .$ . Note that we utilize the validation dataset, but our KL matching method does not require the validation dataset’s labels. That said, our KL matching method is less convenient than our MaxLogit technique, and the two perform similarly. Since this technique requires more data than MaxLogit, we opt to simply use the MaxLogit in the main paper. + +Results. Observe that the proposed MaxLogit method outperforms the maximum softmax probability baseline for all three metrics on both ImageNet and Places365. These results were computed using a ResNet-50 trained on either ImageNet-1K or Places365. In the case of Places365, the AUROC improvement is over $10 \%$ . We note that the utility of the maximum logit could not be appreciated as easily in previous work’s small-scale settings. For example, using the small-scale CIFAR-10 setup of + +Hendrycks et al. Hendrycks et al. (2019a), the MSP attains an average AUROC of $9 0 . 0 8 \%$ while the maximum logit attains $9 0 . 2 2 \%$ , a minor $0 . 1 4 \%$ difference. However, in a large-scale setting, the difference can be over $10 \%$ on individual $\mathcal { D } _ { \mathrm { o u t } }$ datasets. We are not claiming that utilizing the maximum logit is a mathematically innovative formulation, only that it serves as a consistently powerful baseline for large-scale settings that went unappreciated in small-scale settings. In consequence, we suggest using the maximum logit as a new baseline for large-scale multi-class out-of-distribution detection. + +Overview of Other Detection Methods. There are other techniques in out-of-distribution detection which require other assumptions such as more training data. For instance, Hendrycks et al. (2019a); Mohseni et al. (2020) use additional training data labeled as out-of-distribution, and the MaxLogit technique can be naturally extended should such data be available. Hendrycks et al. (2019c) use rotation prediction and self-supervised learning, but we found that scaling this to the ImageNet multiclass setting did not produce strong results. The MSP baseline trained with auxiliary rotation prediction has an AUROC of $5 9 . 1 \%$ , and with MaxLogit it attains a $7 3 . 6 \%$ AUROC, over a $10 \%$ absolute improvement with MaxLogit. Nonetheless this technique did not straightforwardly scale, as the network is better without auxiliary rotation prediction. Likewise, Lee et al. (2018b) propose to use Mahalanobis distances, but in scaling this to 1000 classes, we consistently encountered NaN errors due to high condition numbers. This shows the importance of ensuring that out-of-distribution techniques can scale. + +ODIN Liang et al. (2018) assumes that, for each OOD example source, we can tune hyperparameters for detection. For this reason we do not evaluate with ODIN in the rest of the paper. However, for thoroughness, we evaluate it here. ODIN uses temperature scaling and adds an epsilon perturbation to the input in order to separate the softmax posteriors between in- and out-of-distribution images; we set these hyperparameters following DeVries & Taylor (2018). Then, MaxLogit combined with ODIN results in an FPR95 of 33.6, an AUROC of 88.8 and an AUPR of 51.3 on ImageNet. On Places365, the FPR95 is 35.3, the AUROC is 86.5, and the AUPR is 24.2. Consequently, techniques built with different assumptions can integrate well with MaxLogit. We do not train ImageNet-21K models from scratch with these methods due to limited compute. + +# C MULTI-LABEL OUT-OF-DISTRIBUTION DATASET LIST + +For multi-label classification experiments, we choose the following classes from ImageNet-21K to serve as out-of-distribution data: dolphin (n02069412), deer (n02431122), bat (n02139199), rhino (n02392434), raccoon (n02508213), octopus (n01970164), giant clam (n01959492), leech (n01937909), Venus flytrap (n12782915), cherry tree (n12641413), Japanese cherry blossoms (n12649317), red wood (n12285512), sunflower (n11978713), croissant (n07691650), stick cinnamon (n07814390), cotton (n12176953), rice (n12126084), sugar cane (n12132956), bamboo (n12147226), and tumeric (n12356395). These classes were hand-chosen so that they are distinct from VOC and COCO classes. + +# D OOD SEGMENTATION + +![](images/e830f93c8a5dee8cff942f66fbb348ee36c59204cff44e668c2183b5e4494531.jpg) +Figure 6: ROC curve with VOC as $( \mathcal { D } _ { \mathrm { i n } } )$ and non-overlapping ImageNet classes as $( \mathcal { D } _ { \mathrm { o u t } } ^ { \mathrm { t e s t } } )$ . Curves correspond to an uninformative “Random” detector, Local Outlier Factor, and the MaxLogit detector. + +We cover methods used in the paper in more depth and the modifications necessary to make the methods work with OOD detection in semantic segmentation. We use $f$ to denote the function typically a neural network, $x$ is the input image, and $y _ { i , j }$ is the prediction for pixel $i , j$ . We will denote the output probability distribution per pixel as $P$ and locations $i , j$ as the location of the respective pixel in the output. $f ( x ) _ { i , j }$ denotes the ith row and $j ^ { ; }$ ’th column of the output. + +Confidence Estimation. The method proposed in DeVries & Taylor (2018) works by training a confidence branch added at the end of the neural network. We denote the network predictions as both $P$ and $\hat { c }$ whereby every pixel is assigned a confidence value. + +$$ +\begin{array} { c } { b \sim B ( 0 . 5 ) } \\ { c : = \hat { c } \cdot b + ( 1 - b ) } \\ { P : = P \cdot c + ( 1 - c ) y } \end{array} +$$ + +The confidence estimation denoted by $c$ is given “hints” during training to guide what it is learning. The $B$ is a beta distribution and acts as a regularizer similar to dropout so that the network $f$ does not exclusively rely on the true labels being present. The final loss is modified to include the extra term below: + +![](images/e831207f61f4f9d875c0b02663677b918f457e3a8c337ddc1935c65b9e381231.jpg) +Figure 7: A comparison of lighting consistency in the Fishyscapes anomaly segmentation benchmark and our new StreetHazards dataset. The arrows point in the manually estimated direction of light on parts of the scene. In Fishyscapes, inconsistent lighting allows forensics techniques to detect the anomaly (Johnson & Farid, 2005). Unlike cutand-paste anomalies, the anomalies in our StreetHazards dataset are naturally integrated into their environment with proper lighting and orientation, making them more difficult to detect. + +$$ +\begin{array} { l } { \displaystyle \mathcal { L } _ { \boldsymbol { p } } = \frac { 1 } { | P | } \sum _ { i } - \log ( p _ { i } ) y _ { i } } \\ { \displaystyle \mathcal { L } _ { \boldsymbol { c } } = \frac { 1 } { | P | } \sum _ { i } - \log ( \hat { c } _ { i } ) } \\ { \displaystyle \mathcal { L } = \mathcal { L } _ { \boldsymbol { p } } + \lambda \mathcal { L } _ { \boldsymbol { c } } } \end{array} +$$ + +The reasoning for $\mathcal { L } _ { c }$ is to encourage the network to output confident predictions. Finally $\lambda$ is initialized to 0.1 and is updated by a “budget” parameter which is set to the default of 0.3. The update equation: + +$$ +\left\{ \begin{array} { l l } { { \lambda } / { 0 . 9 9 } } & { \sum \hat { c } _ { i } \leq \mathrm { b u d g e t } } \\ { { \lambda } / { 1 . 0 1 } } & { \sum \hat { c } _ { i } > \mathrm { b u d g e t } } \end{array} \right. +$$ + +This adaptively adjusts the weighting between the two losses and experimentally the update is not sensitive to the budget parameter. + +Semantic Segmentation BDD Anomalies Dataset List. The BDD100K dataset contains 180 instances of the train class, 4296 instances of the motorcycle class, and 10229 instances of the bicycle class. + +StreetHazards 3D Models Dataset List. For semantic segmentation experiments, we choose to use the following classes 3D models from Model Bank Library to serve as out-of-distribution data: Meta-categories: Animals, Vehicles, Weapons, Appliances, Household items (furniture, and kitchen items), Electronics, Instruments, and miscellaneous. The specific animals used are kangaroos, whales, dolphins, cows, lions, frogs, bats, insects, mongooses, scorpions, fish, camels, flamingos, apes, horses, mice, spider, dinosaurs, elephants, moose, shrimps, bats, butterflies, turtles, hippopotamuses, dogs, cats, sheep, seahorse, snail and zebra. The specific vehicles used are military trucks, motorcycles, naval ships, pirate ships, submarines, sailing ships, trolleys, trains, airplanes, helicopters, jets, zeppelin, radar tower, construction vehicles (loaders, dump trucks, bulldozer), farming vehicles (harvester, gantry crane, tractor), fire truck, tank, combat vehicles, and trailers. The specific weapons used are guns, missiles, rocket launchers, and grenades. The appliances used are refrigerators, stoves, washing machines, and ovens. The household items used are cabinets, armoire, grandfather clocks, bathtubs, bureaus, night stand, table, bed, bookcase, office desk, glasses (drinking), throne chair, kitchen utensils (knives, forks, spoons), sofa, clothing iron, plates, sewing machine, and dressing mirror. The electronics used are computer monitor, computer mouse, hair dryer, The instruments category includes bassoon, clarinet, drums, guitar, violin, harp, and keyboard. The miscellaneous category includes rocket, space capsule, space shuttle, lunar module, glasses (wearable), weight machine, balance beam, bench press, bowling ball and pins, and pens. Several categories and instances were excluded from Model Bank Library due to their occurrence in the simulation environment such as playground equipment and various types of foliage and trees. The sizes of instances used in the dataset might not reflect the actual scale that would otherwise naturally occur. Similarly the location of instances in the dataset are not necessarily reflective of where they are likely to occur in nature. \ No newline at end of file diff --git a/parse/dev/vruwp11pWnO/vruwp11pWnO_content_list.json b/parse/dev/vruwp11pWnO/vruwp11pWnO_content_list.json new file mode 100644 index 0000000000000000000000000000000000000000..46f4c8857dff736c5b6d9577b341398bff872252 --- /dev/null +++ b/parse/dev/vruwp11pWnO/vruwp11pWnO_content_list.json @@ -0,0 +1,1615 @@ +[ + { + "type": "text", + "text": "IMPROVING AND ASSESSING ANOMALY DETECTORS FOR LARGE-SCALE SETTINGS ", + "text_level": 1, + "bbox": [ + 176, + 99, + 823, + 146 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Anonymous authors Paper under double-blind review ", + "bbox": [ + 183, + 170, + 398, + 198 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "ABSTRACT ", + "text_level": 1, + "bbox": [ + 454, + 234, + 544, + 251 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Detecting out-of-distribution examples is important for safety-critical machine learning applications such as detecting novel biological phenomena and self-driving cars. However, existing research mainly focuses on simple small-scale settings. To set the stage for more realistic out-of-distribution detection, we depart from small-scale settings and explore large-scale multiclass and multi-label settings with high-resolution images and thousands of classes. To make future work in real-world settings possible, we create new benchmarks for three large-scale settings. To test ImageNet multiclass anomaly detectors, we introduce a new dataset of anomalous species. We leverage ImageNet-21K to evaluate PASCAL VOC and COCO multilabel anomaly detectors. Third, we introduce a new benchmark for anomaly segmentation by introducing a segmentation benchmark with road anomalies. We conduct extensive experiments in these more realistic settings for out-of-distribution detection and find that a surprisingly simple detector based on the maximum logit outperforms prior methods in all the large-scale multi-class, multi-label, and segmentation tasks, establishing a simple new baseline for future work. ", + "bbox": [ + 233, + 266, + 766, + 474 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "1 INTRODUCTION ", + "text_level": 1, + "bbox": [ + 176, + 502, + 336, + 518 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Out-of-distribution (OOD) detection is a valuable tool for developing safe and reliable machine learning (ML) systems. Detecting anomalous inputs allows systems to initiate a conservative fallback policy or defer to human judgment. As an important component of ML Safety (Hendrycks et al., 2021), OOD detection is important for safety-critical applications such as self-driving cars and detecting novel microorganisms. Accordingly, research on out-of-distribution detection has a rich history spanning several decades (Schölkopf et al., 1999; Breunig et al., 2000; Emmott et al., 2015). Recent work leverages deep neural representations for out-of-distribution detection in complex domains, such as image data (Hendrycks & Gimpel, 2017; Lee et al., 2018a; Mohseni et al., 2020; Hendrycks et al., 2019b). However, these works still primarily use small-scale datasets with low-resolution images and few classes. As the community moves towards more realistic, large-scale settings, strong baselines and high-quality benchmarks are imperative for future progress. ", + "bbox": [ + 174, + 535, + 825, + 686 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Large-scale datasets such as ImageNet (Deng et al., 2009) and Places365 (Zhou et al., 2017) present unique challenges not seen in small-scale settings, such as a plethora of fine-grained object classes. We demonstrate that the maximum softmax probability (MSP) detector, a state-of-the-art method for small-scale problems, does not scale well to these challenging conditions. Through extensive experiments, we identify a detector based on the maximum logit (MaxLogit) that greatly outperforms the MSP and other strong baselines in large-scale multi-class anomaly segmentation. To facilitate further research in this setting, we also collect a new out-of-distribution test dataset suitable for models trained on highly diverse datasets. Shown in Figure 2, our Species dataset contains diverse, anomalous species that do not overlap ImageNet-21K which has approximately twenty two thousand classes. Species avoids data leakage and enables a stricter evaluation methodology for ImageNet-21K models. Using Species to conduct more controlled experiments without train-test overlap, we find that contrary to prior claims (Fort et al., 2021; Koner et al., 2021), Vision Transformers (Dosovitskiy et al., 2021a) pre-trained on ImageNet-21K are not substantially better at out-of-distribution detection. ", + "bbox": [ + 174, + 694, + 825, + 875 + ], + "page_idx": 0 + }, + { + "type": "text", + "text": "Moreover, in the common real-world case of multi-label data, the MSP detector cannot naturally be applied in the first place, as it requires softmax probabilities. To enable research into the multi-label setting for anomaly detection, we contribute a multi-label experimental setup and explore various methods on large-scale multi-label datasets. We find that the MaxLogit detector from our investigation into the large-scale multi-class setting generalizes well to multi-label data and again outperforms all other baselines. ", + "bbox": [ + 176, + 882, + 823, + 924 + ], + "page_idx": 0 + }, + { + "type": "image", + "img_path": "images/8595abd6b64885b25129dadf2053a7309f448d326b5e56de08cc3f4b6a1ec9f1.jpg", + "image_caption": [ + "Figure 1: We scale up out-of-distribution detection to large-scale multi-class datasets with thousands of classes, multi-label datasets with complex scenes, and anomaly segmentation in driving environments. We introduce new benchmarks for all three settings. In all of these settings, we find that an OOD detector based on the maximum logit outperforms previous methods, establishing a strong and versatile baseline for future work on large-scale OOD detection. The bottom-right shows a scene from our new anomaly segmentation benchmark and the predicted anomaly using a state-of-the-art detector. " + ], + "image_footnote": [], + "bbox": [ + 173, + 99, + 825, + 375 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "", + "bbox": [ + 176, + 491, + 823, + 534 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "In addition to focusing on small-scale datasets, most existing benchmarks for anomaly detection treat entire images as anomalies. In practice, an image could be anomalous in localized regions while being in-distribution elsewhere. Knowing which regions of an image are anomalous could allow for safer handling of unfamiliar objects in the case of self-driving cars. Creating a benchmark for this task is difficult, though, as simply cutting and pasting anomalous objects into images introduces various unnatural giveaway cues such as edge effects, mismatched orientation, and lighting, all of which trivialize the task of anomaly segmentation (Blum et al., 2019). ", + "bbox": [ + 174, + 541, + 825, + 638 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "To overcome these issues, we utilize a simulated driving environment to create the novel StreetHazards dataset for anomaly segmentation. Using the Unreal Engine and the open-source CARLA simulation environment (Dosovitskiy et al., 2017), we insert a diverse array of foreign objects into driving scenes and re-render the scenes with these novel objects. This enables integration of the foreign objects into their surrounding context with correct lighting and orientation, sidestepping giveaway cues. ", + "bbox": [ + 174, + 645, + 825, + 715 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "To complement the StreetHazards dataset, we convert the BDD100K semantic segmentation dataset (Yu et al., 2018) into an anomaly segmentation dataset, which we call BDD-Anomaly. By leveraging the large scale of BDD100K, we reserve infrequent object classes to be anomalies. We combine this dataset with StreetHazards to form the Combined Anomalous Object Segmentation (CAOS) benchmark. The CAOS benchmark improves over previous evaluations for anomaly segmentation in driving scenes by evaluating detectors on realistic and diverse anomalies. We evaluate several baselines on the CAOS benchmark and discuss problems with porting existing approaches from earlier formulations of out-of-distribution detection. ", + "bbox": [ + 174, + 722, + 825, + 833 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "Despite its simplicity, we find that the MaxLogit detector outperforms all baselines on Species, our multi-class benchmark, and CAOS. In each of these three settings, we discuss why MaxLogit provides superior performance, and we show that these gains are hidden if one looks at small-scale problems alone. The code for our experiments and the Species and CAOS datasets are available at [anonymized]. Our new baseline combined with Species and CAOS benchmarks pave the way for future research on large-scale out-of-distribution detection. ", + "bbox": [ + 174, + 840, + 825, + 924 + ], + "page_idx": 1 + }, + { + "type": "text", + "text": "Anomalous Species Dataset ", + "text_level": 1, + "bbox": [ + 408, + 102, + 591, + 116 + ], + "page_idx": 2 + }, + { + "type": "image", + "img_path": "images/654db4cbaa3ed751b6c6dce51acbc564fb80912d3bccb104063743fd809b6f2d.jpg", + "image_caption": [ + "Figure 2: The Species out-of-distribution dataset is designed for large-scale anomaly detectors pretrained on datasets as diverse as ImageNet-21K. When models are pretrained on ImageNet-21K, many previous OOD detection datasets may overlap with the pretraining set, resulting in erroneous evaluations. To rectify this, Species is comprised of hundreds of anomalous species that are disjoint from ImageNet-21K classes and enables the evaluation of cutting-edge models. " + ], + "image_footnote": [], + "bbox": [ + 174, + 118, + 823, + 357 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "2 RELATED WORK ", + "text_level": 1, + "bbox": [ + 176, + 478, + 344, + 494 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Multi-Class Out-of-Distribution Detection. A recent line of work leverages deep neural representations from multi-class classifiers to perform out-of-distribution (OOD) detection on highdimensional data, including images, text, and speech data. Hendrycks & Gimpel (2017) formulate the task and propose the simple baseline of using the maximum softmax probability of the classifier on an input to gauge whether the input is out-of-distribution. In particular, they formulate the task as distinguishing between examples from an in-distribution dataset and various OOD datasets. Importantly, entire images are treated as out-of-distribution. ", + "bbox": [ + 174, + 520, + 826, + 618 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Continuing this line of work, Lee et al. (2018a) propose to improve the neural representation of the classifier to better separate OOD examples. They use generative adversarial networks to produce neardistribution examples and induce uniform posteriors on these synthetic OOD examples. Hendrycks et al. (2019b) observe that outliers are often easy to obtain in large quantity from diverse, realistic datasets and demonstrate that OOD detectors trained on these outliers generalize to unseen classes of anomalies. Other work investigates improving the anomaly detectors themselves given a fixed classifier (DeVries & Taylor, 2018; Liang et al., 2018). However, as Hendrycks et al. (2019b) observe, many of these works tune hyperparameters on a particular type of anomaly that is also seen at test time, so their evaluation setting is more lenient. In this paper, all anomalies seen at test time come from entirely unseen categories and are not tuned on in any way. Hence, we do not compare to techniques such as ODIN (Liang et al., 2018). Additionally, in a point of departure from prior work, we focus primarily on large-scale images and datasets with many classes. ", + "bbox": [ + 174, + 625, + 825, + 791 + ], + "page_idx": 2 + }, + { + "type": "text", + "text": "Recent work has suggested that stronger representations from Vision Transformers pre-trained on ImageNet-21K can make out-of-distribution detection trivial (Fort et al., 2021; Koner et al., 2021). They evaluate models on detecting CIFAR-10 when fine-tuned on CIFAR-100 or vice versa, using models pretrained on ImageNet-21K. However, over 1,000 classes in ImageNet-21K overlap with CIFAR-10, so it is still unclear how Vision Transformers perform at detecting entirely unseen OOD categories. We create a new OOD test dataset of anomalous species to investigate how well Vision Transformers perform in controlled OOD detection settings without data leakage and overlap. We find that Vision Transformers pre-trained on ImageNet-21K are far from solving OOD detection in large-scale settings. ", + "bbox": [ + 174, + 799, + 825, + 924 + ], + "page_idx": 2 + }, + { + "type": "table", + "img_path": "images/bb7b3019a2b86e9009d5307a4d02ae9f36ffd04663404bc2ebffac2f51127748.jpg", + "table_caption": [], + "table_footnote": [], + "table_body": "
FPR95↓AUROC↑AUPR↑
DinMSPDeVriesMaxLogitMSPDeVries MaxLogitMSPDeVriesMaxLogit
ImageNet44.246.035.884.676.987.238.230.545.8
Places36552.685.836.676.031.185.88.22.019.2
", + "bbox": [ + 173, + 99, + 825, + 160 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "Table 1: Multi-class out-of-distribution detection results using the maximum softmax probability (MSP) baseline (Hendrycks & Gimpel, 2017), the confidence branch detector of DeVries & Taylor (2018), and our maximum logit baseline. All values are percentages and average across five out-ofdistribution test datasets. Full results on individual OOD test datasets are in the Appendix. ", + "bbox": [ + 173, + 171, + 825, + 227 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "Anomaly Segmentation. Several prior works explore segmenting anomalous image regions. One line of work uses the WildDash dataset (Zendel et al., 2018), which contains numerous annotated driving scenes in conditions such as snow, fog, and rain. The WildDash test set contains fifteen “negative images” from different domains for which the goal is to mark the entire image as out-ofdistribution. Thus, while the task is segmentation, the anomalies do not exist as objects within an otherwise in-distribution scene. This setting is similar to that explored by Hendrycks & Gimpel (2017), in which whole images from other datasets serve as out-of-distribution examples. ", + "bbox": [ + 173, + 270, + 825, + 367 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "To approach anomaly segmentation on WildDash, Krešo et al. (2018) train on multiple semantic segmentation domains and treat regions of images from the WildDash driving dataset as out-ofdistribution if they are segmented as regions from different domains, i.e. indoor classes. Bevandic´ et al. (2018) use ILSVRC 2012 images and train their network to segment the entirety of these images as out-of-distribution. ", + "bbox": [ + 174, + 375, + 825, + 444 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "In medical anomaly segmentation and product fault detection, anomalies are regions of otherwise in-distribution images. Baur et al. (2019) segment anomalous regions in brain MRIs using pixelwise reconstruction loss. Similarly, Haselmann et al. (2018) perform product fault detection using pixel-wise reconstruction loss and introduce an expansive dataset for segmentation of product faults. In these relatively simple domains, reconstruction-based approaches work well. In contrast to medical anomaly segmentation and fault detection, we consider complex images from street scenes. These images have high variability in scene layout and lighting, and hence are less amenable to reconstruction-based techniques. ", + "bbox": [ + 174, + 450, + 826, + 563 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "The two works closest to our own are the Lost and Found (Pinggera et al., 2016) and Fishyscapes (Blum et al., 2019) datasets. The Lost and Found dataset consists of real images in a driving environment with small road hazards. The images were collected to mirror the Cityscapes dataset (Cordts et al., 2016) but are only collected from one city and so have less diversity. The dataset contains 35 unique anomalous objects, and methods are allowed to train on many of these. For Lost and Found, only nine unique objects are truly unseen at test time. Crucially, this is a different evaluation setting from our own, where anomalous objects are not revealed at training time, so their dataset is not directly comparable. Nevertheless, the BDD-Anomaly dataset fills several gaps in Lost and Found. First, the images are more diverse, because they are sourced from a more recent and comprehensive semantic segmentation dataset. Second, the anomalies are not restricted to small, sparse road hazards. Concretely, anomalous regions in Lost and Found take up $0 . 1 1 \\%$ of the image on average, whereas anomalous regions in the BDD-Anomaly dataset are larger and fill $0 . 8 3 \\%$ of the image on average. Finally, although the BDD-Anomaly dataset treats three categories as anomalous, compared to Lost and Found it has far more unique anomalous objects. ", + "bbox": [ + 173, + 569, + 825, + 763 + ], + "page_idx": 3 + }, + { + "type": "text", + "text": "The Fishyscapes benchmark for anomaly segmentation consists of cut-and-paste anomalies from out-of-distribution domains. This is problematic, because the anomalies stand out as clearly unnatural in context. For instance, the orientation of anomalous objects is unnatural, and the lighting of the cut-and-paste patch differs from the lighting in the original image, providing an unnatural cue to anomaly detectors that would not exist for real anomalies. Figure 7 shows an example of these inconsistencies. Techniques for detecting image manipulation (Zhou et al., 2018; Johnson & Farid, 2005) are competent at detecting artificial image elements of this kind. Our StreetHazards dataset overcomes these issues by leveraging a simulated driving environment to naturally insert anomalous 3D models into a scene rather than overlaying 2D images. These anomalies are integrated into the scene with proper lighting and orientation, mimicking real-world anomalies and making them significantly more difficult to detect. ", + "bbox": [ + 174, + 771, + 825, + 924 + ], + "page_idx": 3 + }, + { + "type": "image", + "img_path": "images/c3e38c3fa532816e71e5ccfdee9b5bd80f517c0c2915dee5bd921821eb1372be.jpg", + "image_caption": [ + "Figure 3: Small-scale datasets such as CIFAR-10 have relatively disjoint classes, but larger-scale datasets including ImageNet-1K have several classes with high visual similarity to other classes. This implies that large-scale classifiers disperse probability mass among several classes. If the prediction confidence is used for out-of-distribution detection, then images which have similarities to other classes will often wrongly be deemed out-of-distribution due to low and dispersed confidence. This motivates our MaxLogit out-of-distribution detector. " + ], + "image_footnote": [], + "bbox": [ + 169, + 98, + 825, + 339 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "3 MULTI-CLASS PREDICTION FOR OOD DETECTION ", + "text_level": 1, + "bbox": [ + 174, + 443, + 627, + 458 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Problem with existing baselines. Existing baselines for anomaly detection can work well in small-scale settings. However, in more realistic settings image classification networks are often tasked with distinguishing hundreds or thousands of classes, possibly with subtle differences. This is problematic for the maximum softmax probability (MSP) baseline (Hendrycks & Gimpel, 2017), which uses the negative maximum softmax probability as the anomaly score, or $\\begin{array} { r } { \\dot { { \\bf \\varphi } } - \\operatorname* { m a x } _ { k } \\exp f ( x ) _ { k } / \\sum _ { i } \\exp f ( \\bar { x _ { i } } ) _ { i } = - \\operatorname* { m a x } _ { k } \\hat { p } ( y = k \\mid x ) } \\end{array}$ , where $f ( x )$ is the unnormalized logits of classifier $f$ on input $x$ . Classifiers tend to have higher confidence on in-distribution examples than out-of-distribution examples, enabling OOD detection. Assuming single-model evaluation and no access to other anomalies or test-time adaptation, the MSP attains state-of-the-art anomaly detection performance in small-scale settings. However, we show that the MSP is problematic for realistic in-distribution datasets with many classes, such as ImageNet and Places365 (Zhou et al., 2017). Probability mass can be dispersed among visually similar classes, as shown in Figure 3. Consequently, a classifier may produce a low confidence prediction for an in-distribution image, not because the image is unfamiliar, but because the object’s exact class is difficult to determine. To circumvent this problem, we propose using the negative of the maximum unnormalized logit for an anomaly score $- \\operatorname* { m a x } _ { k } f ( x ) _ { k }$ , which we call MaxLogit. Since the logits are unnormalized, they are not affected by the number of classes and can serve as a better baseline for large-scale out-of-distribution detection. ", + "bbox": [ + 174, + 479, + 826, + 728 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "The Species Out-Of-Distribution Dataset. To enable controlled experiments and high-quality evaluations of anomaly detectors in large-scale settings, we create the Species dataset, a new outof-distribution test dataset that has no overlapping classes with ImageNet-21K. The Species dataset is comprised of images scraped from the iNaturalist website and contains hundreds of anomalous species grouped into seven high-level categories: Plants, Microorganisms, Amphibians, Protozoa, Fungi, Arachnids, and Insects. Example images from the Species dataset are in Figure 2. ", + "bbox": [ + 174, + 736, + 825, + 819 + ], + "page_idx": 4 + }, + { + "type": "text", + "text": "Setup. To evaluate the MSP baseline out-of-distribution detector and the MaxLogit detector, we use ImageNet-21K as the in-distribution dataset $\\mathcal { D } _ { \\mathrm { i n } }$ . To obtain representations for anomaly detection, we use models trained on ImageNet-21K-P, a cleaned version of ImageNet-21K with a train/val split (Ridnik et al., 2021a). We evaluate a TResNet-M, ViT-B-16, and Mixer-B-16 (Ridnik et al., 2021b; Dosovitskiy et al., 2021b; Tolstikhin et al., 2021), and the validation split is used for obtaining in-distribution scores. For out-of-distribution test datasets $\\mathcal { D } _ { \\mathrm { o u t } }$ , we use categories from the Species dataset, all of which are unseen during training. Results for these experiments are in Table 2. We also use ImageNet-1K and Places365 as in-distribution datasets $\\mathcal { D } _ { \\mathrm { i n } }$ , for which we use pretrained ResNet-50 models and use several out-of-distribution test datasets $\\mathcal { D } _ { \\mathrm { o u t } }$ . Full results with ImageNet and Places365 as in-distribution are in the Appendix. ", + "bbox": [ + 174, + 825, + 825, + 924 + ], + "page_idx": 4 + }, + { + "type": "table", + "img_path": "images/35cd44e5505cbefb1abbe5d8609e00a686ece733beac6c36b934c08b7457e7a4.jpg", + "table_caption": [ + "Table 2: Results on Species. Models and the processed version of ImageNet-21K (ImageNet-21K-P) are from Ridnik et al. (2021a). All values are percent AUROC. Species enables evaluating anomaly detectors trained on ImageNet-21K and evades class overlap issues present in prior work. Using Species to conduct more controlled experiments without class overlap issues, we find that contrary to recent claims (Fort et al., 2021), simply scaling up Vision Transformers does not make OOD detection trivial. " + ], + "table_footnote": [], + "table_body": "
ResNetViTMLPMixer
DinDtest outMSPMaxLogitMSPMaxLogitMSPMaxLogit
Plants80.387.878.284.880.385.0
Microorganisms77.483.471.182.474.486.0
Amphibians41.848.641.948.844.451.7
Protozoa70.780.469.380.968.077.7
Fungi66.477.464.776.164.176.9
hetArachnids46.956.746.656.848.958.8
Insects47.656.448.054.648.653.8
Mean61.670.160.069.261.270.0
", + "bbox": [ + 171, + 87, + 825, + 231 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 333, + 825, + 376 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Metrics. To evaluate out-of-distribution detectors in large-scale settings, we use three standard metrics of detection performance: area under the ROC curve (AUROC), false positive rate at $9 5 \\%$ recall (FPR95), and area under the precision-recall curve (AUPR). The AUROC and AUPR are important metrics, because they give a holistic measure of performance when the cutoff for detecting anomalies is not a priori obvious or when we want to represent the performance of a detection method across several different cutoffs. ", + "bbox": [ + 174, + 382, + 825, + 465 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "The AUROC can be thought of as the probability that an anomalous example is given a higher score than an ordinary example. Thus, a higher score is better, and an uninformative detector has a AUROC of $50 \\%$ . AUPR provides a metric more attuned to class imbalances, which is relevant in anomaly and failure detection, when the number of anomalies or failures may be relatively small. Last, the FPR95 metric consists of measuring the false positive rate at $9 5 \\%$ . Since these measures are correlated, we occasionally solely present the AUROC for brevity and to preserve space. ", + "bbox": [ + 174, + 472, + 825, + 556 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Results. Results on Species are shown in Table 2. Results with ImageNet-1K and Places365 as in-distribution datasets are in Table 1. We find that the proposed MaxLogit method outperforms the maximum softmax probability baseline on all out-of-distribution test datasets $\\mathcal { D } _ { \\mathrm { o u t } }$ . This holds true for all three models trained on ImageNet-21K. The MSP baseline is not much better than random and is has similar performance for all three model classes. This suggests that contrary to recent claims, (Fort et al., 2021) simply scaling up Vision Transformers does not make OOD detection trivial. ", + "bbox": [ + 174, + 563, + 825, + 647 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "4 MULTI-LABEL PREDICTION FOR OOD DETECTION ", + "text_level": 1, + "bbox": [ + 176, + 667, + 629, + 684 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Current work on out-of-distribution detection primarily considers multi-class or unsupervised settings. Yet as classifiers become more useful in realistic settings, the multi-label formulation becomes increasingly natural. To investigate out-of-distribution detection in multi-label settings, we provide a baseline and evaluation setup. ", + "bbox": [ + 174, + 700, + 825, + 756 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Setup. For multi-label classification we use PASCAL VOC (Everingham et al., 2009) and MSCOCO (Lin et al., 2014) as in-distribution data. To evaluate anomaly detectors for these in-distribution datasets, we use 20 out-of-distribution classes from ImageNet-21K. These classes have no overlap with ImageNet-1K, PASCAL VOC, or MS-COCO. The 20 classes are chosen not to overlap with ImageNet-1K since the multi-label classifiers models are pre-trained on ImageNet-1K. We list the class WordNet IDs in the Appendix. ", + "bbox": [ + 174, + 762, + 825, + 847 + ], + "page_idx": 5 + }, + { + "type": "text", + "text": "Methods. For our experiments, we use a ResNet-101 backbone architecture pre-trained on ImageNet-1K. We replace the final layer with 2 fully connected layers and apply the logistic sigmoid function for multi-label prediction. During training we freeze the batch normalization parameters due to an insufficient number of images for proper mean and variance estimation. We train each model for 50 epochs using the Adam optimizer (Kingma & Ba, 2014) with hyperparameter values $1 0 ^ { - 4 }$ and ", + "bbox": [ + 174, + 854, + 823, + 924 + ], + "page_idx": 5 + }, + { + "type": "table", + "img_path": "images/9bfc4b95d20326f3d05b9f023f7272b1c6d411057166e6c65d7ec8da4d54289b.jpg", + "table_caption": [ + "Table 3: Multi-label out-of-distribution detection comparison of the Isolation Forest (iForest), Local Outlier Factor (LOF), Dropout, logit average, maximum softmax probability, and maximum logit anomaly detectors on PASCAL VOC and MS-COCO. The same network architecture is used for all three detectors. All results shown are percentages. " + ], + "table_footnote": [], + "table_body": "
iForestLOFDropoutLogitAvgMSPMaxLogit
PASCAL VOCFPR9598.684.097.298.282.335.6
AUROC46.368.449.247.974.290.9
AUPR→↑↑37.158.445.341.365.581.2
COCOFPR95√95.678.493.394.581.840.4
AUROC个41.470.258.055.570.790.3
AUPR个63.782.076.374.082.994.0
", + "bbox": [ + 173, + 87, + 826, + 191 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "$1 0 ^ { - 5 }$ for $\\beta _ { 1 }$ and $\\beta _ { 2 }$ respectively. For data augmentation we use standard resizing, random crops, and random flips to obtain images of size $2 5 6 \\times 2 5 6 \\times 3$ . As a result of this training procedure, the mAP of the ResNet-101 on PASCAL VOC is $8 9 . 1 1 \\%$ and $7 2 . 0 \\%$ for MS-COCO. ", + "bbox": [ + 174, + 272, + 825, + 314 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "As there has been little work on out-of-distribution detection in multilabel settings, we include comparisons to classic anomaly detectors for general settings. Isolation Forest, denoted by iForest, works by randomly partitioning the space into half spaces to form a decision tree. The score is determined by how close a point is to the root of the tree. The local outlier factor (LOF) (Breunig et al., 2000) computes a local density ratio between every element and its neighbors. We set the number of neighbors as 20. iForest and LOF are both computed on features from the penultimate layer of the networks. MSP denotes a natural extension of the maximum softmax probability detector in the multi-label setting, obtained by taking the sigmoid of each output score $f ( { \\boldsymbol { x } } ) _ { i }$ and computing $- \\operatorname* { m a x } _ { i } \\sigma ( f ( x ) _ { i } )$ . Alternatively, one can average the logit values, denoted by LogitAvg. These serve as our baseline detectors for multi-label OOD detection. We compare these baselines to the MaxLogit detector that we introduce in Section 3. As in the multi-class case, the MaxLogit anomaly score for multi-label classification is $- \\operatorname* { m a x } _ { i } f ( x ) _ { i }$ . ", + "bbox": [ + 174, + 321, + 825, + 488 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "Results. Results are shown in Table 3. We find that MaxLogit obtains the highest performance in all cases. MaxLogit bears similarity to the MSP baseline (Hendrycks & Gimpel, 2017) but is naturally applicable to multi-label problems. These results establish the MaxLogit as an effective and natural baseline for large-scale multi-label problems. Further, the evaluation setup enables future work in out-of-distribution detection with multi-label datasets. ", + "bbox": [ + 174, + 494, + 825, + 564 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "5 THE CAOS BENCHMARK ", + "text_level": 1, + "bbox": [ + 176, + 585, + 416, + 602 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "The Combined Anomalous Object Segmentation (CAOS) benchmark is comprised of two complementary datasets for evaluating anomaly segmentation systems on diverse, realistic anomalies. First is the StreetHazards dataset, which leverages simulation to provide a large variety of anomalous objects realistically inserted into driving scenes. Second is the BDD-Anomaly dataset, which consists of real images taken from the BDD100K dataset (Yu et al., 2018). StreetHazards contains a highly diverse array of anomalies; BDD-Anomaly contains anomalies in real-world images. Together, these datasets allow researchers to judge techniques on their ability to segment diverse anomalies as well as anomalies in real images. All images have $7 2 0 \\times 1 2 8 0$ resolution. ", + "bbox": [ + 173, + 617, + 825, + 728 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "The StreetHazards Dataset. StreetHazards is an anomaly segmentation dataset that leverages simulation to provide diverse, realistically-inserted anomalous objects. To create the StreetHazards dataset, we use the Unreal Engine along with the CARLA simulation environment (Dosovitskiy et al., 2017). From several months of development and testing including customization of the Unreal Engine and CARLA, we can insert foreign entities into a scene while having them be properly integrated. Unlike previous work, this avoids the issues of inconsistent chromatic aberration, inconsistent lighting, edge effects, and other simple cues that an object is anomalous. Additionally, using a simulated environment allows us to dynamically insert diverse anomalous objects in any location and have them render properly with changes to lighting and weather including time of day, cloudy skies, and rain. ", + "bbox": [ + 174, + 734, + 825, + 861 + ], + "page_idx": 6 + }, + { + "type": "text", + "text": "We use 3 towns from CARLA for training, from which we collect RGB images and their respective semantic segmentation maps to serve as training data for semantic segmentation models. We generate a validation set from the fourth town. Finally, we reserve the fifth and sixth town as our test set. We insert anomalies taken from the Digimation Model Bank Library and semantic ShapeNet (ShapeNetSem) (Savva et al., 2015) into the test set in order to evaluate methods for out-of-distribution detection. In total, we use 250 unique anomaly models of diverse types. There are 12 classes used for training: background, road, street lines, traffic signs, sidewalk, pedestrian, vehicle, building, wall, pole, fence, and vegetation. The thirteenth class is the anomaly class that is only used at test time. We collect 5,125 image and semantic segmentation ground truth pairs for training, 1,031 pairs without anomalies for validation, and 1,500 test pairs with anomalies. ", + "bbox": [ + 174, + 867, + 823, + 924 + ], + "page_idx": 6 + }, + { + "type": "image", + "img_path": "images/9f6a01add5da434aa775b1cb401e714b45aeff0bd523f7abf8d24e9f6ac738fc.jpg", + "image_caption": [ + "Figure 4: A sample of anomalous scenes from the CAOS benchmark with model predictions and anomaly scores. The anomaly scores are thresholded to the top $10 \\%$ of values for visualization. GT is ground truth, the autoencoder model is based on the spatial autoencoder used in Baur et al. (2019), MSP is the maximum softmax probability baseline (Hendrycks & Gimpel, 2017), and MaxLogit is the method we propose as a new baseline for large-scale settings. Compared to baselines, the MaxLogit detector places lower scores on in-distribution image regions, including object outlines, while also doing a better job of highlighting anomalous objects. " + ], + "image_footnote": [], + "bbox": [ + 173, + 102, + 823, + 449 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 571, + 825, + 656 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "The BDD-Anomaly Dataset. BDD-Anomaly is an anomaly segmentation dataset with real images in diverse conditions. We source BDD-Anomaly from BDD100K (Yu et al., 2018), a large-scale semantic segmentation dataset with diverse driving conditions. The original data consists in 7,000 images for training and 1,000 for validation. There are 18 original classes. We choose motorcycle, train, and bicycle as the anomalous object classes and remove all images with these objects from the training and validation sets. This yields 6,280 training pairs, 910 validation pairs without anomalies, and 810 testing pairs with anomalous objects. ", + "bbox": [ + 174, + 662, + 825, + 761 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "5.1 EXPERIMENTS ", + "text_level": 1, + "bbox": [ + 174, + 779, + 315, + 792 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "Evaluation. In anomaly segmentation experiments, each pixel is treated as a prediction, resulting in many predictions to evaluate. To fit these in memory, we compute the metrics on each image and average over the images to obtain final values. ", + "bbox": [ + 174, + 805, + 823, + 847 + ], + "page_idx": 7 + }, + { + "type": "text", + "text": "Methods. Our first baseline is pixel-wise Maximum Softmax Probability (MSP). Introduced by Hendrycks & Gimpel (2017) for multi-class out-of-distribution detection, we directly port this baseline to anomaly segmentation. Alternatively, the background class might serve as an anomaly detector, because it contains everything not in the other classes. To test this hypothesis, “Background” uses the posterior probability of the background class as the anomaly score. The Dropout method leverages MC Dropout (Gal & Ghahramani, 2016) to obtain an epistemic uncertainty estimate. Following Kendall et al. (2015), we compute the pixel-wise posterior variance over multiple dropout masks and average across all classes, which serves as the anomaly score. We also experiment with an autoencoder baseline similar to Baur et al. (2019); Haselmann et al. (2018) where pixel-wise reconstruction loss is used as the anomaly score. This method is called AE. The “Branch” method is a direct port of the confidence branch detector from DeVries & Taylor (2018) to pixel-wise prediction. Finally, we use the MaxLogit method described in earlier sections independently on each pixel. ", + "bbox": [ + 174, + 853, + 825, + 924 + ], + "page_idx": 7 + }, + { + "type": "table", + "img_path": "images/a39f779b992a3a5f4420e180e8d83132c4aaf76e85751c0c198cb34b68e5580a.jpg", + "table_caption": [ + "Table 4: Results on the CAOS benchmark. AUPR is low across the board due to the large class imbalance, but all methods perform substantially better than chance. MaxLogit obtains the best performance. All results are percentages. " + ], + "table_footnote": [], + "table_body": "
MSPBranchBackgroundDropoutAEMaxLogit
StreetHazardsFPR9533.768.469.079.491.726.5
AUROC87.765.758.669.966.189.3
↓↑ AUPR 个6.61.54.57.52.210.6
BDD-Anomaly AUROCFPR9524.525.640.116.674.114.0
87.785.669.790.864.092.6
AUPR3.73.91.14.30.75.4
", + "bbox": [ + 173, + 80, + 826, + 185 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 257, + 825, + 354 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "For all of the baselines except the autoencoder, we train a PSPNet (Zhao et al., 2017) decoder with a ResNet-101 encoder (He et al., 2015) for 20 epochs. We train both the encoder and decoder using SGD with momentum of 0.9, a learning rate of $2 \\times 1 0 ^ { - 2 }$ , and learning rate decay of $1 0 ^ { - 4 }$ . For AE, we use a 4-layer U-Net (Ronneberger et al., 2015) with a spatial latent code as in Baur et al. (2019). The U-Net also uses batch norm and is trained for 10 epochs. Results are in Table 4. ", + "bbox": [ + 173, + 362, + 825, + 431 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "Results and Analysis. MaxLogit outperforms all other methods across the board by a substantial margin. The intuitive baseline of using the posterior for the background class to detect anomalies performs poorly, which suggests that the background class may not align with rare visual features. Even though reconstruction-based scores succeed in product fault segmentation, we find that the AE method performs poorly on the CAOS benchmark, which may be due to the more complex domain. AUPR for all methods is low, indicating that the large class imbalance presents a serious challenge. However, the substantial improvements with the MaxLogit method suggest that progress on this task is possible and there is much room for improvement. A comparison with other datasets is in Figure 5 (Pinggera et al., 2016; Blum et al., 2019; Jung et al., 2021). ", + "bbox": [ + 174, + 438, + 517, + 577 + ], + "page_idx": 8 + }, + { + "type": "table", + "img_path": "images/120ec03391e0f691aa82ae0e7be5996340d88c64bca62f5030a5c6f3232c23af.jpg", + "table_caption": [ + "Figure 5: Auxiliary analysis of the MSP and the MaxLogit AUROCs using prior less comprehensive anomaly segmentation datasets. " + ], + "table_footnote": [], + "table_body": "
MethodMSPMaxLogit
FS Lost and Found87.0%92.0%
Road Anomaly73.8%78.0%
", + "bbox": [ + 529, + 443, + 823, + 502 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 577, + 825, + 632 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "In Figure 4, we see that both MaxLogit and MSP have many false positives, as they assign high anomaly scores to semantic boundaries, a problem also observed in the recent works of (Blum et al., 2019; Angus, 2019). However, the problem is less severe with MaxLogit. A potential explanation for this is that even when the prediction confidence dips at semantic boundaries, the maximum logit can remain the same in a ‘hand-off’ procedure between the classes. Thus, MaxLogit provides a natural mechanism to combat semantic boundary artifacts that could be further explored in future work. ", + "bbox": [ + 173, + 638, + 825, + 723 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "6 CONCLUSION ", + "text_level": 1, + "bbox": [ + 174, + 750, + 318, + 765 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "We scaled out-of-distribution detection to settings with thousands of classes and high-resolution images. We identified an issue faced by existing baselines when scaling to these settings and proposed the maximum logit detector as a natural solution. We introduced the Species dataset to enable more controlled experiments without class overlap and also investigated using multi-label classifiers for OOD detection, establishing an experimental setup for this previously unexplored setting. Finally, we introduced the CAOS benchmark for anomaly segmentation, consisting of diverse, naturally-integrated anomalous objects in driving scenes. Baseline methods on the CAOS benchmark substantially improve on random guessing but are still lacking, indicating potential for future work. Interestingly, the MaxLogit detector also provides consistent and significant gains in the multi-label and anomaly segmentation settings, thereby establishing it as a new baseline in place of the maximum softmax probability baseline on large-scale OOD detection problems. In all, we we hope that our contributions will enable further research on out-of-distribution detection for real-world safety-critical environments. ", + "bbox": [ + 174, + 785, + 825, + 924 + ], + "page_idx": 8 + }, + { + "type": "text", + "text": "", + "bbox": [ + 171, + 103, + 823, + 132 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "REFERENCES ", + "text_level": 1, + "bbox": [ + 174, + 154, + 287, + 170 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Matt Angus. Towards pixel-level ood detection for semantic segmentation, 2019. ", + "bbox": [ + 173, + 178, + 702, + 193 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Christoph Baur, Benedikt Wiestler, Shadi Albarqouni, and Nassir Navab. Deep autoencoding models for unsupervised anomaly segmentation in brain mr images. Lecture Notes in Computer Science, pp. 161–169, 2019. ISSN 1611-3349. doi: 10.1007/978-3-030-11723-8_16. ", + "bbox": [ + 176, + 203, + 823, + 246 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Petra Bevandic, Ivan Krešo, Marin Orši ´ c, and Siniša Šegvi ´ c. Discriminative out-of-distribution ´ detection for semantic segmentation, 2018. ", + "bbox": [ + 173, + 257, + 821, + 286 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Hermann Blum, Paul-Edouard Sarlin, Juan Nieto, Roland Siegwart, and Cesar Cadena. The fishyscapes benchmark: Measuring blind spots in semantic segmentation, 2019. ", + "bbox": [ + 171, + 297, + 823, + 327 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Markus M Breunig, Hans-Peter Kriegel, Raymond T $\\mathrm { N g }$ , and Jörg Sander. Lof: identifying densitybased local outliers. In ACM sigmod record, volume 29, pp. 93–104. ACM, 2000. ", + "bbox": [ + 171, + 337, + 823, + 366 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "M. Cimpoi, S. Maji, I. Kokkinos, S. Mohamed, , and A. Vedaldi. Describing textures in the wild. In Computer Vision and Pattern Recognition, 2014. ", + "bbox": [ + 169, + 376, + 825, + 405 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. The cityscapes dataset for semantic urban scene understanding. In Proc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016. ", + "bbox": [ + 173, + 415, + 825, + 472 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei. ImageNet: A Large-Scale Hierarchical Image Database. In CVPR09, 2009. ", + "bbox": [ + 173, + 482, + 823, + 511 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Terrance DeVries and Graham W Taylor. Learning confidence for out-of-distribution detection in neural networks. arXiv preprint arXiv:1802.04865, 2018. ", + "bbox": [ + 171, + 522, + 825, + 551 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Alexey Dosovitskiy, German Ros, Felipe Codevilla, Antonio Lopez, and Vladlen Koltun. CARLA: An open urban driving simulator. In Proceedings of the 1st Annual Conference on Robot Learning, pp. 1–16, 2017. ", + "bbox": [ + 174, + 560, + 826, + 604 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. ArXiv, abs/2010.11929, 2021a. ", + "bbox": [ + 174, + 614, + 826, + 671 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. In International Conference on Learning Representations, 2021b. URL https://openreview. net/forum?id ${ . } = { }$ YicbFdNTTy. ", + "bbox": [ + 173, + 681, + 828, + 752 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Andrew Emmott, Shubhomoy Das, Thomas Dietterich, Alan Fern, and Weng-Keen Wong. A meta-analysis of the anomaly detection problem, 2015. ", + "bbox": [ + 171, + 762, + 825, + 792 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Mark Everingham, Luc Van Gool, Christopher K. I. Williams, John M. Winn, and Andrew Zisserman. The pascal visual object classes (voc) challenge. International Journal of Computer Vision, 88: 303–338, 2009. ", + "bbox": [ + 173, + 803, + 825, + 844 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Stanislav Fort, Jie Ren, and Balaji Lakshminarayanan. Exploring the limits of out-of-distribution detection. arXiv preprint arXiv:2106.03004, 2021. ", + "bbox": [ + 171, + 856, + 823, + 885 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Yarin Gal and Zoubin Ghahramani. Dropout as a bayesian approximation: Representing model uncertainty in deep learning. International Conference on Machine Learning, 2016. ", + "bbox": [ + 174, + 895, + 823, + 924 + ], + "page_idx": 9 + }, + { + "type": "text", + "text": "Matthias Haselmann, Dieter P Gruber, and Paul Tabatabai. Anomaly detection using deep learning based image completion. In 2018 17th IEEE International Conference on Machine Learning and Applications (ICMLA), pp. 1237–1242. IEEE, 2018. ", + "bbox": [ + 176, + 103, + 823, + 146 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. CVPR, 2015. ", + "bbox": [ + 173, + 155, + 823, + 184 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution examples in neural networks. ICLR, 2017. ", + "bbox": [ + 173, + 193, + 823, + 222 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier exposure. Proceedings of the International Conference on Learning Representations, 2019a. ", + "bbox": [ + 174, + 231, + 823, + 261 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier exposure. In International Conference on Learning Representations, 2019b. ", + "bbox": [ + 176, + 268, + 821, + 299 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Dan Hendrycks, Mantas Mazeika, Saurav Kadavath, and Dawn Song. Using self-supervised learning can improve model robustness and uncertainty. In NeurIPS, 2019c. ", + "bbox": [ + 174, + 306, + 821, + 337 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Dan Hendrycks, Nicholas Carlini, John Schulman, and Jacob Steinhardt. Unsolved problems in ml safety. ArXiv, abs/2109.13916, 2021. ", + "bbox": [ + 173, + 344, + 821, + 375 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Micah K Johnson and Hany Farid. Exposing digital forgeries by detecting inconsistencies in lighting. In Proceedings of the 7th workshop on Multimedia and security, pp. 1–10, 2005. ", + "bbox": [ + 173, + 383, + 820, + 412 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Sanghun Jung, Jungsoo Lee, Daehoon Gwak, Sungha Choi, and Jaegul Choo. Standardized max logits: A simple yet effective approach for identifying unexpected road obstacles in urban-scene segmentation. ArXiv, abs/2107.11264, 2021. ", + "bbox": [ + 174, + 421, + 825, + 464 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Alex Kendall, Vijay Badrinarayanan, and Roberto Cipolla. Bayesian segnet: Model uncertainty in deep convolutional encoder-decoder architectures for scene understanding. ArXiv, abs/1511.02680, 2015. ", + "bbox": [ + 174, + 473, + 826, + 515 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. ICLR, 2014. ", + "bbox": [ + 173, + 525, + 794, + 540 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Rajat Koner, Poulami Sinhamahapatra, Karsten Roscher, Stephan Günnemann, and Volker Tresp. Oodformer: Out-of-distribution detection transformer. ArXiv, 2021. ", + "bbox": [ + 174, + 549, + 823, + 578 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Ivan Krešo, Marin Oršic, Petra Bevandi ´ c, and Siniša Šegvi ´ c. Robust semantic segmentation with ´ ladder-densenet models, 2018. ", + "bbox": [ + 173, + 587, + 823, + 616 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Kimin Lee, Honglak Lee, Kibok Lee, and Jinwoo Shin. Training confidence-calibrated classifiers for detecting out-of-distribution samples. International Conference on Learning Representations, 2018a. ", + "bbox": [ + 174, + 625, + 825, + 667 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. A simple unified framework for detecting out-of-distribution samples and adversarial attacks. ArXiv, abs/1807.03888, 2018b. ", + "bbox": [ + 171, + 676, + 823, + 707 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Shiyu Liang, Yixuan Li, and R. Srikant. Enhancing the reliability of out-of-distribution image detection in neural networks. International Conference on Learning Representations, 2018. ", + "bbox": [ + 169, + 714, + 823, + 744 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Tsung-Yi Lin, Michael Maire, Serge Belongie, Lubomir Bourdev, Ross Girshick, James Hays, Pietro Perona, Deva Ramanan, C. Lawrence Zitnick, and Piotr Dollar. Microsoft COCO: Common objects in context. ECCV, 2014. ", + "bbox": [ + 176, + 752, + 825, + 796 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Sina Mohseni, Mandar Pitale, Jbs Yadawa, and Zhangyang Wang. Self-supervised learning for generalizable out-of-distribution detection. In AAAI, 2020. ", + "bbox": [ + 171, + 804, + 825, + 834 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Peter Pinggera, Sebastian Ramos, Stefan K. Gehrig, Uwe Franke, Carsten Rother, and Rudolf Mester. Lost and found: Detecting small road hazards for self-driving vehicles. 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 1099–1106, 2016. ", + "bbox": [ + 176, + 843, + 823, + 887 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "T. Ridnik, Emanuel Ben-Baruch, Asaf Noy, and Lihi Zelnik-Manor. Imagenet-21k pretraining for the masses. ArXiv, abs/2104.10972, 2021a. ", + "bbox": [ + 174, + 895, + 821, + 924 + ], + "page_idx": 10 + }, + { + "type": "text", + "text": "Tal Ridnik, Hussam Lawen, Asaf Noy, Emanuel Ben Baruch, Gilad Sharir, and Itamar Friedman. Tresnet: High performance gpu-dedicated architecture. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 1400–1409, 2021b. ", + "bbox": [ + 179, + 103, + 823, + 146 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015, pp. 234–241, 2015. ISSN 1611-3349. doi: 10.1007/978-3-319-24574-4_28. ", + "bbox": [ + 178, + 155, + 823, + 196 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Manolis Savva, Angel X. Chang, and Pat Hanrahan. Semantically-Enriched 3D Models for Commonsense Knowledge. CVPR 2015 Workshop on Functionality, Physics, Intentionality and Causality, 2015. ", + "bbox": [ + 174, + 207, + 825, + 248 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Bernhard Schölkopf, Robert Williamson, Alex Smola, John Shawe-Taylor, and John Platt. Support vector method for novelty detection. In Proceedings of the 12th International Conference on Neural Information Processing Systems, NIPS’99, pp. 582–588, Cambridge, MA, USA, 1999. MIT Press. ", + "bbox": [ + 173, + 257, + 826, + 314 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Ilya Tolstikhin, Neil Houlsby, Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Thomas Unterthiner, Jessica Yung, Daniel Keysers, Jakob Uszkoreit, Mario Lucic, et al. Mlp-mixer: An all-mlp architecture for vision. arXiv preprint arXiv:2105.01601, 2021. ", + "bbox": [ + 173, + 323, + 826, + 366 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Fisher Yu, Yinda Zhang, Shuran Song, Ari Seff, and Jianxiong Xiao. LSUN: construction of a large-scale image dataset using deep learning with humans in the loop. CoRR, 2015. ", + "bbox": [ + 173, + 375, + 825, + 404 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Fisher Yu, Wenqi Xian, Yingying Chen, Fangchen Liu, Mike Liao, Vashisht Madhavan, and Trevor Darrell. BDD100K: A diverse driving video database with scalable annotation tooling. CoRR, abs/1805.04687, 2018. ", + "bbox": [ + 174, + 412, + 825, + 455 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Oliver Zendel, Katrin Honauer, Markus Murschitz, Daniel Steininger, and Gustavo Fernandez Dominguez. Wilddash-creating hazard-aware benchmarks. In Proceedings of the European Conference on Computer Vision (ECCV), pp. 402–416, 2018. ", + "bbox": [ + 174, + 464, + 825, + 507 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Hengshuang Zhao, Jianping Shi, Xiaojuan Qi, Xiaogang Wang, and Jiaya Jia. Pyramid scene parsing network. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 6230–6239, 2017. ", + "bbox": [ + 171, + 515, + 823, + 558 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Bolei Zhou, Agata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba. Places: A 10 million image database for scene recognition. PAMI, 2017. ", + "bbox": [ + 169, + 566, + 825, + 597 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "Peng Zhou, Xintong Han, Vlad I Morariu, and Larry S Davis. Learning rich features for image manipulation detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1053–1061, 2018. ", + "bbox": [ + 174, + 604, + 825, + 648 + ], + "page_idx": 11 + }, + { + "type": "text", + "text": "A APPENDIX ", + "text_level": 1, + "bbox": [ + 176, + 103, + 299, + 117 + ], + "page_idx": 12 + }, + { + "type": "table", + "img_path": "images/ac83fc93b0bb49458edab1aad6f2275b1266913fbb8c2964a80965bd28e142e4.jpg", + "table_caption": [ + "Table 5: B is for the maximum softmax probability baseline, M is for maximum logit, D is for the method in DeVries & Taylor (2018), and K is our own KL method described below. Both M and K are ours. Results are on ImageNet and Places365. All values are percentages and are rounded so that 99.95 rounds to 100. " + ], + "table_footnote": [], + "table_body": "
DestFPR95↓AUROC ↑AUPR↑
DinBM DKBMDKBMDK
Gaussian2054100100 9798329252521985579
1eeeeeeRademacher2144158998 9893706254
Blobs26327288079 379917793
Textures685674598087 7685361648
LSUN6663596075 777679221938
Places3656459637279 83797927322446
Mean41.335.84636.185.287.276.988.7 3745.830.559.7
PPssse536Gaussian10671129396 35931624216
Rademacher20109118993 1010011 515.91.688
Blobs59688277298 159341231
Textures86728774657943 7911112
Places6988899291616452 65636
Mean5336.685.840.97685.831.185.819.2230.5
", + "bbox": [ + 173, + 136, + 826, + 351 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "B FULL MULTICLASS OOD DETECTION RESULTS ", + "text_level": 1, + "bbox": [ + 174, + 435, + 604, + 453 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "Datasets. To evaluate the MSP baseline out-of-distribution detector and the MaxLogit detector, we use the ImageNet-1K object recognition dataset and Places365 scene recognition dataset as in-distribution datasets $\\mathcal { D } _ { \\mathrm { i n } }$ . We use several out-of-distribution test datasets $\\mathcal { D } _ { \\mathrm { o u t } }$ , all of which are unseen during training. The first out-of-distribution dataset is Gaussian noise, where each example’s pixels are i.i.d. sampled from $\\mathcal { N } ( 0 , 0 . 5 )$ and clipped to be contained within $[ - 1 , 1 ]$ . Another type of test-time noise is Rademacher noise, in which each pixel is i.i.d. sampled from 2 · Bernoulli $( 0 . 5 ) - 1$ , i.e. each pixel is 1 or $- 1$ with equal probability. Blob examples are more structured than noise; they are algorithmically generated blob images. Meanwhile, Textures is a dataset consisting in images of describable textures (Cimpoi et al., 2014). When evaluating the ImageNet-1K detector, we use LSUN images, a scene recognition dataset (Yu et al., 2015). Our final $\\mathcal { D } _ { \\mathrm { o u t } }$ is Places69, a scene classification dataset that does not share classes with Places365. In all, we evaluate against out-of-distribution examples spanning synthetic and realistic images. ", + "bbox": [ + 173, + 468, + 825, + 635 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "KL Matching Method. To verify our intuitions that led us to develop the MaxLogit detector, we developed a less convenient but similarly powerful technique applicable for the multiclass setting. Recall that some classes tend to be predicted with low confidence and others high confidence. The shape of predicted posterior distributions is often class dependent. ", + "bbox": [ + 174, + 650, + 825, + 705 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "We capture the typical shape of each class’s posterior distribution and form posterior distribution templates for each class. During test time, the network’s softmax posterior distribution is compared to these templates and an anomaly score is generated. More concretely, we compute $k$ different distributions $d _ { k }$ , one for each class. We write $d _ { k } = \\mathbb { E } _ { x ^ { \\prime } \\sim \\mathcal { X } _ { \\mathrm { v a l } } } [ p ( y | x ^ { \\prime } ) ]$ where $k = \\mathrm { a r g m a x } _ { k } p ( y =$ $k \\mid x ^ { \\prime } )$ . Then for a new test input $x$ , we calculate the anomaly score $\\mathrm { { \\ddot { m i n } } } _ { k } ~ \\mathrm { K L } [ p ( y \\mid x ) \\mid \\mid d _ { k } ]$ rather than the MSP baseline $- \\operatorname* { m a x } _ { k } p ( y = k \\mid x ) .$ . Note that we utilize the validation dataset, but our KL matching method does not require the validation dataset’s labels. That said, our KL matching method is less convenient than our MaxLogit technique, and the two perform similarly. Since this technique requires more data than MaxLogit, we opt to simply use the MaxLogit in the main paper. ", + "bbox": [ + 173, + 713, + 825, + 839 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "Results. Observe that the proposed MaxLogit method outperforms the maximum softmax probability baseline for all three metrics on both ImageNet and Places365. These results were computed using a ResNet-50 trained on either ImageNet-1K or Places365. In the case of Places365, the AUROC improvement is over $10 \\%$ . We note that the utility of the maximum logit could not be appreciated as easily in previous work’s small-scale settings. For example, using the small-scale CIFAR-10 setup of ", + "bbox": [ + 174, + 854, + 825, + 924 + ], + "page_idx": 12 + }, + { + "type": "text", + "text": "Hendrycks et al. Hendrycks et al. (2019a), the MSP attains an average AUROC of $9 0 . 0 8 \\%$ while the maximum logit attains $9 0 . 2 2 \\%$ , a minor $0 . 1 4 \\%$ difference. However, in a large-scale setting, the difference can be over $10 \\%$ on individual $\\mathcal { D } _ { \\mathrm { o u t } }$ datasets. We are not claiming that utilizing the maximum logit is a mathematically innovative formulation, only that it serves as a consistently powerful baseline for large-scale settings that went unappreciated in small-scale settings. In consequence, we suggest using the maximum logit as a new baseline for large-scale multi-class out-of-distribution detection. ", + "bbox": [ + 174, + 103, + 825, + 200 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "Overview of Other Detection Methods. There are other techniques in out-of-distribution detection which require other assumptions such as more training data. For instance, Hendrycks et al. (2019a); Mohseni et al. (2020) use additional training data labeled as out-of-distribution, and the MaxLogit technique can be naturally extended should such data be available. Hendrycks et al. (2019c) use rotation prediction and self-supervised learning, but we found that scaling this to the ImageNet multiclass setting did not produce strong results. The MSP baseline trained with auxiliary rotation prediction has an AUROC of $5 9 . 1 \\%$ , and with MaxLogit it attains a $7 3 . 6 \\%$ AUROC, over a $10 \\%$ absolute improvement with MaxLogit. Nonetheless this technique did not straightforwardly scale, as the network is better without auxiliary rotation prediction. Likewise, Lee et al. (2018b) propose to use Mahalanobis distances, but in scaling this to 1000 classes, we consistently encountered NaN errors due to high condition numbers. This shows the importance of ensuring that out-of-distribution techniques can scale. ", + "bbox": [ + 174, + 219, + 825, + 386 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "ODIN Liang et al. (2018) assumes that, for each OOD example source, we can tune hyperparameters for detection. For this reason we do not evaluate with ODIN in the rest of the paper. However, for thoroughness, we evaluate it here. ODIN uses temperature scaling and adds an epsilon perturbation to the input in order to separate the softmax posteriors between in- and out-of-distribution images; we set these hyperparameters following DeVries & Taylor (2018). Then, MaxLogit combined with ODIN results in an FPR95 of 33.6, an AUROC of 88.8 and an AUPR of 51.3 on ImageNet. On Places365, the FPR95 is 35.3, the AUROC is 86.5, and the AUPR is 24.2. Consequently, techniques built with different assumptions can integrate well with MaxLogit. We do not train ImageNet-21K models from scratch with these methods due to limited compute. ", + "bbox": [ + 173, + 393, + 825, + 518 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "C MULTI-LABEL OUT-OF-DISTRIBUTION DATASET LIST ", + "text_level": 1, + "bbox": [ + 176, + 541, + 661, + 558 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "For multi-label classification experiments, we choose the following classes from ImageNet-21K to serve as out-of-distribution data: dolphin (n02069412), deer (n02431122), bat (n02139199), rhino (n02392434), raccoon (n02508213), octopus (n01970164), giant clam (n01959492), leech (n01937909), Venus flytrap (n12782915), cherry tree (n12641413), Japanese cherry blossoms (n12649317), red wood (n12285512), sunflower (n11978713), croissant (n07691650), stick cinnamon (n07814390), cotton (n12176953), rice (n12126084), sugar cane (n12132956), bamboo (n12147226), and tumeric (n12356395). These classes were hand-chosen so that they are distinct from VOC and COCO classes. ", + "bbox": [ + 174, + 575, + 537, + 768 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "D OOD SEGMENTATION ", + "text_level": 1, + "bbox": [ + 174, + 792, + 393, + 809 + ], + "page_idx": 13 + }, + { + "type": "image", + "img_path": "images/e830f93c8a5dee8cff942f66fbb348ee36c59204cff44e668c2183b5e4494531.jpg", + "image_caption": [ + "Figure 6: ROC curve with VOC as $( \\mathcal { D } _ { \\mathrm { i n } } )$ and non-overlapping ImageNet classes as $( \\mathcal { D } _ { \\mathrm { o u t } } ^ { \\mathrm { t e s t } } )$ . Curves correspond to an uninformative “Random” detector, Local Outlier Factor, and the MaxLogit detector. " + ], + "image_footnote": [], + "bbox": [ + 547, + 594, + 818, + 811 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "We cover methods used in the paper in more depth and the modifications necessary to make the methods work with OOD detection in semantic segmentation. We use $f$ to denote the function typically a neural network, $x$ is the input image, and $y _ { i , j }$ is the prediction for pixel $i , j$ . We will denote the output probability distribution per pixel as $P$ and locations $i , j$ as the location of the respective pixel in the output. $f ( x ) _ { i , j }$ denotes the ith row and $j ^ { ; }$ ’th column of the output. ", + "bbox": [ + 174, + 827, + 535, + 924 + ], + "page_idx": 13 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 103, + 535, + 132 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "Confidence Estimation. The method proposed in DeVries & Taylor (2018) works by training a confidence branch added at the end of the neural network. We denote the network predictions as both $P$ and $\\hat { c }$ whereby every pixel is assigned a confidence value. ", + "bbox": [ + 174, + 147, + 485, + 231 + ], + "page_idx": 14 + }, + { + "type": "equation", + "img_path": "images/712f2da0fc9e3316c69cdaeb4970ca6182bdc9b12ec93cb069bfd5f2fb7b72b3.jpg", + "text": "$$\n\\begin{array} { c } { b \\sim B ( 0 . 5 ) } \\\\ { c : = \\hat { c } \\cdot b + ( 1 - b ) } \\\\ { P : = P \\cdot c + ( 1 - c ) y } \\end{array}\n$$", + "text_format": "latex", + "bbox": [ + 253, + 250, + 405, + 308 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "The confidence estimation denoted by $c$ is given “hints” during training to guide what it is learning. The $B$ is a beta distribution and acts as a regularizer similar to dropout so that the network $f$ does not exclusively rely on the true labels being present. The final loss is modified to include the extra term below: ", + "bbox": [ + 173, + 316, + 485, + 414 + ], + "page_idx": 14 + }, + { + "type": "image", + "img_path": "images/e831207f61f4f9d875c0b02663677b918f457e3a8c337ddc1935c65b9e381231.jpg", + "image_caption": [ + "Figure 7: A comparison of lighting consistency in the Fishyscapes anomaly segmentation benchmark and our new StreetHazards dataset. The arrows point in the manually estimated direction of light on parts of the scene. In Fishyscapes, inconsistent lighting allows forensics techniques to detect the anomaly (Johnson & Farid, 2005). Unlike cutand-paste anomalies, the anomalies in our StreetHazards dataset are naturally integrated into their environment with proper lighting and orientation, making them more difficult to detect. " + ], + "image_footnote": [], + "bbox": [ + 504, + 171, + 813, + 407 + ], + "page_idx": 14 + }, + { + "type": "equation", + "img_path": "images/188937f87a242a5c887d29d311cbd567431198a575e25323201226d6d6633277.jpg", + "text": "$$\n\\begin{array} { l } { \\displaystyle \\mathcal { L } _ { \\boldsymbol { p } } = \\frac { 1 } { | P | } \\sum _ { i } - \\log ( p _ { i } ) y _ { i } } \\\\ { \\displaystyle \\mathcal { L } _ { \\boldsymbol { c } } = \\frac { 1 } { | P | } \\sum _ { i } - \\log ( \\hat { c } _ { i } ) } \\\\ { \\displaystyle \\mathcal { L } = \\mathcal { L } _ { \\boldsymbol { p } } + \\lambda \\mathcal { L } _ { \\boldsymbol { c } } } \\end{array}\n$$", + "text_format": "latex", + "bbox": [ + 243, + 429, + 415, + 527 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "The reasoning for $\\mathcal { L } _ { c }$ is to encourage the network to output confident predictions. Finally $\\lambda$ is initialized to 0.1 and is updated by a “budget” parameter which is set to the default of 0.3. The update equation: ", + "bbox": [ + 173, + 535, + 485, + 606 + ], + "page_idx": 14 + }, + { + "type": "equation", + "img_path": "images/c4eed88b17dbc93f4e19087b97ce3896c0eb2ba44f70df785114b20ca53d77a9.jpg", + "text": "$$\n\\left\\{ \\begin{array} { l l } { { \\lambda } / { 0 . 9 9 } } & { \\sum \\hat { c } _ { i } \\leq \\mathrm { b u d g e t } } \\\\ { { \\lambda } / { 1 . 0 1 } } & { \\sum \\hat { c } _ { i } > \\mathrm { b u d g e t } } \\end{array} \\right.\n$$", + "text_format": "latex", + "bbox": [ + 408, + 623, + 588, + 660 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "This adaptively adjusts the weighting between the two losses and experimentally the update is not sensitive to the budget parameter. ", + "bbox": [ + 174, + 670, + 823, + 698 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "Semantic Segmentation BDD Anomalies Dataset List. The BDD100K dataset contains 180 instances of the train class, 4296 instances of the motorcycle class, and 10229 instances of the bicycle class. ", + "bbox": [ + 174, + 712, + 825, + 756 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "StreetHazards 3D Models Dataset List. For semantic segmentation experiments, we choose to use the following classes 3D models from Model Bank Library to serve as out-of-distribution data: Meta-categories: Animals, Vehicles, Weapons, Appliances, Household items (furniture, and kitchen items), Electronics, Instruments, and miscellaneous. The specific animals used are kangaroos, whales, dolphins, cows, lions, frogs, bats, insects, mongooses, scorpions, fish, camels, flamingos, apes, horses, mice, spider, dinosaurs, elephants, moose, shrimps, bats, butterflies, turtles, hippopotamuses, dogs, cats, sheep, seahorse, snail and zebra. The specific vehicles used are military trucks, motorcycles, naval ships, pirate ships, submarines, sailing ships, trolleys, trains, airplanes, helicopters, jets, zeppelin, radar tower, construction vehicles (loaders, dump trucks, bulldozer), farming vehicles (harvester, gantry crane, tractor), fire truck, tank, combat vehicles, and trailers. The specific weapons used are guns, missiles, rocket launchers, and grenades. The appliances used are refrigerators, stoves, washing machines, and ovens. The household items used are cabinets, armoire, grandfather clocks, bathtubs, bureaus, night stand, table, bed, bookcase, office desk, glasses (drinking), throne chair, kitchen utensils (knives, forks, spoons), sofa, clothing iron, plates, sewing machine, and dressing mirror. The electronics used are computer monitor, computer mouse, hair dryer, The instruments category includes bassoon, clarinet, drums, guitar, violin, harp, and keyboard. The miscellaneous category includes rocket, space capsule, space shuttle, lunar module, glasses (wearable), weight machine, balance beam, bench press, bowling ball and pins, and pens. Several categories and instances were excluded from Model Bank Library due to their occurrence in the simulation environment such as playground equipment and various types of foliage and trees. The sizes of instances used in the dataset might not reflect the actual scale that would otherwise naturally occur. Similarly the location of instances in the dataset are not necessarily reflective of where they are likely to occur in nature. ", + "bbox": [ + 173, + 770, + 826, + 924 + ], + "page_idx": 14 + }, + { + "type": "text", + "text": "", + "bbox": [ + 174, + 103, + 825, + 256 + ], + "page_idx": 15 + } +] \ No newline at end of file diff --git a/parse/dev/vruwp11pWnO/vruwp11pWnO_middle.json b/parse/dev/vruwp11pWnO/vruwp11pWnO_middle.json new file mode 100644 index 0000000000000000000000000000000000000000..5b892d080679cce266ed67a3ce0760d96d7dc984 --- /dev/null +++ b/parse/dev/vruwp11pWnO/vruwp11pWnO_middle.json @@ -0,0 +1,39058 @@ +{ + "pdf_info": [ + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 79, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 77, + 505, + 98 + ], + "spans": [ + { + "bbox": [ + 105, + 77, + 505, + 98 + ], + "score": 1.0, + "content": "IMPROVING AND ASSESSING ANOMALY DETECTORS", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 99, + 330, + 118 + ], + "spans": [ + { + "bbox": [ + 105, + 99, + 330, + 118 + ], + "score": 1.0, + "content": "FOR LARGE-SCALE SETTINGS", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 112, + 135, + 244, + 157 + ], + "lines": [ + { + "bbox": [ + 113, + 135, + 201, + 147 + ], + "spans": [ + { + "bbox": [ + 113, + 135, + 201, + 147 + ], + "score": 1.0, + "content": "Anonymous authors", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 111, + 146, + 245, + 158 + ], + "spans": [ + { + "bbox": [ + 111, + 146, + 245, + 158 + ], + "score": 1.0, + "content": "Paper under double-blind review", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5 + }, + { + "type": "title", + "bbox": [ + 278, + 186, + 333, + 199 + ], + "lines": [ + { + "bbox": [ + 277, + 186, + 335, + 200 + ], + "spans": [ + { + "bbox": [ + 277, + 186, + 335, + 200 + ], + "score": 1.0, + "content": "ABSTRACT", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 143, + 211, + 469, + 376 + ], + "lines": [ + { + "bbox": [ + 141, + 212, + 469, + 224 + ], + "spans": [ + { + "bbox": [ + 141, + 212, + 469, + 224 + ], + "score": 1.0, + "content": "Detecting out-of-distribution examples is important for safety-critical machine", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 141, + 222, + 470, + 236 + ], + "spans": [ + { + "bbox": [ + 141, + 222, + 470, + 236 + ], + "score": 1.0, + "content": "learning applications such as detecting novel biological phenomena and self-driving", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 141, + 234, + 470, + 246 + ], + "spans": [ + { + "bbox": [ + 141, + 234, + 470, + 246 + ], + "score": 1.0, + "content": "cars. However, existing research mainly focuses on simple small-scale settings.", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 142, + 245, + 469, + 256 + ], + "spans": [ + { + "bbox": [ + 142, + 245, + 469, + 256 + ], + "score": 1.0, + "content": "To set the stage for more realistic out-of-distribution detection, we depart from", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 141, + 256, + 469, + 267 + ], + "spans": [ + { + "bbox": [ + 141, + 256, + 469, + 267 + ], + "score": 1.0, + "content": "small-scale settings and explore large-scale multiclass and multi-label settings with", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 141, + 267, + 470, + 279 + ], + "spans": [ + { + "bbox": [ + 141, + 267, + 470, + 279 + ], + "score": 1.0, + "content": "high-resolution images and thousands of classes. To make future work in real-world", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 279, + 470, + 290 + ], + "spans": [ + { + "bbox": [ + 141, + 279, + 470, + 290 + ], + "score": 1.0, + "content": "settings possible, we create new benchmarks for three large-scale settings. To test", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 289, + 470, + 300 + ], + "spans": [ + { + "bbox": [ + 141, + 289, + 470, + 300 + ], + "score": 1.0, + "content": "ImageNet multiclass anomaly detectors, we introduce a new dataset of anoma-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 300, + 469, + 311 + ], + "spans": [ + { + "bbox": [ + 141, + 300, + 469, + 311 + ], + "score": 1.0, + "content": "lous species. We leverage ImageNet-21K to evaluate PASCAL VOC and COCO", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 310, + 470, + 323 + ], + "spans": [ + { + "bbox": [ + 141, + 310, + 470, + 323 + ], + "score": 1.0, + "content": "multilabel anomaly detectors. Third, we introduce a new benchmark for anomaly", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 322, + 469, + 333 + ], + "spans": [ + { + "bbox": [ + 141, + 322, + 469, + 333 + ], + "score": 1.0, + "content": "segmentation by introducing a segmentation benchmark with road anomalies. We", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 142, + 333, + 469, + 344 + ], + "spans": [ + { + "bbox": [ + 142, + 333, + 469, + 344 + ], + "score": 1.0, + "content": "conduct extensive experiments in these more realistic settings for out-of-distribution", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 343, + 469, + 356 + ], + "spans": [ + { + "bbox": [ + 141, + 343, + 469, + 356 + ], + "score": 1.0, + "content": "detection and find that a surprisingly simple detector based on the maximum logit", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 353, + 471, + 368 + ], + "spans": [ + { + "bbox": [ + 141, + 353, + 471, + 368 + ], + "score": 1.0, + "content": "outperforms prior methods in all the large-scale multi-class, multi-label, and seg-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 366, + 415, + 377 + ], + "spans": [ + { + "bbox": [ + 141, + 366, + 415, + 377 + ], + "score": 1.0, + "content": "mentation tasks, establishing a simple new baseline for future work.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 12 + }, + { + "type": "title", + "bbox": [ + 108, + 398, + 206, + 411 + ], + "lines": [ + { + "bbox": [ + 105, + 397, + 208, + 414 + ], + "spans": [ + { + "bbox": [ + 105, + 397, + 208, + 414 + ], + "score": 1.0, + "content": "1 INTRODUCTION", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 424, + 505, + 544 + ], + "lines": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "score": 1.0, + "content": "Out-of-distribution (OOD) detection is a valuable tool for developing safe and reliable machine", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 434, + 506, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 447 + ], + "score": 1.0, + "content": "learning (ML) systems. Detecting anomalous inputs allows systems to initiate a conservative fallback", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 445, + 506, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 445, + 506, + 459 + ], + "score": 1.0, + "content": "policy or defer to human judgment. As an important component of ML Safety (Hendrycks et al., 2021),", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 455, + 506, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 506, + 470 + ], + "score": 1.0, + "content": "OOD detection is important for safety-critical applications such as self-driving cars and detecting", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 466, + 505, + 481 + ], + "spans": [ + { + "bbox": [ + 105, + 466, + 505, + 481 + ], + "score": 1.0, + "content": "novel microorganisms. Accordingly, research on out-of-distribution detection has a rich history", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 479, + 506, + 491 + ], + "spans": [ + { + "bbox": [ + 105, + 479, + 506, + 491 + ], + "score": 1.0, + "content": "spanning several decades (Schölkopf et al., 1999; Breunig et al., 2000; Emmott et al., 2015). Recent", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 489, + 506, + 502 + ], + "spans": [ + { + "bbox": [ + 105, + 489, + 506, + 502 + ], + "score": 1.0, + "content": "work leverages deep neural representations for out-of-distribution detection in complex domains,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 500, + 505, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 500, + 505, + 513 + ], + "score": 1.0, + "content": "such as image data (Hendrycks & Gimpel, 2017; Lee et al., 2018a; Mohseni et al., 2020; Hendrycks", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 512, + 504, + 523 + ], + "spans": [ + { + "bbox": [ + 105, + 512, + 504, + 523 + ], + "score": 1.0, + "content": "et al., 2019b). However, these works still primarily use small-scale datasets with low-resolution", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 522, + 505, + 536 + ], + "spans": [ + { + "bbox": [ + 105, + 522, + 505, + 536 + ], + "score": 1.0, + "content": "images and few classes. As the community moves towards more realistic, large-scale settings, strong", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 533, + 403, + 546 + ], + "spans": [ + { + "bbox": [ + 105, + 533, + 403, + 546 + ], + "score": 1.0, + "content": "baselines and high-quality benchmarks are imperative for future progress.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 26 + }, + { + "type": "text", + "bbox": [ + 107, + 550, + 505, + 693 + ], + "lines": [ + { + "bbox": [ + 106, + 551, + 505, + 562 + ], + "spans": [ + { + "bbox": [ + 106, + 551, + 505, + 562 + ], + "score": 1.0, + "content": "Large-scale datasets such as ImageNet (Deng et al., 2009) and Places365 (Zhou et al., 2017) present", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 561, + 506, + 574 + ], + "spans": [ + { + "bbox": [ + 106, + 561, + 506, + 574 + ], + "score": 1.0, + "content": "unique challenges not seen in small-scale settings, such as a plethora of fine-grained object classes.", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 572, + 506, + 585 + ], + "spans": [ + { + "bbox": [ + 106, + 572, + 506, + 585 + ], + "score": 1.0, + "content": "We demonstrate that the maximum softmax probability (MSP) detector, a state-of-the-art method", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 582, + 505, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 582, + 505, + 596 + ], + "score": 1.0, + "content": "for small-scale problems, does not scale well to these challenging conditions. Through extensive", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 594, + 505, + 607 + ], + "spans": [ + { + "bbox": [ + 105, + 594, + 505, + 607 + ], + "score": 1.0, + "content": "experiments, we identify a detector based on the maximum logit (MaxLogit) that greatly outperforms", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 605, + 505, + 617 + ], + "spans": [ + { + "bbox": [ + 106, + 605, + 505, + 617 + ], + "score": 1.0, + "content": "the MSP and other strong baselines in large-scale multi-class anomaly segmentation. To facilitate", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 615, + 506, + 629 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 506, + 629 + ], + "score": 1.0, + "content": "further research in this setting, we also collect a new out-of-distribution test dataset suitable for", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 626, + 506, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 506, + 640 + ], + "score": 1.0, + "content": "models trained on highly diverse datasets. Shown in Figure 2, our Species dataset contains diverse,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 638, + 506, + 651 + ], + "spans": [ + { + "bbox": [ + 105, + 638, + 506, + 651 + ], + "score": 1.0, + "content": "anomalous species that do not overlap ImageNet-21K which has approximately twenty two thousand", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 648, + 505, + 661 + ], + "spans": [ + { + "bbox": [ + 106, + 648, + 505, + 661 + ], + "score": 1.0, + "content": "classes. Species avoids data leakage and enables a stricter evaluation methodology for ImageNet-21K", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 660, + 506, + 673 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 506, + 673 + ], + "score": 1.0, + "content": "models. Using Species to conduct more controlled experiments without train-test overlap, we find that", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 670, + 507, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 507, + 684 + ], + "score": 1.0, + "content": "contrary to prior claims (Fort et al., 2021; Koner et al., 2021), Vision Transformers (Dosovitskiy et al.,", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 682, + 492, + 694 + ], + "spans": [ + { + "bbox": [ + 106, + 682, + 492, + 694 + ], + "score": 1.0, + "content": "2021a) pre-trained on ImageNet-21K are not substantially better at out-of-distribution detection.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 699, + 505, + 710 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 710 + ], + "score": 1.0, + "content": "Moreover, in the common real-world case of multi-label data, the MSP detector cannot naturally be", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "applied in the first place, as it requires softmax probabilities. To enable research into the multi-label", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "score": 1.0, + "content": "setting for anomaly detection, we contribute a multi-label experimental setup and explore various", + "type": "text" + } + ], + "index": 47 + } + ], + "index": 46 + } + ], + "page_idx": 0, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 308, + 761 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 308, + 761 + ], + "score": 1.0, + "content": "1", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 79, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 77, + 505, + 98 + ], + "spans": [ + { + "bbox": [ + 105, + 77, + 505, + 98 + ], + "score": 1.0, + "content": "IMPROVING AND ASSESSING ANOMALY DETECTORS", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 99, + 330, + 118 + ], + "spans": [ + { + "bbox": [ + 105, + 99, + 330, + 118 + ], + "score": 1.0, + "content": "FOR LARGE-SCALE SETTINGS", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 112, + 135, + 244, + 157 + ], + "lines": [ + { + "bbox": [ + 113, + 135, + 201, + 147 + ], + "spans": [ + { + "bbox": [ + 113, + 135, + 201, + 147 + ], + "score": 1.0, + "content": "Anonymous authors", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 111, + 146, + 245, + 158 + ], + "spans": [ + { + "bbox": [ + 111, + 146, + 245, + 158 + ], + "score": 1.0, + "content": "Paper under double-blind review", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 2.5, + "bbox_fs": [ + 111, + 135, + 245, + 158 + ] + }, + { + "type": "title", + "bbox": [ + 278, + 186, + 333, + 199 + ], + "lines": [ + { + "bbox": [ + 277, + 186, + 335, + 200 + ], + "spans": [ + { + "bbox": [ + 277, + 186, + 335, + 200 + ], + "score": 1.0, + "content": "ABSTRACT", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 143, + 211, + 469, + 376 + ], + "lines": [ + { + "bbox": [ + 141, + 212, + 469, + 224 + ], + "spans": [ + { + "bbox": [ + 141, + 212, + 469, + 224 + ], + "score": 1.0, + "content": "Detecting out-of-distribution examples is important for safety-critical machine", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 141, + 222, + 470, + 236 + ], + "spans": [ + { + "bbox": [ + 141, + 222, + 470, + 236 + ], + "score": 1.0, + "content": "learning applications such as detecting novel biological phenomena and self-driving", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 141, + 234, + 470, + 246 + ], + "spans": [ + { + "bbox": [ + 141, + 234, + 470, + 246 + ], + "score": 1.0, + "content": "cars. However, existing research mainly focuses on simple small-scale settings.", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 142, + 245, + 469, + 256 + ], + "spans": [ + { + "bbox": [ + 142, + 245, + 469, + 256 + ], + "score": 1.0, + "content": "To set the stage for more realistic out-of-distribution detection, we depart from", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 141, + 256, + 469, + 267 + ], + "spans": [ + { + "bbox": [ + 141, + 256, + 469, + 267 + ], + "score": 1.0, + "content": "small-scale settings and explore large-scale multiclass and multi-label settings with", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 141, + 267, + 470, + 279 + ], + "spans": [ + { + "bbox": [ + 141, + 267, + 470, + 279 + ], + "score": 1.0, + "content": "high-resolution images and thousands of classes. To make future work in real-world", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 141, + 279, + 470, + 290 + ], + "spans": [ + { + "bbox": [ + 141, + 279, + 470, + 290 + ], + "score": 1.0, + "content": "settings possible, we create new benchmarks for three large-scale settings. To test", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 141, + 289, + 470, + 300 + ], + "spans": [ + { + "bbox": [ + 141, + 289, + 470, + 300 + ], + "score": 1.0, + "content": "ImageNet multiclass anomaly detectors, we introduce a new dataset of anoma-", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 141, + 300, + 469, + 311 + ], + "spans": [ + { + "bbox": [ + 141, + 300, + 469, + 311 + ], + "score": 1.0, + "content": "lous species. We leverage ImageNet-21K to evaluate PASCAL VOC and COCO", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 141, + 310, + 470, + 323 + ], + "spans": [ + { + "bbox": [ + 141, + 310, + 470, + 323 + ], + "score": 1.0, + "content": "multilabel anomaly detectors. Third, we introduce a new benchmark for anomaly", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 141, + 322, + 469, + 333 + ], + "spans": [ + { + "bbox": [ + 141, + 322, + 469, + 333 + ], + "score": 1.0, + "content": "segmentation by introducing a segmentation benchmark with road anomalies. We", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 142, + 333, + 469, + 344 + ], + "spans": [ + { + "bbox": [ + 142, + 333, + 469, + 344 + ], + "score": 1.0, + "content": "conduct extensive experiments in these more realistic settings for out-of-distribution", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 141, + 343, + 469, + 356 + ], + "spans": [ + { + "bbox": [ + 141, + 343, + 469, + 356 + ], + "score": 1.0, + "content": "detection and find that a surprisingly simple detector based on the maximum logit", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 141, + 353, + 471, + 368 + ], + "spans": [ + { + "bbox": [ + 141, + 353, + 471, + 368 + ], + "score": 1.0, + "content": "outperforms prior methods in all the large-scale multi-class, multi-label, and seg-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 141, + 366, + 415, + 377 + ], + "spans": [ + { + "bbox": [ + 141, + 366, + 415, + 377 + ], + "score": 1.0, + "content": "mentation tasks, establishing a simple new baseline for future work.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 12, + "bbox_fs": [ + 141, + 212, + 471, + 377 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 398, + 206, + 411 + ], + "lines": [ + { + "bbox": [ + 105, + 397, + 208, + 414 + ], + "spans": [ + { + "bbox": [ + 105, + 397, + 208, + 414 + ], + "score": 1.0, + "content": "1 INTRODUCTION", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 20 + }, + { + "type": "text", + "bbox": [ + 107, + 424, + 505, + 544 + ], + "lines": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "spans": [ + { + "bbox": [ + 106, + 424, + 505, + 436 + ], + "score": 1.0, + "content": "Out-of-distribution (OOD) detection is a valuable tool for developing safe and reliable machine", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 434, + 506, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 447 + ], + "score": 1.0, + "content": "learning (ML) systems. Detecting anomalous inputs allows systems to initiate a conservative fallback", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 445, + 506, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 445, + 506, + 459 + ], + "score": 1.0, + "content": "policy or defer to human judgment. As an important component of ML Safety (Hendrycks et al., 2021),", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 455, + 506, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 455, + 506, + 470 + ], + "score": 1.0, + "content": "OOD detection is important for safety-critical applications such as self-driving cars and detecting", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 466, + 505, + 481 + ], + "spans": [ + { + "bbox": [ + 105, + 466, + 505, + 481 + ], + "score": 1.0, + "content": "novel microorganisms. Accordingly, research on out-of-distribution detection has a rich history", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 479, + 506, + 491 + ], + "spans": [ + { + "bbox": [ + 105, + 479, + 506, + 491 + ], + "score": 1.0, + "content": "spanning several decades (Schölkopf et al., 1999; Breunig et al., 2000; Emmott et al., 2015). Recent", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 489, + 506, + 502 + ], + "spans": [ + { + "bbox": [ + 105, + 489, + 506, + 502 + ], + "score": 1.0, + "content": "work leverages deep neural representations for out-of-distribution detection in complex domains,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 500, + 505, + 513 + ], + "spans": [ + { + "bbox": [ + 105, + 500, + 505, + 513 + ], + "score": 1.0, + "content": "such as image data (Hendrycks & Gimpel, 2017; Lee et al., 2018a; Mohseni et al., 2020; Hendrycks", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 512, + 504, + 523 + ], + "spans": [ + { + "bbox": [ + 105, + 512, + 504, + 523 + ], + "score": 1.0, + "content": "et al., 2019b). However, these works still primarily use small-scale datasets with low-resolution", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 522, + 505, + 536 + ], + "spans": [ + { + "bbox": [ + 105, + 522, + 505, + 536 + ], + "score": 1.0, + "content": "images and few classes. As the community moves towards more realistic, large-scale settings, strong", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 533, + 403, + 546 + ], + "spans": [ + { + "bbox": [ + 105, + 533, + 403, + 546 + ], + "score": 1.0, + "content": "baselines and high-quality benchmarks are imperative for future progress.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 26, + "bbox_fs": [ + 105, + 424, + 506, + 546 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 550, + 505, + 693 + ], + "lines": [ + { + "bbox": [ + 106, + 551, + 505, + 562 + ], + "spans": [ + { + "bbox": [ + 106, + 551, + 505, + 562 + ], + "score": 1.0, + "content": "Large-scale datasets such as ImageNet (Deng et al., 2009) and Places365 (Zhou et al., 2017) present", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 561, + 506, + 574 + ], + "spans": [ + { + "bbox": [ + 106, + 561, + 506, + 574 + ], + "score": 1.0, + "content": "unique challenges not seen in small-scale settings, such as a plethora of fine-grained object classes.", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 572, + 506, + 585 + ], + "spans": [ + { + "bbox": [ + 106, + 572, + 506, + 585 + ], + "score": 1.0, + "content": "We demonstrate that the maximum softmax probability (MSP) detector, a state-of-the-art method", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 582, + 505, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 582, + 505, + 596 + ], + "score": 1.0, + "content": "for small-scale problems, does not scale well to these challenging conditions. Through extensive", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 594, + 505, + 607 + ], + "spans": [ + { + "bbox": [ + 105, + 594, + 505, + 607 + ], + "score": 1.0, + "content": "experiments, we identify a detector based on the maximum logit (MaxLogit) that greatly outperforms", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 605, + 505, + 617 + ], + "spans": [ + { + "bbox": [ + 106, + 605, + 505, + 617 + ], + "score": 1.0, + "content": "the MSP and other strong baselines in large-scale multi-class anomaly segmentation. To facilitate", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 615, + 506, + 629 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 506, + 629 + ], + "score": 1.0, + "content": "further research in this setting, we also collect a new out-of-distribution test dataset suitable for", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 626, + 506, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 506, + 640 + ], + "score": 1.0, + "content": "models trained on highly diverse datasets. Shown in Figure 2, our Species dataset contains diverse,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 638, + 506, + 651 + ], + "spans": [ + { + "bbox": [ + 105, + 638, + 506, + 651 + ], + "score": 1.0, + "content": "anomalous species that do not overlap ImageNet-21K which has approximately twenty two thousand", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 106, + 648, + 505, + 661 + ], + "spans": [ + { + "bbox": [ + 106, + 648, + 505, + 661 + ], + "score": 1.0, + "content": "classes. Species avoids data leakage and enables a stricter evaluation methodology for ImageNet-21K", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 660, + 506, + 673 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 506, + 673 + ], + "score": 1.0, + "content": "models. Using Species to conduct more controlled experiments without train-test overlap, we find that", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 670, + 507, + 684 + ], + "spans": [ + { + "bbox": [ + 105, + 670, + 507, + 684 + ], + "score": 1.0, + "content": "contrary to prior claims (Fort et al., 2021; Koner et al., 2021), Vision Transformers (Dosovitskiy et al.,", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 682, + 492, + 694 + ], + "spans": [ + { + "bbox": [ + 106, + 682, + 492, + 694 + ], + "score": 1.0, + "content": "2021a) pre-trained on ImageNet-21K are not substantially better at out-of-distribution detection.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 38, + "bbox_fs": [ + 105, + 551, + 507, + 694 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 699, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 699, + 505, + 710 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 710 + ], + "score": 1.0, + "content": "Moreover, in the common real-world case of multi-label data, the MSP detector cannot naturally be", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "applied in the first place, as it requires softmax probabilities. To enable research into the multi-label", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 505, + 733 + ], + "score": 1.0, + "content": "setting for anomaly detection, we contribute a multi-label experimental setup and explore various", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "score": 1.0, + "content": "methods on large-scale multi-label datasets. We find that the MaxLogit detector from our investigation", + "type": "text", + "cross_page": true + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 401, + 505, + 413 + ], + "spans": [ + { + "bbox": [ + 106, + 401, + 505, + 413 + ], + "score": 1.0, + "content": "into the large-scale multi-class setting generalizes well to multi-label data and again outperforms all", + "type": "text", + "cross_page": true + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 411, + 171, + 424 + ], + "spans": [ + { + "bbox": [ + 106, + 411, + 171, + 424 + ], + "score": 1.0, + "content": "other baselines.", + "type": "text", + "cross_page": true + } + ], + "index": 12 + } + ], + "index": 46, + "bbox_fs": [ + 106, + 699, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 79, + 505, + 297 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 79, + 505, + 297 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 79, + 505, + 297 + ], + "spans": [ + { + "bbox": [ + 106, + 79, + 505, + 297 + ], + "score": 0.97, + "type": "image", + "image_path": "8595abd6b64885b25129dadf2053a7309f448d326b5e56de08cc3f4b6a1ec9f1.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 79, + 505, + 151.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 151.66666666666669, + 505, + 224.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 224.33333333333337, + 505, + 297.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 305, + 506, + 381 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 305, + 505, + 317 + ], + "spans": [ + { + "bbox": [ + 106, + 305, + 505, + 317 + ], + "score": 1.0, + "content": "Figure 1: We scale up out-of-distribution detection to large-scale multi-class datasets with thousands", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 315, + 506, + 329 + ], + "spans": [ + { + "bbox": [ + 105, + 315, + 506, + 329 + ], + "score": 1.0, + "content": "of classes, multi-label datasets with complex scenes, and anomaly segmentation in driving environ-", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 327, + 506, + 339 + ], + "spans": [ + { + "bbox": [ + 105, + 327, + 506, + 339 + ], + "score": 1.0, + "content": "ments. We introduce new benchmarks for all three settings. In all of these settings, we find that an", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 338, + 505, + 350 + ], + "spans": [ + { + "bbox": [ + 105, + 338, + 505, + 350 + ], + "score": 1.0, + "content": "OOD detector based on the maximum logit outperforms previous methods, establishing a strong and", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 348, + 506, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 348, + 506, + 361 + ], + "score": 1.0, + "content": "versatile baseline for future work on large-scale OOD detection. The bottom-right shows a scene", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 360, + 506, + 372 + ], + "spans": [ + { + "bbox": [ + 105, + 360, + 506, + 372 + ], + "score": 1.0, + "content": "from our new anomaly segmentation benchmark and the predicted anomaly using a state-of-the-art", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 370, + 144, + 383 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 144, + 383 + ], + "score": 1.0, + "content": "detector.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 6 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 108, + 389, + 504, + 423 + ], + "lines": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 389, + 505, + 403 + ], + "score": 1.0, + "content": "methods on large-scale multi-label datasets. We find that the MaxLogit detector from our investigation", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 401, + 505, + 413 + ], + "spans": [ + { + "bbox": [ + 106, + 401, + 505, + 413 + ], + "score": 1.0, + "content": "into the large-scale multi-class setting generalizes well to multi-label data and again outperforms all", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 411, + 171, + 424 + ], + "spans": [ + { + "bbox": [ + 106, + 411, + 171, + 424 + ], + "score": 1.0, + "content": "other baselines.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11 + }, + { + "type": "text", + "bbox": [ + 107, + 429, + 505, + 506 + ], + "lines": [ + { + "bbox": [ + 105, + 429, + 505, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 429, + 505, + 441 + ], + "score": 1.0, + "content": "In addition to focusing on small-scale datasets, most existing benchmarks for anomaly detection", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 440, + 505, + 452 + ], + "spans": [ + { + "bbox": [ + 105, + 440, + 505, + 452 + ], + "score": 1.0, + "content": "treat entire images as anomalies. In practice, an image could be anomalous in localized regions while", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 451, + 505, + 463 + ], + "spans": [ + { + "bbox": [ + 105, + 451, + 505, + 463 + ], + "score": 1.0, + "content": "being in-distribution elsewhere. Knowing which regions of an image are anomalous could allow", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "score": 1.0, + "content": "for safer handling of unfamiliar objects in the case of self-driving cars. Creating a benchmark for", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 473, + 505, + 485 + ], + "spans": [ + { + "bbox": [ + 105, + 473, + 505, + 485 + ], + "score": 1.0, + "content": "this task is difficult, though, as simply cutting and pasting anomalous objects into images introduces", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 484, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 484, + 506, + 496 + ], + "score": 1.0, + "content": "various unnatural giveaway cues such as edge effects, mismatched orientation, and lighting, all of", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 495, + 387, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 495, + 387, + 507 + ], + "score": 1.0, + "content": "which trivialize the task of anomaly segmentation (Blum et al., 2019).", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 107, + 511, + 505, + 567 + ], + "lines": [ + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "score": 1.0, + "content": "To overcome these issues, we utilize a simulated driving environment to create the novel StreetHazards", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 522, + 505, + 536 + ], + "spans": [ + { + "bbox": [ + 106, + 522, + 505, + 536 + ], + "score": 1.0, + "content": "dataset for anomaly segmentation. Using the Unreal Engine and the open-source CARLA simulation", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 533, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 533, + 505, + 546 + ], + "score": 1.0, + "content": "environment (Dosovitskiy et al., 2017), we insert a diverse array of foreign objects into driving scenes", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 543, + 505, + 558 + ], + "spans": [ + { + "bbox": [ + 105, + 543, + 505, + 558 + ], + "score": 1.0, + "content": "and re-render the scenes with these novel objects. This enables integration of the foreign objects into", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 555, + 473, + 568 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 473, + 568 + ], + "score": 1.0, + "content": "their surrounding context with correct lighting and orientation, sidestepping giveaway cues.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 107, + 572, + 505, + 660 + ], + "lines": [ + { + "bbox": [ + 105, + 571, + 506, + 584 + ], + "spans": [ + { + "bbox": [ + 105, + 571, + 506, + 584 + ], + "score": 1.0, + "content": "To complement the StreetHazards dataset, we convert the BDD100K semantic segmentation dataset", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 582, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 582, + 506, + 596 + ], + "score": 1.0, + "content": "(Yu et al., 2018) into an anomaly segmentation dataset, which we call BDD-Anomaly. By leveraging", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 594, + 505, + 606 + ], + "spans": [ + { + "bbox": [ + 105, + 594, + 505, + 606 + ], + "score": 1.0, + "content": "the large scale of BDD100K, we reserve infrequent object classes to be anomalies. We combine", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 605, + 506, + 617 + ], + "spans": [ + { + "bbox": [ + 105, + 605, + 506, + 617 + ], + "score": 1.0, + "content": "this dataset with StreetHazards to form the Combined Anomalous Object Segmentation (CAOS)", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 615, + 505, + 629 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 505, + 629 + ], + "score": 1.0, + "content": "benchmark. The CAOS benchmark improves over previous evaluations for anomaly segmentation", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 627, + 505, + 639 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 505, + 639 + ], + "score": 1.0, + "content": "in driving scenes by evaluating detectors on realistic and diverse anomalies. We evaluate several", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 637, + 505, + 651 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 505, + 651 + ], + "score": 1.0, + "content": "baselines on the CAOS benchmark and discuss problems with porting existing approaches from", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 648, + 316, + 661 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 316, + 661 + ], + "score": 1.0, + "content": "earlier formulations of out-of-distribution detection.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 28.5 + }, + { + "type": "text", + "bbox": [ + 107, + 666, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 664, + 507, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 507, + 679 + ], + "score": 1.0, + "content": "Despite its simplicity, we find that the MaxLogit detector outperforms all baselines on Species,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "our multi-class benchmark, and CAOS. In each of these three settings, we discuss why MaxLogit", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "provides superior performance, and we show that these gains are hidden if one looks at small-scale", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "problems alone. The code for our experiments and the Species and CAOS datasets are available", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "at [anonymized]. Our new baseline combined with Species and CAOS benchmarks pave the", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 721, + 376, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 376, + 733 + ], + "score": 1.0, + "content": "way for future research on large-scale out-of-distribution detection.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 35.5 + } + ], + "page_idx": 1, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 763 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 763 + ], + "score": 1.0, + "content": "2", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 79, + 505, + 297 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 79, + 505, + 297 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 79, + 505, + 297 + ], + "spans": [ + { + "bbox": [ + 106, + 79, + 505, + 297 + ], + "score": 0.97, + "type": "image", + "image_path": "8595abd6b64885b25129dadf2053a7309f448d326b5e56de08cc3f4b6a1ec9f1.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 79, + 505, + 151.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 151.66666666666669, + 505, + 224.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 224.33333333333337, + 505, + 297.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 305, + 506, + 381 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 305, + 505, + 317 + ], + "spans": [ + { + "bbox": [ + 106, + 305, + 505, + 317 + ], + "score": 1.0, + "content": "Figure 1: We scale up out-of-distribution detection to large-scale multi-class datasets with thousands", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 315, + 506, + 329 + ], + "spans": [ + { + "bbox": [ + 105, + 315, + 506, + 329 + ], + "score": 1.0, + "content": "of classes, multi-label datasets with complex scenes, and anomaly segmentation in driving environ-", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 327, + 506, + 339 + ], + "spans": [ + { + "bbox": [ + 105, + 327, + 506, + 339 + ], + "score": 1.0, + "content": "ments. We introduce new benchmarks for all three settings. In all of these settings, we find that an", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 338, + 505, + 350 + ], + "spans": [ + { + "bbox": [ + 105, + 338, + 505, + 350 + ], + "score": 1.0, + "content": "OOD detector based on the maximum logit outperforms previous methods, establishing a strong and", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 348, + 506, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 348, + 506, + 361 + ], + "score": 1.0, + "content": "versatile baseline for future work on large-scale OOD detection. The bottom-right shows a scene", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 360, + 506, + 372 + ], + "spans": [ + { + "bbox": [ + 105, + 360, + 506, + 372 + ], + "score": 1.0, + "content": "from our new anomaly segmentation benchmark and the predicted anomaly using a state-of-the-art", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 370, + 144, + 383 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 144, + 383 + ], + "score": 1.0, + "content": "detector.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 6 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 108, + 389, + 504, + 423 + ], + "lines": [], + "index": 11, + "bbox_fs": [ + 105, + 389, + 505, + 424 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 429, + 505, + 506 + ], + "lines": [ + { + "bbox": [ + 105, + 429, + 505, + 441 + ], + "spans": [ + { + "bbox": [ + 105, + 429, + 505, + 441 + ], + "score": 1.0, + "content": "In addition to focusing on small-scale datasets, most existing benchmarks for anomaly detection", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 440, + 505, + 452 + ], + "spans": [ + { + "bbox": [ + 105, + 440, + 505, + 452 + ], + "score": 1.0, + "content": "treat entire images as anomalies. In practice, an image could be anomalous in localized regions while", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 451, + 505, + 463 + ], + "spans": [ + { + "bbox": [ + 105, + 451, + 505, + 463 + ], + "score": 1.0, + "content": "being in-distribution elsewhere. Knowing which regions of an image are anomalous could allow", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "spans": [ + { + "bbox": [ + 105, + 461, + 505, + 474 + ], + "score": 1.0, + "content": "for safer handling of unfamiliar objects in the case of self-driving cars. Creating a benchmark for", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 473, + 505, + 485 + ], + "spans": [ + { + "bbox": [ + 105, + 473, + 505, + 485 + ], + "score": 1.0, + "content": "this task is difficult, though, as simply cutting and pasting anomalous objects into images introduces", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 484, + 506, + 496 + ], + "spans": [ + { + "bbox": [ + 105, + 484, + 506, + 496 + ], + "score": 1.0, + "content": "various unnatural giveaway cues such as edge effects, mismatched orientation, and lighting, all of", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 495, + 387, + 507 + ], + "spans": [ + { + "bbox": [ + 105, + 495, + 387, + 507 + ], + "score": 1.0, + "content": "which trivialize the task of anomaly segmentation (Blum et al., 2019).", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 16, + "bbox_fs": [ + 105, + 429, + 506, + 507 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 511, + 505, + 567 + ], + "lines": [ + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 106, + 512, + 505, + 524 + ], + "score": 1.0, + "content": "To overcome these issues, we utilize a simulated driving environment to create the novel StreetHazards", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 522, + 505, + 536 + ], + "spans": [ + { + "bbox": [ + 106, + 522, + 505, + 536 + ], + "score": 1.0, + "content": "dataset for anomaly segmentation. Using the Unreal Engine and the open-source CARLA simulation", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 533, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 533, + 505, + 546 + ], + "score": 1.0, + "content": "environment (Dosovitskiy et al., 2017), we insert a diverse array of foreign objects into driving scenes", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 543, + 505, + 558 + ], + "spans": [ + { + "bbox": [ + 105, + 543, + 505, + 558 + ], + "score": 1.0, + "content": "and re-render the scenes with these novel objects. This enables integration of the foreign objects into", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 555, + 473, + 568 + ], + "spans": [ + { + "bbox": [ + 105, + 555, + 473, + 568 + ], + "score": 1.0, + "content": "their surrounding context with correct lighting and orientation, sidestepping giveaway cues.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 22, + "bbox_fs": [ + 105, + 512, + 505, + 568 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 572, + 505, + 660 + ], + "lines": [ + { + "bbox": [ + 105, + 571, + 506, + 584 + ], + "spans": [ + { + "bbox": [ + 105, + 571, + 506, + 584 + ], + "score": 1.0, + "content": "To complement the StreetHazards dataset, we convert the BDD100K semantic segmentation dataset", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 582, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 582, + 506, + 596 + ], + "score": 1.0, + "content": "(Yu et al., 2018) into an anomaly segmentation dataset, which we call BDD-Anomaly. By leveraging", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 594, + 505, + 606 + ], + "spans": [ + { + "bbox": [ + 105, + 594, + 505, + 606 + ], + "score": 1.0, + "content": "the large scale of BDD100K, we reserve infrequent object classes to be anomalies. We combine", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 605, + 506, + 617 + ], + "spans": [ + { + "bbox": [ + 105, + 605, + 506, + 617 + ], + "score": 1.0, + "content": "this dataset with StreetHazards to form the Combined Anomalous Object Segmentation (CAOS)", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 615, + 505, + 629 + ], + "spans": [ + { + "bbox": [ + 105, + 615, + 505, + 629 + ], + "score": 1.0, + "content": "benchmark. The CAOS benchmark improves over previous evaluations for anomaly segmentation", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 627, + 505, + 639 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 505, + 639 + ], + "score": 1.0, + "content": "in driving scenes by evaluating detectors on realistic and diverse anomalies. We evaluate several", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 637, + 505, + 651 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 505, + 651 + ], + "score": 1.0, + "content": "baselines on the CAOS benchmark and discuss problems with porting existing approaches from", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 648, + 316, + 661 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 316, + 661 + ], + "score": 1.0, + "content": "earlier formulations of out-of-distribution detection.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 28.5, + "bbox_fs": [ + 105, + 571, + 506, + 661 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 666, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 664, + 507, + 679 + ], + "spans": [ + { + "bbox": [ + 105, + 664, + 507, + 679 + ], + "score": 1.0, + "content": "Despite its simplicity, we find that the MaxLogit detector outperforms all baselines on Species,", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "our multi-class benchmark, and CAOS. In each of these three settings, we discuss why MaxLogit", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "provides superior performance, and we show that these gains are hidden if one looks at small-scale", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 711 + ], + "score": 1.0, + "content": "problems alone. The code for our experiments and the Species and CAOS datasets are available", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "at [anonymized]. Our new baseline combined with Species and CAOS benchmarks pave the", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 721, + 376, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 376, + 733 + ], + "score": 1.0, + "content": "way for future research on large-scale out-of-distribution detection.", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 35.5, + "bbox_fs": [ + 105, + 664, + 507, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 250, + 81, + 362, + 92 + ], + "lines": [ + { + "bbox": [ + 248, + 79, + 364, + 94 + ], + "spans": [ + { + "bbox": [ + 248, + 79, + 364, + 94 + ], + "score": 1.0, + "content": "Anomalous Species Dataset", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "image", + "bbox": [ + 107, + 94, + 504, + 283 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 94, + 504, + 283 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 94, + 504, + 283 + ], + "spans": [ + { + "bbox": [ + 107, + 94, + 504, + 283 + ], + "score": 0.973, + "type": "image", + "image_path": "654db4cbaa3ed751b6c6dce51acbc564fb80912d3bccb104063743fd809b6f2d.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 107, + 94, + 504, + 157.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 157.0, + 504, + 220.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 107, + 220.0, + 504, + 283.0 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 292, + 506, + 347 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 293, + 505, + 304 + ], + "spans": [ + { + "bbox": [ + 106, + 293, + 505, + 304 + ], + "score": 1.0, + "content": "Figure 2: The Species out-of-distribution dataset is designed for large-scale anomaly detectors", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 302, + 506, + 316 + ], + "spans": [ + { + "bbox": [ + 105, + 302, + 506, + 316 + ], + "score": 1.0, + "content": "pretrained on datasets as diverse as ImageNet-21K. When models are pretrained on ImageNet-21K,", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 104, + 313, + 506, + 327 + ], + "spans": [ + { + "bbox": [ + 104, + 313, + 506, + 327 + ], + "score": 1.0, + "content": "many previous OOD detection datasets may overlap with the pretraining set, resulting in erroneous", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 325, + 505, + 337 + ], + "spans": [ + { + "bbox": [ + 106, + 325, + 505, + 337 + ], + "score": 1.0, + "content": "evaluations. To rectify this, Species is comprised of hundreds of anomalous species that are disjoint", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 336, + 424, + 348 + ], + "spans": [ + { + "bbox": [ + 106, + 336, + 424, + 348 + ], + "score": 1.0, + "content": "from ImageNet-21K classes and enables the evaluation of cutting-edge models.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 6 + } + ], + "index": 4.0 + }, + { + "type": "title", + "bbox": [ + 108, + 379, + 211, + 392 + ], + "lines": [ + { + "bbox": [ + 104, + 378, + 213, + 394 + ], + "spans": [ + { + "bbox": [ + 104, + 378, + 213, + 394 + ], + "score": 1.0, + "content": "2 RELATED WORK", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 107, + 412, + 506, + 490 + ], + "lines": [ + { + "bbox": [ + 105, + 411, + 506, + 426 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 506, + 426 + ], + "score": 1.0, + "content": "Multi-Class Out-of-Distribution Detection. A recent line of work leverages deep neural rep-", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 424, + 506, + 436 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 506, + 436 + ], + "score": 1.0, + "content": "resentations from multi-class classifiers to perform out-of-distribution (OOD) detection on high-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "score": 1.0, + "content": "dimensional data, including images, text, and speech data. Hendrycks & Gimpel (2017) formulate", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 446, + 506, + 459 + ], + "spans": [ + { + "bbox": [ + 106, + 446, + 506, + 459 + ], + "score": 1.0, + "content": "the task and propose the simple baseline of using the maximum softmax probability of the classi-", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 457, + 506, + 469 + ], + "spans": [ + { + "bbox": [ + 106, + 457, + 506, + 469 + ], + "score": 1.0, + "content": "fier on an input to gauge whether the input is out-of-distribution. In particular, they formulate the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 468, + 507, + 480 + ], + "spans": [ + { + "bbox": [ + 106, + 468, + 507, + 480 + ], + "score": 1.0, + "content": "task as distinguishing between examples from an in-distribution dataset and various OOD datasets.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 479, + 347, + 491 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 347, + 491 + ], + "score": 1.0, + "content": "Importantly, entire images are treated as out-of-distribution.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 13 + }, + { + "type": "text", + "bbox": [ + 107, + 495, + 505, + 627 + ], + "lines": [ + { + "bbox": [ + 105, + 495, + 506, + 509 + ], + "spans": [ + { + "bbox": [ + 105, + 495, + 506, + 509 + ], + "score": 1.0, + "content": "Continuing this line of work, Lee et al. (2018a) propose to improve the neural representation of the", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 506, + 506, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 506, + 519 + ], + "score": 1.0, + "content": "classifier to better separate OOD examples. They use generative adversarial networks to produce near-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 518, + 505, + 530 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 505, + 530 + ], + "score": 1.0, + "content": "distribution examples and induce uniform posteriors on these synthetic OOD examples. Hendrycks", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 528, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 505, + 541 + ], + "score": 1.0, + "content": "et al. (2019b) observe that outliers are often easy to obtain in large quantity from diverse, realistic", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 539, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 539, + 505, + 552 + ], + "score": 1.0, + "content": "datasets and demonstrate that OOD detectors trained on these outliers generalize to unseen classes", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 550, + 506, + 563 + ], + "spans": [ + { + "bbox": [ + 106, + 550, + 506, + 563 + ], + "score": 1.0, + "content": "of anomalies. Other work investigates improving the anomaly detectors themselves given a fixed", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 561, + 507, + 574 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 507, + 574 + ], + "score": 1.0, + "content": "classifier (DeVries & Taylor, 2018; Liang et al., 2018). However, as Hendrycks et al. (2019b) observe,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 572, + 506, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 506, + 586 + ], + "score": 1.0, + "content": "many of these works tune hyperparameters on a particular type of anomaly that is also seen at test", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 583, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 106, + 583, + 506, + 596 + ], + "score": 1.0, + "content": "time, so their evaluation setting is more lenient. In this paper, all anomalies seen at test time come", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 594, + 506, + 607 + ], + "spans": [ + { + "bbox": [ + 105, + 594, + 506, + 607 + ], + "score": 1.0, + "content": "from entirely unseen categories and are not tuned on in any way. Hence, we do not compare to", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 605, + 506, + 617 + ], + "spans": [ + { + "bbox": [ + 106, + 605, + 506, + 617 + ], + "score": 1.0, + "content": "techniques such as ODIN (Liang et al., 2018). Additionally, in a point of departure from prior work,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 617, + 401, + 629 + ], + "spans": [ + { + "bbox": [ + 106, + 617, + 401, + 629 + ], + "score": 1.0, + "content": "we focus primarily on large-scale images and datasets with many classes.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 22.5 + }, + { + "type": "text", + "bbox": [ + 107, + 633, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 632, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 506, + 646 + ], + "score": 1.0, + "content": "Recent work has suggested that stronger representations from Vision Transformers pre-trained on", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 644, + 506, + 656 + ], + "spans": [ + { + "bbox": [ + 106, + 644, + 506, + 656 + ], + "score": 1.0, + "content": "ImageNet-21K can make out-of-distribution detection trivial (Fort et al., 2021; Koner et al., 2021).", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 654, + 506, + 669 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 506, + 669 + ], + "score": 1.0, + "content": "They evaluate models on detecting CIFAR-10 when fine-tuned on CIFAR-100 or vice versa, using", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 666, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 505, + 678 + ], + "score": 1.0, + "content": "models pretrained on ImageNet-21K. However, over 1,000 classes in ImageNet-21K overlap with", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "CIFAR-10, so it is still unclear how Vision Transformers perform at detecting entirely unseen OOD", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "score": 1.0, + "content": "categories. We create a new OOD test dataset of anomalous species to investigate how well Vision", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 698, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 712 + ], + "score": 1.0, + "content": "Transformers perform in controlled OOD detection settings without data leakage and overlap. We", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "score": 1.0, + "content": "find that Vision Transformers pre-trained on ImageNet-21K are far from solving OOD detection in", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 720, + 187, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 187, + 734 + ], + "score": 1.0, + "content": "large-scale settings.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 33 + } + ], + "page_idx": 2, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "spans": [ + { + "bbox": [ + 301, + 750, + 310, + 762 + ], + "score": 1.0, + "content": "3", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 250, + 81, + 362, + 92 + ], + "lines": [ + { + "bbox": [ + 248, + 79, + 364, + 94 + ], + "spans": [ + { + "bbox": [ + 248, + 79, + 364, + 94 + ], + "score": 1.0, + "content": "Anomalous Species Dataset", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "image", + "bbox": [ + 107, + 94, + 504, + 283 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 107, + 94, + 504, + 283 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 107, + 94, + 504, + 283 + ], + "spans": [ + { + "bbox": [ + 107, + 94, + 504, + 283 + ], + "score": 0.973, + "type": "image", + "image_path": "654db4cbaa3ed751b6c6dce51acbc564fb80912d3bccb104063743fd809b6f2d.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 107, + 94, + 504, + 157.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 107, + 157.0, + 504, + 220.0 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 107, + 220.0, + 504, + 283.0 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 292, + 506, + 347 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 293, + 505, + 304 + ], + "spans": [ + { + "bbox": [ + 106, + 293, + 505, + 304 + ], + "score": 1.0, + "content": "Figure 2: The Species out-of-distribution dataset is designed for large-scale anomaly detectors", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 302, + 506, + 316 + ], + "spans": [ + { + "bbox": [ + 105, + 302, + 506, + 316 + ], + "score": 1.0, + "content": "pretrained on datasets as diverse as ImageNet-21K. When models are pretrained on ImageNet-21K,", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 104, + 313, + 506, + 327 + ], + "spans": [ + { + "bbox": [ + 104, + 313, + 506, + 327 + ], + "score": 1.0, + "content": "many previous OOD detection datasets may overlap with the pretraining set, resulting in erroneous", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 325, + 505, + 337 + ], + "spans": [ + { + "bbox": [ + 106, + 325, + 505, + 337 + ], + "score": 1.0, + "content": "evaluations. To rectify this, Species is comprised of hundreds of anomalous species that are disjoint", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 336, + 424, + 348 + ], + "spans": [ + { + "bbox": [ + 106, + 336, + 424, + 348 + ], + "score": 1.0, + "content": "from ImageNet-21K classes and enables the evaluation of cutting-edge models.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 6 + } + ], + "index": 4.0 + }, + { + "type": "title", + "bbox": [ + 108, + 379, + 211, + 392 + ], + "lines": [ + { + "bbox": [ + 104, + 378, + 213, + 394 + ], + "spans": [ + { + "bbox": [ + 104, + 378, + 213, + 394 + ], + "score": 1.0, + "content": "2 RELATED WORK", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 107, + 412, + 506, + 490 + ], + "lines": [ + { + "bbox": [ + 105, + 411, + 506, + 426 + ], + "spans": [ + { + "bbox": [ + 105, + 411, + 506, + 426 + ], + "score": 1.0, + "content": "Multi-Class Out-of-Distribution Detection. A recent line of work leverages deep neural rep-", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 424, + 506, + 436 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 506, + 436 + ], + "score": 1.0, + "content": "resentations from multi-class classifiers to perform out-of-distribution (OOD) detection on high-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "score": 1.0, + "content": "dimensional data, including images, text, and speech data. Hendrycks & Gimpel (2017) formulate", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 446, + 506, + 459 + ], + "spans": [ + { + "bbox": [ + 106, + 446, + 506, + 459 + ], + "score": 1.0, + "content": "the task and propose the simple baseline of using the maximum softmax probability of the classi-", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 457, + 506, + 469 + ], + "spans": [ + { + "bbox": [ + 106, + 457, + 506, + 469 + ], + "score": 1.0, + "content": "fier on an input to gauge whether the input is out-of-distribution. In particular, they formulate the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 468, + 507, + 480 + ], + "spans": [ + { + "bbox": [ + 106, + 468, + 507, + 480 + ], + "score": 1.0, + "content": "task as distinguishing between examples from an in-distribution dataset and various OOD datasets.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 479, + 347, + 491 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 347, + 491 + ], + "score": 1.0, + "content": "Importantly, entire images are treated as out-of-distribution.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 13, + "bbox_fs": [ + 105, + 411, + 507, + 491 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 495, + 505, + 627 + ], + "lines": [ + { + "bbox": [ + 105, + 495, + 506, + 509 + ], + "spans": [ + { + "bbox": [ + 105, + 495, + 506, + 509 + ], + "score": 1.0, + "content": "Continuing this line of work, Lee et al. (2018a) propose to improve the neural representation of the", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 506, + 506, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 506, + 519 + ], + "score": 1.0, + "content": "classifier to better separate OOD examples. They use generative adversarial networks to produce near-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 518, + 505, + 530 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 505, + 530 + ], + "score": 1.0, + "content": "distribution examples and induce uniform posteriors on these synthetic OOD examples. Hendrycks", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 528, + 505, + 541 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 505, + 541 + ], + "score": 1.0, + "content": "et al. (2019b) observe that outliers are often easy to obtain in large quantity from diverse, realistic", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 539, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 539, + 505, + 552 + ], + "score": 1.0, + "content": "datasets and demonstrate that OOD detectors trained on these outliers generalize to unseen classes", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 550, + 506, + 563 + ], + "spans": [ + { + "bbox": [ + 106, + 550, + 506, + 563 + ], + "score": 1.0, + "content": "of anomalies. Other work investigates improving the anomaly detectors themselves given a fixed", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 561, + 507, + 574 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 507, + 574 + ], + "score": 1.0, + "content": "classifier (DeVries & Taylor, 2018; Liang et al., 2018). However, as Hendrycks et al. (2019b) observe,", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 572, + 506, + 586 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 506, + 586 + ], + "score": 1.0, + "content": "many of these works tune hyperparameters on a particular type of anomaly that is also seen at test", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 583, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 106, + 583, + 506, + 596 + ], + "score": 1.0, + "content": "time, so their evaluation setting is more lenient. In this paper, all anomalies seen at test time come", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 594, + 506, + 607 + ], + "spans": [ + { + "bbox": [ + 105, + 594, + 506, + 607 + ], + "score": 1.0, + "content": "from entirely unseen categories and are not tuned on in any way. Hence, we do not compare to", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 605, + 506, + 617 + ], + "spans": [ + { + "bbox": [ + 106, + 605, + 506, + 617 + ], + "score": 1.0, + "content": "techniques such as ODIN (Liang et al., 2018). Additionally, in a point of departure from prior work,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 617, + 401, + 629 + ], + "spans": [ + { + "bbox": [ + 106, + 617, + 401, + 629 + ], + "score": 1.0, + "content": "we focus primarily on large-scale images and datasets with many classes.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 22.5, + "bbox_fs": [ + 105, + 495, + 507, + 629 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 633, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 632, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 105, + 632, + 506, + 646 + ], + "score": 1.0, + "content": "Recent work has suggested that stronger representations from Vision Transformers pre-trained on", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 106, + 644, + 506, + 656 + ], + "spans": [ + { + "bbox": [ + 106, + 644, + 506, + 656 + ], + "score": 1.0, + "content": "ImageNet-21K can make out-of-distribution detection trivial (Fort et al., 2021; Koner et al., 2021).", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 654, + 506, + 669 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 506, + 669 + ], + "score": 1.0, + "content": "They evaluate models on detecting CIFAR-10 when fine-tuned on CIFAR-100 or vice versa, using", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 666, + 505, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 505, + 678 + ], + "score": 1.0, + "content": "models pretrained on ImageNet-21K. However, over 1,000 classes in ImageNet-21K overlap with", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "CIFAR-10, so it is still unclear how Vision Transformers perform at detecting entirely unseen OOD", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 505, + 700 + ], + "score": 1.0, + "content": "categories. We create a new OOD test dataset of anomalous species to investigate how well Vision", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 698, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 505, + 712 + ], + "score": 1.0, + "content": "Transformers perform in controlled OOD detection settings without data leakage and overlap. We", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "score": 1.0, + "content": "find that Vision Transformers pre-trained on ImageNet-21K are far from solving OOD detection in", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 720, + 187, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 187, + 734 + ], + "score": 1.0, + "content": "large-scale settings.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 33, + "bbox_fs": [ + 105, + 632, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 79, + 505, + 127 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 79, + 505, + 127 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 79, + 505, + 127 + ], + "spans": [ + { + "bbox": [ + 106, + 79, + 505, + 127 + ], + "score": 0.966, + "html": "
FPR95↓AUROC↑AUPR↑
DinMSPDeVriesMaxLogitMSPDeVries MaxLogitMSPDeVriesMaxLogit
ImageNet44.246.035.884.676.987.238.230.545.8
Places36552.685.836.676.031.185.88.22.019.2
", + "type": "table", + "image_path": "bb7b3019a2b86e9009d5307a4d02ae9f36ffd04663404bc2ebffac2f51127748.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 79, + 505, + 95.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 95.0, + 505, + 111.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 111.0, + 505, + 127.0 + ], + "spans": [], + "index": 2 + } + ] + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 106, + 136, + 505, + 180 + ], + "lines": [ + { + "bbox": [ + 105, + 135, + 505, + 148 + ], + "spans": [ + { + "bbox": [ + 105, + 135, + 505, + 148 + ], + "score": 1.0, + "content": "Table 1: Multi-class out-of-distribution detection results using the maximum softmax probability", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 146, + 505, + 158 + ], + "spans": [ + { + "bbox": [ + 105, + 146, + 505, + 158 + ], + "score": 1.0, + "content": "(MSP) baseline (Hendrycks & Gimpel, 2017), the confidence branch detector of DeVries & Taylor", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 157, + 507, + 170 + ], + "spans": [ + { + "bbox": [ + 105, + 157, + 507, + 170 + ], + "score": 1.0, + "content": "(2018), and our maximum logit baseline. All values are percentages and average across five out-of-", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 168, + 469, + 181 + ], + "spans": [ + { + "bbox": [ + 105, + 168, + 469, + 181 + ], + "score": 1.0, + "content": "distribution test datasets. Full results on individual OOD test datasets are in the Appendix.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + }, + { + "type": "text", + "bbox": [ + 106, + 214, + 505, + 291 + ], + "lines": [ + { + "bbox": [ + 105, + 214, + 505, + 227 + ], + "spans": [ + { + "bbox": [ + 105, + 214, + 505, + 227 + ], + "score": 1.0, + "content": "Anomaly Segmentation. Several prior works explore segmenting anomalous image regions. One", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 225, + 505, + 238 + ], + "spans": [ + { + "bbox": [ + 105, + 225, + 505, + 238 + ], + "score": 1.0, + "content": "line of work uses the WildDash dataset (Zendel et al., 2018), which contains numerous annotated", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 237, + 505, + 248 + ], + "spans": [ + { + "bbox": [ + 106, + 237, + 505, + 248 + ], + "score": 1.0, + "content": "driving scenes in conditions such as snow, fog, and rain. The WildDash test set contains fifteen", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 246, + 506, + 260 + ], + "spans": [ + { + "bbox": [ + 105, + 246, + 506, + 260 + ], + "score": 1.0, + "content": "“negative images” from different domains for which the goal is to mark the entire image as out-of-", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 258, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 106, + 258, + 505, + 271 + ], + "score": 1.0, + "content": "distribution. Thus, while the task is segmentation, the anomalies do not exist as objects within an", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 270, + 505, + 282 + ], + "spans": [ + { + "bbox": [ + 106, + 270, + 505, + 282 + ], + "score": 1.0, + "content": "otherwise in-distribution scene. This setting is similar to that explored by Hendrycks & Gimpel", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 280, + 463, + 293 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 463, + 293 + ], + "score": 1.0, + "content": "(2017), in which whole images from other datasets serve as out-of-distribution examples.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 10 + }, + { + "type": "text", + "bbox": [ + 107, + 297, + 505, + 352 + ], + "lines": [ + { + "bbox": [ + 105, + 297, + 505, + 310 + ], + "spans": [ + { + "bbox": [ + 105, + 297, + 505, + 310 + ], + "score": 1.0, + "content": "To approach anomaly segmentation on WildDash, Krešo et al. (2018) train on multiple semantic", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 308, + 506, + 320 + ], + "spans": [ + { + "bbox": [ + 105, + 308, + 506, + 320 + ], + "score": 1.0, + "content": "segmentation domains and treat regions of images from the WildDash driving dataset as out-of-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 320, + 504, + 331 + ], + "spans": [ + { + "bbox": [ + 106, + 320, + 504, + 331 + ], + "score": 1.0, + "content": "distribution if they are segmented as regions from different domains, i.e. indoor classes. Bevandic´", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 329, + 505, + 344 + ], + "spans": [ + { + "bbox": [ + 105, + 329, + 505, + 344 + ], + "score": 1.0, + "content": "et al. (2018) use ILSVRC 2012 images and train their network to segment the entirety of these images", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 342, + 195, + 353 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 195, + 353 + ], + "score": 1.0, + "content": "as out-of-distribution.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 16 + }, + { + "type": "text", + "bbox": [ + 107, + 357, + 506, + 446 + ], + "lines": [ + { + "bbox": [ + 106, + 358, + 505, + 370 + ], + "spans": [ + { + "bbox": [ + 106, + 358, + 505, + 370 + ], + "score": 1.0, + "content": "In medical anomaly segmentation and product fault detection, anomalies are regions of otherwise", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 369, + 507, + 381 + ], + "spans": [ + { + "bbox": [ + 106, + 369, + 507, + 381 + ], + "score": 1.0, + "content": "in-distribution images. Baur et al. (2019) segment anomalous regions in brain MRIs using pixel-", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 380, + 505, + 393 + ], + "spans": [ + { + "bbox": [ + 105, + 380, + 505, + 393 + ], + "score": 1.0, + "content": "wise reconstruction loss. Similarly, Haselmann et al. (2018) perform product fault detection using", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 391, + 506, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 391, + 506, + 403 + ], + "score": 1.0, + "content": "pixel-wise reconstruction loss and introduce an expansive dataset for segmentation of product faults.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 402, + 506, + 415 + ], + "spans": [ + { + "bbox": [ + 105, + 402, + 506, + 415 + ], + "score": 1.0, + "content": "In these relatively simple domains, reconstruction-based approaches work well. In contrast to", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 412, + 506, + 425 + ], + "spans": [ + { + "bbox": [ + 105, + 412, + 506, + 425 + ], + "score": 1.0, + "content": "medical anomaly segmentation and fault detection, we consider complex images from street scenes.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 424, + 506, + 437 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 506, + 437 + ], + "score": 1.0, + "content": "These images have high variability in scene layout and lighting, and hence are less amenable to", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 435, + 239, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 239, + 447 + ], + "score": 1.0, + "content": "reconstruction-based techniques.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 22.5 + }, + { + "type": "text", + "bbox": [ + 106, + 451, + 505, + 605 + ], + "lines": [ + { + "bbox": [ + 105, + 451, + 505, + 464 + ], + "spans": [ + { + "bbox": [ + 105, + 451, + 505, + 464 + ], + "score": 1.0, + "content": "The two works closest to our own are the Lost and Found (Pinggera et al., 2016) and Fishyscapes", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 104, + 461, + 506, + 477 + ], + "spans": [ + { + "bbox": [ + 104, + 461, + 506, + 477 + ], + "score": 1.0, + "content": "(Blum et al., 2019) datasets. The Lost and Found dataset consists of real images in a driving", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 473, + 506, + 486 + ], + "spans": [ + { + "bbox": [ + 105, + 473, + 506, + 486 + ], + "score": 1.0, + "content": "environment with small road hazards. The images were collected to mirror the Cityscapes dataset", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 484, + 506, + 497 + ], + "spans": [ + { + "bbox": [ + 105, + 484, + 506, + 497 + ], + "score": 1.0, + "content": "(Cordts et al., 2016) but are only collected from one city and so have less diversity. The dataset", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 496, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 106, + 496, + 506, + 507 + ], + "score": 1.0, + "content": "contains 35 unique anomalous objects, and methods are allowed to train on many of these. For", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 506, + 506, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 506, + 519 + ], + "score": 1.0, + "content": "Lost and Found, only nine unique objects are truly unseen at test time. Crucially, this is a different", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 518, + 506, + 530 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 506, + 530 + ], + "score": 1.0, + "content": "evaluation setting from our own, where anomalous objects are not revealed at training time, so their", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 529, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 106, + 529, + 506, + 540 + ], + "score": 1.0, + "content": "dataset is not directly comparable. Nevertheless, the BDD-Anomaly dataset fills several gaps in Lost", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 539, + 506, + 552 + ], + "spans": [ + { + "bbox": [ + 106, + 539, + 506, + 552 + ], + "score": 1.0, + "content": "and Found. First, the images are more diverse, because they are sourced from a more recent and", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 551, + 506, + 562 + ], + "spans": [ + { + "bbox": [ + 106, + 551, + 506, + 562 + ], + "score": 1.0, + "content": "comprehensive semantic segmentation dataset. Second, the anomalies are not restricted to small,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 561, + 506, + 575 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 424, + 575 + ], + "score": 1.0, + "content": "sparse road hazards. Concretely, anomalous regions in Lost and Found take up", + "type": "text" + }, + { + "bbox": [ + 424, + 561, + 451, + 572 + ], + "score": 0.87, + "content": "0 . 1 1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 452, + 561, + 506, + 575 + ], + "score": 1.0, + "content": "of the image", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 572, + 505, + 585 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 452, + 585 + ], + "score": 1.0, + "content": "on average, whereas anomalous regions in the BDD-Anomaly dataset are larger and fill", + "type": "text" + }, + { + "bbox": [ + 452, + 572, + 479, + 583 + ], + "score": 0.87, + "content": "0 . 8 3 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 572, + 505, + 585 + ], + "score": 1.0, + "content": "of the", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 583, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 583, + 506, + 596 + ], + "score": 1.0, + "content": "image on average. Finally, although the BDD-Anomaly dataset treats three categories as anomalous,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 594, + 392, + 607 + ], + "spans": [ + { + "bbox": [ + 106, + 594, + 392, + 607 + ], + "score": 1.0, + "content": "compared to Lost and Found it has far more unique anomalous objects.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 33.5 + }, + { + "type": "text", + "bbox": [ + 107, + 611, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "score": 1.0, + "content": "The Fishyscapes benchmark for anomaly segmentation consists of cut-and-paste anomalies from", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 622, + 505, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 505, + 634 + ], + "score": 1.0, + "content": "out-of-distribution domains. This is problematic, because the anomalies stand out as clearly unnatural", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "in context. For instance, the orientation of anomalous objects is unnatural, and the lighting of the", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 644, + 505, + 657 + ], + "spans": [ + { + "bbox": [ + 106, + 644, + 505, + 657 + ], + "score": 1.0, + "content": "cut-and-paste patch differs from the lighting in the original image, providing an unnatural cue to", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "score": 1.0, + "content": "anomaly detectors that would not exist for real anomalies. Figure 7 shows an example of these", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "score": 1.0, + "content": "inconsistencies. Techniques for detecting image manipulation (Zhou et al., 2018; Johnson & Farid,", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 106, + 677, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 506, + 689 + ], + "score": 1.0, + "content": "2005) are competent at detecting artificial image elements of this kind. Our StreetHazards dataset", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 106, + 689, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 689, + 505, + 700 + ], + "score": 1.0, + "content": "overcomes these issues by leveraging a simulated driving environment to naturally insert anomalous", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 699, + 504, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 504, + 711 + ], + "score": 1.0, + "content": "3D models into a scene rather than overlaying 2D images. These anomalies are integrated into", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 106, + 710, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 722 + ], + "score": 1.0, + "content": "the scene with proper lighting and orientation, mimicking real-world anomalies and making them", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 721, + 254, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 254, + 733 + ], + "score": 1.0, + "content": "significantly more difficult to detect.", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 46 + } + ], + "page_idx": 3, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 106, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 759 + ], + "lines": [] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 79, + 505, + 127 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 79, + 505, + 127 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 79, + 505, + 127 + ], + "spans": [ + { + "bbox": [ + 106, + 79, + 505, + 127 + ], + "score": 0.966, + "html": "
FPR95↓AUROC↑AUPR↑
DinMSPDeVriesMaxLogitMSPDeVries MaxLogitMSPDeVriesMaxLogit
ImageNet44.246.035.884.676.987.238.230.545.8
Places36552.685.836.676.031.185.88.22.019.2
", + "type": "table", + "image_path": "bb7b3019a2b86e9009d5307a4d02ae9f36ffd04663404bc2ebffac2f51127748.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 79, + 505, + 95.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 95.0, + 505, + 111.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 111.0, + 505, + 127.0 + ], + "spans": [], + "index": 2 + } + ] + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 106, + 136, + 505, + 180 + ], + "lines": [ + { + "bbox": [ + 105, + 135, + 505, + 148 + ], + "spans": [ + { + "bbox": [ + 105, + 135, + 505, + 148 + ], + "score": 1.0, + "content": "Table 1: Multi-class out-of-distribution detection results using the maximum softmax probability", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 146, + 505, + 158 + ], + "spans": [ + { + "bbox": [ + 105, + 146, + 505, + 158 + ], + "score": 1.0, + "content": "(MSP) baseline (Hendrycks & Gimpel, 2017), the confidence branch detector of DeVries & Taylor", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 157, + 507, + 170 + ], + "spans": [ + { + "bbox": [ + 105, + 157, + 507, + 170 + ], + "score": 1.0, + "content": "(2018), and our maximum logit baseline. All values are percentages and average across five out-of-", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 168, + 469, + 181 + ], + "spans": [ + { + "bbox": [ + 105, + 168, + 469, + 181 + ], + "score": 1.0, + "content": "distribution test datasets. Full results on individual OOD test datasets are in the Appendix.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5, + "bbox_fs": [ + 105, + 135, + 507, + 181 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 214, + 505, + 291 + ], + "lines": [ + { + "bbox": [ + 105, + 214, + 505, + 227 + ], + "spans": [ + { + "bbox": [ + 105, + 214, + 505, + 227 + ], + "score": 1.0, + "content": "Anomaly Segmentation. Several prior works explore segmenting anomalous image regions. One", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 225, + 505, + 238 + ], + "spans": [ + { + "bbox": [ + 105, + 225, + 505, + 238 + ], + "score": 1.0, + "content": "line of work uses the WildDash dataset (Zendel et al., 2018), which contains numerous annotated", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 237, + 505, + 248 + ], + "spans": [ + { + "bbox": [ + 106, + 237, + 505, + 248 + ], + "score": 1.0, + "content": "driving scenes in conditions such as snow, fog, and rain. The WildDash test set contains fifteen", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 246, + 506, + 260 + ], + "spans": [ + { + "bbox": [ + 105, + 246, + 506, + 260 + ], + "score": 1.0, + "content": "“negative images” from different domains for which the goal is to mark the entire image as out-of-", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 258, + 505, + 271 + ], + "spans": [ + { + "bbox": [ + 106, + 258, + 505, + 271 + ], + "score": 1.0, + "content": "distribution. Thus, while the task is segmentation, the anomalies do not exist as objects within an", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 270, + 505, + 282 + ], + "spans": [ + { + "bbox": [ + 106, + 270, + 505, + 282 + ], + "score": 1.0, + "content": "otherwise in-distribution scene. This setting is similar to that explored by Hendrycks & Gimpel", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 280, + 463, + 293 + ], + "spans": [ + { + "bbox": [ + 105, + 280, + 463, + 293 + ], + "score": 1.0, + "content": "(2017), in which whole images from other datasets serve as out-of-distribution examples.", + "type": "text" + } + ], + "index": 13 + } + ], + "index": 10, + "bbox_fs": [ + 105, + 214, + 506, + 293 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 297, + 505, + 352 + ], + "lines": [ + { + "bbox": [ + 105, + 297, + 505, + 310 + ], + "spans": [ + { + "bbox": [ + 105, + 297, + 505, + 310 + ], + "score": 1.0, + "content": "To approach anomaly segmentation on WildDash, Krešo et al. (2018) train on multiple semantic", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 308, + 506, + 320 + ], + "spans": [ + { + "bbox": [ + 105, + 308, + 506, + 320 + ], + "score": 1.0, + "content": "segmentation domains and treat regions of images from the WildDash driving dataset as out-of-", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 320, + 504, + 331 + ], + "spans": [ + { + "bbox": [ + 106, + 320, + 504, + 331 + ], + "score": 1.0, + "content": "distribution if they are segmented as regions from different domains, i.e. indoor classes. Bevandic´", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 329, + 505, + 344 + ], + "spans": [ + { + "bbox": [ + 105, + 329, + 505, + 344 + ], + "score": 1.0, + "content": "et al. (2018) use ILSVRC 2012 images and train their network to segment the entirety of these images", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 342, + 195, + 353 + ], + "spans": [ + { + "bbox": [ + 105, + 342, + 195, + 353 + ], + "score": 1.0, + "content": "as out-of-distribution.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 16, + "bbox_fs": [ + 105, + 297, + 506, + 353 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 357, + 506, + 446 + ], + "lines": [ + { + "bbox": [ + 106, + 358, + 505, + 370 + ], + "spans": [ + { + "bbox": [ + 106, + 358, + 505, + 370 + ], + "score": 1.0, + "content": "In medical anomaly segmentation and product fault detection, anomalies are regions of otherwise", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 369, + 507, + 381 + ], + "spans": [ + { + "bbox": [ + 106, + 369, + 507, + 381 + ], + "score": 1.0, + "content": "in-distribution images. Baur et al. (2019) segment anomalous regions in brain MRIs using pixel-", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 380, + 505, + 393 + ], + "spans": [ + { + "bbox": [ + 105, + 380, + 505, + 393 + ], + "score": 1.0, + "content": "wise reconstruction loss. Similarly, Haselmann et al. (2018) perform product fault detection using", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 391, + 506, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 391, + 506, + 403 + ], + "score": 1.0, + "content": "pixel-wise reconstruction loss and introduce an expansive dataset for segmentation of product faults.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 402, + 506, + 415 + ], + "spans": [ + { + "bbox": [ + 105, + 402, + 506, + 415 + ], + "score": 1.0, + "content": "In these relatively simple domains, reconstruction-based approaches work well. In contrast to", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 412, + 506, + 425 + ], + "spans": [ + { + "bbox": [ + 105, + 412, + 506, + 425 + ], + "score": 1.0, + "content": "medical anomaly segmentation and fault detection, we consider complex images from street scenes.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 424, + 506, + 437 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 506, + 437 + ], + "score": 1.0, + "content": "These images have high variability in scene layout and lighting, and hence are less amenable to", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 435, + 239, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 239, + 447 + ], + "score": 1.0, + "content": "reconstruction-based techniques.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 22.5, + "bbox_fs": [ + 105, + 358, + 507, + 447 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 451, + 505, + 605 + ], + "lines": [ + { + "bbox": [ + 105, + 451, + 505, + 464 + ], + "spans": [ + { + "bbox": [ + 105, + 451, + 505, + 464 + ], + "score": 1.0, + "content": "The two works closest to our own are the Lost and Found (Pinggera et al., 2016) and Fishyscapes", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 104, + 461, + 506, + 477 + ], + "spans": [ + { + "bbox": [ + 104, + 461, + 506, + 477 + ], + "score": 1.0, + "content": "(Blum et al., 2019) datasets. The Lost and Found dataset consists of real images in a driving", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 473, + 506, + 486 + ], + "spans": [ + { + "bbox": [ + 105, + 473, + 506, + 486 + ], + "score": 1.0, + "content": "environment with small road hazards. The images were collected to mirror the Cityscapes dataset", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 484, + 506, + 497 + ], + "spans": [ + { + "bbox": [ + 105, + 484, + 506, + 497 + ], + "score": 1.0, + "content": "(Cordts et al., 2016) but are only collected from one city and so have less diversity. The dataset", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 496, + 506, + 507 + ], + "spans": [ + { + "bbox": [ + 106, + 496, + 506, + 507 + ], + "score": 1.0, + "content": "contains 35 unique anomalous objects, and methods are allowed to train on many of these. For", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 506, + 506, + 519 + ], + "spans": [ + { + "bbox": [ + 105, + 506, + 506, + 519 + ], + "score": 1.0, + "content": "Lost and Found, only nine unique objects are truly unseen at test time. Crucially, this is a different", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 518, + 506, + 530 + ], + "spans": [ + { + "bbox": [ + 106, + 518, + 506, + 530 + ], + "score": 1.0, + "content": "evaluation setting from our own, where anomalous objects are not revealed at training time, so their", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 529, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 106, + 529, + 506, + 540 + ], + "score": 1.0, + "content": "dataset is not directly comparable. Nevertheless, the BDD-Anomaly dataset fills several gaps in Lost", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 539, + 506, + 552 + ], + "spans": [ + { + "bbox": [ + 106, + 539, + 506, + 552 + ], + "score": 1.0, + "content": "and Found. First, the images are more diverse, because they are sourced from a more recent and", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 106, + 551, + 506, + 562 + ], + "spans": [ + { + "bbox": [ + 106, + 551, + 506, + 562 + ], + "score": 1.0, + "content": "comprehensive semantic segmentation dataset. Second, the anomalies are not restricted to small,", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 561, + 506, + 575 + ], + "spans": [ + { + "bbox": [ + 105, + 561, + 424, + 575 + ], + "score": 1.0, + "content": "sparse road hazards. Concretely, anomalous regions in Lost and Found take up", + "type": "text" + }, + { + "bbox": [ + 424, + 561, + 451, + 572 + ], + "score": 0.87, + "content": "0 . 1 1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 452, + 561, + 506, + 575 + ], + "score": 1.0, + "content": "of the image", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 572, + 505, + 585 + ], + "spans": [ + { + "bbox": [ + 105, + 572, + 452, + 585 + ], + "score": 1.0, + "content": "on average, whereas anomalous regions in the BDD-Anomaly dataset are larger and fill", + "type": "text" + }, + { + "bbox": [ + 452, + 572, + 479, + 583 + ], + "score": 0.87, + "content": "0 . 8 3 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 572, + 505, + 585 + ], + "score": 1.0, + "content": "of the", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 583, + 506, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 583, + 506, + 596 + ], + "score": 1.0, + "content": "image on average. Finally, although the BDD-Anomaly dataset treats three categories as anomalous,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 594, + 392, + 607 + ], + "spans": [ + { + "bbox": [ + 106, + 594, + 392, + 607 + ], + "score": 1.0, + "content": "compared to Lost and Found it has far more unique anomalous objects.", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 33.5, + "bbox_fs": [ + 104, + 451, + 506, + 607 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 611, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 106, + 611, + 505, + 623 + ], + "score": 1.0, + "content": "The Fishyscapes benchmark for anomaly segmentation consists of cut-and-paste anomalies from", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 622, + 505, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 505, + 634 + ], + "score": 1.0, + "content": "out-of-distribution domains. This is problematic, because the anomalies stand out as clearly unnatural", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "in context. For instance, the orientation of anomalous objects is unnatural, and the lighting of the", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 644, + 505, + 657 + ], + "spans": [ + { + "bbox": [ + 106, + 644, + 505, + 657 + ], + "score": 1.0, + "content": "cut-and-paste patch differs from the lighting in the original image, providing an unnatural cue to", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 655, + 506, + 667 + ], + "score": 1.0, + "content": "anomaly detectors that would not exist for real anomalies. Figure 7 shows an example of these", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 506, + 678 + ], + "score": 1.0, + "content": "inconsistencies. Techniques for detecting image manipulation (Zhou et al., 2018; Johnson & Farid,", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 106, + 677, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 677, + 506, + 689 + ], + "score": 1.0, + "content": "2005) are competent at detecting artificial image elements of this kind. Our StreetHazards dataset", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 106, + 689, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 689, + 505, + 700 + ], + "score": 1.0, + "content": "overcomes these issues by leveraging a simulated driving environment to naturally insert anomalous", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 699, + 504, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 504, + 711 + ], + "score": 1.0, + "content": "3D models into a scene rather than overlaying 2D images. These anomalies are integrated into", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 106, + 710, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 722 + ], + "score": 1.0, + "content": "the scene with proper lighting and orientation, mimicking real-world anomalies and making them", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 106, + 721, + 254, + 733 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 254, + 733 + ], + "score": 1.0, + "content": "significantly more difficult to detect.", + "type": "text" + } + ], + "index": 51 + } + ], + "index": 46, + "bbox_fs": [ + 105, + 611, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 104, + 78, + 505, + 269 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 104, + 78, + 505, + 269 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 104, + 78, + 505, + 269 + ], + "spans": [ + { + "bbox": [ + 104, + 78, + 505, + 269 + ], + "score": 0.962, + "type": "image", + "image_path": "c3e38c3fa532816e71e5ccfdee9b5bd80f517c0c2915dee5bd921821eb1372be.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 104, + 78, + 505, + 141.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 104, + 141.66666666666666, + 505, + 205.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 104, + 205.33333333333331, + 505, + 269.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 277, + 505, + 343 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 277, + 505, + 290 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 505, + 290 + ], + "score": 1.0, + "content": "Figure 3: Small-scale datasets such as CIFAR-10 have relatively disjoint classes, but larger-scale", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 289, + 505, + 300 + ], + "spans": [ + { + "bbox": [ + 106, + 289, + 505, + 300 + ], + "score": 1.0, + "content": "datasets including ImageNet-1K have several classes with high visual similarity to other classes. This", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 299, + 505, + 312 + ], + "spans": [ + { + "bbox": [ + 105, + 299, + 505, + 312 + ], + "score": 1.0, + "content": "implies that large-scale classifiers disperse probability mass among several classes. If the prediction", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 310, + 505, + 322 + ], + "spans": [ + { + "bbox": [ + 106, + 310, + 505, + 322 + ], + "score": 1.0, + "content": "confidence is used for out-of-distribution detection, then images which have similarities to other", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 322, + 505, + 333 + ], + "spans": [ + { + "bbox": [ + 106, + 322, + 505, + 333 + ], + "score": 1.0, + "content": "classes will often wrongly be deemed out-of-distribution due to low and dispersed confidence. This", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 333, + 317, + 344 + ], + "spans": [ + { + "bbox": [ + 106, + 333, + 317, + 344 + ], + "score": 1.0, + "content": "motivates our MaxLogit out-of-distribution detector.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 5.5 + } + ], + "index": 3.25 + }, + { + "type": "title", + "bbox": [ + 107, + 351, + 384, + 363 + ], + "lines": [ + { + "bbox": [ + 104, + 349, + 387, + 366 + ], + "spans": [ + { + "bbox": [ + 104, + 349, + 387, + 366 + ], + "score": 1.0, + "content": "3 MULTI-CLASS PREDICTION FOR OOD DETECTION", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 107, + 380, + 506, + 577 + ], + "lines": [ + { + "bbox": [ + 105, + 379, + 506, + 392 + ], + "spans": [ + { + "bbox": [ + 105, + 379, + 506, + 392 + ], + "score": 1.0, + "content": "Problem with existing baselines. Existing baselines for anomaly detection can work well in", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 391, + 506, + 402 + ], + "spans": [ + { + "bbox": [ + 106, + 391, + 506, + 402 + ], + "score": 1.0, + "content": "small-scale settings. However, in more realistic settings image classification networks are of-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 401, + 507, + 415 + ], + "spans": [ + { + "bbox": [ + 104, + 401, + 507, + 415 + ], + "score": 1.0, + "content": "ten tasked with distinguishing hundreds or thousands of classes, possibly with subtle differences.", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 412, + 506, + 425 + ], + "spans": [ + { + "bbox": [ + 106, + 412, + 506, + 425 + ], + "score": 1.0, + "content": "This is problematic for the maximum softmax probability (MSP) baseline (Hendrycks & Gim-", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 424, + 506, + 436 + ], + "spans": [ + { + "bbox": [ + 106, + 424, + 506, + 436 + ], + "score": 1.0, + "content": "pel, 2017), which uses the negative maximum softmax probability as the anomaly score, or", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 434, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 106, + 434, + 344, + 447 + ], + "score": 0.88, + "content": "\\begin{array} { r } { \\dot { { \\bf \\varphi } } - \\operatorname* { m a x } _ { k } \\exp f ( x ) _ { k } / \\sum _ { i } \\exp f ( \\bar { x _ { i } } ) _ { i } = - \\operatorname* { m a x } _ { k } \\hat { p } ( y = k \\mid x ) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 435, + 375, + 448 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 375, + 435, + 396, + 447 + ], + "score": 0.92, + "content": "f ( x )", + "type": "inline_equation" + }, + { + "bbox": [ + 396, + 435, + 506, + 448 + ], + "score": 1.0, + "content": "is the unnormalized logits", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 445, + 506, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 445, + 157, + 459 + ], + "score": 1.0, + "content": "of classifier", + "type": "text" + }, + { + "bbox": [ + 157, + 447, + 164, + 457 + ], + "score": 0.85, + "content": "f", + "type": "inline_equation" + }, + { + "bbox": [ + 165, + 445, + 203, + 459 + ], + "score": 1.0, + "content": "on input", + "type": "text" + }, + { + "bbox": [ + 203, + 448, + 210, + 456 + ], + "score": 0.71, + "content": "x", + "type": "inline_equation" + }, + { + "bbox": [ + 210, + 445, + 506, + 459 + ], + "score": 1.0, + "content": ". Classifiers tend to have higher confidence on in-distribution examples", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 456, + 505, + 469 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 505, + 469 + ], + "score": 1.0, + "content": "than out-of-distribution examples, enabling OOD detection. Assuming single-model evaluation", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 468, + 505, + 479 + ], + "spans": [ + { + "bbox": [ + 106, + 468, + 505, + 479 + ], + "score": 1.0, + "content": "and no access to other anomalies or test-time adaptation, the MSP attains state-of-the-art anomaly", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 479, + 505, + 490 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 505, + 490 + ], + "score": 1.0, + "content": "detection performance in small-scale settings. However, we show that the MSP is problematic for", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 489, + 506, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 506, + 502 + ], + "score": 1.0, + "content": "realistic in-distribution datasets with many classes, such as ImageNet and Places365 (Zhou et al.,", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 500, + 507, + 513 + ], + "spans": [ + { + "bbox": [ + 106, + 500, + 507, + 513 + ], + "score": 1.0, + "content": "2017). Probability mass can be dispersed among visually similar classes, as shown in Figure 3.", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 510, + 507, + 525 + ], + "spans": [ + { + "bbox": [ + 105, + 510, + 507, + 525 + ], + "score": 1.0, + "content": "Consequently, a classifier may produce a low confidence prediction for an in-distribution image,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 523, + 506, + 534 + ], + "spans": [ + { + "bbox": [ + 106, + 523, + 506, + 534 + ], + "score": 1.0, + "content": "not because the image is unfamiliar, but because the object’s exact class is difficult to determine.", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "score": 1.0, + "content": "To circumvent this problem, we propose using the negative of the maximum unnormalized logit", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 545, + 506, + 557 + ], + "spans": [ + { + "bbox": [ + 105, + 545, + 195, + 557 + ], + "score": 1.0, + "content": "for an anomaly score", + "type": "text" + }, + { + "bbox": [ + 195, + 545, + 255, + 556 + ], + "score": 0.68, + "content": "- \\operatorname* { m a x } _ { k } f ( x ) _ { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 255, + 545, + 506, + 557 + ], + "score": 1.0, + "content": ", which we call MaxLogit. Since the logits are unnormalized,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 555, + 506, + 568 + ], + "spans": [ + { + "bbox": [ + 106, + 555, + 506, + 568 + ], + "score": 1.0, + "content": "they are not affected by the number of classes and can serve as a better baseline for large-scale", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 566, + 223, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 566, + 223, + 578 + ], + "score": 1.0, + "content": "out-of-distribution detection.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 18.5 + }, + { + "type": "text", + "bbox": [ + 107, + 583, + 505, + 649 + ], + "lines": [ + { + "bbox": [ + 105, + 582, + 505, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 582, + 505, + 596 + ], + "score": 1.0, + "content": "The Species Out-Of-Distribution Dataset. To enable controlled experiments and high-quality", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 594, + 506, + 607 + ], + "spans": [ + { + "bbox": [ + 106, + 594, + 506, + 607 + ], + "score": 1.0, + "content": "evaluations of anomaly detectors in large-scale settings, we create the Species dataset, a new out-", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 604, + 506, + 618 + ], + "spans": [ + { + "bbox": [ + 105, + 604, + 506, + 618 + ], + "score": 1.0, + "content": "of-distribution test dataset that has no overlapping classes with ImageNet-21K. The Species dataset", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 617, + 505, + 628 + ], + "spans": [ + { + "bbox": [ + 106, + 617, + 505, + 628 + ], + "score": 1.0, + "content": "is comprised of images scraped from the iNaturalist website and contains hundreds of anomalous", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 627, + 507, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 507, + 640 + ], + "score": 1.0, + "content": "species grouped into seven high-level categories: Plants, Microorganisms, Amphibians, Protozoa,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 638, + 463, + 651 + ], + "spans": [ + { + "bbox": [ + 106, + 638, + 463, + 651 + ], + "score": 1.0, + "content": "Fungi, Arachnids, and Insects. Example images from the Species dataset are in Figure 2.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 30.5 + }, + { + "type": "text", + "bbox": [ + 107, + 654, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 654, + 505, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 505, + 668 + ], + "score": 1.0, + "content": "Setup. To evaluate the MSP baseline out-of-distribution detector and the MaxLogit detector, we", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 666, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 295, + 678 + ], + "score": 1.0, + "content": "use ImageNet-21K as the in-distribution dataset", + "type": "text" + }, + { + "bbox": [ + 296, + 666, + 310, + 677 + ], + "score": 0.89, + "content": "\\mathcal { D } _ { \\mathrm { i n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 311, + 666, + 506, + 678 + ], + "score": 1.0, + "content": ". To obtain representations for anomaly detection,", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 677, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 506, + 689 + ], + "score": 1.0, + "content": "we use models trained on ImageNet-21K-P, a cleaned version of ImageNet-21K with a train/val", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 688, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 506, + 700 + ], + "score": 1.0, + "content": "split (Ridnik et al., 2021a). We evaluate a TResNet-M, ViT-B-16, and Mixer-B-16 (Ridnik et al.,", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 697, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 697, + 506, + 712 + ], + "score": 1.0, + "content": "2021b; Dosovitskiy et al., 2021b; Tolstikhin et al., 2021), and the validation split is used for obtaining", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 340, + 723 + ], + "score": 1.0, + "content": "in-distribution scores. For out-of-distribution test datasets", + "type": "text" + }, + { + "bbox": [ + 340, + 710, + 358, + 721 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 358, + 709, + 506, + 723 + ], + "score": 1.0, + "content": ", we use categories from the Species", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 719, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 505, + 734 + ], + "score": 1.0, + "content": "dataset, all of which are unseen during training. Results for these experiments are in Table 2. We", + "type": "text" + } + ], + "index": 40 + } + ], + "index": 37 + } + ], + "page_idx": 4, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "score": 1.0, + "content": "5", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 104, + 78, + 505, + 269 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 104, + 78, + 505, + 269 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 104, + 78, + 505, + 269 + ], + "spans": [ + { + "bbox": [ + 104, + 78, + 505, + 269 + ], + "score": 0.962, + "type": "image", + "image_path": "c3e38c3fa532816e71e5ccfdee9b5bd80f517c0c2915dee5bd921821eb1372be.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 104, + 78, + 505, + 141.66666666666666 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 104, + 141.66666666666666, + 505, + 205.33333333333331 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 104, + 205.33333333333331, + 505, + 269.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 277, + 505, + 343 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 277, + 505, + 290 + ], + "spans": [ + { + "bbox": [ + 105, + 277, + 505, + 290 + ], + "score": 1.0, + "content": "Figure 3: Small-scale datasets such as CIFAR-10 have relatively disjoint classes, but larger-scale", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 289, + 505, + 300 + ], + "spans": [ + { + "bbox": [ + 106, + 289, + 505, + 300 + ], + "score": 1.0, + "content": "datasets including ImageNet-1K have several classes with high visual similarity to other classes. This", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 299, + 505, + 312 + ], + "spans": [ + { + "bbox": [ + 105, + 299, + 505, + 312 + ], + "score": 1.0, + "content": "implies that large-scale classifiers disperse probability mass among several classes. If the prediction", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 310, + 505, + 322 + ], + "spans": [ + { + "bbox": [ + 106, + 310, + 505, + 322 + ], + "score": 1.0, + "content": "confidence is used for out-of-distribution detection, then images which have similarities to other", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 322, + 505, + 333 + ], + "spans": [ + { + "bbox": [ + 106, + 322, + 505, + 333 + ], + "score": 1.0, + "content": "classes will often wrongly be deemed out-of-distribution due to low and dispersed confidence. This", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 333, + 317, + 344 + ], + "spans": [ + { + "bbox": [ + 106, + 333, + 317, + 344 + ], + "score": 1.0, + "content": "motivates our MaxLogit out-of-distribution detector.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 5.5 + } + ], + "index": 3.25 + }, + { + "type": "title", + "bbox": [ + 107, + 351, + 384, + 363 + ], + "lines": [ + { + "bbox": [ + 104, + 349, + 387, + 366 + ], + "spans": [ + { + "bbox": [ + 104, + 349, + 387, + 366 + ], + "score": 1.0, + "content": "3 MULTI-CLASS PREDICTION FOR OOD DETECTION", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 107, + 380, + 506, + 577 + ], + "lines": [ + { + "bbox": [ + 105, + 379, + 506, + 392 + ], + "spans": [ + { + "bbox": [ + 105, + 379, + 506, + 392 + ], + "score": 1.0, + "content": "Problem with existing baselines. Existing baselines for anomaly detection can work well in", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 391, + 506, + 402 + ], + "spans": [ + { + "bbox": [ + 106, + 391, + 506, + 402 + ], + "score": 1.0, + "content": "small-scale settings. However, in more realistic settings image classification networks are of-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 104, + 401, + 507, + 415 + ], + "spans": [ + { + "bbox": [ + 104, + 401, + 507, + 415 + ], + "score": 1.0, + "content": "ten tasked with distinguishing hundreds or thousands of classes, possibly with subtle differences.", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 412, + 506, + 425 + ], + "spans": [ + { + "bbox": [ + 106, + 412, + 506, + 425 + ], + "score": 1.0, + "content": "This is problematic for the maximum softmax probability (MSP) baseline (Hendrycks & Gim-", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 424, + 506, + 436 + ], + "spans": [ + { + "bbox": [ + 106, + 424, + 506, + 436 + ], + "score": 1.0, + "content": "pel, 2017), which uses the negative maximum softmax probability as the anomaly score, or", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 434, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 106, + 434, + 344, + 447 + ], + "score": 0.88, + "content": "\\begin{array} { r } { \\dot { { \\bf \\varphi } } - \\operatorname* { m a x } _ { k } \\exp f ( x ) _ { k } / \\sum _ { i } \\exp f ( \\bar { x _ { i } } ) _ { i } = - \\operatorname* { m a x } _ { k } \\hat { p } ( y = k \\mid x ) } \\end{array}", + "type": "inline_equation" + }, + { + "bbox": [ + 345, + 435, + 375, + 448 + ], + "score": 1.0, + "content": ", where", + "type": "text" + }, + { + "bbox": [ + 375, + 435, + 396, + 447 + ], + "score": 0.92, + "content": "f ( x )", + "type": "inline_equation" + }, + { + "bbox": [ + 396, + 435, + 506, + 448 + ], + "score": 1.0, + "content": "is the unnormalized logits", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 445, + 506, + 459 + ], + "spans": [ + { + "bbox": [ + 105, + 445, + 157, + 459 + ], + "score": 1.0, + "content": "of classifier", + "type": "text" + }, + { + "bbox": [ + 157, + 447, + 164, + 457 + ], + "score": 0.85, + "content": "f", + "type": "inline_equation" + }, + { + "bbox": [ + 165, + 445, + 203, + 459 + ], + "score": 1.0, + "content": "on input", + "type": "text" + }, + { + "bbox": [ + 203, + 448, + 210, + 456 + ], + "score": 0.71, + "content": "x", + "type": "inline_equation" + }, + { + "bbox": [ + 210, + 445, + 506, + 459 + ], + "score": 1.0, + "content": ". Classifiers tend to have higher confidence on in-distribution examples", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 456, + 505, + 469 + ], + "spans": [ + { + "bbox": [ + 105, + 456, + 505, + 469 + ], + "score": 1.0, + "content": "than out-of-distribution examples, enabling OOD detection. Assuming single-model evaluation", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 468, + 505, + 479 + ], + "spans": [ + { + "bbox": [ + 106, + 468, + 505, + 479 + ], + "score": 1.0, + "content": "and no access to other anomalies or test-time adaptation, the MSP attains state-of-the-art anomaly", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 479, + 505, + 490 + ], + "spans": [ + { + "bbox": [ + 106, + 479, + 505, + 490 + ], + "score": 1.0, + "content": "detection performance in small-scale settings. However, we show that the MSP is problematic for", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 106, + 489, + 506, + 502 + ], + "spans": [ + { + "bbox": [ + 106, + 489, + 506, + 502 + ], + "score": 1.0, + "content": "realistic in-distribution datasets with many classes, such as ImageNet and Places365 (Zhou et al.,", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 500, + 507, + 513 + ], + "spans": [ + { + "bbox": [ + 106, + 500, + 507, + 513 + ], + "score": 1.0, + "content": "2017). Probability mass can be dispersed among visually similar classes, as shown in Figure 3.", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 510, + 507, + 525 + ], + "spans": [ + { + "bbox": [ + 105, + 510, + 507, + 525 + ], + "score": 1.0, + "content": "Consequently, a classifier may produce a low confidence prediction for an in-distribution image,", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 523, + 506, + 534 + ], + "spans": [ + { + "bbox": [ + 106, + 523, + 506, + 534 + ], + "score": 1.0, + "content": "not because the image is unfamiliar, but because the object’s exact class is difficult to determine.", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "spans": [ + { + "bbox": [ + 106, + 533, + 506, + 546 + ], + "score": 1.0, + "content": "To circumvent this problem, we propose using the negative of the maximum unnormalized logit", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 545, + 506, + 557 + ], + "spans": [ + { + "bbox": [ + 105, + 545, + 195, + 557 + ], + "score": 1.0, + "content": "for an anomaly score", + "type": "text" + }, + { + "bbox": [ + 195, + 545, + 255, + 556 + ], + "score": 0.68, + "content": "- \\operatorname* { m a x } _ { k } f ( x ) _ { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 255, + 545, + 506, + 557 + ], + "score": 1.0, + "content": ", which we call MaxLogit. Since the logits are unnormalized,", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 555, + 506, + 568 + ], + "spans": [ + { + "bbox": [ + 106, + 555, + 506, + 568 + ], + "score": 1.0, + "content": "they are not affected by the number of classes and can serve as a better baseline for large-scale", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 566, + 223, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 566, + 223, + 578 + ], + "score": 1.0, + "content": "out-of-distribution detection.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 18.5, + "bbox_fs": [ + 104, + 379, + 507, + 578 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 583, + 505, + 649 + ], + "lines": [ + { + "bbox": [ + 105, + 582, + 505, + 596 + ], + "spans": [ + { + "bbox": [ + 105, + 582, + 505, + 596 + ], + "score": 1.0, + "content": "The Species Out-Of-Distribution Dataset. To enable controlled experiments and high-quality", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 594, + 506, + 607 + ], + "spans": [ + { + "bbox": [ + 106, + 594, + 506, + 607 + ], + "score": 1.0, + "content": "evaluations of anomaly detectors in large-scale settings, we create the Species dataset, a new out-", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 604, + 506, + 618 + ], + "spans": [ + { + "bbox": [ + 105, + 604, + 506, + 618 + ], + "score": 1.0, + "content": "of-distribution test dataset that has no overlapping classes with ImageNet-21K. The Species dataset", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 617, + 505, + 628 + ], + "spans": [ + { + "bbox": [ + 106, + 617, + 505, + 628 + ], + "score": 1.0, + "content": "is comprised of images scraped from the iNaturalist website and contains hundreds of anomalous", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 627, + 507, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 627, + 507, + 640 + ], + "score": 1.0, + "content": "species grouped into seven high-level categories: Plants, Microorganisms, Amphibians, Protozoa,", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 106, + 638, + 463, + 651 + ], + "spans": [ + { + "bbox": [ + 106, + 638, + 463, + 651 + ], + "score": 1.0, + "content": "Fungi, Arachnids, and Insects. Example images from the Species dataset are in Figure 2.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 30.5, + "bbox_fs": [ + 105, + 582, + 507, + 651 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 654, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 654, + 505, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 505, + 668 + ], + "score": 1.0, + "content": "Setup. To evaluate the MSP baseline out-of-distribution detector and the MaxLogit detector, we", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 666, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 295, + 678 + ], + "score": 1.0, + "content": "use ImageNet-21K as the in-distribution dataset", + "type": "text" + }, + { + "bbox": [ + 296, + 666, + 310, + 677 + ], + "score": 0.89, + "content": "\\mathcal { D } _ { \\mathrm { i n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 311, + 666, + 506, + 678 + ], + "score": 1.0, + "content": ". To obtain representations for anomaly detection,", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 677, + 506, + 689 + ], + "spans": [ + { + "bbox": [ + 105, + 677, + 506, + 689 + ], + "score": 1.0, + "content": "we use models trained on ImageNet-21K-P, a cleaned version of ImageNet-21K with a train/val", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 688, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 688, + 506, + 700 + ], + "score": 1.0, + "content": "split (Ridnik et al., 2021a). We evaluate a TResNet-M, ViT-B-16, and Mixer-B-16 (Ridnik et al.,", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 697, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 697, + 506, + 712 + ], + "score": 1.0, + "content": "2021b; Dosovitskiy et al., 2021b; Tolstikhin et al., 2021), and the validation split is used for obtaining", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 709, + 506, + 723 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 340, + 723 + ], + "score": 1.0, + "content": "in-distribution scores. For out-of-distribution test datasets", + "type": "text" + }, + { + "bbox": [ + 340, + 710, + 358, + 721 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 358, + 709, + 506, + 723 + ], + "score": 1.0, + "content": ", we use categories from the Species", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 719, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 105, + 719, + 505, + 734 + ], + "score": 1.0, + "content": "dataset, all of which are unseen during training. Results for these experiments are in Table 2. We", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 263, + 506, + 277 + ], + "spans": [ + { + "bbox": [ + 105, + 263, + 369, + 277 + ], + "score": 1.0, + "content": "also use ImageNet-1K and Places365 as in-distribution datasets", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 369, + 265, + 384, + 275 + ], + "score": 0.89, + "content": "\\mathcal { D } _ { \\mathrm { i n } }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 384, + 263, + 506, + 277 + ], + "score": 1.0, + "content": ", for which we use pretrained", + "type": "text", + "cross_page": true + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 275, + 505, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 275, + 372, + 287 + ], + "score": 1.0, + "content": "ResNet-50 models and use several out-of-distribution test datasets", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 373, + 276, + 391, + 286 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 392, + 275, + 505, + 287 + ], + "score": 1.0, + "content": ". Full results with ImageNet", + "type": "text", + "cross_page": true + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 285, + 320, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 285, + 320, + 298 + ], + "score": 1.0, + "content": "and Places365 as in-distribution are in the Appendix.", + "type": "text", + "cross_page": true + } + ], + "index": 10 + } + ], + "index": 37, + "bbox_fs": [ + 105, + 654, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 105, + 69, + 505, + 183 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 105, + 69, + 505, + 183 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 69, + 505, + 183 + ], + "spans": [ + { + "bbox": [ + 105, + 69, + 505, + 183 + ], + "score": 0.984, + "html": "
ResNetViTMLPMixer
DinDtest outMSPMaxLogitMSPMaxLogitMSPMaxLogit
Plants80.387.878.284.880.385.0
Microorganisms77.483.471.182.474.486.0
Amphibians41.848.641.948.844.451.7
Protozoa70.780.469.380.968.077.7
Fungi66.477.464.776.164.176.9
hetArachnids46.956.746.656.848.958.8
Insects47.656.448.054.648.653.8
Mean61.670.160.069.261.270.0
", + "type": "table", + "image_path": "35cd44e5505cbefb1abbe5d8609e00a686ece733beac6c36b934c08b7457e7a4.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 105, + 69, + 505, + 107.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 105, + 107.0, + 505, + 145.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 105, + 145.0, + 505, + 183.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 106, + 191, + 505, + 247 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 192, + 506, + 204 + ], + "spans": [ + { + "bbox": [ + 106, + 192, + 506, + 204 + ], + "score": 1.0, + "content": "Table 2: Results on Species. Models and the processed version of ImageNet-21K (ImageNet-21K-P)", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 202, + 506, + 216 + ], + "spans": [ + { + "bbox": [ + 106, + 202, + 506, + 216 + ], + "score": 1.0, + "content": "are from Ridnik et al. (2021a). All values are percent AUROC. Species enables evaluating anomaly de-", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 213, + 506, + 227 + ], + "spans": [ + { + "bbox": [ + 105, + 213, + 506, + 227 + ], + "score": 1.0, + "content": "tectors trained on ImageNet-21K and evades class overlap issues present in prior work. Using Species", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 224, + 505, + 237 + ], + "spans": [ + { + "bbox": [ + 105, + 224, + 505, + 237 + ], + "score": 1.0, + "content": "to conduct more controlled experiments without class overlap issues, we find that contrary to recent", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 236, + 505, + 248 + ], + "spans": [ + { + "bbox": [ + 106, + 236, + 505, + 248 + ], + "score": 1.0, + "content": "claims (Fort et al., 2021), simply scaling up Vision Transformers does not make OOD detection trivial.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5 + } + ], + "index": 3.0 + }, + { + "type": "text", + "bbox": [ + 107, + 264, + 505, + 298 + ], + "lines": [ + { + "bbox": [ + 105, + 263, + 506, + 277 + ], + "spans": [ + { + "bbox": [ + 105, + 263, + 369, + 277 + ], + "score": 1.0, + "content": "also use ImageNet-1K and Places365 as in-distribution datasets", + "type": "text" + }, + { + "bbox": [ + 369, + 265, + 384, + 275 + ], + "score": 0.89, + "content": "\\mathcal { D } _ { \\mathrm { i n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 384, + 263, + 506, + 277 + ], + "score": 1.0, + "content": ", for which we use pretrained", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 275, + 505, + 287 + ], + "spans": [ + { + "bbox": [ + 106, + 275, + 372, + 287 + ], + "score": 1.0, + "content": "ResNet-50 models and use several out-of-distribution test datasets", + "type": "text" + }, + { + "bbox": [ + 373, + 276, + 391, + 286 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 392, + 275, + 505, + 287 + ], + "score": 1.0, + "content": ". Full results with ImageNet", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 285, + 320, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 285, + 320, + 298 + ], + "score": 1.0, + "content": "and Places365 as in-distribution are in the Appendix.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 107, + 303, + 505, + 369 + ], + "lines": [ + { + "bbox": [ + 105, + 303, + 505, + 316 + ], + "spans": [ + { + "bbox": [ + 105, + 303, + 505, + 316 + ], + "score": 1.0, + "content": "Metrics. To evaluate out-of-distribution detectors in large-scale settings, we use three standard", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 314, + 505, + 326 + ], + "spans": [ + { + "bbox": [ + 106, + 315, + 484, + 326 + ], + "score": 1.0, + "content": "metrics of detection performance: area under the ROC curve (AUROC), false positive rate at", + "type": "text" + }, + { + "bbox": [ + 484, + 314, + 505, + 325 + ], + "score": 0.9, + "content": "9 5 \\%", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 325, + 505, + 338 + ], + "spans": [ + { + "bbox": [ + 105, + 325, + 505, + 338 + ], + "score": 1.0, + "content": "recall (FPR95), and area under the precision-recall curve (AUPR). The AUROC and AUPR are", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 335, + 506, + 349 + ], + "spans": [ + { + "bbox": [ + 105, + 335, + 506, + 349 + ], + "score": 1.0, + "content": "important metrics, because they give a holistic measure of performance when the cutoff for detecting", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 347, + 505, + 360 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 505, + 360 + ], + "score": 1.0, + "content": "anomalies is not a priori obvious or when we want to represent the performance of a detection", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 358, + 265, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 265, + 369 + ], + "score": 1.0, + "content": "method across several different cutoffs.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 13.5 + }, + { + "type": "text", + "bbox": [ + 107, + 374, + 505, + 441 + ], + "lines": [ + { + "bbox": [ + 105, + 375, + 505, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 505, + 387 + ], + "score": 1.0, + "content": "The AUROC can be thought of as the probability that an anomalous example is given a higher score", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 385, + 505, + 398 + ], + "spans": [ + { + "bbox": [ + 105, + 385, + 505, + 398 + ], + "score": 1.0, + "content": "than an ordinary example. Thus, a higher score is better, and an uninformative detector has a AUROC", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 397, + 505, + 410 + ], + "spans": [ + { + "bbox": [ + 105, + 397, + 117, + 410 + ], + "score": 1.0, + "content": "of", + "type": "text" + }, + { + "bbox": [ + 117, + 397, + 136, + 407 + ], + "score": 0.87, + "content": "50 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 397, + 505, + 410 + ], + "score": 1.0, + "content": ". AUPR provides a metric more attuned to class imbalances, which is relevant in anomaly and", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 408, + 505, + 420 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 505, + 420 + ], + "score": 1.0, + "content": "failure detection, when the number of anomalies or failures may be relatively small. Last, the FPR95", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 419, + 505, + 431 + ], + "spans": [ + { + "bbox": [ + 106, + 420, + 320, + 431 + ], + "score": 1.0, + "content": "metric consists of measuring the false positive rate at", + "type": "text" + }, + { + "bbox": [ + 320, + 419, + 340, + 429 + ], + "score": 0.88, + "content": "9 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 340, + 420, + 505, + 431 + ], + "score": 1.0, + "content": ". Since these measures are correlated, we", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 429, + 402, + 443 + ], + "spans": [ + { + "bbox": [ + 105, + 429, + 402, + 443 + ], + "score": 1.0, + "content": "occasionally solely present the AUROC for brevity and to preserve space.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 19.5 + }, + { + "type": "text", + "bbox": [ + 107, + 446, + 505, + 513 + ], + "lines": [ + { + "bbox": [ + 106, + 446, + 505, + 459 + ], + "spans": [ + { + "bbox": [ + 106, + 446, + 505, + 459 + ], + "score": 1.0, + "content": "Results. Results on Species are shown in Table 2. Results with ImageNet-1K and Places365 as", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 457, + 505, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 457, + 505, + 470 + ], + "score": 1.0, + "content": "in-distribution datasets are in Table 1. We find that the proposed MaxLogit method outperforms the", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 468, + 506, + 481 + ], + "spans": [ + { + "bbox": [ + 106, + 468, + 419, + 481 + ], + "score": 1.0, + "content": "maximum softmax probability baseline on all out-of-distribution test datasets", + "type": "text" + }, + { + "bbox": [ + 419, + 469, + 438, + 480 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 438, + 468, + 506, + 481 + ], + "score": 1.0, + "content": ". This holds true", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 480, + 506, + 491 + ], + "spans": [ + { + "bbox": [ + 106, + 480, + 506, + 491 + ], + "score": 1.0, + "content": "for all three models trained on ImageNet-21K. The MSP baseline is not much better than random and", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 490, + 506, + 502 + ], + "spans": [ + { + "bbox": [ + 105, + 490, + 506, + 502 + ], + "score": 1.0, + "content": "is has similar performance for all three model classes. This suggests that contrary to recent claims,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 502, + 487, + 513 + ], + "spans": [ + { + "bbox": [ + 106, + 502, + 487, + 513 + ], + "score": 1.0, + "content": "(Fort et al., 2021) simply scaling up Vision Transformers does not make OOD detection trivial.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 25.5 + }, + { + "type": "title", + "bbox": [ + 108, + 529, + 385, + 542 + ], + "lines": [ + { + "bbox": [ + 104, + 528, + 387, + 544 + ], + "spans": [ + { + "bbox": [ + 104, + 528, + 387, + 544 + ], + "score": 1.0, + "content": "4 MULTI-LABEL PREDICTION FOR OOD DETECTION", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 555, + 505, + 599 + ], + "lines": [ + { + "bbox": [ + 105, + 554, + 507, + 568 + ], + "spans": [ + { + "bbox": [ + 105, + 554, + 507, + 568 + ], + "score": 1.0, + "content": "Current work on out-of-distribution detection primarily considers multi-class or unsupervised settings.", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 565, + 506, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 506, + 578 + ], + "score": 1.0, + "content": "Yet as classifiers become more useful in realistic settings, the multi-label formulation becomes", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 577, + 506, + 590 + ], + "spans": [ + { + "bbox": [ + 105, + 577, + 506, + 590 + ], + "score": 1.0, + "content": "increasingly natural. To investigate out-of-distribution detection in multi-label settings, we provide a", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 587, + 228, + 601 + ], + "spans": [ + { + "bbox": [ + 105, + 587, + 228, + 601 + ], + "score": 1.0, + "content": "baseline and evaluation setup.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 107, + 604, + 505, + 671 + ], + "lines": [ + { + "bbox": [ + 105, + 605, + 507, + 618 + ], + "spans": [ + { + "bbox": [ + 105, + 605, + 507, + 618 + ], + "score": 1.0, + "content": "Setup. For multi-label classification we use PASCAL VOC (Everingham et al., 2009) and MS-", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 615, + 505, + 628 + ], + "spans": [ + { + "bbox": [ + 106, + 615, + 505, + 628 + ], + "score": 1.0, + "content": "COCO (Lin et al., 2014) as in-distribution data. To evaluate anomaly detectors for these in-distribution", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 626, + 505, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 505, + 640 + ], + "score": 1.0, + "content": "datasets, we use 20 out-of-distribution classes from ImageNet-21K. These classes have no overlap", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 637, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 505, + 650 + ], + "score": 1.0, + "content": "with ImageNet-1K, PASCAL VOC, or MS-COCO. The 20 classes are chosen not to overlap with", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 648, + 505, + 661 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 505, + 661 + ], + "score": 1.0, + "content": "ImageNet-1K since the multi-label classifiers models are pre-trained on ImageNet-1K. We list the", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 660, + 253, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 253, + 672 + ], + "score": 1.0, + "content": "class WordNet IDs in the Appendix.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 36.5 + }, + { + "type": "text", + "bbox": [ + 107, + 677, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "Methods. For our experiments, we use a ResNet-101 backbone architecture pre-trained on", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "ImageNet-1K. We replace the final layer with 2 fully connected layers and apply the logistic sigmoid", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "function for multi-label prediction. During training we freeze the batch normalization parameters due", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "to an insufficient number of images for proper mean and variance estimation. We train each model", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 465, + 733 + ], + "score": 1.0, + "content": "for 50 epochs using the Adam optimizer (Kingma & Ba, 2014) with hyperparameter values", + "type": "text" + }, + { + "bbox": [ + 465, + 720, + 487, + 731 + ], + "score": 0.9, + "content": "1 0 ^ { - 4 }", + "type": "inline_equation" + }, + { + "bbox": [ + 488, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "and", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 42 + } + ], + "page_idx": 5, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 106, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 760 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "score": 1.0, + "content": "6", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 105, + 69, + 505, + 183 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 105, + 69, + 505, + 183 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 69, + 505, + 183 + ], + "spans": [ + { + "bbox": [ + 105, + 69, + 505, + 183 + ], + "score": 0.984, + "html": "
ResNetViTMLPMixer
DinDtest outMSPMaxLogitMSPMaxLogitMSPMaxLogit
Plants80.387.878.284.880.385.0
Microorganisms77.483.471.182.474.486.0
Amphibians41.848.641.948.844.451.7
Protozoa70.780.469.380.968.077.7
Fungi66.477.464.776.164.176.9
hetArachnids46.956.746.656.848.958.8
Insects47.656.448.054.648.653.8
Mean61.670.160.069.261.270.0
", + "type": "table", + "image_path": "35cd44e5505cbefb1abbe5d8609e00a686ece733beac6c36b934c08b7457e7a4.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 105, + 69, + 505, + 107.0 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 105, + 107.0, + 505, + 145.0 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 105, + 145.0, + 505, + 183.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 106, + 191, + 505, + 247 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 192, + 506, + 204 + ], + "spans": [ + { + "bbox": [ + 106, + 192, + 506, + 204 + ], + "score": 1.0, + "content": "Table 2: Results on Species. Models and the processed version of ImageNet-21K (ImageNet-21K-P)", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 202, + 506, + 216 + ], + "spans": [ + { + "bbox": [ + 106, + 202, + 506, + 216 + ], + "score": 1.0, + "content": "are from Ridnik et al. (2021a). All values are percent AUROC. Species enables evaluating anomaly de-", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 213, + 506, + 227 + ], + "spans": [ + { + "bbox": [ + 105, + 213, + 506, + 227 + ], + "score": 1.0, + "content": "tectors trained on ImageNet-21K and evades class overlap issues present in prior work. Using Species", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 224, + 505, + 237 + ], + "spans": [ + { + "bbox": [ + 105, + 224, + 505, + 237 + ], + "score": 1.0, + "content": "to conduct more controlled experiments without class overlap issues, we find that contrary to recent", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 236, + 505, + 248 + ], + "spans": [ + { + "bbox": [ + 106, + 236, + 505, + 248 + ], + "score": 1.0, + "content": "claims (Fort et al., 2021), simply scaling up Vision Transformers does not make OOD detection trivial.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5 + } + ], + "index": 3.0 + }, + { + "type": "text", + "bbox": [ + 107, + 264, + 505, + 298 + ], + "lines": [], + "index": 9, + "bbox_fs": [ + 105, + 263, + 506, + 298 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 303, + 505, + 369 + ], + "lines": [ + { + "bbox": [ + 105, + 303, + 505, + 316 + ], + "spans": [ + { + "bbox": [ + 105, + 303, + 505, + 316 + ], + "score": 1.0, + "content": "Metrics. To evaluate out-of-distribution detectors in large-scale settings, we use three standard", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 314, + 505, + 326 + ], + "spans": [ + { + "bbox": [ + 106, + 315, + 484, + 326 + ], + "score": 1.0, + "content": "metrics of detection performance: area under the ROC curve (AUROC), false positive rate at", + "type": "text" + }, + { + "bbox": [ + 484, + 314, + 505, + 325 + ], + "score": 0.9, + "content": "9 5 \\%", + "type": "inline_equation" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 325, + 505, + 338 + ], + "spans": [ + { + "bbox": [ + 105, + 325, + 505, + 338 + ], + "score": 1.0, + "content": "recall (FPR95), and area under the precision-recall curve (AUPR). The AUROC and AUPR are", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 335, + 506, + 349 + ], + "spans": [ + { + "bbox": [ + 105, + 335, + 506, + 349 + ], + "score": 1.0, + "content": "important metrics, because they give a holistic measure of performance when the cutoff for detecting", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 347, + 505, + 360 + ], + "spans": [ + { + "bbox": [ + 105, + 347, + 505, + 360 + ], + "score": 1.0, + "content": "anomalies is not a priori obvious or when we want to represent the performance of a detection", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 358, + 265, + 369 + ], + "spans": [ + { + "bbox": [ + 105, + 358, + 265, + 369 + ], + "score": 1.0, + "content": "method across several different cutoffs.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 13.5, + "bbox_fs": [ + 105, + 303, + 506, + 369 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 374, + 505, + 441 + ], + "lines": [ + { + "bbox": [ + 105, + 375, + 505, + 387 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 505, + 387 + ], + "score": 1.0, + "content": "The AUROC can be thought of as the probability that an anomalous example is given a higher score", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 385, + 505, + 398 + ], + "spans": [ + { + "bbox": [ + 105, + 385, + 505, + 398 + ], + "score": 1.0, + "content": "than an ordinary example. Thus, a higher score is better, and an uninformative detector has a AUROC", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 397, + 505, + 410 + ], + "spans": [ + { + "bbox": [ + 105, + 397, + 117, + 410 + ], + "score": 1.0, + "content": "of", + "type": "text" + }, + { + "bbox": [ + 117, + 397, + 136, + 407 + ], + "score": 0.87, + "content": "50 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 137, + 397, + 505, + 410 + ], + "score": 1.0, + "content": ". AUPR provides a metric more attuned to class imbalances, which is relevant in anomaly and", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 408, + 505, + 420 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 505, + 420 + ], + "score": 1.0, + "content": "failure detection, when the number of anomalies or failures may be relatively small. Last, the FPR95", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 419, + 505, + 431 + ], + "spans": [ + { + "bbox": [ + 106, + 420, + 320, + 431 + ], + "score": 1.0, + "content": "metric consists of measuring the false positive rate at", + "type": "text" + }, + { + "bbox": [ + 320, + 419, + 340, + 429 + ], + "score": 0.88, + "content": "9 5 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 340, + 420, + 505, + 431 + ], + "score": 1.0, + "content": ". Since these measures are correlated, we", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 429, + 402, + 443 + ], + "spans": [ + { + "bbox": [ + 105, + 429, + 402, + 443 + ], + "score": 1.0, + "content": "occasionally solely present the AUROC for brevity and to preserve space.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 19.5, + "bbox_fs": [ + 105, + 375, + 505, + 443 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 446, + 505, + 513 + ], + "lines": [ + { + "bbox": [ + 106, + 446, + 505, + 459 + ], + "spans": [ + { + "bbox": [ + 106, + 446, + 505, + 459 + ], + "score": 1.0, + "content": "Results. Results on Species are shown in Table 2. Results with ImageNet-1K and Places365 as", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 457, + 505, + 470 + ], + "spans": [ + { + "bbox": [ + 105, + 457, + 505, + 470 + ], + "score": 1.0, + "content": "in-distribution datasets are in Table 1. We find that the proposed MaxLogit method outperforms the", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 468, + 506, + 481 + ], + "spans": [ + { + "bbox": [ + 106, + 468, + 419, + 481 + ], + "score": 1.0, + "content": "maximum softmax probability baseline on all out-of-distribution test datasets", + "type": "text" + }, + { + "bbox": [ + 419, + 469, + 438, + 480 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 438, + 468, + 506, + 481 + ], + "score": 1.0, + "content": ". This holds true", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 480, + 506, + 491 + ], + "spans": [ + { + "bbox": [ + 106, + 480, + 506, + 491 + ], + "score": 1.0, + "content": "for all three models trained on ImageNet-21K. The MSP baseline is not much better than random and", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 490, + 506, + 502 + ], + "spans": [ + { + "bbox": [ + 105, + 490, + 506, + 502 + ], + "score": 1.0, + "content": "is has similar performance for all three model classes. This suggests that contrary to recent claims,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 502, + 487, + 513 + ], + "spans": [ + { + "bbox": [ + 106, + 502, + 487, + 513 + ], + "score": 1.0, + "content": "(Fort et al., 2021) simply scaling up Vision Transformers does not make OOD detection trivial.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 25.5, + "bbox_fs": [ + 105, + 446, + 506, + 513 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 529, + 385, + 542 + ], + "lines": [ + { + "bbox": [ + 104, + 528, + 387, + 544 + ], + "spans": [ + { + "bbox": [ + 104, + 528, + 387, + 544 + ], + "score": 1.0, + "content": "4 MULTI-LABEL PREDICTION FOR OOD DETECTION", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 555, + 505, + 599 + ], + "lines": [ + { + "bbox": [ + 105, + 554, + 507, + 568 + ], + "spans": [ + { + "bbox": [ + 105, + 554, + 507, + 568 + ], + "score": 1.0, + "content": "Current work on out-of-distribution detection primarily considers multi-class or unsupervised settings.", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 565, + 506, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 506, + 578 + ], + "score": 1.0, + "content": "Yet as classifiers become more useful in realistic settings, the multi-label formulation becomes", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 577, + 506, + 590 + ], + "spans": [ + { + "bbox": [ + 105, + 577, + 506, + 590 + ], + "score": 1.0, + "content": "increasingly natural. To investigate out-of-distribution detection in multi-label settings, we provide a", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 587, + 228, + 601 + ], + "spans": [ + { + "bbox": [ + 105, + 587, + 228, + 601 + ], + "score": 1.0, + "content": "baseline and evaluation setup.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 31.5, + "bbox_fs": [ + 105, + 554, + 507, + 601 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 604, + 505, + 671 + ], + "lines": [ + { + "bbox": [ + 105, + 605, + 507, + 618 + ], + "spans": [ + { + "bbox": [ + 105, + 605, + 507, + 618 + ], + "score": 1.0, + "content": "Setup. For multi-label classification we use PASCAL VOC (Everingham et al., 2009) and MS-", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 615, + 505, + 628 + ], + "spans": [ + { + "bbox": [ + 106, + 615, + 505, + 628 + ], + "score": 1.0, + "content": "COCO (Lin et al., 2014) as in-distribution data. To evaluate anomaly detectors for these in-distribution", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 626, + 505, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 505, + 640 + ], + "score": 1.0, + "content": "datasets, we use 20 out-of-distribution classes from ImageNet-21K. These classes have no overlap", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 637, + 505, + 650 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 505, + 650 + ], + "score": 1.0, + "content": "with ImageNet-1K, PASCAL VOC, or MS-COCO. The 20 classes are chosen not to overlap with", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 648, + 505, + 661 + ], + "spans": [ + { + "bbox": [ + 105, + 648, + 505, + 661 + ], + "score": 1.0, + "content": "ImageNet-1K since the multi-label classifiers models are pre-trained on ImageNet-1K. We list the", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 660, + 253, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 253, + 672 + ], + "score": 1.0, + "content": "class WordNet IDs in the Appendix.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 36.5, + "bbox_fs": [ + 105, + 605, + 507, + 672 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 677, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "Methods. For our experiments, we use a ResNet-101 backbone architecture pre-trained on", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 700 + ], + "score": 1.0, + "content": "ImageNet-1K. We replace the final layer with 2 fully connected layers and apply the logistic sigmoid", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "function for multi-label prediction. During training we freeze the batch normalization parameters due", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "to an insufficient number of images for proper mean and variance estimation. We train each model", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 465, + 733 + ], + "score": 1.0, + "content": "for 50 epochs using the Adam optimizer (Kingma & Ba, 2014) with hyperparameter values", + "type": "text" + }, + { + "bbox": [ + 465, + 720, + 487, + 731 + ], + "score": 0.9, + "content": "1 0 ^ { - 4 }", + "type": "inline_equation" + }, + { + "bbox": [ + 488, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "and", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 42, + "bbox_fs": [ + 105, + 676, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 69, + 506, + 152 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 69, + 506, + 152 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 69, + 506, + 152 + ], + "spans": [ + { + "bbox": [ + 106, + 69, + 506, + 152 + ], + "score": 0.982, + "html": "
iForestLOFDropoutLogitAvgMSPMaxLogit
PASCAL VOCFPR9598.684.097.298.282.335.6
AUROC46.368.449.247.974.290.9
AUPR→↑↑37.158.445.341.365.581.2
COCOFPR95√95.678.493.394.581.840.4
AUROC个41.470.258.055.570.790.3
AUPR个63.782.076.374.082.994.0
", + "type": "table", + "image_path": "9bfc4b95d20326f3d05b9f023f7272b1c6d411057166e6c65d7ec8da4d54289b.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 69, + 506, + 96.66666666666667 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 96.66666666666667, + 506, + 124.33333333333334 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 124.33333333333334, + 506, + 152.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 106, + 159, + 505, + 204 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 160, + 505, + 172 + ], + "spans": [ + { + "bbox": [ + 105, + 160, + 505, + 172 + ], + "score": 1.0, + "content": "Table 3: Multi-label out-of-distribution detection comparison of the Isolation Forest (iForest), Local", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 170, + 506, + 184 + ], + "spans": [ + { + "bbox": [ + 105, + 170, + 506, + 184 + ], + "score": 1.0, + "content": "Outlier Factor (LOF), Dropout, logit average, maximum softmax probability, and maximum logit", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 182, + 505, + 194 + ], + "spans": [ + { + "bbox": [ + 106, + 182, + 505, + 194 + ], + "score": 1.0, + "content": "anomaly detectors on PASCAL VOC and MS-COCO. The same network architecture is used for all", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 192, + 308, + 206 + ], + "spans": [ + { + "bbox": [ + 105, + 192, + 308, + 206 + ], + "score": 1.0, + "content": "three detectors. All results shown are percentages.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "text", + "bbox": [ + 107, + 216, + 505, + 249 + ], + "lines": [ + { + "bbox": [ + 106, + 214, + 505, + 229 + ], + "spans": [ + { + "bbox": [ + 106, + 215, + 128, + 227 + ], + "score": 0.89, + "content": "1 0 ^ { - 5 }", + "type": "inline_equation" + }, + { + "bbox": [ + 128, + 214, + 143, + 229 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 143, + 216, + 154, + 227 + ], + "score": 0.89, + "content": "\\beta _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 214, + 172, + 229 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 172, + 216, + 184, + 227 + ], + "score": 0.89, + "content": "\\beta _ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 184, + 214, + 505, + 229 + ], + "score": 1.0, + "content": "respectively. For data augmentation we use standard resizing, random crops, and", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 227, + 504, + 239 + ], + "spans": [ + { + "bbox": [ + 105, + 227, + 254, + 239 + ], + "score": 1.0, + "content": "random flips to obtain images of size", + "type": "text" + }, + { + "bbox": [ + 254, + 227, + 315, + 238 + ], + "score": 0.91, + "content": "2 5 6 \\times 2 5 6 \\times 3", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 227, + 504, + 239 + ], + "score": 1.0, + "content": ". As a result of this training procedure, the mAP", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 237, + 409, + 250 + ], + "spans": [ + { + "bbox": [ + 105, + 237, + 266, + 250 + ], + "score": 1.0, + "content": "of the ResNet-101 on PASCAL VOC is", + "type": "text" + }, + { + "bbox": [ + 267, + 239, + 299, + 249 + ], + "score": 0.86, + "content": "8 9 . 1 1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 299, + 237, + 316, + 250 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 317, + 238, + 344, + 249 + ], + "score": 0.87, + "content": "7 2 . 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 344, + 237, + 409, + 250 + ], + "score": 1.0, + "content": "for MS-COCO.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 107, + 255, + 505, + 387 + ], + "lines": [ + { + "bbox": [ + 106, + 255, + 506, + 267 + ], + "spans": [ + { + "bbox": [ + 106, + 255, + 506, + 267 + ], + "score": 1.0, + "content": "As there has been little work on out-of-distribution detection in multilabel settings, we include", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 266, + 506, + 279 + ], + "spans": [ + { + "bbox": [ + 106, + 266, + 506, + 279 + ], + "score": 1.0, + "content": "comparisons to classic anomaly detectors for general settings. Isolation Forest, denoted by iForest,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 277, + 505, + 289 + ], + "spans": [ + { + "bbox": [ + 106, + 277, + 505, + 289 + ], + "score": 1.0, + "content": "works by randomly partitioning the space into half spaces to form a decision tree. The score is", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 104, + 286, + 506, + 302 + ], + "spans": [ + { + "bbox": [ + 104, + 286, + 506, + 302 + ], + "score": 1.0, + "content": "determined by how close a point is to the root of the tree. The local outlier factor (LOF) (Breunig", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 299, + 505, + 311 + ], + "spans": [ + { + "bbox": [ + 106, + 299, + 505, + 311 + ], + "score": 1.0, + "content": "et al., 2000) computes a local density ratio between every element and its neighbors. We set the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 309, + 505, + 322 + ], + "spans": [ + { + "bbox": [ + 105, + 309, + 505, + 322 + ], + "score": 1.0, + "content": "number of neighbors as 20. iForest and LOF are both computed on features from the penultimate", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 320, + 506, + 334 + ], + "spans": [ + { + "bbox": [ + 105, + 320, + 506, + 334 + ], + "score": 1.0, + "content": "layer of the networks. MSP denotes a natural extension of the maximum softmax probability detector", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 330, + 506, + 345 + ], + "spans": [ + { + "bbox": [ + 105, + 330, + 417, + 345 + ], + "score": 1.0, + "content": "in the multi-label setting, obtained by taking the sigmoid of each output score", + "type": "text" + }, + { + "bbox": [ + 418, + 332, + 442, + 344 + ], + "score": 0.93, + "content": "f ( { \\boldsymbol { x } } ) _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 442, + 330, + 506, + 345 + ], + "score": 1.0, + "content": "and computing", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 343, + 505, + 355 + ], + "spans": [ + { + "bbox": [ + 106, + 343, + 177, + 355 + ], + "score": 0.86, + "content": "- \\operatorname* { m a x } _ { i } \\sigma ( f ( x ) _ { i } )", + "type": "inline_equation" + }, + { + "bbox": [ + 178, + 343, + 505, + 355 + ], + "score": 1.0, + "content": ". Alternatively, one can average the logit values, denoted by LogitAvg. These serve", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 353, + 506, + 367 + ], + "spans": [ + { + "bbox": [ + 105, + 353, + 506, + 367 + ], + "score": 1.0, + "content": "as our baseline detectors for multi-label OOD detection. We compare these baselines to the MaxLogit", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 364, + 506, + 377 + ], + "spans": [ + { + "bbox": [ + 105, + 364, + 506, + 377 + ], + "score": 1.0, + "content": "detector that we introduce in Section 3. As in the multi-class case, the MaxLogit anomaly score for", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 375, + 277, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 216, + 388 + ], + "score": 1.0, + "content": "multi-label classification is", + "type": "text" + }, + { + "bbox": [ + 216, + 375, + 273, + 388 + ], + "score": 0.85, + "content": "- \\operatorname* { m a x } _ { i } f ( x ) _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 273, + 375, + 277, + 388 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 15.5 + }, + { + "type": "text", + "bbox": [ + 107, + 392, + 505, + 447 + ], + "lines": [ + { + "bbox": [ + 106, + 391, + 505, + 405 + ], + "spans": [ + { + "bbox": [ + 106, + 391, + 505, + 405 + ], + "score": 1.0, + "content": "Results. Results are shown in Table 3. We find that MaxLogit obtains the highest performance", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 403, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 105, + 403, + 505, + 416 + ], + "score": 1.0, + "content": "in all cases. MaxLogit bears similarity to the MSP baseline (Hendrycks & Gimpel, 2017) but is", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 414, + 505, + 427 + ], + "spans": [ + { + "bbox": [ + 105, + 414, + 505, + 427 + ], + "score": 1.0, + "content": "naturally applicable to multi-label problems. These results establish the MaxLogit as an effective and", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 425, + 505, + 438 + ], + "spans": [ + { + "bbox": [ + 105, + 425, + 505, + 438 + ], + "score": 1.0, + "content": "natural baseline for large-scale multi-label problems. Further, the evaluation setup enables future", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 435, + 357, + 449 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 357, + 449 + ], + "score": 1.0, + "content": "work in out-of-distribution detection with multi-label datasets.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 24 + }, + { + "type": "title", + "bbox": [ + 108, + 464, + 255, + 477 + ], + "lines": [ + { + "bbox": [ + 105, + 463, + 257, + 479 + ], + "spans": [ + { + "bbox": [ + 105, + 463, + 257, + 479 + ], + "score": 1.0, + "content": "5 THE CAOS BENCHMARK", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 106, + 489, + 505, + 577 + ], + "lines": [ + { + "bbox": [ + 105, + 489, + 507, + 502 + ], + "spans": [ + { + "bbox": [ + 105, + 489, + 507, + 502 + ], + "score": 1.0, + "content": "The Combined Anomalous Object Segmentation (CAOS) benchmark is comprised of two comple-", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 501, + 506, + 512 + ], + "spans": [ + { + "bbox": [ + 105, + 501, + 506, + 512 + ], + "score": 1.0, + "content": "mentary datasets for evaluating anomaly segmentation systems on diverse, realistic anomalies. First", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 511, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 105, + 511, + 505, + 524 + ], + "score": 1.0, + "content": "is the StreetHazards dataset, which leverages simulation to provide a large variety of anomalous", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 523, + 505, + 535 + ], + "spans": [ + { + "bbox": [ + 106, + 523, + 505, + 535 + ], + "score": 1.0, + "content": "objects realistically inserted into driving scenes. Second is the BDD-Anomaly dataset, which consists", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 532, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 105, + 532, + 505, + 546 + ], + "score": 1.0, + "content": "of real images taken from the BDD100K dataset (Yu et al., 2018). StreetHazards contains a highly", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 544, + 505, + 557 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 505, + 557 + ], + "score": 1.0, + "content": "diverse array of anomalies; BDD-Anomaly contains anomalies in real-world images. Together, these", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 555, + 506, + 568 + ], + "spans": [ + { + "bbox": [ + 106, + 555, + 506, + 568 + ], + "score": 1.0, + "content": "datasets allow researchers to judge techniques on their ability to segment diverse anomalies as well", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 566, + 383, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 567, + 288, + 578 + ], + "score": 1.0, + "content": "as anomalies in real images. All images have", + "type": "text" + }, + { + "bbox": [ + 289, + 566, + 337, + 577 + ], + "score": 0.9, + "content": "7 2 0 \\times 1 2 8 0", + "type": "inline_equation" + }, + { + "bbox": [ + 338, + 567, + 383, + 578 + ], + "score": 1.0, + "content": "resolution.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 107, + 582, + 505, + 682 + ], + "lines": [ + { + "bbox": [ + 106, + 582, + 505, + 595 + ], + "spans": [ + { + "bbox": [ + 106, + 582, + 505, + 595 + ], + "score": 1.0, + "content": "The StreetHazards Dataset. StreetHazards is an anomaly segmentation dataset that leverages", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 594, + 505, + 605 + ], + "spans": [ + { + "bbox": [ + 106, + 594, + 505, + 605 + ], + "score": 1.0, + "content": "simulation to provide diverse, realistically-inserted anomalous objects. To create the StreetHazards", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 604, + 506, + 618 + ], + "spans": [ + { + "bbox": [ + 105, + 604, + 506, + 618 + ], + "score": 1.0, + "content": "dataset, we use the Unreal Engine along with the CARLA simulation environment (Dosovitskiy et al.,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 614, + 505, + 629 + ], + "spans": [ + { + "bbox": [ + 105, + 614, + 505, + 629 + ], + "score": 1.0, + "content": "2017). From several months of development and testing including customization of the Unreal Engine", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 626, + 506, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 506, + 640 + ], + "score": 1.0, + "content": "and CARLA, we can insert foreign entities into a scene while having them be properly integrated.", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 637, + 507, + 652 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 507, + 652 + ], + "score": 1.0, + "content": "Unlike previous work, this avoids the issues of inconsistent chromatic aberration, inconsistent lighting,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 649, + 505, + 661 + ], + "spans": [ + { + "bbox": [ + 106, + 649, + 505, + 661 + ], + "score": 1.0, + "content": "edge effects, and other simple cues that an object is anomalous. Additionally, using a simulated", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 660, + 504, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 504, + 672 + ], + "score": 1.0, + "content": "environment allows us to dynamically insert diverse anomalous objects in any location and have them", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 671, + 502, + 684 + ], + "spans": [ + { + "bbox": [ + 106, + 671, + 502, + 684 + ], + "score": 1.0, + "content": "render properly with changes to lighting and weather including time of day, cloudy skies, and rain.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 40 + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "We use 3 towns from CARLA for training, from which we collect RGB images and their respective", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "score": 1.0, + "content": "semantic segmentation maps to serve as training data for semantic segmentation models. We generate", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "a validation set from the fourth town. Finally, we reserve the fifth and sixth town as our test", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 106, + 721, + 506, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 506, + 732 + ], + "score": 1.0, + "content": "set. We insert anomalies taken from the Digimation Model Bank Library and semantic ShapeNet", + "type": "text" + } + ], + "index": 48 + } + ], + "index": 46.5 + } + ], + "page_idx": 6, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 309, + 763 + ], + "score": 1.0, + "content": "7", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 69, + 506, + 152 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 69, + 506, + 152 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 69, + 506, + 152 + ], + "spans": [ + { + "bbox": [ + 106, + 69, + 506, + 152 + ], + "score": 0.982, + "html": "
iForestLOFDropoutLogitAvgMSPMaxLogit
PASCAL VOCFPR9598.684.097.298.282.335.6
AUROC46.368.449.247.974.290.9
AUPR→↑↑37.158.445.341.365.581.2
COCOFPR95√95.678.493.394.581.840.4
AUROC个41.470.258.055.570.790.3
AUPR个63.782.076.374.082.994.0
", + "type": "table", + "image_path": "9bfc4b95d20326f3d05b9f023f7272b1c6d411057166e6c65d7ec8da4d54289b.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 69, + 506, + 96.66666666666667 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 96.66666666666667, + 506, + 124.33333333333334 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 124.33333333333334, + 506, + 152.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 106, + 159, + 505, + 204 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 160, + 505, + 172 + ], + "spans": [ + { + "bbox": [ + 105, + 160, + 505, + 172 + ], + "score": 1.0, + "content": "Table 3: Multi-label out-of-distribution detection comparison of the Isolation Forest (iForest), Local", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 170, + 506, + 184 + ], + "spans": [ + { + "bbox": [ + 105, + 170, + 506, + 184 + ], + "score": 1.0, + "content": "Outlier Factor (LOF), Dropout, logit average, maximum softmax probability, and maximum logit", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 182, + 505, + 194 + ], + "spans": [ + { + "bbox": [ + 106, + 182, + 505, + 194 + ], + "score": 1.0, + "content": "anomaly detectors on PASCAL VOC and MS-COCO. The same network architecture is used for all", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 192, + 308, + 206 + ], + "spans": [ + { + "bbox": [ + 105, + 192, + 308, + 206 + ], + "score": 1.0, + "content": "three detectors. All results shown are percentages.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 4.5 + } + ], + "index": 2.75 + }, + { + "type": "text", + "bbox": [ + 107, + 216, + 505, + 249 + ], + "lines": [ + { + "bbox": [ + 106, + 214, + 505, + 229 + ], + "spans": [ + { + "bbox": [ + 106, + 215, + 128, + 227 + ], + "score": 0.89, + "content": "1 0 ^ { - 5 }", + "type": "inline_equation" + }, + { + "bbox": [ + 128, + 214, + 143, + 229 + ], + "score": 1.0, + "content": "for", + "type": "text" + }, + { + "bbox": [ + 143, + 216, + 154, + 227 + ], + "score": 0.89, + "content": "\\beta _ { 1 }", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 214, + 172, + 229 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 172, + 216, + 184, + 227 + ], + "score": 0.89, + "content": "\\beta _ { 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 184, + 214, + 505, + 229 + ], + "score": 1.0, + "content": "respectively. For data augmentation we use standard resizing, random crops, and", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 227, + 504, + 239 + ], + "spans": [ + { + "bbox": [ + 105, + 227, + 254, + 239 + ], + "score": 1.0, + "content": "random flips to obtain images of size", + "type": "text" + }, + { + "bbox": [ + 254, + 227, + 315, + 238 + ], + "score": 0.91, + "content": "2 5 6 \\times 2 5 6 \\times 3", + "type": "inline_equation" + }, + { + "bbox": [ + 315, + 227, + 504, + 239 + ], + "score": 1.0, + "content": ". As a result of this training procedure, the mAP", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 237, + 409, + 250 + ], + "spans": [ + { + "bbox": [ + 105, + 237, + 266, + 250 + ], + "score": 1.0, + "content": "of the ResNet-101 on PASCAL VOC is", + "type": "text" + }, + { + "bbox": [ + 267, + 239, + 299, + 249 + ], + "score": 0.86, + "content": "8 9 . 1 1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 299, + 237, + 316, + 250 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 317, + 238, + 344, + 249 + ], + "score": 0.87, + "content": "7 2 . 0 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 344, + 237, + 409, + 250 + ], + "score": 1.0, + "content": "for MS-COCO.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 8, + "bbox_fs": [ + 105, + 214, + 505, + 250 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 255, + 505, + 387 + ], + "lines": [ + { + "bbox": [ + 106, + 255, + 506, + 267 + ], + "spans": [ + { + "bbox": [ + 106, + 255, + 506, + 267 + ], + "score": 1.0, + "content": "As there has been little work on out-of-distribution detection in multilabel settings, we include", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 266, + 506, + 279 + ], + "spans": [ + { + "bbox": [ + 106, + 266, + 506, + 279 + ], + "score": 1.0, + "content": "comparisons to classic anomaly detectors for general settings. Isolation Forest, denoted by iForest,", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 106, + 277, + 505, + 289 + ], + "spans": [ + { + "bbox": [ + 106, + 277, + 505, + 289 + ], + "score": 1.0, + "content": "works by randomly partitioning the space into half spaces to form a decision tree. The score is", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 104, + 286, + 506, + 302 + ], + "spans": [ + { + "bbox": [ + 104, + 286, + 506, + 302 + ], + "score": 1.0, + "content": "determined by how close a point is to the root of the tree. The local outlier factor (LOF) (Breunig", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 299, + 505, + 311 + ], + "spans": [ + { + "bbox": [ + 106, + 299, + 505, + 311 + ], + "score": 1.0, + "content": "et al., 2000) computes a local density ratio between every element and its neighbors. We set the", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 309, + 505, + 322 + ], + "spans": [ + { + "bbox": [ + 105, + 309, + 505, + 322 + ], + "score": 1.0, + "content": "number of neighbors as 20. iForest and LOF are both computed on features from the penultimate", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 320, + 506, + 334 + ], + "spans": [ + { + "bbox": [ + 105, + 320, + 506, + 334 + ], + "score": 1.0, + "content": "layer of the networks. MSP denotes a natural extension of the maximum softmax probability detector", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 330, + 506, + 345 + ], + "spans": [ + { + "bbox": [ + 105, + 330, + 417, + 345 + ], + "score": 1.0, + "content": "in the multi-label setting, obtained by taking the sigmoid of each output score", + "type": "text" + }, + { + "bbox": [ + 418, + 332, + 442, + 344 + ], + "score": 0.93, + "content": "f ( { \\boldsymbol { x } } ) _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 442, + 330, + 506, + 345 + ], + "score": 1.0, + "content": "and computing", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 343, + 505, + 355 + ], + "spans": [ + { + "bbox": [ + 106, + 343, + 177, + 355 + ], + "score": 0.86, + "content": "- \\operatorname* { m a x } _ { i } \\sigma ( f ( x ) _ { i } )", + "type": "inline_equation" + }, + { + "bbox": [ + 178, + 343, + 505, + 355 + ], + "score": 1.0, + "content": ". Alternatively, one can average the logit values, denoted by LogitAvg. These serve", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 353, + 506, + 367 + ], + "spans": [ + { + "bbox": [ + 105, + 353, + 506, + 367 + ], + "score": 1.0, + "content": "as our baseline detectors for multi-label OOD detection. We compare these baselines to the MaxLogit", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 364, + 506, + 377 + ], + "spans": [ + { + "bbox": [ + 105, + 364, + 506, + 377 + ], + "score": 1.0, + "content": "detector that we introduce in Section 3. As in the multi-class case, the MaxLogit anomaly score for", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 375, + 277, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 216, + 388 + ], + "score": 1.0, + "content": "multi-label classification is", + "type": "text" + }, + { + "bbox": [ + 216, + 375, + 273, + 388 + ], + "score": 0.85, + "content": "- \\operatorname* { m a x } _ { i } f ( x ) _ { i }", + "type": "inline_equation" + }, + { + "bbox": [ + 273, + 375, + 277, + 388 + ], + "score": 1.0, + "content": ".", + "type": "text" + } + ], + "index": 21 + } + ], + "index": 15.5, + "bbox_fs": [ + 104, + 255, + 506, + 388 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 392, + 505, + 447 + ], + "lines": [ + { + "bbox": [ + 106, + 391, + 505, + 405 + ], + "spans": [ + { + "bbox": [ + 106, + 391, + 505, + 405 + ], + "score": 1.0, + "content": "Results. Results are shown in Table 3. We find that MaxLogit obtains the highest performance", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 403, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 105, + 403, + 505, + 416 + ], + "score": 1.0, + "content": "in all cases. MaxLogit bears similarity to the MSP baseline (Hendrycks & Gimpel, 2017) but is", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 414, + 505, + 427 + ], + "spans": [ + { + "bbox": [ + 105, + 414, + 505, + 427 + ], + "score": 1.0, + "content": "naturally applicable to multi-label problems. These results establish the MaxLogit as an effective and", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 425, + 505, + 438 + ], + "spans": [ + { + "bbox": [ + 105, + 425, + 505, + 438 + ], + "score": 1.0, + "content": "natural baseline for large-scale multi-label problems. Further, the evaluation setup enables future", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 435, + 357, + 449 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 357, + 449 + ], + "score": 1.0, + "content": "work in out-of-distribution detection with multi-label datasets.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 24, + "bbox_fs": [ + 105, + 391, + 505, + 449 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 464, + 255, + 477 + ], + "lines": [ + { + "bbox": [ + 105, + 463, + 257, + 479 + ], + "spans": [ + { + "bbox": [ + 105, + 463, + 257, + 479 + ], + "score": 1.0, + "content": "5 THE CAOS BENCHMARK", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 27 + }, + { + "type": "text", + "bbox": [ + 106, + 489, + 505, + 577 + ], + "lines": [ + { + "bbox": [ + 105, + 489, + 507, + 502 + ], + "spans": [ + { + "bbox": [ + 105, + 489, + 507, + 502 + ], + "score": 1.0, + "content": "The Combined Anomalous Object Segmentation (CAOS) benchmark is comprised of two comple-", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 501, + 506, + 512 + ], + "spans": [ + { + "bbox": [ + 105, + 501, + 506, + 512 + ], + "score": 1.0, + "content": "mentary datasets for evaluating anomaly segmentation systems on diverse, realistic anomalies. First", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 511, + 505, + 524 + ], + "spans": [ + { + "bbox": [ + 105, + 511, + 505, + 524 + ], + "score": 1.0, + "content": "is the StreetHazards dataset, which leverages simulation to provide a large variety of anomalous", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 523, + 505, + 535 + ], + "spans": [ + { + "bbox": [ + 106, + 523, + 505, + 535 + ], + "score": 1.0, + "content": "objects realistically inserted into driving scenes. Second is the BDD-Anomaly dataset, which consists", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 532, + 505, + 546 + ], + "spans": [ + { + "bbox": [ + 105, + 532, + 505, + 546 + ], + "score": 1.0, + "content": "of real images taken from the BDD100K dataset (Yu et al., 2018). StreetHazards contains a highly", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 544, + 505, + 557 + ], + "spans": [ + { + "bbox": [ + 105, + 544, + 505, + 557 + ], + "score": 1.0, + "content": "diverse array of anomalies; BDD-Anomaly contains anomalies in real-world images. Together, these", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 555, + 506, + 568 + ], + "spans": [ + { + "bbox": [ + 106, + 555, + 506, + 568 + ], + "score": 1.0, + "content": "datasets allow researchers to judge techniques on their ability to segment diverse anomalies as well", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 566, + 383, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 567, + 288, + 578 + ], + "score": 1.0, + "content": "as anomalies in real images. All images have", + "type": "text" + }, + { + "bbox": [ + 289, + 566, + 337, + 577 + ], + "score": 0.9, + "content": "7 2 0 \\times 1 2 8 0", + "type": "inline_equation" + }, + { + "bbox": [ + 338, + 567, + 383, + 578 + ], + "score": 1.0, + "content": "resolution.", + "type": "text" + } + ], + "index": 35 + } + ], + "index": 31.5, + "bbox_fs": [ + 105, + 489, + 507, + 578 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 582, + 505, + 682 + ], + "lines": [ + { + "bbox": [ + 106, + 582, + 505, + 595 + ], + "spans": [ + { + "bbox": [ + 106, + 582, + 505, + 595 + ], + "score": 1.0, + "content": "The StreetHazards Dataset. StreetHazards is an anomaly segmentation dataset that leverages", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 594, + 505, + 605 + ], + "spans": [ + { + "bbox": [ + 106, + 594, + 505, + 605 + ], + "score": 1.0, + "content": "simulation to provide diverse, realistically-inserted anomalous objects. To create the StreetHazards", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 604, + 506, + 618 + ], + "spans": [ + { + "bbox": [ + 105, + 604, + 506, + 618 + ], + "score": 1.0, + "content": "dataset, we use the Unreal Engine along with the CARLA simulation environment (Dosovitskiy et al.,", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 614, + 505, + 629 + ], + "spans": [ + { + "bbox": [ + 105, + 614, + 505, + 629 + ], + "score": 1.0, + "content": "2017). From several months of development and testing including customization of the Unreal Engine", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 626, + 506, + 640 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 506, + 640 + ], + "score": 1.0, + "content": "and CARLA, we can insert foreign entities into a scene while having them be properly integrated.", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 637, + 507, + 652 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 507, + 652 + ], + "score": 1.0, + "content": "Unlike previous work, this avoids the issues of inconsistent chromatic aberration, inconsistent lighting,", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 649, + 505, + 661 + ], + "spans": [ + { + "bbox": [ + 106, + 649, + 505, + 661 + ], + "score": 1.0, + "content": "edge effects, and other simple cues that an object is anomalous. Additionally, using a simulated", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 660, + 504, + 672 + ], + "spans": [ + { + "bbox": [ + 106, + 660, + 504, + 672 + ], + "score": 1.0, + "content": "environment allows us to dynamically insert diverse anomalous objects in any location and have them", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 106, + 671, + 502, + 684 + ], + "spans": [ + { + "bbox": [ + 106, + 671, + 502, + 684 + ], + "score": 1.0, + "content": "render properly with changes to lighting and weather including time of day, cloudy skies, and rain.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 40, + "bbox_fs": [ + 105, + 582, + 507, + 684 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 687, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "We use 3 towns from CARLA for training, from which we collect RGB images and their respective", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 506, + 711 + ], + "score": 1.0, + "content": "semantic segmentation maps to serve as training data for semantic segmentation models. We generate", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "a validation set from the fourth town. Finally, we reserve the fifth and sixth town as our test", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 106, + 721, + 506, + 732 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 506, + 732 + ], + "score": 1.0, + "content": "set. We insert anomalies taken from the Digimation Model Bank Library and semantic ShapeNet", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 106, + 454, + 505, + 466 + ], + "spans": [ + { + "bbox": [ + 106, + 454, + 505, + 466 + ], + "score": 1.0, + "content": "(ShapeNetSem) (Savva et al., 2015) into the test set in order to evaluate methods for out-of-distribution", + "type": "text", + "cross_page": true + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 465, + 506, + 477 + ], + "spans": [ + { + "bbox": [ + 105, + 465, + 506, + 477 + ], + "score": 1.0, + "content": "detection. In total, we use 250 unique anomaly models of diverse types. There are 12 classes used for", + "type": "text", + "cross_page": true + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 476, + 505, + 488 + ], + "spans": [ + { + "bbox": [ + 105, + 476, + 505, + 488 + ], + "score": 1.0, + "content": "training: background, road, street lines, traffic signs, sidewalk, pedestrian, vehicle, building, wall,", + "type": "text", + "cross_page": true + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 487, + 506, + 499 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 506, + 499 + ], + "score": 1.0, + "content": "pole, fence, and vegetation. The thirteenth class is the anomaly class that is only used at test time. We", + "type": "text", + "cross_page": true + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 498, + 505, + 509 + ], + "spans": [ + { + "bbox": [ + 106, + 498, + 505, + 509 + ], + "score": 1.0, + "content": "collect 5,125 image and semantic segmentation ground truth pairs for training, 1,031 pairs without", + "type": "text", + "cross_page": true + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 509, + 353, + 521 + ], + "spans": [ + { + "bbox": [ + 106, + 509, + 353, + 521 + ], + "score": 1.0, + "content": "anomalies for validation, and 1,500 test pairs with anomalies.", + "type": "text", + "cross_page": true + } + ], + "index": 15 + } + ], + "index": 46.5, + "bbox_fs": [ + 105, + 688, + 506, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 81, + 504, + 356 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 81, + 504, + 356 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 81, + 504, + 356 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 504, + 356 + ], + "score": 0.97, + "type": "image", + "image_path": "9f6a01add5da434aa775b1cb401e714b45aeff0bd523f7abf8d24e9f6ac738fc.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 81, + 504, + 172.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 172.66666666666669, + 504, + 264.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 264.33333333333337, + 504, + 356.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 365, + 506, + 442 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 365, + 506, + 377 + ], + "spans": [ + { + "bbox": [ + 106, + 365, + 506, + 377 + ], + "score": 1.0, + "content": "Figure 4: A sample of anomalous scenes from the CAOS benchmark with model predictions and", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 376, + 506, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 376, + 360, + 388 + ], + "score": 1.0, + "content": "anomaly scores. The anomaly scores are thresholded to the top", + "type": "text" + }, + { + "bbox": [ + 360, + 376, + 379, + 386 + ], + "score": 0.86, + "content": "10 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 380, + 376, + 506, + 388 + ], + "score": 1.0, + "content": "of values for visualization. GT", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 387, + 506, + 399 + ], + "spans": [ + { + "bbox": [ + 105, + 387, + 506, + 399 + ], + "score": 1.0, + "content": "is ground truth, the autoencoder model is based on the spatial autoencoder used in Baur et al. (2019),", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 397, + 506, + 410 + ], + "spans": [ + { + "bbox": [ + 105, + 397, + 506, + 410 + ], + "score": 1.0, + "content": "MSP is the maximum softmax probability baseline (Hendrycks & Gimpel, 2017), and MaxLogit", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 408, + 505, + 421 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 505, + 421 + ], + "score": 1.0, + "content": "is the method we propose as a new baseline for large-scale settings. Compared to baselines, the", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 419, + 506, + 432 + ], + "spans": [ + { + "bbox": [ + 105, + 419, + 506, + 432 + ], + "score": 1.0, + "content": "MaxLogit detector places lower scores on in-distribution image regions, including object outlines,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 430, + 363, + 443 + ], + "spans": [ + { + "bbox": [ + 106, + 430, + 363, + 443 + ], + "score": 1.0, + "content": "while also doing a better job of highlighting anomalous objects.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 6 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 107, + 453, + 505, + 520 + ], + "lines": [ + { + "bbox": [ + 106, + 454, + 505, + 466 + ], + "spans": [ + { + "bbox": [ + 106, + 454, + 505, + 466 + ], + "score": 1.0, + "content": "(ShapeNetSem) (Savva et al., 2015) into the test set in order to evaluate methods for out-of-distribution", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 105, + 465, + 506, + 477 + ], + "spans": [ + { + "bbox": [ + 105, + 465, + 506, + 477 + ], + "score": 1.0, + "content": "detection. In total, we use 250 unique anomaly models of diverse types. There are 12 classes used for", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 476, + 505, + 488 + ], + "spans": [ + { + "bbox": [ + 105, + 476, + 505, + 488 + ], + "score": 1.0, + "content": "training: background, road, street lines, traffic signs, sidewalk, pedestrian, vehicle, building, wall,", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 105, + 487, + 506, + 499 + ], + "spans": [ + { + "bbox": [ + 105, + 487, + 506, + 499 + ], + "score": 1.0, + "content": "pole, fence, and vegetation. The thirteenth class is the anomaly class that is only used at test time. We", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 106, + 498, + 505, + 509 + ], + "spans": [ + { + "bbox": [ + 106, + 498, + 505, + 509 + ], + "score": 1.0, + "content": "collect 5,125 image and semantic segmentation ground truth pairs for training, 1,031 pairs without", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 509, + 353, + 521 + ], + "spans": [ + { + "bbox": [ + 106, + 509, + 353, + 521 + ], + "score": 1.0, + "content": "anomalies for validation, and 1,500 test pairs with anomalies.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 12.5 + }, + { + "type": "text", + "bbox": [ + 107, + 525, + 505, + 603 + ], + "lines": [ + { + "bbox": [ + 105, + 524, + 505, + 539 + ], + "spans": [ + { + "bbox": [ + 105, + 524, + 505, + 539 + ], + "score": 1.0, + "content": "The BDD-Anomaly Dataset. BDD-Anomaly is an anomaly segmentation dataset with real images", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 537, + 505, + 549 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 549 + ], + "score": 1.0, + "content": "in diverse conditions. We source BDD-Anomaly from BDD100K (Yu et al., 2018), a large-scale", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 548, + 505, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 548, + 505, + 560 + ], + "score": 1.0, + "content": "semantic segmentation dataset with diverse driving conditions. The original data consists in 7,000", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 559, + 506, + 571 + ], + "spans": [ + { + "bbox": [ + 105, + 559, + 506, + 571 + ], + "score": 1.0, + "content": "images for training and 1,000 for validation. There are 18 original classes. We choose motorcycle,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 569, + 505, + 582 + ], + "spans": [ + { + "bbox": [ + 105, + 569, + 505, + 582 + ], + "score": 1.0, + "content": "train, and bicycle as the anomalous object classes and remove all images with these objects from the", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 580, + 507, + 594 + ], + "spans": [ + { + "bbox": [ + 105, + 580, + 507, + 594 + ], + "score": 1.0, + "content": "training and validation sets. This yields 6,280 training pairs, 910 validation pairs without anomalies,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 592, + 291, + 604 + ], + "spans": [ + { + "bbox": [ + 106, + 592, + 291, + 604 + ], + "score": 1.0, + "content": "and 810 testing pairs with anomalous objects.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 19 + }, + { + "type": "title", + "bbox": [ + 107, + 617, + 193, + 628 + ], + "lines": [ + { + "bbox": [ + 105, + 616, + 194, + 629 + ], + "spans": [ + { + "bbox": [ + 105, + 616, + 194, + 629 + ], + "score": 1.0, + "content": "5.1 EXPERIMENTS", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 23 + }, + { + "type": "text", + "bbox": [ + 107, + 638, + 504, + 671 + ], + "lines": [ + { + "bbox": [ + 105, + 637, + 505, + 651 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 505, + 651 + ], + "score": 1.0, + "content": "Evaluation. In anomaly segmentation experiments, each pixel is treated as a prediction, resulting", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 649, + 505, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 505, + 662 + ], + "score": 1.0, + "content": "in many predictions to evaluate. To fit these in memory, we compute the metrics on each image and", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 660, + 293, + 672 + ], + "spans": [ + { + "bbox": [ + 105, + 660, + 293, + 672 + ], + "score": 1.0, + "content": "average over the images to obtain final values.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 107, + 676, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "Methods. Our first baseline is pixel-wise Maximum Softmax Probability (MSP). Introduced by", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "Hendrycks & Gimpel (2017) for multi-class out-of-distribution detection, we directly port this", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "baseline to anomaly segmentation. Alternatively, the background class might serve as an anomaly", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "detector, because it contains everything not in the other classes. To test this hypothesis, “Background”", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "uses the posterior probability of the background class as the anomaly score. The Dropout method", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 29 + } + ], + "page_idx": 7, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 752, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 750, + 309, + 761 + ], + "spans": [ + { + "bbox": [ + 302, + 750, + 309, + 761 + ], + "score": 1.0, + "content": "8", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 191, + 69, + 420, + 80 + ], + "lines": [ + { + "bbox": [ + 191, + 69, + 420, + 82 + ], + "spans": [ + { + "bbox": [ + 191, + 69, + 420, + 82 + ], + "score": 1.0, + "content": "Examples and Predictions for Our StreetHazards Dataset", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "image", + "bbox": [ + 106, + 81, + 504, + 356 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 106, + 81, + 504, + 356 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 81, + 504, + 356 + ], + "spans": [ + { + "bbox": [ + 106, + 81, + 504, + 356 + ], + "score": 0.97, + "type": "image", + "image_path": "9f6a01add5da434aa775b1cb401e714b45aeff0bd523f7abf8d24e9f6ac738fc.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 81, + 504, + 172.66666666666669 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 172.66666666666669, + 504, + 264.33333333333337 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 264.33333333333337, + 504, + 356.00000000000006 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 106, + 365, + 506, + 442 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 365, + 506, + 377 + ], + "spans": [ + { + "bbox": [ + 106, + 365, + 506, + 377 + ], + "score": 1.0, + "content": "Figure 4: A sample of anomalous scenes from the CAOS benchmark with model predictions and", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 376, + 506, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 376, + 360, + 388 + ], + "score": 1.0, + "content": "anomaly scores. The anomaly scores are thresholded to the top", + "type": "text" + }, + { + "bbox": [ + 360, + 376, + 379, + 386 + ], + "score": 0.86, + "content": "10 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 380, + 376, + 506, + 388 + ], + "score": 1.0, + "content": "of values for visualization. GT", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 387, + 506, + 399 + ], + "spans": [ + { + "bbox": [ + 105, + 387, + 506, + 399 + ], + "score": 1.0, + "content": "is ground truth, the autoencoder model is based on the spatial autoencoder used in Baur et al. (2019),", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 397, + 506, + 410 + ], + "spans": [ + { + "bbox": [ + 105, + 397, + 506, + 410 + ], + "score": 1.0, + "content": "MSP is the maximum softmax probability baseline (Hendrycks & Gimpel, 2017), and MaxLogit", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 408, + 505, + 421 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 505, + 421 + ], + "score": 1.0, + "content": "is the method we propose as a new baseline for large-scale settings. Compared to baselines, the", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 419, + 506, + 432 + ], + "spans": [ + { + "bbox": [ + 105, + 419, + 506, + 432 + ], + "score": 1.0, + "content": "MaxLogit detector places lower scores on in-distribution image regions, including object outlines,", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 430, + 363, + 443 + ], + "spans": [ + { + "bbox": [ + 106, + 430, + 363, + 443 + ], + "score": 1.0, + "content": "while also doing a better job of highlighting anomalous objects.", + "type": "text" + } + ], + "index": 9 + } + ], + "index": 6 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 107, + 453, + 505, + 520 + ], + "lines": [], + "index": 12.5, + "bbox_fs": [ + 105, + 454, + 506, + 521 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 525, + 505, + 603 + ], + "lines": [ + { + "bbox": [ + 105, + 524, + 505, + 539 + ], + "spans": [ + { + "bbox": [ + 105, + 524, + 505, + 539 + ], + "score": 1.0, + "content": "The BDD-Anomaly Dataset. BDD-Anomaly is an anomaly segmentation dataset with real images", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 537, + 505, + 549 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 549 + ], + "score": 1.0, + "content": "in diverse conditions. We source BDD-Anomaly from BDD100K (Yu et al., 2018), a large-scale", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 548, + 505, + 560 + ], + "spans": [ + { + "bbox": [ + 105, + 548, + 505, + 560 + ], + "score": 1.0, + "content": "semantic segmentation dataset with diverse driving conditions. The original data consists in 7,000", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 105, + 559, + 506, + 571 + ], + "spans": [ + { + "bbox": [ + 105, + 559, + 506, + 571 + ], + "score": 1.0, + "content": "images for training and 1,000 for validation. There are 18 original classes. We choose motorcycle,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 569, + 505, + 582 + ], + "spans": [ + { + "bbox": [ + 105, + 569, + 505, + 582 + ], + "score": 1.0, + "content": "train, and bicycle as the anomalous object classes and remove all images with these objects from the", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 105, + 580, + 507, + 594 + ], + "spans": [ + { + "bbox": [ + 105, + 580, + 507, + 594 + ], + "score": 1.0, + "content": "training and validation sets. This yields 6,280 training pairs, 910 validation pairs without anomalies,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 592, + 291, + 604 + ], + "spans": [ + { + "bbox": [ + 106, + 592, + 291, + 604 + ], + "score": 1.0, + "content": "and 810 testing pairs with anomalous objects.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 19, + "bbox_fs": [ + 105, + 524, + 507, + 604 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 617, + 193, + 628 + ], + "lines": [ + { + "bbox": [ + 105, + 616, + 194, + 629 + ], + "spans": [ + { + "bbox": [ + 105, + 616, + 194, + 629 + ], + "score": 1.0, + "content": "5.1 EXPERIMENTS", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 23 + }, + { + "type": "text", + "bbox": [ + 107, + 638, + 504, + 671 + ], + "lines": [ + { + "bbox": [ + 105, + 637, + 505, + 651 + ], + "spans": [ + { + "bbox": [ + 105, + 637, + 505, + 651 + ], + "score": 1.0, + "content": "Evaluation. In anomaly segmentation experiments, each pixel is treated as a prediction, resulting", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 649, + 505, + 662 + ], + "spans": [ + { + "bbox": [ + 105, + 649, + 505, + 662 + ], + "score": 1.0, + "content": "in many predictions to evaluate. To fit these in memory, we compute the metrics on each image and", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 105, + 660, + 293, + 672 + ], + "spans": [ + { + "bbox": [ + 105, + 660, + 293, + 672 + ], + "score": 1.0, + "content": "average over the images to obtain final values.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25, + "bbox_fs": [ + 105, + 637, + 505, + 672 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 676, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 505, + 690 + ], + "score": 1.0, + "content": "Methods. Our first baseline is pixel-wise Maximum Softmax Probability (MSP). Introduced by", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 106, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "Hendrycks & Gimpel (2017) for multi-class out-of-distribution detection, we directly port this", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 106, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "baseline to anomaly segmentation. Alternatively, the background class might serve as an anomaly", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "score": 1.0, + "content": "detector, because it contains everything not in the other classes. To test this hypothesis, “Background”", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 720, + 505, + 734 + ], + "spans": [ + { + "bbox": [ + 106, + 720, + 505, + 734 + ], + "score": 1.0, + "content": "uses the posterior probability of the background class as the anomaly score. The Dropout method", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 105, + 203, + 507, + 218 + ], + "spans": [ + { + "bbox": [ + 105, + 203, + 507, + 218 + ], + "score": 1.0, + "content": "leverages MC Dropout (Gal & Ghahramani, 2016) to obtain an epistemic uncertainty estimate.", + "type": "text", + "cross_page": true + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 214, + 506, + 228 + ], + "spans": [ + { + "bbox": [ + 105, + 214, + 506, + 228 + ], + "score": 1.0, + "content": "Following Kendall et al. (2015), we compute the pixel-wise posterior variance over multiple dropout", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 226, + 505, + 238 + ], + "spans": [ + { + "bbox": [ + 105, + 226, + 505, + 238 + ], + "score": 1.0, + "content": "masks and average across all classes, which serves as the anomaly score. We also experiment with", + "type": "text", + "cross_page": true + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 237, + 506, + 249 + ], + "spans": [ + { + "bbox": [ + 105, + 237, + 506, + 249 + ], + "score": 1.0, + "content": "an autoencoder baseline similar to Baur et al. (2019); Haselmann et al. (2018) where pixel-wise", + "type": "text", + "cross_page": true + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 248, + 506, + 260 + ], + "spans": [ + { + "bbox": [ + 105, + 248, + 506, + 260 + ], + "score": 1.0, + "content": "reconstruction loss is used as the anomaly score. This method is called AE. The “Branch” method is", + "type": "text", + "cross_page": true + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 259, + 506, + 271 + ], + "spans": [ + { + "bbox": [ + 106, + 259, + 506, + 271 + ], + "score": 1.0, + "content": "a direct port of the confidence branch detector from DeVries & Taylor (2018) to pixel-wise prediction.", + "type": "text", + "cross_page": true + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 270, + 488, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 270, + 488, + 282 + ], + "score": 1.0, + "content": "Finally, we use the MaxLogit method described in earlier sections independently on each pixel.", + "type": "text", + "cross_page": true + } + ], + "index": 12 + } + ], + "index": 29, + "bbox_fs": [ + 105, + 676, + 506, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 64, + 506, + 147 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 64, + 506, + 147 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 64, + 506, + 147 + ], + "spans": [ + { + "bbox": [ + 106, + 64, + 506, + 147 + ], + "score": 0.979, + "html": "
MSPBranchBackgroundDropoutAEMaxLogit
StreetHazardsFPR9533.768.469.079.491.726.5
AUROC87.765.758.669.966.189.3
↓↑ AUPR 个6.61.54.57.52.210.6
BDD-Anomaly AUROCFPR9524.525.640.116.674.114.0
87.785.669.790.864.092.6
AUPR3.73.91.14.30.75.4
", + "type": "table", + "image_path": "a39f779b992a3a5f4420e180e8d83132c4aaf76e85751c0c198cb34b68e5580a.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 64, + 506, + 91.66666666666667 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 91.66666666666667, + 506, + 119.33333333333334 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 119.33333333333334, + 506, + 147.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 106, + 155, + 506, + 188 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 154, + 505, + 167 + ], + "spans": [ + { + "bbox": [ + 105, + 154, + 505, + 167 + ], + "score": 1.0, + "content": "Table 4: Results on the CAOS benchmark. AUPR is low across the board due to the large class", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 165, + 506, + 178 + ], + "spans": [ + { + "bbox": [ + 105, + 165, + 506, + 178 + ], + "score": 1.0, + "content": "imbalance, but all methods perform substantially better than chance. MaxLogit obtains the best", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 177, + 272, + 190 + ], + "spans": [ + { + "bbox": [ + 105, + 177, + 272, + 190 + ], + "score": 1.0, + "content": "performance. All results are percentages.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 107, + 204, + 505, + 281 + ], + "lines": [ + { + "bbox": [ + 105, + 203, + 507, + 218 + ], + "spans": [ + { + "bbox": [ + 105, + 203, + 507, + 218 + ], + "score": 1.0, + "content": "leverages MC Dropout (Gal & Ghahramani, 2016) to obtain an epistemic uncertainty estimate.", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 214, + 506, + 228 + ], + "spans": [ + { + "bbox": [ + 105, + 214, + 506, + 228 + ], + "score": 1.0, + "content": "Following Kendall et al. (2015), we compute the pixel-wise posterior variance over multiple dropout", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 105, + 226, + 505, + 238 + ], + "spans": [ + { + "bbox": [ + 105, + 226, + 505, + 238 + ], + "score": 1.0, + "content": "masks and average across all classes, which serves as the anomaly score. We also experiment with", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 237, + 506, + 249 + ], + "spans": [ + { + "bbox": [ + 105, + 237, + 506, + 249 + ], + "score": 1.0, + "content": "an autoencoder baseline similar to Baur et al. (2019); Haselmann et al. (2018) where pixel-wise", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 105, + 248, + 506, + 260 + ], + "spans": [ + { + "bbox": [ + 105, + 248, + 506, + 260 + ], + "score": 1.0, + "content": "reconstruction loss is used as the anomaly score. This method is called AE. The “Branch” method is", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 259, + 506, + 271 + ], + "spans": [ + { + "bbox": [ + 106, + 259, + 506, + 271 + ], + "score": 1.0, + "content": "a direct port of the confidence branch detector from DeVries & Taylor (2018) to pixel-wise prediction.", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 270, + 488, + 282 + ], + "spans": [ + { + "bbox": [ + 105, + 270, + 488, + 282 + ], + "score": 1.0, + "content": "Finally, we use the MaxLogit method described in earlier sections independently on each pixel.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 9 + }, + { + "type": "text", + "bbox": [ + 106, + 287, + 505, + 342 + ], + "lines": [ + { + "bbox": [ + 106, + 287, + 506, + 299 + ], + "spans": [ + { + "bbox": [ + 106, + 287, + 506, + 299 + ], + "score": 1.0, + "content": "For all of the baselines except the autoencoder, we train a PSPNet (Zhao et al., 2017) decoder with a", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 296, + 506, + 311 + ], + "spans": [ + { + "bbox": [ + 105, + 296, + 506, + 311 + ], + "score": 1.0, + "content": "ResNet-101 encoder (He et al., 2015) for 20 epochs. We train both the encoder and decoder using", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 308, + 506, + 321 + ], + "spans": [ + { + "bbox": [ + 106, + 308, + 296, + 321 + ], + "score": 1.0, + "content": "SGD with momentum of 0.9, a learning rate of", + "type": "text" + }, + { + "bbox": [ + 297, + 308, + 335, + 319 + ], + "score": 0.92, + "content": "2 \\times 1 0 ^ { - 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 335, + 308, + 445, + 321 + ], + "score": 1.0, + "content": ", and learning rate decay of", + "type": "text" + }, + { + "bbox": [ + 446, + 308, + 467, + 319 + ], + "score": 0.89, + "content": "1 0 ^ { - 4 }", + "type": "inline_equation" + }, + { + "bbox": [ + 467, + 308, + 506, + 321 + ], + "score": 1.0, + "content": ". For AE,", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 320, + 506, + 332 + ], + "spans": [ + { + "bbox": [ + 106, + 320, + 506, + 332 + ], + "score": 1.0, + "content": "we use a 4-layer U-Net (Ronneberger et al., 2015) with a spatial latent code as in Baur et al. (2019).", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 330, + 444, + 343 + ], + "spans": [ + { + "bbox": [ + 105, + 330, + 444, + 343 + ], + "score": 1.0, + "content": "The U-Net also uses batch norm and is trained for 10 epochs. Results are in Table 4.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 15 + }, + { + "type": "text", + "bbox": [ + 107, + 347, + 317, + 457 + ], + "lines": [ + { + "bbox": [ + 106, + 347, + 317, + 360 + ], + "spans": [ + { + "bbox": [ + 106, + 347, + 317, + 360 + ], + "score": 1.0, + "content": "Results and Analysis. MaxLogit outperforms all", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 358, + 318, + 370 + ], + "spans": [ + { + "bbox": [ + 106, + 358, + 318, + 370 + ], + "score": 1.0, + "content": "other methods across the board by a substantial mar-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 369, + 318, + 382 + ], + "spans": [ + { + "bbox": [ + 105, + 369, + 318, + 382 + ], + "score": 1.0, + "content": "gin. The intuitive baseline of using the posterior", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 380, + 318, + 393 + ], + "spans": [ + { + "bbox": [ + 106, + 380, + 318, + 393 + ], + "score": 1.0, + "content": "for the background class to detect anomalies per-", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 392, + 317, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 392, + 317, + 403 + ], + "score": 1.0, + "content": "forms poorly, which suggests that the background", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 402, + 317, + 414 + ], + "spans": [ + { + "bbox": [ + 106, + 402, + 317, + 414 + ], + "score": 1.0, + "content": "class may not align with rare visual features. Even", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 413, + 318, + 425 + ], + "spans": [ + { + "bbox": [ + 105, + 413, + 318, + 425 + ], + "score": 1.0, + "content": "though reconstruction-based scores succeed in prod-", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 424, + 317, + 435 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 317, + 435 + ], + "score": 1.0, + "content": "uct fault segmentation, we find that the AE method", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 435, + 317, + 446 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 317, + 446 + ], + "score": 1.0, + "content": "performs poorly on the CAOS benchmark, which", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 446, + 317, + 457 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 317, + 457 + ], + "score": 1.0, + "content": "may be due to the more complex domain. AUPR for", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 25.0 + }, + { + "type": "table", + "bbox": [ + 324, + 351, + 504, + 398 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 324, + 351, + 504, + 398 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 324, + 351, + 504, + 398 + ], + "spans": [ + { + "bbox": [ + 324, + 351, + 504, + 398 + ], + "score": 0.965, + "html": "
MethodMSPMaxLogit
FS Lost and Found87.0%92.0%
Road Anomaly73.8%78.0%
", + "type": "table", + "image_path": "120ec03391e0f691aa82ae0e7be5996340d88c64bca62f5030a5c6f3232c23af.jpg" + } + ] + } + ], + "index": 22, + "virtual_lines": [ + { + "bbox": [ + 324, + 351, + 504, + 366.6666666666667 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 324, + 366.6666666666667, + 504, + 382.33333333333337 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 324, + 382.33333333333337, + 504, + 398.00000000000006 + ], + "spans": [], + "index": 25 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 324, + 410, + 505, + 443 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 324, + 411, + 505, + 422 + ], + "spans": [ + { + "bbox": [ + 324, + 411, + 505, + 422 + ], + "score": 1.0, + "content": "Figure 5: Auxiliary analysis of the MSP and", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 324, + 420, + 506, + 434 + ], + "spans": [ + { + "bbox": [ + 324, + 420, + 506, + 434 + ], + "score": 1.0, + "content": "the MaxLogit AUROCs using prior less com-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 324, + 432, + 498, + 444 + ], + "spans": [ + { + "bbox": [ + 324, + 432, + 498, + 444 + ], + "score": 1.0, + "content": "prehensive anomaly segmentation datasets.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31 + } + ], + "index": 26.5 + }, + { + "type": "text", + "bbox": [ + 107, + 457, + 505, + 501 + ], + "lines": [ + { + "bbox": [ + 106, + 457, + 506, + 469 + ], + "spans": [ + { + "bbox": [ + 106, + 457, + 506, + 469 + ], + "score": 1.0, + "content": "all methods is low, indicating that the large class imbalance presents a serious challenge. However,", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 468, + 505, + 480 + ], + "spans": [ + { + "bbox": [ + 105, + 468, + 505, + 480 + ], + "score": 1.0, + "content": "the substantial improvements with the MaxLogit method suggest that progress on this task is possible", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 479, + 505, + 492 + ], + "spans": [ + { + "bbox": [ + 105, + 479, + 505, + 492 + ], + "score": 1.0, + "content": "and there is much room for improvement. A comparison with other datasets is in Figure 5 (Pinggera", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 489, + 304, + 503 + ], + "spans": [ + { + "bbox": [ + 105, + 489, + 304, + 503 + ], + "score": 1.0, + "content": "et al., 2016; Blum et al., 2019; Jung et al., 2021).", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 35.5 + }, + { + "type": "text", + "bbox": [ + 106, + 506, + 505, + 573 + ], + "lines": [ + { + "bbox": [ + 105, + 507, + 505, + 520 + ], + "spans": [ + { + "bbox": [ + 105, + 507, + 505, + 520 + ], + "score": 1.0, + "content": "In Figure 4, we see that both MaxLogit and MSP have many false positives, as they assign high", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 518, + 506, + 531 + ], + "spans": [ + { + "bbox": [ + 105, + 518, + 506, + 531 + ], + "score": 1.0, + "content": "anomaly scores to semantic boundaries, a problem also observed in the recent works of (Blum et al.,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 528, + 506, + 542 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 506, + 542 + ], + "score": 1.0, + "content": "2019; Angus, 2019). However, the problem is less severe with MaxLogit. A potential explanation for", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 539, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 539, + 505, + 552 + ], + "score": 1.0, + "content": "this is that even when the prediction confidence dips at semantic boundaries, the maximum logit can", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 551, + 505, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 551, + 505, + 563 + ], + "score": 1.0, + "content": "remain the same in a ‘hand-off’ procedure between the classes. Thus, MaxLogit provides a natural", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 562, + 490, + 573 + ], + "spans": [ + { + "bbox": [ + 106, + 562, + 490, + 573 + ], + "score": 1.0, + "content": "mechanism to combat semantic boundary artifacts that could be further explored in future work.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 40.5 + }, + { + "type": "title", + "bbox": [ + 107, + 594, + 195, + 606 + ], + "lines": [ + { + "bbox": [ + 105, + 592, + 197, + 610 + ], + "spans": [ + { + "bbox": [ + 105, + 592, + 197, + 610 + ], + "score": 1.0, + "content": "6 CONCLUSION", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 44 + }, + { + "type": "text", + "bbox": [ + 107, + 622, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 621, + 505, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 621, + 505, + 634 + ], + "score": 1.0, + "content": "We scaled out-of-distribution detection to settings with thousands of classes and high-resolution", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 633, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 506, + 646 + ], + "score": 1.0, + "content": "images. We identified an issue faced by existing baselines when scaling to these settings and proposed", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "score": 1.0, + "content": "the maximum logit detector as a natural solution. We introduced the Species dataset to enable more", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 106, + 655, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 506, + 668 + ], + "score": 1.0, + "content": "controlled experiments without class overlap and also investigated using multi-label classifiers for", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "score": 1.0, + "content": "OOD detection, establishing an experimental setup for this previously unexplored setting. Finally, we", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "introduced the CAOS benchmark for anomaly segmentation, consisting of diverse, naturally-integrated", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 687, + 506, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 701 + ], + "score": 1.0, + "content": "anomalous objects in driving scenes. Baseline methods on the CAOS benchmark substantially im-", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "prove on random guessing but are still lacking, indicating potential for future work. Interestingly,", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "the MaxLogit detector also provides consistent and significant gains in the multi-label and anomaly", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 105, + 721, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 506, + 733 + ], + "score": 1.0, + "content": "segmentation settings, thereby establishing it as a new baseline in place of the maximum softmax prob-", + "type": "text" + } + ], + "index": 54 + } + ], + "index": 49.5 + } + ], + "page_idx": 8, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 302, + 751, + 308, + 759 + ], + "lines": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "spans": [ + { + "bbox": [ + 302, + 751, + 309, + 762 + ], + "score": 1.0, + "content": "9", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "table", + "bbox": [ + 106, + 64, + 506, + 147 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 64, + 506, + 147 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 64, + 506, + 147 + ], + "spans": [ + { + "bbox": [ + 106, + 64, + 506, + 147 + ], + "score": 0.979, + "html": "
MSPBranchBackgroundDropoutAEMaxLogit
StreetHazardsFPR9533.768.469.079.491.726.5
AUROC87.765.758.669.966.189.3
↓↑ AUPR 个6.61.54.57.52.210.6
BDD-Anomaly AUROCFPR9524.525.640.116.674.114.0
87.785.669.790.864.092.6
AUPR3.73.91.14.30.75.4
", + "type": "table", + "image_path": "a39f779b992a3a5f4420e180e8d83132c4aaf76e85751c0c198cb34b68e5580a.jpg" + } + ] + } + ], + "index": 1, + "virtual_lines": [ + { + "bbox": [ + 106, + 64, + 506, + 91.66666666666667 + ], + "spans": [], + "index": 0 + }, + { + "bbox": [ + 106, + 91.66666666666667, + 506, + 119.33333333333334 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 119.33333333333334, + 506, + 147.0 + ], + "spans": [], + "index": 2 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 106, + 155, + 506, + 188 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 154, + 505, + 167 + ], + "spans": [ + { + "bbox": [ + 105, + 154, + 505, + 167 + ], + "score": 1.0, + "content": "Table 4: Results on the CAOS benchmark. AUPR is low across the board due to the large class", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 165, + 506, + 178 + ], + "spans": [ + { + "bbox": [ + 105, + 165, + 506, + 178 + ], + "score": 1.0, + "content": "imbalance, but all methods perform substantially better than chance. MaxLogit obtains the best", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 177, + 272, + 190 + ], + "spans": [ + { + "bbox": [ + 105, + 177, + 272, + 190 + ], + "score": 1.0, + "content": "performance. All results are percentages.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + } + ], + "index": 2.5 + }, + { + "type": "text", + "bbox": [ + 107, + 204, + 505, + 281 + ], + "lines": [], + "index": 9, + "bbox_fs": [ + 105, + 203, + 507, + 282 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 106, + 287, + 505, + 342 + ], + "lines": [ + { + "bbox": [ + 106, + 287, + 506, + 299 + ], + "spans": [ + { + "bbox": [ + 106, + 287, + 506, + 299 + ], + "score": 1.0, + "content": "For all of the baselines except the autoencoder, we train a PSPNet (Zhao et al., 2017) decoder with a", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 296, + 506, + 311 + ], + "spans": [ + { + "bbox": [ + 105, + 296, + 506, + 311 + ], + "score": 1.0, + "content": "ResNet-101 encoder (He et al., 2015) for 20 epochs. We train both the encoder and decoder using", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 106, + 308, + 506, + 321 + ], + "spans": [ + { + "bbox": [ + 106, + 308, + 296, + 321 + ], + "score": 1.0, + "content": "SGD with momentum of 0.9, a learning rate of", + "type": "text" + }, + { + "bbox": [ + 297, + 308, + 335, + 319 + ], + "score": 0.92, + "content": "2 \\times 1 0 ^ { - 2 }", + "type": "inline_equation" + }, + { + "bbox": [ + 335, + 308, + 445, + 321 + ], + "score": 1.0, + "content": ", and learning rate decay of", + "type": "text" + }, + { + "bbox": [ + 446, + 308, + 467, + 319 + ], + "score": 0.89, + "content": "1 0 ^ { - 4 }", + "type": "inline_equation" + }, + { + "bbox": [ + 467, + 308, + 506, + 321 + ], + "score": 1.0, + "content": ". For AE,", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 106, + 320, + 506, + 332 + ], + "spans": [ + { + "bbox": [ + 106, + 320, + 506, + 332 + ], + "score": 1.0, + "content": "we use a 4-layer U-Net (Ronneberger et al., 2015) with a spatial latent code as in Baur et al. (2019).", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 105, + 330, + 444, + 343 + ], + "spans": [ + { + "bbox": [ + 105, + 330, + 444, + 343 + ], + "score": 1.0, + "content": "The U-Net also uses batch norm and is trained for 10 epochs. Results are in Table 4.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 15, + "bbox_fs": [ + 105, + 287, + 506, + 343 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 347, + 317, + 457 + ], + "lines": [ + { + "bbox": [ + 106, + 347, + 317, + 360 + ], + "spans": [ + { + "bbox": [ + 106, + 347, + 317, + 360 + ], + "score": 1.0, + "content": "Results and Analysis. MaxLogit outperforms all", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 358, + 318, + 370 + ], + "spans": [ + { + "bbox": [ + 106, + 358, + 318, + 370 + ], + "score": 1.0, + "content": "other methods across the board by a substantial mar-", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 369, + 318, + 382 + ], + "spans": [ + { + "bbox": [ + 105, + 369, + 318, + 382 + ], + "score": 1.0, + "content": "gin. The intuitive baseline of using the posterior", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 106, + 380, + 318, + 393 + ], + "spans": [ + { + "bbox": [ + 106, + 380, + 318, + 393 + ], + "score": 1.0, + "content": "for the background class to detect anomalies per-", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 392, + 317, + 403 + ], + "spans": [ + { + "bbox": [ + 105, + 392, + 317, + 403 + ], + "score": 1.0, + "content": "forms poorly, which suggests that the background", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 106, + 402, + 317, + 414 + ], + "spans": [ + { + "bbox": [ + 106, + 402, + 317, + 414 + ], + "score": 1.0, + "content": "class may not align with rare visual features. Even", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 413, + 318, + 425 + ], + "spans": [ + { + "bbox": [ + 105, + 413, + 318, + 425 + ], + "score": 1.0, + "content": "though reconstruction-based scores succeed in prod-", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 424, + 317, + 435 + ], + "spans": [ + { + "bbox": [ + 105, + 424, + 317, + 435 + ], + "score": 1.0, + "content": "uct fault segmentation, we find that the AE method", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 105, + 435, + 317, + 446 + ], + "spans": [ + { + "bbox": [ + 105, + 435, + 317, + 446 + ], + "score": 1.0, + "content": "performs poorly on the CAOS benchmark, which", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 446, + 317, + 457 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 317, + 457 + ], + "score": 1.0, + "content": "may be due to the more complex domain. AUPR for", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 457, + 506, + 469 + ], + "spans": [ + { + "bbox": [ + 106, + 457, + 506, + 469 + ], + "score": 1.0, + "content": "all methods is low, indicating that the large class imbalance presents a serious challenge. However,", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 468, + 505, + 480 + ], + "spans": [ + { + "bbox": [ + 105, + 468, + 505, + 480 + ], + "score": 1.0, + "content": "the substantial improvements with the MaxLogit method suggest that progress on this task is possible", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 479, + 505, + 492 + ], + "spans": [ + { + "bbox": [ + 105, + 479, + 505, + 492 + ], + "score": 1.0, + "content": "and there is much room for improvement. A comparison with other datasets is in Figure 5 (Pinggera", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 489, + 304, + 503 + ], + "spans": [ + { + "bbox": [ + 105, + 489, + 304, + 503 + ], + "score": 1.0, + "content": "et al., 2016; Blum et al., 2019; Jung et al., 2021).", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 25.0, + "bbox_fs": [ + 105, + 347, + 318, + 457 + ] + }, + { + "type": "table", + "bbox": [ + 324, + 351, + 504, + 398 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 324, + 351, + 504, + 398 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 324, + 351, + 504, + 398 + ], + "spans": [ + { + "bbox": [ + 324, + 351, + 504, + 398 + ], + "score": 0.965, + "html": "
MethodMSPMaxLogit
FS Lost and Found87.0%92.0%
Road Anomaly73.8%78.0%
", + "type": "table", + "image_path": "120ec03391e0f691aa82ae0e7be5996340d88c64bca62f5030a5c6f3232c23af.jpg" + } + ] + } + ], + "index": 22, + "virtual_lines": [ + { + "bbox": [ + 324, + 351, + 504, + 366.6666666666667 + ], + "spans": [], + "index": 20 + }, + { + "bbox": [ + 324, + 366.6666666666667, + 504, + 382.33333333333337 + ], + "spans": [], + "index": 22 + }, + { + "bbox": [ + 324, + 382.33333333333337, + 504, + 398.00000000000006 + ], + "spans": [], + "index": 25 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 324, + 410, + 505, + 443 + ], + "group_id": 1, + "lines": [ + { + "bbox": [ + 324, + 411, + 505, + 422 + ], + "spans": [ + { + "bbox": [ + 324, + 411, + 505, + 422 + ], + "score": 1.0, + "content": "Figure 5: Auxiliary analysis of the MSP and", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 324, + 420, + 506, + 434 + ], + "spans": [ + { + "bbox": [ + 324, + 420, + 506, + 434 + ], + "score": 1.0, + "content": "the MaxLogit AUROCs using prior less com-", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 324, + 432, + 498, + 444 + ], + "spans": [ + { + "bbox": [ + 324, + 432, + 498, + 444 + ], + "score": 1.0, + "content": "prehensive anomaly segmentation datasets.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31 + } + ], + "index": 26.5 + }, + { + "type": "text", + "bbox": [ + 107, + 457, + 505, + 501 + ], + "lines": [], + "index": 35.5, + "bbox_fs": [ + 105, + 457, + 506, + 503 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 106, + 506, + 505, + 573 + ], + "lines": [ + { + "bbox": [ + 105, + 507, + 505, + 520 + ], + "spans": [ + { + "bbox": [ + 105, + 507, + 505, + 520 + ], + "score": 1.0, + "content": "In Figure 4, we see that both MaxLogit and MSP have many false positives, as they assign high", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 105, + 518, + 506, + 531 + ], + "spans": [ + { + "bbox": [ + 105, + 518, + 506, + 531 + ], + "score": 1.0, + "content": "anomaly scores to semantic boundaries, a problem also observed in the recent works of (Blum et al.,", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 105, + 528, + 506, + 542 + ], + "spans": [ + { + "bbox": [ + 105, + 528, + 506, + 542 + ], + "score": 1.0, + "content": "2019; Angus, 2019). However, the problem is less severe with MaxLogit. A potential explanation for", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 539, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 105, + 539, + 505, + 552 + ], + "score": 1.0, + "content": "this is that even when the prediction confidence dips at semantic boundaries, the maximum logit can", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 551, + 505, + 563 + ], + "spans": [ + { + "bbox": [ + 105, + 551, + 505, + 563 + ], + "score": 1.0, + "content": "remain the same in a ‘hand-off’ procedure between the classes. Thus, MaxLogit provides a natural", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 106, + 562, + 490, + 573 + ], + "spans": [ + { + "bbox": [ + 106, + 562, + 490, + 573 + ], + "score": 1.0, + "content": "mechanism to combat semantic boundary artifacts that could be further explored in future work.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 40.5, + "bbox_fs": [ + 105, + 507, + 506, + 573 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 594, + 195, + 606 + ], + "lines": [ + { + "bbox": [ + 105, + 592, + 197, + 610 + ], + "spans": [ + { + "bbox": [ + 105, + 592, + 197, + 610 + ], + "score": 1.0, + "content": "6 CONCLUSION", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 44 + }, + { + "type": "text", + "bbox": [ + 107, + 622, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 621, + 505, + 634 + ], + "spans": [ + { + "bbox": [ + 106, + 621, + 505, + 634 + ], + "score": 1.0, + "content": "We scaled out-of-distribution detection to settings with thousands of classes and high-resolution", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 633, + 506, + 646 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 506, + 646 + ], + "score": 1.0, + "content": "images. We identified an issue faced by existing baselines when scaling to these settings and proposed", + "type": "text" + } + ], + "index": 46 + }, + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "score": 1.0, + "content": "the maximum logit detector as a natural solution. We introduced the Species dataset to enable more", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 106, + 655, + 506, + 668 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 506, + 668 + ], + "score": 1.0, + "content": "controlled experiments without class overlap and also investigated using multi-label classifiers for", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "spans": [ + { + "bbox": [ + 105, + 665, + 506, + 678 + ], + "score": 1.0, + "content": "OOD detection, establishing an experimental setup for this previously unexplored setting. Finally, we", + "type": "text" + } + ], + "index": 49 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "introduced the CAOS benchmark for anomaly segmentation, consisting of diverse, naturally-integrated", + "type": "text" + } + ], + "index": 50 + }, + { + "bbox": [ + 105, + 687, + 506, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 701 + ], + "score": 1.0, + "content": "anomalous objects in driving scenes. Baseline methods on the CAOS benchmark substantially im-", + "type": "text" + } + ], + "index": 51 + }, + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 506, + 712 + ], + "score": 1.0, + "content": "prove on random guessing but are still lacking, indicating potential for future work. Interestingly,", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 505, + 722 + ], + "score": 1.0, + "content": "the MaxLogit detector also provides consistent and significant gains in the multi-label and anomaly", + "type": "text" + } + ], + "index": 53 + }, + { + "bbox": [ + 105, + 721, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 721, + 506, + 733 + ], + "score": 1.0, + "content": "segmentation settings, thereby establishing it as a new baseline in place of the maximum softmax prob-", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 106, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "ability baseline on large-scale OOD detection problems. In all, we we hope that our contributions will", + "type": "text", + "cross_page": true + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 92, + 502, + 107 + ], + "spans": [ + { + "bbox": [ + 105, + 92, + 502, + 107 + ], + "score": 1.0, + "content": "enable further research on out-of-distribution detection for real-world safety-critical environments.", + "type": "text", + "cross_page": true + } + ], + "index": 1 + } + ], + "index": 49.5, + "bbox_fs": [ + 105, + 621, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 82, + 504, + 105 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "ability baseline on large-scale OOD detection problems. In all, we we hope that our contributions will", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 92, + 502, + 107 + ], + "spans": [ + { + "bbox": [ + 105, + 92, + 502, + 107 + ], + "score": 1.0, + "content": "enable further research on out-of-distribution detection for real-world safety-critical environments.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "title", + "bbox": [ + 107, + 122, + 176, + 135 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 176, + 135 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 176, + 135 + ], + "score": 1.0, + "content": "REFERENCES", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 2 + }, + { + "type": "text", + "bbox": [ + 106, + 141, + 430, + 153 + ], + "lines": [ + { + "bbox": [ + 106, + 140, + 432, + 154 + ], + "spans": [ + { + "bbox": [ + 106, + 140, + 432, + 154 + ], + "score": 1.0, + "content": "Matt Angus. Towards pixel-level ood detection for semantic segmentation, 2019.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 108, + 161, + 504, + 195 + ], + "lines": [ + { + "bbox": [ + 106, + 160, + 505, + 174 + ], + "spans": [ + { + "bbox": [ + 106, + 160, + 505, + 174 + ], + "score": 1.0, + "content": "Christoph Baur, Benedikt Wiestler, Shadi Albarqouni, and Nassir Navab. Deep autoencoding models", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 171, + 507, + 186 + ], + "spans": [ + { + "bbox": [ + 115, + 171, + 507, + 186 + ], + "score": 1.0, + "content": "for unsupervised anomaly segmentation in brain mr images. Lecture Notes in Computer Science,", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 183, + 425, + 196 + ], + "spans": [ + { + "bbox": [ + 115, + 183, + 425, + 196 + ], + "score": 1.0, + "content": "pp. 161–169, 2019. ISSN 1611-3349. doi: 10.1007/978-3-030-11723-8_16.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5 + }, + { + "type": "text", + "bbox": [ + 106, + 204, + 503, + 227 + ], + "lines": [ + { + "bbox": [ + 106, + 204, + 505, + 217 + ], + "spans": [ + { + "bbox": [ + 106, + 204, + 505, + 217 + ], + "score": 1.0, + "content": "Petra Bevandic, Ivan Krešo, Marin Orši ´ c, and Siniša Šegvi ´ c. Discriminative out-of-distribution ´", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 116, + 216, + 290, + 227 + ], + "spans": [ + { + "bbox": [ + 116, + 216, + 290, + 227 + ], + "score": 1.0, + "content": "detection for semantic segmentation, 2018.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 105, + 236, + 504, + 259 + ], + "lines": [ + { + "bbox": [ + 106, + 236, + 505, + 249 + ], + "spans": [ + { + "bbox": [ + 106, + 236, + 505, + 249 + ], + "score": 1.0, + "content": "Hermann Blum, Paul-Edouard Sarlin, Juan Nieto, Roland Siegwart, and Cesar Cadena. The", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 116, + 247, + 437, + 260 + ], + "spans": [ + { + "bbox": [ + 116, + 247, + 437, + 260 + ], + "score": 1.0, + "content": "fishyscapes benchmark: Measuring blind spots in semantic segmentation, 2019.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "text", + "bbox": [ + 105, + 267, + 504, + 290 + ], + "lines": [ + { + "bbox": [ + 105, + 267, + 505, + 280 + ], + "spans": [ + { + "bbox": [ + 105, + 267, + 318, + 280 + ], + "score": 1.0, + "content": "Markus M Breunig, Hans-Peter Kriegel, Raymond T", + "type": "text" + }, + { + "bbox": [ + 318, + 268, + 333, + 279 + ], + "score": 0.6, + "content": "\\mathrm { N g }", + "type": "inline_equation" + }, + { + "bbox": [ + 333, + 267, + 505, + 280 + ], + "score": 1.0, + "content": ", and Jörg Sander. Lof: identifying density-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 116, + 278, + 444, + 291 + ], + "spans": [ + { + "bbox": [ + 116, + 278, + 444, + 291 + ], + "score": 1.0, + "content": "based local outliers. In ACM sigmod record, volume 29, pp. 93–104. ACM, 2000.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5 + }, + { + "type": "text", + "bbox": [ + 104, + 298, + 505, + 321 + ], + "lines": [ + { + "bbox": [ + 105, + 298, + 506, + 312 + ], + "spans": [ + { + "bbox": [ + 105, + 298, + 506, + 312 + ], + "score": 1.0, + "content": "M. Cimpoi, S. Maji, I. Kokkinos, S. Mohamed, , and A. Vedaldi. Describing textures in the wild. In", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 116, + 309, + 313, + 322 + ], + "spans": [ + { + "bbox": [ + 116, + 309, + 313, + 322 + ], + "score": 1.0, + "content": "Computer Vision and Pattern Recognition, 2014.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5 + }, + { + "type": "text", + "bbox": [ + 106, + 329, + 505, + 374 + ], + "lines": [ + { + "bbox": [ + 105, + 329, + 505, + 343 + ], + "spans": [ + { + "bbox": [ + 105, + 329, + 505, + 343 + ], + "score": 1.0, + "content": "Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 116, + 341, + 505, + 353 + ], + "spans": [ + { + "bbox": [ + 116, + 341, + 505, + 353 + ], + "score": 1.0, + "content": "Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. The cityscapes dataset for semantic", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 116, + 352, + 505, + 364 + ], + "spans": [ + { + "bbox": [ + 116, + 352, + 505, + 364 + ], + "score": 1.0, + "content": "urban scene understanding. In Proc. of the IEEE Conference on Computer Vision and Pattern", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 116, + 363, + 228, + 375 + ], + "spans": [ + { + "bbox": [ + 116, + 363, + 228, + 375 + ], + "score": 1.0, + "content": "Recognition (CVPR), 2016.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 16.5 + }, + { + "type": "text", + "bbox": [ + 106, + 382, + 504, + 405 + ], + "lines": [ + { + "bbox": [ + 105, + 382, + 505, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 382, + 505, + 396 + ], + "score": 1.0, + "content": "J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei. ImageNet: A Large-Scale Hierarchical", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 394, + 262, + 405 + ], + "spans": [ + { + "bbox": [ + 115, + 394, + 262, + 405 + ], + "score": 1.0, + "content": "Image Database. In CVPR09, 2009.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5 + }, + { + "type": "text", + "bbox": [ + 105, + 414, + 505, + 437 + ], + "lines": [ + { + "bbox": [ + 106, + 414, + 505, + 426 + ], + "spans": [ + { + "bbox": [ + 106, + 414, + 505, + 426 + ], + "score": 1.0, + "content": "Terrance DeVries and Graham W Taylor. Learning confidence for out-of-distribution detection in", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 425, + 348, + 437 + ], + "spans": [ + { + "bbox": [ + 115, + 425, + 348, + 437 + ], + "score": 1.0, + "content": "neural networks. arXiv preprint arXiv:1802.04865, 2018.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5 + }, + { + "type": "text", + "bbox": [ + 107, + 444, + 506, + 479 + ], + "lines": [ + { + "bbox": [ + 106, + 445, + 507, + 457 + ], + "spans": [ + { + "bbox": [ + 106, + 445, + 507, + 457 + ], + "score": 1.0, + "content": "Alexey Dosovitskiy, German Ros, Felipe Codevilla, Antonio Lopez, and Vladlen Koltun. CARLA:", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 115, + 456, + 507, + 470 + ], + "spans": [ + { + "bbox": [ + 115, + 456, + 507, + 470 + ], + "score": 1.0, + "content": "An open urban driving simulator. In Proceedings of the 1st Annual Conference on Robot Learning,", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 114, + 468, + 181, + 479 + ], + "spans": [ + { + "bbox": [ + 114, + 468, + 181, + 479 + ], + "score": 1.0, + "content": "pp. 1–16, 2017.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24 + }, + { + "type": "text", + "bbox": [ + 107, + 487, + 506, + 532 + ], + "lines": [ + { + "bbox": [ + 106, + 488, + 505, + 499 + ], + "spans": [ + { + "bbox": [ + 106, + 488, + 505, + 499 + ], + "score": 1.0, + "content": "Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 116, + 499, + 505, + 511 + ], + "spans": [ + { + "bbox": [ + 116, + 499, + 505, + 511 + ], + "score": 1.0, + "content": "Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 509, + 506, + 522 + ], + "spans": [ + { + "bbox": [ + 115, + 509, + 506, + 522 + ], + "score": 1.0, + "content": "and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale.", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 115, + 521, + 242, + 533 + ], + "spans": [ + { + "bbox": [ + 115, + 521, + 242, + 533 + ], + "score": 1.0, + "content": "ArXiv, abs/2010.11929, 2021a.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 106, + 540, + 507, + 596 + ], + "lines": [ + { + "bbox": [ + 107, + 541, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 107, + 541, + 505, + 552 + ], + "score": 1.0, + "content": "Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 116, + 551, + 506, + 565 + ], + "spans": [ + { + "bbox": [ + 116, + 551, + 506, + 565 + ], + "score": 1.0, + "content": "Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit,", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 562, + 505, + 576 + ], + "spans": [ + { + "bbox": [ + 115, + 562, + 505, + 576 + ], + "score": 1.0, + "content": "and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. In", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 114, + 572, + 509, + 588 + ], + "spans": [ + { + "bbox": [ + 114, + 572, + 509, + 588 + ], + "score": 1.0, + "content": "International Conference on Learning Representations, 2021b. URL https://openreview.", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 584, + 259, + 598 + ], + "spans": [ + { + "bbox": [ + 115, + 584, + 188, + 598 + ], + "score": 1.0, + "content": "net/forum?id", + "type": "text" + }, + { + "bbox": [ + 189, + 586, + 195, + 594 + ], + "score": 0.47, + "content": "{ . } = { }", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 584, + 259, + 598 + ], + "score": 1.0, + "content": "YicbFdNTTy.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 32 + }, + { + "type": "text", + "bbox": [ + 105, + 604, + 505, + 628 + ], + "lines": [ + { + "bbox": [ + 105, + 603, + 506, + 618 + ], + "spans": [ + { + "bbox": [ + 105, + 603, + 506, + 618 + ], + "score": 1.0, + "content": "Andrew Emmott, Shubhomoy Das, Thomas Dietterich, Alan Fern, and Weng-Keen Wong. A", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 616, + 339, + 628 + ], + "spans": [ + { + "bbox": [ + 115, + 616, + 339, + 628 + ], + "score": 1.0, + "content": "meta-analysis of the anomaly detection problem, 2015.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5 + }, + { + "type": "text", + "bbox": [ + 106, + 636, + 505, + 669 + ], + "lines": [ + { + "bbox": [ + 105, + 635, + 506, + 649 + ], + "spans": [ + { + "bbox": [ + 105, + 635, + 506, + 649 + ], + "score": 1.0, + "content": "Mark Everingham, Luc Van Gool, Christopher K. I. Williams, John M. Winn, and Andrew Zisserman.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 116, + 647, + 506, + 659 + ], + "spans": [ + { + "bbox": [ + 116, + 647, + 506, + 659 + ], + "score": 1.0, + "content": "The pascal visual object classes (voc) challenge. International Journal of Computer Vision, 88:", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 116, + 658, + 181, + 669 + ], + "spans": [ + { + "bbox": [ + 116, + 658, + 181, + 669 + ], + "score": 1.0, + "content": "303–338, 2009.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38 + }, + { + "type": "text", + "bbox": [ + 105, + 678, + 504, + 701 + ], + "lines": [ + { + "bbox": [ + 106, + 678, + 505, + 691 + ], + "spans": [ + { + "bbox": [ + 106, + 678, + 505, + 691 + ], + "score": 1.0, + "content": "Stanislav Fort, Jie Ren, and Balaji Lakshminarayanan. Exploring the limits of out-of-distribution", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 116, + 689, + 322, + 702 + ], + "spans": [ + { + "bbox": [ + 116, + 689, + 322, + 702 + ], + "score": 1.0, + "content": "detection. arXiv preprint arXiv:2106.03004, 2021.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 40.5 + }, + { + "type": "text", + "bbox": [ + 107, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 708, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 708, + 505, + 722 + ], + "score": 1.0, + "content": "Yarin Gal and Zoubin Ghahramani. Dropout as a bayesian approximation: Representing model", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 116, + 721, + 453, + 732 + ], + "spans": [ + { + "bbox": [ + 116, + 721, + 453, + 732 + ], + "score": 1.0, + "content": "uncertainty in deep learning. International Conference on Machine Learning, 2016.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42.5 + } + ], + "page_idx": 9, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "10", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 105, + 82, + 504, + 105 + ], + "lines": [], + "index": 0.5, + "bbox_fs": [ + 105, + 82, + 505, + 107 + ], + "lines_deleted": true + }, + { + "type": "title", + "bbox": [ + 107, + 122, + 176, + 135 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 176, + 135 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 176, + 135 + ], + "score": 1.0, + "content": "REFERENCES", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 2 + }, + { + "type": "text", + "bbox": [ + 106, + 141, + 430, + 153 + ], + "lines": [ + { + "bbox": [ + 106, + 140, + 432, + 154 + ], + "spans": [ + { + "bbox": [ + 106, + 140, + 432, + 154 + ], + "score": 1.0, + "content": "Matt Angus. Towards pixel-level ood detection for semantic segmentation, 2019.", + "type": "text" + } + ], + "index": 3 + } + ], + "index": 3, + "bbox_fs": [ + 106, + 140, + 432, + 154 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 161, + 504, + 195 + ], + "lines": [ + { + "bbox": [ + 106, + 160, + 505, + 174 + ], + "spans": [ + { + "bbox": [ + 106, + 160, + 505, + 174 + ], + "score": 1.0, + "content": "Christoph Baur, Benedikt Wiestler, Shadi Albarqouni, and Nassir Navab. Deep autoencoding models", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 171, + 507, + 186 + ], + "spans": [ + { + "bbox": [ + 115, + 171, + 507, + 186 + ], + "score": 1.0, + "content": "for unsupervised anomaly segmentation in brain mr images. Lecture Notes in Computer Science,", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 183, + 425, + 196 + ], + "spans": [ + { + "bbox": [ + 115, + 183, + 425, + 196 + ], + "score": 1.0, + "content": "pp. 161–169, 2019. ISSN 1611-3349. doi: 10.1007/978-3-030-11723-8_16.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5, + "bbox_fs": [ + 106, + 160, + 507, + 196 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 204, + 503, + 227 + ], + "lines": [ + { + "bbox": [ + 106, + 204, + 505, + 217 + ], + "spans": [ + { + "bbox": [ + 106, + 204, + 505, + 217 + ], + "score": 1.0, + "content": "Petra Bevandic, Ivan Krešo, Marin Orši ´ c, and Siniša Šegvi ´ c. Discriminative out-of-distribution ´", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 116, + 216, + 290, + 227 + ], + "spans": [ + { + "bbox": [ + 116, + 216, + 290, + 227 + ], + "score": 1.0, + "content": "detection for semantic segmentation, 2018.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5, + "bbox_fs": [ + 106, + 204, + 505, + 227 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 236, + 504, + 259 + ], + "lines": [ + { + "bbox": [ + 106, + 236, + 505, + 249 + ], + "spans": [ + { + "bbox": [ + 106, + 236, + 505, + 249 + ], + "score": 1.0, + "content": "Hermann Blum, Paul-Edouard Sarlin, Juan Nieto, Roland Siegwart, and Cesar Cadena. The", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 116, + 247, + 437, + 260 + ], + "spans": [ + { + "bbox": [ + 116, + 247, + 437, + 260 + ], + "score": 1.0, + "content": "fishyscapes benchmark: Measuring blind spots in semantic segmentation, 2019.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5, + "bbox_fs": [ + 106, + 236, + 505, + 260 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 267, + 504, + 290 + ], + "lines": [ + { + "bbox": [ + 105, + 267, + 505, + 280 + ], + "spans": [ + { + "bbox": [ + 105, + 267, + 318, + 280 + ], + "score": 1.0, + "content": "Markus M Breunig, Hans-Peter Kriegel, Raymond T", + "type": "text" + }, + { + "bbox": [ + 318, + 268, + 333, + 279 + ], + "score": 0.6, + "content": "\\mathrm { N g }", + "type": "inline_equation" + }, + { + "bbox": [ + 333, + 267, + 505, + 280 + ], + "score": 1.0, + "content": ", and Jörg Sander. Lof: identifying density-", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 116, + 278, + 444, + 291 + ], + "spans": [ + { + "bbox": [ + 116, + 278, + 444, + 291 + ], + "score": 1.0, + "content": "based local outliers. In ACM sigmod record, volume 29, pp. 93–104. ACM, 2000.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5, + "bbox_fs": [ + 105, + 267, + 505, + 291 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 298, + 505, + 321 + ], + "lines": [ + { + "bbox": [ + 105, + 298, + 506, + 312 + ], + "spans": [ + { + "bbox": [ + 105, + 298, + 506, + 312 + ], + "score": 1.0, + "content": "M. Cimpoi, S. Maji, I. Kokkinos, S. Mohamed, , and A. Vedaldi. Describing textures in the wild. In", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 116, + 309, + 313, + 322 + ], + "spans": [ + { + "bbox": [ + 116, + 309, + 313, + 322 + ], + "score": 1.0, + "content": "Computer Vision and Pattern Recognition, 2014.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5, + "bbox_fs": [ + 105, + 298, + 506, + 322 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 329, + 505, + 374 + ], + "lines": [ + { + "bbox": [ + 105, + 329, + 505, + 343 + ], + "spans": [ + { + "bbox": [ + 105, + 329, + 505, + 343 + ], + "score": 1.0, + "content": "Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 116, + 341, + 505, + 353 + ], + "spans": [ + { + "bbox": [ + 116, + 341, + 505, + 353 + ], + "score": 1.0, + "content": "Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. The cityscapes dataset for semantic", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 116, + 352, + 505, + 364 + ], + "spans": [ + { + "bbox": [ + 116, + 352, + 505, + 364 + ], + "score": 1.0, + "content": "urban scene understanding. In Proc. of the IEEE Conference on Computer Vision and Pattern", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 116, + 363, + 228, + 375 + ], + "spans": [ + { + "bbox": [ + 116, + 363, + 228, + 375 + ], + "score": 1.0, + "content": "Recognition (CVPR), 2016.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 16.5, + "bbox_fs": [ + 105, + 329, + 505, + 375 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 382, + 504, + 405 + ], + "lines": [ + { + "bbox": [ + 105, + 382, + 505, + 396 + ], + "spans": [ + { + "bbox": [ + 105, + 382, + 505, + 396 + ], + "score": 1.0, + "content": "J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei. ImageNet: A Large-Scale Hierarchical", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 394, + 262, + 405 + ], + "spans": [ + { + "bbox": [ + 115, + 394, + 262, + 405 + ], + "score": 1.0, + "content": "Image Database. In CVPR09, 2009.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19.5, + "bbox_fs": [ + 105, + 382, + 505, + 405 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 414, + 505, + 437 + ], + "lines": [ + { + "bbox": [ + 106, + 414, + 505, + 426 + ], + "spans": [ + { + "bbox": [ + 106, + 414, + 505, + 426 + ], + "score": 1.0, + "content": "Terrance DeVries and Graham W Taylor. Learning confidence for out-of-distribution detection in", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 115, + 425, + 348, + 437 + ], + "spans": [ + { + "bbox": [ + 115, + 425, + 348, + 437 + ], + "score": 1.0, + "content": "neural networks. arXiv preprint arXiv:1802.04865, 2018.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21.5, + "bbox_fs": [ + 106, + 414, + 505, + 437 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 444, + 506, + 479 + ], + "lines": [ + { + "bbox": [ + 106, + 445, + 507, + 457 + ], + "spans": [ + { + "bbox": [ + 106, + 445, + 507, + 457 + ], + "score": 1.0, + "content": "Alexey Dosovitskiy, German Ros, Felipe Codevilla, Antonio Lopez, and Vladlen Koltun. CARLA:", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 115, + 456, + 507, + 470 + ], + "spans": [ + { + "bbox": [ + 115, + 456, + 507, + 470 + ], + "score": 1.0, + "content": "An open urban driving simulator. In Proceedings of the 1st Annual Conference on Robot Learning,", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 114, + 468, + 181, + 479 + ], + "spans": [ + { + "bbox": [ + 114, + 468, + 181, + 479 + ], + "score": 1.0, + "content": "pp. 1–16, 2017.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24, + "bbox_fs": [ + 106, + 445, + 507, + 479 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 487, + 506, + 532 + ], + "lines": [ + { + "bbox": [ + 106, + 488, + 505, + 499 + ], + "spans": [ + { + "bbox": [ + 106, + 488, + 505, + 499 + ], + "score": 1.0, + "content": "Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 116, + 499, + 505, + 511 + ], + "spans": [ + { + "bbox": [ + 116, + 499, + 505, + 511 + ], + "score": 1.0, + "content": "Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit,", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 509, + 506, + 522 + ], + "spans": [ + { + "bbox": [ + 115, + 509, + 506, + 522 + ], + "score": 1.0, + "content": "and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale.", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 115, + 521, + 242, + 533 + ], + "spans": [ + { + "bbox": [ + 115, + 521, + 242, + 533 + ], + "score": 1.0, + "content": "ArXiv, abs/2010.11929, 2021a.", + "type": "text" + } + ], + "index": 29 + } + ], + "index": 27.5, + "bbox_fs": [ + 106, + 488, + 506, + 533 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 540, + 507, + 596 + ], + "lines": [ + { + "bbox": [ + 107, + 541, + 505, + 552 + ], + "spans": [ + { + "bbox": [ + 107, + 541, + 505, + 552 + ], + "score": 1.0, + "content": "Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 116, + 551, + 506, + 565 + ], + "spans": [ + { + "bbox": [ + 116, + 551, + 506, + 565 + ], + "score": 1.0, + "content": "Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit,", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 562, + 505, + 576 + ], + "spans": [ + { + "bbox": [ + 115, + 562, + 505, + 576 + ], + "score": 1.0, + "content": "and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. In", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 114, + 572, + 509, + 588 + ], + "spans": [ + { + "bbox": [ + 114, + 572, + 509, + 588 + ], + "score": 1.0, + "content": "International Conference on Learning Representations, 2021b. URL https://openreview.", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 115, + 584, + 259, + 598 + ], + "spans": [ + { + "bbox": [ + 115, + 584, + 188, + 598 + ], + "score": 1.0, + "content": "net/forum?id", + "type": "text" + }, + { + "bbox": [ + 189, + 586, + 195, + 594 + ], + "score": 0.47, + "content": "{ . } = { }", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 584, + 259, + 598 + ], + "score": 1.0, + "content": "YicbFdNTTy.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 32, + "bbox_fs": [ + 107, + 541, + 509, + 598 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 604, + 505, + 628 + ], + "lines": [ + { + "bbox": [ + 105, + 603, + 506, + 618 + ], + "spans": [ + { + "bbox": [ + 105, + 603, + 506, + 618 + ], + "score": 1.0, + "content": "Andrew Emmott, Shubhomoy Das, Thomas Dietterich, Alan Fern, and Weng-Keen Wong. A", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 616, + 339, + 628 + ], + "spans": [ + { + "bbox": [ + 115, + 616, + 339, + 628 + ], + "score": 1.0, + "content": "meta-analysis of the anomaly detection problem, 2015.", + "type": "text" + } + ], + "index": 36 + } + ], + "index": 35.5, + "bbox_fs": [ + 105, + 603, + 506, + 628 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 636, + 505, + 669 + ], + "lines": [ + { + "bbox": [ + 105, + 635, + 506, + 649 + ], + "spans": [ + { + "bbox": [ + 105, + 635, + 506, + 649 + ], + "score": 1.0, + "content": "Mark Everingham, Luc Van Gool, Christopher K. I. Williams, John M. Winn, and Andrew Zisserman.", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 116, + 647, + 506, + 659 + ], + "spans": [ + { + "bbox": [ + 116, + 647, + 506, + 659 + ], + "score": 1.0, + "content": "The pascal visual object classes (voc) challenge. International Journal of Computer Vision, 88:", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 116, + 658, + 181, + 669 + ], + "spans": [ + { + "bbox": [ + 116, + 658, + 181, + 669 + ], + "score": 1.0, + "content": "303–338, 2009.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38, + "bbox_fs": [ + 105, + 635, + 506, + 669 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 678, + 504, + 701 + ], + "lines": [ + { + "bbox": [ + 106, + 678, + 505, + 691 + ], + "spans": [ + { + "bbox": [ + 106, + 678, + 505, + 691 + ], + "score": 1.0, + "content": "Stanislav Fort, Jie Ren, and Balaji Lakshminarayanan. Exploring the limits of out-of-distribution", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 116, + 689, + 322, + 702 + ], + "spans": [ + { + "bbox": [ + 116, + 689, + 322, + 702 + ], + "score": 1.0, + "content": "detection. arXiv preprint arXiv:2106.03004, 2021.", + "type": "text" + } + ], + "index": 41 + } + ], + "index": 40.5, + "bbox_fs": [ + 106, + 678, + 505, + 702 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 709, + 504, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 708, + 505, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 708, + 505, + 722 + ], + "score": 1.0, + "content": "Yarin Gal and Zoubin Ghahramani. Dropout as a bayesian approximation: Representing model", + "type": "text" + } + ], + "index": 42 + }, + { + "bbox": [ + 116, + 721, + 453, + 732 + ], + "spans": [ + { + "bbox": [ + 116, + 721, + 453, + 732 + ], + "score": 1.0, + "content": "uncertainty in deep learning. International Conference on Machine Learning, 2016.", + "type": "text" + } + ], + "index": 43 + } + ], + "index": 42.5, + "bbox_fs": [ + 106, + 708, + 505, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 108, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 505, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 505, + 96 + ], + "score": 1.0, + "content": "Matthias Haselmann, Dieter P Gruber, and Paul Tabatabai. Anomaly detection using deep learning", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 116, + 94, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 116, + 94, + 505, + 106 + ], + "score": 1.0, + "content": "based image completion. In 2018 17th IEEE International Conference on Machine Learning and", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 104, + 326, + 117 + ], + "spans": [ + { + "bbox": [ + 115, + 104, + 326, + 117 + ], + "score": 1.0, + "content": "Applications (ICMLA), pp. 1237–1242. IEEE, 2018.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 106, + 123, + 504, + 146 + ], + "lines": [ + { + "bbox": [ + 105, + 121, + 506, + 138 + ], + "spans": [ + { + "bbox": [ + 105, + 121, + 506, + 138 + ], + "score": 1.0, + "content": "Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 135, + 223, + 147 + ], + "spans": [ + { + "bbox": [ + 115, + 135, + 223, + 147 + ], + "score": 1.0, + "content": "recognition. CVPR, 2015.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5 + }, + { + "type": "text", + "bbox": [ + 106, + 153, + 504, + 176 + ], + "lines": [ + { + "bbox": [ + 106, + 153, + 505, + 166 + ], + "spans": [ + { + "bbox": [ + 106, + 153, + 505, + 166 + ], + "score": 1.0, + "content": "Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 165, + 288, + 176 + ], + "spans": [ + { + "bbox": [ + 115, + 165, + 288, + 176 + ], + "score": 1.0, + "content": "examples in neural networks. ICLR, 2017.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5 + }, + { + "type": "text", + "bbox": [ + 107, + 183, + 504, + 207 + ], + "lines": [ + { + "bbox": [ + 105, + 183, + 505, + 197 + ], + "spans": [ + { + "bbox": [ + 105, + 183, + 505, + 197 + ], + "score": 1.0, + "content": "Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 115, + 194, + 488, + 208 + ], + "spans": [ + { + "bbox": [ + 115, + 194, + 488, + 208 + ], + "score": 1.0, + "content": "exposure. Proceedings of the International Conference on Learning Representations, 2019a.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5 + }, + { + "type": "text", + "bbox": [ + 108, + 213, + 503, + 237 + ], + "lines": [ + { + "bbox": [ + 106, + 213, + 505, + 226 + ], + "spans": [ + { + "bbox": [ + 106, + 213, + 505, + 226 + ], + "score": 1.0, + "content": "Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 115, + 225, + 422, + 237 + ], + "spans": [ + { + "bbox": [ + 115, + 225, + 422, + 237 + ], + "score": 1.0, + "content": "exposure. In International Conference on Learning Representations, 2019b.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5 + }, + { + "type": "text", + "bbox": [ + 107, + 243, + 503, + 267 + ], + "lines": [ + { + "bbox": [ + 105, + 241, + 505, + 258 + ], + "spans": [ + { + "bbox": [ + 105, + 241, + 505, + 258 + ], + "score": 1.0, + "content": "Dan Hendrycks, Mantas Mazeika, Saurav Kadavath, and Dawn Song. Using self-supervised learning", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 116, + 255, + 386, + 267 + ], + "spans": [ + { + "bbox": [ + 116, + 255, + 386, + 267 + ], + "score": 1.0, + "content": "can improve model robustness and uncertainty. In NeurIPS, 2019c.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5 + }, + { + "type": "text", + "bbox": [ + 106, + 273, + 503, + 297 + ], + "lines": [ + { + "bbox": [ + 106, + 273, + 505, + 286 + ], + "spans": [ + { + "bbox": [ + 106, + 273, + 505, + 286 + ], + "score": 1.0, + "content": "Dan Hendrycks, Nicholas Carlini, John Schulman, and Jacob Steinhardt. Unsolved problems in ml", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 285, + 268, + 297 + ], + "spans": [ + { + "bbox": [ + 115, + 285, + 268, + 297 + ], + "score": 1.0, + "content": "safety. ArXiv, abs/2109.13916, 2021.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5 + }, + { + "type": "text", + "bbox": [ + 106, + 304, + 502, + 327 + ], + "lines": [ + { + "bbox": [ + 105, + 303, + 504, + 317 + ], + "spans": [ + { + "bbox": [ + 105, + 303, + 504, + 317 + ], + "score": 1.0, + "content": "Micah K Johnson and Hany Farid. Exposing digital forgeries by detecting inconsistencies in lighting.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 315, + 440, + 328 + ], + "spans": [ + { + "bbox": [ + 115, + 315, + 440, + 328 + ], + "score": 1.0, + "content": "In Proceedings of the 7th workshop on Multimedia and security, pp. 1–10, 2005.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5 + }, + { + "type": "text", + "bbox": [ + 107, + 334, + 505, + 368 + ], + "lines": [ + { + "bbox": [ + 106, + 334, + 506, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 334, + 506, + 347 + ], + "score": 1.0, + "content": "Sanghun Jung, Jungsoo Lee, Daehoon Gwak, Sungha Choi, and Jaegul Choo. Standardized max", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 115, + 345, + 505, + 358 + ], + "spans": [ + { + "bbox": [ + 115, + 345, + 505, + 358 + ], + "score": 1.0, + "content": "logits: A simple yet effective approach for identifying unexpected road obstacles in urban-scene", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 357, + 297, + 368 + ], + "spans": [ + { + "bbox": [ + 115, + 357, + 297, + 368 + ], + "score": 1.0, + "content": "segmentation. ArXiv, abs/2107.11264, 2021.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18 + }, + { + "type": "text", + "bbox": [ + 107, + 375, + 506, + 408 + ], + "lines": [ + { + "bbox": [ + 105, + 375, + 505, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 505, + 388 + ], + "score": 1.0, + "content": "Alex Kendall, Vijay Badrinarayanan, and Roberto Cipolla. Bayesian segnet: Model uncertainty in", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 115, + 387, + 506, + 399 + ], + "spans": [ + { + "bbox": [ + 115, + 387, + 506, + 399 + ], + "score": 1.0, + "content": "deep convolutional encoder-decoder architectures for scene understanding. ArXiv, abs/1511.02680,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 114, + 396, + 143, + 410 + ], + "spans": [ + { + "bbox": [ + 114, + 396, + 143, + 410 + ], + "score": 1.0, + "content": "2015.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21 + }, + { + "type": "text", + "bbox": [ + 106, + 416, + 486, + 428 + ], + "lines": [ + { + "bbox": [ + 106, + 416, + 487, + 430 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 487, + 430 + ], + "score": 1.0, + "content": "Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. ICLR, 2014.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 23 + }, + { + "type": "text", + "bbox": [ + 107, + 435, + 504, + 458 + ], + "lines": [ + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "score": 1.0, + "content": "Rajat Koner, Poulami Sinhamahapatra, Karsten Roscher, Stephan Günnemann, and Volker Tresp.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 117, + 447, + 389, + 458 + ], + "spans": [ + { + "bbox": [ + 117, + 447, + 389, + 458 + ], + "score": 1.0, + "content": "Oodformer: Out-of-distribution detection transformer. ArXiv, 2021.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24.5 + }, + { + "type": "text", + "bbox": [ + 106, + 465, + 504, + 488 + ], + "lines": [ + { + "bbox": [ + 106, + 465, + 505, + 478 + ], + "spans": [ + { + "bbox": [ + 106, + 465, + 505, + 478 + ], + "score": 1.0, + "content": "Ivan Krešo, Marin Oršic, Petra Bevandi ´ c, and Siniša Šegvi ´ c. Robust semantic segmentation with ´", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 117, + 477, + 241, + 488 + ], + "spans": [ + { + "bbox": [ + 117, + 477, + 241, + 488 + ], + "score": 1.0, + "content": "ladder-densenet models, 2018.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26.5 + }, + { + "type": "text", + "bbox": [ + 107, + 495, + 505, + 529 + ], + "lines": [ + { + "bbox": [ + 106, + 496, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 106, + 496, + 505, + 508 + ], + "score": 1.0, + "content": "Kimin Lee, Honglak Lee, Kibok Lee, and Jinwoo Shin. Training confidence-calibrated classifiers", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 115, + 506, + 506, + 520 + ], + "spans": [ + { + "bbox": [ + 115, + 506, + 506, + 520 + ], + "score": 1.0, + "content": "for detecting out-of-distribution samples. International Conference on Learning Representations,", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 115, + 516, + 147, + 531 + ], + "spans": [ + { + "bbox": [ + 115, + 516, + 147, + 531 + ], + "score": 1.0, + "content": "2018a.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 105, + 536, + 504, + 560 + ], + "lines": [ + { + "bbox": [ + 105, + 536, + 506, + 551 + ], + "spans": [ + { + "bbox": [ + 105, + 536, + 506, + 551 + ], + "score": 1.0, + "content": "Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. A simple unified framework for detecting", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 547, + 451, + 560 + ], + "spans": [ + { + "bbox": [ + 115, + 547, + 451, + 560 + ], + "score": 1.0, + "content": "out-of-distribution samples and adversarial attacks. ArXiv, abs/1807.03888, 2018b.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5 + }, + { + "type": "text", + "bbox": [ + 104, + 566, + 504, + 590 + ], + "lines": [ + { + "bbox": [ + 105, + 565, + 505, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 505, + 581 + ], + "score": 1.0, + "content": "Shiyu Liang, Yixuan Li, and R. Srikant. Enhancing the reliability of out-of-distribution image", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 116, + 578, + 484, + 591 + ], + "spans": [ + { + "bbox": [ + 116, + 578, + 484, + 591 + ], + "score": 1.0, + "content": "detection in neural networks. International Conference on Learning Representations, 2018.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5 + }, + { + "type": "text", + "bbox": [ + 108, + 596, + 505, + 631 + ], + "lines": [ + { + "bbox": [ + 106, + 596, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 106, + 596, + 505, + 610 + ], + "score": 1.0, + "content": "Tsung-Yi Lin, Michael Maire, Serge Belongie, Lubomir Bourdev, Ross Girshick, James Hays, Pietro", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 608, + 505, + 621 + ], + "spans": [ + { + "bbox": [ + 115, + 608, + 505, + 621 + ], + "score": 1.0, + "content": "Perona, Deva Ramanan, C. Lawrence Zitnick, and Piotr Dollar. Microsoft COCO: Common objects", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 619, + 217, + 630 + ], + "spans": [ + { + "bbox": [ + 115, + 619, + 217, + 630 + ], + "score": 1.0, + "content": "in context. ECCV, 2014.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36 + }, + { + "type": "text", + "bbox": [ + 105, + 637, + 505, + 661 + ], + "lines": [ + { + "bbox": [ + 106, + 637, + 506, + 651 + ], + "spans": [ + { + "bbox": [ + 106, + 637, + 506, + 651 + ], + "score": 1.0, + "content": "Sina Mohseni, Mandar Pitale, Jbs Yadawa, and Zhangyang Wang. Self-supervised learning for", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 649, + 353, + 662 + ], + "spans": [ + { + "bbox": [ + 115, + 649, + 353, + 662 + ], + "score": 1.0, + "content": "generalizable out-of-distribution detection. In AAAI, 2020.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5 + }, + { + "type": "text", + "bbox": [ + 108, + 668, + 504, + 703 + ], + "lines": [ + { + "bbox": [ + 105, + 668, + 506, + 682 + ], + "spans": [ + { + "bbox": [ + 105, + 668, + 506, + 682 + ], + "score": 1.0, + "content": "Peter Pinggera, Sebastian Ramos, Stefan K. Gehrig, Uwe Franke, Carsten Rother, and Rudolf", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 116, + 679, + 505, + 692 + ], + "spans": [ + { + "bbox": [ + 116, + 679, + 505, + 692 + ], + "score": 1.0, + "content": "Mester. Lost and found: Detecting small road hazards for self-driving vehicles. 2016 IEEE/RSJ", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 116, + 690, + 479, + 703 + ], + "spans": [ + { + "bbox": [ + 116, + 690, + 479, + 703 + ], + "score": 1.0, + "content": "International Conference on Intelligent Robots and Systems (IROS), pp. 1099–1106, 2016.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41 + }, + { + "type": "text", + "bbox": [ + 107, + 709, + 503, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 504, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 504, + 722 + ], + "score": 1.0, + "content": "T. Ridnik, Emanuel Ben-Baruch, Asaf Noy, and Lihi Zelnik-Manor. Imagenet-21k pretraining for the", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 115, + 720, + 277, + 733 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 277, + 733 + ], + "score": 1.0, + "content": "masses. ArXiv, abs/2104.10972, 2021a.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 43.5 + } + ], + "page_idx": 10, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 309, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 765 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 765 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 15, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 108, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 105, + 81, + 505, + 96 + ], + "spans": [ + { + "bbox": [ + 105, + 81, + 505, + 96 + ], + "score": 1.0, + "content": "Matthias Haselmann, Dieter P Gruber, and Paul Tabatabai. Anomaly detection using deep learning", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 116, + 94, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 116, + 94, + 505, + 106 + ], + "score": 1.0, + "content": "based image completion. In 2018 17th IEEE International Conference on Machine Learning and", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 115, + 104, + 326, + 117 + ], + "spans": [ + { + "bbox": [ + 115, + 104, + 326, + 117 + ], + "score": 1.0, + "content": "Applications (ICMLA), pp. 1237–1242. IEEE, 2018.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1, + "bbox_fs": [ + 105, + 81, + 505, + 117 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 123, + 504, + 146 + ], + "lines": [ + { + "bbox": [ + 105, + 121, + 506, + 138 + ], + "spans": [ + { + "bbox": [ + 105, + 121, + 506, + 138 + ], + "score": 1.0, + "content": "Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 135, + 223, + 147 + ], + "spans": [ + { + "bbox": [ + 115, + 135, + 223, + 147 + ], + "score": 1.0, + "content": "recognition. CVPR, 2015.", + "type": "text" + } + ], + "index": 4 + } + ], + "index": 3.5, + "bbox_fs": [ + 105, + 121, + 506, + 147 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 153, + 504, + 176 + ], + "lines": [ + { + "bbox": [ + 106, + 153, + 505, + 166 + ], + "spans": [ + { + "bbox": [ + 106, + 153, + 505, + 166 + ], + "score": 1.0, + "content": "Dan Hendrycks and Kevin Gimpel. A baseline for detecting misclassified and out-of-distribution", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 115, + 165, + 288, + 176 + ], + "spans": [ + { + "bbox": [ + 115, + 165, + 288, + 176 + ], + "score": 1.0, + "content": "examples in neural networks. ICLR, 2017.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 5.5, + "bbox_fs": [ + 106, + 153, + 505, + 176 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 183, + 504, + 207 + ], + "lines": [ + { + "bbox": [ + 105, + 183, + 505, + 197 + ], + "spans": [ + { + "bbox": [ + 105, + 183, + 505, + 197 + ], + "score": 1.0, + "content": "Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 115, + 194, + 488, + 208 + ], + "spans": [ + { + "bbox": [ + 115, + 194, + 488, + 208 + ], + "score": 1.0, + "content": "exposure. Proceedings of the International Conference on Learning Representations, 2019a.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7.5, + "bbox_fs": [ + 105, + 183, + 505, + 208 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 213, + 503, + 237 + ], + "lines": [ + { + "bbox": [ + 106, + 213, + 505, + 226 + ], + "spans": [ + { + "bbox": [ + 106, + 213, + 505, + 226 + ], + "score": 1.0, + "content": "Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. Deep anomaly detection with outlier", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 115, + 225, + 422, + 237 + ], + "spans": [ + { + "bbox": [ + 115, + 225, + 422, + 237 + ], + "score": 1.0, + "content": "exposure. In International Conference on Learning Representations, 2019b.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 9.5, + "bbox_fs": [ + 106, + 213, + 505, + 237 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 243, + 503, + 267 + ], + "lines": [ + { + "bbox": [ + 105, + 241, + 505, + 258 + ], + "spans": [ + { + "bbox": [ + 105, + 241, + 505, + 258 + ], + "score": 1.0, + "content": "Dan Hendrycks, Mantas Mazeika, Saurav Kadavath, and Dawn Song. Using self-supervised learning", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 116, + 255, + 386, + 267 + ], + "spans": [ + { + "bbox": [ + 116, + 255, + 386, + 267 + ], + "score": 1.0, + "content": "can improve model robustness and uncertainty. In NeurIPS, 2019c.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 11.5, + "bbox_fs": [ + 105, + 241, + 505, + 267 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 273, + 503, + 297 + ], + "lines": [ + { + "bbox": [ + 106, + 273, + 505, + 286 + ], + "spans": [ + { + "bbox": [ + 106, + 273, + 505, + 286 + ], + "score": 1.0, + "content": "Dan Hendrycks, Nicholas Carlini, John Schulman, and Jacob Steinhardt. Unsolved problems in ml", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 115, + 285, + 268, + 297 + ], + "spans": [ + { + "bbox": [ + 115, + 285, + 268, + 297 + ], + "score": 1.0, + "content": "safety. ArXiv, abs/2109.13916, 2021.", + "type": "text" + } + ], + "index": 14 + } + ], + "index": 13.5, + "bbox_fs": [ + 106, + 273, + 505, + 297 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 304, + 502, + 327 + ], + "lines": [ + { + "bbox": [ + 105, + 303, + 504, + 317 + ], + "spans": [ + { + "bbox": [ + 105, + 303, + 504, + 317 + ], + "score": 1.0, + "content": "Micah K Johnson and Hany Farid. Exposing digital forgeries by detecting inconsistencies in lighting.", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 115, + 315, + 440, + 328 + ], + "spans": [ + { + "bbox": [ + 115, + 315, + 440, + 328 + ], + "score": 1.0, + "content": "In Proceedings of the 7th workshop on Multimedia and security, pp. 1–10, 2005.", + "type": "text" + } + ], + "index": 16 + } + ], + "index": 15.5, + "bbox_fs": [ + 105, + 303, + 504, + 328 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 334, + 505, + 368 + ], + "lines": [ + { + "bbox": [ + 106, + 334, + 506, + 347 + ], + "spans": [ + { + "bbox": [ + 106, + 334, + 506, + 347 + ], + "score": 1.0, + "content": "Sanghun Jung, Jungsoo Lee, Daehoon Gwak, Sungha Choi, and Jaegul Choo. Standardized max", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 115, + 345, + 505, + 358 + ], + "spans": [ + { + "bbox": [ + 115, + 345, + 505, + 358 + ], + "score": 1.0, + "content": "logits: A simple yet effective approach for identifying unexpected road obstacles in urban-scene", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 357, + 297, + 368 + ], + "spans": [ + { + "bbox": [ + 115, + 357, + 297, + 368 + ], + "score": 1.0, + "content": "segmentation. ArXiv, abs/2107.11264, 2021.", + "type": "text" + } + ], + "index": 19 + } + ], + "index": 18, + "bbox_fs": [ + 106, + 334, + 506, + 368 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 375, + 506, + 408 + ], + "lines": [ + { + "bbox": [ + 105, + 375, + 505, + 388 + ], + "spans": [ + { + "bbox": [ + 105, + 375, + 505, + 388 + ], + "score": 1.0, + "content": "Alex Kendall, Vijay Badrinarayanan, and Roberto Cipolla. Bayesian segnet: Model uncertainty in", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 115, + 387, + 506, + 399 + ], + "spans": [ + { + "bbox": [ + 115, + 387, + 506, + 399 + ], + "score": 1.0, + "content": "deep convolutional encoder-decoder architectures for scene understanding. ArXiv, abs/1511.02680,", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 114, + 396, + 143, + 410 + ], + "spans": [ + { + "bbox": [ + 114, + 396, + 143, + 410 + ], + "score": 1.0, + "content": "2015.", + "type": "text" + } + ], + "index": 22 + } + ], + "index": 21, + "bbox_fs": [ + 105, + 375, + 506, + 410 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 416, + 486, + 428 + ], + "lines": [ + { + "bbox": [ + 106, + 416, + 487, + 430 + ], + "spans": [ + { + "bbox": [ + 106, + 416, + 487, + 430 + ], + "score": 1.0, + "content": "Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. ICLR, 2014.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 23, + "bbox_fs": [ + 106, + 416, + 487, + 430 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 435, + 504, + 458 + ], + "lines": [ + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 105, + 434, + 506, + 448 + ], + "score": 1.0, + "content": "Rajat Koner, Poulami Sinhamahapatra, Karsten Roscher, Stephan Günnemann, and Volker Tresp.", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 117, + 447, + 389, + 458 + ], + "spans": [ + { + "bbox": [ + 117, + 447, + 389, + 458 + ], + "score": 1.0, + "content": "Oodformer: Out-of-distribution detection transformer. ArXiv, 2021.", + "type": "text" + } + ], + "index": 25 + } + ], + "index": 24.5, + "bbox_fs": [ + 105, + 434, + 506, + 458 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 465, + 504, + 488 + ], + "lines": [ + { + "bbox": [ + 106, + 465, + 505, + 478 + ], + "spans": [ + { + "bbox": [ + 106, + 465, + 505, + 478 + ], + "score": 1.0, + "content": "Ivan Krešo, Marin Oršic, Petra Bevandi ´ c, and Siniša Šegvi ´ c. Robust semantic segmentation with ´", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 117, + 477, + 241, + 488 + ], + "spans": [ + { + "bbox": [ + 117, + 477, + 241, + 488 + ], + "score": 1.0, + "content": "ladder-densenet models, 2018.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 26.5, + "bbox_fs": [ + 106, + 465, + 505, + 488 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 495, + 505, + 529 + ], + "lines": [ + { + "bbox": [ + 106, + 496, + 505, + 508 + ], + "spans": [ + { + "bbox": [ + 106, + 496, + 505, + 508 + ], + "score": 1.0, + "content": "Kimin Lee, Honglak Lee, Kibok Lee, and Jinwoo Shin. Training confidence-calibrated classifiers", + "type": "text" + } + ], + "index": 28 + }, + { + "bbox": [ + 115, + 506, + 506, + 520 + ], + "spans": [ + { + "bbox": [ + 115, + 506, + 506, + 520 + ], + "score": 1.0, + "content": "for detecting out-of-distribution samples. International Conference on Learning Representations,", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 115, + 516, + 147, + 531 + ], + "spans": [ + { + "bbox": [ + 115, + 516, + 147, + 531 + ], + "score": 1.0, + "content": "2018a.", + "type": "text" + } + ], + "index": 30 + } + ], + "index": 29, + "bbox_fs": [ + 106, + 496, + 506, + 531 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 536, + 504, + 560 + ], + "lines": [ + { + "bbox": [ + 105, + 536, + 506, + 551 + ], + "spans": [ + { + "bbox": [ + 105, + 536, + 506, + 551 + ], + "score": 1.0, + "content": "Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. A simple unified framework for detecting", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 115, + 547, + 451, + 560 + ], + "spans": [ + { + "bbox": [ + 115, + 547, + 451, + 560 + ], + "score": 1.0, + "content": "out-of-distribution samples and adversarial attacks. ArXiv, abs/1807.03888, 2018b.", + "type": "text" + } + ], + "index": 32 + } + ], + "index": 31.5, + "bbox_fs": [ + 105, + 536, + 506, + 560 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 566, + 504, + 590 + ], + "lines": [ + { + "bbox": [ + 105, + 565, + 505, + 581 + ], + "spans": [ + { + "bbox": [ + 105, + 565, + 505, + 581 + ], + "score": 1.0, + "content": "Shiyu Liang, Yixuan Li, and R. Srikant. Enhancing the reliability of out-of-distribution image", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 116, + 578, + 484, + 591 + ], + "spans": [ + { + "bbox": [ + 116, + 578, + 484, + 591 + ], + "score": 1.0, + "content": "detection in neural networks. International Conference on Learning Representations, 2018.", + "type": "text" + } + ], + "index": 34 + } + ], + "index": 33.5, + "bbox_fs": [ + 105, + 565, + 505, + 591 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 596, + 505, + 631 + ], + "lines": [ + { + "bbox": [ + 106, + 596, + 505, + 610 + ], + "spans": [ + { + "bbox": [ + 106, + 596, + 505, + 610 + ], + "score": 1.0, + "content": "Tsung-Yi Lin, Michael Maire, Serge Belongie, Lubomir Bourdev, Ross Girshick, James Hays, Pietro", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 115, + 608, + 505, + 621 + ], + "spans": [ + { + "bbox": [ + 115, + 608, + 505, + 621 + ], + "score": 1.0, + "content": "Perona, Deva Ramanan, C. Lawrence Zitnick, and Piotr Dollar. Microsoft COCO: Common objects", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 115, + 619, + 217, + 630 + ], + "spans": [ + { + "bbox": [ + 115, + 619, + 217, + 630 + ], + "score": 1.0, + "content": "in context. ECCV, 2014.", + "type": "text" + } + ], + "index": 37 + } + ], + "index": 36, + "bbox_fs": [ + 106, + 596, + 505, + 630 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 637, + 505, + 661 + ], + "lines": [ + { + "bbox": [ + 106, + 637, + 506, + 651 + ], + "spans": [ + { + "bbox": [ + 106, + 637, + 506, + 651 + ], + "score": 1.0, + "content": "Sina Mohseni, Mandar Pitale, Jbs Yadawa, and Zhangyang Wang. Self-supervised learning for", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 115, + 649, + 353, + 662 + ], + "spans": [ + { + "bbox": [ + 115, + 649, + 353, + 662 + ], + "score": 1.0, + "content": "generalizable out-of-distribution detection. In AAAI, 2020.", + "type": "text" + } + ], + "index": 39 + } + ], + "index": 38.5, + "bbox_fs": [ + 106, + 637, + 506, + 662 + ] + }, + { + "type": "text", + "bbox": [ + 108, + 668, + 504, + 703 + ], + "lines": [ + { + "bbox": [ + 105, + 668, + 506, + 682 + ], + "spans": [ + { + "bbox": [ + 105, + 668, + 506, + 682 + ], + "score": 1.0, + "content": "Peter Pinggera, Sebastian Ramos, Stefan K. Gehrig, Uwe Franke, Carsten Rother, and Rudolf", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 116, + 679, + 505, + 692 + ], + "spans": [ + { + "bbox": [ + 116, + 679, + 505, + 692 + ], + "score": 1.0, + "content": "Mester. Lost and found: Detecting small road hazards for self-driving vehicles. 2016 IEEE/RSJ", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 116, + 690, + 479, + 703 + ], + "spans": [ + { + "bbox": [ + 116, + 690, + 479, + 703 + ], + "score": 1.0, + "content": "International Conference on Intelligent Robots and Systems (IROS), pp. 1099–1106, 2016.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 41, + "bbox_fs": [ + 105, + 668, + 506, + 703 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 709, + 503, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 709, + 504, + 722 + ], + "spans": [ + { + "bbox": [ + 106, + 709, + 504, + 722 + ], + "score": 1.0, + "content": "T. Ridnik, Emanuel Ben-Baruch, Asaf Noy, and Lihi Zelnik-Manor. Imagenet-21k pretraining for the", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 115, + 720, + 277, + 733 + ], + "spans": [ + { + "bbox": [ + 115, + 720, + 277, + 733 + ], + "score": 1.0, + "content": "masses. ArXiv, abs/2104.10972, 2021a.", + "type": "text" + } + ], + "index": 44 + } + ], + "index": 43.5, + "bbox_fs": [ + 106, + 709, + 504, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 110, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 107, + 81, + 506, + 96 + ], + "spans": [ + { + "bbox": [ + 107, + 81, + 506, + 96 + ], + "score": 1.0, + "content": "Tal Ridnik, Hussam Lawen, Asaf Noy, Emanuel Ben Baruch, Gilad Sharir, and Itamar Friedman.", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 116, + 94, + 506, + 106 + ], + "spans": [ + { + "bbox": [ + 116, + 94, + 506, + 106 + ], + "score": 1.0, + "content": "Tresnet: High performance gpu-dedicated architecture. In Proceedings of the IEEE/CVF Winter", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 116, + 104, + 407, + 118 + ], + "spans": [ + { + "bbox": [ + 116, + 104, + 407, + 118 + ], + "score": 1.0, + "content": "Conference on Applications of Computer Vision, pp. 1400–1409, 2021b.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1 + }, + { + "type": "text", + "bbox": [ + 109, + 123, + 504, + 156 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "score": 1.0, + "content": "Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 135, + 505, + 147 + ], + "spans": [ + { + "bbox": [ + 115, + 135, + 505, + 147 + ], + "score": 1.0, + "content": "image segmentation. Medical Image Computing and Computer-Assisted Intervention – MICCAI", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 145, + 449, + 158 + ], + "spans": [ + { + "bbox": [ + 115, + 145, + 449, + 158 + ], + "score": 1.0, + "content": "2015, pp. 234–241, 2015. ISSN 1611-3349. doi: 10.1007/978-3-319-24574-4_28.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4 + }, + { + "type": "text", + "bbox": [ + 107, + 164, + 505, + 197 + ], + "lines": [ + { + "bbox": [ + 106, + 164, + 506, + 176 + ], + "spans": [ + { + "bbox": [ + 106, + 164, + 506, + 176 + ], + "score": 1.0, + "content": "Manolis Savva, Angel X. Chang, and Pat Hanrahan. Semantically-Enriched 3D Models for Common-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 114, + 174, + 507, + 188 + ], + "spans": [ + { + "bbox": [ + 114, + 174, + 507, + 188 + ], + "score": 1.0, + "content": "sense Knowledge. CVPR 2015 Workshop on Functionality, Physics, Intentionality and Causality,", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 114, + 184, + 143, + 199 + ], + "spans": [ + { + "bbox": [ + 114, + 184, + 143, + 199 + ], + "score": 1.0, + "content": "2015.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7 + }, + { + "type": "text", + "bbox": [ + 106, + 204, + 506, + 249 + ], + "lines": [ + { + "bbox": [ + 105, + 204, + 506, + 218 + ], + "spans": [ + { + "bbox": [ + 105, + 204, + 506, + 218 + ], + "score": 1.0, + "content": "Bernhard Schölkopf, Robert Williamson, Alex Smola, John Shawe-Taylor, and John Platt. Support", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 116, + 216, + 505, + 228 + ], + "spans": [ + { + "bbox": [ + 116, + 216, + 505, + 228 + ], + "score": 1.0, + "content": "vector method for novelty detection. In Proceedings of the 12th International Conference on", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 114, + 226, + 506, + 240 + ], + "spans": [ + { + "bbox": [ + 114, + 226, + 506, + 240 + ], + "score": 1.0, + "content": "Neural Information Processing Systems, NIPS’99, pp. 582–588, Cambridge, MA, USA, 1999. MIT", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 115, + 237, + 144, + 251 + ], + "spans": [ + { + "bbox": [ + 115, + 237, + 144, + 251 + ], + "score": 1.0, + "content": "Press.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 10.5 + }, + { + "type": "text", + "bbox": [ + 106, + 256, + 506, + 290 + ], + "lines": [ + { + "bbox": [ + 105, + 256, + 506, + 269 + ], + "spans": [ + { + "bbox": [ + 105, + 256, + 506, + 269 + ], + "score": 1.0, + "content": "Ilya Tolstikhin, Neil Houlsby, Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Thomas Un-", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 116, + 268, + 505, + 280 + ], + "spans": [ + { + "bbox": [ + 116, + 268, + 505, + 280 + ], + "score": 1.0, + "content": "terthiner, Jessica Yung, Daniel Keysers, Jakob Uszkoreit, Mario Lucic, et al. Mlp-mixer: An", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 115, + 277, + 405, + 291 + ], + "spans": [ + { + "bbox": [ + 115, + 277, + 405, + 291 + ], + "score": 1.0, + "content": "all-mlp architecture for vision. arXiv preprint arXiv:2105.01601, 2021.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14 + }, + { + "type": "text", + "bbox": [ + 106, + 297, + 505, + 320 + ], + "lines": [ + { + "bbox": [ + 105, + 297, + 506, + 310 + ], + "spans": [ + { + "bbox": [ + 105, + 297, + 506, + 310 + ], + "score": 1.0, + "content": "Fisher Yu, Yinda Zhang, Shuran Song, Ari Seff, and Jianxiong Xiao. LSUN: construction of a", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 308, + 456, + 321 + ], + "spans": [ + { + "bbox": [ + 115, + 308, + 456, + 321 + ], + "score": 1.0, + "content": "large-scale image dataset using deep learning with humans in the loop. CoRR, 2015.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5 + }, + { + "type": "text", + "bbox": [ + 107, + 327, + 505, + 361 + ], + "lines": [ + { + "bbox": [ + 105, + 327, + 505, + 339 + ], + "spans": [ + { + "bbox": [ + 105, + 327, + 505, + 339 + ], + "score": 1.0, + "content": "Fisher Yu, Wenqi Xian, Yingying Chen, Fangchen Liu, Mike Liao, Vashisht Madhavan, and Trevor", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 338, + 506, + 351 + ], + "spans": [ + { + "bbox": [ + 115, + 338, + 506, + 351 + ], + "score": 1.0, + "content": "Darrell. BDD100K: A diverse driving video database with scalable annotation tooling. CoRR,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 349, + 210, + 361 + ], + "spans": [ + { + "bbox": [ + 115, + 349, + 210, + 361 + ], + "score": 1.0, + "content": "abs/1805.04687, 2018.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19 + }, + { + "type": "text", + "bbox": [ + 107, + 368, + 505, + 402 + ], + "lines": [ + { + "bbox": [ + 106, + 367, + 506, + 381 + ], + "spans": [ + { + "bbox": [ + 106, + 367, + 506, + 381 + ], + "score": 1.0, + "content": "Oliver Zendel, Katrin Honauer, Markus Murschitz, Daniel Steininger, and Gustavo Fernan-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 117, + 380, + 504, + 392 + ], + "spans": [ + { + "bbox": [ + 117, + 380, + 504, + 392 + ], + "score": 1.0, + "content": "dez Dominguez. Wilddash-creating hazard-aware benchmarks. In Proceedings of the European", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 117, + 390, + 363, + 403 + ], + "spans": [ + { + "bbox": [ + 117, + 390, + 363, + 403 + ], + "score": 1.0, + "content": "Conference on Computer Vision (ECCV), pp. 402–416, 2018.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22 + }, + { + "type": "text", + "bbox": [ + 105, + 408, + 504, + 442 + ], + "lines": [ + { + "bbox": [ + 105, + 408, + 505, + 423 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 505, + 423 + ], + "score": 1.0, + "content": "Hengshuang Zhao, Jianping Shi, Xiaojuan Qi, Xiaogang Wang, and Jiaya Jia. Pyramid scene", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 419, + 507, + 434 + ], + "spans": [ + { + "bbox": [ + 115, + 419, + 507, + 434 + ], + "score": 1.0, + "content": "parsing network. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 117, + 432, + 191, + 443 + ], + "spans": [ + { + "bbox": [ + 117, + 432, + 191, + 443 + ], + "score": 1.0, + "content": "6230–6239, 2017.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25 + }, + { + "type": "text", + "bbox": [ + 104, + 449, + 505, + 473 + ], + "lines": [ + { + "bbox": [ + 105, + 450, + 505, + 462 + ], + "spans": [ + { + "bbox": [ + 105, + 450, + 505, + 462 + ], + "score": 1.0, + "content": "Bolei Zhou, Agata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba. Places: A 10", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 461, + 355, + 473 + ], + "spans": [ + { + "bbox": [ + 115, + 461, + 355, + 473 + ], + "score": 1.0, + "content": "million image database for scene recognition. PAMI, 2017.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27.5 + }, + { + "type": "text", + "bbox": [ + 107, + 479, + 505, + 514 + ], + "lines": [ + { + "bbox": [ + 105, + 478, + 505, + 494 + ], + "spans": [ + { + "bbox": [ + 105, + 478, + 505, + 494 + ], + "score": 1.0, + "content": "Peng Zhou, Xintong Han, Vlad I Morariu, and Larry S Davis. Learning rich features for image", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 491, + 505, + 504 + ], + "spans": [ + { + "bbox": [ + 116, + 491, + 505, + 504 + ], + "score": 1.0, + "content": "manipulation detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 501, + 259, + 515 + ], + "spans": [ + { + "bbox": [ + 115, + 501, + 259, + 515 + ], + "score": 1.0, + "content": "Recognition, pp. 1053–1061, 2018.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30 + } + ], + "page_idx": 11, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 311, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "12", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 110, + 82, + 504, + 116 + ], + "lines": [ + { + "bbox": [ + 107, + 81, + 506, + 96 + ], + "spans": [ + { + "bbox": [ + 107, + 81, + 506, + 96 + ], + "score": 1.0, + "content": "Tal Ridnik, Hussam Lawen, Asaf Noy, Emanuel Ben Baruch, Gilad Sharir, and Itamar Friedman.", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 116, + 94, + 506, + 106 + ], + "spans": [ + { + "bbox": [ + 116, + 94, + 506, + 106 + ], + "score": 1.0, + "content": "Tresnet: High performance gpu-dedicated architecture. In Proceedings of the IEEE/CVF Winter", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 116, + 104, + 407, + 118 + ], + "spans": [ + { + "bbox": [ + 116, + 104, + 407, + 118 + ], + "score": 1.0, + "content": "Conference on Applications of Computer Vision, pp. 1400–1409, 2021b.", + "type": "text" + } + ], + "index": 2 + } + ], + "index": 1, + "bbox_fs": [ + 107, + 81, + 506, + 118 + ] + }, + { + "type": "text", + "bbox": [ + 109, + 123, + 504, + 156 + ], + "lines": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "spans": [ + { + "bbox": [ + 106, + 123, + 505, + 135 + ], + "score": 1.0, + "content": "Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 115, + 135, + 505, + 147 + ], + "spans": [ + { + "bbox": [ + 115, + 135, + 505, + 147 + ], + "score": 1.0, + "content": "image segmentation. Medical Image Computing and Computer-Assisted Intervention – MICCAI", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 115, + 145, + 449, + 158 + ], + "spans": [ + { + "bbox": [ + 115, + 145, + 449, + 158 + ], + "score": 1.0, + "content": "2015, pp. 234–241, 2015. ISSN 1611-3349. doi: 10.1007/978-3-319-24574-4_28.", + "type": "text" + } + ], + "index": 5 + } + ], + "index": 4, + "bbox_fs": [ + 106, + 123, + 505, + 158 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 164, + 505, + 197 + ], + "lines": [ + { + "bbox": [ + 106, + 164, + 506, + 176 + ], + "spans": [ + { + "bbox": [ + 106, + 164, + 506, + 176 + ], + "score": 1.0, + "content": "Manolis Savva, Angel X. Chang, and Pat Hanrahan. Semantically-Enriched 3D Models for Common-", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 114, + 174, + 507, + 188 + ], + "spans": [ + { + "bbox": [ + 114, + 174, + 507, + 188 + ], + "score": 1.0, + "content": "sense Knowledge. CVPR 2015 Workshop on Functionality, Physics, Intentionality and Causality,", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 114, + 184, + 143, + 199 + ], + "spans": [ + { + "bbox": [ + 114, + 184, + 143, + 199 + ], + "score": 1.0, + "content": "2015.", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 7, + "bbox_fs": [ + 106, + 164, + 507, + 199 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 204, + 506, + 249 + ], + "lines": [ + { + "bbox": [ + 105, + 204, + 506, + 218 + ], + "spans": [ + { + "bbox": [ + 105, + 204, + 506, + 218 + ], + "score": 1.0, + "content": "Bernhard Schölkopf, Robert Williamson, Alex Smola, John Shawe-Taylor, and John Platt. Support", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 116, + 216, + 505, + 228 + ], + "spans": [ + { + "bbox": [ + 116, + 216, + 505, + 228 + ], + "score": 1.0, + "content": "vector method for novelty detection. In Proceedings of the 12th International Conference on", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 114, + 226, + 506, + 240 + ], + "spans": [ + { + "bbox": [ + 114, + 226, + 506, + 240 + ], + "score": 1.0, + "content": "Neural Information Processing Systems, NIPS’99, pp. 582–588, Cambridge, MA, USA, 1999. MIT", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 115, + 237, + 144, + 251 + ], + "spans": [ + { + "bbox": [ + 115, + 237, + 144, + 251 + ], + "score": 1.0, + "content": "Press.", + "type": "text" + } + ], + "index": 12 + } + ], + "index": 10.5, + "bbox_fs": [ + 105, + 204, + 506, + 251 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 256, + 506, + 290 + ], + "lines": [ + { + "bbox": [ + 105, + 256, + 506, + 269 + ], + "spans": [ + { + "bbox": [ + 105, + 256, + 506, + 269 + ], + "score": 1.0, + "content": "Ilya Tolstikhin, Neil Houlsby, Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Thomas Un-", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 116, + 268, + 505, + 280 + ], + "spans": [ + { + "bbox": [ + 116, + 268, + 505, + 280 + ], + "score": 1.0, + "content": "terthiner, Jessica Yung, Daniel Keysers, Jakob Uszkoreit, Mario Lucic, et al. Mlp-mixer: An", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 115, + 277, + 405, + 291 + ], + "spans": [ + { + "bbox": [ + 115, + 277, + 405, + 291 + ], + "score": 1.0, + "content": "all-mlp architecture for vision. arXiv preprint arXiv:2105.01601, 2021.", + "type": "text" + } + ], + "index": 15 + } + ], + "index": 14, + "bbox_fs": [ + 105, + 256, + 506, + 291 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 297, + 505, + 320 + ], + "lines": [ + { + "bbox": [ + 105, + 297, + 506, + 310 + ], + "spans": [ + { + "bbox": [ + 105, + 297, + 506, + 310 + ], + "score": 1.0, + "content": "Fisher Yu, Yinda Zhang, Shuran Song, Ari Seff, and Jianxiong Xiao. LSUN: construction of a", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 115, + 308, + 456, + 321 + ], + "spans": [ + { + "bbox": [ + 115, + 308, + 456, + 321 + ], + "score": 1.0, + "content": "large-scale image dataset using deep learning with humans in the loop. CoRR, 2015.", + "type": "text" + } + ], + "index": 17 + } + ], + "index": 16.5, + "bbox_fs": [ + 105, + 297, + 506, + 321 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 327, + 505, + 361 + ], + "lines": [ + { + "bbox": [ + 105, + 327, + 505, + 339 + ], + "spans": [ + { + "bbox": [ + 105, + 327, + 505, + 339 + ], + "score": 1.0, + "content": "Fisher Yu, Wenqi Xian, Yingying Chen, Fangchen Liu, Mike Liao, Vashisht Madhavan, and Trevor", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 115, + 338, + 506, + 351 + ], + "spans": [ + { + "bbox": [ + 115, + 338, + 506, + 351 + ], + "score": 1.0, + "content": "Darrell. BDD100K: A diverse driving video database with scalable annotation tooling. CoRR,", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 115, + 349, + 210, + 361 + ], + "spans": [ + { + "bbox": [ + 115, + 349, + 210, + 361 + ], + "score": 1.0, + "content": "abs/1805.04687, 2018.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 19, + "bbox_fs": [ + 105, + 327, + 506, + 361 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 368, + 505, + 402 + ], + "lines": [ + { + "bbox": [ + 106, + 367, + 506, + 381 + ], + "spans": [ + { + "bbox": [ + 106, + 367, + 506, + 381 + ], + "score": 1.0, + "content": "Oliver Zendel, Katrin Honauer, Markus Murschitz, Daniel Steininger, and Gustavo Fernan-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 117, + 380, + 504, + 392 + ], + "spans": [ + { + "bbox": [ + 117, + 380, + 504, + 392 + ], + "score": 1.0, + "content": "dez Dominguez. Wilddash-creating hazard-aware benchmarks. In Proceedings of the European", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 117, + 390, + 363, + 403 + ], + "spans": [ + { + "bbox": [ + 117, + 390, + 363, + 403 + ], + "score": 1.0, + "content": "Conference on Computer Vision (ECCV), pp. 402–416, 2018.", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 22, + "bbox_fs": [ + 106, + 367, + 506, + 403 + ] + }, + { + "type": "text", + "bbox": [ + 105, + 408, + 504, + 442 + ], + "lines": [ + { + "bbox": [ + 105, + 408, + 505, + 423 + ], + "spans": [ + { + "bbox": [ + 105, + 408, + 505, + 423 + ], + "score": 1.0, + "content": "Hengshuang Zhao, Jianping Shi, Xiaojuan Qi, Xiaogang Wang, and Jiaya Jia. Pyramid scene", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 115, + 419, + 507, + 434 + ], + "spans": [ + { + "bbox": [ + 115, + 419, + 507, + 434 + ], + "score": 1.0, + "content": "parsing network. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 117, + 432, + 191, + 443 + ], + "spans": [ + { + "bbox": [ + 117, + 432, + 191, + 443 + ], + "score": 1.0, + "content": "6230–6239, 2017.", + "type": "text" + } + ], + "index": 26 + } + ], + "index": 25, + "bbox_fs": [ + 105, + 408, + 507, + 443 + ] + }, + { + "type": "text", + "bbox": [ + 104, + 449, + 505, + 473 + ], + "lines": [ + { + "bbox": [ + 105, + 450, + 505, + 462 + ], + "spans": [ + { + "bbox": [ + 105, + 450, + 505, + 462 + ], + "score": 1.0, + "content": "Bolei Zhou, Agata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba. Places: A 10", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 115, + 461, + 355, + 473 + ], + "spans": [ + { + "bbox": [ + 115, + 461, + 355, + 473 + ], + "score": 1.0, + "content": "million image database for scene recognition. PAMI, 2017.", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 27.5, + "bbox_fs": [ + 105, + 450, + 505, + 473 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 479, + 505, + 514 + ], + "lines": [ + { + "bbox": [ + 105, + 478, + 505, + 494 + ], + "spans": [ + { + "bbox": [ + 105, + 478, + 505, + 494 + ], + "score": 1.0, + "content": "Peng Zhou, Xintong Han, Vlad I Morariu, and Larry S Davis. Learning rich features for image", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 116, + 491, + 505, + 504 + ], + "spans": [ + { + "bbox": [ + 116, + 491, + 505, + 504 + ], + "score": 1.0, + "content": "manipulation detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 115, + 501, + 259, + 515 + ], + "spans": [ + { + "bbox": [ + 115, + 501, + 259, + 515 + ], + "score": 1.0, + "content": "Recognition, pp. 1053–1061, 2018.", + "type": "text" + } + ], + "index": 31 + } + ], + "index": 30, + "bbox_fs": [ + 105, + 478, + 505, + 515 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 82, + 183, + 93 + ], + "lines": [ + { + "bbox": [ + 105, + 79, + 185, + 97 + ], + "spans": [ + { + "bbox": [ + 105, + 79, + 185, + 97 + ], + "score": 1.0, + "content": "A APPENDIX", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table", + "bbox": [ + 106, + 108, + 506, + 278 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 108, + 506, + 278 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 108, + 506, + 278 + ], + "spans": [ + { + "bbox": [ + 106, + 108, + 506, + 278 + ], + "score": 0.985, + "html": "
DestFPR95↓AUROC ↑AUPR↑
DinBM DKBMDKBMDK
Gaussian2054100100 9798329252521985579
1eeeeeeRademacher2144158998 9893706254
Blobs26327288079 379917793
Textures685674598087 7685361648
LSUN6663596075 777679221938
Places3656459637279 83797927322446
Mean41.335.84636.185.287.276.988.7 3745.830.559.7
PPssse536Gaussian10671129396 35931624216
Rademacher20109118993 1010011 515.91.688
Blobs59688277298 159341231
Textures86728774657943 7911112
Places6988899291616452 65636
Mean5336.685.840.97685.831.185.819.2230.5
", + "type": "table", + "image_path": "ac83fc93b0bb49458edab1aad6f2275b1266913fbb8c2964a80965bd28e142e4.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 106, + 108, + 506, + 164.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 164.66666666666666, + 506, + 221.33333333333331 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 106, + 221.33333333333331, + 506, + 278.0 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 106, + 286, + 506, + 330 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 286, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 286, + 505, + 298 + ], + "score": 1.0, + "content": "Table 5: B is for the maximum softmax probability baseline, M is for maximum logit, D is for the", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 297, + 506, + 309 + ], + "spans": [ + { + "bbox": [ + 105, + 297, + 506, + 309 + ], + "score": 1.0, + "content": "method in DeVries & Taylor (2018), and K is our own KL method described below. Both M and K", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 309, + 506, + 322 + ], + "spans": [ + { + "bbox": [ + 105, + 309, + 506, + 322 + ], + "score": 1.0, + "content": "are ours. Results are on ImageNet and Places365. All values are percentages and are rounded so that", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 319, + 191, + 331 + ], + "spans": [ + { + "bbox": [ + 106, + 319, + 191, + 331 + ], + "score": 1.0, + "content": "99.95 rounds to 100.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5.5 + } + ], + "index": 3.75 + }, + { + "type": "title", + "bbox": [ + 107, + 345, + 370, + 359 + ], + "lines": [ + { + "bbox": [ + 105, + 344, + 371, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 371, + 361 + ], + "score": 1.0, + "content": "B FULL MULTICLASS OOD DETECTION RESULTS", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 106, + 371, + 505, + 503 + ], + "lines": [ + { + "bbox": [ + 105, + 370, + 506, + 384 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 506, + 384 + ], + "score": 1.0, + "content": "Datasets. To evaluate the MSP baseline out-of-distribution detector and the MaxLogit detector,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 383, + 506, + 395 + ], + "spans": [ + { + "bbox": [ + 106, + 383, + 506, + 395 + ], + "score": 1.0, + "content": "we use the ImageNet-1K object recognition dataset and Places365 scene recognition dataset as", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 393, + 505, + 406 + ], + "spans": [ + { + "bbox": [ + 106, + 393, + 202, + 406 + ], + "score": 1.0, + "content": "in-distribution datasets", + "type": "text" + }, + { + "bbox": [ + 202, + 393, + 217, + 404 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { i n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 217, + 393, + 415, + 406 + ], + "score": 1.0, + "content": ". We use several out-of-distribution test datasets", + "type": "text" + }, + { + "bbox": [ + 416, + 394, + 434, + 404 + ], + "score": 0.89, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 393, + 505, + 406 + ], + "score": 1.0, + "content": ", all of which are", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 405, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 105, + 405, + 505, + 416 + ], + "score": 1.0, + "content": "unseen during training. The first out-of-distribution dataset is Gaussian noise, where each example’s", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 415, + 506, + 428 + ], + "spans": [ + { + "bbox": [ + 106, + 415, + 225, + 428 + ], + "score": 1.0, + "content": "pixels are i.i.d. sampled from", + "type": "text" + }, + { + "bbox": [ + 225, + 415, + 266, + 427 + ], + "score": 0.92, + "content": "\\mathcal { N } ( 0 , 0 . 5 )", + "type": "inline_equation" + }, + { + "bbox": [ + 266, + 415, + 406, + 428 + ], + "score": 1.0, + "content": "and clipped to be contained within", + "type": "text" + }, + { + "bbox": [ + 407, + 415, + 435, + 427 + ], + "score": 0.88, + "content": "[ - 1 , 1 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 415, + 506, + 428 + ], + "score": 1.0, + "content": ". Another type of", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 426, + 506, + 438 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 465, + 438 + ], + "score": 1.0, + "content": "test-time noise is Rademacher noise, in which each pixel is i.i.d. sampled from 2 · Bernoulli", + "type": "text" + }, + { + "bbox": [ + 466, + 426, + 502, + 438 + ], + "score": 0.66, + "content": "( 0 . 5 ) - 1", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 426, + 506, + 438 + ], + "score": 1.0, + "content": ",", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 104, + 436, + 506, + 451 + ], + "spans": [ + { + "bbox": [ + 104, + 436, + 194, + 451 + ], + "score": 1.0, + "content": "i.e. each pixel is 1 or", + "type": "text" + }, + { + "bbox": [ + 194, + 438, + 208, + 448 + ], + "score": 0.76, + "content": "- 1", + "type": "inline_equation" + }, + { + "bbox": [ + 208, + 436, + 506, + 451 + ], + "score": 1.0, + "content": "with equal probability. Blob examples are more structured than noise; they", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 447, + 506, + 461 + ], + "spans": [ + { + "bbox": [ + 105, + 447, + 506, + 461 + ], + "score": 1.0, + "content": "are algorithmically generated blob images. Meanwhile, Textures is a dataset consisting in images of", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 459, + 506, + 471 + ], + "spans": [ + { + "bbox": [ + 106, + 459, + 506, + 471 + ], + "score": 1.0, + "content": "describable textures (Cimpoi et al., 2014). When evaluating the ImageNet-1K detector, we use LSUN", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 469, + 506, + 483 + ], + "spans": [ + { + "bbox": [ + 105, + 469, + 352, + 483 + ], + "score": 1.0, + "content": "images, a scene recognition dataset (Yu et al., 2015). Our final", + "type": "text" + }, + { + "bbox": [ + 352, + 470, + 370, + 481 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 371, + 469, + 506, + 483 + ], + "score": 1.0, + "content": "is Places69, a scene classification", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 481, + 505, + 493 + ], + "spans": [ + { + "bbox": [ + 106, + 481, + 505, + 493 + ], + "score": 1.0, + "content": "dataset that does not share classes with Places365. In all, we evaluate against out-of-distribution", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 491, + 308, + 505 + ], + "spans": [ + { + "bbox": [ + 105, + 491, + 308, + 505 + ], + "score": 1.0, + "content": "examples spanning synthetic and realistic images.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 14.5 + }, + { + "type": "text", + "bbox": [ + 107, + 515, + 505, + 559 + ], + "lines": [ + { + "bbox": [ + 105, + 514, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 514, + 505, + 528 + ], + "score": 1.0, + "content": "KL Matching Method. To verify our intuitions that led us to develop the MaxLogit detector, we", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 526, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 506, + 540 + ], + "score": 1.0, + "content": "developed a less convenient but similarly powerful technique applicable for the multiclass setting.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 537, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 550 + ], + "score": 1.0, + "content": "Recall that some classes tend to be predicted with low confidence and others high confidence. The", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 549, + 372, + 561 + ], + "spans": [ + { + "bbox": [ + 106, + 549, + 372, + 561 + ], + "score": 1.0, + "content": "shape of predicted posterior distributions is often class dependent.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 22.5 + }, + { + "type": "text", + "bbox": [ + 106, + 565, + 505, + 665 + ], + "lines": [ + { + "bbox": [ + 106, + 566, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 106, + 566, + 505, + 577 + ], + "score": 1.0, + "content": "We capture the typical shape of each class’s posterior distribution and form posterior distribution", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 577, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 106, + 577, + 505, + 588 + ], + "score": 1.0, + "content": "templates for each class. During test time, the network’s softmax posterior distribution is compared", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 588, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 459, + 600 + ], + "score": 1.0, + "content": "to these templates and an anomaly score is generated. More concretely, we compute", + "type": "text" + }, + { + "bbox": [ + 459, + 588, + 466, + 597 + ], + "score": 0.8, + "content": "k", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 588, + 505, + 600 + ], + "score": 1.0, + "content": "different", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 597, + 506, + 612 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 160, + 612 + ], + "score": 1.0, + "content": "distributions", + "type": "text" + }, + { + "bbox": [ + 160, + 599, + 171, + 609 + ], + "score": 0.88, + "content": "d _ { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 172, + 597, + 297, + 612 + ], + "score": 1.0, + "content": ", one for each class. We write", + "type": "text" + }, + { + "bbox": [ + 297, + 598, + 390, + 610 + ], + "score": 0.92, + "content": "d _ { k } = \\mathbb { E } _ { x ^ { \\prime } \\sim \\mathcal { X } _ { \\mathrm { v a l } } } [ p ( y | x ^ { \\prime } ) ]", + "type": "inline_equation" + }, + { + "bbox": [ + 391, + 597, + 420, + 612 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 420, + 598, + 506, + 610 + ], + "score": 0.92, + "content": "k = \\mathrm { a r g m a x } _ { k } p ( y =", + "type": "inline_equation" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 608, + 506, + 622 + ], + "spans": [ + { + "bbox": [ + 106, + 609, + 133, + 621 + ], + "score": 0.9, + "content": "k \\mid x ^ { \\prime } )", + "type": "inline_equation" + }, + { + "bbox": [ + 134, + 608, + 242, + 622 + ], + "score": 1.0, + "content": ". Then for a new test input", + "type": "text" + }, + { + "bbox": [ + 243, + 612, + 249, + 619 + ], + "score": 0.73, + "content": "x", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 608, + 382, + 622 + ], + "score": 1.0, + "content": ", we calculate the anomaly score", + "type": "text" + }, + { + "bbox": [ + 382, + 609, + 478, + 621 + ], + "score": 0.74, + "content": "\\mathrm { { \\ddot { m i n } } } _ { k } ~ \\mathrm { K L } [ p ( y \\mid x ) \\mid \\mid d _ { k } ]", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 608, + 506, + 622 + ], + "score": 1.0, + "content": "rather", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 619, + 506, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 619, + 199, + 633 + ], + "score": 1.0, + "content": "than the MSP baseline", + "type": "text" + }, + { + "bbox": [ + 199, + 620, + 285, + 632 + ], + "score": 0.89, + "content": "- \\operatorname* { m a x } _ { k } p ( y = k \\mid x ) .", + "type": "inline_equation" + }, + { + "bbox": [ + 286, + 619, + 506, + 633 + ], + "score": 1.0, + "content": ". Note that we utilize the validation dataset, but our KL", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 631, + 506, + 644 + ], + "spans": [ + { + "bbox": [ + 105, + 631, + 506, + 644 + ], + "score": 1.0, + "content": "matching method does not require the validation dataset’s labels. That said, our KL matching method", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 104, + 640, + 506, + 656 + ], + "spans": [ + { + "bbox": [ + 104, + 640, + 506, + 656 + ], + "score": 1.0, + "content": "is less convenient than our MaxLogit technique, and the two perform similarly. Since this technique", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 652, + 464, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 652, + 464, + 667 + ], + "score": 1.0, + "content": "requires more data than MaxLogit, we opt to simply use the MaxLogit in the main paper.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 29 + }, + { + "type": "text", + "bbox": [ + 107, + 677, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "Results. Observe that the proposed MaxLogit method outperforms the maximum softmax probabil-", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 686, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 686, + 505, + 701 + ], + "score": 1.0, + "content": "ity baseline for all three metrics on both ImageNet and Places365. These results were computed using", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "a ResNet-50 trained on either ImageNet-1K or Places365. In the case of Places365, the AUROC", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 190, + 722 + ], + "score": 1.0, + "content": "improvement is over", + "type": "text" + }, + { + "bbox": [ + 190, + 710, + 209, + 720 + ], + "score": 0.87, + "content": "10 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 209, + 709, + 506, + 722 + ], + "score": 1.0, + "content": ". We note that the utility of the maximum logit could not be appreciated as", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "easily in previous work’s small-scale settings. For example, using the small-scale CIFAR-10 setup of", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 36 + } + ], + "page_idx": 12, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "title", + "bbox": [ + 108, + 82, + 183, + 93 + ], + "lines": [ + { + "bbox": [ + 105, + 79, + 185, + 97 + ], + "spans": [ + { + "bbox": [ + 105, + 79, + 185, + 97 + ], + "score": 1.0, + "content": "A APPENDIX", + "type": "text" + } + ], + "index": 0 + } + ], + "index": 0 + }, + { + "type": "table", + "bbox": [ + 106, + 108, + 506, + 278 + ], + "blocks": [ + { + "type": "table_body", + "bbox": [ + 106, + 108, + 506, + 278 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 106, + 108, + 506, + 278 + ], + "spans": [ + { + "bbox": [ + 106, + 108, + 506, + 278 + ], + "score": 0.985, + "html": "
DestFPR95↓AUROC ↑AUPR↑
DinBM DKBMDKBMDK
Gaussian2054100100 9798329252521985579
1eeeeeeRademacher2144158998 9893706254
Blobs26327288079 379917793
Textures685674598087 7685361648
LSUN6663596075 777679221938
Places3656459637279 83797927322446
Mean41.335.84636.185.287.276.988.7 3745.830.559.7
PPssse536Gaussian10671129396 35931624216
Rademacher20109118993 1010011 515.91.688
Blobs59688277298 159341231
Textures86728774657943 7911112
Places6988899291616452 65636
Mean5336.685.840.97685.831.185.819.2230.5
", + "type": "table", + "image_path": "ac83fc93b0bb49458edab1aad6f2275b1266913fbb8c2964a80965bd28e142e4.jpg" + } + ] + } + ], + "index": 2, + "virtual_lines": [ + { + "bbox": [ + 106, + 108, + 506, + 164.66666666666666 + ], + "spans": [], + "index": 1 + }, + { + "bbox": [ + 106, + 164.66666666666666, + 506, + 221.33333333333331 + ], + "spans": [], + "index": 2 + }, + { + "bbox": [ + 106, + 221.33333333333331, + 506, + 278.0 + ], + "spans": [], + "index": 3 + } + ] + }, + { + "type": "table_caption", + "bbox": [ + 106, + 286, + 506, + 330 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 105, + 286, + 505, + 298 + ], + "spans": [ + { + "bbox": [ + 105, + 286, + 505, + 298 + ], + "score": 1.0, + "content": "Table 5: B is for the maximum softmax probability baseline, M is for maximum logit, D is for the", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 297, + 506, + 309 + ], + "spans": [ + { + "bbox": [ + 105, + 297, + 506, + 309 + ], + "score": 1.0, + "content": "method in DeVries & Taylor (2018), and K is our own KL method described below. Both M and K", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 309, + 506, + 322 + ], + "spans": [ + { + "bbox": [ + 105, + 309, + 506, + 322 + ], + "score": 1.0, + "content": "are ours. Results are on ImageNet and Places365. All values are percentages and are rounded so that", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 319, + 191, + 331 + ], + "spans": [ + { + "bbox": [ + 106, + 319, + 191, + 331 + ], + "score": 1.0, + "content": "99.95 rounds to 100.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 5.5 + } + ], + "index": 3.75 + }, + { + "type": "title", + "bbox": [ + 107, + 345, + 370, + 359 + ], + "lines": [ + { + "bbox": [ + 105, + 344, + 371, + 361 + ], + "spans": [ + { + "bbox": [ + 105, + 344, + 371, + 361 + ], + "score": 1.0, + "content": "B FULL MULTICLASS OOD DETECTION RESULTS", + "type": "text" + } + ], + "index": 8 + } + ], + "index": 8 + }, + { + "type": "text", + "bbox": [ + 106, + 371, + 505, + 503 + ], + "lines": [ + { + "bbox": [ + 105, + 370, + 506, + 384 + ], + "spans": [ + { + "bbox": [ + 105, + 370, + 506, + 384 + ], + "score": 1.0, + "content": "Datasets. To evaluate the MSP baseline out-of-distribution detector and the MaxLogit detector,", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 383, + 506, + 395 + ], + "spans": [ + { + "bbox": [ + 106, + 383, + 506, + 395 + ], + "score": 1.0, + "content": "we use the ImageNet-1K object recognition dataset and Places365 scene recognition dataset as", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 393, + 505, + 406 + ], + "spans": [ + { + "bbox": [ + 106, + 393, + 202, + 406 + ], + "score": 1.0, + "content": "in-distribution datasets", + "type": "text" + }, + { + "bbox": [ + 202, + 393, + 217, + 404 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { i n } }", + "type": "inline_equation" + }, + { + "bbox": [ + 217, + 393, + 415, + 406 + ], + "score": 1.0, + "content": ". We use several out-of-distribution test datasets", + "type": "text" + }, + { + "bbox": [ + 416, + 394, + 434, + 404 + ], + "score": 0.89, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 434, + 393, + 505, + 406 + ], + "score": 1.0, + "content": ", all of which are", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 405, + 505, + 416 + ], + "spans": [ + { + "bbox": [ + 105, + 405, + 505, + 416 + ], + "score": 1.0, + "content": "unseen during training. The first out-of-distribution dataset is Gaussian noise, where each example’s", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 415, + 506, + 428 + ], + "spans": [ + { + "bbox": [ + 106, + 415, + 225, + 428 + ], + "score": 1.0, + "content": "pixels are i.i.d. sampled from", + "type": "text" + }, + { + "bbox": [ + 225, + 415, + 266, + 427 + ], + "score": 0.92, + "content": "\\mathcal { N } ( 0 , 0 . 5 )", + "type": "inline_equation" + }, + { + "bbox": [ + 266, + 415, + 406, + 428 + ], + "score": 1.0, + "content": "and clipped to be contained within", + "type": "text" + }, + { + "bbox": [ + 407, + 415, + 435, + 427 + ], + "score": 0.88, + "content": "[ - 1 , 1 ]", + "type": "inline_equation" + }, + { + "bbox": [ + 435, + 415, + 506, + 428 + ], + "score": 1.0, + "content": ". Another type of", + "type": "text" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 426, + 506, + 438 + ], + "spans": [ + { + "bbox": [ + 105, + 426, + 465, + 438 + ], + "score": 1.0, + "content": "test-time noise is Rademacher noise, in which each pixel is i.i.d. sampled from 2 · Bernoulli", + "type": "text" + }, + { + "bbox": [ + 466, + 426, + 502, + 438 + ], + "score": 0.66, + "content": "( 0 . 5 ) - 1", + "type": "inline_equation" + }, + { + "bbox": [ + 503, + 426, + 506, + 438 + ], + "score": 1.0, + "content": ",", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 104, + 436, + 506, + 451 + ], + "spans": [ + { + "bbox": [ + 104, + 436, + 194, + 451 + ], + "score": 1.0, + "content": "i.e. each pixel is 1 or", + "type": "text" + }, + { + "bbox": [ + 194, + 438, + 208, + 448 + ], + "score": 0.76, + "content": "- 1", + "type": "inline_equation" + }, + { + "bbox": [ + 208, + 436, + 506, + 451 + ], + "score": 1.0, + "content": "with equal probability. Blob examples are more structured than noise; they", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 447, + 506, + 461 + ], + "spans": [ + { + "bbox": [ + 105, + 447, + 506, + 461 + ], + "score": 1.0, + "content": "are algorithmically generated blob images. Meanwhile, Textures is a dataset consisting in images of", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 459, + 506, + 471 + ], + "spans": [ + { + "bbox": [ + 106, + 459, + 506, + 471 + ], + "score": 1.0, + "content": "describable textures (Cimpoi et al., 2014). When evaluating the ImageNet-1K detector, we use LSUN", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 469, + 506, + 483 + ], + "spans": [ + { + "bbox": [ + 105, + 469, + 352, + 483 + ], + "score": 1.0, + "content": "images, a scene recognition dataset (Yu et al., 2015). Our final", + "type": "text" + }, + { + "bbox": [ + 352, + 470, + 370, + 481 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 371, + 469, + 506, + 483 + ], + "score": 1.0, + "content": "is Places69, a scene classification", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 481, + 505, + 493 + ], + "spans": [ + { + "bbox": [ + 106, + 481, + 505, + 493 + ], + "score": 1.0, + "content": "dataset that does not share classes with Places365. In all, we evaluate against out-of-distribution", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 491, + 308, + 505 + ], + "spans": [ + { + "bbox": [ + 105, + 491, + 308, + 505 + ], + "score": 1.0, + "content": "examples spanning synthetic and realistic images.", + "type": "text" + } + ], + "index": 20 + } + ], + "index": 14.5, + "bbox_fs": [ + 104, + 370, + 506, + 505 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 515, + 505, + 559 + ], + "lines": [ + { + "bbox": [ + 105, + 514, + 505, + 528 + ], + "spans": [ + { + "bbox": [ + 105, + 514, + 505, + 528 + ], + "score": 1.0, + "content": "KL Matching Method. To verify our intuitions that led us to develop the MaxLogit detector, we", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 526, + 506, + 540 + ], + "spans": [ + { + "bbox": [ + 105, + 526, + 506, + 540 + ], + "score": 1.0, + "content": "developed a less convenient but similarly powerful technique applicable for the multiclass setting.", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 537, + 505, + 550 + ], + "spans": [ + { + "bbox": [ + 105, + 537, + 505, + 550 + ], + "score": 1.0, + "content": "Recall that some classes tend to be predicted with low confidence and others high confidence. The", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 106, + 549, + 372, + 561 + ], + "spans": [ + { + "bbox": [ + 106, + 549, + 372, + 561 + ], + "score": 1.0, + "content": "shape of predicted posterior distributions is often class dependent.", + "type": "text" + } + ], + "index": 24 + } + ], + "index": 22.5, + "bbox_fs": [ + 105, + 514, + 506, + 561 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 565, + 505, + 665 + ], + "lines": [ + { + "bbox": [ + 106, + 566, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 106, + 566, + 505, + 577 + ], + "score": 1.0, + "content": "We capture the typical shape of each class’s posterior distribution and form posterior distribution", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 106, + 577, + 505, + 588 + ], + "spans": [ + { + "bbox": [ + 106, + 577, + 505, + 588 + ], + "score": 1.0, + "content": "templates for each class. During test time, the network’s softmax posterior distribution is compared", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 106, + 588, + 505, + 600 + ], + "spans": [ + { + "bbox": [ + 106, + 588, + 459, + 600 + ], + "score": 1.0, + "content": "to these templates and an anomaly score is generated. More concretely, we compute", + "type": "text" + }, + { + "bbox": [ + 459, + 588, + 466, + 597 + ], + "score": 0.8, + "content": "k", + "type": "inline_equation" + }, + { + "bbox": [ + 466, + 588, + 505, + 600 + ], + "score": 1.0, + "content": "different", + "type": "text" + } + ], + "index": 27 + }, + { + "bbox": [ + 105, + 597, + 506, + 612 + ], + "spans": [ + { + "bbox": [ + 105, + 597, + 160, + 612 + ], + "score": 1.0, + "content": "distributions", + "type": "text" + }, + { + "bbox": [ + 160, + 599, + 171, + 609 + ], + "score": 0.88, + "content": "d _ { k }", + "type": "inline_equation" + }, + { + "bbox": [ + 172, + 597, + 297, + 612 + ], + "score": 1.0, + "content": ", one for each class. We write", + "type": "text" + }, + { + "bbox": [ + 297, + 598, + 390, + 610 + ], + "score": 0.92, + "content": "d _ { k } = \\mathbb { E } _ { x ^ { \\prime } \\sim \\mathcal { X } _ { \\mathrm { v a l } } } [ p ( y | x ^ { \\prime } ) ]", + "type": "inline_equation" + }, + { + "bbox": [ + 391, + 597, + 420, + 612 + ], + "score": 1.0, + "content": "where", + "type": "text" + }, + { + "bbox": [ + 420, + 598, + 506, + 610 + ], + "score": 0.92, + "content": "k = \\mathrm { a r g m a x } _ { k } p ( y =", + "type": "inline_equation" + } + ], + "index": 28 + }, + { + "bbox": [ + 106, + 608, + 506, + 622 + ], + "spans": [ + { + "bbox": [ + 106, + 609, + 133, + 621 + ], + "score": 0.9, + "content": "k \\mid x ^ { \\prime } )", + "type": "inline_equation" + }, + { + "bbox": [ + 134, + 608, + 242, + 622 + ], + "score": 1.0, + "content": ". Then for a new test input", + "type": "text" + }, + { + "bbox": [ + 243, + 612, + 249, + 619 + ], + "score": 0.73, + "content": "x", + "type": "inline_equation" + }, + { + "bbox": [ + 249, + 608, + 382, + 622 + ], + "score": 1.0, + "content": ", we calculate the anomaly score", + "type": "text" + }, + { + "bbox": [ + 382, + 609, + 478, + 621 + ], + "score": 0.74, + "content": "\\mathrm { { \\ddot { m i n } } } _ { k } ~ \\mathrm { K L } [ p ( y \\mid x ) \\mid \\mid d _ { k } ]", + "type": "inline_equation" + }, + { + "bbox": [ + 478, + 608, + 506, + 622 + ], + "score": 1.0, + "content": "rather", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 619, + 506, + 633 + ], + "spans": [ + { + "bbox": [ + 105, + 619, + 199, + 633 + ], + "score": 1.0, + "content": "than the MSP baseline", + "type": "text" + }, + { + "bbox": [ + 199, + 620, + 285, + 632 + ], + "score": 0.89, + "content": "- \\operatorname* { m a x } _ { k } p ( y = k \\mid x ) .", + "type": "inline_equation" + }, + { + "bbox": [ + 286, + 619, + 506, + 633 + ], + "score": 1.0, + "content": ". Note that we utilize the validation dataset, but our KL", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 105, + 631, + 506, + 644 + ], + "spans": [ + { + "bbox": [ + 105, + 631, + 506, + 644 + ], + "score": 1.0, + "content": "matching method does not require the validation dataset’s labels. That said, our KL matching method", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 104, + 640, + 506, + 656 + ], + "spans": [ + { + "bbox": [ + 104, + 640, + 506, + 656 + ], + "score": 1.0, + "content": "is less convenient than our MaxLogit technique, and the two perform similarly. Since this technique", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 652, + 464, + 667 + ], + "spans": [ + { + "bbox": [ + 105, + 652, + 464, + 667 + ], + "score": 1.0, + "content": "requires more data than MaxLogit, we opt to simply use the MaxLogit in the main paper.", + "type": "text" + } + ], + "index": 33 + } + ], + "index": 29, + "bbox_fs": [ + 104, + 566, + 506, + 667 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 677, + 505, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "Results. Observe that the proposed MaxLogit method outperforms the maximum softmax probabil-", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 105, + 686, + 505, + 701 + ], + "spans": [ + { + "bbox": [ + 105, + 686, + 505, + 701 + ], + "score": 1.0, + "content": "ity baseline for all three metrics on both ImageNet and Places365. These results were computed using", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 699, + 505, + 711 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 505, + 711 + ], + "score": 1.0, + "content": "a ResNet-50 trained on either ImageNet-1K or Places365. In the case of Places365, the AUROC", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 105, + 709, + 506, + 722 + ], + "spans": [ + { + "bbox": [ + 105, + 709, + 190, + 722 + ], + "score": 1.0, + "content": "improvement is over", + "type": "text" + }, + { + "bbox": [ + 190, + 710, + 209, + 720 + ], + "score": 0.87, + "content": "10 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 209, + 709, + 506, + 722 + ], + "score": 1.0, + "content": ". We note that the utility of the maximum logit could not be appreciated as", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "spans": [ + { + "bbox": [ + 105, + 720, + 506, + 733 + ], + "score": 1.0, + "content": "easily in previous work’s small-scale settings. For example, using the small-scale CIFAR-10 setup of", + "type": "text" + } + ], + "index": 38 + } + ], + "index": 36, + "bbox_fs": [ + 105, + 676, + 506, + 733 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 159 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 445, + 95 + ], + "score": 1.0, + "content": "Hendrycks et al. Hendrycks et al. (2019a), the MSP attains an average AUROC of", + "type": "text" + }, + { + "bbox": [ + 446, + 82, + 479, + 93 + ], + "score": 0.88, + "content": "9 0 . 0 8 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "while", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 93, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 105, + 93, + 213, + 106 + ], + "score": 1.0, + "content": "the maximum logit attains", + "type": "text" + }, + { + "bbox": [ + 213, + 94, + 245, + 104 + ], + "score": 0.88, + "content": "9 0 . 2 2 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 245, + 93, + 281, + 106 + ], + "score": 1.0, + "content": ", a minor", + "type": "text" + }, + { + "bbox": [ + 282, + 94, + 309, + 104 + ], + "score": 0.87, + "content": "0 . 1 4 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 93, + 505, + 106 + ], + "score": 1.0, + "content": "difference. However, in a large-scale setting, the", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 506, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 198, + 117 + ], + "score": 1.0, + "content": "difference can be over", + "type": "text" + }, + { + "bbox": [ + 198, + 105, + 217, + 115 + ], + "score": 0.87, + "content": "10 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 217, + 104, + 273, + 117 + ], + "score": 1.0, + "content": "on individual", + "type": "text" + }, + { + "bbox": [ + 273, + 105, + 292, + 115 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 292, + 104, + 506, + 117 + ], + "score": 1.0, + "content": "datasets. We are not claiming that utilizing the maxi-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 115, + 505, + 128 + ], + "spans": [ + { + "bbox": [ + 105, + 115, + 505, + 128 + ], + "score": 1.0, + "content": "mum logit is a mathematically innovative formulation, only that it serves as a consistently powerful", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 126, + 505, + 139 + ], + "spans": [ + { + "bbox": [ + 105, + 126, + 505, + 139 + ], + "score": 1.0, + "content": "baseline for large-scale settings that went unappreciated in small-scale settings. In consequence, we", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "spans": [ + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "score": 1.0, + "content": "suggest using the maximum logit as a new baseline for large-scale multi-class out-of-distribution", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 147, + 148, + 161 + ], + "spans": [ + { + "bbox": [ + 105, + 147, + 148, + 161 + ], + "score": 1.0, + "content": "detection.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 3 + }, + { + "type": "text", + "bbox": [ + 107, + 174, + 505, + 306 + ], + "lines": [ + { + "bbox": [ + 106, + 174, + 505, + 186 + ], + "spans": [ + { + "bbox": [ + 106, + 174, + 505, + 186 + ], + "score": 1.0, + "content": "Overview of Other Detection Methods. There are other techniques in out-of-distribution detection", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 186, + 505, + 197 + ], + "spans": [ + { + "bbox": [ + 106, + 186, + 505, + 197 + ], + "score": 1.0, + "content": "which require other assumptions such as more training data. For instance, Hendrycks et al. (2019a);", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 195, + 505, + 209 + ], + "spans": [ + { + "bbox": [ + 105, + 195, + 505, + 209 + ], + "score": 1.0, + "content": "Mohseni et al. (2020) use additional training data labeled as out-of-distribution, and the MaxLogit", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 207, + 505, + 219 + ], + "spans": [ + { + "bbox": [ + 106, + 207, + 505, + 219 + ], + "score": 1.0, + "content": "technique can be naturally extended should such data be available. Hendrycks et al. (2019c) use", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 219, + 505, + 231 + ], + "spans": [ + { + "bbox": [ + 106, + 219, + 505, + 231 + ], + "score": 1.0, + "content": "rotation prediction and self-supervised learning, but we found that scaling this to the ImageNet", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 229, + 505, + 241 + ], + "spans": [ + { + "bbox": [ + 105, + 229, + 505, + 241 + ], + "score": 1.0, + "content": "multiclass setting did not produce strong results. The MSP baseline trained with auxiliary rotation", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 240, + 505, + 253 + ], + "spans": [ + { + "bbox": [ + 106, + 240, + 228, + 253 + ], + "score": 1.0, + "content": "prediction has an AUROC of", + "type": "text" + }, + { + "bbox": [ + 228, + 240, + 256, + 251 + ], + "score": 0.87, + "content": "5 9 . 1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 256, + 240, + 387, + 253 + ], + "score": 1.0, + "content": ", and with MaxLogit it attains a", + "type": "text" + }, + { + "bbox": [ + 387, + 240, + 415, + 251 + ], + "score": 0.87, + "content": "7 3 . 6 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 416, + 240, + 484, + 253 + ], + "score": 1.0, + "content": "AUROC, over a", + "type": "text" + }, + { + "bbox": [ + 485, + 240, + 505, + 251 + ], + "score": 0.86, + "content": "10 \\%", + "type": "inline_equation" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 250, + 507, + 264 + ], + "spans": [ + { + "bbox": [ + 105, + 250, + 507, + 264 + ], + "score": 1.0, + "content": "absolute improvement with MaxLogit. Nonetheless this technique did not straightforwardly scale,", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 261, + 506, + 275 + ], + "spans": [ + { + "bbox": [ + 105, + 261, + 506, + 275 + ], + "score": 1.0, + "content": "as the network is better without auxiliary rotation prediction. Likewise, Lee et al. (2018b) propose", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 273, + 505, + 285 + ], + "spans": [ + { + "bbox": [ + 105, + 273, + 505, + 285 + ], + "score": 1.0, + "content": "to use Mahalanobis distances, but in scaling this to 1000 classes, we consistently encountered NaN", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 284, + 505, + 297 + ], + "spans": [ + { + "bbox": [ + 106, + 284, + 505, + 297 + ], + "score": 1.0, + "content": "errors due to high condition numbers. This shows the importance of ensuring that out-of-distribution", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 295, + 193, + 307 + ], + "spans": [ + { + "bbox": [ + 106, + 295, + 193, + 307 + ], + "score": 1.0, + "content": "techniques can scale.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 12.5 + }, + { + "type": "text", + "bbox": [ + 106, + 312, + 505, + 411 + ], + "lines": [ + { + "bbox": [ + 106, + 312, + 505, + 324 + ], + "spans": [ + { + "bbox": [ + 106, + 312, + 505, + 324 + ], + "score": 1.0, + "content": "ODIN Liang et al. (2018) assumes that, for each OOD example source, we can tune hyperparameters", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 322, + 505, + 335 + ], + "spans": [ + { + "bbox": [ + 105, + 322, + 505, + 335 + ], + "score": 1.0, + "content": "for detection. For this reason we do not evaluate with ODIN in the rest of the paper. However, for", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 334, + 505, + 346 + ], + "spans": [ + { + "bbox": [ + 106, + 334, + 505, + 346 + ], + "score": 1.0, + "content": "thoroughness, we evaluate it here. ODIN uses temperature scaling and adds an epsilon perturbation", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 104, + 343, + 506, + 359 + ], + "spans": [ + { + "bbox": [ + 104, + 343, + 506, + 359 + ], + "score": 1.0, + "content": "to the input in order to separate the softmax posteriors between in- and out-of-distribution images;", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 356, + 506, + 368 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 506, + 368 + ], + "score": 1.0, + "content": "we set these hyperparameters following DeVries & Taylor (2018). Then, MaxLogit combined with", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 365, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 105, + 365, + 505, + 379 + ], + "score": 1.0, + "content": "ODIN results in an FPR95 of 33.6, an AUROC of 88.8 and an AUPR of 51.3 on ImageNet. On", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 376, + 505, + 391 + ], + "spans": [ + { + "bbox": [ + 105, + 376, + 505, + 391 + ], + "score": 1.0, + "content": "Places365, the FPR95 is 35.3, the AUROC is 86.5, and the AUPR is 24.2. Consequently, techniques", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 104, + 387, + 506, + 401 + ], + "spans": [ + { + "bbox": [ + 104, + 387, + 506, + 401 + ], + "score": 1.0, + "content": "built with different assumptions can integrate well with MaxLogit. We do not train ImageNet-21K", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 399, + 367, + 413 + ], + "spans": [ + { + "bbox": [ + 105, + 399, + 367, + 413 + ], + "score": 1.0, + "content": "models from scratch with these methods due to limited compute.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 23 + }, + { + "type": "title", + "bbox": [ + 108, + 429, + 405, + 442 + ], + "lines": [ + { + "bbox": [ + 106, + 428, + 406, + 444 + ], + "spans": [ + { + "bbox": [ + 106, + 428, + 406, + 444 + ], + "score": 1.0, + "content": "C MULTI-LABEL OUT-OF-DISTRIBUTION DATASET LIST", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 107, + 456, + 329, + 609 + ], + "lines": [ + { + "bbox": [ + 106, + 455, + 329, + 468 + ], + "spans": [ + { + "bbox": [ + 106, + 455, + 329, + 468 + ], + "score": 1.0, + "content": "For multi-label classification experiments, we choose", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 466, + 329, + 479 + ], + "spans": [ + { + "bbox": [ + 105, + 466, + 329, + 479 + ], + "score": 1.0, + "content": "the following classes from ImageNet-21K to serve as", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 478, + 329, + 490 + ], + "spans": [ + { + "bbox": [ + 106, + 478, + 329, + 490 + ], + "score": 1.0, + "content": "out-of-distribution data: dolphin (n02069412), deer", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 488, + 330, + 501 + ], + "spans": [ + { + "bbox": [ + 106, + 488, + 330, + 501 + ], + "score": 1.0, + "content": "(n02431122), bat (n02139199), rhino (n02392434),", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 500, + 330, + 512 + ], + "spans": [ + { + "bbox": [ + 105, + 500, + 330, + 512 + ], + "score": 1.0, + "content": "raccoon (n02508213), octopus (n01970164), giant", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 510, + 330, + 523 + ], + "spans": [ + { + "bbox": [ + 106, + 510, + 330, + 523 + ], + "score": 1.0, + "content": "clam (n01959492), leech (n01937909), Venus fly-", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 522, + 330, + 534 + ], + "spans": [ + { + "bbox": [ + 106, + 522, + 330, + 534 + ], + "score": 1.0, + "content": "trap (n12782915), cherry tree (n12641413), Japanese", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 531, + 330, + 545 + ], + "spans": [ + { + "bbox": [ + 105, + 531, + 330, + 545 + ], + "score": 1.0, + "content": "cherry blossoms (n12649317), red wood (n12285512),", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 543, + 329, + 555 + ], + "spans": [ + { + "bbox": [ + 106, + 543, + 329, + 555 + ], + "score": 1.0, + "content": "sunflower (n11978713), croissant (n07691650), stick", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 554, + 329, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 554, + 329, + 567 + ], + "score": 1.0, + "content": "cinnamon (n07814390), cotton (n12176953), rice", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 565, + 329, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 329, + 578 + ], + "score": 1.0, + "content": "(n12126084), sugar cane (n12132956), bamboo", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 576, + 329, + 588 + ], + "spans": [ + { + "bbox": [ + 106, + 576, + 329, + 588 + ], + "score": 1.0, + "content": "(n12147226), and tumeric (n12356395). These classes", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 587, + 329, + 599 + ], + "spans": [ + { + "bbox": [ + 105, + 587, + 329, + 599 + ], + "score": 1.0, + "content": "were hand-chosen so that they are distinct from VOC", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 598, + 186, + 610 + ], + "spans": [ + { + "bbox": [ + 106, + 598, + 186, + 610 + ], + "score": 1.0, + "content": "and COCO classes.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 35.5 + }, + { + "type": "title", + "bbox": [ + 107, + 628, + 241, + 641 + ], + "lines": [ + { + "bbox": [ + 105, + 626, + 243, + 644 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 243, + 644 + ], + "score": 1.0, + "content": "D OOD SEGMENTATION", + "type": "text" + } + ], + "index": 55 + } + ], + "index": 55 + }, + { + "type": "image", + "bbox": [ + 335, + 471, + 501, + 643 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 335, + 471, + 501, + 643 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 335, + 471, + 501, + 643 + ], + "spans": [ + { + "bbox": [ + 335, + 471, + 501, + 643 + ], + "score": 0.966, + "type": "image", + "image_path": "e830f93c8a5dee8cff942f66fbb348ee36c59204cff44e668c2183b5e4494531.jpg" + } + ] + } + ], + "index": 49, + "virtual_lines": [ + { + "bbox": [ + 335, + 471, + 501, + 484.2307692307692 + ], + "spans": [], + "index": 43 + }, + { + "bbox": [ + 335, + 484.2307692307692, + 501, + 497.46153846153845 + ], + "spans": [], + "index": 44 + }, + { + "bbox": [ + 335, + 497.46153846153845, + 501, + 510.6923076923077 + ], + "spans": [], + "index": 45 + }, + { + "bbox": [ + 335, + 510.6923076923077, + 501, + 523.9230769230769 + ], + "spans": [], + "index": 46 + }, + { + "bbox": [ + 335, + 523.9230769230769, + 501, + 537.1538461538462 + ], + "spans": [], + "index": 47 + }, + { + "bbox": [ + 335, + 537.1538461538462, + 501, + 550.3846153846155 + ], + "spans": [], + "index": 48 + }, + { + "bbox": [ + 335, + 550.3846153846155, + 501, + 563.6153846153848 + ], + "spans": [], + "index": 49 + }, + { + "bbox": [ + 335, + 563.6153846153848, + 501, + 576.846153846154 + ], + "spans": [], + "index": 50 + }, + { + "bbox": [ + 335, + 576.846153846154, + 501, + 590.0769230769233 + ], + "spans": [], + "index": 51 + }, + { + "bbox": [ + 335, + 590.0769230769233, + 501, + 603.3076923076926 + ], + "spans": [], + "index": 52 + }, + { + "bbox": [ + 335, + 603.3076923076926, + 501, + 616.5384615384619 + ], + "spans": [], + "index": 53 + }, + { + "bbox": [ + 335, + 616.5384615384619, + 501, + 629.7692307692312 + ], + "spans": [], + "index": 54 + }, + { + "bbox": [ + 335, + 629.7692307692312, + 501, + 643.0000000000005 + ], + "spans": [], + "index": 63 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 336, + 656, + 505, + 711 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 335, + 655, + 504, + 669 + ], + "spans": [ + { + "bbox": [ + 335, + 655, + 484, + 669 + ], + "score": 1.0, + "content": "Figure 6: ROC curve with VOC as", + "type": "text" + }, + { + "bbox": [ + 485, + 657, + 504, + 668 + ], + "score": 0.82, + "content": "( \\mathcal { D } _ { \\mathrm { i n } } )", + "type": "inline_equation" + } + ], + "index": 64 + }, + { + "bbox": [ + 335, + 666, + 505, + 680 + ], + "spans": [ + { + "bbox": [ + 335, + 666, + 505, + 680 + ], + "score": 1.0, + "content": "and non-overlapping ImageNet classes as", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 336, + 674, + 509, + 695 + ], + "spans": [ + { + "bbox": [ + 336, + 678, + 361, + 690 + ], + "score": 0.84, + "content": "( \\mathcal { D } _ { \\mathrm { o u t } } ^ { \\mathrm { t e s t } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 674, + 509, + 695 + ], + "score": 1.0, + "content": ". Curves correspond to an uninfor-", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 335, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 335, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "mative “Random” detector, Local Outlier", + "type": "text" + } + ], + "index": 67 + }, + { + "bbox": [ + 335, + 699, + 476, + 712 + ], + "spans": [ + { + "bbox": [ + 335, + 699, + 476, + 712 + ], + "score": 1.0, + "content": "Factor, and the MaxLogit detector.", + "type": "text" + } + ], + "index": 68 + } + ], + "index": 66 + } + ], + "index": 57.5 + }, + { + "type": "text", + "bbox": [ + 107, + 655, + 328, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 655, + 329, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 329, + 667 + ], + "score": 1.0, + "content": "We cover methods used in the paper in more depth and", + "type": "text" + } + ], + "index": 56 + }, + { + "bbox": [ + 106, + 666, + 329, + 677 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 329, + 677 + ], + "score": 1.0, + "content": "the modifications necessary to make the methods work", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 106, + 676, + 329, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 329, + 689 + ], + "score": 1.0, + "content": "with OOD detection in semantic segmentation. We use", + "type": "text" + } + ], + "index": 58 + }, + { + "bbox": [ + 107, + 687, + 327, + 699 + ], + "spans": [ + { + "bbox": [ + 107, + 688, + 114, + 699 + ], + "score": 0.85, + "content": "f", + "type": "inline_equation" + }, + { + "bbox": [ + 114, + 687, + 320, + 699 + ], + "score": 1.0, + "content": "to denote the function typically a neural network,", + "type": "text" + }, + { + "bbox": [ + 321, + 690, + 327, + 698 + ], + "score": 0.65, + "content": "x", + "type": "inline_equation" + } + ], + "index": 59 + }, + { + "bbox": [ + 105, + 698, + 329, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 204, + 712 + ], + "score": 1.0, + "content": "is the input image, and", + "type": "text" + }, + { + "bbox": [ + 204, + 700, + 219, + 711 + ], + "score": 0.88, + "content": "y _ { i , j }", + "type": "inline_equation" + }, + { + "bbox": [ + 220, + 698, + 329, + 712 + ], + "score": 1.0, + "content": "is the prediction for pixel", + "type": "text" + } + ], + "index": 60 + }, + { + "bbox": [ + 107, + 709, + 329, + 722 + ], + "spans": [ + { + "bbox": [ + 107, + 710, + 120, + 721 + ], + "score": 0.85, + "content": "i , j", + "type": "inline_equation" + }, + { + "bbox": [ + 120, + 709, + 329, + 722 + ], + "score": 1.0, + "content": ". We will denote the output probability distribution", + "type": "text" + } + ], + "index": 61 + }, + { + "bbox": [ + 104, + 720, + 329, + 732 + ], + "spans": [ + { + "bbox": [ + 104, + 720, + 156, + 732 + ], + "score": 1.0, + "content": "per pixel as", + "type": "text" + }, + { + "bbox": [ + 157, + 721, + 166, + 730 + ], + "score": 0.82, + "content": "P", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 720, + 225, + 732 + ], + "score": 1.0, + "content": "and locations", + "type": "text" + }, + { + "bbox": [ + 225, + 721, + 239, + 732 + ], + "score": 0.89, + "content": "i , j", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 720, + 329, + 732 + ], + "score": 1.0, + "content": "as the location of the", + "type": "text" + } + ], + "index": 62 + } + ], + "index": 59 + } + ], + "page_idx": 13, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 25, + 309, + 39 + ], + "spans": [ + { + "bbox": [ + 106, + 25, + 309, + 39 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 159 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 505, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 445, + 95 + ], + "score": 1.0, + "content": "Hendrycks et al. Hendrycks et al. (2019a), the MSP attains an average AUROC of", + "type": "text" + }, + { + "bbox": [ + 446, + 82, + 479, + 93 + ], + "score": 0.88, + "content": "9 0 . 0 8 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 479, + 82, + 505, + 95 + ], + "score": 1.0, + "content": "while", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 93, + 505, + 106 + ], + "spans": [ + { + "bbox": [ + 105, + 93, + 213, + 106 + ], + "score": 1.0, + "content": "the maximum logit attains", + "type": "text" + }, + { + "bbox": [ + 213, + 94, + 245, + 104 + ], + "score": 0.88, + "content": "9 0 . 2 2 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 245, + 93, + 281, + 106 + ], + "score": 1.0, + "content": ", a minor", + "type": "text" + }, + { + "bbox": [ + 282, + 94, + 309, + 104 + ], + "score": 0.87, + "content": "0 . 1 4 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 309, + 93, + 505, + 106 + ], + "score": 1.0, + "content": "difference. However, in a large-scale setting, the", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 104, + 506, + 117 + ], + "spans": [ + { + "bbox": [ + 105, + 104, + 198, + 117 + ], + "score": 1.0, + "content": "difference can be over", + "type": "text" + }, + { + "bbox": [ + 198, + 105, + 217, + 115 + ], + "score": 0.87, + "content": "10 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 217, + 104, + 273, + 117 + ], + "score": 1.0, + "content": "on individual", + "type": "text" + }, + { + "bbox": [ + 273, + 105, + 292, + 115 + ], + "score": 0.9, + "content": "\\mathcal { D } _ { \\mathrm { o u t } }", + "type": "inline_equation" + }, + { + "bbox": [ + 292, + 104, + 506, + 117 + ], + "score": 1.0, + "content": "datasets. We are not claiming that utilizing the maxi-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 115, + 505, + 128 + ], + "spans": [ + { + "bbox": [ + 105, + 115, + 505, + 128 + ], + "score": 1.0, + "content": "mum logit is a mathematically innovative formulation, only that it serves as a consistently powerful", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 105, + 126, + 505, + 139 + ], + "spans": [ + { + "bbox": [ + 105, + 126, + 505, + 139 + ], + "score": 1.0, + "content": "baseline for large-scale settings that went unappreciated in small-scale settings. In consequence, we", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "spans": [ + { + "bbox": [ + 105, + 137, + 505, + 150 + ], + "score": 1.0, + "content": "suggest using the maximum logit as a new baseline for large-scale multi-class out-of-distribution", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 147, + 148, + 161 + ], + "spans": [ + { + "bbox": [ + 105, + 147, + 148, + 161 + ], + "score": 1.0, + "content": "detection.", + "type": "text" + } + ], + "index": 6 + } + ], + "index": 3, + "bbox_fs": [ + 105, + 82, + 506, + 161 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 174, + 505, + 306 + ], + "lines": [ + { + "bbox": [ + 106, + 174, + 505, + 186 + ], + "spans": [ + { + "bbox": [ + 106, + 174, + 505, + 186 + ], + "score": 1.0, + "content": "Overview of Other Detection Methods. There are other techniques in out-of-distribution detection", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 186, + 505, + 197 + ], + "spans": [ + { + "bbox": [ + 106, + 186, + 505, + 197 + ], + "score": 1.0, + "content": "which require other assumptions such as more training data. For instance, Hendrycks et al. (2019a);", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 105, + 195, + 505, + 209 + ], + "spans": [ + { + "bbox": [ + 105, + 195, + 505, + 209 + ], + "score": 1.0, + "content": "Mohseni et al. (2020) use additional training data labeled as out-of-distribution, and the MaxLogit", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 207, + 505, + 219 + ], + "spans": [ + { + "bbox": [ + 106, + 207, + 505, + 219 + ], + "score": 1.0, + "content": "technique can be naturally extended should such data be available. Hendrycks et al. (2019c) use", + "type": "text" + } + ], + "index": 10 + }, + { + "bbox": [ + 106, + 219, + 505, + 231 + ], + "spans": [ + { + "bbox": [ + 106, + 219, + 505, + 231 + ], + "score": 1.0, + "content": "rotation prediction and self-supervised learning, but we found that scaling this to the ImageNet", + "type": "text" + } + ], + "index": 11 + }, + { + "bbox": [ + 105, + 229, + 505, + 241 + ], + "spans": [ + { + "bbox": [ + 105, + 229, + 505, + 241 + ], + "score": 1.0, + "content": "multiclass setting did not produce strong results. The MSP baseline trained with auxiliary rotation", + "type": "text" + } + ], + "index": 12 + }, + { + "bbox": [ + 106, + 240, + 505, + 253 + ], + "spans": [ + { + "bbox": [ + 106, + 240, + 228, + 253 + ], + "score": 1.0, + "content": "prediction has an AUROC of", + "type": "text" + }, + { + "bbox": [ + 228, + 240, + 256, + 251 + ], + "score": 0.87, + "content": "5 9 . 1 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 256, + 240, + 387, + 253 + ], + "score": 1.0, + "content": ", and with MaxLogit it attains a", + "type": "text" + }, + { + "bbox": [ + 387, + 240, + 415, + 251 + ], + "score": 0.87, + "content": "7 3 . 6 \\%", + "type": "inline_equation" + }, + { + "bbox": [ + 416, + 240, + 484, + 253 + ], + "score": 1.0, + "content": "AUROC, over a", + "type": "text" + }, + { + "bbox": [ + 485, + 240, + 505, + 251 + ], + "score": 0.86, + "content": "10 \\%", + "type": "inline_equation" + } + ], + "index": 13 + }, + { + "bbox": [ + 105, + 250, + 507, + 264 + ], + "spans": [ + { + "bbox": [ + 105, + 250, + 507, + 264 + ], + "score": 1.0, + "content": "absolute improvement with MaxLogit. Nonetheless this technique did not straightforwardly scale,", + "type": "text" + } + ], + "index": 14 + }, + { + "bbox": [ + 105, + 261, + 506, + 275 + ], + "spans": [ + { + "bbox": [ + 105, + 261, + 506, + 275 + ], + "score": 1.0, + "content": "as the network is better without auxiliary rotation prediction. Likewise, Lee et al. (2018b) propose", + "type": "text" + } + ], + "index": 15 + }, + { + "bbox": [ + 105, + 273, + 505, + 285 + ], + "spans": [ + { + "bbox": [ + 105, + 273, + 505, + 285 + ], + "score": 1.0, + "content": "to use Mahalanobis distances, but in scaling this to 1000 classes, we consistently encountered NaN", + "type": "text" + } + ], + "index": 16 + }, + { + "bbox": [ + 106, + 284, + 505, + 297 + ], + "spans": [ + { + "bbox": [ + 106, + 284, + 505, + 297 + ], + "score": 1.0, + "content": "errors due to high condition numbers. This shows the importance of ensuring that out-of-distribution", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 106, + 295, + 193, + 307 + ], + "spans": [ + { + "bbox": [ + 106, + 295, + 193, + 307 + ], + "score": 1.0, + "content": "techniques can scale.", + "type": "text" + } + ], + "index": 18 + } + ], + "index": 12.5, + "bbox_fs": [ + 105, + 174, + 507, + 307 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 312, + 505, + 411 + ], + "lines": [ + { + "bbox": [ + 106, + 312, + 505, + 324 + ], + "spans": [ + { + "bbox": [ + 106, + 312, + 505, + 324 + ], + "score": 1.0, + "content": "ODIN Liang et al. (2018) assumes that, for each OOD example source, we can tune hyperparameters", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 322, + 505, + 335 + ], + "spans": [ + { + "bbox": [ + 105, + 322, + 505, + 335 + ], + "score": 1.0, + "content": "for detection. For this reason we do not evaluate with ODIN in the rest of the paper. However, for", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 106, + 334, + 505, + 346 + ], + "spans": [ + { + "bbox": [ + 106, + 334, + 505, + 346 + ], + "score": 1.0, + "content": "thoroughness, we evaluate it here. ODIN uses temperature scaling and adds an epsilon perturbation", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 104, + 343, + 506, + 359 + ], + "spans": [ + { + "bbox": [ + 104, + 343, + 506, + 359 + ], + "score": 1.0, + "content": "to the input in order to separate the softmax posteriors between in- and out-of-distribution images;", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 105, + 356, + 506, + 368 + ], + "spans": [ + { + "bbox": [ + 105, + 356, + 506, + 368 + ], + "score": 1.0, + "content": "we set these hyperparameters following DeVries & Taylor (2018). Then, MaxLogit combined with", + "type": "text" + } + ], + "index": 23 + }, + { + "bbox": [ + 105, + 365, + 505, + 379 + ], + "spans": [ + { + "bbox": [ + 105, + 365, + 505, + 379 + ], + "score": 1.0, + "content": "ODIN results in an FPR95 of 33.6, an AUROC of 88.8 and an AUPR of 51.3 on ImageNet. On", + "type": "text" + } + ], + "index": 24 + }, + { + "bbox": [ + 105, + 376, + 505, + 391 + ], + "spans": [ + { + "bbox": [ + 105, + 376, + 505, + 391 + ], + "score": 1.0, + "content": "Places365, the FPR95 is 35.3, the AUROC is 86.5, and the AUPR is 24.2. Consequently, techniques", + "type": "text" + } + ], + "index": 25 + }, + { + "bbox": [ + 104, + 387, + 506, + 401 + ], + "spans": [ + { + "bbox": [ + 104, + 387, + 506, + 401 + ], + "score": 1.0, + "content": "built with different assumptions can integrate well with MaxLogit. We do not train ImageNet-21K", + "type": "text" + } + ], + "index": 26 + }, + { + "bbox": [ + 105, + 399, + 367, + 413 + ], + "spans": [ + { + "bbox": [ + 105, + 399, + 367, + 413 + ], + "score": 1.0, + "content": "models from scratch with these methods due to limited compute.", + "type": "text" + } + ], + "index": 27 + } + ], + "index": 23, + "bbox_fs": [ + 104, + 312, + 506, + 413 + ] + }, + { + "type": "title", + "bbox": [ + 108, + 429, + 405, + 442 + ], + "lines": [ + { + "bbox": [ + 106, + 428, + 406, + 444 + ], + "spans": [ + { + "bbox": [ + 106, + 428, + 406, + 444 + ], + "score": 1.0, + "content": "C MULTI-LABEL OUT-OF-DISTRIBUTION DATASET LIST", + "type": "text" + } + ], + "index": 28 + } + ], + "index": 28 + }, + { + "type": "text", + "bbox": [ + 107, + 456, + 329, + 609 + ], + "lines": [ + { + "bbox": [ + 106, + 455, + 329, + 468 + ], + "spans": [ + { + "bbox": [ + 106, + 455, + 329, + 468 + ], + "score": 1.0, + "content": "For multi-label classification experiments, we choose", + "type": "text" + } + ], + "index": 29 + }, + { + "bbox": [ + 105, + 466, + 329, + 479 + ], + "spans": [ + { + "bbox": [ + 105, + 466, + 329, + 479 + ], + "score": 1.0, + "content": "the following classes from ImageNet-21K to serve as", + "type": "text" + } + ], + "index": 30 + }, + { + "bbox": [ + 106, + 478, + 329, + 490 + ], + "spans": [ + { + "bbox": [ + 106, + 478, + 329, + 490 + ], + "score": 1.0, + "content": "out-of-distribution data: dolphin (n02069412), deer", + "type": "text" + } + ], + "index": 31 + }, + { + "bbox": [ + 106, + 488, + 330, + 501 + ], + "spans": [ + { + "bbox": [ + 106, + 488, + 330, + 501 + ], + "score": 1.0, + "content": "(n02431122), bat (n02139199), rhino (n02392434),", + "type": "text" + } + ], + "index": 32 + }, + { + "bbox": [ + 105, + 500, + 330, + 512 + ], + "spans": [ + { + "bbox": [ + 105, + 500, + 330, + 512 + ], + "score": 1.0, + "content": "raccoon (n02508213), octopus (n01970164), giant", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 106, + 510, + 330, + 523 + ], + "spans": [ + { + "bbox": [ + 106, + 510, + 330, + 523 + ], + "score": 1.0, + "content": "clam (n01959492), leech (n01937909), Venus fly-", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 106, + 522, + 330, + 534 + ], + "spans": [ + { + "bbox": [ + 106, + 522, + 330, + 534 + ], + "score": 1.0, + "content": "trap (n12782915), cherry tree (n12641413), Japanese", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 105, + 531, + 330, + 545 + ], + "spans": [ + { + "bbox": [ + 105, + 531, + 330, + 545 + ], + "score": 1.0, + "content": "cherry blossoms (n12649317), red wood (n12285512),", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 106, + 543, + 329, + 555 + ], + "spans": [ + { + "bbox": [ + 106, + 543, + 329, + 555 + ], + "score": 1.0, + "content": "sunflower (n11978713), croissant (n07691650), stick", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 105, + 554, + 329, + 567 + ], + "spans": [ + { + "bbox": [ + 105, + 554, + 329, + 567 + ], + "score": 1.0, + "content": "cinnamon (n07814390), cotton (n12176953), rice", + "type": "text" + } + ], + "index": 38 + }, + { + "bbox": [ + 106, + 565, + 329, + 578 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 329, + 578 + ], + "score": 1.0, + "content": "(n12126084), sugar cane (n12132956), bamboo", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 106, + 576, + 329, + 588 + ], + "spans": [ + { + "bbox": [ + 106, + 576, + 329, + 588 + ], + "score": 1.0, + "content": "(n12147226), and tumeric (n12356395). These classes", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 105, + 587, + 329, + 599 + ], + "spans": [ + { + "bbox": [ + 105, + 587, + 329, + 599 + ], + "score": 1.0, + "content": "were hand-chosen so that they are distinct from VOC", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 106, + 598, + 186, + 610 + ], + "spans": [ + { + "bbox": [ + 106, + 598, + 186, + 610 + ], + "score": 1.0, + "content": "and COCO classes.", + "type": "text" + } + ], + "index": 42 + } + ], + "index": 35.5, + "bbox_fs": [ + 105, + 455, + 330, + 610 + ] + }, + { + "type": "title", + "bbox": [ + 107, + 628, + 241, + 641 + ], + "lines": [ + { + "bbox": [ + 105, + 626, + 243, + 644 + ], + "spans": [ + { + "bbox": [ + 105, + 626, + 243, + 644 + ], + "score": 1.0, + "content": "D OOD SEGMENTATION", + "type": "text" + } + ], + "index": 55 + } + ], + "index": 55 + }, + { + "type": "image", + "bbox": [ + 335, + 471, + 501, + 643 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 335, + 471, + 501, + 643 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 335, + 471, + 501, + 643 + ], + "spans": [ + { + "bbox": [ + 335, + 471, + 501, + 643 + ], + "score": 0.966, + "type": "image", + "image_path": "e830f93c8a5dee8cff942f66fbb348ee36c59204cff44e668c2183b5e4494531.jpg" + } + ] + } + ], + "index": 49, + "virtual_lines": [ + { + "bbox": [ + 335, + 471, + 501, + 484.2307692307692 + ], + "spans": [], + "index": 43 + }, + { + "bbox": [ + 335, + 484.2307692307692, + 501, + 497.46153846153845 + ], + "spans": [], + "index": 44 + }, + { + "bbox": [ + 335, + 497.46153846153845, + 501, + 510.6923076923077 + ], + "spans": [], + "index": 45 + }, + { + "bbox": [ + 335, + 510.6923076923077, + 501, + 523.9230769230769 + ], + "spans": [], + "index": 46 + }, + { + "bbox": [ + 335, + 523.9230769230769, + 501, + 537.1538461538462 + ], + "spans": [], + "index": 47 + }, + { + "bbox": [ + 335, + 537.1538461538462, + 501, + 550.3846153846155 + ], + "spans": [], + "index": 48 + }, + { + "bbox": [ + 335, + 550.3846153846155, + 501, + 563.6153846153848 + ], + "spans": [], + "index": 49 + }, + { + "bbox": [ + 335, + 563.6153846153848, + 501, + 576.846153846154 + ], + "spans": [], + "index": 50 + }, + { + "bbox": [ + 335, + 576.846153846154, + 501, + 590.0769230769233 + ], + "spans": [], + "index": 51 + }, + { + "bbox": [ + 335, + 590.0769230769233, + 501, + 603.3076923076926 + ], + "spans": [], + "index": 52 + }, + { + "bbox": [ + 335, + 603.3076923076926, + 501, + 616.5384615384619 + ], + "spans": [], + "index": 53 + }, + { + "bbox": [ + 335, + 616.5384615384619, + 501, + 629.7692307692312 + ], + "spans": [], + "index": 54 + }, + { + "bbox": [ + 335, + 629.7692307692312, + 501, + 643.0000000000005 + ], + "spans": [], + "index": 63 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 336, + 656, + 505, + 711 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 335, + 655, + 504, + 669 + ], + "spans": [ + { + "bbox": [ + 335, + 655, + 484, + 669 + ], + "score": 1.0, + "content": "Figure 6: ROC curve with VOC as", + "type": "text" + }, + { + "bbox": [ + 485, + 657, + 504, + 668 + ], + "score": 0.82, + "content": "( \\mathcal { D } _ { \\mathrm { i n } } )", + "type": "inline_equation" + } + ], + "index": 64 + }, + { + "bbox": [ + 335, + 666, + 505, + 680 + ], + "spans": [ + { + "bbox": [ + 335, + 666, + 505, + 680 + ], + "score": 1.0, + "content": "and non-overlapping ImageNet classes as", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 336, + 674, + 509, + 695 + ], + "spans": [ + { + "bbox": [ + 336, + 678, + 361, + 690 + ], + "score": 0.84, + "content": "( \\mathcal { D } _ { \\mathrm { o u t } } ^ { \\mathrm { t e s t } } )", + "type": "inline_equation" + }, + { + "bbox": [ + 362, + 674, + 509, + 695 + ], + "score": 1.0, + "content": ". Curves correspond to an uninfor-", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 335, + 688, + 505, + 700 + ], + "spans": [ + { + "bbox": [ + 335, + 688, + 505, + 700 + ], + "score": 1.0, + "content": "mative “Random” detector, Local Outlier", + "type": "text" + } + ], + "index": 67 + }, + { + "bbox": [ + 335, + 699, + 476, + 712 + ], + "spans": [ + { + "bbox": [ + 335, + 699, + 476, + 712 + ], + "score": 1.0, + "content": "Factor, and the MaxLogit detector.", + "type": "text" + } + ], + "index": 68 + } + ], + "index": 66 + } + ], + "index": 57.5 + }, + { + "type": "text", + "bbox": [ + 107, + 655, + 328, + 732 + ], + "lines": [ + { + "bbox": [ + 106, + 655, + 329, + 667 + ], + "spans": [ + { + "bbox": [ + 106, + 655, + 329, + 667 + ], + "score": 1.0, + "content": "We cover methods used in the paper in more depth and", + "type": "text" + } + ], + "index": 56 + }, + { + "bbox": [ + 106, + 666, + 329, + 677 + ], + "spans": [ + { + "bbox": [ + 106, + 666, + 329, + 677 + ], + "score": 1.0, + "content": "the modifications necessary to make the methods work", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 106, + 676, + 329, + 689 + ], + "spans": [ + { + "bbox": [ + 106, + 676, + 329, + 689 + ], + "score": 1.0, + "content": "with OOD detection in semantic segmentation. We use", + "type": "text" + } + ], + "index": 58 + }, + { + "bbox": [ + 107, + 687, + 327, + 699 + ], + "spans": [ + { + "bbox": [ + 107, + 688, + 114, + 699 + ], + "score": 0.85, + "content": "f", + "type": "inline_equation" + }, + { + "bbox": [ + 114, + 687, + 320, + 699 + ], + "score": 1.0, + "content": "to denote the function typically a neural network,", + "type": "text" + }, + { + "bbox": [ + 321, + 690, + 327, + 698 + ], + "score": 0.65, + "content": "x", + "type": "inline_equation" + } + ], + "index": 59 + }, + { + "bbox": [ + 105, + 698, + 329, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 698, + 204, + 712 + ], + "score": 1.0, + "content": "is the input image, and", + "type": "text" + }, + { + "bbox": [ + 204, + 700, + 219, + 711 + ], + "score": 0.88, + "content": "y _ { i , j }", + "type": "inline_equation" + }, + { + "bbox": [ + 220, + 698, + 329, + 712 + ], + "score": 1.0, + "content": "is the prediction for pixel", + "type": "text" + } + ], + "index": 60 + }, + { + "bbox": [ + 107, + 709, + 329, + 722 + ], + "spans": [ + { + "bbox": [ + 107, + 710, + 120, + 721 + ], + "score": 0.85, + "content": "i , j", + "type": "inline_equation" + }, + { + "bbox": [ + 120, + 709, + 329, + 722 + ], + "score": 1.0, + "content": ". We will denote the output probability distribution", + "type": "text" + } + ], + "index": 61 + }, + { + "bbox": [ + 104, + 720, + 329, + 732 + ], + "spans": [ + { + "bbox": [ + 104, + 720, + 156, + 732 + ], + "score": 1.0, + "content": "per pixel as", + "type": "text" + }, + { + "bbox": [ + 157, + 721, + 166, + 730 + ], + "score": 0.82, + "content": "P", + "type": "inline_equation" + }, + { + "bbox": [ + 167, + 720, + 225, + 732 + ], + "score": 1.0, + "content": "and locations", + "type": "text" + }, + { + "bbox": [ + 225, + 721, + 239, + 732 + ], + "score": 0.89, + "content": "i , j", + "type": "inline_equation" + }, + { + "bbox": [ + 239, + 720, + 329, + 732 + ], + "score": 1.0, + "content": "as the location of the", + "type": "text" + } + ], + "index": 62 + }, + { + "bbox": [ + 106, + 82, + 328, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 233, + 95 + ], + "score": 1.0, + "content": "respective pixel in the output.", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 233, + 82, + 263, + 95 + ], + "score": 0.93, + "content": "f ( x ) _ { i , j }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 264, + 82, + 328, + 95 + ], + "score": 1.0, + "content": "denotes the ith", + "type": "text", + "cross_page": true + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 93, + 247, + 106 + ], + "spans": [ + { + "bbox": [ + 106, + 93, + 141, + 106 + ], + "score": 1.0, + "content": "row and", + "type": "text", + "cross_page": true + }, + { + "bbox": [ + 141, + 94, + 148, + 105 + ], + "score": 0.81, + "content": "j ^ { ; }", + "type": "inline_equation", + "cross_page": true + }, + { + "bbox": [ + 148, + 93, + 247, + 106 + ], + "score": 1.0, + "content": "’th column of the output.", + "type": "text", + "cross_page": true + } + ], + "index": 1 + } + ], + "index": 59, + "bbox_fs": [ + 104, + 655, + 329, + 732 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 328, + 105 + ], + "lines": [ + { + "bbox": [ + 106, + 82, + 328, + 95 + ], + "spans": [ + { + "bbox": [ + 106, + 82, + 233, + 95 + ], + "score": 1.0, + "content": "respective pixel in the output.", + "type": "text" + }, + { + "bbox": [ + 233, + 82, + 263, + 95 + ], + "score": 0.93, + "content": "f ( x ) _ { i , j }", + "type": "inline_equation" + }, + { + "bbox": [ + 264, + 82, + 328, + 95 + ], + "score": 1.0, + "content": "denotes the ith", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 106, + 93, + 247, + 106 + ], + "spans": [ + { + "bbox": [ + 106, + 93, + 141, + 106 + ], + "score": 1.0, + "content": "row and", + "type": "text" + }, + { + "bbox": [ + 141, + 94, + 148, + 105 + ], + "score": 0.81, + "content": "j ^ { ; }", + "type": "inline_equation" + }, + { + "bbox": [ + 148, + 93, + 247, + 106 + ], + "score": 1.0, + "content": "’th column of the output.", + "type": "text" + } + ], + "index": 1 + } + ], + "index": 0.5 + }, + { + "type": "text", + "bbox": [ + 107, + 117, + 297, + 183 + ], + "lines": [ + { + "bbox": [ + 106, + 116, + 298, + 129 + ], + "spans": [ + { + "bbox": [ + 106, + 116, + 298, + 129 + ], + "score": 1.0, + "content": "Confidence Estimation. The method pro-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 128, + 298, + 141 + ], + "spans": [ + { + "bbox": [ + 105, + 128, + 298, + 141 + ], + "score": 1.0, + "content": "posed in DeVries & Taylor (2018) works by", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 139, + 298, + 150 + ], + "spans": [ + { + "bbox": [ + 106, + 139, + 298, + 150 + ], + "score": 1.0, + "content": "training a confidence branch added at the end", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 150, + 297, + 162 + ], + "spans": [ + { + "bbox": [ + 106, + 150, + 297, + 162 + ], + "score": 1.0, + "content": "of the neural network. We denote the network", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 160, + 298, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 160, + 182, + 174 + ], + "score": 1.0, + "content": "predictions as both", + "type": "text" + }, + { + "bbox": [ + 182, + 162, + 191, + 171 + ], + "score": 0.82, + "content": "P", + "type": "inline_equation" + }, + { + "bbox": [ + 192, + 160, + 209, + 174 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 209, + 162, + 215, + 171 + ], + "score": 0.78, + "content": "\\hat { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 215, + 160, + 298, + 174 + ], + "score": 1.0, + "content": "whereby every pixel", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 172, + 231, + 184 + ], + "spans": [ + { + "bbox": [ + 105, + 172, + 231, + 184 + ], + "score": 1.0, + "content": "is assigned a confidence value.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 4.5 + }, + { + "type": "interline_equation", + "bbox": [ + 155, + 198, + 248, + 244 + ], + "lines": [ + { + "bbox": [ + 155, + 198, + 248, + 244 + ], + "spans": [ + { + "bbox": [ + 155, + 198, + 248, + 244 + ], + "score": 0.9, + "content": "\\begin{array} { c } { b \\sim B ( 0 . 5 ) } \\\\ { c : = \\hat { c } \\cdot b + ( 1 - b ) } \\\\ { P : = P \\cdot c + ( 1 - c ) y } \\end{array}", + "type": "interline_equation", + "image_path": "712f2da0fc9e3316c69cdaeb4970ca6182bdc9b12ec93cb069bfd5f2fb7b72b3.jpg" + } + ] + } + ], + "index": 13.5, + "virtual_lines": [ + { + "bbox": [ + 155, + 198, + 248, + 221.0 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 155, + 221.0, + 248, + 244.0 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 251, + 297, + 328 + ], + "lines": [ + { + "bbox": [ + 105, + 250, + 297, + 263 + ], + "spans": [ + { + "bbox": [ + 105, + 250, + 257, + 263 + ], + "score": 1.0, + "content": "The confidence estimation denoted by", + "type": "text" + }, + { + "bbox": [ + 258, + 253, + 263, + 261 + ], + "score": 0.75, + "content": "c", + "type": "inline_equation" + }, + { + "bbox": [ + 264, + 250, + 297, + 263 + ], + "score": 1.0, + "content": "is given", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 261, + 298, + 274 + ], + "spans": [ + { + "bbox": [ + 105, + 261, + 298, + 274 + ], + "score": 1.0, + "content": "“hints” during training to guide what it is learn-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 272, + 298, + 285 + ], + "spans": [ + { + "bbox": [ + 106, + 272, + 145, + 285 + ], + "score": 1.0, + "content": "ing. The", + "type": "text" + }, + { + "bbox": [ + 145, + 273, + 154, + 283 + ], + "score": 0.81, + "content": "B", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 272, + 298, + 285 + ], + "score": 1.0, + "content": "is a beta distribution and acts as a", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 285, + 298, + 295 + ], + "spans": [ + { + "bbox": [ + 105, + 285, + 298, + 295 + ], + "score": 1.0, + "content": "regularizer similar to dropout so that the network", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 107, + 295, + 298, + 307 + ], + "spans": [ + { + "bbox": [ + 107, + 295, + 114, + 307 + ], + "score": 0.84, + "content": "f", + "type": "inline_equation" + }, + { + "bbox": [ + 114, + 295, + 298, + 307 + ], + "score": 1.0, + "content": "does not exclusively rely on the true labels be-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 306, + 297, + 317 + ], + "spans": [ + { + "bbox": [ + 105, + 306, + 297, + 317 + ], + "score": 1.0, + "content": "ing present. The final loss is modified to include", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 317, + 193, + 328 + ], + "spans": [ + { + "bbox": [ + 106, + 317, + 193, + 328 + ], + "score": 1.0, + "content": "the extra term below:", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 20 + }, + { + "type": "image", + "bbox": [ + 309, + 136, + 498, + 323 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 309, + 136, + 498, + 323 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 309, + 136, + 498, + 323 + ], + "spans": [ + { + "bbox": [ + 309, + 136, + 498, + 323 + ], + "score": 0.976, + "type": "image", + "image_path": "e831207f61f4f9d875c0b02663677b918f457e3a8c337ddc1935c65b9e381231.jpg" + } + ] + } + ], + "index": 24, + "virtual_lines": [ + { + "bbox": [ + 309, + 136, + 498, + 148.46666666666667 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 309, + 148.46666666666667, + 498, + 160.93333333333334 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 309, + 160.93333333333334, + 498, + 173.4 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 309, + 173.4, + 498, + 185.86666666666667 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 309, + 185.86666666666667, + 498, + 198.33333333333334 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 309, + 198.33333333333334, + 498, + 210.8 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 309, + 210.8, + 498, + 223.26666666666668 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 309, + 223.26666666666668, + 498, + 235.73333333333335 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 309, + 235.73333333333335, + 498, + 248.20000000000002 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 309, + 248.20000000000002, + 498, + 260.6666666666667 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 309, + 260.6666666666667, + 498, + 273.1333333333333 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 309, + 273.1333333333333, + 498, + 285.59999999999997 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 309, + 285.59999999999997, + 498, + 298.0666666666666 + ], + "spans": [], + "index": 29 + }, + { + "bbox": [ + 309, + 298.0666666666666, + 498, + 310.53333333333325 + ], + "spans": [], + "index": 30 + }, + { + "bbox": [ + 309, + 310.53333333333325, + 498, + 322.9999999999999 + ], + "spans": [], + "index": 31 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 304, + 338, + 505, + 459 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 303, + 337, + 505, + 351 + ], + "spans": [ + { + "bbox": [ + 303, + 337, + 505, + 351 + ], + "score": 1.0, + "content": "Figure 7: A comparison of lighting consistency in", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 304, + 349, + 505, + 361 + ], + "spans": [ + { + "bbox": [ + 304, + 349, + 505, + 361 + ], + "score": 1.0, + "content": "the Fishyscapes anomaly segmentation benchmark", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 304, + 360, + 505, + 371 + ], + "spans": [ + { + "bbox": [ + 304, + 360, + 505, + 371 + ], + "score": 1.0, + "content": "and our new StreetHazards dataset. The arrows", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 303, + 371, + 505, + 383 + ], + "spans": [ + { + "bbox": [ + 303, + 371, + 505, + 383 + ], + "score": 1.0, + "content": "point in the manually estimated direction of light", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 304, + 382, + 506, + 393 + ], + "spans": [ + { + "bbox": [ + 304, + 382, + 506, + 393 + ], + "score": 1.0, + "content": "on parts of the scene. In Fishyscapes, inconsis-", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 303, + 393, + 506, + 405 + ], + "spans": [ + { + "bbox": [ + 303, + 393, + 506, + 405 + ], + "score": 1.0, + "content": "tent lighting allows forensics techniques to detect", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 303, + 403, + 507, + 415 + ], + "spans": [ + { + "bbox": [ + 303, + 403, + 507, + 415 + ], + "score": 1.0, + "content": "the anomaly (Johnson & Farid, 2005). Unlike cut-", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 303, + 415, + 507, + 426 + ], + "spans": [ + { + "bbox": [ + 303, + 415, + 507, + 426 + ], + "score": 1.0, + "content": "and-paste anomalies, the anomalies in our Street-", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 303, + 425, + 506, + 437 + ], + "spans": [ + { + "bbox": [ + 303, + 425, + 506, + 437 + ], + "score": 1.0, + "content": "Hazards dataset are naturally integrated into their", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 304, + 437, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 304, + 437, + 506, + 448 + ], + "score": 1.0, + "content": "environment with proper lighting and orientation,", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 304, + 447, + 455, + 459 + ], + "spans": [ + { + "bbox": [ + 304, + 447, + 455, + 459 + ], + "score": 1.0, + "content": "making them more difficult to detect.", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 39 + } + ], + "index": 31.5 + }, + { + "type": "interline_equation", + "bbox": [ + 149, + 340, + 254, + 418 + ], + "lines": [ + { + "bbox": [ + 149, + 340, + 254, + 418 + ], + "spans": [ + { + "bbox": [ + 149, + 340, + 254, + 418 + ], + "score": 0.9, + "content": "\\begin{array} { l } { \\displaystyle \\mathcal { L } _ { \\boldsymbol { p } } = \\frac { 1 } { | P | } \\sum _ { i } - \\log ( p _ { i } ) y _ { i } } \\\\ { \\displaystyle \\mathcal { L } _ { \\boldsymbol { c } } = \\frac { 1 } { | P | } \\sum _ { i } - \\log ( \\hat { c } _ { i } ) } \\\\ { \\displaystyle \\mathcal { L } = \\mathcal { L } _ { \\boldsymbol { p } } + \\lambda \\mathcal { L } _ { \\boldsymbol { c } } } \\end{array}", + "type": "interline_equation", + "image_path": "188937f87a242a5c887d29d311cbd567431198a575e25323201226d6d6633277.jpg" + } + ] + } + ], + "index": 35.0, + "virtual_lines": [ + { + "bbox": [ + 149, + 340, + 254, + 379.0 + ], + "spans": [], + "index": 32 + }, + { + "bbox": [ + 149, + 379.0, + 254, + 418.0 + ], + "spans": [], + "index": 38 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 424, + 297, + 480 + ], + "lines": [ + { + "bbox": [ + 106, + 425, + 299, + 437 + ], + "spans": [ + { + "bbox": [ + 106, + 425, + 183, + 437 + ], + "score": 1.0, + "content": "The reasoning for", + "type": "text" + }, + { + "bbox": [ + 184, + 425, + 196, + 436 + ], + "score": 0.88, + "content": "\\mathcal { L } _ { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 425, + 299, + 437 + ], + "score": 1.0, + "content": "is to encourage the net-", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 436, + 296, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 436, + 289, + 447 + ], + "score": 1.0, + "content": "work to output confident predictions. Finally", + "type": "text" + }, + { + "bbox": [ + 289, + 436, + 296, + 446 + ], + "score": 0.7, + "content": "\\lambda", + "type": "inline_equation" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 446, + 299, + 458 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 299, + 458 + ], + "score": 1.0, + "content": "is initialized to 0.1 and is updated by a “budget”", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 458, + 297, + 469 + ], + "spans": [ + { + "bbox": [ + 105, + 458, + 297, + 469 + ], + "score": 1.0, + "content": "parameter which is set to the default of 0.3. The", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 469, + 176, + 481 + ], + "spans": [ + { + "bbox": [ + 106, + 469, + 176, + 481 + ], + "score": 1.0, + "content": "update equation:", + "type": "text" + } + ], + "index": 46 + } + ], + "index": 43 + }, + { + "type": "interline_equation", + "bbox": [ + 250, + 494, + 360, + 523 + ], + "lines": [ + { + "bbox": [ + 250, + 494, + 360, + 523 + ], + "spans": [ + { + "bbox": [ + 250, + 494, + 360, + 523 + ], + "score": 0.86, + "content": "\\left\\{ \\begin{array} { l l } { { \\lambda } / { 0 . 9 9 } } & { \\sum \\hat { c } _ { i } \\leq \\mathrm { b u d g e t } } \\\\ { { \\lambda } / { 1 . 0 1 } } & { \\sum \\hat { c } _ { i } > \\mathrm { b u d g e t } } \\end{array} \\right.", + "type": "interline_equation", + "image_path": "c4eed88b17dbc93f4e19087b97ce3896c0eb2ba44f70df785114b20ca53d77a9.jpg" + } + ] + } + ], + "index": 50.5, + "virtual_lines": [ + { + "bbox": [ + 250, + 494, + 360, + 508.5 + ], + "spans": [], + "index": 50 + }, + { + "bbox": [ + 250, + 508.5, + 360, + 523.0 + ], + "spans": [], + "index": 51 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 531, + 504, + 553 + ], + "lines": [ + { + "bbox": [ + 105, + 530, + 506, + 544 + ], + "spans": [ + { + "bbox": [ + 105, + 530, + 506, + 544 + ], + "score": 1.0, + "content": "This adaptively adjusts the weighting between the two losses and experimentally the update is not", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 106, + 542, + 241, + 553 + ], + "spans": [ + { + "bbox": [ + 106, + 542, + 241, + 553 + ], + "score": 1.0, + "content": "sensitive to the budget parameter.", + "type": "text" + } + ], + "index": 53 + } + ], + "index": 52.5 + }, + { + "type": "text", + "bbox": [ + 107, + 564, + 505, + 599 + ], + "lines": [ + { + "bbox": [ + 106, + 565, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 505, + 577 + ], + "score": 1.0, + "content": "Semantic Segmentation BDD Anomalies Dataset List. The BDD100K dataset contains 180", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 106, + 577, + 505, + 589 + ], + "spans": [ + { + "bbox": [ + 106, + 577, + 505, + 589 + ], + "score": 1.0, + "content": "instances of the train class, 4296 instances of the motorcycle class, and 10229 instances of the bicycle", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 105, + 586, + 132, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 132, + 600 + ], + "score": 1.0, + "content": "class.", + "type": "text" + } + ], + "index": 56 + } + ], + "index": 55 + }, + { + "type": "text", + "bbox": [ + 106, + 610, + 506, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 610, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 505, + 623 + ], + "score": 1.0, + "content": "StreetHazards 3D Models Dataset List. For semantic segmentation experiments, we choose to", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 105, + 622, + 507, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 507, + 634 + ], + "score": 1.0, + "content": "use the following classes 3D models from Model Bank Library to serve as out-of-distribution data:", + "type": "text" + } + ], + "index": 58 + }, + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "Meta-categories: Animals, Vehicles, Weapons, Appliances, Household items (furniture, and kitchen", + "type": "text" + } + ], + "index": 59 + }, + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "score": 1.0, + "content": "items), Electronics, Instruments, and miscellaneous. The specific animals used are kangaroos, whales,", + "type": "text" + } + ], + "index": 60 + }, + { + "bbox": [ + 105, + 654, + 507, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 507, + 668 + ], + "score": 1.0, + "content": "dolphins, cows, lions, frogs, bats, insects, mongooses, scorpions, fish, camels, flamingos, apes, horses,", + "type": "text" + } + ], + "index": 61 + }, + { + "bbox": [ + 105, + 666, + 507, + 680 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 507, + 680 + ], + "score": 1.0, + "content": "mice, spider, dinosaurs, elephants, moose, shrimps, bats, butterflies, turtles, hippopotamuses, dogs,", + "type": "text" + } + ], + "index": 62 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "cats, sheep, seahorse, snail and zebra. The specific vehicles used are military trucks, motorcycles,", + "type": "text" + } + ], + "index": 63 + }, + { + "bbox": [ + 105, + 687, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 702 + ], + "score": 1.0, + "content": "naval ships, pirate ships, submarines, sailing ships, trolleys, trains, airplanes, helicopters, jets,", + "type": "text" + } + ], + "index": 64 + }, + { + "bbox": [ + 105, + 699, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 505, + 712 + ], + "score": 1.0, + "content": "zeppelin, radar tower, construction vehicles (loaders, dump trucks, bulldozer), farming vehicles", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 106, + 710, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 723 + ], + "score": 1.0, + "content": "(harvester, gantry crane, tractor), fire truck, tank, combat vehicles, and trailers. The specific weapons", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 106, + 721, + 506, + 734 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 506, + 734 + ], + "score": 1.0, + "content": "used are guns, missiles, rocket launchers, and grenades. The appliances used are refrigerators, stoves,", + "type": "text" + } + ], + "index": 67 + } + ], + "index": 62 + } + ], + "page_idx": 14, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 106, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 106, + 26, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 751, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 328, + 105 + ], + "lines": [], + "index": 0.5, + "bbox_fs": [ + 106, + 82, + 328, + 106 + ], + "lines_deleted": true + }, + { + "type": "text", + "bbox": [ + 107, + 117, + 297, + 183 + ], + "lines": [ + { + "bbox": [ + 106, + 116, + 298, + 129 + ], + "spans": [ + { + "bbox": [ + 106, + 116, + 298, + 129 + ], + "score": 1.0, + "content": "Confidence Estimation. The method pro-", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 105, + 128, + 298, + 141 + ], + "spans": [ + { + "bbox": [ + 105, + 128, + 298, + 141 + ], + "score": 1.0, + "content": "posed in DeVries & Taylor (2018) works by", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 139, + 298, + 150 + ], + "spans": [ + { + "bbox": [ + 106, + 139, + 298, + 150 + ], + "score": 1.0, + "content": "training a confidence branch added at the end", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 150, + 297, + 162 + ], + "spans": [ + { + "bbox": [ + 106, + 150, + 297, + 162 + ], + "score": 1.0, + "content": "of the neural network. We denote the network", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 105, + 160, + 298, + 174 + ], + "spans": [ + { + "bbox": [ + 105, + 160, + 182, + 174 + ], + "score": 1.0, + "content": "predictions as both", + "type": "text" + }, + { + "bbox": [ + 182, + 162, + 191, + 171 + ], + "score": 0.82, + "content": "P", + "type": "inline_equation" + }, + { + "bbox": [ + 192, + 160, + 209, + 174 + ], + "score": 1.0, + "content": "and", + "type": "text" + }, + { + "bbox": [ + 209, + 162, + 215, + 171 + ], + "score": 0.78, + "content": "\\hat { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 215, + 160, + 298, + 174 + ], + "score": 1.0, + "content": "whereby every pixel", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 105, + 172, + 231, + 184 + ], + "spans": [ + { + "bbox": [ + 105, + 172, + 231, + 184 + ], + "score": 1.0, + "content": "is assigned a confidence value.", + "type": "text" + } + ], + "index": 7 + } + ], + "index": 4.5, + "bbox_fs": [ + 105, + 116, + 298, + 184 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 155, + 198, + 248, + 244 + ], + "lines": [ + { + "bbox": [ + 155, + 198, + 248, + 244 + ], + "spans": [ + { + "bbox": [ + 155, + 198, + 248, + 244 + ], + "score": 0.9, + "content": "\\begin{array} { c } { b \\sim B ( 0 . 5 ) } \\\\ { c : = \\hat { c } \\cdot b + ( 1 - b ) } \\\\ { P : = P \\cdot c + ( 1 - c ) y } \\end{array}", + "type": "interline_equation", + "image_path": "712f2da0fc9e3316c69cdaeb4970ca6182bdc9b12ec93cb069bfd5f2fb7b72b3.jpg" + } + ] + } + ], + "index": 13.5, + "virtual_lines": [ + { + "bbox": [ + 155, + 198, + 248, + 221.0 + ], + "spans": [], + "index": 13 + }, + { + "bbox": [ + 155, + 221.0, + 248, + 244.0 + ], + "spans": [], + "index": 14 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 251, + 297, + 328 + ], + "lines": [ + { + "bbox": [ + 105, + 250, + 297, + 263 + ], + "spans": [ + { + "bbox": [ + 105, + 250, + 257, + 263 + ], + "score": 1.0, + "content": "The confidence estimation denoted by", + "type": "text" + }, + { + "bbox": [ + 258, + 253, + 263, + 261 + ], + "score": 0.75, + "content": "c", + "type": "inline_equation" + }, + { + "bbox": [ + 264, + 250, + 297, + 263 + ], + "score": 1.0, + "content": "is given", + "type": "text" + } + ], + "index": 17 + }, + { + "bbox": [ + 105, + 261, + 298, + 274 + ], + "spans": [ + { + "bbox": [ + 105, + 261, + 298, + 274 + ], + "score": 1.0, + "content": "“hints” during training to guide what it is learn-", + "type": "text" + } + ], + "index": 18 + }, + { + "bbox": [ + 106, + 272, + 298, + 285 + ], + "spans": [ + { + "bbox": [ + 106, + 272, + 145, + 285 + ], + "score": 1.0, + "content": "ing. The", + "type": "text" + }, + { + "bbox": [ + 145, + 273, + 154, + 283 + ], + "score": 0.81, + "content": "B", + "type": "inline_equation" + }, + { + "bbox": [ + 155, + 272, + 298, + 285 + ], + "score": 1.0, + "content": "is a beta distribution and acts as a", + "type": "text" + } + ], + "index": 19 + }, + { + "bbox": [ + 105, + 285, + 298, + 295 + ], + "spans": [ + { + "bbox": [ + 105, + 285, + 298, + 295 + ], + "score": 1.0, + "content": "regularizer similar to dropout so that the network", + "type": "text" + } + ], + "index": 20 + }, + { + "bbox": [ + 107, + 295, + 298, + 307 + ], + "spans": [ + { + "bbox": [ + 107, + 295, + 114, + 307 + ], + "score": 0.84, + "content": "f", + "type": "inline_equation" + }, + { + "bbox": [ + 114, + 295, + 298, + 307 + ], + "score": 1.0, + "content": "does not exclusively rely on the true labels be-", + "type": "text" + } + ], + "index": 21 + }, + { + "bbox": [ + 105, + 306, + 297, + 317 + ], + "spans": [ + { + "bbox": [ + 105, + 306, + 297, + 317 + ], + "score": 1.0, + "content": "ing present. The final loss is modified to include", + "type": "text" + } + ], + "index": 22 + }, + { + "bbox": [ + 106, + 317, + 193, + 328 + ], + "spans": [ + { + "bbox": [ + 106, + 317, + 193, + 328 + ], + "score": 1.0, + "content": "the extra term below:", + "type": "text" + } + ], + "index": 23 + } + ], + "index": 20, + "bbox_fs": [ + 105, + 250, + 298, + 328 + ] + }, + { + "type": "image", + "bbox": [ + 309, + 136, + 498, + 323 + ], + "blocks": [ + { + "type": "image_body", + "bbox": [ + 309, + 136, + 498, + 323 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 309, + 136, + 498, + 323 + ], + "spans": [ + { + "bbox": [ + 309, + 136, + 498, + 323 + ], + "score": 0.976, + "type": "image", + "image_path": "e831207f61f4f9d875c0b02663677b918f457e3a8c337ddc1935c65b9e381231.jpg" + } + ] + } + ], + "index": 24, + "virtual_lines": [ + { + "bbox": [ + 309, + 136, + 498, + 148.46666666666667 + ], + "spans": [], + "index": 8 + }, + { + "bbox": [ + 309, + 148.46666666666667, + 498, + 160.93333333333334 + ], + "spans": [], + "index": 9 + }, + { + "bbox": [ + 309, + 160.93333333333334, + 498, + 173.4 + ], + "spans": [], + "index": 10 + }, + { + "bbox": [ + 309, + 173.4, + 498, + 185.86666666666667 + ], + "spans": [], + "index": 11 + }, + { + "bbox": [ + 309, + 185.86666666666667, + 498, + 198.33333333333334 + ], + "spans": [], + "index": 12 + }, + { + "bbox": [ + 309, + 198.33333333333334, + 498, + 210.8 + ], + "spans": [], + "index": 15 + }, + { + "bbox": [ + 309, + 210.8, + 498, + 223.26666666666668 + ], + "spans": [], + "index": 16 + }, + { + "bbox": [ + 309, + 223.26666666666668, + 498, + 235.73333333333335 + ], + "spans": [], + "index": 24 + }, + { + "bbox": [ + 309, + 235.73333333333335, + 498, + 248.20000000000002 + ], + "spans": [], + "index": 25 + }, + { + "bbox": [ + 309, + 248.20000000000002, + 498, + 260.6666666666667 + ], + "spans": [], + "index": 26 + }, + { + "bbox": [ + 309, + 260.6666666666667, + 498, + 273.1333333333333 + ], + "spans": [], + "index": 27 + }, + { + "bbox": [ + 309, + 273.1333333333333, + 498, + 285.59999999999997 + ], + "spans": [], + "index": 28 + }, + { + "bbox": [ + 309, + 285.59999999999997, + 498, + 298.0666666666666 + ], + "spans": [], + "index": 29 + }, + { + "bbox": [ + 309, + 298.0666666666666, + 498, + 310.53333333333325 + ], + "spans": [], + "index": 30 + }, + { + "bbox": [ + 309, + 310.53333333333325, + 498, + 322.9999999999999 + ], + "spans": [], + "index": 31 + } + ] + }, + { + "type": "image_caption", + "bbox": [ + 304, + 338, + 505, + 459 + ], + "group_id": 0, + "lines": [ + { + "bbox": [ + 303, + 337, + 505, + 351 + ], + "spans": [ + { + "bbox": [ + 303, + 337, + 505, + 351 + ], + "score": 1.0, + "content": "Figure 7: A comparison of lighting consistency in", + "type": "text" + } + ], + "index": 33 + }, + { + "bbox": [ + 304, + 349, + 505, + 361 + ], + "spans": [ + { + "bbox": [ + 304, + 349, + 505, + 361 + ], + "score": 1.0, + "content": "the Fishyscapes anomaly segmentation benchmark", + "type": "text" + } + ], + "index": 34 + }, + { + "bbox": [ + 304, + 360, + 505, + 371 + ], + "spans": [ + { + "bbox": [ + 304, + 360, + 505, + 371 + ], + "score": 1.0, + "content": "and our new StreetHazards dataset. The arrows", + "type": "text" + } + ], + "index": 35 + }, + { + "bbox": [ + 303, + 371, + 505, + 383 + ], + "spans": [ + { + "bbox": [ + 303, + 371, + 505, + 383 + ], + "score": 1.0, + "content": "point in the manually estimated direction of light", + "type": "text" + } + ], + "index": 36 + }, + { + "bbox": [ + 304, + 382, + 506, + 393 + ], + "spans": [ + { + "bbox": [ + 304, + 382, + 506, + 393 + ], + "score": 1.0, + "content": "on parts of the scene. In Fishyscapes, inconsis-", + "type": "text" + } + ], + "index": 37 + }, + { + "bbox": [ + 303, + 393, + 506, + 405 + ], + "spans": [ + { + "bbox": [ + 303, + 393, + 506, + 405 + ], + "score": 1.0, + "content": "tent lighting allows forensics techniques to detect", + "type": "text" + } + ], + "index": 39 + }, + { + "bbox": [ + 303, + 403, + 507, + 415 + ], + "spans": [ + { + "bbox": [ + 303, + 403, + 507, + 415 + ], + "score": 1.0, + "content": "the anomaly (Johnson & Farid, 2005). Unlike cut-", + "type": "text" + } + ], + "index": 40 + }, + { + "bbox": [ + 303, + 415, + 507, + 426 + ], + "spans": [ + { + "bbox": [ + 303, + 415, + 507, + 426 + ], + "score": 1.0, + "content": "and-paste anomalies, the anomalies in our Street-", + "type": "text" + } + ], + "index": 44 + }, + { + "bbox": [ + 303, + 425, + 506, + 437 + ], + "spans": [ + { + "bbox": [ + 303, + 425, + 506, + 437 + ], + "score": 1.0, + "content": "Hazards dataset are naturally integrated into their", + "type": "text" + } + ], + "index": 47 + }, + { + "bbox": [ + 304, + 437, + 506, + 448 + ], + "spans": [ + { + "bbox": [ + 304, + 437, + 506, + 448 + ], + "score": 1.0, + "content": "environment with proper lighting and orientation,", + "type": "text" + } + ], + "index": 48 + }, + { + "bbox": [ + 304, + 447, + 455, + 459 + ], + "spans": [ + { + "bbox": [ + 304, + 447, + 455, + 459 + ], + "score": 1.0, + "content": "making them more difficult to detect.", + "type": "text" + } + ], + "index": 49 + } + ], + "index": 39 + } + ], + "index": 31.5 + }, + { + "type": "interline_equation", + "bbox": [ + 149, + 340, + 254, + 418 + ], + "lines": [ + { + "bbox": [ + 149, + 340, + 254, + 418 + ], + "spans": [ + { + "bbox": [ + 149, + 340, + 254, + 418 + ], + "score": 0.9, + "content": "\\begin{array} { l } { \\displaystyle \\mathcal { L } _ { \\boldsymbol { p } } = \\frac { 1 } { | P | } \\sum _ { i } - \\log ( p _ { i } ) y _ { i } } \\\\ { \\displaystyle \\mathcal { L } _ { \\boldsymbol { c } } = \\frac { 1 } { | P | } \\sum _ { i } - \\log ( \\hat { c } _ { i } ) } \\\\ { \\displaystyle \\mathcal { L } = \\mathcal { L } _ { \\boldsymbol { p } } + \\lambda \\mathcal { L } _ { \\boldsymbol { c } } } \\end{array}", + "type": "interline_equation", + "image_path": "188937f87a242a5c887d29d311cbd567431198a575e25323201226d6d6633277.jpg" + } + ] + } + ], + "index": 35.0, + "virtual_lines": [ + { + "bbox": [ + 149, + 340, + 254, + 379.0 + ], + "spans": [], + "index": 32 + }, + { + "bbox": [ + 149, + 379.0, + 254, + 418.0 + ], + "spans": [], + "index": 38 + } + ] + }, + { + "type": "text", + "bbox": [ + 106, + 424, + 297, + 480 + ], + "lines": [ + { + "bbox": [ + 106, + 425, + 299, + 437 + ], + "spans": [ + { + "bbox": [ + 106, + 425, + 183, + 437 + ], + "score": 1.0, + "content": "The reasoning for", + "type": "text" + }, + { + "bbox": [ + 184, + 425, + 196, + 436 + ], + "score": 0.88, + "content": "\\mathcal { L } _ { c }", + "type": "inline_equation" + }, + { + "bbox": [ + 196, + 425, + 299, + 437 + ], + "score": 1.0, + "content": "is to encourage the net-", + "type": "text" + } + ], + "index": 41 + }, + { + "bbox": [ + 105, + 436, + 296, + 447 + ], + "spans": [ + { + "bbox": [ + 105, + 436, + 289, + 447 + ], + "score": 1.0, + "content": "work to output confident predictions. Finally", + "type": "text" + }, + { + "bbox": [ + 289, + 436, + 296, + 446 + ], + "score": 0.7, + "content": "\\lambda", + "type": "inline_equation" + } + ], + "index": 42 + }, + { + "bbox": [ + 105, + 446, + 299, + 458 + ], + "spans": [ + { + "bbox": [ + 105, + 446, + 299, + 458 + ], + "score": 1.0, + "content": "is initialized to 0.1 and is updated by a “budget”", + "type": "text" + } + ], + "index": 43 + }, + { + "bbox": [ + 105, + 458, + 297, + 469 + ], + "spans": [ + { + "bbox": [ + 105, + 458, + 297, + 469 + ], + "score": 1.0, + "content": "parameter which is set to the default of 0.3. The", + "type": "text" + } + ], + "index": 45 + }, + { + "bbox": [ + 106, + 469, + 176, + 481 + ], + "spans": [ + { + "bbox": [ + 106, + 469, + 176, + 481 + ], + "score": 1.0, + "content": "update equation:", + "type": "text" + } + ], + "index": 46 + } + ], + "index": 43, + "bbox_fs": [ + 105, + 425, + 299, + 481 + ] + }, + { + "type": "interline_equation", + "bbox": [ + 250, + 494, + 360, + 523 + ], + "lines": [ + { + "bbox": [ + 250, + 494, + 360, + 523 + ], + "spans": [ + { + "bbox": [ + 250, + 494, + 360, + 523 + ], + "score": 0.86, + "content": "\\left\\{ \\begin{array} { l l } { { \\lambda } / { 0 . 9 9 } } & { \\sum \\hat { c } _ { i } \\leq \\mathrm { b u d g e t } } \\\\ { { \\lambda } / { 1 . 0 1 } } & { \\sum \\hat { c } _ { i } > \\mathrm { b u d g e t } } \\end{array} \\right.", + "type": "interline_equation", + "image_path": "c4eed88b17dbc93f4e19087b97ce3896c0eb2ba44f70df785114b20ca53d77a9.jpg" + } + ] + } + ], + "index": 50.5, + "virtual_lines": [ + { + "bbox": [ + 250, + 494, + 360, + 508.5 + ], + "spans": [], + "index": 50 + }, + { + "bbox": [ + 250, + 508.5, + 360, + 523.0 + ], + "spans": [], + "index": 51 + } + ] + }, + { + "type": "text", + "bbox": [ + 107, + 531, + 504, + 553 + ], + "lines": [ + { + "bbox": [ + 105, + 530, + 506, + 544 + ], + "spans": [ + { + "bbox": [ + 105, + 530, + 506, + 544 + ], + "score": 1.0, + "content": "This adaptively adjusts the weighting between the two losses and experimentally the update is not", + "type": "text" + } + ], + "index": 52 + }, + { + "bbox": [ + 106, + 542, + 241, + 553 + ], + "spans": [ + { + "bbox": [ + 106, + 542, + 241, + 553 + ], + "score": 1.0, + "content": "sensitive to the budget parameter.", + "type": "text" + } + ], + "index": 53 + } + ], + "index": 52.5, + "bbox_fs": [ + 105, + 530, + 506, + 553 + ] + }, + { + "type": "text", + "bbox": [ + 107, + 564, + 505, + 599 + ], + "lines": [ + { + "bbox": [ + 106, + 565, + 505, + 577 + ], + "spans": [ + { + "bbox": [ + 106, + 565, + 505, + 577 + ], + "score": 1.0, + "content": "Semantic Segmentation BDD Anomalies Dataset List. The BDD100K dataset contains 180", + "type": "text" + } + ], + "index": 54 + }, + { + "bbox": [ + 106, + 577, + 505, + 589 + ], + "spans": [ + { + "bbox": [ + 106, + 577, + 505, + 589 + ], + "score": 1.0, + "content": "instances of the train class, 4296 instances of the motorcycle class, and 10229 instances of the bicycle", + "type": "text" + } + ], + "index": 55 + }, + { + "bbox": [ + 105, + 586, + 132, + 600 + ], + "spans": [ + { + "bbox": [ + 105, + 586, + 132, + 600 + ], + "score": 1.0, + "content": "class.", + "type": "text" + } + ], + "index": 56 + } + ], + "index": 55, + "bbox_fs": [ + 105, + 565, + 505, + 600 + ] + }, + { + "type": "text", + "bbox": [ + 106, + 610, + 506, + 732 + ], + "lines": [ + { + "bbox": [ + 105, + 610, + 505, + 623 + ], + "spans": [ + { + "bbox": [ + 105, + 610, + 505, + 623 + ], + "score": 1.0, + "content": "StreetHazards 3D Models Dataset List. For semantic segmentation experiments, we choose to", + "type": "text" + } + ], + "index": 57 + }, + { + "bbox": [ + 105, + 622, + 507, + 634 + ], + "spans": [ + { + "bbox": [ + 105, + 622, + 507, + 634 + ], + "score": 1.0, + "content": "use the following classes 3D models from Model Bank Library to serve as out-of-distribution data:", + "type": "text" + } + ], + "index": 58 + }, + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "spans": [ + { + "bbox": [ + 106, + 633, + 505, + 645 + ], + "score": 1.0, + "content": "Meta-categories: Animals, Vehicles, Weapons, Appliances, Household items (furniture, and kitchen", + "type": "text" + } + ], + "index": 59 + }, + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "spans": [ + { + "bbox": [ + 105, + 643, + 506, + 657 + ], + "score": 1.0, + "content": "items), Electronics, Instruments, and miscellaneous. The specific animals used are kangaroos, whales,", + "type": "text" + } + ], + "index": 60 + }, + { + "bbox": [ + 105, + 654, + 507, + 668 + ], + "spans": [ + { + "bbox": [ + 105, + 654, + 507, + 668 + ], + "score": 1.0, + "content": "dolphins, cows, lions, frogs, bats, insects, mongooses, scorpions, fish, camels, flamingos, apes, horses,", + "type": "text" + } + ], + "index": 61 + }, + { + "bbox": [ + 105, + 666, + 507, + 680 + ], + "spans": [ + { + "bbox": [ + 105, + 666, + 507, + 680 + ], + "score": 1.0, + "content": "mice, spider, dinosaurs, elephants, moose, shrimps, bats, butterflies, turtles, hippopotamuses, dogs,", + "type": "text" + } + ], + "index": 62 + }, + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "spans": [ + { + "bbox": [ + 105, + 676, + 506, + 690 + ], + "score": 1.0, + "content": "cats, sheep, seahorse, snail and zebra. The specific vehicles used are military trucks, motorcycles,", + "type": "text" + } + ], + "index": 63 + }, + { + "bbox": [ + 105, + 687, + 506, + 702 + ], + "spans": [ + { + "bbox": [ + 105, + 687, + 506, + 702 + ], + "score": 1.0, + "content": "naval ships, pirate ships, submarines, sailing ships, trolleys, trains, airplanes, helicopters, jets,", + "type": "text" + } + ], + "index": 64 + }, + { + "bbox": [ + 105, + 699, + 505, + 712 + ], + "spans": [ + { + "bbox": [ + 105, + 699, + 505, + 712 + ], + "score": 1.0, + "content": "zeppelin, radar tower, construction vehicles (loaders, dump trucks, bulldozer), farming vehicles", + "type": "text" + } + ], + "index": 65 + }, + { + "bbox": [ + 106, + 710, + 505, + 723 + ], + "spans": [ + { + "bbox": [ + 106, + 710, + 505, + 723 + ], + "score": 1.0, + "content": "(harvester, gantry crane, tractor), fire truck, tank, combat vehicles, and trailers. The specific weapons", + "type": "text" + } + ], + "index": 66 + }, + { + "bbox": [ + 106, + 721, + 506, + 734 + ], + "spans": [ + { + "bbox": [ + 106, + 721, + 506, + 734 + ], + "score": 1.0, + "content": "used are guns, missiles, rocket launchers, and grenades. The appliances used are refrigerators, stoves,", + "type": "text" + } + ], + "index": 67 + }, + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "score": 1.0, + "content": "washing machines, and ovens. The household items used are cabinets, armoire, grandfather clocks,", + "type": "text", + "cross_page": true + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 93, + 506, + 106 + ], + "spans": [ + { + "bbox": [ + 105, + 93, + 506, + 106 + ], + "score": 1.0, + "content": "bathtubs, bureaus, night stand, table, bed, bookcase, office desk, glasses (drinking), throne chair,", + "type": "text", + "cross_page": true + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 102, + 506, + 119 + ], + "spans": [ + { + "bbox": [ + 105, + 102, + 506, + 119 + ], + "score": 1.0, + "content": "kitchen utensils (knives, forks, spoons), sofa, clothing iron, plates, sewing machine, and dressing", + "type": "text", + "cross_page": true + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 115, + 505, + 128 + ], + "spans": [ + { + "bbox": [ + 106, + 115, + 505, + 128 + ], + "score": 1.0, + "content": "mirror. The electronics used are computer monitor, computer mouse, hair dryer, The instruments", + "type": "text", + "cross_page": true + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 126, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 106, + 126, + 505, + 138 + ], + "score": 1.0, + "content": "category includes bassoon, clarinet, drums, guitar, violin, harp, and keyboard. The miscellaneous", + "type": "text", + "cross_page": true + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 138, + 505, + 150 + ], + "spans": [ + { + "bbox": [ + 106, + 138, + 505, + 150 + ], + "score": 1.0, + "content": "category includes rocket, space capsule, space shuttle, lunar module, glasses (wearable), weight", + "type": "text", + "cross_page": true + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 149, + 505, + 161 + ], + "spans": [ + { + "bbox": [ + 106, + 149, + 505, + 161 + ], + "score": 1.0, + "content": "machine, balance beam, bench press, bowling ball and pins, and pens. Several categories and instances", + "type": "text", + "cross_page": true + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 159, + 505, + 172 + ], + "spans": [ + { + "bbox": [ + 106, + 159, + 505, + 172 + ], + "score": 1.0, + "content": "were excluded from Model Bank Library due to their occurrence in the simulation environment such", + "type": "text", + "cross_page": true + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 171, + 505, + 183 + ], + "spans": [ + { + "bbox": [ + 106, + 171, + 505, + 183 + ], + "score": 1.0, + "content": "as playground equipment and various types of foliage and trees. The sizes of instances used in the", + "type": "text", + "cross_page": true + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 181, + 505, + 194 + ], + "spans": [ + { + "bbox": [ + 106, + 181, + 505, + 194 + ], + "score": 1.0, + "content": "dataset might not reflect the actual scale that would otherwise naturally occur. Similarly the location", + "type": "text", + "cross_page": true + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 191, + 499, + 204 + ], + "spans": [ + { + "bbox": [ + 106, + 191, + 499, + 204 + ], + "score": 1.0, + "content": "of instances in the dataset are not necessarily reflective of where they are likely to occur in nature.", + "type": "text", + "cross_page": true + } + ], + "index": 10 + } + ], + "index": 62, + "bbox_fs": [ + 105, + 610, + 507, + 734 + ] + } + ] + }, + { + "preproc_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 203 + ], + "lines": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "spans": [ + { + "bbox": [ + 105, + 82, + 506, + 95 + ], + "score": 1.0, + "content": "washing machines, and ovens. The household items used are cabinets, armoire, grandfather clocks,", + "type": "text" + } + ], + "index": 0 + }, + { + "bbox": [ + 105, + 93, + 506, + 106 + ], + "spans": [ + { + "bbox": [ + 105, + 93, + 506, + 106 + ], + "score": 1.0, + "content": "bathtubs, bureaus, night stand, table, bed, bookcase, office desk, glasses (drinking), throne chair,", + "type": "text" + } + ], + "index": 1 + }, + { + "bbox": [ + 105, + 102, + 506, + 119 + ], + "spans": [ + { + "bbox": [ + 105, + 102, + 506, + 119 + ], + "score": 1.0, + "content": "kitchen utensils (knives, forks, spoons), sofa, clothing iron, plates, sewing machine, and dressing", + "type": "text" + } + ], + "index": 2 + }, + { + "bbox": [ + 106, + 115, + 505, + 128 + ], + "spans": [ + { + "bbox": [ + 106, + 115, + 505, + 128 + ], + "score": 1.0, + "content": "mirror. The electronics used are computer monitor, computer mouse, hair dryer, The instruments", + "type": "text" + } + ], + "index": 3 + }, + { + "bbox": [ + 106, + 126, + 505, + 138 + ], + "spans": [ + { + "bbox": [ + 106, + 126, + 505, + 138 + ], + "score": 1.0, + "content": "category includes bassoon, clarinet, drums, guitar, violin, harp, and keyboard. The miscellaneous", + "type": "text" + } + ], + "index": 4 + }, + { + "bbox": [ + 106, + 138, + 505, + 150 + ], + "spans": [ + { + "bbox": [ + 106, + 138, + 505, + 150 + ], + "score": 1.0, + "content": "category includes rocket, space capsule, space shuttle, lunar module, glasses (wearable), weight", + "type": "text" + } + ], + "index": 5 + }, + { + "bbox": [ + 106, + 149, + 505, + 161 + ], + "spans": [ + { + "bbox": [ + 106, + 149, + 505, + 161 + ], + "score": 1.0, + "content": "machine, balance beam, bench press, bowling ball and pins, and pens. Several categories and instances", + "type": "text" + } + ], + "index": 6 + }, + { + "bbox": [ + 106, + 159, + 505, + 172 + ], + "spans": [ + { + "bbox": [ + 106, + 159, + 505, + 172 + ], + "score": 1.0, + "content": "were excluded from Model Bank Library due to their occurrence in the simulation environment such", + "type": "text" + } + ], + "index": 7 + }, + { + "bbox": [ + 106, + 171, + 505, + 183 + ], + "spans": [ + { + "bbox": [ + 106, + 171, + 505, + 183 + ], + "score": 1.0, + "content": "as playground equipment and various types of foliage and trees. The sizes of instances used in the", + "type": "text" + } + ], + "index": 8 + }, + { + "bbox": [ + 106, + 181, + 505, + 194 + ], + "spans": [ + { + "bbox": [ + 106, + 181, + 505, + 194 + ], + "score": 1.0, + "content": "dataset might not reflect the actual scale that would otherwise naturally occur. Similarly the location", + "type": "text" + } + ], + "index": 9 + }, + { + "bbox": [ + 106, + 191, + 499, + 204 + ], + "spans": [ + { + "bbox": [ + 106, + 191, + 499, + 204 + ], + "score": 1.0, + "content": "of instances in the dataset are not necessarily reflective of where they are likely to occur in nature.", + "type": "text" + } + ], + "index": 10 + } + ], + "index": 5 + } + ], + "page_idx": 15, + "page_size": [ + 612, + 792 + ], + "discarded_blocks": [ + { + "type": "discarded", + "bbox": [ + 107, + 27, + 308, + 37 + ], + "lines": [ + { + "bbox": [ + 107, + 25, + 308, + 38 + ], + "spans": [ + { + "bbox": [ + 107, + 25, + 308, + 38 + ], + "score": 1.0, + "content": "Under review as a conference paper at ICLR 2022", + "type": "text" + } + ] + } + ] + }, + { + "type": "discarded", + "bbox": [ + 300, + 752, + 310, + 760 + ], + "lines": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "spans": [ + { + "bbox": [ + 299, + 750, + 312, + 764 + ], + "score": 1.0, + "content": "", + "type": "text", + "height": 14, + "width": 13 + } + ] + } + ] + } + ], + "para_blocks": [ + { + "type": "text", + "bbox": [ + 107, + 82, + 505, + 203 + ], + "lines": [], + "index": 5, + "bbox_fs": [ + 105, + 82, + 506, + 204 + ], + "lines_deleted": true + } + ] + } + ], + "_backend": "pipeline", + "_version_name": "2.2.2" +} \ No newline at end of file diff --git a/parse/dev/wlEOsQ917F/images/2688ecd25ca7cb86a395fda5ce9b688673c5e37d3422795d1fc2b1b4c85e6aab.jpg b/parse/dev/wlEOsQ917F/images/2688ecd25ca7cb86a395fda5ce9b688673c5e37d3422795d1fc2b1b4c85e6aab.jpg new file mode 100644 index 0000000000000000000000000000000000000000..60b19843c1c65d7b184eb0dc198374758f561a16 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/2688ecd25ca7cb86a395fda5ce9b688673c5e37d3422795d1fc2b1b4c85e6aab.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7330c335f0da6aceafaf75fb743345b7397021ecca467502a3a5771e80fb7d3c +size 51310 diff --git a/parse/dev/wlEOsQ917F/images/2cad3bc99892a0519dcf64b4d51cc53bbeebfb43daa1744f6954f349acaee224.jpg b/parse/dev/wlEOsQ917F/images/2cad3bc99892a0519dcf64b4d51cc53bbeebfb43daa1744f6954f349acaee224.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9902eca09976026c0709ca0ef350a5ee1f5039cd --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/2cad3bc99892a0519dcf64b4d51cc53bbeebfb43daa1744f6954f349acaee224.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:581ad4878dce2ddaff4fa62a88950da1d844f1fdcc901b4b3d3606f883a29c44 +size 18658 diff --git a/parse/dev/wlEOsQ917F/images/48952885d7753c9b07c3aebab5746a0ccea5fa554b21835026246a566fa9b494.jpg b/parse/dev/wlEOsQ917F/images/48952885d7753c9b07c3aebab5746a0ccea5fa554b21835026246a566fa9b494.jpg new file mode 100644 index 0000000000000000000000000000000000000000..493947fd5148c183508a44dc1c6e832fd37a4318 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/48952885d7753c9b07c3aebab5746a0ccea5fa554b21835026246a566fa9b494.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5deb03c719dfc15d3ae5aed132a94668d10a52bc6b1d9c31d7079cd33d30a55b +size 8250 diff --git a/parse/dev/wlEOsQ917F/images/5b8583b8f4e3127a6f57e87a8eec2bdb05bebdf958e042d951a51cd676a39614.jpg b/parse/dev/wlEOsQ917F/images/5b8583b8f4e3127a6f57e87a8eec2bdb05bebdf958e042d951a51cd676a39614.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ad4a617e52eb8754d101f17c930acfb35b9dfcb5 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/5b8583b8f4e3127a6f57e87a8eec2bdb05bebdf958e042d951a51cd676a39614.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d8e4f7a824aa3d825d3f475d0457cb70c066e8563228421a41a7c3e0e5a3c0ff +size 18637 diff --git a/parse/dev/wlEOsQ917F/images/60a8b5e303f000db113358f398206514b13d198f9e3b9f31ffa2140a730db22f.jpg b/parse/dev/wlEOsQ917F/images/60a8b5e303f000db113358f398206514b13d198f9e3b9f31ffa2140a730db22f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3eff3b10c8c61539869253ecbc24ff75a5200882 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/60a8b5e303f000db113358f398206514b13d198f9e3b9f31ffa2140a730db22f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a13e67bb05414d2b505c27aa49948b4b4dbf5c9566db28aa6857416aa9c2dba +size 9196 diff --git a/parse/dev/wlEOsQ917F/images/673a643345889d4e3248fc0df9e2e6740aec277c118909e511938888551c02a8.jpg b/parse/dev/wlEOsQ917F/images/673a643345889d4e3248fc0df9e2e6740aec277c118909e511938888551c02a8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..30de515c14892ddbd943271979ee0102acaaadb8 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/673a643345889d4e3248fc0df9e2e6740aec277c118909e511938888551c02a8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f5abe70a4e121f7b6ca60f24296c397cc6e7a3fe515b02d7f168f61437284377 +size 8398 diff --git a/parse/dev/wlEOsQ917F/images/753ead0f5c30f5763cf9f2e08983156c9b1f3ab03e9fcb8c1002682f604aed00.jpg b/parse/dev/wlEOsQ917F/images/753ead0f5c30f5763cf9f2e08983156c9b1f3ab03e9fcb8c1002682f604aed00.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a1339badc286ee04a36f555bd7b776b5c8232ac2 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/753ead0f5c30f5763cf9f2e08983156c9b1f3ab03e9fcb8c1002682f604aed00.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79acde961a01aed1dd9bf3fec7c489e33dd8ba81f5ca5286809761692e711e15 +size 5899 diff --git a/parse/dev/wlEOsQ917F/images/76826469efea4ca34fad1e18a86a761b926bb4edffb1f2b761a265b1f6c9ee23.jpg b/parse/dev/wlEOsQ917F/images/76826469efea4ca34fad1e18a86a761b926bb4edffb1f2b761a265b1f6c9ee23.jpg new file mode 100644 index 0000000000000000000000000000000000000000..52ac7c932a09c8d2462f1956e79c15480c20af14 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/76826469efea4ca34fad1e18a86a761b926bb4edffb1f2b761a265b1f6c9ee23.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:29d79546af20e9f66490258b07d7f6ed89068a47311632b433cc5f9bd2a170af +size 5240 diff --git a/parse/dev/wlEOsQ917F/images/7a75d9ca6f4e1eb9d69ea5b9354eec61b127209a03944608b0531a9e1cefad59.jpg b/parse/dev/wlEOsQ917F/images/7a75d9ca6f4e1eb9d69ea5b9354eec61b127209a03944608b0531a9e1cefad59.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e0603c16c3167f36a004880544deda199d4ce903 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/7a75d9ca6f4e1eb9d69ea5b9354eec61b127209a03944608b0531a9e1cefad59.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d75beeebbd39f03f4f100defbf6233c396255bc7b05e72a8b1372a727ebe0e98 +size 14231 diff --git a/parse/dev/wlEOsQ917F/images/851627f4797c8bee48086eb1f575679b6b9eb4c71c728ebde4ec66df75a4e654.jpg b/parse/dev/wlEOsQ917F/images/851627f4797c8bee48086eb1f575679b6b9eb4c71c728ebde4ec66df75a4e654.jpg new file mode 100644 index 0000000000000000000000000000000000000000..89aa26dbf680dfa8b782d8d2274f7f87255bb8b5 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/851627f4797c8bee48086eb1f575679b6b9eb4c71c728ebde4ec66df75a4e654.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3afa19982e4691a8e1e8189070fb5a6d205187aaea6d229083325833afdaa6fe +size 12714 diff --git a/parse/dev/wlEOsQ917F/images/90ad97bdf05e07bbf69e0840bb834a6e93d616567a411c7bf450c6ae5a96dfac.jpg b/parse/dev/wlEOsQ917F/images/90ad97bdf05e07bbf69e0840bb834a6e93d616567a411c7bf450c6ae5a96dfac.jpg new file mode 100644 index 0000000000000000000000000000000000000000..79429c40c15f187e2dcc09448f9ca39c0aebb70a --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/90ad97bdf05e07bbf69e0840bb834a6e93d616567a411c7bf450c6ae5a96dfac.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2304b833a22289206fd3357a351721c9ae2252f4165bed12585a759c03200d3b +size 6381 diff --git a/parse/dev/wlEOsQ917F/images/996cecb29016fbe6174834ee4b2af1d24afbda38b250f2689bcf387c5fafa554.jpg b/parse/dev/wlEOsQ917F/images/996cecb29016fbe6174834ee4b2af1d24afbda38b250f2689bcf387c5fafa554.jpg new file mode 100644 index 0000000000000000000000000000000000000000..80fd52c3b0510c32802e22c256cf5e258a3e03da --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/996cecb29016fbe6174834ee4b2af1d24afbda38b250f2689bcf387c5fafa554.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7d176b638b62f61ccf4c47316a031c7155007256b36b1b8b852fa686b1c3e1dc +size 5145 diff --git a/parse/dev/wlEOsQ917F/images/9da4ca1859e2f4486e6a59330ee7e0b44553df77ed6dbacadf22987a95fb4189.jpg b/parse/dev/wlEOsQ917F/images/9da4ca1859e2f4486e6a59330ee7e0b44553df77ed6dbacadf22987a95fb4189.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ff58837fadf9727eed8e85dde527297ee780a693 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/9da4ca1859e2f4486e6a59330ee7e0b44553df77ed6dbacadf22987a95fb4189.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c7fe5f5a7a47dd99f43fbdd1b13d91539e774d60884f077887152d5c31c8c34f +size 5156 diff --git a/parse/dev/wlEOsQ917F/images/c4a75ecef04aabb2255a9fa5dcc236b7c35d1ef10951a5f31261e1ea5640cad6.jpg b/parse/dev/wlEOsQ917F/images/c4a75ecef04aabb2255a9fa5dcc236b7c35d1ef10951a5f31261e1ea5640cad6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2844c0dd466db137c4ad7d50fd38867832bd6898 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/c4a75ecef04aabb2255a9fa5dcc236b7c35d1ef10951a5f31261e1ea5640cad6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:311626c61ec1ed1e813238c4ecc3eef40456e4ef1426ecc12963dff9aa1c4bf9 +size 5121 diff --git a/parse/dev/wlEOsQ917F/images/c8ea2e88e87a5719ad7c39dc1f9c5aa334772c691070e1fa10a524f99dac7eab.jpg b/parse/dev/wlEOsQ917F/images/c8ea2e88e87a5719ad7c39dc1f9c5aa334772c691070e1fa10a524f99dac7eab.jpg new file mode 100644 index 0000000000000000000000000000000000000000..af776154529e0f3d5865c07f0284788fc1c54aa4 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/c8ea2e88e87a5719ad7c39dc1f9c5aa334772c691070e1fa10a524f99dac7eab.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d7c9907399f5d1758872118e9d4796aaf4d53f81456fe3c9779965d17ed6c2f +size 3304 diff --git a/parse/dev/wlEOsQ917F/images/d171f703a0c21f685a2182fd2abaa97c57ea243855b8b097d1c3c03eb9ed610a.jpg b/parse/dev/wlEOsQ917F/images/d171f703a0c21f685a2182fd2abaa97c57ea243855b8b097d1c3c03eb9ed610a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1ed3c8f5443a7a1af5f01f4507a45dcd4f33b9c9 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/d171f703a0c21f685a2182fd2abaa97c57ea243855b8b097d1c3c03eb9ed610a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daf5f10b2233a5bbc5ef55f0a26143b8ab4863f75311367aa1c8513bdcfedda7 +size 4577 diff --git a/parse/dev/wlEOsQ917F/images/f65f526ac4a0f42f4512be04e32029c311e51a7726b022d8c22f66a6cb7df06b.jpg b/parse/dev/wlEOsQ917F/images/f65f526ac4a0f42f4512be04e32029c311e51a7726b022d8c22f66a6cb7df06b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8644d1ae55100e5d7a1647bf2163580719d45d47 --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/f65f526ac4a0f42f4512be04e32029c311e51a7726b022d8c22f66a6cb7df06b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:06c5a5a95ea7f477ee612086a6713138525b0542d05e529f775a96b67097eead +size 20214 diff --git a/parse/dev/wlEOsQ917F/images/fa9814b49bb946cfb9c109ef1f9ae2cd5d2de464d1eeead1b48bf982b7720ed3.jpg b/parse/dev/wlEOsQ917F/images/fa9814b49bb946cfb9c109ef1f9ae2cd5d2de464d1eeead1b48bf982b7720ed3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..89edf704e446a47eaed8de6dd4a28fb0128175cf --- /dev/null +++ b/parse/dev/wlEOsQ917F/images/fa9814b49bb946cfb9c109ef1f9ae2cd5d2de464d1eeead1b48bf982b7720ed3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:edfe4a74e38002f5c9d1bee9a32f8cb4775508da3a6c91fd9ddc0a203a7c2e01 +size 5943 diff --git a/parse/dev/wv79UiY5U7/images/2679cf42c5ca7afde4a76a523590adcad544a86c1101f2b35addc5e22e0d29de.jpg b/parse/dev/wv79UiY5U7/images/2679cf42c5ca7afde4a76a523590adcad544a86c1101f2b35addc5e22e0d29de.jpg new file mode 100644 index 0000000000000000000000000000000000000000..33407a61b39e3c7ae8d4aa2a07335d6358b4dde9 --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/2679cf42c5ca7afde4a76a523590adcad544a86c1101f2b35addc5e22e0d29de.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59a6318ba4504d5fca58e08f4940c54a14d8bdcadbd92b24ec13ac6e9dcc3e69 +size 25533 diff --git a/parse/dev/wv79UiY5U7/images/5880a1e3bd6cd3c1d7c81a867381c35d10f2f48c5a760f279138df1cf0bdf3a4.jpg b/parse/dev/wv79UiY5U7/images/5880a1e3bd6cd3c1d7c81a867381c35d10f2f48c5a760f279138df1cf0bdf3a4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bd89e2bf11dc1d320642bdba6b35a470454cf63a --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/5880a1e3bd6cd3c1d7c81a867381c35d10f2f48c5a760f279138df1cf0bdf3a4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b5f30c0e45072cd24b3d0235f288c286c8d26368f9b985d61688c024cdfa6393 +size 111204 diff --git a/parse/dev/wv79UiY5U7/images/59cb387eaa11efcf1fe9990004718392ee957360c7b38bc20eb8747030037cea.jpg b/parse/dev/wv79UiY5U7/images/59cb387eaa11efcf1fe9990004718392ee957360c7b38bc20eb8747030037cea.jpg new file mode 100644 index 0000000000000000000000000000000000000000..96b6e349c0d67ff7f06e624912720bdffaa3e75b --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/59cb387eaa11efcf1fe9990004718392ee957360c7b38bc20eb8747030037cea.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c8277b156e62bc5c812dcdb6f4b89771c72c947954eea0846894687d65e12a1e +size 124052 diff --git a/parse/dev/wv79UiY5U7/images/9109d63e5c77bff9c3902a83a05ce5457cdefa8e3688c476533290a693e205f7.jpg b/parse/dev/wv79UiY5U7/images/9109d63e5c77bff9c3902a83a05ce5457cdefa8e3688c476533290a693e205f7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..53b3b07d00046ca4c8f7f10fef52dd22d4f8a71c --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/9109d63e5c77bff9c3902a83a05ce5457cdefa8e3688c476533290a693e205f7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bfe03afe2adc2ab0cee1ab4307f429ade8e7e398d5f7397cbda1326b64ef4b0b +size 23080 diff --git a/parse/dev/wv79UiY5U7/images/a6f75623bf6c27b9b39c97e30c9dccdd5ceeb53188a88c396f0b67988df825f7.jpg b/parse/dev/wv79UiY5U7/images/a6f75623bf6c27b9b39c97e30c9dccdd5ceeb53188a88c396f0b67988df825f7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a79cb3a590e0b2f1104c25586dc61675a57581fe --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/a6f75623bf6c27b9b39c97e30c9dccdd5ceeb53188a88c396f0b67988df825f7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d97ddfff8d042decc4d58cc8ba74c314e7686dc9f58d3edd8daf4e3f935bd97 +size 16267 diff --git a/parse/dev/wv79UiY5U7/images/b19f84fd0a3b52e2e475622b0bb0a3e1d862d65ede861c5eb52a84372a704122.jpg b/parse/dev/wv79UiY5U7/images/b19f84fd0a3b52e2e475622b0bb0a3e1d862d65ede861c5eb52a84372a704122.jpg new file mode 100644 index 0000000000000000000000000000000000000000..55e9746fb3410620189cc72dc0df3573a835265b --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/b19f84fd0a3b52e2e475622b0bb0a3e1d862d65ede861c5eb52a84372a704122.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:424d062ef065b06ac0ed4db855a35f7e2418303060f9eaf59fd46b423544ee01 +size 15567 diff --git a/parse/dev/wv79UiY5U7/images/b724073f5d31a2a624b4d55823d3db3c56d82fc197f5992637a482159ecf6e7b.jpg b/parse/dev/wv79UiY5U7/images/b724073f5d31a2a624b4d55823d3db3c56d82fc197f5992637a482159ecf6e7b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2504aba730d1f4e36bc1a7e49e28714ccd9af390 --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/b724073f5d31a2a624b4d55823d3db3c56d82fc197f5992637a482159ecf6e7b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f8d472a91d4524b415caa1cf1fe1deb163da9e973084475a8c078679ca4fa5b0 +size 16411 diff --git a/parse/dev/wv79UiY5U7/images/b7f9ce0a993f619a4c9e733eba30581534866055483bee3d24f984c1a430c635.jpg b/parse/dev/wv79UiY5U7/images/b7f9ce0a993f619a4c9e733eba30581534866055483bee3d24f984c1a430c635.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f6a7651ee18b46fd7c3d9dcd679dfed176356ccb --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/b7f9ce0a993f619a4c9e733eba30581534866055483bee3d24f984c1a430c635.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4f8f1b1b4778935a33b6dc7a481dffdf3d9b281477a91a703cb5b12b6cbbc063 +size 19084 diff --git a/parse/dev/wv79UiY5U7/images/bd0d9934e6787c5f94594cb8179e809d60cfc4a1ff3bb7ada286ba3543b5e6bd.jpg b/parse/dev/wv79UiY5U7/images/bd0d9934e6787c5f94594cb8179e809d60cfc4a1ff3bb7ada286ba3543b5e6bd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..517b677d54506e7b640db2d61afea6bb3055a039 --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/bd0d9934e6787c5f94594cb8179e809d60cfc4a1ff3bb7ada286ba3543b5e6bd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a08c54a0b21fa0915b6bb5bdb1fe32630f1b2014975e47a2be5f14623fab48c8 +size 53089 diff --git a/parse/dev/wv79UiY5U7/images/dd71c6cf0bdb2bfce4295a9727ef57802e3b4792986d2731663ccef0b54643af.jpg b/parse/dev/wv79UiY5U7/images/dd71c6cf0bdb2bfce4295a9727ef57802e3b4792986d2731663ccef0b54643af.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ebfde38e263c62cbce10949c9da668dbf7ce317c --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/dd71c6cf0bdb2bfce4295a9727ef57802e3b4792986d2731663ccef0b54643af.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27cad2e2c584e0c37945de5512dbcce3e2576c93eeec6e0718b172fe5e164187 +size 37523 diff --git a/parse/dev/wv79UiY5U7/images/ec55786e780ac343012b7e186f7424d0f65b7f4749737d805de26a12cfaadb2f.jpg b/parse/dev/wv79UiY5U7/images/ec55786e780ac343012b7e186f7424d0f65b7f4749737d805de26a12cfaadb2f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..545521138852a6cf4ddb9adcdc57a33b32e06e7e --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/ec55786e780ac343012b7e186f7424d0f65b7f4749737d805de26a12cfaadb2f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd4c41f0f2e55439aee7892790717f33a5f8ef1b7d50a702da620e40a9e07365 +size 82962 diff --git a/parse/dev/wv79UiY5U7/images/f93276185cbc74760b08a71a948ec69ff823864460eba48e32ddb37a732b76e3.jpg b/parse/dev/wv79UiY5U7/images/f93276185cbc74760b08a71a948ec69ff823864460eba48e32ddb37a732b76e3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0d070621abc660ebae06b7c25c4161ff6b13192a --- /dev/null +++ b/parse/dev/wv79UiY5U7/images/f93276185cbc74760b08a71a948ec69ff823864460eba48e32ddb37a732b76e3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:180567174fbdfed10a878043976c74436d2f4d8e9545c893989b901c6d8b6362 +size 15446 diff --git a/parse/dev/xozJw0kZXF/images/9f5738f9018fc3c1a12e477eeef466e79440d7ff2419e54786471887fcedf4ca.jpg b/parse/dev/xozJw0kZXF/images/9f5738f9018fc3c1a12e477eeef466e79440d7ff2419e54786471887fcedf4ca.jpg new file mode 100644 index 0000000000000000000000000000000000000000..533b86c67e6b9069f755f01b4468014d3d857854 --- /dev/null +++ b/parse/dev/xozJw0kZXF/images/9f5738f9018fc3c1a12e477eeef466e79440d7ff2419e54786471887fcedf4ca.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6890f658661bf34fe2e5629fcf78cbc47d604e42000798650d614a02f991b952 +size 93081 diff --git a/parse/dev/xpjsOQtKqx/images/002a28dc73dc68df915b2c8dacea3d261b797644f7abcb4d4bcf23bdd65de831.jpg b/parse/dev/xpjsOQtKqx/images/002a28dc73dc68df915b2c8dacea3d261b797644f7abcb4d4bcf23bdd65de831.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cf607e64282b5efbbe4b63362ce02ae37675bc0e --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/002a28dc73dc68df915b2c8dacea3d261b797644f7abcb4d4bcf23bdd65de831.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:542fe1a05c1b68025b86ca0a504459c667def019246e93050008ab99dfd163c1 +size 26835 diff --git a/parse/dev/xpjsOQtKqx/images/02255fc30b598ef650d6282d3b7243349dbdaa23443936f92ce99feb579d14f1.jpg b/parse/dev/xpjsOQtKqx/images/02255fc30b598ef650d6282d3b7243349dbdaa23443936f92ce99feb579d14f1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..60a5e963b11e1fce28ecb8279deeb25a110e0406 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/02255fc30b598ef650d6282d3b7243349dbdaa23443936f92ce99feb579d14f1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20fcbd36a0cf33887fbc45a8f3cb89f12c42891bec62031d6c67dcbb7d5ecca0 +size 6844 diff --git a/parse/dev/xpjsOQtKqx/images/0ddee6db81d7855ba9e8ad0ff3739ea1486d1f60a99bae3a24296d750fdd0655.jpg b/parse/dev/xpjsOQtKqx/images/0ddee6db81d7855ba9e8ad0ff3739ea1486d1f60a99bae3a24296d750fdd0655.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ade86a89231f58e210841b83c5ed31adb3843a00 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/0ddee6db81d7855ba9e8ad0ff3739ea1486d1f60a99bae3a24296d750fdd0655.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:282a2a375a7e43e4caa38a6732518627af4b36d412b038572b8cd7c2d48f26f1 +size 8159 diff --git a/parse/dev/xpjsOQtKqx/images/14fb8cd6600be93b610c11ce4c3e4fa43d9843deed34949738450a74ff290091.jpg b/parse/dev/xpjsOQtKqx/images/14fb8cd6600be93b610c11ce4c3e4fa43d9843deed34949738450a74ff290091.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fc55617ffbc0a8ee8707702c655f4ff7ff984f21 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/14fb8cd6600be93b610c11ce4c3e4fa43d9843deed34949738450a74ff290091.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2f75b86f98009dd07666ca29f3397911ae5a709224a5bf137fcec48c11fc95cb +size 7527 diff --git a/parse/dev/xpjsOQtKqx/images/223a9b603d24154808889d5f6e9e7106453e97b2b9a46c7c1ab9ddc4b68aa801.jpg b/parse/dev/xpjsOQtKqx/images/223a9b603d24154808889d5f6e9e7106453e97b2b9a46c7c1ab9ddc4b68aa801.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ef34528f69d5084a98f82ae92ab3233bb6d71edd --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/223a9b603d24154808889d5f6e9e7106453e97b2b9a46c7c1ab9ddc4b68aa801.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85e43261b19e56a7febb268b0b64add8e0f8eb127e726ec757c7b2ca4081be89 +size 117582 diff --git a/parse/dev/xpjsOQtKqx/images/2ee6c07f62d3c22df75ef62c56ba6340fe06713e74d00a7b40b88c223456aa0b.jpg b/parse/dev/xpjsOQtKqx/images/2ee6c07f62d3c22df75ef62c56ba6340fe06713e74d00a7b40b88c223456aa0b.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e04e6266571bfbdc17cbd372a1a7913a7de46d95 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/2ee6c07f62d3c22df75ef62c56ba6340fe06713e74d00a7b40b88c223456aa0b.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1a1be8a0b2c2a8b14139e034435d12268d7260d55f9c8ec8cc15370483998878 +size 45292 diff --git a/parse/dev/xpjsOQtKqx/images/2f8b18a1141bfe1abf560f53435da0974f1f0fde461fe1eb5df085276e77830d.jpg b/parse/dev/xpjsOQtKqx/images/2f8b18a1141bfe1abf560f53435da0974f1f0fde461fe1eb5df085276e77830d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6299e10500addf5b28811897b07aec1396f0663e --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/2f8b18a1141bfe1abf560f53435da0974f1f0fde461fe1eb5df085276e77830d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25e8c8a33e8dd53f62afb0429cbe8fc871fd4356fe337d59f47d16c99225b6a7 +size 21532 diff --git a/parse/dev/xpjsOQtKqx/images/40903baebe2a3275e99f2534d26007a50c5ca400c704d03f9284490538064796.jpg b/parse/dev/xpjsOQtKqx/images/40903baebe2a3275e99f2534d26007a50c5ca400c704d03f9284490538064796.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6d6ebad15322b38fc00628d2785136318d8182b5 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/40903baebe2a3275e99f2534d26007a50c5ca400c704d03f9284490538064796.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66ad7022a8952cba8464e64306acead1130f0848f559e9480a9a389b14ff5805 +size 42272 diff --git a/parse/dev/xpjsOQtKqx/images/41591c9293e885cad6a5af8367f563220491b53039f3c3dd9fb0ce98edac512a.jpg b/parse/dev/xpjsOQtKqx/images/41591c9293e885cad6a5af8367f563220491b53039f3c3dd9fb0ce98edac512a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c252cb491f8ea7d57e0784867f0480093f5e66e7 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/41591c9293e885cad6a5af8367f563220491b53039f3c3dd9fb0ce98edac512a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:73231bbc46cdf762f45aecf1b8c6b761eb5afd473b55d10344f3a8d8b83e240b +size 49340 diff --git a/parse/dev/xpjsOQtKqx/images/470d710e98c206524d3c54f0851c3ac3165c2279f50ecbbd309d8eb678c23677.jpg b/parse/dev/xpjsOQtKqx/images/470d710e98c206524d3c54f0851c3ac3165c2279f50ecbbd309d8eb678c23677.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a11d49ebb98293e1591ff6527a670d51d3313232 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/470d710e98c206524d3c54f0851c3ac3165c2279f50ecbbd309d8eb678c23677.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23f0f799874e1549b224d8bc0b07aece06ec05dcd1795382b451bd27efddee8f +size 53985 diff --git a/parse/dev/xpjsOQtKqx/images/480f8816a4b1838967137464bcd0f4556d6b02f376b67287e69986d1d1db5d1f.jpg b/parse/dev/xpjsOQtKqx/images/480f8816a4b1838967137464bcd0f4556d6b02f376b67287e69986d1d1db5d1f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..771995b27bd2a1a965ef9a54ef5982841dce5d7d --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/480f8816a4b1838967137464bcd0f4556d6b02f376b67287e69986d1d1db5d1f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:baeea6bd97575a5748799afa7582dd9bcbbd423e4954726af22bcbaef2f51c8a +size 62266 diff --git a/parse/dev/xpjsOQtKqx/images/4b09f206ff668e3150ed696bf68252c2938408dc627239893a8c04baf6b3b538.jpg b/parse/dev/xpjsOQtKqx/images/4b09f206ff668e3150ed696bf68252c2938408dc627239893a8c04baf6b3b538.jpg new file mode 100644 index 0000000000000000000000000000000000000000..93c486b25415a42c75d4d0e825d98494627bb680 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/4b09f206ff668e3150ed696bf68252c2938408dc627239893a8c04baf6b3b538.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b11a3887c6131cd52ee32cb8d3a90b609136d5d05979d362106642ad47eb8b47 +size 22693 diff --git a/parse/dev/xpjsOQtKqx/images/516e33c67936a7597a700a7e7512343428cb495ea22ec5c1bfd55c4898034dfe.jpg b/parse/dev/xpjsOQtKqx/images/516e33c67936a7597a700a7e7512343428cb495ea22ec5c1bfd55c4898034dfe.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2059af88517ec042db8f5d3ec3bbb7cf32f6fd0d --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/516e33c67936a7597a700a7e7512343428cb495ea22ec5c1bfd55c4898034dfe.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f449e7acecd6db7045024d19b380012630a8e8193663a15357a0022664418e3 +size 19162 diff --git a/parse/dev/xpjsOQtKqx/images/528be597e2708937b3ccc30e087fe838fc20541c8d9e96e13b6047db547eab11.jpg b/parse/dev/xpjsOQtKqx/images/528be597e2708937b3ccc30e087fe838fc20541c8d9e96e13b6047db547eab11.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5e89c8d6ceaeb9c487ceee81f5cad0e1f74e2f86 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/528be597e2708937b3ccc30e087fe838fc20541c8d9e96e13b6047db547eab11.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db7f9a3fe1969669a28e8654f6f7fbfcc5bd6468dbbaddeb937212d871eb69bb +size 53736 diff --git a/parse/dev/xpjsOQtKqx/images/589ffeeaf00995080bd70185ed6cc9fbde8eb356666ff31732dc5e902c4fdddf.jpg b/parse/dev/xpjsOQtKqx/images/589ffeeaf00995080bd70185ed6cc9fbde8eb356666ff31732dc5e902c4fdddf.jpg new file mode 100644 index 0000000000000000000000000000000000000000..31a7ef085ba0b02f6a5145296006a3a33bb3814a --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/589ffeeaf00995080bd70185ed6cc9fbde8eb356666ff31732dc5e902c4fdddf.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e0d4fa3468c7c3cbc6cd42bd070d37bdee8166bea3e2a17c649f24c0e6d0544 +size 94945 diff --git a/parse/dev/xpjsOQtKqx/images/58b883cbc7e57666ee6374666b0705e46b7cd74a19dc15b02f0ff1133ba3337c.jpg b/parse/dev/xpjsOQtKqx/images/58b883cbc7e57666ee6374666b0705e46b7cd74a19dc15b02f0ff1133ba3337c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3fcb13bdc8c44527d0ee3ee5f4bce13eb39f30ad --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/58b883cbc7e57666ee6374666b0705e46b7cd74a19dc15b02f0ff1133ba3337c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d148cae6a353c27b4832c37009ed42a3cfa1b8e75b700b43384572ae2c7d05f5 +size 17775 diff --git a/parse/dev/xpjsOQtKqx/images/598effc7797b2348e189f4351828745743c50f65c79a152e0ccd3d9368f44c41.jpg b/parse/dev/xpjsOQtKqx/images/598effc7797b2348e189f4351828745743c50f65c79a152e0ccd3d9368f44c41.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5edb374946d4cba748ffed173c2acf55ffb087ff --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/598effc7797b2348e189f4351828745743c50f65c79a152e0ccd3d9368f44c41.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e92093fe77f4f101a27d771153d1560a2d6bb239e1dba2cb074a4c4017df8dbd +size 5547 diff --git a/parse/dev/xpjsOQtKqx/images/61c79d2f6435c6d84b5a2e9688283d26300b844e9195b07b45b4ef50cfc46969.jpg b/parse/dev/xpjsOQtKqx/images/61c79d2f6435c6d84b5a2e9688283d26300b844e9195b07b45b4ef50cfc46969.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2e662d8a72619f1b440b07a99a6c83891beeddd0 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/61c79d2f6435c6d84b5a2e9688283d26300b844e9195b07b45b4ef50cfc46969.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5712bb661cf9773ee8b7be5dde266e16d4700ea40b4598834d2db9547c54646d +size 126097 diff --git a/parse/dev/xpjsOQtKqx/images/6d06f4242c25e6df1b4ef63750851817332366bcec09c2991c9d7b95fc0e6e33.jpg b/parse/dev/xpjsOQtKqx/images/6d06f4242c25e6df1b4ef63750851817332366bcec09c2991c9d7b95fc0e6e33.jpg new file mode 100644 index 0000000000000000000000000000000000000000..74237886f138420fac5cf1170b76620cec5324b1 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/6d06f4242c25e6df1b4ef63750851817332366bcec09c2991c9d7b95fc0e6e33.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f70788357af23dcc5121ad17b5a3a0b8c29404f20b52515699e81c5a52e8956b +size 24815 diff --git a/parse/dev/xpjsOQtKqx/images/80fe969f8081eb6d00f3a40176303d51d89c5b83ce8ff36433e5f662d6a1fc04.jpg b/parse/dev/xpjsOQtKqx/images/80fe969f8081eb6d00f3a40176303d51d89c5b83ce8ff36433e5f662d6a1fc04.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f2658c757ee7dd0970b26cfa851b3f5e45fcadf7 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/80fe969f8081eb6d00f3a40176303d51d89c5b83ce8ff36433e5f662d6a1fc04.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3733500a394f462e41d98a57a4428137442a3946e156d6f4c4b07396c8738099 +size 69214 diff --git a/parse/dev/xpjsOQtKqx/images/89c7321b8ba95c8f9960fb0e006e001a0d2033256d52011a40ba9dffa3c2853c.jpg b/parse/dev/xpjsOQtKqx/images/89c7321b8ba95c8f9960fb0e006e001a0d2033256d52011a40ba9dffa3c2853c.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f3aafe37d10f38bc505e78c54ed918401a4e8eeb --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/89c7321b8ba95c8f9960fb0e006e001a0d2033256d52011a40ba9dffa3c2853c.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:156dcb357b6a41b352ddcff6e00b5680e234c25ff00131aae25d61fb550a737d +size 6564 diff --git a/parse/dev/xpjsOQtKqx/images/90b9dadbfe06080b8fd3f05ede3cb312b2f9e59d5c9abd49a95bdeecf2e6f8b6.jpg b/parse/dev/xpjsOQtKqx/images/90b9dadbfe06080b8fd3f05ede3cb312b2f9e59d5c9abd49a95bdeecf2e6f8b6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..568c773d7aaec712ff1459f0996e3e11a49ff4f1 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/90b9dadbfe06080b8fd3f05ede3cb312b2f9e59d5c9abd49a95bdeecf2e6f8b6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:71cf3011cae7cd5f3cc0c4be5cca9e3ef7add96a7089b03146c4c7608853baa4 +size 153255 diff --git a/parse/dev/xpjsOQtKqx/images/948a0359240336ea5a5d303f31b8499558e13513e5bb7149202b84a33925f5fe.jpg b/parse/dev/xpjsOQtKqx/images/948a0359240336ea5a5d303f31b8499558e13513e5bb7149202b84a33925f5fe.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9036641cdf2493a07bf45b17e1f2c7435f2040b3 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/948a0359240336ea5a5d303f31b8499558e13513e5bb7149202b84a33925f5fe.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:360094d2dd2373336983be0b49977198abb9221b7738965a27871c9394841e73 +size 47587 diff --git a/parse/dev/xpjsOQtKqx/images/9ed884a0f0d5c2b21661aace7c5dbeb206734bf3a2ef0348ff2e1be16e0efed3.jpg b/parse/dev/xpjsOQtKqx/images/9ed884a0f0d5c2b21661aace7c5dbeb206734bf3a2ef0348ff2e1be16e0efed3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..acebc66702198f75d8560100d4b3a465fdeb2b73 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/9ed884a0f0d5c2b21661aace7c5dbeb206734bf3a2ef0348ff2e1be16e0efed3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1dc4914bd270b925ad09543786571998f97e75c247556b6fd15e93fc77d3bcd +size 11906 diff --git a/parse/dev/xpjsOQtKqx/images/9ef1085726097eb8e837598a08bd2cd6bf27c8d4e1e5920d69dedbd2c2428910.jpg b/parse/dev/xpjsOQtKqx/images/9ef1085726097eb8e837598a08bd2cd6bf27c8d4e1e5920d69dedbd2c2428910.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f5ef8a0dc126b68d36760c017b3068b2b6536118 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/9ef1085726097eb8e837598a08bd2cd6bf27c8d4e1e5920d69dedbd2c2428910.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b88fba3a88736f14856c12ac36c9c7dc1a2033afecd06ac6152723d81be196b5 +size 33089 diff --git a/parse/dev/xpjsOQtKqx/images/a030f4824bbd63fbe41ce9dc960e5df81997946d91d87eb70b80a6451cfed878.jpg b/parse/dev/xpjsOQtKqx/images/a030f4824bbd63fbe41ce9dc960e5df81997946d91d87eb70b80a6451cfed878.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f009e92d826d622f290027627b2c8d516202e675 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/a030f4824bbd63fbe41ce9dc960e5df81997946d91d87eb70b80a6451cfed878.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d7e59e55ef3379d12dc8e8b0e9934afa5e3066c86004cf947b2a8b212b7c59f +size 4484 diff --git a/parse/dev/xpjsOQtKqx/images/b27833e368ad1296c62d42c051b5102623b5ba609fb9bd1456608facb3058cb4.jpg b/parse/dev/xpjsOQtKqx/images/b27833e368ad1296c62d42c051b5102623b5ba609fb9bd1456608facb3058cb4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..90082aec94c722890ca85d4a60480e6cc76b562f --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/b27833e368ad1296c62d42c051b5102623b5ba609fb9bd1456608facb3058cb4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:445956f0ee6a644db2c838dfc3caeee3bd82f10bead1565990fbb9f51efbb764 +size 4868 diff --git a/parse/dev/xpjsOQtKqx/images/b73329ddacd75a14dea10ad0277990cca5ab2604f298c4ccaa39315686003529.jpg b/parse/dev/xpjsOQtKqx/images/b73329ddacd75a14dea10ad0277990cca5ab2604f298c4ccaa39315686003529.jpg new file mode 100644 index 0000000000000000000000000000000000000000..51109551fc4c0c75f96cd3655f7beba5d95f661f --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/b73329ddacd75a14dea10ad0277990cca5ab2604f298c4ccaa39315686003529.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:604b7d57ebb98782ebf7539a7f217d9a7a6b3ff31205d1671d838bc95895786c +size 34805 diff --git a/parse/dev/xpjsOQtKqx/images/b872c7f94f3f9d6a2241c43ea5e1572fb1926b17e518cf7e68772de3d1942c4a.jpg b/parse/dev/xpjsOQtKqx/images/b872c7f94f3f9d6a2241c43ea5e1572fb1926b17e518cf7e68772de3d1942c4a.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0e36b663f1b6f7fd27e2ac3e380dd0d17f6ee9a2 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/b872c7f94f3f9d6a2241c43ea5e1572fb1926b17e518cf7e68772de3d1942c4a.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b8d6597934d886a597a894102ddfcf5b0fe89ad45ff5c52158c6c60ba4602b5 +size 66671 diff --git a/parse/dev/xpjsOQtKqx/images/bf1621778e4da87c131efb21ad3eb3569e9a4cf0c23a1792c3db948d13a47206.jpg b/parse/dev/xpjsOQtKqx/images/bf1621778e4da87c131efb21ad3eb3569e9a4cf0c23a1792c3db948d13a47206.jpg new file mode 100644 index 0000000000000000000000000000000000000000..35f6db24408be2c04760f84d0128e3453b7db7f2 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/bf1621778e4da87c131efb21ad3eb3569e9a4cf0c23a1792c3db948d13a47206.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82ce77fc6e34984db2c6bdd6d982a89b466bd7d0eb075fccf5ca4579e8b12ecb +size 4341 diff --git a/parse/dev/xpjsOQtKqx/images/daf17c409febf7e7b38e344d1472902136f30e1a34e702cfa0a71299ecf0a412.jpg b/parse/dev/xpjsOQtKqx/images/daf17c409febf7e7b38e344d1472902136f30e1a34e702cfa0a71299ecf0a412.jpg new file mode 100644 index 0000000000000000000000000000000000000000..80c199f5ec6d33bee93d0c68bc1c2e3bd317cb74 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/daf17c409febf7e7b38e344d1472902136f30e1a34e702cfa0a71299ecf0a412.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c1fa70d35f9d2f519c602a52956667fe8854f673cfdaeb6ac9264cdc6e0dfea9 +size 43841 diff --git a/parse/dev/xpjsOQtKqx/images/e1cd25bda20810b52b0730a43a3814630602c4d8908a9f64812badea0db8d2be.jpg b/parse/dev/xpjsOQtKqx/images/e1cd25bda20810b52b0730a43a3814630602c4d8908a9f64812badea0db8d2be.jpg new file mode 100644 index 0000000000000000000000000000000000000000..27042313d24dd8d535caad498b6702c28cdb21b4 --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/e1cd25bda20810b52b0730a43a3814630602c4d8908a9f64812badea0db8d2be.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d20c7ee0fb89723b74875513d2142a89a617dc88d51788f95cf759eed077bef +size 52224 diff --git a/parse/dev/xpjsOQtKqx/images/f323dd885d5ac7b8647e39e17496dd8cf7ea21783810e7613bc5977eb0760c7f.jpg b/parse/dev/xpjsOQtKqx/images/f323dd885d5ac7b8647e39e17496dd8cf7ea21783810e7613bc5977eb0760c7f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ee79acfc68304949c6e7e7161449da6d3cfb837a --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/f323dd885d5ac7b8647e39e17496dd8cf7ea21783810e7613bc5977eb0760c7f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79c264cdfe5d80a2b80e6b71ec733d1651d308844d9e2be52fd1ea7f422bf299 +size 34590 diff --git a/parse/dev/xpjsOQtKqx/images/f54a5f70e767e96b4c186df75219531fbbff18e204d7b14aac0c74c47d74af92.jpg b/parse/dev/xpjsOQtKqx/images/f54a5f70e767e96b4c186df75219531fbbff18e204d7b14aac0c74c47d74af92.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e8a2ac8d7996e6de206fafbb133684eb07c3680e --- /dev/null +++ b/parse/dev/xpjsOQtKqx/images/f54a5f70e767e96b4c186df75219531fbbff18e204d7b14aac0c74c47d74af92.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:64d1bb4d17037209738641ddf4247800ef7bb7a84cb4eb8d14d003b1dc0e8dfc +size 7191 diff --git a/parse/dev/zf_Ll3HZWgy/images/0904bbb5b354f16b6aaf4fd4aa38a7f6927d73ac8e56bc73fad0091e33d1d9dc.jpg b/parse/dev/zf_Ll3HZWgy/images/0904bbb5b354f16b6aaf4fd4aa38a7f6927d73ac8e56bc73fad0091e33d1d9dc.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bb414be395d91235980e906afba8e920fe140e84 --- /dev/null +++ b/parse/dev/zf_Ll3HZWgy/images/0904bbb5b354f16b6aaf4fd4aa38a7f6927d73ac8e56bc73fad0091e33d1d9dc.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc006e5450b5cfa588aee779e0d867c4656cb8296087199ad9b7a58a91ff35cb +size 17915 diff --git a/parse/dev/zf_Ll3HZWgy/images/29c8e0bbe3f3f4dd7b236933079a03fbbdd62bff838cf7b56bc80ba65f3c3a52.jpg b/parse/dev/zf_Ll3HZWgy/images/29c8e0bbe3f3f4dd7b236933079a03fbbdd62bff838cf7b56bc80ba65f3c3a52.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7ca2b1997e16cd5a22c0b0b6f41525176bae418c --- /dev/null +++ b/parse/dev/zf_Ll3HZWgy/images/29c8e0bbe3f3f4dd7b236933079a03fbbdd62bff838cf7b56bc80ba65f3c3a52.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e54da08d0a3b08ed83889aa321ca8c71a1a81e715bc546ef91741abd94875be8 +size 60856 diff --git a/parse/dev/zf_Ll3HZWgy/images/2bcec2615f05a7c4fd3a2f722ee6ec371946b04b3963b6ddcac66e80cea8a859.jpg b/parse/dev/zf_Ll3HZWgy/images/2bcec2615f05a7c4fd3a2f722ee6ec371946b04b3963b6ddcac66e80cea8a859.jpg new file mode 100644 index 0000000000000000000000000000000000000000..79d5963af13ed2c3ccde566623bc1ecacab3824e --- /dev/null +++ b/parse/dev/zf_Ll3HZWgy/images/2bcec2615f05a7c4fd3a2f722ee6ec371946b04b3963b6ddcac66e80cea8a859.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82eab8238a1c028ef31af16224c06b99f7e20ff54b25b1c0834967308fcc09aa +size 59634 diff --git a/parse/dev/zf_Ll3HZWgy/images/410aacf2cc6356ee03650e2497e263a1cf81601a78c706c489dc26eb2b5f4189.jpg b/parse/dev/zf_Ll3HZWgy/images/410aacf2cc6356ee03650e2497e263a1cf81601a78c706c489dc26eb2b5f4189.jpg new file mode 100644 index 0000000000000000000000000000000000000000..04d5f2bc1436966e451c4758fc41788e7e1ba710 --- /dev/null +++ b/parse/dev/zf_Ll3HZWgy/images/410aacf2cc6356ee03650e2497e263a1cf81601a78c706c489dc26eb2b5f4189.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cef41c75924b344eb5db49e3a3c21132b128c876592aa8118052efe0d16ec789 +size 70310 diff --git a/parse/dev/zf_Ll3HZWgy/images/69304161d163f20f2b3fdbfaf6cc463c20c73b3378e0d031301496f64166b383.jpg b/parse/dev/zf_Ll3HZWgy/images/69304161d163f20f2b3fdbfaf6cc463c20c73b3378e0d031301496f64166b383.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a3da4a1ff54c24d08bcf863fdc36584bd2265c59 --- /dev/null +++ b/parse/dev/zf_Ll3HZWgy/images/69304161d163f20f2b3fdbfaf6cc463c20c73b3378e0d031301496f64166b383.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2fc21f2ca66bbef1d1e205c83d7c4f587900e7d2fd6826d0521ed611ee6d7692 +size 87953 diff --git a/parse/dev/zf_Ll3HZWgy/images/6d379ef9eaa75cbe8eaf312c397619b388462aabed45b59206119d112ee1dc09.jpg b/parse/dev/zf_Ll3HZWgy/images/6d379ef9eaa75cbe8eaf312c397619b388462aabed45b59206119d112ee1dc09.jpg new file mode 100644 index 0000000000000000000000000000000000000000..437289d5d64e2dd525b0d9e856834af8fb465cd8 --- /dev/null +++ b/parse/dev/zf_Ll3HZWgy/images/6d379ef9eaa75cbe8eaf312c397619b388462aabed45b59206119d112ee1dc09.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:09b844bc768249e6e5703ed085f84dc1b7b47fde570193142da459bdcdffa9a5 +size 63141 diff --git a/parse/dev/zf_Ll3HZWgy/images/8ef436e304f5471ba237448c4581a59150f732398c0942288b06f8697f9d2ef0.jpg b/parse/dev/zf_Ll3HZWgy/images/8ef436e304f5471ba237448c4581a59150f732398c0942288b06f8697f9d2ef0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..987413e760cfeba1866aedcf732235f3be0e477a --- /dev/null +++ b/parse/dev/zf_Ll3HZWgy/images/8ef436e304f5471ba237448c4581a59150f732398c0942288b06f8697f9d2ef0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:93f2678385a47c2a3f1174c640631ec9b78e9dd1252a1ff1ab08e3331ca04aee +size 15230 diff --git a/parse/dev/zf_Ll3HZWgy/images/b6f93693c54192d261696a23111cb568bcb278b5db5e4003e4326f570f71d813.jpg b/parse/dev/zf_Ll3HZWgy/images/b6f93693c54192d261696a23111cb568bcb278b5db5e4003e4326f570f71d813.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1814f2d9ca6f3669874a74e27c993d8a46acfc85 --- /dev/null +++ b/parse/dev/zf_Ll3HZWgy/images/b6f93693c54192d261696a23111cb568bcb278b5db5e4003e4326f570f71d813.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ac2b287a022b8a53e70c552ff57f97a8271c927d9f5df4b8f2130c6b693353a +size 26843 diff --git a/parse/dev/zf_Ll3HZWgy/images/de6dcbe3c3a12b265c69b1d922bf4d616ba7d5ff3613014d377b7dcf753221d8.jpg b/parse/dev/zf_Ll3HZWgy/images/de6dcbe3c3a12b265c69b1d922bf4d616ba7d5ff3613014d377b7dcf753221d8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fbb57437e3a32f25eae3ad5550e927db73d555c4 --- /dev/null +++ b/parse/dev/zf_Ll3HZWgy/images/de6dcbe3c3a12b265c69b1d922bf4d616ba7d5ff3613014d377b7dcf753221d8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:303ccb4376f339c310af6e063c4d190dba6dcf9fde9afa318346da70fd0464cc +size 24511 diff --git a/parse/dev/zq1iJkNk3uN/images/05899bf9ce8b7ecbf860f3ef3c370d4df60211f7f85b8b68a64cbf2208b142fb.jpg b/parse/dev/zq1iJkNk3uN/images/05899bf9ce8b7ecbf860f3ef3c370d4df60211f7f85b8b68a64cbf2208b142fb.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a7975cfc4c869904eabd9f1ee931f30bae0fd418 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/05899bf9ce8b7ecbf860f3ef3c370d4df60211f7f85b8b68a64cbf2208b142fb.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:baad7b08971820b0d006d3a54ce09be61e270a75a5fcaee266c9e10ad330a98e +size 31461 diff --git a/parse/dev/zq1iJkNk3uN/images/099e973feb436fd6ca674d188bd97dc50d7da4c983895863c106332cf6dab9c0.jpg b/parse/dev/zq1iJkNk3uN/images/099e973feb436fd6ca674d188bd97dc50d7da4c983895863c106332cf6dab9c0.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5b3b200a4a224b3f0e82c26489db9bc8d7452c85 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/099e973feb436fd6ca674d188bd97dc50d7da4c983895863c106332cf6dab9c0.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:98f64f614b8ecde791e60055a9a09fa599b458174731f120437a5f2a294a2ea8 +size 8313 diff --git a/parse/dev/zq1iJkNk3uN/images/0a212f39308583b1cf69a8c17dd30e076d9672d7286f2297521508281ad81028.jpg b/parse/dev/zq1iJkNk3uN/images/0a212f39308583b1cf69a8c17dd30e076d9672d7286f2297521508281ad81028.jpg new file mode 100644 index 0000000000000000000000000000000000000000..018ae8b92a85f937e37d3f3067980b2e69b291c0 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/0a212f39308583b1cf69a8c17dd30e076d9672d7286f2297521508281ad81028.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c324d76be1210dbf0fae8ebf55dbe4441413c90abaf00da96ec1424cb9a58a3 +size 29842 diff --git a/parse/dev/zq1iJkNk3uN/images/0ce1d9446a4c27c1beac26f91cc23c8f585923fe59a3017bc43c6e9fe2aab4fe.jpg b/parse/dev/zq1iJkNk3uN/images/0ce1d9446a4c27c1beac26f91cc23c8f585923fe59a3017bc43c6e9fe2aab4fe.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d723e8bd4c5c5030f4419a527524c5a875a2a451 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/0ce1d9446a4c27c1beac26f91cc23c8f585923fe59a3017bc43c6e9fe2aab4fe.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ab99114324f287cb5a6b230f0615c7d6d9bbfab78a8e3590b6033a99e8c0b817 +size 81715 diff --git a/parse/dev/zq1iJkNk3uN/images/24124e253e38c6704603134c4ef33a12601e11c766185c26503b711bcee1bced.jpg b/parse/dev/zq1iJkNk3uN/images/24124e253e38c6704603134c4ef33a12601e11c766185c26503b711bcee1bced.jpg new file mode 100644 index 0000000000000000000000000000000000000000..100da2d34c5927efa567e6fe10105a596f559662 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/24124e253e38c6704603134c4ef33a12601e11c766185c26503b711bcee1bced.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a40a0cc3ea798912403b19882609cb81e26f5801401603e3dbc71044153dacc +size 9997 diff --git a/parse/dev/zq1iJkNk3uN/images/2419d1a278807a209c104291cc45f9107ffa7faa3350cc9fbaf8bbe17baf2a26.jpg b/parse/dev/zq1iJkNk3uN/images/2419d1a278807a209c104291cc45f9107ffa7faa3350cc9fbaf8bbe17baf2a26.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ef5adfd8172f64bef89d03d6a055b8951dcf004c --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/2419d1a278807a209c104291cc45f9107ffa7faa3350cc9fbaf8bbe17baf2a26.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b2ab7e7caf103a2cd6a6c094f11295efa964ba45ae6364b60e49e30904d5df8 +size 64025 diff --git a/parse/dev/zq1iJkNk3uN/images/2ce5145e595e67cb713313239ce738b67507991bf096300d17a37e96f75693e5.jpg b/parse/dev/zq1iJkNk3uN/images/2ce5145e595e67cb713313239ce738b67507991bf096300d17a37e96f75693e5.jpg new file mode 100644 index 0000000000000000000000000000000000000000..71362520382bf2e76e75641ba6deeb09813923e0 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/2ce5145e595e67cb713313239ce738b67507991bf096300d17a37e96f75693e5.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b4ae0bfc1fefff775d7855bbb0807a5ba0617f8d88a41b0d69cef12fef8640d8 +size 27653 diff --git a/parse/dev/zq1iJkNk3uN/images/3d2b4004d15eb6f48d3f5e80d9d1488c241c13b497c0d9fef0536ead1748173d.jpg b/parse/dev/zq1iJkNk3uN/images/3d2b4004d15eb6f48d3f5e80d9d1488c241c13b497c0d9fef0536ead1748173d.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f611e8a0e408a505ff950a4f747ef0c172c6d1ff --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/3d2b4004d15eb6f48d3f5e80d9d1488c241c13b497c0d9fef0536ead1748173d.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0c199141595a1fe920cd6b7393382b2124e61e6032128c089968cbdcd48f27b +size 53469 diff --git a/parse/dev/zq1iJkNk3uN/images/5122ea60d58db20432ff26bdbb4bbe9a554035589c526eeff41d31ea91e0641f.jpg b/parse/dev/zq1iJkNk3uN/images/5122ea60d58db20432ff26bdbb4bbe9a554035589c526eeff41d31ea91e0641f.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d5c95a38a07625423a64d470e1dbbfa2cb822681 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/5122ea60d58db20432ff26bdbb4bbe9a554035589c526eeff41d31ea91e0641f.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be8427dc7365f6ee25b931d4fe0d589dbd29deffd1ae1476753e6acb9725167c +size 94818 diff --git a/parse/dev/zq1iJkNk3uN/images/53ff2d3a432b12e62a157fe31dc337181959190e86b844012553188c822353c6.jpg b/parse/dev/zq1iJkNk3uN/images/53ff2d3a432b12e62a157fe31dc337181959190e86b844012553188c822353c6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f5c38b5a574444662b06863434ba7e6f1f604e15 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/53ff2d3a432b12e62a157fe31dc337181959190e86b844012553188c822353c6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91d37d1a6f8feb560d0b44d57cf104140e0f1c185914f4c9d786edfc2345410d +size 33979 diff --git a/parse/dev/zq1iJkNk3uN/images/7597d0083fb5b3c88dc12cc3288664828aa98bb8384c0b068b46b4ce50440dfa.jpg b/parse/dev/zq1iJkNk3uN/images/7597d0083fb5b3c88dc12cc3288664828aa98bb8384c0b068b46b4ce50440dfa.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f2b3f1f0b94eb5ce0001fe15ad7d0ce1961a2f2c --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/7597d0083fb5b3c88dc12cc3288664828aa98bb8384c0b068b46b4ce50440dfa.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:83f090afe2b2d912ea9f0ef3c051a20e7a5062b6af94e21abc747ff8a6600713 +size 181091 diff --git a/parse/dev/zq1iJkNk3uN/images/796081aa69bcfc094c57ebf682e4ad8fe85bc76f651c50febc1462a8b3dfd2f4.jpg b/parse/dev/zq1iJkNk3uN/images/796081aa69bcfc094c57ebf682e4ad8fe85bc76f651c50febc1462a8b3dfd2f4.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2bb8137387f514a108a790f13b08cbfc5bf8060b --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/796081aa69bcfc094c57ebf682e4ad8fe85bc76f651c50febc1462a8b3dfd2f4.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8f719a183bde1bc770477a4b4cfe7e1beb2854ddf33f17325da6fa5cbb98fbe +size 16623 diff --git a/parse/dev/zq1iJkNk3uN/images/8662742de9a8a9ec0d5242984fd33aa7e005d4a664aa011675880d70caa2e1cd.jpg b/parse/dev/zq1iJkNk3uN/images/8662742de9a8a9ec0d5242984fd33aa7e005d4a664aa011675880d70caa2e1cd.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d57eefd7212ef7eb2b34a41f6d1d8628e6695adf --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/8662742de9a8a9ec0d5242984fd33aa7e005d4a664aa011675880d70caa2e1cd.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7947226b913eb88fb29be3335efb443d8fd041015ee128dca0290d1d5ac3d570 +size 70036 diff --git a/parse/dev/zq1iJkNk3uN/images/9cf73d38023a7786008f4c6ac9a8a422d164c277356e8beea2be9652a80ddb72.jpg b/parse/dev/zq1iJkNk3uN/images/9cf73d38023a7786008f4c6ac9a8a422d164c277356e8beea2be9652a80ddb72.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a945683b2546f26a7eb14e508b046ac5f27413af --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/9cf73d38023a7786008f4c6ac9a8a422d164c277356e8beea2be9652a80ddb72.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:53ac0ee7ce879f5278f2035d37a21dbc552f583bfa1ca147148bcdf6e69c6195 +size 64162 diff --git a/parse/dev/zq1iJkNk3uN/images/a199ca3e8ed6e778fb5e867454a918ea7578fff4a0d309b8192582da8d2d66b6.jpg b/parse/dev/zq1iJkNk3uN/images/a199ca3e8ed6e778fb5e867454a918ea7578fff4a0d309b8192582da8d2d66b6.jpg new file mode 100644 index 0000000000000000000000000000000000000000..babab29861fa212873c8b21b14b3e1b7b2ee3afd --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/a199ca3e8ed6e778fb5e867454a918ea7578fff4a0d309b8192582da8d2d66b6.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3f7f254135c1933bd4466d71a42798b5211e66ca9de0c5a24e5d8a093a8a3573 +size 24086 diff --git a/parse/dev/zq1iJkNk3uN/images/a1c88e67ffd3076e07cf6bd1348a5c72a05f0560bd752c84a44e89c351ccc327.jpg b/parse/dev/zq1iJkNk3uN/images/a1c88e67ffd3076e07cf6bd1348a5c72a05f0560bd752c84a44e89c351ccc327.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6ff6b3d74d931e8c01f8ed8a70c6faad0aeb6466 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/a1c88e67ffd3076e07cf6bd1348a5c72a05f0560bd752c84a44e89c351ccc327.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2bf986489a11434c556b32c4252cf9b72679c19363973b8ff92f775be6559415 +size 37619 diff --git a/parse/dev/zq1iJkNk3uN/images/bc5e47ac7e18c08e37f2312818f8ebdaedd2ea57aaef0cfcf10e0f1d8daa11ee.jpg b/parse/dev/zq1iJkNk3uN/images/bc5e47ac7e18c08e37f2312818f8ebdaedd2ea57aaef0cfcf10e0f1d8daa11ee.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c36030f3acece91a9cc067bed1cbaa6cd3d9296a --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/bc5e47ac7e18c08e37f2312818f8ebdaedd2ea57aaef0cfcf10e0f1d8daa11ee.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20f46acf5aa4ce210f1bef83540fb206535d236c868b0fafd0324e7acabca53c +size 29532 diff --git a/parse/dev/zq1iJkNk3uN/images/e685a6a711b7ad54e01e952638e18b72f3ce6a258ddcd267d907ad4e151b4ac2.jpg b/parse/dev/zq1iJkNk3uN/images/e685a6a711b7ad54e01e952638e18b72f3ce6a258ddcd267d907ad4e151b4ac2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7193202137900099d56f6b1ff3787ba077abb706 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/e685a6a711b7ad54e01e952638e18b72f3ce6a258ddcd267d907ad4e151b4ac2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:99727966f2f10d1306de003b00641335d5b5d978152fd4a57b9fd6696ba3e983 +size 81142 diff --git a/parse/dev/zq1iJkNk3uN/images/efca79d801146a93e7ca9561d754ff7a39aecdc8ab47135af616f0ab59f5eb72.jpg b/parse/dev/zq1iJkNk3uN/images/efca79d801146a93e7ca9561d754ff7a39aecdc8ab47135af616f0ab59f5eb72.jpg new file mode 100644 index 0000000000000000000000000000000000000000..008086369a1ad106f5aa5f7f1c7541665c2878c5 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/efca79d801146a93e7ca9561d754ff7a39aecdc8ab47135af616f0ab59f5eb72.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2310d0f491992dbb888edaa01b277db2baab77bf9a6d9061b9834f95aa21319f +size 30204 diff --git a/parse/dev/zq1iJkNk3uN/images/f61344fb96af452036cbd5e7de88d22ad6f3e802c2d5da0e26f0d73a004c59c8.jpg b/parse/dev/zq1iJkNk3uN/images/f61344fb96af452036cbd5e7de88d22ad6f3e802c2d5da0e26f0d73a004c59c8.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9dcdb73e5e002f15f241f5fa6ba54f35284b3f0a --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/f61344fb96af452036cbd5e7de88d22ad6f3e802c2d5da0e26f0d73a004c59c8.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2354c8ebb91152350bc7654ac6c5176eb83a8d9c5e8fbd79879c6af65188ca44 +size 90945 diff --git a/parse/dev/zq1iJkNk3uN/images/ff256bd190f5fa5cb1563d8163a4724b11aa7664d82de5778c462e23266912f7.jpg b/parse/dev/zq1iJkNk3uN/images/ff256bd190f5fa5cb1563d8163a4724b11aa7664d82de5778c462e23266912f7.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9f9f4229a85077ecd7a46c79ff5aae59ff33a739 --- /dev/null +++ b/parse/dev/zq1iJkNk3uN/images/ff256bd190f5fa5cb1563d8163a4724b11aa7664d82de5778c462e23266912f7.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b6a7aa2e3697a5ecd67ecd7ed78b586393e8b2983637c0a38bcc89c8e0e1264 +size 40506