shdragron commited on
Commit
35644fc
路
verified 路
1 Parent(s): 3dc1fbb

Add segmentation checkpoints (GaussianLSS, CVT)

Browse files
README.md CHANGED
@@ -8,6 +8,7 @@ tags:
8
  - multi-camera
9
  - robustness
10
  - 3d-object-detection
 
11
  ---
12
 
13
  # RoboGeo checkpoints
@@ -16,8 +17,8 @@ Trained checkpoints for **RoboGeo: Benchmarking Camera-Geometry Robustness in Mu
16
  (WACV 2027) 路 [Project page](https://robogeo2027.github.io/) 路 [Dataset](https://huggingface.co/datasets/shdragron/RoboGeo)
17
 
18
  Each model is trained on one platform of RoboGeo. The Sedan models are evaluated under viewpoint
19
- perturbation and cross-platform transfer; the SUV and Bus models are trained on the target platforms
20
- and serve as the reference for the Cross-Platform Transfer Score (CTS).
21
 
22
  ## 3D object detection
23
 
@@ -30,13 +31,25 @@ and serve as the reference for the Cross-Platform Transfer Score (CTS).
30
  | CAPE | SUV | `detection/cape/cape_suv.pth` | 246 MB | `6150f66701eb1036` |
31
  | CAPE | Bus | `detection/cape/cape_bus.pth` | 246 MB | `469c5c49ac5622bb` |
32
 
33
- Full checksums are in `SHA256SUMS`. Each checkpoint holds `state_dict` and `meta.CLASSES` and loads with
34
- `torch.load(path, weights_only=True)`. Training and evaluation code:
35
- [shdragron/RoboGeo2027](https://github.com/shdragron/RoboGeo2027) (coming soon).
 
 
 
 
 
 
 
 
 
 
 
36
 
37
  ## License
38
 
39
  Released under [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/), the license of the
40
  RoboGeo dataset the models are trained on. The model architectures come from BEVFormer and CAPE
41
- (Apache-2.0). When using these checkpoints, please cite RoboGeo, SimBEV and CARLA as listed in the
 
42
  [dataset card](https://huggingface.co/datasets/shdragron/RoboGeo).
 
8
  - multi-camera
9
  - robustness
10
  - 3d-object-detection
11
+ - bev-segmentation
12
  ---
13
 
14
  # RoboGeo checkpoints
 
17
  (WACV 2027) 路 [Project page](https://robogeo2027.github.io/) 路 [Dataset](https://huggingface.co/datasets/shdragron/RoboGeo)
18
 
19
  Each model is trained on one platform of RoboGeo. The Sedan models are evaluated under viewpoint
20
+ perturbation and cross-platform transfer; the SUV and Bus models are the models trained on the target
21
+ platforms of cross-platform transfer.
22
 
23
  ## 3D object detection
24
 
 
31
  | CAPE | SUV | `detection/cape/cape_suv.pth` | 246 MB | `6150f66701eb1036` |
32
  | CAPE | Bus | `detection/cape/cape_bus.pth` | 246 MB | `469c5c49ac5622bb` |
33
 
34
+ ## BEV segmentation
35
+
36
+ | model | platform | file | size | sha256 (first 16) |
37
+ |---|---|---|---|---|
38
+ | GaussianLSS | Sedan | `segmentation/gaussianlss/gaussianlss_sedan.pth` | 33.8 MB | `4ca7cf0901b99c34` |
39
+ | GaussianLSS | SUV | `segmentation/gaussianlss/gaussianlss_suv.pth` | 33.8 MB | `087171e00240fcaf` |
40
+ | GaussianLSS | Bus | `segmentation/gaussianlss/gaussianlss_bus.pth` | 33.8 MB | `149d3bd36c34cd17` |
41
+ | CVT | Sedan | `segmentation/cvt/cvt_sedan.pth` | 4.5 MB | `ded6a2c934a89edb` |
42
+ | CVT | SUV | `segmentation/cvt/cvt_suv.pth` | 4.5 MB | `946356200d2cdff1` |
43
+ | CVT | Bus | `segmentation/cvt/cvt_bus.pth` | 4.5 MB | `257bdae34378407e` |
44
+
45
+ Full checksums are in `SHA256SUMS`. Detection checkpoints hold `state_dict` and `meta.CLASSES`;
46
+ segmentation checkpoints hold `state_dict`. All load with `torch.load(path, weights_only=True)`.
47
+ Training and evaluation code: [shdragron/RoboGeo2027](https://github.com/shdragron/RoboGeo2027) (coming soon).
48
 
49
  ## License
50
 
51
  Released under [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/), the license of the
52
  RoboGeo dataset the models are trained on. The model architectures come from BEVFormer and CAPE
53
+ (Apache-2.0), and from GaussianLSS and Cross View Transformers (MIT). When using these checkpoints,
54
+ please cite RoboGeo, SimBEV and CARLA as listed in the
55
  [dataset card](https://huggingface.co/datasets/shdragron/RoboGeo).
SHA256SUMS CHANGED
@@ -4,3 +4,9 @@ e166f0aec95b1cff602e7cd4e1493082b906a215d261377a6f4a721fdc636f43 detection/bevf
4
  b7d6fa8f60de29c0165312d168ebfe5457740a8a6f301f5a938daa4bb37874b3 detection/cape/cape_sedan.pth
5
  6150f66701eb10368edfe32d94e17c5ed1b56ce033ca241c25d9a387c727c891 detection/cape/cape_suv.pth
6
  469c5c49ac5622bb0b6dcab6a35af99239ac0146b0c9117e72abefda46f75c70 detection/cape/cape_bus.pth
 
 
 
 
 
 
 
4
  b7d6fa8f60de29c0165312d168ebfe5457740a8a6f301f5a938daa4bb37874b3 detection/cape/cape_sedan.pth
5
  6150f66701eb10368edfe32d94e17c5ed1b56ce033ca241c25d9a387c727c891 detection/cape/cape_suv.pth
6
  469c5c49ac5622bb0b6dcab6a35af99239ac0146b0c9117e72abefda46f75c70 detection/cape/cape_bus.pth
7
+ 4ca7cf0901b99c34a49fc5ef9cfeaf3fcd2c68d315b2d57a3007f2fb62d36e22 segmentation/gaussianlss/gaussianlss_sedan.pth
8
+ 087171e00240fcaff15a64017cedaf1a977669566e09360b1f804ad91da26081 segmentation/gaussianlss/gaussianlss_suv.pth
9
+ 149d3bd36c34cd17578798b68311060421fd1f0fcc83d4bdb4832a6dfa802943 segmentation/gaussianlss/gaussianlss_bus.pth
10
+ ded6a2c934a89edbff845166c5edf92872d35bf99f61139ddaa591b514d40d36 segmentation/cvt/cvt_sedan.pth
11
+ 946356200d2cdff1621815ec8b6934360afe46865131ee7d6cde074c70c5796d segmentation/cvt/cvt_suv.pth
12
+ 257bdae34378407e54b31ec0b9692cb4c0a554d36246340a09b8f3295f9ddfe6 segmentation/cvt/cvt_bus.pth
segmentation/cvt/cvt_bus.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:257bdae34378407e54b31ec0b9692cb4c0a554d36246340a09b8f3295f9ddfe6
3
+ size 4538073
segmentation/cvt/cvt_sedan.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ded6a2c934a89edbff845166c5edf92872d35bf99f61139ddaa591b514d40d36
3
+ size 4539467
segmentation/cvt/cvt_suv.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:946356200d2cdff1621815ec8b6934360afe46865131ee7d6cde074c70c5796d
3
+ size 4538073
segmentation/gaussianlss/gaussianlss_bus.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:149d3bd36c34cd17578798b68311060421fd1f0fcc83d4bdb4832a6dfa802943
3
+ size 33797707
segmentation/gaussianlss/gaussianlss_sedan.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ca7cf0901b99c34a49fc5ef9cfeaf3fcd2c68d315b2d57a3007f2fb62d36e22
3
+ size 33799243
segmentation/gaussianlss/gaussianlss_suv.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:087171e00240fcaff15a64017cedaf1a977669566e09360b1f804ad91da26081
3
+ size 33797707