Organize PROWBench inputs for browsing

README.md

@@ -41,17 +41,17 @@ PROWBench provides structured conditioning inputs for video/world-model evaluati
 | `unverified_first_frame` | 90 | 90 | 270 | 0 |
 | `neon` | 20 | 20 | 60 | 20 |
 
-## Representative images
+## Visual samples
 
-These examples show the distinction between provided RGB first frames and frames extracted from proxy input videos. The same images are embedded in the Parquet `preview_image` column for the selected samples.
+One representative image per episode is indexed in the Parquet `preview_image` column. These five compact examples show both kinds of image; click one to inspect it at full size.
 
-| Group | Preview | Source |
-|---|---|---|
-| `verified_first_frame` | ![City crossing RGB first frame](https://huggingface.co/datasets/AlayaLab/PROWBench/resolve/main/images/verified_first_frame/dv200-001-city-crossing/follow.png) | RGB first frame |
-| `neon` | ![Neon RGB first frame](https://huggingface.co/datasets/AlayaLab/PROWBench/resolve/main/images/neon/ndm-01-cam-dynamic_obj-static-death/follow.png) | RGB first frame |
-| `long_video` | ![Long-video proxy preview](https://huggingface.co/datasets/AlayaLab/PROWBench/resolve/main/previews/long_video/long30-001-forest-long-return.png) | Proxy-video frame |
-| `multi_view` | ![Multi-view proxy preview](https://huggingface.co/datasets/AlayaLab/PROWBench/resolve/main/previews/multi_view/multi4-001-forest-staggered-follow.png) | Proxy-video frame |
-| `unverified_first_frame` | ![Unverified-first-frame group proxy preview](https://huggingface.co/datasets/AlayaLab/PROWBench/resolve/main/previews/unverified_first_frame/dv200-003-city-traffic.png) | Proxy-video frame |
+| RGB first frame · city crossing | RGB first frame · neon |
+|:---:|:---:|
+| <a href="./images/verified_first_frame/dv200-001-city-crossing/follow.png"><img src="./images/verified_first_frame/dv200-001-city-crossing/follow.png" width="240" alt="City crossing RGB first frame"></a> | <a href="./images/neon/ndm-01-cam-dynamic_obj-static-death/follow.png"><img src="./images/neon/ndm-01-cam-dynamic_obj-static-death/follow.png" width="240" alt="Neon RGB first frame"></a> |
+| Proxy frame · long video | Proxy frame · multi-view |
+| <a href="./previews/long_video/long30-001-forest-long-return.png"><img src="./previews/long_video/long30-001-forest-long-return.png" width="240" alt="Long-video proxy preview"></a> | <a href="./previews/multi_view/multi4-001-forest-staggered-follow.png"><img src="./previews/multi_view/multi4-001-forest-staggered-follow.png" width="240" alt="Multi-view proxy preview"></a> |
+| Proxy frame · unverified-first-frame group |  |
+| <a href="./previews/unverified_first_frame/dv200-003-city-traffic.png"><img src="./previews/unverified_first_frame/dv200-003-city-traffic.png" width="240" alt="Unverified-first-frame group proxy preview"></a> |  |
 
 The videos in this repository are **proxy input videos**, not generated model results. RGB first frames are included only where recorded in the source catalog. Depth and generated result videos are not part of this release.
 

SHA256SUMS

@@ -1,4 +1,4 @@
-b203e95bb72b34b41c88c1da6a45c532504d3bdad73266e1674041035d2da08d  README.md
+19336d62828b920aec2b1eaf49cb3fc170be3d7b9c23330afbbee6ced1444f91  README.md
 c190eb114e83677c00e0527c663c99d7435c37972951806f40654d7e23a1dca3  archives/PROWBench-no-depth-20261008T152517Z.tar.zst
 0c48e186c946fcf267a4dc45203d70dfdab014243aeca8ece082d9a9665f4d9a  organization.json
 55b8613d2e5f18d68a8b17d288961d57684c5966c0d0d228f9bb7e2393727629  organized-files.jsonl.gz