Upload README.md with huggingface_hub

README.md

@@ -9,7 +9,26 @@ tags:
 
 # ExtremeOcc-3D
 
-A benchmark of 264 single-view scenes for amodal 3D generation under extreme occlusion. Each sample provides an occluded scene render, an object mask, an isolated observed-object render, a category text label, and ground-truth meshes / multi-view renders.
+ExtremeOcc-3D is a benchmark for amodal 3D generation under severe object
+occlusion, introduced with [RelaxFlow: Text-Driven Amodal 3D
+Generation](https://arxiv.org/abs/2603.05425), an ICML 2026 Spotlight paper.
+
+The benchmark contains 264 single-view indoor scenes. Each sample provides an
+occluded scene render, a target-object mask, an isolated render of the visible
+object part, a category text label, and ground-truth 3D assets for evaluation.
+
+| Item | Value |
+|---|---|
+| Paper | [RelaxFlow: Text-Driven Amodal 3D Generation](https://arxiv.org/abs/2603.05425) |
+| Venue | ICML 2026 Spotlight |
+| Task | Text-guided amodal 3D generation under extreme occlusion |
+| Samples | 264 single-view scenes |
+| Inputs | Occluded RGB render, target mask, observed-object render, category text |
+| Ground truth | Meshes and multi-view renders for target objects |
+| Code | [viridityzhu/RelaxFlow](https://github.com/viridityzhu/RelaxFlow) |
+
+Use `manifest.json` as the entry point. All paths in the manifest are relative
+to this dataset root.
 
 ## Folder layout
 
@@ -22,7 +41,7 @@ ExtremeOcc-3D/
     └── <object_hash>/{mesh.ply, view_0000.png, ...}
 ```
 
-## Manifest fields
+## Manifest
 
 Each entry contains:
 
@@ -38,6 +57,13 @@ Each entry contains:
 
 All paths are relative to this `ExtremeOcc-3D/` directory.
 
+## Usage
+
+Download this repository, then pass `manifest.json` to the RelaxFlow batch
+runner. If you generate prior images separately, use
+`prepare_manifest_with_priors.py` from the RelaxFlow release to attach them to a
+copy of the manifest.
+
 ## Adding prior images
 
 The manifest does not include a `prior_images` field. Generate priors with any image generator you prefer, organize them per sample, then run the helper to produce a complete manifest: