nepfaff's picture
Initial upload: meshes + V-HACD/CoACD baseline archives
fd0c0af verified
---
license: apache-2.0
tags:
- robotics
- simulation
- 3d-assets
- convex-decomposition
- collision-geometry
- benchmark
size_categories:
- 1K<n<10K
---
# Visibility Decomposition Benchmark
Benchmark dataset of diverse 3D meshes for evaluating convex decomposition methods. Used by
[visibility-decomposition](https://github.com/wernerpe/visibility-decomposition) to compare
visibility-based decomposition against [V-HACD](https://github.com/kmammou/v-hacd) and
[CoACD](https://github.com/SarahWeiii/CoACD).
## Dataset Composition
| Subset | Meshes | Format | Description |
|--------|--------|--------|-------------|
| CoACD Examples | 5 | .obj | Smoke-test meshes from the [CoACD repo](https://github.com/SarahWeiii/CoACD) |
| ArtVIP | 528 | .obj | Per-link meshes from 205 articulated objects across 5 categories |
| SceneSmith | 500 | .glb | Stratified sample of generated objects from [SceneSmith](https://scenesmith.github.io/) scenes |
**Total: 1,033 meshes**
## Directory Structure
```
coacd_examples/{object_name}/original.obj
artvip/{category}/{object_name}/{link_name}/original.obj
scenesmith/{category}/{object_name}/original.glb
manifest.json
```
## Usage
```bash
git clone https://github.com/wernerpe/visibility-decomposition
cd visibility-decomposition
python scripts/download_benchmark_dataset.py
python scripts/run_benchmark.py
```
## Acknowledgments
### ArtVIP
The articulated object meshes are derived from the
[ArtVIP](https://huggingface.co/datasets/x-humanoid-robomind/ArtVIP) dataset
(Apache License 2.0).
> Zhao Jin, Zhengping Che, Tao Li, Zhen Zhao, Kun Wu, Yuheng Zhang, Yinuo Zhao,
> Zehui Liu, Qiang Zhang, Xiaozhu Ju, Jing Tian, Yousong Xue, Jian Tang.
> *ArtVIP: Articulated Digital Assets of Visual Realism, Modular Interaction,
> and Physical Fidelity for Robot Learning.* 2025.
### SceneSmith
The generated object meshes are from scenes produced by
[SceneSmith](https://scenesmith.github.io/), sourced from
[scenesmith-example-scenes](https://huggingface.co/datasets/nepfaff/scenesmith-example-scenes).
Preprocessed ArtVIP assets used by SceneSmith are available at
[scenesmith-preprocessed-data](https://huggingface.co/datasets/nepfaff/scenesmith-preprocessed-data).
### CoACD
The CoACD example meshes are from the [CoACD](https://github.com/SarahWeiii/CoACD) repository.