File size: 3,876 Bytes
b6fb9f9 5ebe21a b6fb9f9 5ebe21a b6fb9f9 5ebe21a b6fb9f9 5ebe21a b6fb9f9 f704036 b6fb9f9 27bf61c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 |
---
license: apache-2.0
tags:
- robotics
- simulation
- 3d-assets
- articulated-objects
- drake
- sdf
pretty_name: SceneSmith Preprocessed Data
---
# SceneSmith Preprocessed Data
Preprocessed 3D assets for use with [SceneSmith](https://github.com/nepfaff/scenesmith), a VLM-agent-based system for generating physically realistic, interactive indoor scenes.
## ArtVIP (Articulated Objects)
Simulation-ready articulated objects (cabinets, drawers, appliances, etc.) converted from the [ArtVIP](https://huggingface.co/datasets/x-humanoid-robomind/ArtVIP) dataset. Assets have been converted from USD to Drake SDFormat using [mesh-to-sim-asset](https://github.com/nepfaff/mesh-to-sim-asset), with:
- **Drake SDFormat (.sdf)** model files with articulated joints
- **Visual meshes** in glTF and OBJ formats
- **Collision geometries** via convex decomposition (two variants available)
- **Estimated physical properties** (mass, inertia per link)
- **Pre-computed CLIP embeddings** for text-based retrieval
### Variants
| Archive | Collision Method | Description |
|---------|-----------------|-------------|
| `artvip/artvip_vhacd.tar.gz` | VHACD | Tighter collision geometries (recommended) |
| `artvip/artvip_coacd.tar.gz` | CoACD | Can produce faster simulations |
Both variants include identical CLIP embeddings for retrieval.
### Usage with SceneSmith
Download your preferred variant and extract to `data/artvip_sdf/`:
```sh
# Download VHACD variant (recommended)
huggingface-cli download nepfaff/scenesmith-preprocessed-data \
artvip/artvip_vhacd.tar.gz --repo-type dataset --local-dir .
mkdir -p data/artvip_sdf
tar xzf artvip/artvip_vhacd.tar.gz -C data/artvip_sdf
rm -rf artvip
# Or download CoACD variant
huggingface-cli download nepfaff/scenesmith-preprocessed-data \
artvip/artvip_coacd.tar.gz --repo-type dataset --local-dir .
mkdir -p data/artvip_sdf
tar xzf artvip/artvip_coacd.tar.gz -C data/artvip_sdf
rm -rf artvip
```
### Asset Categories
| Category | Description |
|----------|-------------|
| `large_furniture/` | Cupboards, wardrobes, etc. |
| `small_furniture/` | Nightstands, side tables, etc. |
| `major_appliances/` | Refrigerators, washing machines, etc. |
| `small_appliances/` | Microwaves, toasters, etc. |
| `household_items/` | Storage boxes, bins, etc. |
### Manual Preprocessing
To preprocess ArtVIP assets yourself (or to process updated versions), use [mesh-to-sim-asset](https://github.com/nepfaff/mesh-to-sim-asset) to convert from USD to Drake SDFormat.
## AmbientCG Material Embeddings
Pre-computed CLIP embeddings for [AmbientCG](https://ambientcg.com/) PBR materials, enabling text-based material retrieval for walls, floors, and surfaces.
The embeddings are computed from AmbientCG preview images using CLIP (ViT-H-14-378-quickgelu, 1024D).
> **Note:** You still need to download the actual material textures from AmbientCG separately (see SceneSmith README). This dataset only contains the pre-computed embeddings to skip the computation step.
### Usage with SceneSmith
```sh
huggingface-cli download nepfaff/scenesmith-preprocessed-data \
--repo-type dataset \
--include "ambientcg/embeddings/**" \
--local-dir data/scenesmith-preprocessed-data
mv data/scenesmith-preprocessed-data/ambientcg/embeddings data/materials/embeddings
rm -rf data/scenesmith-preprocessed-data
```
## Attribution
The articulated object assets are derived from the [ArtVIP](https://huggingface.co/datasets/x-humanoid-robomind/ArtVIP) dataset, licensed under the [Apache License 2.0](https://www.apache.org/licenses/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.
|