README.md

---
license: other
license_name: mixed-source
license_link: "https://huggingface.co/datasets/spatialgym/SpatialGym_Figural#sources-and-licensing"
pretty_name: "SpatialGym Figural Objects"
tags:
  - 3d
  - usd
  - objects
  - spatial-reasoning
task_categories:
  - other
viewer: false
---

# SpatialGym Figural Objects

Object packs for the figural (tabletop) tasks of SpatialGym — **Pose Matching**, **Mental Rotation**
and **Active Closure**. Each pack is a set of self-contained USD objects baked from one curated
catalog of source meshes (RoboTwin 2.0 rigid objects + YCB `google_16k` scans), plus a `manifest.json`
the environments read. The other six figural packs (polycubes, CLEVR objects, spatial-relation solids,
assembly, net folding) are generated locally by `assets/download.sh` and are not part of this repository.

## Overview

| Pack | Objects | Sources | What the environment gets | Download | Unpacked |
|------|--------:|---------|---------------------------|---------:|---------:|
| **pose_matching** | 411 | RoboTwin 354 · YCB 57 | rest pose, metric size, PBR textures | 1.4 GB | 1.4 GB |
| **mental_rotation** | 151 | RoboTwin 132 · YCB 19 | upright, longest axis 1.7 m, yaw-asymmetric subset | 534 MB | 527 MB |
| **active_closure** | 307 | RoboTwin + YCB | voxelised body (cube `PointInstancer`) + 4-way options | 4 MB | 20 MB |

## Files

```
SpatialGym_Figural/
├── README.md
├── pose_matching.tar.gz      # unpacks to pose_matching/{manifest.json, objects/<id>.usd, objects/textures/}
├── mental_rotation.tar.gz    # unpacks to mental_rotation/{manifest.json, objects/<id>.usd, objects/textures/}
└── active_closure.tar.gz     # unpacks to active_closure/{manifest.json, <id>.usd}
```

`SpatialGym_IsaacLab/assets/download.sh figural` in the SpatialGym repository fetches the three
packs into `assets/data/figural/`, builds the six generated packs next to them and checks every
pack against the version its code expects.

## Packs

**pose_matching** — `objects/<id>.usd`, one per catalog entry, in its canonical rest pose at real-world
size; textures externalised under `objects/textures/`.

<p align="center"><img src="figures/pose_matching.png" width="100%" /></p>
<p align="center"><sub>Pose Matching: one task frame per scene, difficulty 1–3.</sub></p>

| Key | Meaning |
|-----|---------|
| `id`, `usd` | Object id and USD path relative to the pack |
| `label_l1`, `label_l2` | Coarse / fine object labels shown to the agent |
| `footprint_cm`, `height_cm` | Rest-pose footprint `[w, l]` and height |
| `yaw_symmetric` | `true` ⇒ the object looks the same at every yaw and is excluded from the manipulable pool |
| `source`, `orig_id`, `variant` | Provenance: source dataset, original id, texture / shape variant |

**mental_rotation** — the yaw-asymmetric subset of the same catalog, re-baked upright (rest rotation
zeroed) and rescaled so every object's longest axis is 1.7 m. No per-object difficulty: object complexity
is not an axis of this task.

<p align="center"><img src="figures/mental_rotation.png" width="100%" /></p>
<p align="center"><sub>Mental Rotation: top row the locally generated polycube pool (not in this repository), bottom row this pack's curated objects.</sub></p>

| Key | Meaning |
|-----|---------|
| `id`, `usd`, `label_l1`, `label_l2`, `source`, `orig_id`, `variant` | As above |
| `dims_m` | Bounding-box extent `[x, y, z]` in metres, measured off the baked USD |

**active_closure** — one `<id>.usd` per object holding a `PointInstancer` of cubes (the voxelised body,
348–30 221 voxels), floated in a booth for the closure question.

<p align="center"><img src="figures/active_closure.png" width="100%" /></p>
<p align="center"><sub>Active Closure: one object per column, visibility rungs 1–3 down the column.</sub></p>

| Key | Meaning |
|-----|---------|
| `id`, `usd`, `category` | Object id, USD path, category label |
| `num_voxels`, `size_m` | Voxel count and longest-axis size of the voxelised body |
| `options`, `answer_idx` | A 4-way option set and the correct index. The environment draws its own option set at load; these are kept for older code |

All figures are frames the tasks themselves render (oracle runs, the environments' own camera and lighting).

## Sources and licensing

The objects are derived from two source datasets:

- **RoboTwin 2.0** object library — code and dataset released under the MIT license.
  Chen et al., *RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization
  for Robust Bimanual Robotic Manipulation*, arXiv:2506.18088, 2025.
- **YCB Object and Model Set** (`google_16k` scans) — Calli et al., *The YCB Object and Model Set:
  Towards Common Benchmarks for Manipulation Research*, ICAR 2015. Model files are distributed under
  CC BY 4.0.

The baked USD packs in this repository are provided for research use; please cite SpatialGym and
the source datasets above when you use them.