# DexCap: Hand Motion Capture Data for Dexterous Manipulation

## Facts

- Provider: Chen Wang et al. (Stanford University)
- License: CC-BY-4.0 (https://huggingface.co/datasets/chenwangj/DexCap-Data)
- Size: 6 files
- Format: HDF5 (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/chenwangj/DexCap-Data
- Access: public, readable without an account
- Category: Robotics
- Source: https://arxiv.org/abs/2403.07788
- Page: https://scigantic.com/archive/69246c17-6a91-58d3-bfea-fd836ab2a562
- Keywords: Robotics, robot learning, HDF5, Hugging Face

## About

The official dataset collected with DexCap, a portable hand motion capture system for learning dexterous robot manipulation from human demonstrations (Wang et al., arXiv:2403.07788). It contains 30 minutes of mocap data for a wiping task and 60 minutes of in-the-wild mocap data for a packaging task. Raw data are zip archives in which each frame folder holds the chest camera RGB image (JPEG) and depth image (PNG), the chest camera 6-DoF pose in the world frame, left and right hand 6-DoF poses, and 3D hand joint positions in the palm frame as text files. Postprocessed training sets are HDF5 files with 10,000-point point clouds and a 5-step action gap, readable with h5py and the training scripts in the DexCap GitHub repository. Measured 2026-11-14 at about 108 GB across two raw zips and two HDF5 files. Useful for human-to-robot retargeting, point-cloud imitation learning and dexterous bimanual manipulation. The card releases the dataset under CC BY 4.0; cite the DexCap paper.

## Load the data

```python
from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="chenwangj/DexCap-Data", filename="hand_packaging_wild_1-20_5actiongap_10000points.hdf5", repo_type="dataset")
print(path)
```

## Data structure

File types in sampled files: HDF5

Example file: `hand_packaging_wild_1-20_5actiongap_10000points.hdf5`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/69246c17-6a91-58d3-bfea-fd836ab2a562
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/69246c17-6a91-58d3-bfea-fd836ab2a562
- Over MCP: `get_schema_card` with id `69246c17-6a91-58d3-bfea-fd836ab2a562` at https://api.scigantic.com/mcp

## Related topics

- [HDF5 datasets](https://scigantic.com/topics/hdf5)
- [Robotics datasets](https://scigantic.com/topics/robotics)
- [Robot learning datasets](https://scigantic.com/topics/robot-learning)
