DexCanvas: Dexterous Manipulation Dataset with 30M Frames

DexCanvas is a large-scale hybrid dataset for robotic hand-object interaction research, combining real human demonstrations with physics-validated simulation data. The v0.1 test release contains approximately 30 million multi-view RGB-D frames, representing about 70 hours of dexterous hand-object interaction. It was created by DEXROBOT and last updated on February 2, 2026.

Use Cases

• Training imitation learning models based on real human demonstration data.
• Developing and validating physics simulators using the physics-validated simulation data.
• Researching multi-view perception for robotics based on the RGB-D frames.
• Benchmarking dexterous manipulation algorithms on a large-scale dataset.

Strengths

• Large scale with approximately 30 million multi-view RGB-D frames.
• Hybrid nature combining real human demonstrations with simulation data.
• Substantial duration of about 70 hours of interaction data.

Limitations

• This is a test release containing only 1% of the full dataset.
• Contact force data is not included in v0.1.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

DEXROBOT

Collection Method

Combines real human demonstrations with physics-validated simulation data.

Freshness

Last updated 2026-02-02 07:59:59; freshness should be verified.