ego-steerper-dataset
EgoSteer Egocentric Human Hand-Motion Dataset
A collection of 40 episodes of bimanual human hand motion from 8 public egocentric video datasets, with estimated 3D wrist poses, fingertip positions, camera poses, and free-form English instructions in a 74-d state/action layout.
Downloads277
Episodes40
Why This Matters for Physical AI
This dataset provides large-scale egocentric hand motion data from diverse real-world manipulation tasks, enabling training of vision-language-action models for dexterous bimanual manipulation with metric 3D pose supervision.
Technical Profile
- Modalities
- rgbhand_poseproprioceptionlanguagecamera_pose
- Action Space
- joint_positions
- Environment
- labkitchenfactoryhome
- Task Types
- dexterous_manipulationbimanual_manipulationtool_useobject_manipulationassemblyrepairkitchen_activities
- Episodes
- 40
- Data Format
- LeRobot
- Annotation Types
- language_instructionshand_posecamera_posehand_presence_labels
- License
- per-dataset
Access
Need custom rgb data?
Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.
Request a Sample Pack