MIT-Media-Labcc-by-nc-4.0

HOI4D VITRA Streaming v1

A vision-language-action dataset converting HOI4D hand-object interaction data into a VLA-compatible streaming format with egocentric RGB-D video, MANO hand poses, camera trajectories, and action annotations.

Downloads34
Episodes1682

Why This Matters for Physical AI

Provides egocentric hand-object interaction data with precise MANO hand pose tracking and camera-relative action representations essential for training vision-language-action models for dexterous manipulation and hand tracking tasks.

Technical Profile

Modalities
rgbdepthlanguageproprioceptionpoint_cloud
Action Space
angle
Environment
lab
Task Types
hand-object-interactionmanipulation
Episodes
1682
Data Format
npy
Annotation Types
language_instructionsaction_labelshand_posesegmentationcamera_extrinsicsdepthkeypoints
License
cc-by-nc-4.0
Part of the HOI4D family

Access

Need custom rgb data?

Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets