Egocentric Video Dataset for Physical AI and Robotics

239.3 hours of first-person video data for egocentric vision and tracking. The dataset features multimodal data from egocentric views, including annotations and motion capture for extracting 3D poses, and provides detailed 3D objects and scenes using visual data from VR headsets. It was authored by UniDataPro and last updated on 2026-05-12.

Use Cases

• Train 3D pose estimation models based on motion capture annotations.
• Analyze hand motions and interactions based on first-person video data.
• Develop scene understanding algorithms based on detailed 3D object and scene reconstructions.
• Advance egocentric tracking systems based on multimodal data from VR headsets.

Strengths

• Contains 239.3 hours of first-person video data.
• Includes motion capture data for extracting 3D poses.
• Provides detailed 3D object and scene reconstructions from VR headset data.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file size are unknown, which may limit suitability assessment.
• Freshness should be verified as the last update timestamp is in the future (2026-05-12).

Provenance

Source

UniDataPro via Hugging Face.

Collection Method

Likely gathered from VR headsets for egocentric vision research.

Freshness

Last updated 2026-05-12 00:59:48.