10K-hour egocentric video dataset released: 72,754 head-mounted episodes for robot learning
rohanpaul_ai · x · 2026-11-17
A large-scale egocentric video dataset has shipped its first release: 10,124 hours (1.09B frames) of head-mounted footage for robot/embodied AI training.
- Scale: 72,754 episodes (median 4.9 min, up to 30 min) from 12,227 contributors.
- Environments: 15,548 distinct environments across 11 setting types; 83% of hours filmed in workplaces.
- Tasks: 62,000+ unique task descriptions spanning 56 task types, 30 categories, 35 action families.
- Modalities: 1920x1080 30fps ultra-wide head-mounted video with hands in frame 95%+ of the time; 100Hz IMU aligned to every frame plus magnetometer, barometer and other native streams.
- Format: LeRobot v3 + MCAP with per-episode catalogs and annotations.
Related event: Hub open-sources HubEgo-100K, a 100K-hour egocentric skill video dataset(5 posts)→
More from Embodied
- UMA Robotics bets on Europe first: survey shows EU edges US on talent, customers, cost — RemiCadene · 2026-11-18
- Tesla letter: FSD hits 1.74M paid users, Robotaxi in 7 cities ahead of Dec 3 vote — yunta_tsai · 2026-11-18
- Open-Source Robot Lamp Preps Skill Store: Install Physical Skills Like Apps — dee_hw · 2026-11-17
- Genesis AI raises $500M Series A at $3.5B valuation for general-purpose robots — chris_j_paxton · 2026-11-17
- Robotics researchers spar: pick-and-place with dexterous hands is 'so boring' — chris_j_paxton · 2026-11-17
- Open-Source reBot Arm B601-DM Robotic Arm on Sale for $1,090 — philfung · 2026-11-17