Ego2Robot: Synthesizing 18,500+ Hours of Robot Training Data from Egocentric Videos
Ye Wang · hf · 2026-08-06
Ego2Robot introduces a scalable pipeline that converts egocentric human manipulation videos into robot training data.
- Core Method: It utilizes action retargeting, robot-arm visual synthesis, and multi-level quality curation to transform in-the-wild human videos into usable robot demonstrations.
- Data Scale: The pipeline produces 18,561 hours of training data across 15 robot morphologies, creating the largest ego-to-robot dataset to date.
- Evaluation: The authors extend RoboTwin2.0 with disentangled perturbation axes (visuals, layout, morphology, semantics). Joint pretraining with this data consistently improves out-of-distribution generalization for VLA models, validated by real-robot deployments.
More from Embodied
- Developer Shows Sneak Peek of New Vibescape Environment for Apple Vision Pro — dreamwieber · 2026-08-06
- 100k Hours of Real-World Data Yields Only 57% Success in Robotics — eigenron · 2026-08-06
- Dongfang Starlink Launches AI Hyperspectral Satellites, Plans 2M Lunar Robots for Data Center — 新智元 · 2026-08-06
- USTC Study: AI Agents Achieve Only 3.3% Zero-Intervention Execution in Robot Labs — 新智元 · 2026-08-06
- Tsinghua Startup Uses 'Uncertain Differential Geometry' to Challenge End-to-End Robot Models — 机器之心 · 2026-08-06
- Indie Dev Uses AI to Design Circuit Boards, Aims to Crack Production and Sell Hardware — pramodk73 · 2026-08-06