Ant Group releases 2,000 hours of egocentric manipulation data for embodied learning
burny_tech · x · 2026-07-26
Open-AoE opens a 2,000-hour egocentric manipulation dataset and training pipeline
Ant Group and Ant Digital Technologies released Open-AoE, an open egocentric manipulation dataset and full toolchain for embodied learning.
- The first release includes about 2,000 hours of manipulation videos captured in natural environments.
- Data comes from 500+ contributors and 400+ smartphones.
- The package includes text annotations, MANO-based hand poses, camera trajectories, and temporally localized atomic actions.
- It also ships a processing pipeline that turns raw recordings into structured samples through action segmentation, semantic annotation, hand reconstruction, and camera trajectory reconstruction.
- The downstream toolchain supports visualization, cross-embodiment retargeting, model-specific data conversion, and training recipes for VLA policies, WAMs, and World Models.
The project aims to lower the barrier for open data contribution and reuse in embodied AI, human-to-robot transfer, and world modeling.
More from Embodied
- Rumor says Flux 3 unifies image, video, audio, and action prediction — theteknosaur · 2026-07-26
- Chinese robotics team trains humanoids with kicks, pushes, and knockdowns — 2C_ornot2C · 2026-07-26
- D-Propulse previews an air-breathing rotating detonation engine with aerospike nozzle — Paimaamu · 2026-07-26
- A reMarkable Paper Pro demo turns an e-ink tablet into an AI notebook — Olivier__OG · 2026-07-26
- T-Rex lifts dexterous hand success to 65% by giving touch its own fast path — 量子位 · 2026-07-26
- Chinese robot dog switches between wheels and legs autonomously — RoboBalaji · 2026-07-26