Most egocentric data at scale is private. We think that's worth changing. With Hugging Face, we've released the first 10,000 hours of EgoSuite-Open100K — an open egocentric human dataset for Physical AI. #Robotics

The full release will total 100k hours across 15k+ tasks and 15k+ real-world scenes. 128 scene types across 7 environments. Hand pose, body pose, and event-level semantic annotation, released in LeRobot v3 and MCAP.

Not sure where to start? EgoDemo is a 50-hour sample covering every annotated subset plus two raw-video variants. Explore the full collection here:

@huggingface You should get @TexavieInc 5k1n tactile gloves. Without it you miss critical tactile data, which is the real frontier in dexterity and object manipulation, that is every real valuable task.

@huggingface Any eta for the 100k ?

@huggingface No official eta yet! We’re getting feedback for the initial 10k first.

@huggingface @MaxScore

@huggingface Awesome effort thanks @bgxc

@huggingface 10,000 open hours is huge. the robot failing weirdly might be the best data in there

@huggingface Open egocentric data is a double-edged sword. How do you address the privacy of the people whose daily lives are recorded?

@huggingface This is an impressive step forward for Physical AI and robotics. We’d love to explore a collaboration with Lightwheel to help amplify this launch and bring it to a wider, relevant audience. Open to discussing promotional opportunities?

@huggingface Hey! 👋 Your AI project looks promising. Are you open to a quick growth discussion?