Gen-HumanEgo
Description
Gen-HumanEgo contains 1,847.7 hours and 44,632 episodes of first-person human demonstrations across 10,257 tasks in home, business, industry, and agriculture settings. Six synchronized DAS-Ego cameras capture 1600 × 1300 RGB at 30 fps; GenRobot’s processing pipeline adds reconstructed 3D hand keypoints, MANO parameters and hand meshes, estimated depth, and hierarchical video-, task-, and subtask-level annotations. Episodes are packaged in MCAP. This is distinct from the publisher’s handheld-gripper RealOmni dataset; access is automatically granted after submitting an intended-use form.
License
The publisher’s Hugging Face dataset card declares CC BY-SA 4.0; attribution and share-alike conditions apply. Access also requires the publisher’s intended-use form.*
* Double-check the publisher's current license and usage terms before using this dataset.