Robot learning depends on large and diverse data, yet robot data collection remains expensive and difficult to scale. HanVerse lets anyone contribute egocentric video of everyday tasks and earn HAN tokens, powering the next generation of robot learning. Whitepaper →
Capture
Record egocentric video of daily tasks using the HanCapture mobile app. The app guides framing and ensures data quality on-device.
Validate
Data is scored for quality, annotated with dense language labels, 6-DoF camera poses, and 3D hand poses (21 keypoints per hand). Provenance is registered on-chain.
Earn
Contributors receive HAN tokens based on quality scores, task diversity, scene novelty, and annotation density. Every data ID is a verifiable on-chain asset.
Train
Research labs and robotics companies access curated datasets to train generalizable manipulation policies across diverse robot embodiments.
Georgia Tech RL2 Lab
Stanford REAL Lab
Mecka AI
EgoVerse