Human2Any Transfers Human Video Skills to Robots with Zero Real-World Training Data
danfei_xu · x · 2026-11-08
Human2Any, a Georgia Tech × NVIDIA framework accepted at CoRL 2026, tackles both what and how to transfer from human videos to robots.
- Learns object-centric interaction priors from human videos, representing manipulation via object-object interaction motion that strips away embodiment-specific details
- Composes these priors with robot-side feasibility reasoning and motion planning, adapting to different embodiments, scene layouts, and task contexts via test-time reasoning and planning
- Deploys on a real Franka arm and an RBY-1 humanoid mobile robot with no real-world robot demonstrations, generalizing to new scenes and long-horizon tasks
More from Embodied
- PyTorch ATX and Austin Robotics host robot learning meetup alongside CoRL — kscottz · 2026-11-08
- Samsung Robotics Proposes ε4P: Route Imperfect Robot Data by Noise Level for Precision Manipulation — jiqizhixin · 2026-11-08
- Amazon FAR's 'Pick Anything' grasps 66 object types with a single policy — JitendraMalikCV · 2026-11-08
- A group of humanoid robots walks through Figure AI HQ, Westworld vibes — CyberRobooo · 2026-11-08
- CoRL 2026 sells out: record 2,094 submissions, 687 accepted papers — Majumdar_Ani · 2026-11-08
- Clockwork uConsole kit boots on first assembly, but the screen stays dark — vincent_koc · 2026-11-08