CordViP
CordViP research team
A correspondence-based policy implementation with data collection, training and real-robot evaluation code.
What you can explore
Inspect the object/hand geometry pipeline and how it feeds manipulation learning.
Before you use it
Uses separate pose-estimation weights, demo data and hardware integration. FoundationPose, policy components and upstream data have their own dependencies and terms.
Original contributors
Yankai Fu, Qiuxuan Feng, Ning Chen and collaborators
License / access: Code: MIT · check upstream models, data and dependency terms
Editorial reference · checked 2026-09-22