KeyGen: Unsupervised Keypoint based Object-Centric Representations for Category-Level Policy Generalization
KeyGen is introduced, a framework that learns canonicalized semantic 3D keypoints from point clouds and uses them as structured object-centric representations for policy learning to predict full manipulation trajectories, enabling consistent geometric correspondence across object instances.
Shu-Xin Cao, Liquang Wang, Masoud Moghani et al.
· 1 citation