Jul 2026
Pix2Act: Image-Space Manipulation Policies with Equivariant Augmentation
P Pix2Act is proposed, an imitation learning method that addresses high-dimensional 3D control as a simpler, more learnable 2D prediction problem by generating continuous image-space keypoint trajectories in each camera plane and losslessly recovering end-effector poses via triangulation.
Hao-zhe Huang, Lin-Feng Zhao, Haotian Liu et al.
· arXiv.org · 1 citation