Leveraging SE(3) Equivariance for Self-supervised Category-Level Object Pose Estimation from Point Clouds