To address these issues, this paper proposes CAPED, an end-to-end robust Category-level Articulated object Pose Estimator integrated differentiable rendering. Given partial point cloud as input, CAPED ...