Learning View-Disentangled Human Pose Representation by Contrastive Cross-View Mutual Information Maximization | IEEE Conference Publication | IEEE Xplore