BKinD-3D: Self-Supervised 3D Keypoint Discovery from Multi-View Videos | IEEE Conference Publication | IEEE Xplore