Learning to Predict Keypoints and Structure of Articulated Objects without Supervision | IEEE Conference Publication | IEEE Xplore