KTPose: Keypoint-Based Tokens in Vision Transformer for Human Pose Estimation | IEEE Conference Publication | IEEE Xplore