KTPose A novel keypoint-based tokens in Vision Transformer for human pose estimation 结构图如下: 主要创新点: 模型效果: