Virtual Pose Coach: A Motion-Retargeting Approach for Pose Training

Tzu-Chun Chiu, Ming-Han Lee, Kun-Ru Wu, Yu-Shuen Wang, Yu-Chee Tseng; Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) Workshops, 2025, pp. 5926-5934

In sports training, effective learning necessitates that students accurately replicate their coach's movements. However, anatomical differences, such as variations in limb length and skeletal proportions, can make it difficult to achieve proper pose alignment and may diminish training effectiveness. To tackle this issue, we propose a virtual pose coach framework. This innovative approach uses motion retargeting to create personalized virtual poses tailored to each student's body structure, which enhances their ability to imitate the coach's movements accurately. Unlike current methods that adjust joint rotations between characters and often experience discrepancies in rotation distributions during training and testing -- due to the countless possible transitions between two poses --our framework focuses on joint positions for retargeting. This strategy eliminates ambiguity and enables effective motion retargeting across different body structures, facilitating motion transfer between skeletons with varying anatomical designs. We illustrate the advantages of our system through a case study that shows how our retargeting method significantly enhances students' ability to replicate a coach's movements, indicating its potential to improve sports training outcomes.