Video Annotation Tool using Human Pose Estimation for Sports Training
João Diogo, Rui Rodrigues, Rui Neves Madeira, Nuno M. R. Correia · 2022
This paper presents and discusses the integration of human pose estimation techniques into an existing web-based multimodal video annotation tool, applying it to the sports context, where basketball is the first case study. The relevance of video analysis extends across many fields of work (e.g., professional sports, education). In sports, systematic, detailed analysis using videos of players and teams is vital to evaluate many aspects of both training and competition. MotionNotes annotation tool now combines human pose and motion information with existing traditional annotation mechanisms (e.g., text and drawings annotations), allowing users to add further details to their annotation work. The paper reports feedback from a pilot study based on a participatory workshop involving people with relevant competitive experience in basketball. Based on this use case feedback, we conclude with an outlook of future iterations for our video annotation tool.