Multi-Person Human-Centric Motion Understanding for Complex Coordinated Activities

Understanding coordinated human activities involving multiple interacting people — such as team sports, group dance, or collaborative physical tasks — requires reasoning not just about individual pose and motion but about the relational and temporal coordination between multiple people simultaneously. This project proposes a multi-person, human-centric motion understanding framework that jointly models individual pose dynamics and inter-person coordination patterns to recognize and analyze complex coordinated activities from video.