Sports Science AI
Sports Science AI is a multi-modal AI platform that combines biomechanical motion analysis with large language models to automatically translate complex sports performance data into accessible, evidence-based coaching feedback across fitness, baseball, and badminton.
Project Description
Sports Science AI is a multi-modal artificial intelligence platform designed to enable scalable, high-precision sports science by integrating motion capture biomechanics, video-based pose estimation, and large language models for automated coaching feedback. The platform processes three-dimensional kinematic features extracted from human movement and translates biomechanical deviations into interpretable, context-specific training guidance. Through this architecture, Sports Science AI supports both performance optimization and injury risk mitigation, providing quantitative assessment alongside automated natural-language explanation. This project operates as a unified, expandable ecosystem comprising fitness biomechanics, baseball skill assessment, and badminton technique analysis. The fitness module evaluates exercise execution in terms of joint loading, movement efficiency, and compensatory patterns during strength and conditioning tasks. The baseball module quantifies batting motion in adolescent athletes through comparative biomechanical modeling against expert swing templates, enabling age-appropriate skill correction. The badminton module applies pose-based inference to characterize stroke mechanics, footwork timing, and tactical movement quality. In combination, these capabilities demonstrate SportWell’s potential as a comprehensive AI framework capable of translating professional sports science into accessible, evidence-based digital coaching for diverse athletic populations.
System Overiew
Faculty Leads
Prof. Ming-Chun Huang
Associate Professor, Duke Kunshan University
Contributors
Zhuohan Zhou
Undergrad Student, Duke Kunshan University
Chengyi Li
Undergrad Student, Duke Kunshan University