Sports Analysis & AI Research

Sports Research Overview

This research focuses on video-based sports analytics for understanding, evaluating, and improving human movement in athletic activities. By leveraging computer vision, deep learning, and 3D human pose estimation, the work aims to extract fine-grained motion representations from video data and translate them into actionable insights for performance analysis and injury prevention.

A key emphasis is placed on technique-aware and phase-aware motion modeling, enabling the analysis of complex sports movements such as cyclic actions, rapid transitions, and whole-body coordination. These models are designed to capture both spatial posture and temporal dynamics, which are essential for assessing movement quality and efficiency in sports.

Technically, the research investigates multimodal learning frameworks that integrate RGB video, optical flow, skeletal keypoints, and biomechanical cues. In addition, multi-view 3D reconstruction and temporal refinement techniques are explored to achieve robust motion estimation under unconstrained outdoor and competition environments.

The ultimate goal is to develop interpretable and scalable sports AI systems that support athlete training, performance optimization, and movement safety, bridging advanced computer vision research with real-world sports science applications.