Human Understanding in Real-World Sports

Abstract
Human understanding in computer vision goes beyond simply detecting people in images. It focuses on interpreting body movements, analysing interactions, and describing motion in natural language. Key tasks include 2D and 3D pose estimation, human mesh recovery, and action recognition. These capabilities enable applications such as AI fitness coaching, human–robot interaction, rehabilitation monitoring, and immersive augmented or virtual reality experiences.

Sports scenarios in real-world settings introduce additional challenges, including multiple people moving simultaneously, frequent occlusions, high-speed motions, and diverse body shapes and clothing. This research aims to develop robust methods that can accurately understand complex human behaviours under these uncontrolled conditions.

Bio
Qingzheng Xu is a Ph.D. student at the School of EECS, University of Queensland, under the supervision of Dr Xin Yu and A/Prof Sen Wang. His research focuses on human understanding in real-world sports.