A Comparison of Accuracy in Pose Estimation for Sprinters based on Skeleton Recognition

Abstract

Traditionally, optical motion capture systems have been widely utilized to accurately obtain athletes’ pose data at the laboratory settings. However, it is impossible to use such wearable motion capture in real races to analyze the performance of participating athletes. In this study, we analyzed sprinters’ form using markerless posture recognition technology, specifically OpenPose, and investigated the feasibility of applying pose estimation to sprint running.