Stick/club/racquet sports

Document Type



This pilot study assessed the accuracy of markerless motion capture to estimate the external mechanical work performed during tennis serves. One tennis player performed 9 serves whilst motion data were captured concurrently with a criterion marker-based and a custom markerless system (utilising HRNet and OpenPose). Centre of mass kinetic and potential energy were calculated and used to compute external mechanical work for all 3 approaches. Markerless methods yielded differences of 2-3% (HRNet) and 6-9% (OpenPose) for external work compared to the criterion measure, with the former regarded as a high level of agreement. Our markerless system, paired with HRNet, shows promise for accurately estimating external work during the tennis serve and has potential to provide players and coaches with a non-invasive tool for monitoring training ‘load’ in the field.