Markerless Motion Capture Data of Fifteen Tennis Players Completing an On-Court Fatiguing Protocol

This dataset contains 3D keypoint data and kinematic data from a custom markerless motion capture system of 15 tennis players completing an on-court fatiguing protocol. Each trial in this dataset consists of a serve and eight groundstrokes (alternate sides of the court), followed by a maximal sprint. Additionally, there is a baseline maximal sprint. For each player, a scaled OpenSim model is provided. For each trial, two files are provided: Keypoints from the markerless motion capture system and kinematics derived from OpenSim.Keypoints: 3D filtered keypoint data. Processing: Video data captured from 8 cameras at 200 Hz; 2D pose estimation performed with a pre-trained HRNet model through MMPose; 2D keypoints reconstructed in 3D, gap-filled and filtered with a Kalman Smoother; data converted to trc format to be compatible with OpenSim.Kinematics: Kinematics (position / orientation & velocity / angular velocity) for all body segments and centre of mass. Processing: Keypoints data used to scale an OpenSim model (available at https://github.com/‌‌julieemmerson/​‌‌‌‌‌‌opensim_tennis_model) and run inverse kinematics for each trial; OpenSim analyse tool used to filter joint angles (Butterworth; 6Hz cut-off) and export kinematics.