This validation study involved the simultaneous collection of video and marker data using a single iPhone and an 8-camera Qualisys marker-based motion capture system, during acceleration and maximum velocity sprinting, across three athletes.
SprintAI's pose estimation core is built on Sports2D and Pose2Sim, open-source tools developed by David Pagnon and colleagues that combine deep-learning pose estimation with OpenSim musculoskeletal inverse kinematics. Their accuracy and robustness have been reported in peer-reviewed studies.
The Pose2Sim studies used multi-camera 3D reconstruction. SprintAI uses a single camera, so its own accuracy against Qualisys is reported below.
Hip, knee and ankle angles through the full step, SprintAI against 8-camera Qualisys motion capture.
RMSE beside published OpenCap values (2 cameras) as reported.
Between-system SEM, beside Crespeau et al. (2026), sprinting with an 8‑camera markerless system.
Agreement between SprintAI and Qualisys for spatiotemporal and kinematic KPIs, grouped by category. Values are the median absolute bias of the KPIs in each category, in native units and as a percentage of the Qualisys mean. Select a category to view individual KPIs.