Markerless Motion Capture in Sport and Exercise: A Systematic Review of Applications, Validity, and Reliability

Markerless motion capture is increasingly used in sport and exercise biomechanics because it reduces preparation time and equipment burden and permits assessment in more ecologically valid environments than marker-based analysis. This systematic review synthesised evidence on the applications, validity, and reliability of markerless motion capture systems in sport and exercise. In accordance with PRISMA guidelines, Web of Science, Scopus, EBSCO, and PubMed were searched for peer-reviewed studies published between 2015 and 2025. Forty-six studies met the inclusion criteria. Theia3D, OpenPose, and OpenCap were the most frequently investigated systems. Applications included walking, running, squatting, jumping, landing, throwing, kicking, striking, cutting, cycling, and tennis strokes. Under controlled conditions, the extracted sagittal-plane joint-angle outcomes showed root mean square errors of 3.20–16.41° and root mean square differences of 2.20–20.00°. Within- and between-session reliability depended on the task and joint assessed. Errors varied across upper-limb and rotational outcomes, with a transverse-plane shoulder root mean square error of 56.61° reported during cutting. Markerless motion capture shows potential for selected sport and exercise biomechanical assessments and athlete monitoring. Task-specific validation in representative sporting environments, together with assessment of between-session repeatability and measurement error, is needed before measurements are used to detect longitudinal changes or inform clinical, coaching, injury-risk, or return-to-sport decisions.