Comparison of Lower-Limb Joint Angles Obtained From Single-View Markerless and Marker-Based Three-Dimensional Motion Capture Systems Across Camera Locations

This study compared lower-limb joint angles obtained from single-view markerless three-dimensional (3-D) motion capture (MoCap) systems with those obtained from a marker-based MoCap system across different camera locations. Three markerless MoCap systems, MediaPipe, PAFUSE, and RTMW3D, were used, and the accuracy and agreement of joint angles estimated during range-of-motion (ROM) tasks of the hip, knee, and ankle joints were evaluated using statistical analyses. The results showed that joint-angle accuracy varied substantially depending on the markerless MoCap system, camera location, joint, and movement type. RTMW3D generally demonstrated superior performance for hip and knee flexion–extension, whereas PAFUSE demonstrated relatively better performance in hip rotation. In contrast, MediaPipe showed limited agreement with the marker-based MoCap system under most conditions. Agreement analyses further indicated that favorable performance was observed only for specific combinations of joint motion, model, and camera location. These findings suggest that single-view markerless MoCap systems should not be applied uniformly across all lower-limb ROM tasks. However, when the model and camera location are strategically selected, single-view markerless MoCap may provide useful joint-specific ROM information in settings where marker-based MoCap systems are impractical.