<![CDATA[The rapid advancement of computer vision (CV) technology has created new opportunities for analyzing and improving motor movement accuracy in competitive sports. This narrative review synthesizes theoretical and empirical evidence on the application of CV-based systems, including markerless motion capture, pose estimation algorithms, and real-time feedback technologies, to enhance athletic performance through motor learning processes. Drawing exclusively on peer-reviewed Scopus-indexed studies, the review examines motor learning theories, the technical foundations of CV systems, empirical evidence of their effectiveness, and key implementation challenges. The literature indicates that markerless CV systems can achieve kinematic accuracy comparable to traditional marker-based motion capture methods, particularly in sagittal-plane movement analysis. Advances in deep learning architectures, including convolutional neural networks (CNNs), graph convolutional networks (GCNs), and transformer-based models, have substantially improved real-time pose estimation accuracy. When integrated with motor learning frameworks such as Optimal Feedback Control Theory and the Optimal Theory of Motor Learning, CV-based feedback systems can enhance movement precision, support biomechanical correction, and facilitate skill acquisition. However, challenges related to occlusion, computational latency, environmental variability, and limited sport-specific datasets remain. Future research should emphasize standardized evaluation protocols, longitudinal validation studies, and ethical frameworks for AI-assisted coaching applications.]]>
