Determining the accuracy of sports training using computer vision and deep learning techniques

A systematic analysis was made on the accuracy of sports activities in terms of sports biomechanics within the scope of machine learning algorithms of human pose estimation data obtained from real-time images with the MediaPipe Pose Estimation model, which was developed based on deep learning. The study examined how the MediaPipe Pose Estimation model predicts biomechanical parameters in different sports activities based on body position. The performance of the model was evaluated by comparing real-time camera data with training using various machine learning algorithms such as regression and classification. In addition, this study is a pioneering research on the applicability of computer vision assisted deep learning techniques in sports training and pose estimation. By training the data obtained with the human pose estimation model of Mediapipe with machine learning algorithms, it has been turned into an application that can be used to accurately track the positions of the athletes. This thesis abstract examines the impact of data from real-time images on the accuracy of sports activities using Medipipe's human pose estimation model. It will also guide how this tool can be used to improve the performance of athletes. Efforts have been made to demonstrate the suitability and efficacy of the MediaPipe Exposure Estimation model as a valuable tool in the field of sports biomechanics.