Abstract
Currently, increasing the effectiveness of the training process, reducing the risk of injuries, and personalizing training and education are key challenges in the fields of sports and physical education. In this context, research focused on the automated analysis and evaluation of physical exercises using artificial intelligence (AI) and computer vision technologies is rapidly developing. These technologies enable real-time tracking of human movement and accurate determination of biomechanical parameters. This paper presents the development of an intelligent computer vision – based fitness assistant designed to recognize physical activity, analyze exercise execution techniques, and provide users with real-time feedback. The proposed system is implemented using deep learning methods based on the Ultralytics YOLO architecture and the OpenCV library. The system detects key body landmarks and joints, analyzes their spatial positions, evaluates the correctness of exercise execution, counts repetitions, and determines performance metrics. During the study, a dataset of video recordings containing various exercise movements was collected, preprocessed, and annotated. Deep neural network models were trained, and their accuracy and performance were evaluated. In addition, biomechanical modeling was applied to calculate loads on individual body segments and to generate corrective recommendations aimed at improving movement quality. Experimental results demonstrate that the proposed system achieves high accuracy in exercise recognition and operates efficiently in real time. The developed software solution can be used by individuals training at home, athletes, coaches, and physical education professionals as an additional tool for monitoring and optimizing the training process. The research findings confirm the promising potential for widespread application of artificial intelligence and computer vision technologies in the fields of fitness and sports.