PhysioNet: Vision based Human Exercise Action Recognition leveraging Transfer Learning

Technology-driven solutions are increasingly important tools for those who want to maintain an active lifestyle in a time when personal health and fitness are given more attention than ever before. This study offers a fresh solution to the problem of workout scoring and identification. The main goal is to develop an automated system that can recognize and evaluate a wide variety of workouts done in video recordings with an emphasis on improving exercise form, posture, and safety. To accomplish reliable exercise identification, the suggested system makes use of Convolutional Neural Networks (CNNs), YoLo and transfer learning. The model is given access to a large amount of visual data thanks to the collection and preprocessing of an extensive dataset of workout videos. The customized CNN architecture provides insights into the feature extraction and learning processes, in addition to visualizations. The creation of an automated exercise detection and scoring system that tackles the issues of exercise diversity, real-time processing, accuracy, generalization, and model interpretability is one of the research's major accomplishments. This initiative provides users with quick feedback to enhance the quality and safety of their exercises by fusing cutting-edge deep learning algorithms with real-world applications in fitness monitoring. The outcomes show a notable improvement in activity identification accuracy, supporting the system's potential to assist people in leading healthier, more active lives. This study advances the nexus between technology and fitness by providing new opportunities for personalized fitness tracking, real-time exercise coaching, and more extensive uses in health and wellbeing.