Enhancing Squat Safety and Performance with Computer Vision and Deep Learning Model

Exercise is good for one's health and fitness, however, it can also be ineffective and potentially dangerous if done improperly by the user. Exercise mistakes are made when users don't use Correct form or pose. Poor squat posture for example can damage the knee health for a long period. Thus, maintaining a healthy squat posture is crucial for a person to workout effectively. This project introduces the use of computer vision to develop a model using the MediaPipe Pose, that recognizes and classifies the best squat posture and provides recommendations on how users can improve the form. The data is collected from exercise videos of correct squat posture by a professional coach. The developed algorithm successfully classifies correct posture with overall accuracy of 85%.