BeFit—A Real-Time Workout Analyzer

Maintaining one’s physical fitness is of utmost importance. Exercising regularly is very important as it helps improve the quality of life. However, incorrect posture during exercises may lead to severe long-term injuries such as back pain, Tendinitis or even hamstring strains. Hence, this application BeFit is proposed that analyzes the posture of the user performing a particular workout by comparing their workout to the reference image or video provided by the system. The system will analyze the angles between the limbs of the body and compare it to the reference video or image using the Cosine rule. After synchronizing user and reference image or video the system gives a green skeleton if the user posture is correct and a red skeleton if the user posture is incorrect. This model has been achieved using the PoseNet library on Tensorflow. The maximum score that the PoseNet model achieves ranges from 0.92874 to 0.98325 for all the key points. With the help of this model, fitness enthusiasts can perform a particular workout accurately at the comfort of their home without getting injured and with proper guidance.