“AI FIT” Virtual Gym With Real-Time Exercise Form Detection

The growing importance of health and fitness in the wake of the COVID-19 pandemic has caused many people to focus on their well-being. While the fitness industry has digitalized its services, there still exists a lack of experienced human touch and a gymnasium environment in these systems. The proposed system aims to provide an integrated solution for exercise enthusiasts with all the features that can replicate an offline gym. It utilizes the process of Skeletonization using the MediaPipe library for body key point detection. It further uses a Long-term Recurrent Convolutional Network (LRCN) model composed of the Conv 2D layer and Long Short-Term Memory (LSTM) layer to detect a person’s exercise form in real-time, analyse it, and display their progress in the Analysis section. With the integration of modules such as a Training Space (for understanding different exercises), Real-time Exercise Form Detection, Analysis and Statistics(to provide a detailed analysis of your workout), and Blog Space(for the community to interact), this system provides a solution for exercise enthusiasts with all the features that can replicate an offline gym. The LRCN model trained on skeletonized image frames showed an accuracy of 96.49% on the train set and 95.50% on the test set for real-time form detection.