A Deep Learning Based Posture Correction System

A computerized system developed using machine learning to monitor correct posture for an exercise being performed. The posture for exercise is important for the physical benefit of it. The system for detecting exercise is implemented with a combination of a Convolutional Neural Network (CNN) model, Long Short-Term Memory(LSTM) model and Recurrent Neural Networks(RNN) model which is trained on a different set of exercise dataset. The machine learning models identify the exercise type which is being performed in the input source. The posture of the exercise is being estimated using the OpenPose's BlazePose library. The correctness of the exercise is being evaluated by comparing the body joint angles with the predefined pose threshold joint angles and real-time feedback is provided. Additionally, if the posture of the exercise is found correct the repetitions are being counted to ensure accuracy and consistency in the workouts. The system helps in enhancing the exercise performance and minimize the chances of injury by providing precise posture analysis and form correction during exercise routines.