Development of an AI Enabled Yoga Posture (Aasans) Prediction System Using Deep Neural Network Model

“Yoga does not just change the way we see things; it transforms the person who sees”.This is a statement that highlights not just suggests the greatness we have in the practice which was once seen as just a hobby - YOGA In this paper, we propose a method to detect correct yoga postures using deep learning techniques. Our approach involves using a dataset of various yoga asanas to train our model to detect any abnormalities between the desired pose and the user's real time pose. By using this approach, we hope to provide a useful tool for individuals who want to learn and practice yoga at home, without the need for expensive trainers or yoga centers. AI Trainer will provide an interactive platform for beginners to learn about the fundamentals of yoga and asanas, as well as provide a recommendation system for users based on their postures. With the help of MoveNet, the human pose estimation can be done fast and can be easily deployed considering and keeping the inference times as low as possible during that. The outcome of this is a model which can implement and can achieve accurate key points for estimating a wide variety of gestures, movements, environments, and setup of hardware. The proposed model is giving an accuracy of 0.9929% and we have prepared an application is built using react javascript in which the pose estimation model is integrated. This application will help the user to easily maneuver and interact with the pose estimation model.