AI-Based System for Real-Time Yoga Pose Detection and Correction
Khushi Sikarwar, Nookala Venu, Aditya Dubey, Dhananjay Bisen · 2025
Although yoga is recognized for its positive effects on both physical and mental health, injuries can occur from improper positioning while practicing. To tackle this issue, I introduce a real-time yoga pose detection and correction system based on AI. This system employs a convolutional neural network (CNN) for categorizing five popular yoga poses Downward Dog, Goddess, Plank, Tree, and Warrior II. An available dataset was utilized and divided into 80% for training and 20% for testing. Methods like resizing, rotating, magnifying, and mirroring images were utilized to improve the model's effectiveness. The camera-based system monitors body movements and gives instant feedback on pose accuracy, being a helpful tool for yoga enthusiasts. The accuracy of the model on the validation set was 87%, with real-time testing confirming effective pose correction. Future efforts will concentrate on enlarging the dataset and backing up more intricate poses.