YogaWise: Enhancing Yoga with Intelligent Real Time Tracking using TensorFlow MoveNet
Amey Parle, Rugved Shinde, Rahul Chougule, Shruti Agrawal · 2024
This work proposes a novel method that uses image processing and ensemble learning to analyze yoga practices thoroughly, in response to the growing need for accurate assessment in the field of yoga. The main goal is to help users execute yoga postures efficiently and in real time. With the use of the MoveNet model, the system was able to obtain a testing accuracy of 98% as well as strong precision, recall, and F1 Score metrics, demonstrating its ability to recognize and classify a variety of yoga poses with a low number of false positives and negatives. Methodologically, a large-scale dataset was gathered, cleaned up, and used to train a deep learning model that was then smoothly incorporated into an easy-to-use online application. The proposed system offers pre-designed yoga sessions, different types of pain relief routines, and individual pose exploration through an intuitive user interface. It also offers real-time feedback, posture prediction all of which improve users’ yoga experiences and help them move closer to their health and fitness objectives.