YogMaster: Detection and Correction of Yoga Postures Using Augmented Reality
Omkar Parekh, Siddhesh Kondaskar, Yashar Shaikh, Neha Shende, Nilesh T. Deotale · 2023
Yoga is becoming increasingly popular due to its positive effects on the body and mind. Although many individuals engage in yoga practices, it is important to use proper instruction from an experienced instructor to avoid potential health issues. The proposed solution aims to address this issue by developing a web-based application that uses a camera to capture the user's yoga poses. The TensorFlow MoveNet model then processes the captured images which identifies 17 key body points. This information is then fed to a Neural Network model for Classification which verifies the accuracy of the pose and if necessary, provides correction through a Correction model. The user will be alerted to any corrections through both visual and audio cues. The results demonstrate that the model has 99.47% accuracy.