Yoga Pose Detection and Correction

Yoga Pose Detection and Correction

S.C. Sethi1, Sampada Bhosale2, Prachi Dawkhar3, Mitali Rote4

1Assistant Professor, Department of Computer Engineering, NBNSTIC, Pune, India

2,3,4 Student, Department of Computer Engineering, NBNSTIC, Pune, India

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - Yoga is a practice that connects body, breath, and mind. Yoga is extremely popular due to its numerous mental and physical health benefits. However, it is important to perform yoga poses correctly to maximize the benefits of yoga and prevent injury. To address this requirement, a yoga pose estimation and correction model was developed using real-time video input and pose detection and correction algorithms. Yoga Pose Detection and Correction is an innovative project that uses computer vision and artificial intelligence to improve yoga practice. The purpose of this research is to develop deep learning-based technology to estimate the correct posture of the practitioner and make the task easier. To implement this strategy, the study used his four different deep learning architectures: MediaPipe, OpenPose, PoseNet, and EpipolarPose. The main components of the project include advanced video processing algorithms and pose recognition models. CNN is a network architecture for deep learning algorithms and is used for image recognition. Comparing the estimation accuracy of all architectures in the study, the MediaPipe architecture provides the highest estimation accuracy.

Key Words:  Artificial intelligence, deep learning, machine learning techniques, pose estimation techniques, skeleton and yoga.