YOGA POSE OPTIMIZATION USING DEEP LEARNING TECHNIQUES

The second phase, Real-Time Feedback, enables immediate pose correction by comparing a user’s posture with thecorrect form, ensuring a 99.88% accuracy rate in assessment. This instant responsiveness gives it a significant edge overtraditional models like Pose Net CNN, which often lack precision and speed. The Y_PN_MSSD model is a cutting-edgesolution that enhances virtual yoga instruction by combining Pose-Net and Mobile Net SSD (TF Lite Move Net) forprecise, real-time pose assessment. One of the major challenges in online yoga training is ensuring that practitionersreceive accurate, real-time feedback to prevent incorrect posture, discomfort, or injury. Traditional learning methods,such as instructional videos and articles, lack immediate correction, making live teacher feedback essential.Y_PN_MSSD addresses this gap by using advanced machine learning and computer vision to track human movementwith exceptional accuracy. The model operates through three key phases. The first phase, Data Collection & Training,involves gathering yoga pose data from diverse users and integrating it with an extensive open database of seven keyyoga postures. This allows the system to recognize different body structures and skill levels, ensuring a more inclusiveand adaptable learning experience. By analyzing anatomical key points, the model distinguishes between correct andincorrect alignments, refining its ability to provide accurate feedback.