Enhanced Yoga Pose Classification Using Advanced CNN Architectures and MediaPipe

An advanced yoga poses classification system based on the integration of advanced Convolutional Neural Network architectures and MediaPipe. Yoga is characterized by a variety of poses which are beneficial for health hence it has to be classified well to avoid poses being done incorrectly or injures occurring. Conventional classification techniques that are based on visual observation or simple machine learning aren't accurate or dependable. The new system employs the use of MediaPipe for the real-time image capturing, detecting pose key points from the video frames that are then sorted out through a deep CNN for classification. The system's approach takes into consideration the merits of both technologies with an intention of achieving high accuracy and performance in real time. The structure and experimental design of the system are presented along with results proving its practicality and usefulness for personal training, online teaching, and rehabilitation. The discussion includes the elaborated technique and its applications and a plan for the future works is proposed.