Real-Time Yoga Pose Detection and Classification Using YOLO and Computer Vision

This research introduces a real-time yoga pose classification system, aimed at enhancing self-guided yoga practice and promoting accessibility across web and mobile platforms. Key Features: Pose Estimation: Utilizes MediaPipe to extract key body landmarks and skeleton data. Machine Learning Models: Includes CNN, SVM, and MLP for accurate pose classification with up to 94% accuracy. Feature Extraction: Computes joint angles and distances for normalization and structural pattern recognition. Real-Time Feedback: Provides instant posture corrections using skeleton overlays, text-based guidance, and color-coded visual cues. Accessibility: Deployable on web browsers and mobile devices, supporting offline functionality. Privacy-Conscious Design: Preserves user anonymity through skeleton-based recognition. Technologies Used: MediaPipe, TensorFlow, OpenCV Python for backend development React.js for user interface Android Studio for mobile application Benefits: Promotes safer yoga practices with instructor-like guidance. Encourages personalized and interactive self-improvement. Bridges traditional fitness practices with cutting-edge AI technology.