Shapesense: Body Type Classification Using MediaPipe Pose Landmark Detection Model
Keywords:
Body shape detection, Deep learning, Digital image analysis, Error correction algorithms, Personalized health tracking, Personalized services, Posture analysis, Shoulder-to-Hip Ratio (SHR), Waist-to-Hip Ratio (WHR)Abstract
This project implements MediaPipe solutions to classify body types while overcoming various challenges like body shapes, clothing, and lighting conditions. For precise proportional analysis, MediaPipe’s pre-trained models and pose estimation options are used to scan key body landmarks. The system successfully distinguishes body types (pear, apple, rectangle, hourglass) while maintaining user privacy. By offering actionable insights across domains, from fashion and fitness to health, it enhances user experiences with personalized recommendations and contextually relevant guidance. More safeguards and fallback methods will be built to strengthen this methodology. This encompasses but is not limited to novel error correction procedures, redundancy layers, and the implementation of rigorous fail-safes to maintain precision even in sub-optimal scenarios.
