HUMAN POSE ESTIMATION USING MACHINE LEARNING IN PYTHON

Abstract

This study presents a Virtual Gym Trainer system that utilizes cutting-edge technologies,
including MobileNetV3 Transfer Learning and a Custom Convolutional Neural Network (CNN), to
enhance yoga practice through precise pose detection. The MobileNetV3 Transfer Learning model is ????inetuned
on a diverse dataset of yoga poses, enabling real-time and accurate recognition of various postures.
This lightweight model is optimized for deployment on mobile devices, ensuring accessibility and
convenience for users. Complementing the MobileNetV3 Transfer Learning approach, the Custom CNN is
speci????ically designed to capture intricate details of yoga postures, enhancing identi????ication and
classi????ication accuracy even in challenging conditions. Trained on a comprehensive dataset, the Custom
CNN further improves the Virtual Gym Trainer system's effectiveness in providing personalized feedback
and guidance during yoga sessions. The integration of MobileNetV3 Transfer Learning and Custom CNN
in our Virtual Gym Trainer represents a signi????icant advancement in ????itness technology, offering users a
seamless and interactive yoga experience. This system not only improves form and alignment but also
tracks progress and provides tailored recommendations, making yoga practice more engaging, effective,
and accessible to individuals seeking optimal health and wellness