Applying Pose-Guided Deep Learning for Real--Time Virtual Try-On

Abstract

This paper presents a real-time virtual try-on and size estimation system that allows users to preview garments directly on their live webcam feed, without generating 3D avatars or synthetic models. The goal is to replicate a mirror- like try-on experience on a web platform using computer vision techniques. The system integrates real-time body pose detection, size estimation from key points, garment overlay alignment, and a personalized virtual closet. The proposed approach ensures efficient rendering, user privacy, and compatibility with standard consumer devices. Experimental results demonstrate that the system achieves stable garment alignment and accurate body measurement estimation with a mean error of 3—4 cm. This work provides a practical solution for improving online shopping experiences through non-intrusive, real-time visualization.