Real-Time Virtual Try-On Using AI and Augmented Reality for Personalized Fashion Experience

Real-Time Virtual Try-On Using AI and Augmented Reality for Personalized Fashion Experience

Sharvari Dubey, Renuka Randhir, Siddhi Jadhav,Saniya Chopade, Anil S. Londhe

Datta Meghe College of Engineering University of Mumbai

Abstract—The rise of digital fashion commerce has increased the demand for interactive and personalized shopping experi- ences. Virtual Try-On (VTO) technology addresses this need by allowing users to visualize garments on their own images, reducing uncertainty around fit and appearance. This paper presents a deep learning-based VTO system that uses a 2D-to-2D transformation pipeline, leveraging U²-Net for human segmenta- tion and a Segmind-powered diffusion model for realistic garment synthesis. The system includes a lightweight Flask backend, a chatbot interface for collecting user traits such as gender, skin tone, and body type, and a recommendation engine for contextual outfit suggestions. A trend visualization module further supports decision-making by displaying gender-specific clothing trends. Despite being trained on a limited dataset ( 2.5 GB), the system shows improved overlay accuracy and personalization compared to earlier models like VITON. Designed with modularity and scalability in mind, this implementation demonstrates the poten- tial of AI-driven virtual try-on in enhancing online fashion retail.

Index Terms—Virtual Try-On, Human Segmentation, Garment Recommendation, Fashion Technology, Deep Learning, Diffusion Models, Interactive E-Commerce