ClothMeUp: AI-powered virtual try-on

Abstract

The rapid growth of online fashion retail underscores the demand for effective Virtual Try-On (VTON) solutions balancing realistic garment visualization and intuitive user interaction. Existing systems frequently suffer from low-resolution outputs, garment detail loss, and cumbersome workflows, negatively impacting consumer confidence. This report introduces ClothMeUp, an innovative AI-driven VTON framework that overcomes these limitations with three core advancements: (1) an integrated pipeline combining IDM-VTON, a diffusion-based model achieving accurate garment alignment, and Real-ESRGAN for artifact-free, high-resolution (2K) upscaling; (2) a Chrome extension with context-aware right-click garment selection and an unobtrusive side- panel interface; and (3) dynamic cropping techniques enhancing inpainting fidelity for detailed edits. ClothMeUp utilizes a modular backend architecture with FastAPI for processing, Supabase for authentication, and Docker for scalable deployment. Empirical evaluation involving 29 participants indicated substantial improvements in user satisfaction; 90% reported increased purchase confidence due to enhanced visual realism (average interface rating: 4.7/5). Real-ESRGAN-processed images achieved higher preference scores (4.3/5) compared to baselines, affirming superior detail preservation. The Chrome extension streamlined the try-on workflow, aligning closely with user expectations.