FitVTON: Fit-aware Virtual Try-On via Body-Garment Size Control
Researchers have developed FitVTON, a new virtual try-on model that prioritizes realistic garment fit over mere texture preservation. Unlike previous methods that treated try-on as 2D inpainting, FitVTON encodes garment and body size information through structured text prompts and learns from simulated data. The model includes auxiliary heads for predicting garment and body masks, and a texture rectification stage to enhance realism. Experiments show FitVTON achieves superior fitting fidelity, sizing accuracy, and shape preservation compared to existing state-of-the-art methods. AI
IMPACT Enhances realism in virtual try-on applications by focusing on accurate garment fit and sizing.