VTON 360: High-Fidelity Virtual Try-On from Any Viewing Direction
This addresses the problem of realistic digital clothing visualization for e-commerce and fashion design, representing an incremental advancement in 3D virtual try-on technology.
The paper tackles the challenge of achieving high-fidelity virtual try-on that supports rendering from any viewing direction by proposing VTON 360, a 3D method that extends 2D VTON models with multi-view inputs and novel techniques, resulting in effective performance demonstrated on large-scale real datasets and e-commerce images.
Virtual Try-On (VTON) is a transformative technology in e-commerce and fashion design, enabling realistic digital visualization of clothing on individuals. In this work, we propose VTON 360, a novel 3D VTON method that addresses the open challenge of achieving high-fidelity VTON that supports any-view rendering. Specifically, we leverage the equivalence between a 3D model and its rendered multi-view 2D images, and reformulate 3D VTON as an extension of 2D VTON that ensures 3D consistent results across multiple views. To achieve this, we extend 2D VTON models to include multi-view garments and clothing-agnostic human body images as input, and propose several novel techniques to enhance them, including: i) a pseudo-3D pose representation using normal maps derived from the SMPL-X 3D human model, ii) a multi-view spatial attention mechanism that models the correlations between features from different viewing angles, and iii) a multi-view CLIP embedding that enhances the garment CLIP features used in 2D VTON with camera information. Extensive experiments on large-scale real datasets and clothing images from e-commerce platforms demonstrate the effectiveness of our approach. Project page: https://scnuhealthy.github.io/VTON360.