Virtual Try-On-Enhancing Fashion Exploration for Gen-Z

Deep Learning & Generative Models

Virtual Try-On (VTON) is a rapidly emerging technology designed to digitally visualize how garments might appear when worn by individuals, thus transforming traditional fashion and retail experiences. This survey paper meticulously explores various state-of-he-art techniques employed in VTON systems, primarily focusing on image-based, 3D-based, and hybrid approaches. Initially, the paper introduces fundamental concepts of virtual try-on systems, tracing their historical progression and relevance within modern e-commerce and fashion industries. It systematically categorizes methodologies into distinct groups, highlighting pioneering approaches such as warping methods, generative adversarial networks (GAN), and advanced 3D garment simulation techniques. The paper further emphasizes crucial technologies and datasets pivotal to VTON advancements, including GANs, transformers, diffusion models, and benchmarking datasets like DeepFashion and VITON. In addressing existing limitations, this survey underscores critical challenges such as achieving photorealistic rendering, effectively handling occlusions and diverse human poses, ensuring real-time processing, and generalizing across various fabric textures and garment styles. Moreover, recent innovations and their implications on commercial and real-time applications are thoroughly discussed. Finally, the paper delineates future research directions aimed at enhancing system realism, scalability, personalization, and integration of emerging generative AI methodologies, highlighting the significant potential for continued innovation and application in the digital retail landscape.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.