GAN Based Model for Virtual Try On of Clothes

E-commerce platforms focused on fashion, imagebased virtual try-on solutions have shown considerable potential for enhancing the user experience and increasing customer satisfaction. The $2 D$ solutions that exist fail to provide a complete satisfactory experience as they lack the capability to take into consideration customer-specific postures and modifications. They rely majorly on generative models due to which reconstructionof high-quality images of the clothing is difficult. The clothing in $3 D V T$ solutions is created as $3 D$ models, which are thendraped on $3 D$ human avatars with various body types. A huge catalog of digitized $3 D$ garment and clothing models is a crucial requirement for a $3 D$ VT system, but this requirement requires expensive multiview capture/scan equipment and cannot scaleup for big catalog volumes, especially in the current fast-fashion settings. Hence, we propose a video virtual Try-On model which is based on fully learnable characteristics preserving virtual try on network $C P$ VTON [3]. It first forms a TSP for transforming clothes to fit the target body shape via Geometric Matching Module.Then use a GAN based model for pose transfer.The accuracy resulting from such a model could help in decreasing the waiting time in traditional stores, and also help in decreasing sanitizing costs.