Fashion Design Assistant Network Based on Transformer and Score Matching

Against the backdrop of consumption upgrading, Consumers' needs for personalized clothing customization are growing increasingly. However, fashion design is highly reliant on designers' experience,which severely restricts the development of personalized design. Although existing research has utilized CV algorithms to address issues related to local features of fashion design, little attention has been paid to the overall design. This paper proposes the SWIN - transofrmer - DFTscorematchin g network, aiming to solve the problems of overall clothing structure and color style design. This network transforms the preliminary structure by matching SWIN into a finalized structure,meanwhile generates and corrects the style with the help of DFTscore - matching. The experiments are carried out based on a dataset containing 20,000 samples, and models are built using SWIN - B, VGG - 16, and U - Net. The results show that this network performs well in the overall fashion design, and ablation experiments verify the effectiveness of each module. This research provides new ideas and methods for the development of fashion design assistant networks.