Research on image synthesis of fabric replacement in suit customization
Qiubo Huang, Dachuan Ma, Yanting Zhang, Guangwei Xu · 2022
With the development of the Internet, online clothing customization has become a trend. Fabric replacement is an important part of clothing customization. If the high-reso-lution image of clothes can be rendered after replacing a speci-fied fabric instantly online, it can greatly enhance the consum-er's purchase. To address this need, this paper proposes an image synthesis method based on pix2pixHD network for suit fab-ric replacement. In this paper, we modify the generator to sepa-rate texture generation and color generation of suit images to reduce the interference of color information to texture infor-mation generation. Meanwhile, we generate and discriminate the image in brightness channel and color channel, and intro-duce additional texture loss and color loss in the discriminator to guide the generator to generate images with higher quality and restore the real fabric texture and color. The effect of a se-ries of different loss functions on the quality of image synthesis is also compared and analyzed. The experimental results show that the suit image synthesized by this method has high resolution and good visual impression, and can restore the genuine texture of the fabric well, which meets the needs of suit customization system.