Deep Fabric Print Generation for Fashion

Extracting feelings from the mood board is a necessary step for designers to create new illustrations. However, the process of feature extraction from the mood board and obtaining a new print is both time-consuming and creatively challenging. A style transfer process based on a deep neural network is described here to solve this problem by creatively generating an entirely new print. In this paper, the AAST (Aesthetic-Aware Image Transfer) method is utilized to extract the color and texture features from the mood board and transfer it into the content images. This algorithm can generate numerous results, fusing content and mood board at high resolution. Unlike other previous style transfer methods, which transfer the style and color features simultaneously, a novel style transfer block, i.e., Aesthetic-Aware Image Transfer divides the color and texture features into two independent feature maps, thus creating a preferred print. Specifically, colors and textures extracted from multi-images are two instinct paths in the training section from which the generator obtains and transfers color and texture separately. Compared with the filters in Illustrator and Photoshop, this method can extract the feelings of arbitrary images as filters and generate prints from them. Extensive experimentation has demonstrated the feasibility and effectiveness of the method.