Research on Clothing Sketch Conversion Based on Generative Adversarial Networks

The image conversion of clothing sketches has enormous application value in fields such as visual editing, clothing design, and human-computer interaction. At present, most of the objects converted from sketches are relatively simple shapes, such as faces, human postures, animal postures, shoes, and bags. At present, sketch conversion-based work is difficult to handle images with complex texture structures, making it challenging to solve the task of image conversion for clothing sketches. Additionally, the lack of clothing datasets is also a difficult problem to solve. In order to solve the above problems, this paper proposes using an edge detection algorithm to generate clothing simulation sketches in reverse, solving the problem of missing clothing sketch datasets. Then, combined with generative adversarial networks in the field of artificial intelligence, the Cloth Sketch GAN (CSG) network framework is proposed. It improves the generator and discriminator on the basis of the pix2pix network, and enhances the image generation effect by adding an attention module in the generator of the U-Net network structure. Two scales of discriminators are used to discriminate the images separately, and multiple layers of feature extraction are introduced based on the two scales of discriminators. The design of the feature matching loss function and perception loss function is added to make the generated images more realistic and clear, thus completing the conversion from clothing sketch to clothing image. Applied research. Finally, experiments were conducted on the clothing dataset constructed in this article, and the results showed that Cloth Sketch GAN had a significant improvement in the performance of generating clothing images. Moreover, with the improved network proposed in this article, the task of converting clothing sketches to real clothing images can be effectively accomplished on the clothing dataset.