LayoutGAN for Automated Layout Design in Graphic Design: An Application of Generative Adversarial Networks

The current field of visual communication design is particularly influenced by artificial intelligence, with automatic processing applications for images and color schemes having emerged, and a number of design models for intelligent layout have been proposed. However, the statistical learning framework model will face the limitation of black box, the process of graphic design cannot be separated from the real-time participation of designers, so intelligent graphic design is still essentially a semi intelligent system. In this paper, LayoutGAN is used to study the layout generation problem in graphic design. The network structure is designed in accordance with the idea of LayoutGAN. Then experiments are conducted on MINST dataset to verify the effectiveness of the network structure. The experimental results show that the wireframe rendering discriminator is better than the relationship-based discriminator, and then only the LayoutGAN with wireframe rendering discriminator is used to experiment on the room floor plan dataset, and finally the automatic floor plan generation is realized.