Method of Campus Planning Automatic Generation Based On Gans and Diffusion Models: Take The Middle School Campus As An Example

Conventional campus design techniques frequently lack efficiency and have difficulty incorporating the performance of green buildings.  Although previous studies have made some progress in using Generative Adversarial Networks (GANs) for campus layout creation, there are still issues such as unstable training, limited diversity in results, and inadequate image quality. To overcome these limitations, this paper presents an innovative approach that integrates GANs with Diffusion Models, termed GAN-SD. The experimental results demonstrate that our proposed method offers significant advantages. It not only diversifies the outcomes of campus layout generation but also significantly reduces the training duration compared to using GANs alone. Additionally, it enhances image quality and stabilizes the generation performance when compared to the independent use of Diffusion Models. This method provides a valuable reference for the rapid design of campus layout.