In this paper, we propose a topology optimization method based on Generative Adversarial Networks (GANs), which considers a two-dimensional topology as an image generation problem. The quality of the generated image is improved by a weighted loss function of the adversarial loss and the mean square error. Analyze the parametric constraint solving method, describe the geometric elements and their constraint relationships through geometric constraint graph (GCG), define the concepts of degrees of freedom and constraints, classify the constraint problems, and clarify the evaluation criteria of the algorithms. Analyze the generative graphic design process and emphasize the central role of algorithms in graphic generation. Verify the visual quality and creative advantages of the generated graphics of this paper’s method by means of digital graphic design practice and comparative evaluation of effects. The results show that the overall visual effect of the generated graphics of this paper’s method has an aesthetic score between 2 and 5. The scores are higher than those of the comparison method in the seven indicators of the detailed layout performance of the generated graphics. When the embedded coded information reaches 205000 bits, the average values of PSNR, SSIM, and LPIPS are 26.20334, 0.98112, and 0.00424, respectively, with good quality of visual perception. And the generated graphs are more non-orderly and have excellent creative effect.
