摘要
探讨了生成对抗网络(GAN)中的LayoutGAN模型,分析其生成器、基于关系判别器和线框渲染判别器在创意图像生成中的潜力,通过实验对比基于关系判别器的LayoutGAN与采用线框渲染判别器的LayoutGAN在MNIST数据集上的表现。结果显示,线框渲染判别器在生成创意图像方面展现出了显著优势,其生成的图像细节更丰富,更具创造性和艺术性。这一结果表明线框渲染判别器在图像生成中具有卓越的性能,可为计算机平面设计提供新的可能性和方向。
The study discusses the LayoutGAN model in generative adversative network(GAN);analyzes the potential of its generator,relation-based discriminator and wire-frame rendering discriminator in creative image generation;and compares the performance of relation-based discriminator LayoutGAN and wire-frame rendering discriminator on MNIST dataset through experiments.The results show that wire-frame render discriminators show significant advantages in generating creative images,which are richer in detail,more creative and artistic.This result shows that the wire-frame rendering discriminator has excellent performance in image generation,and it can provide new possibilities and directions for computer graphic design.
作者
邢磊
马玉冰
Xing Lei;Ma Yubing(Shandong Huayu University of Technology,Dezhou 253034,China)
出处
《黑龙江科学》
2024年第16期113-115,119,共4页
Heilongjiang Science
关键词
生成对抗网络
高校计算机
平面设计
创意图像生成
Generate adversarial network
University computer
Graphic design
Creative image generation