基于生成对抗网络的织物图案生成方法  

Fabric Pattern Generation Based on Generative Adversarial Network

作　　者：李锋[1] 邵健 LI Feng;SHAO Jian(College of Computer Science and Technology,Donghua University,Shanghai 201620,China)

机构地区：[1]东华大学计算机科学与技术学院,上海201620

出　　处：《智能计算机与应用》2020年第10期32-36,43,共6页Intelligent Computer and Applications

基　　金：国家重点研发计划项目(2017YFB0309800)。

摘　　要：织物图案设计作为一种传统工艺,在当代艺术的环境下需要更加多元化的设计思路。本文基于深度卷积生成对抗网络(DCGAN)进行织物图案设计,为了消除DCGAN中的反卷积操作产生的棋盘效应,使用缩放卷积对原模型生成器中的反卷积操作进行优化,并在缩放卷积的上采样过程中分别采用最近邻插值和双线性插值,同时与原始DCGAN模型的生成样本进行实验对比。实验结果表明,使用缩放卷积改进的DCGAN可以更有效的提升生成图像的质量。As a traditional craft,fabric pattern design requires more diversified design ideas in the context of modern and contemporary art. This paper designs fabric patterns based on the Deep Convolution Generated Adver-sarial Network( DCGAN). In order to eliminate the checkerboard artifacts produced by deconvolution op-eration in DCGAN,this paper uses resize convolution to optimize the deconvolution in the generator of DCGAN. In the up-sampling process of resize convolution,nearest neighbor interpolation and bilinear interpolation are used respectively,and experimental comparison with the generated images of the original DCGAN is performed. Experiments show that the improved DCGAN using resize convolution can improve the quality of generated images more effectively.

关 键 词：织物图案设计 深度卷积生成对抗网络 棋盘效应 缩放卷积 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]