文献类型：期刊文献

英文题名：The generative adversarial network improved by channel relationship learning mechanisms

作者：Yue, Danyang[1];Luo, Jianxu[1];Li, Hongyi[2]

机构：[1]East China Univ Sci & Technol, Minist Educ, Key Lab Adv Control & Optimizat Chem Proc, Shanghai, Peoples R China;[2]Fudan Univ, Sch Comp Sci, Shanghai, Peoples R China

年份：2021

卷号：454

起止页码：1

外文期刊名：NEUROCOMPUTING

收录：;EI(收录号：20212410501651);WOS:【SCI-EXPANDED(收录号:WOS:000672469900001)】；

基金：The authors gratefully acknowledge the financial supports by the Science and Technology Commission of Shanghai Municipality under Grant No. 19511121203.

语种：英文

外文关键词：Channel relationship; Squeeze-and-Excitation; Dual-attention; Generative Adversarial Networks

摘要：Although recent deep generative models are able to generate high-resolution, diverse natural samples from complex datasets, the generated samples still exist some problems in terms of images structure and detailed texture. In this paper, we propose a novel network architecture-SEDA-GAN that can learn the potential relationship in the dimension of the channel to enhance the generation performance of GAN. The proposed architecture applies Squeeze-and-Excitation(SE) block for feature recalibration to model channel-interdependencies within the GAN feature, and it also incorporates a dual-attention (DA) model with a channel attention mechanism in the GAN framework that can obtain global dependencies between channels. After conducting some comparative experiments on CIFAR and ImageNet datasets by using model BIGGAN as a baseline, our model performance has a certain improvement when evaluating on Frechet Inception Distance(FID) and Inception Score(IS) respectively. (C) 2021 Elsevier B.V. All rights reserved.