中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
首页 机构 成果 学者
  
  1. CAS IR Grid
  2. 自动化研究所
  3. 中国科学院自动化研究所
  4. 智能制造技术与系统研究中心
  5. 多维数据分析团队
Asymmetric CycleGAN for image-to-image translations with uneven complexities

文献类型：期刊论文

作者Dou, Hao1,3; Chen, Chen1,3; Hu, Xiyuan4; Jia, Libang1,3; Peng, Silong1,2,3
刊名NEUROCOMPUTING
出版日期2020-11-20
卷号415页码:114-122
关键词Unpaired Image Translation CycleGAN Asymmetric Translation Average Image Entropy Edge-retain Prior
ISSN号0925-2312
DOI10.1016/j.neucom.2020.07.044
通讯作者Hu, Xiyuan(xiyuan.hu@foxmail.com)
英文摘要CycleGAN is one of the famous and basic methods for unpaired image-to-image translation tasks. Inspired by the experiments of the NIR-RGB translation, which is a kind of translation where images are translated from simple to complex or vice versa, we concluded the definition of asymmetric translation task. Because of the complexity difference between two domains, the complexity inequality in bidirectional translations is significant. We analyzed and witnessed the limitation of the original CycleGAN in asymmetric translation tasks and proposed an Asymmetric CycleGAN model with generators of unequal sizes to adapt to the asymmetric need in asymmetric translations. An empirical metric was also given to determine the asymmetric task from the aspect of image entropy and could be treated as the auxiliary guidance to design the asymmetric generators. Besides, the edge-retain loss between the input and the generated images was introduced to enhance the structural visual quality. Residual-block-net based and U-net based generators were both applied here to verify the Asymmetric CycleGAN. The performance of different depth of generators for Asymmetric CycleGAN was also discussed on the basis of experiments. The qualitative visual evaluation demonstrated that our model had achieved great improvements compared to original CycleGAN. (C) 2020 Elsevier B.V. All rights reserved.
资助项目National Nature Science Foundation of China[61906194] ; National Nature Science Foundation of China[61571438] ; Liaoning Collaboration Innovation Center For CSLE
WOS研究方向Computer Science
语种英语
WOS记录号WOS:000579808700011
出版者ELSEVIER
资助机构National Nature Science Foundation of China ; Liaoning Collaboration Innovation Center For CSLE
源URL[http://ir.ia.ac.cn/handle/173211/42122]  
专题自动化研究所_智能制造技术与系统研究中心_多维数据分析团队
通讯作者Hu, Xiyuan
作者单位1.Univ Chinese Acad Sci, Beijing, Peoples R China
2.Beijing Visyst Co Ltd, Beijing, Peoples R China
3.Chinese Acad Sci, Inst Automat, Beijing, Peoples R China
4.Nanjing Univ Sci & Technol, Sch Comp Sci & Engn, Nanjing, Peoples R China
推荐引用方式
GB/T 7714		 Dou, Hao,Chen, Chen,Hu, Xiyuan,et al. Asymmetric CycleGAN for image-to-image translations with uneven complexities[J]. NEUROCOMPUTING,2020,415:114-122.
APA Dou, Hao,Chen, Chen,Hu, Xiyuan,Jia, Libang,&Peng, Silong.(2020).Asymmetric CycleGAN for image-to-image translations with uneven complexities.NEUROCOMPUTING,415,114-122.
MLA Dou, Hao,et al."Asymmetric CycleGAN for image-to-image translations with uneven complexities".NEUROCOMPUTING 415(2020):114-122.

入库方式： OAI收割

来源：自动化研究所

浏览0
下载0
收藏0
其他版本

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。