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. 多模态人工智能系统全国重点实验室
  5. 三维可视计算
Pixel-wise Dense Detector for Image Inpainting

文献类型：期刊论文

作者Zhang, Ruisong3,4; Quan, Weize3,4; Wu, Baoyuan1,2; Li, Zhifeng5; Yan, Dong-Ming3,4
刊名COMPUTER GRAPHICS FORUM
出版日期2020-10-01
卷号39期号:7页码:471-482
ISSN号0167-7055
DOI10.1111/cgf.14160
通讯作者Yan, Dong-Ming()
英文摘要Recent GAN-based image inpainting approaches adopt an average strategy to discriminate the generated image and output a scalar, which inevitably lose the position information of visual artifacts. Moreover, the adversarial loss and reconstruction loss (e.g., l(1) loss) are combined with tradeoff weights, which are also difficult to tune. In this paper, we propose a novel detection-based generative framework for image inpainting, which adopts the min-max strategy in an adversarial process. The generator follows an encoder-decoder architecture to fill the missing regions, and the detector using weakly supervised learning localizes the position of artifacts in a pixel-wise manner. Such position information makes the generator pay attention to artifacts and further enhance them. More importantly, we explicitly insert the output of the detector into the reconstruction loss with a weighting criterion, which balances the weight of the adversarial loss and reconstruction loss automatically rather than manual operation. Experiments on multiple public datasets show the superior performance of the proposed framework. The source code is available at https://github.com/Evergrow/GDN_Inpainting.
资助项目National Key R&D Program of China[2019YFB2204104] ; National Natural Science Foundation of China[61772523] ; Beijing Natural Science Foundation[L182059] ; Tencent AI Lab Rhino-Bird Focused Research Program[JR202023] ; Open Research Fund Program of State key Laboratory of Hydroscience and Engineering, Tsinghua University[sklhse-2020-D-07]
WOS研究方向Computer Science
语种英语
WOS记录号WOS:000594502700040
出版者WILEY
资助机构National Key R&D Program of China ; National Natural Science Foundation of China ; Beijing Natural Science Foundation ; Tencent AI Lab Rhino-Bird Focused Research Program ; Open Research Fund Program of State key Laboratory of Hydroscience and Engineering, Tsinghua University
源URL[http://ir.ia.ac.cn/handle/173211/43222]  
专题模式识别国家重点实验室_三维可视计算
通讯作者Yan, Dong-Ming
作者单位1.Shenzhen Res Inst Big Data, Secure Comp Lab Big Data, Shenzhen, Peoples R China
2.Chinese Univ Hong Kong, Sch Data Sci, Shenzhen, Peoples R China
3.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing 100049, Peoples R China
4.Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
5.Tencent Lab, Shenzhen, Peoples R China
推荐引用方式
GB/T 7714		 Zhang, Ruisong,Quan, Weize,Wu, Baoyuan,et al. Pixel-wise Dense Detector for Image Inpainting[J]. COMPUTER GRAPHICS FORUM,2020,39(7):471-482.
APA Zhang, Ruisong,Quan, Weize,Wu, Baoyuan,Li, Zhifeng,&Yan, Dong-Ming.(2020).Pixel-wise Dense Detector for Image Inpainting.COMPUTER GRAPHICS FORUM,39(7),471-482.
MLA Zhang, Ruisong,et al."Pixel-wise Dense Detector for Image Inpainting".COMPUTER GRAPHICS FORUM 39.7(2020):471-482.

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。