Anisotropic Convolution for Image Classification

doi:10.1109/TIP.2020.2985875

CORC > 自动化研究所 > 中国科学院自动化研究所 > 模式识别国家重点实验室 > 视频内容安全团队

	Anisotropic Convolution for Image Classification
	Li, Wenjuan 5; Li, Bing 2,5; Yuan, Chunfeng 5; Li, Yangxi 3; Wu, Haohao 6; Hu, Weiming 1,4,5; Wang, Fangshi 6
刊名	IEEE TRANSACTIONS ON IMAGE PROCESSING
	2020
卷号	29 页码:5584-5595
关键词	Convolution Shape Kernel Feature extraction Task analysis Training Neural networks Anisotropic convolution image classification object localization
ISSN号	1057-7149
DOI	10.1109/TIP.2020.2985875
通讯作者	Li, Bing(bli@nlpr.ia.ac.cn)
英文摘要	Convolutional neural networks are built upon simple but useful convolution modules. The traditional convolution has a limitation on feature extraction and object localization due to its fixed scale and geometric structure. Besides, the loss of spatial information also restricts the networks' performance and depth. To overcome these limitations, this paper proposes a novel anisotropic convolution by adding a scale factor and a shape factor into the traditional convolution. The anisotropic convolution augments the receptive fields flexibly and dynamically depending on the valid sizes of objects. In addition, the anisotropic convolution is a generalized convolution. The traditional convolution, dilated convolution and deformable convolution can be viewed as its special cases. Furthermore, in order to improve the training efficiency and avoid falling into a local optimum, this paper introduces a simplified implementation of the anisotropic convolution. The anisotropic convolution can be applied to arbitrary convolutional networks and the enhanced networks are called ACNs (anisotropic convolutional networks). Experimental results show that ACNs achieve better performance than many state-of-the-art methods and the baseline networks in tasks of image classification and object localization, especially in classification task of tiny images.
资助项目	Beijing Natural Science Foundation[JQ18018] ; Beijing Natural Science Foundation[L172051] ; Beijing Natural Science Foundation[L182058] ; National Key RD Plan[2017YFB1002801] ; Natural Science Foundation of China[U1936204] ; Natural Science Foundation of China[U1803119] ; Natural Science Foundation of China[U1736106] ; Natural Science Foundation of China[61876100] ; Natural Science Foundation of China[61751212] ; Natural Science Foundation of China[61721004] ; Natural Science Foundation of China[61906192] ; Natural Science Foundation of China[61972397] ; Natural Science Foundation of China[61772225] ; NSFC-General Technology Collaborative Fund for Basic Research[U1636218] ; Key Research Program of Frontier Sciences, CAS[YZDJ-SSW-JSC040] ; Science and Technology Service Network Initiative, CAS[KFJ-STS-SCYD-317] ; CAS External Cooperation Key Project ; Youth Innovation Promotion Association, CAS
WOS研究方向	Computer Science ; Engineering
语种	英语
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
WOS记录号	WOS:000529943000003
资助机构	Beijing Natural Science Foundation ; National Key RD Plan ; Natural Science Foundation of China ; NSFC-General Technology Collaborative Fund for Basic Research ; Key Research Program of Frontier Sciences, CAS ; Science and Technology Service Network Initiative, CAS ; CAS External Cooperation Key Project ; Youth Innovation Promotion Association, CAS
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/39389]
专题	自动化研究所_模式识别国家重点实验室_视频内容安全团队
通讯作者	Li, Bing
作者单位	1.Chinese Acad Sci, CAS Ctr Excellence Brain Sci & Intelligence Techn, Beijing 100190, Peoples R China 2.Peoples Daily Online, State Key Lab Commun Content Cognit, Beijing 100733, Peoples R China 3.Coordinat Ctr China CNCERT CC, Natl Comp Network Emergency Response Tech Team, Beijing 100029, Peoples R China 4.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing 100039, Peoples R China 5.Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China 6.Beijing Jiaotong Univ, Sch Software Engn, Beijing 100093, Peoples R China
推荐引用方式 GB/T 7714	Li, Wenjuan,Li, Bing,Yuan, Chunfeng,et al. Anisotropic Convolution for Image Classification[J]. IEEE TRANSACTIONS ON IMAGE PROCESSING,2020,29:5584-5595.
APA	Li, Wenjuan.,Li, Bing.,Yuan, Chunfeng.,Li, Yangxi.,Wu, Haohao.,...&Wang, Fangshi.(2020).Anisotropic Convolution for Image Classification.IEEE TRANSACTIONS ON IMAGE PROCESSING,29,5584-5595.
MLA	Li, Wenjuan,et al."Anisotropic Convolution for Image Classification".IEEE TRANSACTIONS ON IMAGE PROCESSING 29(2020):5584-5595.