Personalized Convolution for Face Recognition

doi:10.1007/s11263-021-01536-x

	Personalized Convolution for Face Recognition
	Han, Chunrui 1,2; Shan, Shiguang 1,2,3; Kan, Meina 1,2; Wu, Shuzhe 1,2; Chen, Xilin 1,2
	2022-01-04
发表期刊	INTERNATIONAL JOURNAL OF COMPUTER VISION
ISSN	0920-5691
页码	19
摘要	Face recognition has been significantly advanced by deep learning based methods. In all face recognition methods based on convolutional neural network (CNN), the convolutional kernels for feature extraction are fixed regardless of the input face once the training stage is finished. By contrast, we humans are usually impressed by some unique characteristics of different persons, such as one's blue eyes while another one's crooked nose, or even someone's naevus at specific location. Inspired by this observation, we propose a personalized convolution method which aims to extract special distinguishing characteristics of each person for more accurate face recognition. Specifically, given a face, we adaptively generate a set of kernels for him/her, named by us ordinary kernel, which is further analytically decomposed into two orthogonal components, i.e., the commonality component and the specialty component. The former characterizes the commonality among subjects which is optimized on a reference set. The latter is the residual part by filtering out the commonality component from the ordinary kernel, so as to capture those special characteristics, named by us personalized kernel. The CNNs with personalized kernels for convolution can highlight those specialty of a person's distinguishing characteristics while suppress his/her commonality with others, leading to better distinguishing of different faces. Additionally, as a by-product, the reference set also facilitates the adaptation of our method to different scenarios by simply selecting faces of a particular population. Extensive experiments on the challenging LFW, IJB-A and IJB-C datasets validate that our proposed personalized convolution achieves significant improvement over the conventional CNN, and also other existing methods for face recognition.
关键词	Face recognition Personalized convolution Personalized kernel
DOI	10.1007/s11263-021-01536-x
收录类别	SCI
语种	英语
资助项目	National KeyResearch andDevelopment Program of China[2017YFA0700800] ; Natural Science Foundation of China[61772496] ; Beijing Nova Program[Z191100001119123]
WOS研究方向	Computer Science
WOS类目	Computer Science, Artificial Intelligence
WOS记录号	WOS:000738560100003
出版者	SPRINGER
引用统计	被引频次：6[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/18383
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Kan, Meina
作者单位	1.Chinese Acad Sci, Inst Comp Technol, CAS, Key Lab Intelligent Informat Proc, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Beijing 100049, Peoples R China 3.CAS Ctr Excellence Brain Sci & Intelligence Techn, Shanghai, Peoples R China
推荐引用方式 GB/T 7714	Han, Chunrui,Shan, Shiguang,Kan, Meina,et al. Personalized Convolution for Face Recognition[J]. INTERNATIONAL JOURNAL OF COMPUTER VISION,2022:19.
APA	Han, Chunrui,Shan, Shiguang,Kan, Meina,Wu, Shuzhe,&Chen, Xilin.(2022).Personalized Convolution for Face Recognition.INTERNATIONAL JOURNAL OF COMPUTER VISION,19.
MLA	Han, Chunrui,et al."Personalized Convolution for Face Recognition".INTERNATIONAL JOURNAL OF COMPUTER VISION (2022):19.