Multi-objective evolutionary 3D face reconstruction based on improved encoder-decoder network

doi:10.1016/j.ins.2021.09.024

CASIA OpenIR > 多模态人工智能系统全国重点实验室 > 人工智能与机器学习（杨雪冰）-技术团队

	Multi-objective evolutionary 3D face reconstruction based on improved encoder-decoder network
	Cai, Xingjuan 1; Cao, Yihao 1; Ren, Yeqing 2; Cui, Zhihua 1; Zhang, Wensheng3
发表期刊	INFORMATION SCIENCES
ISSN	0020-0255
	2021-12-01
卷号	581 页码:233-248
通讯作者	Cui, Zhihua(zhihua.cui@hotmail.com)
摘要	Three-dimensional (3D) face application has attracted significant attention the field of multimedia and image processing. However, the end-to-end 3D face reconstruction method is still immature, and there are some problems, such as overfitting caused by too few training sets and unacceptable 3D face texture alignment performance. Therefore, we design a novel approach to construct a 3D face, named multi-objective evo-lutionary 3D face reconstruction based on improved encoder-decoder network (MoEDN). This study introduces a regularization algorithm named feature map distortion (Disout); whose purpose is to strengthen the network generalization ability. Based on this, we con-struct a multi-objective evolutionary 3D face reconstruction model, in which decision vari-ables are distortion probability, distorted block size, distorted intensity, probability step, and learning rate; and objective functions are loss and structural similarity (SSIM). We use four multi-objective evolutionary algorithms (NSGA-II, AGEII, NSLS, and MOEA/D) to optimize the proposed model. Experimental results demonstrate that NSLS has the best performance. In addition, compared with position map regression network (PRNet), 2D-assisted self-supervised learning (2DASL) and other state-of-the-art, the proposed model achieves better loss values and NME values. Therefore, the proposed multi-objective evo-lutionary 3D face reconstruction model has outstanding 3D facial reconstruction perfor-mance in large poses and face expression. (c) 2021 Elsevier Inc. All rights reserved.
关键词	3D face reconstruction Multi-objective evolutionary Regularization algorithm Computer vision
DOI	10.1016/j.ins.2021.09.024
关键词[WOS]	IMAGE ; ALGORITHM
收录类别	SCI
语种	英语
资助项目	National Key Research and Development Program of China[2018YFC1604000] ; National Natural Science Foundation of China[61806138] ; Key R&D program of Shanxi Province (International Cooperation)[201903D421048] ; Key R&D program of Shanxi Province (High Technology)[201903D121119]
项目资助者	National Key Research and Development Program of China ; National Natural Science Foundation of China ; Key R&D program of Shanxi Province (International Cooperation) ; Key R&D program of Shanxi Province (High Technology)
WOS研究方向	Computer Science
WOS类目	Computer Science, Information Systems
WOS记录号	WOS:000705058500005
出版者	ELSEVIER SCIENCE INC
引用统计	被引频次：18[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://ir.ia.ac.cn/handle/173211/46204
专题	多模态人工智能系统全国重点实验室_人工智能与机器学习（杨雪冰）-技术团队
通讯作者	Cui, Zhihua
作者单位	1.Taiyuan Univ Sci & Technol, Sch Comp Sci & Technol, Taiyuan, Peoples R China 2.Beijing Univ Posts & Telecommun, Beijing 100876, Peoples R China 3.Chinese Acad Sci, Inst Automat, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Cai, Xingjuan,Cao, Yihao,Ren, Yeqing,et al. Multi-objective evolutionary 3D face reconstruction based on improved encoder-decoder network[J]. INFORMATION SCIENCES,2021,581:233-248.
APA	Cai, Xingjuan,Cao, Yihao,Ren, Yeqing,Cui, Zhihua,&Zhang, Wensheng.(2021).Multi-objective evolutionary 3D face reconstruction based on improved encoder-decoder network.INFORMATION SCIENCES,581,233-248.
MLA	Cai, Xingjuan,et al."Multi-objective evolutionary 3D face reconstruction based on improved encoder-decoder network".INFORMATION SCIENCES 581(2021):233-248.