SG-SRNs: Superpixel-Guided Scene Representation Networks

doi:10.1109/LSP.2022.3209147

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 > 机器人视觉团队

	SG-SRNs: Superpixel-Guided Scene Representation Networks
	Liu, Qiang 1,6; Lu, Xiao 2; Dong, Qiulei 3,4,5; Zhang, Yangyong 1,6; Wang, Haixia 1,6
刊名	IEEE SIGNAL PROCESSING LETTERS
	2022
卷号	29 页码:2038-2042
关键词	Image segmentation Three-dimensional displays Task analysis Image color analysis Image reconstruction Distortion Cameras Scene representation networks self-supervised multi-task learning superpixel-guided superpixel regularization
ISSN号	1070-9908
DOI	10.1109/LSP.2022.3209147
通讯作者	Wang, Haixia(hxwang@sdust.edu.cn)
英文摘要	Recently, Scene Representation Networks (SRNs) have attracted increasing attention in computer vision, due to their continuous and light-weight scene representation ability. However, SRNs generally perform poorly on low-texture image regions. Addressing this problem, we propose superpixel-guided scene representation networks in this paper, called SG-SRNs, consisting of a backbone module (SRNs), a superpixel segmentation module, and a superpixel regularization module. In the proposed method, except for the novel view synthesis task, the task of representation-aware superpixel segmentation mask generation is realized by the proposed superpixel segmentation module. Then, the superpixel regularization module utilizes the superpixel segmentation mask to guide the backbone to be learned in a locally smooth way, and optimizes the scene representations of the local regions to indirectly alleviate the structure distortion of low-texture regions in a self-supervised manner. Extensive experimental results on both our constructed datasets and the public Synthetic-NeRF dataset demonstrated that the proposed SG-SRNs achieved a significantly better 3D structure representing performance.
资助项目	National Natural Science Foundation of China[62073199] ; National Natural Science Foundation of China[62273213] ; National Natural Science Foundation of China[U1805264] ; Strategic Priority Research Program of the Chinese Academy of Sciences[XDB32050100] ; Beijing Municipal Science and Technology Project[Z211100011021004] ; Natural Science Foundation of Shandong Province[ZR2020MF095] ; Taishan Scholarship Construction Engineering
WOS研究方向	Engineering
语种	英语
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
WOS记录号	WOS:000866498400005
资助机构	National Natural Science Foundation of China ; Strategic Priority Research Program of the Chinese Academy of Sciences ; Beijing Municipal Science and Technology Project ; Natural Science Foundation of Shandong Province ; Taishan Scholarship Construction Engineering
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/50334]
专题	自动化研究所_模式识别国家重点实验室_机器人视觉团队
通讯作者	Wang, Haixia
作者单位	1.Shandong Univ Sci & Technol, Coll Elect Engn & Automat, Qingdao 266590, Peoples R China 2.Shandong Univ Sci & Technol, Coll Energy Storage Technol, Qingdao 266590, Peoples R China 3.Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China 4.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing 100049, Peoples R China 5.Chinese Acad Sci, Ctr Excellence Brain Sci & Intelligence Technol, Beijing 100190, Peoples R China 6.Shandong Univ Sci & Technol, Key Lab Robot & Intelligent Technol Shandong Prov, Qingdao 266590, Peoples R China
推荐引用方式 GB/T 7714	Liu, Qiang,Lu, Xiao,Dong, Qiulei,et al. SG-SRNs: Superpixel-Guided Scene Representation Networks[J]. IEEE SIGNAL PROCESSING LETTERS,2022,29:2038-2042.
APA	Liu, Qiang,Lu, Xiao,Dong, Qiulei,Zhang, Yangyong,&Wang, Haixia.(2022).SG-SRNs: Superpixel-Guided Scene Representation Networks.IEEE SIGNAL PROCESSING LETTERS,29,2038-2042.
MLA	Liu, Qiang,et al."SG-SRNs: Superpixel-Guided Scene Representation Networks".IEEE SIGNAL PROCESSING LETTERS 29(2022):2038-2042.