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
RFMNet: Robust Deep Functional Maps for unsupervised non-rigid shape correspondence

文献类型：期刊论文

作者Hu, Ling1; Li, Qinsong4,5; Liu, Shengjun5; Yan, Dong-Ming2,3; Xu, Haojun5; Liu, Xinru5
刊名GRAPHICAL MODELS
出版日期2023-10-01
卷号129页码:11
ISSN号1524-0703
关键词Shape correspondence Functional maps Unsupervised learning Optimal transport
DOI10.1016/j.gmod.2023.101189
通讯作者Liu, Shengjun(shjliu.cg@csu.edu.cn)
英文摘要In traditional deep functional maps for non-rigid shape correspondence, estimating a functional map including high-frequency information requires enough linearly independent features via the least square method, which is prone to be violated in practice, especially at an early stage of training, or costly post-processing, e.g. ZoomOut. In this paper, we propose a novel method called RFMNet (Robust Deep Functional Map Networks), which jointly considers training stability and more geometric shape features than previous works. We directly first produce a pointwise map by resorting to optimal transport and then convert it to an initial functional map. Such a mechanism mitigates the requirements for the descriptor and avoids the training instabilities resulting from the least square solver. Benefitting from the novel strategy, we successfully integrate a state-of -the-art geometric regularization for further optimizing the functional map, which substantially filters the initial functional map. We show our novel computing functional map module brings more stable training even under encoding the functional map with high-frequency information and faster convergence speed. Considering the pointwise and functional maps, an unsupervised loss is presented for penalizing the correspondence distortion of Delta functions between shapes. To catch discretization-resistant and orientation-aware shape features with our network, we utilize DiffusionNet as a feature extractor. Experimental results demonstrate our apparent superiority in correspondence quality and generalization across various shape discretizations and different datasets compared to the state-of-the-art learning methods.
资助项目National Key Research and Development Program, China[2019YFB2204104] ; Hunan Provincial Natural Science Foundation of China[2021JJ30172] ; Hunan Provincial Natural Science Foundation of China[2023JJ40769] ; Natural Science Founda-tion of China[62172447] ; Natural Science Founda-tion of China[61876191] ; Natural Science Founda-tion of China[62172415] ; Open Project Program of the State Key Laboratory of Multimodal Artificial Intelligence Systems, China[202200025]
WOS研究方向Computer Science
语种英语
出版者ACADEMIC PRESS INC ELSEVIER SCIENCE
WOS记录号WOS:001048855000001
资助机构National Key Research and Development Program, China ; Hunan Provincial Natural Science Foundation of China ; Natural Science Founda-tion of China ; Open Project Program of the State Key Laboratory of Multimodal Artificial Intelligence Systems, China
源URL[http://ir.ia.ac.cn/handle/173211/53999]  
专题多模态人工智能系统全国重点实验室
通讯作者Liu, Shengjun
作者单位1.Hunan First Normal Univ, Sch Math & Stat, Changsha, Peoples R China
2.Chinese Acad Sci, Inst Automat, State Key Lab Multimodal Artificial Intelligence S, Beijing, Peoples R China
3.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing, Peoples R China
4.Cent South Univ, Big Data Inst, Changsha, Peoples R China
5.Cent South Univ, Inst Engn Modeling & Sci Comp, Changsha, Peoples R China
推荐引用方式
GB/T 7714		 Hu, Ling,Li, Qinsong,Liu, Shengjun,et al. RFMNet: Robust Deep Functional Maps for unsupervised non-rigid shape correspondence[J]. GRAPHICAL MODELS,2023,129:11.
APA Hu, Ling,Li, Qinsong,Liu, Shengjun,Yan, Dong-Ming,Xu, Haojun,&Liu, Xinru.(2023).RFMNet: Robust Deep Functional Maps for unsupervised non-rigid shape correspondence.GRAPHICAL MODELS,129,11.
MLA Hu, Ling,et al."RFMNet: Robust Deep Functional Maps for unsupervised non-rigid shape correspondence".GRAPHICAL MODELS 129(2023):11.

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。