CASIA OpenIR  > 智能感知与计算
Few-shot learning with unsupervised part discovery and part-aligned similarity
Chen, Wentao1,2; Zhang, Zhang2,3,4; Wang, Wei2,3; Wang, Liang2,3; Wang, Zilei1; Tan, Tieniu1,2,3
Source PublicationPATTERN RECOGNITION
ISSN0031-3203
2023
Volume133Pages:12
Corresponding AuthorZhang, Zhang(zzhang@nlpr.ia.ac.cn)
AbstractFew-shot learning aims to recognize novel concepts with only a few examples. To this end, previous studies resort to acquiring a strong inductive bias via meta-learning on a group of similar tasks, which however needs a large labeled base dataset to sample training tasks. In this paper, we show that such inductive bias can be learned from a flat collection of unlabeled images, and instantiated as transfer-able representations among seen and unseen classes. Specifically, we propose a novel unsupervised Part Discovery Network (PDN) to learn transferable representations from unlabeled images, which automat-ically selects the most discriminative part from an input image and then maximizes its similarities to the global view of the input and other neighbors with similar semantics. To better leverage the learned representations for few-shot learning, we further propose Part-Aligned Similarity (PAS), the key of which is to measure image similarities based on a set of discriminative and aligned parts. We conduct extensive studies on five popular few-shot learning datasets to evaluate our approach. The experimental results show that our approach outperforms previous unsupervised methods by a large margin and is even com-parable with state-of-the-art supervised methods.(c) 2022 Elsevier Ltd. All rights reserved.
KeywordFew-shot learning Self-supervised learning Part discovery network Part-aligned similarity
DOI10.1016/j.patcog.2022.108986
Indexed BySCI
Language英语
Funding ProjectNational Natural Science Foundation of China[61721004] ; National Natural Science Foundation of China[61836008] ; National Natural Science Foundation of China[61976214] ; National Natural Science Foundation of China[62076078] ; CAS -AIR
Funding OrganizationNational Natural Science Foundation of China ; CAS -AIR
WOS Research AreaComputer Science ; Engineering
WOS SubjectComputer Science, Artificial Intelligence ; Engineering, Electrical & Electronic
WOS IDWOS:000863094500003
PublisherELSEVIER SCI LTD
Citation statistics
Document Type期刊论文
Identifierhttp://ir.ia.ac.cn/handle/173211/50312
Collection智能感知与计算
Corresponding AuthorZhang, Zhang
Affiliation1.Univ Sci & Technol China, Hefei, Peoples R China
2.CASIA, NLPR, Ctr Res Intelligent Percept & Comp, Beijing, Peoples R China
3.Univ Chinese Acad Sci, Beijing, Peoples R China
4.Inst Automat, 95 Zhongguancun East Rd, Beijing, Peoples R China
First Author AffilicationChinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
Corresponding Author AffilicationChinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
Recommended Citation
GB/T 7714
Chen, Wentao,Zhang, Zhang,Wang, Wei,et al. Few-shot learning with unsupervised part discovery and part-aligned similarity[J]. PATTERN RECOGNITION,2023,133:12.
APA Chen, Wentao,Zhang, Zhang,Wang, Wei,Wang, Liang,Wang, Zilei,&Tan, Tieniu.(2023).Few-shot learning with unsupervised part discovery and part-aligned similarity.PATTERN RECOGNITION,133,12.
MLA Chen, Wentao,et al."Few-shot learning with unsupervised part discovery and part-aligned similarity".PATTERN RECOGNITION 133(2023):12.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[Chen, Wentao]'s Articles
[Zhang, Zhang]'s Articles
[Wang, Wei]'s Articles
Baidu academic
Similar articles in Baidu academic
[Chen, Wentao]'s Articles
[Zhang, Zhang]'s Articles
[Wang, Wei]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Chen, Wentao]'s Articles
[Zhang, Zhang]'s Articles
[Wang, Wei]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.