Few-Shot Image Classification: Current Status and Research Trends
Conventional image classification methods usually require a large number of training samples for the training model. However, in practical scenarios, the amount of available sample data is often insufficient, which easily leads to overfitting in network construction. Few-shot learning provides an ef...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-05-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/11/11/1752 |
| _version_ | 1797493725539074048 |
|---|---|
| author | Ying Liu Hengchang Zhang Weidong Zhang Guojun Lu Qi Tian Nam Ling |
| author_facet | Ying Liu Hengchang Zhang Weidong Zhang Guojun Lu Qi Tian Nam Ling |
| author_sort | Ying Liu |
| collection | DOAJ |
| description | Conventional image classification methods usually require a large number of training samples for the training model. However, in practical scenarios, the amount of available sample data is often insufficient, which easily leads to overfitting in network construction. Few-shot learning provides an effective solution to this problem and has been a hot research topic. This paper provides an intensive survey on the state-of-the-art techniques in image classification based on few-shot learning. According to the different deep learning mechanisms, the existing algorithms are divided into four categories: transfer learning based, meta-learning based, data augmentation based, and multimodal based methods. Transfer learning based methods transfer useful prior knowledge from the source domain to the target domain. Meta-learning based methods employ past prior knowledge to guide the learning of new tasks. Data augmentation based methods expand the amount of sample data with auxiliary information. Multimodal based methods use the information of the auxiliary modal to facilitate the implementation of image classification tasks. This paper also summarizes the few-shot image datasets available in the literature, and experimental results tested by some representative algorithms are provided to compare their performance and analyze their pros and cons. In addition, the application of existing research outcomes on few-shot image classification in different practical fields are discussed. Finally, a few future research directions are identified. |
| first_indexed | 2024-03-10T01:24:10Z |
| format | Article |
| id | doaj.art-b2d17c100fc0427cb1ab19381f1e4dd9 |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-10T01:24:10Z |
| publishDate | 2022-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-b2d17c100fc0427cb1ab19381f1e4dd92023-11-23T13:55:19ZengMDPI AGElectronics2079-92922022-05-011111175210.3390/electronics11111752Few-Shot Image Classification: Current Status and Research TrendsYing Liu0Hengchang Zhang1Weidong Zhang2Guojun Lu3Qi Tian4Nam Ling5Center for Image and Information Processing, Xi’an University of Posts and Telecommunications, Xi’an 710121, ChinaCenter for Image and Information Processing, Xi’an University of Posts and Telecommunications, Xi’an 710121, ChinaCenter for Image and Information Processing, Xi’an University of Posts and Telecommunications, Xi’an 710121, ChinaSchool of Engineering, Information Technology and Physical Sciences, Federation University Australia, Gippsland 3841, AustraliaHuawei Technologies Co., Ltd., Shenzhen 518000, ChinaDepartment of Computer Science and Engineering, Santa Clara University, Santa Clara, CA 950053, USAConventional image classification methods usually require a large number of training samples for the training model. However, in practical scenarios, the amount of available sample data is often insufficient, which easily leads to overfitting in network construction. Few-shot learning provides an effective solution to this problem and has been a hot research topic. This paper provides an intensive survey on the state-of-the-art techniques in image classification based on few-shot learning. According to the different deep learning mechanisms, the existing algorithms are divided into four categories: transfer learning based, meta-learning based, data augmentation based, and multimodal based methods. Transfer learning based methods transfer useful prior knowledge from the source domain to the target domain. Meta-learning based methods employ past prior knowledge to guide the learning of new tasks. Data augmentation based methods expand the amount of sample data with auxiliary information. Multimodal based methods use the information of the auxiliary modal to facilitate the implementation of image classification tasks. This paper also summarizes the few-shot image datasets available in the literature, and experimental results tested by some representative algorithms are provided to compare their performance and analyze their pros and cons. In addition, the application of existing research outcomes on few-shot image classification in different practical fields are discussed. Finally, a few future research directions are identified.https://www.mdpi.com/2079-9292/11/11/1752few-shot learningtransfer learningmeta-learningdata augmentationmultimodal |
| spellingShingle | Ying Liu Hengchang Zhang Weidong Zhang Guojun Lu Qi Tian Nam Ling Few-Shot Image Classification: Current Status and Research Trends Electronics few-shot learning transfer learning meta-learning data augmentation multimodal |
| title | Few-Shot Image Classification: Current Status and Research Trends |
| title_full | Few-Shot Image Classification: Current Status and Research Trends |
| title_fullStr | Few-Shot Image Classification: Current Status and Research Trends |
| title_full_unstemmed | Few-Shot Image Classification: Current Status and Research Trends |
| title_short | Few-Shot Image Classification: Current Status and Research Trends |
| title_sort | few shot image classification current status and research trends |
| topic | few-shot learning transfer learning meta-learning data augmentation multimodal |
| url | https://www.mdpi.com/2079-9292/11/11/1752 |
| work_keys_str_mv | AT yingliu fewshotimageclassificationcurrentstatusandresearchtrends AT hengchangzhang fewshotimageclassificationcurrentstatusandresearchtrends AT weidongzhang fewshotimageclassificationcurrentstatusandresearchtrends AT guojunlu fewshotimageclassificationcurrentstatusandresearchtrends AT qitian fewshotimageclassificationcurrentstatusandresearchtrends AT namling fewshotimageclassificationcurrentstatusandresearchtrends |