DAC: Disentanglement-and-Calibration Module for Cross-Domain Few-Shot Classification
Cross-domain few-shot classification (CD-FSC) aims to develop few-shot classification models trained on seen domains but tested on unseen domains. However, the cross-domain setup poses a challenge in the form of domain shift between the training and testing domains. Previous research has demonstrate...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10182258/ |
| _version_ | 1797742984680177664 |
|---|---|
| author | Hao Zheng Qiang Zhang Asako Kanezaki |
| author_facet | Hao Zheng Qiang Zhang Asako Kanezaki |
| author_sort | Hao Zheng |
| collection | DOAJ |
| description | Cross-domain few-shot classification (CD-FSC) aims to develop few-shot classification models trained on seen domains but tested on unseen domains. However, the cross-domain setup poses a challenge in the form of domain shift between the training and testing domains. Previous research has demonstrated that the encoder can disentangle features into domain-shared and domain-specific features. However, poorly estimated domain-specific features can lead to inadequate generalization on the unseen domain. This paper proposes a disentanglement-and-calibration module (DAC) to address this issue. The disentanglement component separates the features into domain-shared and domain-specific features, while the calibration component corrects the domain-specific features. We demonstrate that the DAC module can significantly enhance the generalization capability of several baseline methods. Furthermore, we show that MatchingNet with the DAC module outperforms existing state-of-the-art methods by 10%-11% when trained on mini-ImageNet, CUB-200, Cars196, Places365 and tested on Plantae dataset. |
| first_indexed | 2024-03-12T14:48:04Z |
| format | Article |
| id | doaj.art-ac25d389a03943fdad5ea96385f59365 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-03-12T14:48:04Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-ac25d389a03943fdad5ea96385f593652023-08-15T23:01:34ZengIEEEIEEE Access2169-35362023-01-0111826658267310.1109/ACCESS.2023.329498410182258DAC: Disentanglement-and-Calibration Module for Cross-Domain Few-Shot ClassificationHao Zheng0https://orcid.org/0009-0002-0950-3600Qiang Zhang1Asako Kanezaki2https://orcid.org/0000-0003-3217-1405Tokyo Institute of Technology, Meguro City, Tokyo, JapanThe Hong Kong University of Science and Technology (Guangzhou), Nansha, ChinaTokyo Institute of Technology, Meguro City, Tokyo, JapanCross-domain few-shot classification (CD-FSC) aims to develop few-shot classification models trained on seen domains but tested on unseen domains. However, the cross-domain setup poses a challenge in the form of domain shift between the training and testing domains. Previous research has demonstrated that the encoder can disentangle features into domain-shared and domain-specific features. However, poorly estimated domain-specific features can lead to inadequate generalization on the unseen domain. This paper proposes a disentanglement-and-calibration module (DAC) to address this issue. The disentanglement component separates the features into domain-shared and domain-specific features, while the calibration component corrects the domain-specific features. We demonstrate that the DAC module can significantly enhance the generalization capability of several baseline methods. Furthermore, we show that MatchingNet with the DAC module outperforms existing state-of-the-art methods by 10%-11% when trained on mini-ImageNet, CUB-200, Cars196, Places365 and tested on Plantae dataset.https://ieeexplore.ieee.org/document/10182258/Cross-domain few-shot classificationdisentanglementdomain shiftrepresentation learning |
| spellingShingle | Hao Zheng Qiang Zhang Asako Kanezaki DAC: Disentanglement-and-Calibration Module for Cross-Domain Few-Shot Classification IEEE Access Cross-domain few-shot classification disentanglement domain shift representation learning |
| title | DAC: Disentanglement-and-Calibration Module for Cross-Domain Few-Shot Classification |
| title_full | DAC: Disentanglement-and-Calibration Module for Cross-Domain Few-Shot Classification |
| title_fullStr | DAC: Disentanglement-and-Calibration Module for Cross-Domain Few-Shot Classification |
| title_full_unstemmed | DAC: Disentanglement-and-Calibration Module for Cross-Domain Few-Shot Classification |
| title_short | DAC: Disentanglement-and-Calibration Module for Cross-Domain Few-Shot Classification |
| title_sort | dac disentanglement and calibration module for cross domain few shot classification |
| topic | Cross-domain few-shot classification disentanglement domain shift representation learning |
| url | https://ieeexplore.ieee.org/document/10182258/ |
| work_keys_str_mv | AT haozheng dacdisentanglementandcalibrationmoduleforcrossdomainfewshotclassification AT qiangzhang dacdisentanglementandcalibrationmoduleforcrossdomainfewshotclassification AT asakokanezaki dacdisentanglementandcalibrationmoduleforcrossdomainfewshotclassification |