Few-Shot Learning Based on Double Pooling Squeeze and Excitation Attention
Training a generalized reliable model is a great challenge since sufficiently labeled data are unavailable in some open application scenarios. Few-shot learning (FSL) aims to learn new problems with only a few examples that can tackle this problem and attract extensive attention. This paper proposes...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-12-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/12/1/27 |
| _version_ | 1797626081677672448 |
|---|---|
| author | Qiuyu Xu Jie Su Ying Wang Jing Zhang Yixin Zhong |
| author_facet | Qiuyu Xu Jie Su Ying Wang Jing Zhang Yixin Zhong |
| author_sort | Qiuyu Xu |
| collection | DOAJ |
| description | Training a generalized reliable model is a great challenge since sufficiently labeled data are unavailable in some open application scenarios. Few-shot learning (FSL) aims to learn new problems with only a few examples that can tackle this problem and attract extensive attention. This paper proposes a novel few-shot learning method based on double pooling squeeze and excitation attention (dSE) for the purpose of improving the discriminative ability of the model by proposing a novel feature expression. Specifically, the proposed dSE module adopts two types of pooling to emphasize features responding to foreground object channels. We employed both the pixel descriptor and channel descriptor to capture locally identifiable channel features and pixel features of an image (as opposed to traditional few-shot learning methods). Additionally, in order to improve the robustness of the model, we designed a new loss function. To verify the performance of the method, a large number of experiments were performed on multiple standard few-shot image benchmark datasets, showing that our framework can outperform several existing approaches. Moreover, we performed extensive experiments on three more challenging fine-grained few-shot datasets, the experimental results demonstrate that the proposed method achieves state-of-the-art performances. In particular, this work achieves 92.36% accuracy under the 5-way–5-shot classification setting of the Stanford Cars dataset. |
| first_indexed | 2024-03-11T10:05:25Z |
| format | Article |
| id | doaj.art-f46266f994844197964be106ff98ab85 |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-11T10:05:25Z |
| publishDate | 2022-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-f46266f994844197964be106ff98ab852023-11-16T15:10:06ZengMDPI AGElectronics2079-92922022-12-011212710.3390/electronics12010027Few-Shot Learning Based on Double Pooling Squeeze and Excitation AttentionQiuyu Xu0Jie Su1Ying Wang2Jing Zhang3Yixin Zhong4School of Information Science and Engineering, University of Jinan, Jinan 250022, ChinaSchool of Information Science and Engineering, University of Jinan, Jinan 250022, ChinaSchool of Information Science and Engineering, University of Jinan, Jinan 250022, ChinaSchool of Information Science and Engineering, University of Jinan, Jinan 250022, ChinaArtificial Intelligence Research Institute, University of Jinan, Jinan 250022, ChinaTraining a generalized reliable model is a great challenge since sufficiently labeled data are unavailable in some open application scenarios. Few-shot learning (FSL) aims to learn new problems with only a few examples that can tackle this problem and attract extensive attention. This paper proposes a novel few-shot learning method based on double pooling squeeze and excitation attention (dSE) for the purpose of improving the discriminative ability of the model by proposing a novel feature expression. Specifically, the proposed dSE module adopts two types of pooling to emphasize features responding to foreground object channels. We employed both the pixel descriptor and channel descriptor to capture locally identifiable channel features and pixel features of an image (as opposed to traditional few-shot learning methods). Additionally, in order to improve the robustness of the model, we designed a new loss function. To verify the performance of the method, a large number of experiments were performed on multiple standard few-shot image benchmark datasets, showing that our framework can outperform several existing approaches. Moreover, we performed extensive experiments on three more challenging fine-grained few-shot datasets, the experimental results demonstrate that the proposed method achieves state-of-the-art performances. In particular, this work achieves 92.36% accuracy under the 5-way–5-shot classification setting of the Stanford Cars dataset.https://www.mdpi.com/2079-9292/12/1/27few-shot learningmetric learningimage classificationattention mechanismfeature representation |
| spellingShingle | Qiuyu Xu Jie Su Ying Wang Jing Zhang Yixin Zhong Few-Shot Learning Based on Double Pooling Squeeze and Excitation Attention Electronics few-shot learning metric learning image classification attention mechanism feature representation |
| title | Few-Shot Learning Based on Double Pooling Squeeze and Excitation Attention |
| title_full | Few-Shot Learning Based on Double Pooling Squeeze and Excitation Attention |
| title_fullStr | Few-Shot Learning Based on Double Pooling Squeeze and Excitation Attention |
| title_full_unstemmed | Few-Shot Learning Based on Double Pooling Squeeze and Excitation Attention |
| title_short | Few-Shot Learning Based on Double Pooling Squeeze and Excitation Attention |
| title_sort | few shot learning based on double pooling squeeze and excitation attention |
| topic | few-shot learning metric learning image classification attention mechanism feature representation |
| url | https://www.mdpi.com/2079-9292/12/1/27 |
| work_keys_str_mv | AT qiuyuxu fewshotlearningbasedondoublepoolingsqueezeandexcitationattention AT jiesu fewshotlearningbasedondoublepoolingsqueezeandexcitationattention AT yingwang fewshotlearningbasedondoublepoolingsqueezeandexcitationattention AT jingzhang fewshotlearningbasedondoublepoolingsqueezeandexcitationattention AT yixinzhong fewshotlearningbasedondoublepoolingsqueezeandexcitationattention |