MS-ALN: Multiscale Attention Learning Network for Pest Recognition
Complex backgrounds, occlusions, and non-uniform classes present great challenges to pest recognition in practical applications. In this paper, we propose a multiscale attention learning network to address these problems. This network recursively locates discriminative regions and learns region-base...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2022-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9757218/ |
| _version_ | 1811332238720106496 |
|---|---|
| author | Fuxiang Feng Hanlin Dong Youmei Zhang Yu Zhang Bin Li |
| author_facet | Fuxiang Feng Hanlin Dong Youmei Zhang Yu Zhang Bin Li |
| author_sort | Fuxiang Feng |
| collection | DOAJ |
| description | Complex backgrounds, occlusions, and non-uniform classes present great challenges to pest recognition in practical applications. In this paper, we propose a multiscale attention learning network to address these problems. This network recursively locates discriminative regions and learns region-based feature representation in four branches. Three newly designed modules, which are target localization, attention detection, and attention removal connect two feature extracting sub-networks in adjacent branches to generate images of different scales. The target localization and attention detection modules locate the discriminative regions to filter out complex backgrounds while the attention removal module randomly removes the discriminative region to encourage the model to tackle occlusions. Thereafter, the parameter-shared classification sub-network follows the feature extracting sub-network in every branch for pest recognition. A decoupled learning strategy is adopted to address the problem of non-uniform classes. We experimented on the widely used IP-102 dataset and achieved state-of-the-art performance. |
| first_indexed | 2024-04-13T16:33:49Z |
| format | Article |
| id | doaj.art-eec7eb7667a14853b7296ea0fcc597e9 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-04-13T16:33:49Z |
| publishDate | 2022-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-eec7eb7667a14853b7296ea0fcc597e92022-12-22T02:39:30ZengIEEEIEEE Access2169-35362022-01-0110408884089810.1109/ACCESS.2022.31673979757218MS-ALN: Multiscale Attention Learning Network for Pest RecognitionFuxiang Feng0https://orcid.org/0000-0003-3058-4089Hanlin Dong1https://orcid.org/0000-0002-4395-6112Youmei Zhang2https://orcid.org/0000-0003-4185-0127Yu Zhang3Bin Li4https://orcid.org/0000-0002-4028-0938School of Mathematics and Statistics, Qilu University of Technology (Shandong Academy of Sciences), Jinan, ChinaSchool of Mathematics and Statistics, Qilu University of Technology (Shandong Academy of Sciences), Jinan, ChinaSchool of Mathematics and Statistics, Qilu University of Technology (Shandong Academy of Sciences), Jinan, ChinaSchool of Mathematics and Statistics, Qilu University of Technology (Shandong Academy of Sciences), Jinan, ChinaSchool of Mathematics and Statistics, Qilu University of Technology (Shandong Academy of Sciences), Jinan, ChinaComplex backgrounds, occlusions, and non-uniform classes present great challenges to pest recognition in practical applications. In this paper, we propose a multiscale attention learning network to address these problems. This network recursively locates discriminative regions and learns region-based feature representation in four branches. Three newly designed modules, which are target localization, attention detection, and attention removal connect two feature extracting sub-networks in adjacent branches to generate images of different scales. The target localization and attention detection modules locate the discriminative regions to filter out complex backgrounds while the attention removal module randomly removes the discriminative region to encourage the model to tackle occlusions. Thereafter, the parameter-shared classification sub-network follows the feature extracting sub-network in every branch for pest recognition. A decoupled learning strategy is adopted to address the problem of non-uniform classes. We experimented on the widely used IP-102 dataset and achieved state-of-the-art performance.https://ieeexplore.ieee.org/document/9757218/Pest recognitionmultiscale attention learning networktarget localization moduleattention detection moduleattention removal module |
| spellingShingle | Fuxiang Feng Hanlin Dong Youmei Zhang Yu Zhang Bin Li MS-ALN: Multiscale Attention Learning Network for Pest Recognition IEEE Access Pest recognition multiscale attention learning network target localization module attention detection module attention removal module |
| title | MS-ALN: Multiscale Attention Learning Network for Pest Recognition |
| title_full | MS-ALN: Multiscale Attention Learning Network for Pest Recognition |
| title_fullStr | MS-ALN: Multiscale Attention Learning Network for Pest Recognition |
| title_full_unstemmed | MS-ALN: Multiscale Attention Learning Network for Pest Recognition |
| title_short | MS-ALN: Multiscale Attention Learning Network for Pest Recognition |
| title_sort | ms aln multiscale attention learning network for pest recognition |
| topic | Pest recognition multiscale attention learning network target localization module attention detection module attention removal module |
| url | https://ieeexplore.ieee.org/document/9757218/ |
| work_keys_str_mv | AT fuxiangfeng msalnmultiscaleattentionlearningnetworkforpestrecognition AT hanlindong msalnmultiscaleattentionlearningnetworkforpestrecognition AT youmeizhang msalnmultiscaleattentionlearningnetworkforpestrecognition AT yuzhang msalnmultiscaleattentionlearningnetworkforpestrecognition AT binli msalnmultiscaleattentionlearningnetworkforpestrecognition |