Adaptive Feature Fusion for Small Object Detection
In order to alleviate the situation that small objects are prone to missed detection and false detection in natural scenes, this paper proposed a small object detection algorithm for adaptive feature fusion, referred to as MMF-YOLO. First, aiming at the problem that small object pixels are easy to l...
| Váldodahkkit: | , , |
|---|---|
| Materiálatiipa: | Artihkal |
| Giella: | English |
| Almmustuhtton: |
MDPI AG
2022-11-01
|
| Ráidu: | Applied Sciences |
| Fáttát: | |
| Liŋkkat: | https://www.mdpi.com/2076-3417/12/22/11854 |
| _version_ | 1827645016989761536 |
|---|---|
| author | Qi Zhang Hongying Zhang Xiuwen Lu |
| author_facet | Qi Zhang Hongying Zhang Xiuwen Lu |
| author_sort | Qi Zhang |
| collection | DOAJ |
| description | In order to alleviate the situation that small objects are prone to missed detection and false detection in natural scenes, this paper proposed a small object detection algorithm for adaptive feature fusion, referred to as MMF-YOLO. First, aiming at the problem that small object pixels are easy to lose, a multi-branch cross-scale feature fusion module with fusion factor was proposed, where each fusion path has an adaptive fusion factor, which can allow the network to independently adjust the importance of features according to the learned weights. Then, aiming at the problem that small objects are similar to background information and small objects overlap in complex scenes, the M-CBAM attention mechanism was proposed, which was added to the feature reinforcement extraction module to reduce feature redundancy. Finally, in light of the problem of small object size and large size span, the size of the object detection head was modified to adapt to the small object size. Experiments on the VisDrone2019 dataset showed that the mAP of the proposed algorithm could reach 42.23%, and the parameter quantity was only 29.33 MB, which is 9.13% ± 0.07% higher than the benchmark network mAP, and the network model was reduced by 5.22 MB. |
| first_indexed | 2024-03-09T18:29:36Z |
| format | Article |
| id | doaj.art-b34f0a2afbff44f595c0aad08697de79 |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-09T18:29:36Z |
| publishDate | 2022-11-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-b34f0a2afbff44f595c0aad08697de792023-11-24T07:42:01ZengMDPI AGApplied Sciences2076-34172022-11-0112221185410.3390/app122211854Adaptive Feature Fusion for Small Object DetectionQi Zhang0Hongying Zhang1Xiuwen Lu2School of Information Engineering, Southwest University of Science and Technology, Mianyang 621000, ChinaSchool of Information Engineering, Southwest University of Science and Technology, Mianyang 621000, ChinaSchool of Information Engineering, Southwest University of Science and Technology, Mianyang 621000, ChinaIn order to alleviate the situation that small objects are prone to missed detection and false detection in natural scenes, this paper proposed a small object detection algorithm for adaptive feature fusion, referred to as MMF-YOLO. First, aiming at the problem that small object pixels are easy to lose, a multi-branch cross-scale feature fusion module with fusion factor was proposed, where each fusion path has an adaptive fusion factor, which can allow the network to independently adjust the importance of features according to the learned weights. Then, aiming at the problem that small objects are similar to background information and small objects overlap in complex scenes, the M-CBAM attention mechanism was proposed, which was added to the feature reinforcement extraction module to reduce feature redundancy. Finally, in light of the problem of small object size and large size span, the size of the object detection head was modified to adapt to the small object size. Experiments on the VisDrone2019 dataset showed that the mAP of the proposed algorithm could reach 42.23%, and the parameter quantity was only 29.33 MB, which is 9.13% ± 0.07% higher than the benchmark network mAP, and the network model was reduced by 5.22 MB.https://www.mdpi.com/2076-3417/12/22/11854multi-scale feature fusionadaptive fusion factorattention mechanismsmall object detection |
| spellingShingle | Qi Zhang Hongying Zhang Xiuwen Lu Adaptive Feature Fusion for Small Object Detection Applied Sciences multi-scale feature fusion adaptive fusion factor attention mechanism small object detection |
| title | Adaptive Feature Fusion for Small Object Detection |
| title_full | Adaptive Feature Fusion for Small Object Detection |
| title_fullStr | Adaptive Feature Fusion for Small Object Detection |
| title_full_unstemmed | Adaptive Feature Fusion for Small Object Detection |
| title_short | Adaptive Feature Fusion for Small Object Detection |
| title_sort | adaptive feature fusion for small object detection |
| topic | multi-scale feature fusion adaptive fusion factor attention mechanism small object detection |
| url | https://www.mdpi.com/2076-3417/12/22/11854 |
| work_keys_str_mv | AT qizhang adaptivefeaturefusionforsmallobjectdetection AT hongyingzhang adaptivefeaturefusionforsmallobjectdetection AT xiuwenlu adaptivefeaturefusionforsmallobjectdetection |