AFRE-Net: Adaptive Feature Representation Enhancement for Arbitrary Oriented Object Detection
Arbitrary-oriented object detection (AOOD) is a crucial task in aerial image analysis but is also faced with significant challenges. In current AOOD detectors, commonly used multi-scale feature fusion modules fall short in spatial and semantic information complement between scales. Additionally, fix...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-10-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/15/20/4965 |
| _version_ | 1797572377577521152 |
|---|---|
| author | Tianwei Zhang Xu Sun Lina Zhuang Xiaoyu Dong Jianjun Sha Bing Zhang Ke Zheng |
| author_facet | Tianwei Zhang Xu Sun Lina Zhuang Xiaoyu Dong Jianjun Sha Bing Zhang Ke Zheng |
| author_sort | Tianwei Zhang |
| collection | DOAJ |
| description | Arbitrary-oriented object detection (AOOD) is a crucial task in aerial image analysis but is also faced with significant challenges. In current AOOD detectors, commonly used multi-scale feature fusion modules fall short in spatial and semantic information complement between scales. Additionally, fixed feature extraction structures are usually used following a fusion model, resulting in the inability of detectors to self-adjust. At the same time, feature fusion and extraction modules are designed in isolation and the internal synergy between them is ignored. The above problems result in feature representation deficiency, thus affecting the overall detection precision. To solve these problems, we first create a fine-grained feature pyramid network (FG-FPN) that not only provides richer spatial and semantic features, but also completes neighbor scale features in a self-learning mode. Subsequently, we propose a novel feature enhancement module (FEM) to fit FG-FPN. FEM authorizes the detection unit to automatically adjust the sensing area and adaptively suppress background interference, thereby generating stronger feature representations. Our proposed solution was tested through extensive experiments on challenging datasets, including DOTA (77.44% mAP), HRSC2016 (97.82% mAP), UCAS-AOD (91.34% mAP), as well as ICDAR2015 (86.27% F-score) and its effectiveness and high applicability are verified on all the above datasets. |
| first_indexed | 2024-03-10T20:55:25Z |
| format | Article |
| id | doaj.art-293a1ce101f64c5a89d57c6603c2494d |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T20:55:25Z |
| publishDate | 2023-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-293a1ce101f64c5a89d57c6603c2494d2023-11-19T17:58:56ZengMDPI AGRemote Sensing2072-42922023-10-011520496510.3390/rs15204965AFRE-Net: Adaptive Feature Representation Enhancement for Arbitrary Oriented Object DetectionTianwei Zhang0Xu Sun1Lina Zhuang2Xiaoyu Dong3Jianjun Sha4Bing Zhang5Ke Zheng6Key Laboratory of Computational Optical Imaging Technology, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, ChinaKey Laboratory of Digital Earth Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, ChinaKey Laboratory of Digital Earth Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, ChinaDepartment of Complexity Science and Engineering, The University of Tokyo, Tokyo 277-8561, JapanCollege of Intelligent Systems Science and Engineering, Harbin Engineering University, Harbin 150001, ChinaCollege of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, ChinaCollege of Geography and Environment, Liaocheng University, Liaocheng 252059, ChinaArbitrary-oriented object detection (AOOD) is a crucial task in aerial image analysis but is also faced with significant challenges. In current AOOD detectors, commonly used multi-scale feature fusion modules fall short in spatial and semantic information complement between scales. Additionally, fixed feature extraction structures are usually used following a fusion model, resulting in the inability of detectors to self-adjust. At the same time, feature fusion and extraction modules are designed in isolation and the internal synergy between them is ignored. The above problems result in feature representation deficiency, thus affecting the overall detection precision. To solve these problems, we first create a fine-grained feature pyramid network (FG-FPN) that not only provides richer spatial and semantic features, but also completes neighbor scale features in a self-learning mode. Subsequently, we propose a novel feature enhancement module (FEM) to fit FG-FPN. FEM authorizes the detection unit to automatically adjust the sensing area and adaptively suppress background interference, thereby generating stronger feature representations. Our proposed solution was tested through extensive experiments on challenging datasets, including DOTA (77.44% mAP), HRSC2016 (97.82% mAP), UCAS-AOD (91.34% mAP), as well as ICDAR2015 (86.27% F-score) and its effectiveness and high applicability are verified on all the above datasets.https://www.mdpi.com/2072-4292/15/20/4965deep learningobject detectionremote sensingfeature representation enhancement |
| spellingShingle | Tianwei Zhang Xu Sun Lina Zhuang Xiaoyu Dong Jianjun Sha Bing Zhang Ke Zheng AFRE-Net: Adaptive Feature Representation Enhancement for Arbitrary Oriented Object Detection Remote Sensing deep learning object detection remote sensing feature representation enhancement |
| title | AFRE-Net: Adaptive Feature Representation Enhancement for Arbitrary Oriented Object Detection |
| title_full | AFRE-Net: Adaptive Feature Representation Enhancement for Arbitrary Oriented Object Detection |
| title_fullStr | AFRE-Net: Adaptive Feature Representation Enhancement for Arbitrary Oriented Object Detection |
| title_full_unstemmed | AFRE-Net: Adaptive Feature Representation Enhancement for Arbitrary Oriented Object Detection |
| title_short | AFRE-Net: Adaptive Feature Representation Enhancement for Arbitrary Oriented Object Detection |
| title_sort | afre net adaptive feature representation enhancement for arbitrary oriented object detection |
| topic | deep learning object detection remote sensing feature representation enhancement |
| url | https://www.mdpi.com/2072-4292/15/20/4965 |
| work_keys_str_mv | AT tianweizhang afrenetadaptivefeaturerepresentationenhancementforarbitraryorientedobjectdetection AT xusun afrenetadaptivefeaturerepresentationenhancementforarbitraryorientedobjectdetection AT linazhuang afrenetadaptivefeaturerepresentationenhancementforarbitraryorientedobjectdetection AT xiaoyudong afrenetadaptivefeaturerepresentationenhancementforarbitraryorientedobjectdetection AT jianjunsha afrenetadaptivefeaturerepresentationenhancementforarbitraryorientedobjectdetection AT bingzhang afrenetadaptivefeaturerepresentationenhancementforarbitraryorientedobjectdetection AT kezheng afrenetadaptivefeaturerepresentationenhancementforarbitraryorientedobjectdetection |