Attentional feature pyramid network for small object detection
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Min, Kyungseo | - |
dc.contributor.author | Lee, Gun-Hee | - |
dc.contributor.author | Lee, Seong-Whan | - |
dc.date.accessioned | 2022-12-08T20:41:49Z | - |
dc.date.available | 2022-12-08T20:41:49Z | - |
dc.date.created | 2022-12-08 | - |
dc.date.issued | 2022-11 | - |
dc.identifier.issn | 0893-6080 | - |
dc.identifier.uri | https://scholar.korea.ac.kr/handle/2021.sw.korea/146516 | - |
dc.description.abstract | Recent state-of-the-art detectors generally exploit the Feature Pyramid Networks (FPN) due to its advantage of detecting objects at different scales. Despite significant advances in object detection owing to the design of feature pyramids, it is still challenging to detect small objects with low resolution and dense distribution in complex scenes. To address these problems, we propose Atten-tional Feature Pyramid Network, a new feature pyramid architecture named AFPN which consists of three components to enhance the small object detection ability, specifically: Dynamic Texture Attention, Foreground-Aware Co-Attention, and Detail Context Attention. First, Dynamic Texture Attention augments the texture features dynamically by filtering out redundant semantics to highlight small objects in lower layers and amplifying credible details to emphasize large objects in higher layers. Then, Foreground-Aware Co-Attention is explored to detect densely arranged small objects by enhancing the objects feature via foreground-correlated contexts and suppressing the background noise. Finally, to better capture the features of small objects, Detail Context Attention adaptively aggregates detail cues of RoI features with different scales for a more accurate feature representation. By substituting FPN with AFPN in Faster R-CNN, our method performs on par with the state-of-the-art performance on Tsinghua-Tencent 100K. Furthermore, we achieve highly competitive results on small category of both PASCAL VOC and MS COCO. (c) 2022 Elsevier Ltd. All rights reserved. | - |
dc.language | English | - |
dc.language.iso | en | - |
dc.publisher | PERGAMON-ELSEVIER SCIENCE LTD | - |
dc.subject | RECOGNITION | - |
dc.title | Attentional feature pyramid network for small object detection | - |
dc.type | Article | - |
dc.contributor.affiliatedAuthor | Lee, Seong-Whan | - |
dc.identifier.doi | 10.1016/j.neunet.2022.08.029 | - |
dc.identifier.scopusid | 2-s2.0-85138203026 | - |
dc.identifier.wosid | 000865423600013 | - |
dc.identifier.bibliographicCitation | NEURAL NETWORKS, v.155, pp.439 - 450 | - |
dc.relation.isPartOf | NEURAL NETWORKS | - |
dc.citation.title | NEURAL NETWORKS | - |
dc.citation.volume | 155 | - |
dc.citation.startPage | 439 | - |
dc.citation.endPage | 450 | - |
dc.type.rims | ART | - |
dc.type.docType | Article | - |
dc.description.journalClass | 1 | - |
dc.description.journalRegisteredClass | scie | - |
dc.description.journalRegisteredClass | scopus | - |
dc.relation.journalResearchArea | Computer Science | - |
dc.relation.journalResearchArea | Neurosciences & Neurology | - |
dc.relation.journalWebOfScienceCategory | Computer Science, Artificial Intelligence | - |
dc.relation.journalWebOfScienceCategory | Neurosciences | - |
dc.subject.keywordPlus | RECOGNITION | - |
dc.subject.keywordAuthor | Object detection | - |
dc.subject.keywordAuthor | Small object detection | - |
dc.subject.keywordAuthor | Feature pyramid network | - |
dc.subject.keywordAuthor | Attention mechanism | - |
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.
(02841) 서울특별시 성북구 안암로 14502-3290-1114
COPYRIGHT © 2021 Korea University. All Rights Reserved.
Certain data included herein are derived from the © Web of Science of Clarivate Analytics. All rights reserved.
You may not copy or re-distribute this material in whole or in part without the prior written consent of Clarivate Analytics.