Toward Better Accuracy-Efficiency Tradeoffs for Oriented SAR Ship Object Detection

In oriented synthetic aperture radar (SAR) ship detection task, convolutional neural network based detectors have dramatically improved the detection performance, but enormous parameters make it difficult to realize model lightweighting. Recently, DETR and its variants have demonstrated excellent pe...

Full description

Saved in:
Bibliographic Details
Main Authors: Moran Ju, Buniu Niu, Mulin Li, Tengkai Mao, Si-nian Jin
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10944503/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850203060384563200
author Moran Ju
Buniu Niu
Mulin Li
Tengkai Mao
Si-nian Jin
author_facet Moran Ju
Buniu Niu
Mulin Li
Tengkai Mao
Si-nian Jin
author_sort Moran Ju
collection DOAJ
description In oriented synthetic aperture radar (SAR) ship detection task, convolutional neural network based detectors have dramatically improved the detection performance, but enormous parameters make it difficult to realize model lightweighting. Recently, DETR and its variants have demonstrated excellent performance in object detection task, while model construction through linear layers has great potential in terms of model lightweighting. However, DETR-based models are rarely applied to oriented object detection task, while the network structure relies on manual experience and cannot be designed automatically. In this article, we propose a novel neural architecture search based lightweight detector in polar coordinate system with DETR as search space for oriented SAR ship detection, where oriented bounding boxes are encoded and decoded in polar coordinate system to cope with boundary discontinuity problems, and the weight entanglement strategy is adopted to realize automatic and lightweight design of DETR. Meanwhile, we design an oriented multiscale attention to alleviate the problem of sampling a large amount of background due to offset learning. Furthermore, we introduce a downsampling feedforward network to significantly reduce network floating point operations. Finally, we transplant FPDDet head as auxiliary head to improve encoder potential ship feature learning and decoder cross-attention learning. Experimental results show that our models not only achieve DETR lightweighting and real-time detection, but also improve detection performance. Our base models achieve state-of-the-art performance on both RSSDD and RSDD datasets compared to previous best models, with 1.36% and 2.28% improvement in mAP with 32.67 and 32.14 GFLOPs, respectively.
format Article
id doaj-art-40bbdc09fbf848cea62c1982527b9e30
institution OA Journals
issn 1939-1404
2151-1535
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-40bbdc09fbf848cea62c1982527b9e302025-08-20T02:11:37ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-01189666968110.1109/JSTARS.2025.355533010944503Toward Better Accuracy-Efficiency Tradeoffs for Oriented SAR Ship Object DetectionMoran Ju0https://orcid.org/0000-0002-3158-4956Buniu Niu1https://orcid.org/0009-0004-2472-959XMulin Li2https://orcid.org/0009-0009-6744-294XTengkai Mao3https://orcid.org/0009-0002-7044-0733Si-nian Jin4https://orcid.org/0000-0003-3226-365XCollege of Information Science and Technology, Dalian Maritime University, Dalian, ChinaCollege of Information Science and Technology, Dalian Maritime University, Dalian, ChinaCollege of Information Science and Technology, Dalian Maritime University, Dalian, ChinaCollege of Information Science and Technology, Dalian Maritime University, Dalian, ChinaCollege of Information Science and Technology, Dalian Maritime University, Dalian, ChinaIn oriented synthetic aperture radar (SAR) ship detection task, convolutional neural network based detectors have dramatically improved the detection performance, but enormous parameters make it difficult to realize model lightweighting. Recently, DETR and its variants have demonstrated excellent performance in object detection task, while model construction through linear layers has great potential in terms of model lightweighting. However, DETR-based models are rarely applied to oriented object detection task, while the network structure relies on manual experience and cannot be designed automatically. In this article, we propose a novel neural architecture search based lightweight detector in polar coordinate system with DETR as search space for oriented SAR ship detection, where oriented bounding boxes are encoded and decoded in polar coordinate system to cope with boundary discontinuity problems, and the weight entanglement strategy is adopted to realize automatic and lightweight design of DETR. Meanwhile, we design an oriented multiscale attention to alleviate the problem of sampling a large amount of background due to offset learning. Furthermore, we introduce a downsampling feedforward network to significantly reduce network floating point operations. Finally, we transplant FPDDet head as auxiliary head to improve encoder potential ship feature learning and decoder cross-attention learning. Experimental results show that our models not only achieve DETR lightweighting and real-time detection, but also improve detection performance. Our base models achieve state-of-the-art performance on both RSSDD and RSDD datasets compared to previous best models, with 1.36% and 2.28% improvement in mAP with 32.67 and 32.14 GFLOPs, respectively.https://ieeexplore.ieee.org/document/10944503/DETRneural architecture search (NAS)oriented ship detectionsynthetic aperture radar (SAR)
spellingShingle Moran Ju
Buniu Niu
Mulin Li
Tengkai Mao
Si-nian Jin
Toward Better Accuracy-Efficiency Tradeoffs for Oriented SAR Ship Object Detection
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
DETR
neural architecture search (NAS)
oriented ship detection
synthetic aperture radar (SAR)
title Toward Better Accuracy-Efficiency Tradeoffs for Oriented SAR Ship Object Detection
title_full Toward Better Accuracy-Efficiency Tradeoffs for Oriented SAR Ship Object Detection
title_fullStr Toward Better Accuracy-Efficiency Tradeoffs for Oriented SAR Ship Object Detection
title_full_unstemmed Toward Better Accuracy-Efficiency Tradeoffs for Oriented SAR Ship Object Detection
title_short Toward Better Accuracy-Efficiency Tradeoffs for Oriented SAR Ship Object Detection
title_sort toward better accuracy efficiency tradeoffs for oriented sar ship object detection
topic DETR
neural architecture search (NAS)
oriented ship detection
synthetic aperture radar (SAR)
url https://ieeexplore.ieee.org/document/10944503/
work_keys_str_mv AT moranju towardbetteraccuracyefficiencytradeoffsfororientedsarshipobjectdetection
AT buniuniu towardbetteraccuracyefficiencytradeoffsfororientedsarshipobjectdetection
AT mulinli towardbetteraccuracyefficiencytradeoffsfororientedsarshipobjectdetection
AT tengkaimao towardbetteraccuracyefficiencytradeoffsfororientedsarshipobjectdetection
AT sinianjin towardbetteraccuracyefficiencytradeoffsfororientedsarshipobjectdetection