A Lightweight SAR Ship Detection Network Based on Deep Multiscale Grouped Convolution, Network Pruning, and Knowledge Distillation
Deep learning has proven to be highly effective in synthetic aperture radar (SAR) image target detection. However, many latest deep learning models have predominantly focused on increasing depth and size to enhance detection accuracy, often ignoring the balance between accuracy and detection speed,...
Saved in:
| Main Authors: | , |
|---|---|
| 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/10759720/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850041467410579456 |
|---|---|
| author | Boyi Hu Hongxia Miao |
| author_facet | Boyi Hu Hongxia Miao |
| author_sort | Boyi Hu |
| collection | DOAJ |
| description | Deep learning has proven to be highly effective in synthetic aperture radar (SAR) image target detection. However, many latest deep learning models have predominantly focused on increasing depth and size to enhance detection accuracy, often ignoring the balance between accuracy and detection speed, as well as the practical deployment of these models on hardware platforms. Therefore, a lightweight algorithm for SAR ship detection is designed in this article. First, we propose a preliminary lightweight scheme, including a multiscale feature learning augmented backbone, a lightweight feature fusion neck, and a parameter-sharing lightweight detection head. Second, unimportant branches of the network are pruned to further compress the model. Finally, the detection accuracy of the model is enhanced by knowledge distillation without augmenting the model volume, which compensates for the accuracy loss caused by model compression. Experimental validation is conducted on three SAR image ship detection datasets (SSDD, high-resolution SAR images dataset, large-scale SAR ship detection dataset-v1.0) to thoroughly assess the effectiveness of the proposed lightweight algorithm. Experimental results on the three datasets demonstrate that the proposed method achieves a model volume reduction to one-third of the baseline while maintaining a minimal decrease in detection accuracy. In SSDD, the proposed method achieved 98.7 accuracy, 0.92M parameters, 3.1G FLOPS and 2.1 MB size of 1.5X pruning rate. Furthermore, it outperforms other state-of-the-art lightweight detectors. |
| format | Article |
| id | doaj-art-cdfad6852ddb4cf59e9ac89ea2c593e3 |
| institution | DOAJ |
| 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-cdfad6852ddb4cf59e9ac89ea2c593e32025-08-20T02:55:46ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-01182190220710.1109/JSTARS.2024.350217210759720A Lightweight SAR Ship Detection Network Based on Deep Multiscale Grouped Convolution, Network Pruning, and Knowledge DistillationBoyi Hu0https://orcid.org/0009-0004-0865-2596Hongxia Miao1https://orcid.org/0000-0003-4808-0989State Key Laboratory of Networking and Switching Technology, School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing, ChinaState Key Laboratory of Networking and Switching Technology, School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing, ChinaDeep learning has proven to be highly effective in synthetic aperture radar (SAR) image target detection. However, many latest deep learning models have predominantly focused on increasing depth and size to enhance detection accuracy, often ignoring the balance between accuracy and detection speed, as well as the practical deployment of these models on hardware platforms. Therefore, a lightweight algorithm for SAR ship detection is designed in this article. First, we propose a preliminary lightweight scheme, including a multiscale feature learning augmented backbone, a lightweight feature fusion neck, and a parameter-sharing lightweight detection head. Second, unimportant branches of the network are pruned to further compress the model. Finally, the detection accuracy of the model is enhanced by knowledge distillation without augmenting the model volume, which compensates for the accuracy loss caused by model compression. Experimental validation is conducted on three SAR image ship detection datasets (SSDD, high-resolution SAR images dataset, large-scale SAR ship detection dataset-v1.0) to thoroughly assess the effectiveness of the proposed lightweight algorithm. Experimental results on the three datasets demonstrate that the proposed method achieves a model volume reduction to one-third of the baseline while maintaining a minimal decrease in detection accuracy. In SSDD, the proposed method achieved 98.7 accuracy, 0.92M parameters, 3.1G FLOPS and 2.1 MB size of 1.5X pruning rate. Furthermore, it outperforms other state-of-the-art lightweight detectors.https://ieeexplore.ieee.org/document/10759720/Convolutional neural network (CNN)knowledge distillation (KD)lightweightnetwork pruningsynthetic aperture radar (SAR) target detection |
| spellingShingle | Boyi Hu Hongxia Miao A Lightweight SAR Ship Detection Network Based on Deep Multiscale Grouped Convolution, Network Pruning, and Knowledge Distillation IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Convolutional neural network (CNN) knowledge distillation (KD) lightweight network pruning synthetic aperture radar (SAR) target detection |
| title | A Lightweight SAR Ship Detection Network Based on Deep Multiscale Grouped Convolution, Network Pruning, and Knowledge Distillation |
| title_full | A Lightweight SAR Ship Detection Network Based on Deep Multiscale Grouped Convolution, Network Pruning, and Knowledge Distillation |
| title_fullStr | A Lightweight SAR Ship Detection Network Based on Deep Multiscale Grouped Convolution, Network Pruning, and Knowledge Distillation |
| title_full_unstemmed | A Lightweight SAR Ship Detection Network Based on Deep Multiscale Grouped Convolution, Network Pruning, and Knowledge Distillation |
| title_short | A Lightweight SAR Ship Detection Network Based on Deep Multiscale Grouped Convolution, Network Pruning, and Knowledge Distillation |
| title_sort | lightweight sar ship detection network based on deep multiscale grouped convolution network pruning and knowledge distillation |
| topic | Convolutional neural network (CNN) knowledge distillation (KD) lightweight network pruning synthetic aperture radar (SAR) target detection |
| url | https://ieeexplore.ieee.org/document/10759720/ |
| work_keys_str_mv | AT boyihu alightweightsarshipdetectionnetworkbasedondeepmultiscalegroupedconvolutionnetworkpruningandknowledgedistillation AT hongxiamiao alightweightsarshipdetectionnetworkbasedondeepmultiscalegroupedconvolutionnetworkpruningandknowledgedistillation AT boyihu lightweightsarshipdetectionnetworkbasedondeepmultiscalegroupedconvolutionnetworkpruningandknowledgedistillation AT hongxiamiao lightweightsarshipdetectionnetworkbasedondeepmultiscalegroupedconvolutionnetworkpruningandknowledgedistillation |