A Few-shot Learning Algorithm for Underwater Small Object Detection based on the Transformer Architecture
DOI:
https://doi.org/10.15837/ijccc.2026.4.6915Keywords:
sonar image, underwater small object detection, MP-DETR, transfer learningAbstract
This paper proposes a few-shot learning approach based on the transformer model to tackle the challenges of underwater small object detection in sonar images using deep learning techniques. We analyze the difficulties encountered in underwater small object detection with deep learning methods when training samples are limited and propose corresponding solutions to address these challenges. To address the challenges of unclear object contours and blurred feature details in sonar images, we employ a transformer-based object detector that leverages the attention mechanism to effectively utilize global image information for precise object localization and identification. To handle the prevalence of small targets, we design a Multi-Branch Feature Extraction Module that aggregates feature maps with different receptive fields, thereby enhancing the effective utilization of small object features. In addition, the core design of the RT-DETR model is incorporated as the baseline architecture, which significantly improves the real-time performance of the detector. To overcome the challenge of limited training data, we apply transfer learning by first training the entire MP-DETR (More Precise DETR) network on a large-scale, general-purpose dataset. Then the IOU-aware query selection module and detection head of the MP-DETR network is finetuned using a self-compiled underwater small sample sonar image training dataset. The proposed approach achieves a deep neural network suitable for underwater object detection based on sonar images. Experiments were conducted on a self-constructed underwater small-sample sonar image dataset, and the proposed MP-DETR achieved 98.5% mean average precision (MAP) and 53 frames per second (FPS) real-time performance, which provides higher detection accuracy and real-time performance compared with existing methods.
References
Sixian Cai, Guocheng Li, and Yuan Shan. Underwater object detection using collaborative weakly supervision. Computers and Electrical Engineering, 102:108159, 2022. https://doi.org/10.1016/j.compeleceng.2022.108159
Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to-end object detection with transformers. In European conference on computer vision, pages 213-229. Springer, 2020. https://doi.org/10.1007/978-3-030-58452-8_13
Linhui Dai, Hong Liu, Hao Tang, Zhiwei Wu, and Pinhao Song. Ao2-detr: Arbitrary-oriented object detection transformer. IEEE Transactions on Circuits and Systems for Video Technology, 33(5):2342-2356, 2022. https://doi.org/10.1109/TCSVT.2022.3222906
Xiyang Dai, Yinpeng Chen, Jianwei Yang, Pengchuan Zhang, Lu Yuan, and Lei Zhang. Dynamic detr: End-to-end object detection with dynamic attention. In Proceedings of the IEEE/CVF international conference on computer vision, pages 2988-2997, 2021.
Xinnan Fan, Liang Lu, Pengfei Shi, and Xuewu Zhang. A novel sonar target detection and classification algorithm. Multimedia Tools and Applications, 81(7):10091-10106, 2022. https://doi.org/10.1007/s11042-022-12054-4
Sheezan Fayaz, Shabir A Parah, and GJ Qureshi. Underwater object detection: architectures and algorithms-a comprehensive review. Multimedia Tools and Applications, 81(15):20871-20916, 2022. https://doi.org/10.1007/s11042-022-12502-1
Z Ge. Yolox: Exceeding yolo series in 2021. arXiv preprint arXiv:2107.08430, 2021.
Stanisław Hożyń. A review of underwater mine detection and classification in sonar imagery. Electronics, 10(23):2943, 2021. https://doi.org/10.3390/electronics10232943
Guanying Huo, Ziyin Wu, and Jiabiao Li. Underwater object classification in sidescan sonar images using deep transfer learning and semisynthetic training data. IEEE access, 8:47407-47418, 2020. https://doi.org/10.1109/ACCESS.2020.2978880
Glenn Jocher, Ayush Chaurasia, Alex Stoken, Jirka Borovec, Yonghye Kwon, Kalen Michael, Jiacong Fang, Zeng Yifu, Colin Wong, Diego Montes, et al. ultralytics/yolov5: v7. 0-yolov5 sota realtime instance segmentation. Zenodo, 2022.
Divas Karimanzira, Helge Renkewitz, David Shea, and Jan Albiez. Object detection in sonar images. Electronics, 9(7):1180, 2020. https://doi.org/10.3390/electronics9071180
Jaskirat Kaur and Williamjeet Singh. Tools, techniques, datasets and application areas for object detection in an image: a review. Multimedia Tools and Applications, 81(27):38297-38351, 2022. https://doi.org/10.1007/s11042-022-13153-y
Hoang Thanh Le, Son Lam Phung, Philip B Chapple, Abdesselam Bouzerdoum, Christian H Ritz, et al. Deep gabor neural network for automatic detection of mine-like objects in sonar imagery. IEEE Access, 8:94126-94139, 2020. https://doi.org/10.1109/ACCESS.2020.2995390
Chuyi Li, Lulu Li, Hongliang Jiang, Kaiheng Weng, Yifei Geng, Liang Li, Zaidan Ke, Qingyuan Li, Meng Cheng, Weiqiang Nie, et al. Yolov6: A single-stage object detection framework for industrial applications. arXiv preprint arXiv:2209.02976, 2022.
Feng Li, Hao Zhang, Shilong Liu, Jian Guo, Lionel M Ni, and Lei Zhang. Dn-detr: Accelerate detr training by introducing query denoising. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 13619-13627, 2022.
Yujie Li, Shinya Takahashi, and Seiichi Serikawa. Cognitive ocean of things: a comprehensive review and future trends. Wireless Networks, pages 1-10, 2022.
Kun Liu, Lei Peng, and Shanran Tang. Underwater object detection using tc-yolo with attention mechanisms. Sensors, 23(5):2567, 2023. https://doi.org/10.3390/s23052567
Shilong Liu, Feng Li, Hao Zhang, Xiao Yang, Xianbiao Qi, Hang Su, Jun Zhu, and Lei Zhang. Dab-detr: Dynamic anchor boxes are better queries for detr. arXiv preprint arXiv:2201.12329, 2022.
Yen-Cheng Liu, Chih-Yao Ma, Zijian He, Chia-Wen Kuo, Kan Chen, Peizhao Zhang, Bichen Wu, Zsolt Kira, and Peter Vajda. Unbiased teacher for semi-supervised object detection. arXiv preprint arXiv:2102.09480, 2021.
Huimin Lu, Dong Wang, Yujie Li, Jianru Li, Xin Li, Hyoungseop Kim, Seiichi Serikawa, and Iztok Humar. Conet: A cognitive ocean network. IEEE Wireless Communications, 26(3):90-96, 2019. https://doi.org/10.1109/MWC.2019.1800325
Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, and Jingdong Wang. Conditional detr for fast training convergence. In Proceedings of the IEEE/CVF international conference on computer vision, pages 3651-3660, 2021. https://doi.org/10.1109/ICCV48922.2021.00363
Ali Tariq Nagi, Mazhar Javed Awan, Mazin Abed Mohammed, Amena Mahmoud, Arnab Majumdar, and Orawit Thinnukool. Performance analysis for covid-19 diagnosis using custom and state-of-the-art deep learning models. Applied Sciences, 12(13):6364, 2022. https://doi.org/10.3390/app12136364
Duong Nguyen, Matthieu Simonin, Guillaume Hajduch, Rodolphe Vadaine, Cédric Tedeschi, and Ronan Fablet. Detection of abnormal vessel behaviours from ais data using geotracknet: from the laboratory to the ocean. In 2020 21st IEEE International Conference on Mobile Data Management (MDM), pages 264-268. IEEE, 2020. https://doi.org/10.1109/MDM48529.2020.00061
Son Lam Phung, Thi Nhat Anh Nguyen, Hoang Thanh Le, Philip B Chapple, Christian H Ritz, Abdesselam Bouzerdoum, and Le Chung Tran. Mine-like object sensing in sonar imagery with a compact deep learning architecture for scarce data. In 2019 Digital Image Computing: Techniques and Applications (DICTA), pages 1-7. IEEE, 2019. https://doi.org/10.1109/DICTA47822.2019.8945982
Joseph Redmon. Yolov3: An incremental improvement. arXiv preprint arXiv:1804.02767, 2018.
Byungseok Roh, JaeWoong Shin, Wuhyun Shin, and Saehoon Kim. Sparse detr: Efficient endto- end object detection with learnable sparsity. arXiv preprint arXiv:2111.14330, 2021.
Akshita Saini and Mantosh Biswas. Object detection in underwater image by detecting edges using adaptive thresholding. In 2019 3rd International Conference on Trends in Electronics and Informatics (ICOEI), pages 628-632. IEEE, 2019. https://doi.org/10.1109/ICOEI.2019.8862794
Steve Schmit. Cut and Capture System Technology for Demilitarization of Underwater Munitions. IEEE, 2022. https://doi.org/10.1109/OCEANS47191.2022.9977009
Pengfei Shi, Huanru Sun, Xinnan Fan, Qi He, Xuan Zhou, and Liang Lu. An effective automatic object detection algorithm for continuous sonar image sequences. Multimedia Tools and Applications, 83(4):10233-10246, 2024. https://doi.org/10.1007/s11042-023-15837-5
Pinhao Song, Pengteng Li, Linhui Dai, Tao Wang, and Zhan Chen. Boosting r-cnn: Reweighting r-cnn samples by rpn's error for underwater object detection. Neurocomputing, 530:150-164, 2023. https://doi.org/10.1016/j.neucom.2023.01.088
Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and ZbigniewWojna. Rethinking the inception architecture for computer vision. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2818-2826, 2016. https://doi.org/10.1109/CVPR.2016.308
Hesham Tarek, Hesham Aly, Saleh Eisa, and Mohamed Abul-Soud. Optimized deep learning algorithms for tomato leaf disease detection with hardware deployment. Electronics, 11(1):140, 2022. https://doi.org/10.3390/electronics11010140
Yi Wang, Syed Muhammad Arsalan Bashir, Mahrukh Khan, Qudrat Ullah, Rui Wang, Yilin Song, Zhe Guo, and Yilong Niu. Remote sensing image super-resolution and object detection: Benchmark and state of the art. Expert Systems with Applications, 197:116793, 2022. https://doi.org/10.1016/j.eswa.2022.116793
Yingchun Wang, Jingyi Wang, Weizhan Zhang, Yufeng Zhan, Song Guo, Qinghua Zheng, and XuanyuWang. A survey on deploying mobile deep learning applications: A systemic and technical perspective. Digital Communications and Networks, 8(1):1-17, 2022. https://doi.org/10.1016/j.dcan.2021.06.001
Yingming Wang, Xiangyu Zhang, Tong Yang, and Jian Sun. Anchor detr: Query design for transformer-based detector. In Proceedings of the AAAI conference on artificial intelligence, volume 36, pages 2567-2575, 2022. https://doi.org/10.1609/aaai.v36i3.20158
Zhen Wang, Jianxin Guo, Leya Zeng, Chuanlei Zhang, and Buhong Wang. Mlffnet: Multilevel feature fusion network for object detection in sonar images. IEEE Transactions on Geoscience and Remote Sensing, 60:1-19, 2022. https://doi.org/10.1109/TGRS.2022.3214748
Mengde Xu, Zheng Zhang, Han Hu, Jianfeng Wang, Lijuan Wang, Fangyun Wei, Xiang Bai, and Zicheng Liu. End-to-end semi-supervised object detection with soft teacher. In Proceedings of the IEEE/CVF international conference on computer vision, pages 3060-3069, 2021.
Shubo Xu, Minghua Zhang, Wei Song, Haibin Mei, Qi He, and Antonio Liotta. A systematic review and analysis of deep learning-based underwater object detection. Neurocomputing, 527: 204-232, 2023. https://doi.org/10.1016/j.neucom.2023.01.056
Chao Yang, Yongpeng Li, Longyu Jiang, and Jianxing Huang. Foreground enhancement network for object detection in sonar images. Machine Vision and Applications, 34(4):56, 2023. https://doi.org/10.1007/s00138-023-01406-1
Honghui Yang, Zili Liu, Xiaopei Wu, Wenxiao Wang, Wei Qian, Xiaofei He, and Deng Cai. Graph r-cnn: Towards accurate 3d object detection with semantic-decorated local graph. In European Conference on Computer Vision, pages 662-679. Springer, 2022. https://doi.org/10.1007/978-3-031-20074-8_38
Li Yuanzi, Ye Xiufen, and Zhang Weizheng. Transyolo: high-performance object detector for forward looking sonar images. IEEE Signal Processing Letters, 29:2098-2102, 2022. https://doi.org/10.1109/LSP.2022.3210839
Tang Yulin, Shaohua Jin, Gang Bian, and Yonghou Zhang. Shipwreck target recognition in side-scan sonar images by improved yolov3 model based on transfer learning. IEEE Access, 8: 173450-173460, 2020. https://doi.org/10.1109/ACCESS.2020.3024813
Hao Zhang, Feng Li, Shilong Liu, Lei Zhang, Hang Su, Jun Zhu, Lionel M Ni, and Heung-Yeung Shum. Dino: Detr with improved denoising anchor boxes for end-to-end object detection. arXiv preprint arXiv:2203.03605, 2022.
Haoting Zhang, Mei Tian, Gaoping Shao, Juan Cheng, and Jingjing Liu. Target detection of forward-looking sonar image based on improved yolov5. IEEE Access, 10:18023-18034, 2022. https://doi.org/10.1109/ACCESS.2022.3150339
Lanyong Zhang, Chengyu Li, and Hongfang Sun. Object detection/tracking toward underwater photographs by remotely operated vehicles (rovs). Future Generation Computer Systems, 126: 163-168, 2022. https://doi.org/10.1016/j.future.2021.07.011
Yian Zhao, Wenyu Lv, Shangliang Xu, Jinman Wei, Guanzhong Wang, Qingqing Dang, Yi Liu, and Jie Chen. Detrs beat yolos on real-time object detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16965-16974, 2024. https://doi.org/10.1109/CVPR52733.2024.01605
Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, and Jifeng Dai. Deformable detr: Deformable transformers for end-to-end object detection. arXiv preprint arXiv:2010.04159, 2020.
Zhuofan Zong, Guanglu Song, and Yu Liu. Detrs with collaborative hybrid assignments training. In Proceedings of the IEEE/CVF international conference on computer vision, pages 6748-6758, 2023. https://doi.org/10.1109/ICCV51070.2023.00621
Additional Files
Published
Issue
Section
License
Copyright (c) 2026 Xiaobing Wu, Xiaoyu Zhang, Panlong Tan
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
ONLINE OPEN ACCES: Acces to full text of each article and each issue are allowed for free in respect of Attribution-NonCommercial 4.0 International (CC BY-NC 4.0.
You are free to:
-Share: copy and redistribute the material in any medium or format;
-Adapt: remix, transform, and build upon the material.
The licensor cannot revoke these freedoms as long as you follow the license terms.
DISCLAIMER: The author(s) of each article appearing in International Journal of Computers Communications & Control is/are solely responsible for the content thereof; the publication of an article shall not constitute or be deemed to constitute any representation by the Editors or Agora University Press that the data presented therein are original, correct or sufficient to support the conclusions reached or that the experiment design or methodology is adequate.
