Bi2F-YOLO: a novel framework for underwater object detection based on YOLOv7

Abstract

Underwater object detection faces significant challenges, including ambiguity and occlusion, which greatly undermine the accuracy of traditional algorithms. To address these issues, we propose Bi2F-YOLO, an algorithm, specifically designed for underwater environments. Bi2F-YOLO integrates the BiFormer module into the YOLOv7 backbone, utilizing Bi-Level routing attention (BRA) to focus on key features such as object edges and textures. This effectively addresses the problem of object ambiguity. In the detection head, we replace the conventional ELAN component with the FasterNet module. This update enhances detection efficiency and accuracy through the use of partial convolution (PConv), which redistributes the convolution kernel weights based on the sparsity of the input feature map. By doing so, it prevents the dilution of critical underwater object features caused by interference from irrelevant data. This effectively resolves the occlusion problem in underwater target detection while simultaneously reducing model parameters and computational costs. The experimental results show that Bi2F-YOLO achieves 87.3%

{\mathit{mAP}}_{50}

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$${{mAP}}_{50}$$\end{document}

and 86.8%

{\mathit{AP}}_{50}

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$${{AP}}_{50}$$\end{document}

on the RUOD dataset. Compared to existing methods, it reduces model parameters by 14.3% fewer parameters (Params) and lowers computational costs by 5.5%, highlighting its efficiency and accuracy in underwater object detection.

Keywords

FasterNet / BiFormer / YOLO / Loss function / Underwater object detection / Information and Computing Sciences / Artificial Intelligence and Image Processing

References

[1] Chen GQ, Mao ZY, Wang K, Shen JE (2023a) HTDet: a hybrid transformer-based approach for underwater small object detection. Remote Sens 15(4):1076

[2] Chen JR, Kao SH, He H, Zhuo WP, Wen S, Lee CH et al (2023b) Run, don’t walk: chasing higher FLOPS for faster neural networks. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Vancouver, pp 12021–12031

[3] ChenL, ZhouFX, WangSK, DongJY, LiN, MaHP, et al. . SWIPENET: object detection in noisy underwater scenes. Pattern Recognit, 2022, 132: 108926

[4] ChenZ, GaoHM, ZhangZ, ZhouHL, WangX, TianY. Underwater salient object detection by combining 2D and 3D visual features. Neurocomputing, 2020, 391: 249-259

[5] Dosovitskiy A, Beyer L, Kolesnikov A, Weissenborn D, Zhai XH, Unterthiner T et al (2020) An image is worth 16x16 words: transformers for image recognition at scale. Preprint at arXiv:2010.11929

[6] Fu CP, Fan X, Xiao JW, Yuan WQ, Liu RS, Luo ZX (2023a) Learning heavily-degraded prior for underwater object detection. IEEE Trans Circuits Syst Video Technol 33(11):6887–6896

[7] Fu CP, Liu RS, Fan X, Chen PY, Fu H, Yuan WQ et al (2023b) Rethinking general underwater object detection: datasets, challenges, and solutions. Neurocomputing 517:243–256

[8] GaoJX, ZhangYH, GengX, TangH, BhattiUA. PE-Transformer: path enhanced transformer for improving underwater object detection. Expert Syst Appl, 2024, 246: 123253

[9] HuaX, CuiXP, XuXH, QiuSH, LiangYJ, BaoXQ, et al. . Underwater object detection algorithm based on feature enhancement and progressive dynamic aggregation strategy. Pattern Recognit, 2023, 139: 109511

[10] JiaJQ, FuM, LiuXF, ZhengB. Underwater object detection based on improved efficientdet. Remote Sens, 2022, 14184487

[11] JianMW, LiuXY, LuoHJ, LuXW, YuH, DongJY. Underwater image processing and analysis: a review. Signal Proc-Image Commun, 2021, 91: 116088

[12] Jocher G, Stoken A, Borovec J, Changyu L, Hogan A, Diaconu L et al (2020) Ultralytics yolov5. https://github.com/ultralytics/yolov5. Accessed 19 Dec 2024

[13] LiAL, YuL, TianSW. Underwater biological detection based on YOLOv4 combined with channel attention. J Mar Sci Eng, 2022, 104469

[14] Liang XT, Song PH (2022) Excavating RoI attention for underwater object detection. In: 2022 IEEE International Conference on Image Processing (ICIP), Bordeaux, pp 2651–2655

[15] Lin TY, Maire M, Belongie S, Hays J, Perona P, Ramanan D et al (2014) Microsoft COCO: common objects in context. In: 13th European Conference on Computer Vision (ECCV), Zurich, pp 740–755

[16] Lin WH, Zhong JX, Liu S, Li T, Li G (2020) ROIMIX: proposal-fusion among multiple images for underwater object detection. In: 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Barcelona, pp 2588–2592

[17] LiuJ, LiuS, XuSJ, ZhouCJ. Two-stage underwater object detection network using swin transformer. IEEE Access, 2022, 10: 117235-117247

[18] LiuK, PengL, TangSR. Underwater object detection using TC-YOLO with attention mechanisms. Sensors, 2023, 2352567

[19] Liu Z, Lin YT, Cao Y, Hu H, Wei YX, Zhang Z et al (2021) Swin Transformer: hierarchical vision Transformer using shifted windows. In: 18th IEEE/CVF International Conference on Computer Vision (ICCV), Montreal, pp 10012–10022

[20] LiuZY, WangB, LiY, HeJX, LiYF. UnitModule: a lightweight joint image enhancement module for underwater object detection. Pattern Recognit, 2024, 151: 110435

[21] MaHP, ZhangYJ, SunSY, ZhangWJ, FeiMR, ZhouHY. Weighted multi-error information entropy based you only look once network for underwater object detection. Eng Appl Artif Intell, 2024, 130: 107766

[22] Rezatofighi H, Tsoi N, Gwak J, Sadeghian A, Reid I, Savarese S (2019) Generalized intersection over union: a metric and a loss for bounding box regression. In: 32nd IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, pp 658–666

[23] ShenX, SunXD, WangHB, FuXP. Multi-dimensional, multi-functional and multi-level attention in YOLO for underwater object detection. Neural Comput Appl, 2023, 352719935-19960

[24] SongPH, LiPT, DaiLH, WangT, ChenZ. Boosting R-CNN: reweighting R-CNN samples by RPN’s error for underwater object detection. Neurocomputing, 2023, 530: 150-164

[25] Sun YQ, Wang X, Zheng Y, Yao L, Qi SH, Tang LL (2022) Underwater object detection with swin transformer. In: 2022 4th International Conference on Data Intelligence and Security (ICDIS), Shenzhen, pp 422–427

[26] Wang BW, Wang XQ, Wang YJ (2023a) UOD-FPN: an effective feature pyramid network for underwater object detection. In: 2023 International Conference on New Trends in Computational Intelligence (NTCI), Qingdao, pp 122–126

[27] Wang CY, Bochkovskiy A, Liao HYM (2023b) YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. Preprint at arXiv:2207.02696

[28] WenG, LiSB, LiuFC, LuoXY, ErMJ, MahmudM, et al. . YOLOv5s-CA: a modified YOLOv5s network with coordinate attention for underwater target detection. Sensors, 2023, 2373367

[29] XueLZ, ZengXY, JinAQ. A novel deep-learning method with channel attention mechanism for underwater target recognition. Sensors, 2022, 22155492

[30] YehCH, LinCH, KangLW, HuangCH, LinMH, ChangCY, et al. . Lightweight deep neural network for joint learning of underwater object detection and color conversion. IEEE Trans Neural Netw Learn Syst, 2022, 33116129-6143

[31] ZhangMH, XuSB, SongW, HeQ, WeiQM. Lightweight underwater object detection based on YOLO v4 and multi-scale attentional feature fusion. Remote Sens, 2021, 13224706

[32] ZhangYF, RenWQ, ZhangZ, JiaZ, WangL, TanTN. Focal and efficient IOU loss for accurate bounding box regression. Neurocomputing, 2022, 506: 146-157

[33] Zheng ZH, Wang P, Liu W, Li JZ, Ye RG, Ren DW (2020) Distance-IoU loss: faster and better learning for bounding box regression. In: 34th AAAI Conference on Artificial Intelligence, New York, pp 12993–13000

[34] Zhou JC, He ZX, Lam KM, Wang YD, Zhang WS, Guo CL et al (2024) AMSP-UOD: when vortex convolution and stochastic perturbation meet underwater object detection. In: 38th AAAI Conference on Artificial Intelligence, Vancouver, pp 7659–7667

[35] Zhu L, Wang XJ, Ke ZH, Zhang W, Lau R (2023) BiFormer: vision transformer with Bi-level routing attention. In: 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Vancouver, pp 10323–10333

RIGHTS & PERMISSIONS

The Author(s)