摘 要：文章针对小目标检测存在的可利用特征少、定位精度要求高、数据集小目标占比少、样本不均衡和小目标对象聚集等问题，提出将 coordinate attention 注意力嵌入 YOLOv5 模型。Coordinate attention 注意力机制通过获取位置感知和方向感知的信息，能使 YOLOv5 模型更准确地识别和定位感兴趣的目标。YOLOv5 改进模型采用木虱和 VisDrone2019 数据集开展实验验证，实验结果表明嵌入 coordinate attention 能有效提高 YOLOv5 的算法性能。
关键词：目标检测；YOLOv5；coordinate attention；注意力机制
DOI:10.19850/j.cnki.2096-4706.2023.03.013
中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2023）03-0055-07
Research on Improved Algorithm of Small Target Detection Based on YOLOv5
CHEN Furong1, XIAO Mingming2
(1.College of Information Science and Technology, Zhongkai University of Agricultural and Engineering, Guangzhou 510225, China; 2.College of Information and Communication Engineering, Guangzhou Maritime University, Guangzhou 510725, China)
Abstract: Aiming at the problems of small target detection, such as few available features, requirement of high positioning accuracy, small proportion of small target in data set, unbalanced samples and small target aggregation, this paper proposes to embed coordinate attention into YOLOv5 model. Coordinated attention mechanism can enable YOLOv5 model to identify and locate interested targets more accurately by obtaining information of location awareness and direction awareness. The improved YOLOv5 model uses psyllid and VisDrone 2019 datasets to carry out experiments to verify, and the experimental results show that embedding coordinate attention can effectively improve the algorithm performance of YOLOv5.
Keywords: target detection; YOLOv5; coordinate attention; attention mechanism
参考文献：
[1] KISANTAL M，WOJNA Z，MURAWSKI J，et al. Augmentation for small object detection [J/OL].arXiv：1902.07296 [cs. CV].（2019-02-19）.https://arxiv.org/abs/1902.07296v1.
[2] LIN T Y，DOLLAR P,GIRSHICK R. Feature Pyramid Networks for Object Detection [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu：IEEE，2017：936-944.
[3] NOH J，BAE W，LEE W，et al. Better to Follow，Follow to Be Better：Towards Precise Supervision of Feature Super-Resolution for Small Object Detection[C]//2019 IEEE/CVF International Conference on 图 13 增加 coordinate attention 的YOLOv5 的平均精度均值图 14 YOLOv5 的召回率 图 15 增加 coordinate attention 的YOLOv5 的召回率Computer Vision （ICCV）.Seoul：IEEE，2019：9724-9733.
[4] 贾可心，马正华，朱蓉，等 . 注意力机制改进轻量 SSD模型的海面小目标检测 [J]. 中国图象图形学报，2022，27（4）：1161-1175.
[5] ZHU X，SU W，LU L，et al. Deformable DETR： Deformable Transformers for End-to-End Object Detection [J/OL]. arXiv：2010.04159 [cs.CV].（2020-10-08）.https://arxiv.org/ abs/2010.04159.
[6] XU C，WANG J，YANG W，et al. RFLA：Gaussian Receptive Field Based Label Assignment for Tiny Object Detection [C]// Computer Vision-ECCV 2022.Cham：Springer，2022：526-543.
[7] WANG J，XU C，YANG W，et al. A Normalized Gaussian Wasserstein Distance for Tiny Object Detection [J/OL].arXiv： 2110.13389 [cs.CV].（2021-10-26）.https://arxiv.org/abs/2110.13389v2.
[8] YANG C，HUANG Z，WANG N. QueryDet：Cascaded Sparse Query for Accelerating High-Resolution Small Object Detection [C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）.IEEE：New Orleans，2022：13658-13667.
[9] 窦其龙，颜明重，朱大奇 . 基于 YOLO-v5 的星载 SAR 图像海洋小目标检测 [J]. 应用科技，2021，48（6）：1-7.
[10] 刘闪亮，吴仁彪，屈景怡，等 . 基于 A-YOLOv5s 的机场小目标检测方法 [J/OL]. 安全与环境学报：1-8[2022-08-01].DOI：10.13637/j.issn.1009-6094.2022.0819.
[11] 田枫，贾昊鹏，刘芳 . 改进 YOLOv5 的油田作业现场安全着装小目标检测 [J]. 计算机系统应用，2022，31（3）：159-168.
[12] 奉志强，谢志军，包正伟，等 . 基于改进 YOLOv5 的无人机实时密集小目标检测算法 [J/OL]. 航空学报：1-15[2022-09-01].http://kns.cnki.net/kcms/detail/11.1929.V.20220509.2316.010.html.
[13] HOU Q，ZHOU D，FENG J. Coordinate Attention for Efficient Mobile Network Design [C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）.Nashville： IEEE，2021：13708-13717.
[14] HU J，SHEN L，ALBANIE S，et al. Squeeze-andExcitation [J].Networks IEEE Transactions on Pattern Analysis and Machine Intelligence，2019，42（8）：2011-2023.
[15] WOO S，PARK J，LEE J，et al. CBAM：Convolutional Block Attention Module[C]//ECCV 2018.Munich：Springer，2018：3-19.
[16] GIRSHICK R. Fast R-CNN [C]//2015 IEEE International Conference on Computer Vision （ICCV）.Santiago：IEEE，2015：1440-1448.
[17] HE K，GKIOXARI G，DOLLAR P，et al. Mask R-CNN [C]//2017 IEEE International Conference on Computer Vision （ICCV）. Venice：IEEE，2017：2980-2988.
[18] BAI Y，ZHANG Y，DING M，et al. SOD-MTGAN：Small Object Detection via Multi-Task Generative Adversarial Network [C]// Computer Vision–ECCV 2018.Munich：Springer，2018：210–226.
[19] LI J，LIANG X，WEI Y，et al. Perceptual Generative Adversarial Networks for Small Object Detection [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. Honolulu：IEEE，2017：1951-1959.
[20] CHEN G，CHOI W，YU X，et al. Learning efficient object detection models with knowledge distillation [C]//NIPS'17：Proceedings of the 31st International Conference on Neural Information Processing Systems.Long Beach：Curran Associates Inc.，2017：742-751.
[21] CARION N，MASSA F，SYNNAEVE G，et al. End-toEnd Object Detection with Transformers [C]//Computer Vision-ECCV 2020.Cham：Springer，2020：213–229.
[22] LIN T，DOLLAR P，GIRSHICK R，et al. Feature Pyramid Networks for Object Detection [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition（CVPR）.Honolulu：IEEE，2017：936-994.
[23] LUC C，MINH-TAN P，SEBASTIEN L. Small Object Detection in Remote Sensing Images Based on Super-Resolution with Auxiliary Generative Adversarial Networks [J/OL].Remote Sensing， 2020，12（19）：3152（2020-08-25）.https://doi.org/10.3390/ rs12193152.
[24] HE K，ZHANG X，REN S，et al. Deep Residual Learning for Image Recognition [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition（CVPR）.Las Vegas：IEEE，2016：770-778.
作者简介：陈富荣（1995—），男，汉族，硕士研究生在读，研究方向：计算机视觉；通讯作者：肖明明（1972—），男，汉族，广东三水人，教授，博士研究生，研究方向：计算机视觉。