[1]潘烨新,黄启鹏,韦 超,等.基于注意力机制的 YOLOv5 优化模型[J].计算机技术与发展,2023,33(12):163-170.[doi:10. 3969 / j. issn. 1673-629X. 2023. 12. 023]
　PAN Ye-xin,HUANG Qi-peng,WEI Chao,et al.YOLOv5 Optimization Model Based on Attention Mechanism[J].,2023,33(12):163-170.[doi:10. 3969 / j. issn. 1673-629X. 2023. 12. 023]
点击复制

基于注意力机制的 YOLOv5 优化模型()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
33
期数:
2023年12期
页码:
163-170
栏目:
人工智能
出版日期:
2023-12-10

文章信息/Info

Title:
YOLOv5 Optimization Model Based on Attention Mechanism
文章编号:
1673-629X(2023)12-0163-08
作者:
潘烨新1; 2 ; 黄启鹏1; 2 ; 韦 超1; 2 ; 杨 哲1; 2
1. 苏州大学 计算机科学与技术学院,江苏 苏州 215006;
2. 省计算机信息处理技术重点实验室,江苏 苏州 215006
Author(s):
PAN Ye-xin1; 2 ; HUANG Qi-peng1; 2 ; WEI Chao1; 2 ; YANG Zhe1; 2
1. Department of Computer Science and Technology,Soochow University,Suzhou 215006,China;
2. Jiangsu Provincial Key Laboratory for Computer Information Processing Technology,Suzhou 215006,China
关键词:
机器视觉; 深度学习; 目标检测; 注意力机制; 损失函数
Keywords:
computer vision; deep learning; object detection; attention mechanism; loss function
分类号:
TP391
DOI:
10. 3969 / j. issn. 1673-629X. 2023. 12. 023
摘要:
目标检测是机器视觉研究中的重要分支。 目前在工业生态中应用广泛的 YOLOv5 模型经过版本迭代,在预测权重大小以及检测精度方面都有所优化,但模型的处理速度仍然较低,尤其是对于小目标及遮挡目标的检测效果有待改进。该文提出一种基于注意力机制的 YOLO v5 改进模型。 首先,通过引入维度关联注意力机制模块进行特征融合,提升主干网络的特征提取能力,达到改善小目标与遮挡目标的检测效果;其次,采用 SIoU 损失函数代替 CIoU 损失函数,作为新的边界框回归参数的损失函数,提高边界框的定位精度以及检测速度。 实验结果显示,优化模型的平均精度均值达到 87. 8% ,相比于 YOLOv5 提高了 4. 7 百分点,在单 GPU 上模型的检测速度达到 83. 3 FPS。
Abstract:
With the development of machine vision technology,target detection has become an important branch. At present,the YOLOv5model,which is widely used in the industrial ecology,has undergone version iterations and has been optimized in terms of predictionweight and detection accuracy,but the processing speed of the model is still not high,especially for small targets and occluded objects.The detection effect needs to be improved. We propose an improved model of YOLO v5 based on attention mechanism. First of all,byintroducing the dimension related attention mechanism module for feature fusion,the feature extraction ability of the backbone network isimproved to improve the detection effect of small targets and occluded objects; secondly,the SIoU loss function is used instead of theCIoU loss function as a new bounding box regression parameter. The loss function improves the positioning accuracy and detection speedof the bounding box. The experimental results show that the average precision of the optimized model reaches 87. 8% ,which is 4. 7 percentage points higher than that of YOLO v5,and the detection speed of the model on a single GPU reaches 83. 3 FPS.

相似文献/References:

[1]王明平 宋丽梅.基于计算机视觉的车架号采集系统[J].计算机技术与发展,2008,(04):239.
　WANG Ming-ping,SONG Li-mei.Vehicle Identify Number Acquisition System Based on Machine-Vision[J].,2008,(12):239.
[2]蒋恩松 肖辉军 孙刘杰 熊清廉.基于机器视觉的套印误差自动检测系统设计[J].计算机技术与发展,2008,(07):173.
　JIANG En-song,XIAO Hui-jun,SUN Liu-jie,et al.Design of Automatic Detecting Printing Registration Deviation System Based on Machine Vision[J].,2008,(12):173.
[3]张锦娟 师军 于佳丽 卢照.SpikeNet的研究及其在快速人脸识别中的应用[J].计算机技术与发展,2010,(07):235.
　ZHANG Jin-juan,SHI Jun,YU Jia-li,et al.Research on SpikeNet and Its Application in Quick Face Recognition[J].,2010,(12):235.
[4]方鹤鹤 冯宏伟 马煜.基于形态滤波和分水岭变换的边缘检测方法[J].计算机技术与发展,2006,(01):49.
　FANG He-he,FENG Hong-wei,MA Yu.Edge Detection Method Based on Morphology Filter and Watershed[J].,2006,(12):49.
[5]李雪晨 汪仁煌 艾星芳.改进的弯曲度算法在阶梯修剪检测中的应用[J].计算机技术与发展,2012,(03):166.
　LI Xue-chen,WANG Ren-huang,AI Xing-fang.Use of Improved Tortuosity Algorithm in Step Form Disfigurement Test[J].,2012,(12):166.
[6]赵立双 冯莹 曹毓.单目视觉定位中SURF算法参数的优化[J].计算机技术与发展,2012,(06):6.
　ZHAO Li-shuang,FENG Ying,CAO Yu.Optimization of SURF Parameters in Monocular Visual Odometry[J].,2012,(12):6.
[7]徐自越 李战明 李二超.OpenCV在焊缝实时检测与处理系统中的应用[J].计算机技术与发展,2012,(08):170.
　XU Zi-yue,LI Zhan-ming,LI Er-chao.Application of OpenCV on Real-time Detection and Processing System of Seam[J].,2012,(12):170.
[8]杨杰,卢盛林,赵晓芳.机器视觉在钢化玻璃缺陷检测中的应用研究[J].计算机技术与发展,2013,(03):211.
　YANG Jie,LU Sheng-lin,ZHAO Xiao-fang.Application and Research of Machine Vision in Tempered Glass Defect Inspection[J].,2013,(12):211.
[9]陈鹏宇[],孙文奇[],赵忠龙[]. 基于机器视觉的印刷质量检测研究[J].计算机技术与发展,2014,24(07):103.
　CHEN Peng-yu[],SUN Wen-qi[],ZHAO Zhong-long[]. rinting Quality Detection Based on Machine Vision[J].,2014,24(12):103.
[10]洪磊[],嵇保健[],洪峰[]. 一种基于亚像素角点的 SIFT 立体匹配算法研究[J].计算机技术与发展,2016,26(01):48.
　HONG Lei[],JI Bao-jian[],HONG Feng[]. Research on a SIFT Stereo Matching Algorithm Based on Sub-pixel Corners[J].,2016,26(12):48.
[11]李 郸,梁洲源,马慧宇,等.机器视觉在卷烟生产中的应用与展望[J].计算机技术与发展,2024,34(01):213.[doi:10. 3969 / j. issn. 1673-629X. 2024. 01. 031]
　LI Dan,LIANG Zhou-yuan,MA Hui-yu,et al.Application and Prospect of Machine Vision in Cigarette Production[J].,2024,34(12):213.[doi:10. 3969 / j. issn. 1673-629X. 2024. 01. 031]

更新日期/Last Update: 2023-12-10