基于改进YOLOv5s的目标检测算法  

Target Detection Algorithm Based on Improved YOLOv5s

作　　者：宋为 王赫莹[1] 郭忠峰[1] 胡克腾 SONG Wei;WANG Heying;GUO Zhongfeng;HU Keteng(School of Mechanical Engineering,Shenyang University of Technology,Shenyang 110870,China)

机构地区：[1]沈阳工业大学机械工程学院,沈阳110870

出　　处：《机械工程师》2024年第9期64-67,共4页Mechanical Engineer

摘　　要：针对YOLOv5s算法定位能力不足等问题,在YOLOv5s的基础上提出一种改进算法。该算法首先使用CIOU_Loss替换YOLOv5的GIOU_Loss作为算法的边界回归损失函数,考虑到预测框与真实边界框的长宽比与中心距离,可以在提升定位精度的同时提高算法的收敛速度;最后使用轻量化卷积模块GhostConv替换YOLOv5s特征融合层的卷积,在轻量化的同时,可以提升检测精度。使用VOC2007数据集来验证算法的有效性,经试验表明,改进后的算法精度达到了95.8%,比原算法提高了2.3%。Aiming at the problem of insufficient positioning ability of YOLOv5s algorithm,this paper proposes an improved algorithm based on YOLOv5s.Firstly,CIOU_Loss is used to replace GIOU_Loss of YOLOv5 as the boundary regression loss function of the algorithm.Considering the length-width ratio and center distance between the prediction frame and the real boundary frame,the positioning accuracy and convergence speed of the algorithm can be improved.Finally,lightweight convolution module GhostConv is used to replace the convolution of YOLOv5s feature fusion layer,which can improve the detection accuracy while being lightweight.VOC2007 data set is used to verify the effectiveness of the algorithm.The experiment shows that the accuracy of the improved algorithm reaches 95.8%,which is 2.3%higher than the original algorithm.

关 键 词：目标检测 YOLOv5s 损失函数 GhostConv 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]