基于改进的yolov5n模型车辆检测算法  

Vehicle Detection Algorithm Based on Improved YOLOv5nModel

作　　者：王文上 刘晓群[1] 郝娟[1] WANG Wenshang;LIU Xiaoqun;HAO Juan Hebei(University of Architecture College,Zhangjiakou Hebei 075000,China)

机构地区：[1]河北建筑工程学院,河北张家口075000

出　　处：《长江信息通信》2024年第11期32-35,共4页Changjiang Information & Communications

摘　　要：随着智能交通系统的发展,车辆检测技术在交通安全和智能监控中发挥着重要作用。文章通过改进yolov5n的模型,引入Coordatt(Coordinate Attention)注意力机制让模型更加聚焦于关键信息,提高对目标物体的定位和识别能力,同时又添加了EIOU损失函数,该损失函数有助于模型更准确地预测边界框并且加快模型的收敛速度。改进后的模型YOLO-CE在自制数据集上的检测上取得了显著的性能提升,与原始的模型相比,精度提高了1.4%,召回率提高了3.5%,mAP值提高了1.7%。With the development of intelligent transportation systems,vehicle detection technology plays an important role in traffic safety and intelligent monitoring.This article improves the YOLOv5N model by introducing the Coordinate Attention mechanism to focus more on key information,improving the localization and recognition ability of target objects.At the same time,an EIOU loss function is added,which helps the model predict bounding boxes more accurately and accelerates the convergence speed of the model.The improved model YOLO-CE achieved significant performance improvement in detection on self-made datasets,with an accuracy increase of 1.4%,a recall increase of 3.5%,and an mAP value increase of 1.7%compared to the original model.

关 键 词：olov5n 车辆检测 注意力机制 损失函数 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]