摘要
“行程码”的有效使用需要解决图像模糊、网络权重大等问题。本文提出一种基于损失函数改进的轻量化模型LL_YOLO(Lightweight Loss YOLO)。LL_YOLO基于YOLOv5模型,通过接口在线调用图像增强函数进行画质增强、改进损失函数,提高检测精度,轻量化压缩模型。实验结果表明,LL_YOLO在图像增强与损失函数模块等的作用下,识别精度提高到91.82%,参数量降低为2.8M。因此LL_YOLO具有低参数量和计算量的优势,对高算力硬件的依赖性低,能够极大地降低应用部署成本。
The effective use of travel code needs to solve the problems of blurred image and heavy network weight.In this paper,we propose a lightweight model LL_YOLO(Lightweight Loss YOLO)based on improved loss function.It is based on the YOLOv5 model,and improves the detection accuracy and lightweight compression model by calling the image enhancement function online through the interface.The experimental results show that the recognition accuracy of LL_YOLO is improved to 91.82%and the number of parameters is reduced to 2.8M under the effect of image enhancement and loss function module.LL_YOLO has the advantages of low number of parameters and computation,low dependence on high-computing hardware,and can greatly reduce the application deployment cost.
作者
王媛媛
田海阳
朱俊勋
严少峰
黄佳泷
宋照渝
Wang Yuanyuan;Tian Haiyang;Zhu Junxun;Yan Shaofeng;Huang Jialong;Song Zhaoyu(School of Computer and Software Engineering,Huaiyin Institute of Technology,Huaian,Jiangsu 223001,China;Jiangsu Engineering Laboratory of iot Mobile Internet Technology)
出处
《计算机时代》
2023年第4期116-119,共4页
Computer Era
基金
淮阴工学院—淮安经济技术开发区产学研合作项目(Z413H21522)
江苏省自然科学基金面上项目(BK20211365)
大学生创新创业训练计划项目(202211049087Y、202211049268XJ)。
关键词
行程码
损失函数
轻量化
图像增强
travel code
loss function
lightweight
image enhancement
