基于改进YOLOv5的X光卷烟检测
X-ray Cigarette Detection Based on Improved YOLOv5
任宝峰 1祁卫国 1肖占云 1撒兴涛 1贾然1
作者信息
- 1. 承德市烟草专卖局(公司),河北 承德 067000
- 折叠
摘要
为解决基于X光的卷烟检测存在的预测精度低和主观性强的问题,介绍一种改进的YOLOv5算法,并将其应用到基于X光的卷烟目标检测中;该方法将YOLOv5的骨干网络进行改动,增添卷积块注意力模块(Convolutional Block Attention Module,CBAM);CBAM是一个用于前馈卷积神经网络的简单而有效的注意力模块,能够沿着通道和空间两个单独的维度依次推断注意映射;然后,通过将注意力映射乘以输入特征映射,进行自适应特征细化;最后,将该方法应用某地区卷烟X光的测试数据;试验结果表明:该模型拥有更好的预测精度,为卷烟X光检测提供了一种新思路.
Abstract
In order to solve the problems of low prediction accuracy and strong subjectivity in X-ray cigarette detection,this paper introduced an X-ray cigarette detection method based on improved YOLOv5.In this method,the Convolutional Block Attention Module(CBAM)is added into the backbone network of YOLOv5.CBAM is a simple and effective attention module for feedforward con-volutional neural networks that is able to sequentially infer attention maps along two separate dimen-sions,namely channel and space.Then adaptive feature refinement is performed by multiplying the attention map by the input feature map.Finally,the method is applied to the test data of cigarette X-ray in certain area.The experimental results show that the model has better prediction accuracy and provides a new idea for cigarette X-ray detection.
关键词
X光图片/卷烟检测/YOLOv5/卷积块注意力模块Key words
X-ray images/cigarette detection/YOLOv5/Convolutional Block Attention Module引用本文复制引用
出版年
2024
NETL
NSTL
万方数据