融合深度学习的烟草品牌识别模型  

A Tobacco Brand Recognition Model Based on Deep Learning

作　　者：董佳 陈杜勇 刘凯平 杨斌 钱漫 祁晓君 叶子健 钟培泉 李岱峰 DONG Jia;CHEN Duyong;LIU Kaiping;YANG Bin;QIAN Man;QI Xiaojun;Li Daifeng(Guangdong Tobacco Dongguan Co.,LTD.,Dongguan,China 523000;School of Information Management,Sun Yat-sen University,Guangzhou,China,510006)

机构地区：[1]广东烟草东莞市有限公司,广东东莞523000 [2]中山大学信息管理学院,广东广州510006

出　　处：《福建电脑》2024年第3期34-38,共5页Journal of Fujian Computer

基　　金：2023年广东省烟草专卖局(公司)科技项目专项基于人工智能的专卖智慧监管系统研究(粤烟科项202206)资助。

摘　　要：为提升烟草品牌识别的效率和准确性,本文设计了基于目标检测和多分类模块的烟草品牌识别方案。首先采用Faster-RCNN目标检测模型优化主干网络,然后使用ResNeXt多分类模型对烟盒进行具体品牌的分类识别。实验的结果表明,本方案在烟草品牌识别方面具有较高的准确率,可以为烟草品牌管理者提供良好的品牌识别服务。To improve the efficiency and accuracy of tobacco brand recognition,this paper designs a tobacco brand recognition scheme based on object detection and multi classification modules.Firstly,the Faster RCNN object detection model is used to optimize the backbone network,and then the ResNeXt multi classification model is used to classify and recognize specific brands of cigarette boxes.The experimental results indicate that this scheme has high accuracy in tobacco brand recognition and can provide good brand recognition services for tobacco brand managers.

关 键 词：烟草 品牌识别 分类模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]