基于深度神经网络的可定制图像识别研究  

Research on Customizable Image Recognition Based on Deep Neural Network

作　　者：吕秋雨 李健 李棋波 高荣[1] 卢守东[1] Lyu Qiuyu;Li Jian;Li Qibo;Gao Rong;Lu Shoudong(School of Information and Statistics,Guangxi University of Finance and Economics,Nanning Guangxi 530000,China)

机构地区：[1]广西财经学院信息与统计学院,广西南宁530000

出　　处：《信息与电脑》2020年第11期103-105,共3页Information & Computer

基　　金：国家级大学生创新创业训练计划支持项目(项目编号:201911548030)。

摘　　要：随着IT技术的迅速发展,图像识别技术已经慢慢融入并改变了人们的生活。为了使训练模型更加合理并帮助有特定图像识别需求的用户,此次研究采用卷积神经网络搭建一个可定制训练模型和服务的平台,并对2762种不同品牌香烟进行测试分类识别。通过对不同品牌的香烟图片进行定制化训练,识别率平均可达96%,验证了该方案在实际应用中的有效性。With the rapid development of IT technology,image recognition technology has gradually integrated into and changed people's lives.In order to make the training model more reasonable and help users with specific image recognition needs,this study uses convolutional neural network to build a platform for customized training model and services,and tests and classifies 2762 different brands of cigarettes.Through customized training for different brands of cigarette pictures,the recognition rate can reach 96%on average,which verifies the effectiveness of the scheme in practical application.

关 键 词：图像识别 可定制化 深度神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]