基于深度迁移学习的垃圾分类系统设计与实现  

Design and Implementation of Garbage Classification System Based on Deep Transfer Learning

作　　者：武凌[1] 王浩[1] 张晓春[1] 周健[1] 段爱华[1] WU Ling;WANG Hao;ZHANG Xiaochun;ZHOU Jian;DUAN Aihua(School of Management Science and Engineering, Anhui University of Finance & Economics, Bengbu 233030, China)

机构地区：[1]安徽财经大学管理科学与工程学院,安徽蚌埠233030

出　　处：《沈阳大学学报（自然科学版）》2020年第6期496-502,共7页Journal of Shenyang University：Natural Science

基　　金：安徽省高校自然科学研究重点项目(KJ2020A0012).

摘　　要：设计了一种基于深度迁移学习模型,对多种常见的可回收垃圾图像进行识别分类的垃圾图像分类系统.对比了VGG16、InceptionV3、InceptionResnetV2预训练模型的性能,最优识别正确率达到了90%以上,并进一步设计了基于Flask的Web应用调用模型.A garbage image classification system based on the deep transfer learning model was designed to classify recyclable garbage images.VGG16,InceptionV3 and InceptionResnetV2 pre-training models were used to train 6 kinds of garbage.The verification and performance comparison were carried out,and the classification accuracy was 90%.The Flask-based web application was designed to use the model.

关 键 词：垃圾分类 图像分类 深度学习 预训练 模型 迁移学习 InceptionResnetV2 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]