基于深度学习的垃圾分类方法研究

北京理工大学珠海学院2020届本科生毕业论文Research on garbage classification based on deep learningAbstractThis paper studies garbage classification based on deep learning methods.Since convolu-tional neural networks are particularly prominent in image recognition applications,it usesconvolution operations,pooling operations,local connections,and weight sharing to formthe classified model which is a brand new convolutional neural network.Training,recogni-tion,classification and prediction of five kinds of garbage images were analyzed.With thesame experimental conditions,compared with the traditional artificial neural network,theresults show that CNN improves the recognition accuracy by about 15%.It finds that theimage resolution is proportional to the time.In this model,the recognition of each categoryis studied separately,which shows that the predicted accuracy of paper and plastic is morethan 90%,which is easier to predict success than other categories.Keywords:Convolutional neural network Artificial neural networks Pooling layer Lo-cal connection