Research On Household Garbage Classification Method Based On Deep Convolutional Neural Network

With the improvement of people’s living standards,the amount of household garbage is rapidly increasing.A large amount of garbage poses a serious threat to human health,environmental damage and economic development.It is urgent to solve the problem of garbage,and it has also become an urgent topic for all countries to study.At present,in the aspect of garbage disposal,manual sorting mainly depends on manual sorting,which has disadvantages such as low sorting efficiency and poor effect.However,using deep learning technology to carry out intelligent and automatic garbage processing will become a trend,which can improve the efficiency and economic value of garbage disposal.The key technology to solve the problem of intelligent garbage disposal is intelligent garbage classification.With the rapid development of deep learning technology,it is possible to realize intelligent garbage classification by using deep learning technology.In this paper,based on the analysis of image recognition,garbage classification model and target detection algorithm,aiming at the accuracy of garbage classification,deep convolutional neural network is adopted to study the household garbage classification method.After experimental verification,the experimental results show that the classification accuracy and speed have been improved.The main research contents of this article include:(1)Construct a garbage classification data set,the source of the data set: the first is the Huawei data set,which is a public data set provided in the garbage classification competition organized by Huawei;the second is the crawler data set,which uses crawler technology to crawl life garbage pictures,the pictures that do not meet the requirements are eliminated;the third is the target detection data set,and some common garbage pictures are selected for manual marking.(2)For the problem of garbage classification,two image classification models of Res Net and Efficient Net are studied,the model parameters are fine-tuned and the fully connected layer is reconstructed,and the transfer learning method is used to conduct experimental analysis on the two models.Through experimental verification,it is found that there are mainly images Insufficient feature extraction and the inability to classify multiple objects have a certain impact on the accuracy and speed of garbage image classification.(3)according to the model above problem,this paper proposes a garbage classification model based on attention mechanism,to improve the model by introducing mechanism of attention,and Res Net50 + CBAM selected as the target detection model of backbone network,garbage classification model is verified through experiment on the classification accuracy and speed increased,reached the design target.The results of the above algorithms are compared through experiments.The experimental results show that the garbage classification model based on attention mechanism is more in line with the requirements of automatic development in classification results,and can meet the requirements of real-time while ensuring accuracy.