Research On Domestic Waste Image Classification Algorithm Based On Deep Learning

How to adopt reasonable and effective harmless treatment of domestic garbage is an urgent problem to be solved,which is of great significance to renewable resources,improving ecology,and promoting production and economic feasibility.At present,there are still many limitations in garbage classification solutions,such as slow classification speed,low image recognition accuracy,low degree of intelligence and automation of algorithms,which can not be built on intelligent terminal equipment and so on.Aiming at the problems of low accuracy in existing garbage classification models,such as difficulty in identifying small objects with multiple targets.This paper studies the domestic garbage identification and classification model based on deep learning,which provides new ideas for the intelligence and automation of garbage classification.Specific research contents include:(1)A garbage image classification model based on attention mechanism is proposed.Firstly,aiming at the imperfect garbage classification model based on deep learning,this paper uses the parameters and calculation amount of deep separable convolution model based on Res Net50 model;Secondly,CBAM attention mechanism is added to enhance the extraction of local and global features to obtain more complete and effective feature information;Then,Focal Loss is used instead of cross entropy loss function to deal with the problem of sample imbalance in data set to improve the classification accuracy of the model.Experimental results show that the classification accuracy of the improved model on garbage image data sets reaches 92.27%.Compared with other classical image classification models,this model can better recognize different kinds of garbage and has the classification accuracy,which is suitable for the research of garbage classification field.(2)An improved multi-target garbage image detection model based on YOLOv3 is proposed.Firstly,a multi-target garbage detection model based on multi-scale feature fusion is proposed to solve the problem that it is difficult to identify multiple small-size garbage in one image.This paper introduces Mobile Netv3 network on the basis of YOLOv3 model to replace the complexity of Darknet 53 model of YOLOv3 backbone network and ensure the accuracy of the model;Secondly,four different scales are used to enhance the detection ability of small target objects,so that the positioning of regression box is more accurate;Then,CIOU loss function is used to replace the original loss function to further improve the accuracy of the model Experimental results show that the improved model can effectively detect different types of garbage.The location of classification targets is more accurate,and its m AP value is 65.21%,which improves the detection accuracy and speed.Compared with YOLOv3,the m AP value of this model is increased by 2.5% when detecting all kinds of garbage,which effectively improves the performance of the model and the application requirements of edge computing devices.