Garbage Images Classification Method Based On Deep Learning

Tremendous amount of garbages has exceeded the bearing capacity of nature compared with the natural decomposition in the past,which has not only the critical effect on human survival but also extraordinary destruction to the surroundings.How to deal with these garbages efficiently is a realistic and critical challenge.One of excellent ways of transforming trash into treasure reasonably is garbage classification.However,many folks don’t know the specific categories of each kind of garbage,which could lead to the mis-throwing of garbage,the increase of labor costs,and the aggravation of resource waste in our daily life.As an old proverb puts it," Sharp tools make good work." We must make great use of the power of sci-tech to further promote garbage classification.It is coming true that utilizing Artificial Intelligence(AI)to help garbage classification.Currently,AI is presenting the outstanding competition,and has extensive application prospects and social benefits.AI is a good choice for the problems such as long time,high labor intensity and low precision of manual in traditional garbage classification.It has a series of advantages such as relatively mature technology,high recognition degree,strong automation,and light labor intensity.Therefore,many researchers are using AI to collaborate and promote the study of garbage classification.Under the above background,a garbage classification method integrates with attention mechanism is presented in this context,which not merely is desirable in quickly discriminate different types of garbage but speeds up classification process.Firstly,this thesis surveys previous researches and shows latest status quo in these fields of image classification and garbage classification,as well as the common image classification models.Secondly,a novel network model named ResNet-ECA is established by combining the attention mechanism ECANet and the residual network ResNet101,which can improve the network performance to extract the features of garbage images.In the meantime,the ResNet-ECA model at two garbage image datasets is verified.ResNet-ECA model can promote garbage images classification accuracy and accelerate the efficiency of garbage classification by actual instance.Herein,recognition accuracy is above 90%.Finally,a prototype system of garbage image classification is developed utilizing Gradio.This thesis provides a reference for the research and development of integrated garbage classification methods and the application of Gradio framework.The research results are innovative and referential to some extent.