Household garbage classification based on deep learning

Abstract

Garbage classification plays an essential role in protecting the earth’s ecological environment and promoting economic development. Before computer vision technology was developed, waste classification was mostly carried out by manual sorting, which has some disadvantages such as high labor intensity, low sorting efficiency, and poor working environment.

In recent years, the success of deep learning technology in computer vision has spurred significant progress in image classification. Many researchers are exploring the use of deep learning technology for garbage classification and have put forward some effective methods. Currently, a lot of automatic garbage classification methods have been proposed and can be divided into traditional machine learning methods and deep learning methods.

In this project, a comprehensive survey was conducted to review the existing garbage classification methods based on traditional machine learning approaches and on deep learning methods. The performance and characteristics of a variety methods are analyzed and compared to show the advantages and disadvantages of each other. In addition, the dissertation also introduces the existing public datasets of garbage classification used in different researches.

Moreover, a deep learning network (ResNeXt101) is applied to perform household garbage classification in this dissertation. The detailed structure of the network is introduced and the effectiveness of the algorithm is verified by testing with garbage images collected in real life.