Author(s)

Hongwu Qin, Zhe Su, Xiuqin Ma, Yibo Wang, Jemal H. Abawajy

Corresponding Author

Xiuqin Ma

ABSTRACT

Garbage classification is of great significance to maximize recycling and disposal of domestic garbage. In this paper, we propose deep convolutional neural network (CNN) based domestic garbage classification model. Although deep learning has been applied in the field of image recognition and classification with great success, it requires very large and comprehensive training dataset to classify images accurately. The robustness of deep learning model is significantly affected for domestic garbage where the amount and types of data are small to meet the needs of deep learning model training. To solve this problem, we propose a method based on Deep Convolutional Generative Adversarial Network (DCGAN) to enhance the domestic garbage dataset. We validate the proposed domestic garbage classification approach experimentally and compare it with exiting approaches. The experimental results show that the proposed method can effectively expand the data set and significantly improve the classification accuracy, compared with the exiting data enhancement methods based on geometric transformation image operation. In detail, the improvement of data size of training set is up to 81.7% compared with the original training set. And then the improvement of the average accuracy of classification is 7.27% in comparison with no data enhancement method.

KEYWORDS

convolutional neural network, DCGAN, Classification