Garbage Classification Based on GC-Net

Abstract

Garbage classification is a symbol of social progress and ecological civilization. However, garbage classification is still facing severe problems. Aiming at the problems of garbage “mixed loading, mixed transportation and mixed processing”, a garbage classification approach is proposed in this paper, where convolutional neural network, transfer learning and ensemble learning are used to construct an image classification framework. Moreover, the parallel hybrid attention mechanism PHAM and FReLU activation function are introduced into the framework, and the GC-Net model is created. RuiJie Dataset and a public garbage classification dataset on the Kaggle platform were used to train and test the GC-Net model. GC-Net combines the advantages of many image algorithms to improve the accuracy of garbage classification further and enhance the model’s generalization performance.