Mask Recognition Method Based on Graph Convolutional Network

Abstract

The rapid spread of COVID-19 worldwide has exacerbated the health crisis and affected our daily lives. Medical researchers have shown that wearing masks in public places is essential to reduce the spread of COVID-19 infection. The increase in the accuracy of the identification of masks in public places can effectively prevent the further spread of the epidemic. In this paper, we propose a mask recognition network based on graph convolutional network, D-GCN. The network adopts the method of combining convolutional neural network and graph convolutional neural network. First, Dense Net 101 is used to extract features of the real-time image to be tested, and then GCN processes the training label information to form a directed graph through the word embedding vector, it forms a new classifier based on the label connection between the training pictures, and processes the extracted feature vectors. Finally the classification result is output and completing the recognition. The experiment is carried out on the mask dataset and MAFA dataset, and the recognition accuracy is significantly improved compared with previous methods.