Comparison of Neural Network Models for Facial Recognition with Wearing Masks

Facial recognition with masks aims at identification in case of the COVID-19 coronavirus epidemic, which is challenging since masks obscure a large number of feature points of faces. Comparing the performance of multiple neural network models trained for face recognition with masks can help to find a better solution for facial recognition with masks. In this paper, we analyze and compare three neural networks, including Siamese Network, Alex Network and Resnet_CBAM Network for facial recognition with masks to come up with a better facial recognition method with masks for study and discussion. Specifically, we use VGG16 and Resnet18 models as feature extraction networks of the Siamese Network. Accuracy, precision, recall and AUC are used to evaluate the performances of the related models. The experimental results show that Resnet_CBAM has better performance on facial recognition with masks.