Convolutional Neural Networks for Multiclass Classification of Masks

Recently, the COVID-19 pandemic has caused many human infections and spread the virus around the world. Research has shown that wearing masks goes a long way toward preventing the spread of infections, so correct use is very important. Today, wearing masks in public places is a common practice, but the virus will continue to spread if not used correctly. This work aims to correctly detect and classify the use of masks, for which the use of cumulative neural networks has been proposed. Deep learning is most effective at detecting whether a person is wearing a mask correctly. The model is trained with the MaskedFace-Net dataset and evaluated with different images of it. We use the Caffe model for face detection and preprocess the images to extract features. These images are entered into a convolutional neural network, where they are classified into the use of masks, non-use of masks or incorrect use of masks. The proposed model achieved an accuracy of 99.6865% on the test data improving the percentage compared to other authors.