Mask Detection and Classification in Thermal Face Images

Face masks are recommended to reduce the transmission of many viruses, especially SARS-CoV-2. Therefore, automatic detection whether there is a mask on face, what type worn, and how it worn an important research topic. In this work, use thermal imaging was considered analyze possibility detecting (localizing) as well check possible classify face. The previously proposed dataset images extended annotated with description location within Different deep learning models were adapted. best model for face turned out be Yolov5 in "nano" version, reaching mAP higher than 97% precision about 95%. High accuracy also obtained classification. results convolutional neural network built autoencoder initially trained image reconstruction problem. pretrained encoder used train classifier which achieved 91%.