Detecting Facial Images In Public With And Without Masks Using VGG And FR-TSVM Models

Since 2019, Covid-19 has become a common problem affecting all mankind. The disease has successfully spread all over the world. Wearing a mask can practically protect the infection. Thus, detecting people wearing and not wearing masks in public is essential. However, there is still some room to improve detection accuracy of the present methods. In this paper, the transfer learning model and FR-TSVM model are used to study the latest data of pneumonia epidemic situation in Covid-19. First, a data set of 12,000 facial images wearing masks and not wearing masks in public was collected for training, testing, and validation. The pictures will be put into the improved VGG model. Then the structure of VGG model was used to extract the features of images. These features were trained by FR-TSVM with fuzzy concept included. This approach can achieve 95.5% accuracy, and it is also higher than the detection results of other methods.