Face Mask Identification using Transfer Learning Techniques

Coronavirus (COVID-19) has become the major health issues in the current scenario. Symptoms start with the common cold and end with fever. These viruses affect the respiratory part of the human body. Hence, wearing of face mask is essential. Usage of masks in public places has become mandatory. Due to the significance of the face mask in public places, a transfer learning approach has been proposed for the identification of face masks. The system comprises feature extraction and classification. For feature extraction, transfer learning is used and for classification, Machine Learning is used (ML). Mobile Net and VGG 16 (Visual geometric Group) is used as feature extractor, similarly, SVM and Random Forest are used as a classifier. Performance metrics are evaluated to determine the highest accuracy from the resultant model. The high optimal model is determined by comparing its accuracy. Of the four models, VGG-16_SVM gives high accuracy of 89% and Mobile Net_ random Forest gives the second most accuracy of 87%. Out of all VGG-16,_ SVM gives the best classification output, which will be the most optimum model for face mask identification. The dataset is procured from Masked Face - Net dataset, it consists of human faces with correct and incorrect mask data based on Flicker-Faces-HQ (FFHQ).