To Örebro University

With the number of positive cases of Covid-19 infection is increasing, it is essential for everyone to wear a face mask and prevent the spread of Covid. As people are gathering in a large number at different locations, it is quite important for everyone to wear a face mask and prevent the covid spread. With the increase in the crowd gathering, it is often hard to see who is not wearing a mask. Although various techniques have been proposed earlier for face mask detection, the results have not been effective. This paper proposes region-based deep learning detection techniques for face mask detection using Faster R-CNN. The proposed model uses ResNet-50 as RPN which generates anchors and output region proposals. Later, ROI pooling is used to map the feature map in proposal to target dimensions. Finally, a classifier is used to output the final class and bounding box around the face. The proposed work attained a final mean average precision (mAP) of 45% over 30 epochs and achieved satisfactory performance.