Face mask detection using semantic region based convolutional neural networks

ABSTRACT

Due to Covid-19, most of the health care organizations and governments have ordered their citizens to wear face mask to protect themself. The proposed novel research presents a tactical methodology for rapid detection of whether a person is wearing a face mask or not. It is entirely different from the existing system, aims at training the deep learning model with a minimum number of image samples and to operate face mask instance segmentation along with object box detection. The system proposes a novel and semantic pixel-to-pixel region based deep learning network, which can detect number of face mask instances in different categories pixel wise to organize the segment bounding box and the confidence of various categories for each pixel. The system experimentally demonstrates that it can effectively and precisely detect the face mask with multi-feature combination. It is also reported that the proposed application performance outperforms the existing system. © 2022 Author(s).