Yolov4FaceMask: COVID-19 Mask Detector

The rampant COVID-19 disease has brought global crisis with its deadly spread to most countries in the world, and about over 58 million confirmed cases along with 1.4 million deaths on April 2021. The absence of any active therapeutic agents and the lack of immunity against COVID-19 increase the population's vulnerability. Since there are efficient vaccines but it is not available to all worldwide, Especially the inhabitants of the third world, in addition the appearance of mutated copies (Indian version, Nigerian version ⋯), properly wearing masks in public areas is one of the major protection methods for people. Although it is important to wear the facemask correctly, there are not much research studies about facemask detection based on image processing. In this work, we propose Yolov4FaceMask, which is a model with high accuracy and can detect Masked/IncorrectMask/UnMasked faces. The main challenging point of this task is the absence of practical datasets. For this, we introduce our new practical dataset, which contains 14409 images. Experiments show that the Yolov4FaceMask achieves 88.82% mAP on our proposed dataset. This paper presents a Deep Learning (DL) based framework for automating face-masked/incorrect mask/unmasked people's detection using images and surveillance video and the library OpenCV for image processing. Our proposed model uses the object detection model named YOLOv4 to segregate human's faces from the background. The Yolo4FaceMask results are further compared with other popular state-of-the-art models of face mask detection, e.g. Face-mask InceptionV3 and SSDMNV2 and RetinaFaceMask in terms of mean Average Precision (mAP).