Abstract
Face recognition is the most reliable security measure in smartphones. Face masks make face detection more challenging. In most countries the wearing of facial masks is mandatory on public transport and in public areas to prevent the spread of Covid-19. In smartphone security, facial detection and authorization is a recent measure implemented along with fingerprints, passwords, and pins. This research aims to identify the person while the face is covered with a facial mask with only eyes and forehead being exposed. Four transfer learning object detection models namely YOLOv3, YOLOv4, YOLOv3 tiny, and YOLOv4 tiny have been implemented in identifying the masked and unmasked faces. The YOLO V3 object detection model is the best performing model with a detection time of 0.015 seconds, F1 score of 0.85 and mAP score of 0.88. Further, fast, and accurate masked face detection suitable for deployment on a smartphone hardware architecture is demonstrated on a webcam.
