Face Detection Method with Mask by Improved YOLOv5

Abstract

Facial recognition has become a major challenge today as more and more individuals wear masks to avoid contracting the COVID-19 virus. The rapid spread of the COVID-19 pandemic has made it necessary for people to use a face mask, especially in public places, to prevent the spread of this disease. Therefore, recognizing faces and distinguishing a person's identity has become a problem that cannot be easily recognized, as many researches have proposed finding solutions to detect faces. But faces wearing a mask were not accurately detected, so in this research it was proposed to use a deep learning algorithm, which is the improved YOLOv5, which is a YOLO model that is characterized by accuracy and speed compared to YOLO models a deep learning algorithm. The YOLOv5 algorithm is proposed here from YOLO Network to detect and recognize faces with and without wearing a mask. It is an advanced and fast system for detecting faces in real time. As we reviewed most of the experiences with previous versions of YOLO, we noticed that YOLOv5 is a better model than previous YOLO models at detecting faces while wearing a mask, but needs to improve accuracy. As face detection is of great importance in various fields in terms of security in all public places and requires accuracy in detection. It is known that there is very little data available on images of wearing masks. So the training and evaluation was performed on the dataset available on Google Colab to the improved YOLOv5 algorithm in this paper.