Masked or Unmasked Face Detection from Online Video using Learning Aided Pattern Recognition Method

Security and surveillance are major concerns for every organization. During the Covid-19 pandemic situation, wearing of face mask has been recommended to be essential while accessing a sanitized place. Demand for developing surveillance mechanism for pandemic compliant infrastructure has arisen. This paper discusses a computer vision technique for detection of unmasked or masked faces from video sequence and describes a method for unmasked and masked face detection from custom data set using YOLOv5. The network is trained from scratch using a custom dataset and images are labeled into two groups. The model is trained from scratch with a public dataset and is capable of detecting human faces with and without mask correctly with high rate of accuracy. The system has detected multiple occurrences of such faces. The model has also worked properly with video input.