ABSTRACT
The COVID19 pandemic has had a significant impact on people's social lives. Due to this pandemic, almost every office, institution, organization in the world suffered a great deal from being practically closed. The World Health Organization (W.H.O) recommended everyone wear a mask whenever they step outside or in a public place. Therefore, it is mandatory to cover your face with a mask at public places, social gatherings, etc. Facemask detection has recently become one of the most important tasks to help society. The advancement of technology has proven that deep learning has shown its effectiveness in recognition and classification through image processing. There are many face detection models created by using several algorithms and techniques. Find whether a person has puton a mask properly or not and identify that person who didn’t puton a mask properly employing their age and gender. The combination of the face mask detection module and age & gender detection module is used. In our paper, the Haar cascade classifier was implemented to detect faces from the input images in the face mask recognition module. We train this module using CNN. We can recognize faces in this model using the Voila Jones technique and Haar-like features. The face detection module and age & gender detection module is trained by using a Convolutional neural network. A model trained by Tal Hassner and Gil Levi is used to implement Age and Gender detection;an alert sound will be a part of the outcome if the person is not wearing a mask properly. For the facemask detection module, the dataset is taken from Kaggle;images of people wearing masks and not wearing masks are gathered from different sources and formed into a dataset to train the model. In this paper, we used the Adience dataset to train age & gender detection and a dataset from Kaggle containing pictures of people’s faces with and without a mask. The model attains an accuracy of 93.42 %for face mask detection and an accuracy of 91.23% for Age and Gender detection. © 2006-2022 Asian Research Publishing Network (ARPN). All rights reserved.