IoT-Deep Learning Based Face Mask Detection System for Entrance and Exit Door

Abstract

During the pandemic, it has been seen that the global population follows the guidelines issued by the health organization regarding wearing face masks, but some people do not take care of this and do not use masks. The objective of the proposed system, Wollega University Face Mask Detection System (WUFMDS), is to restrict people who are not wearing a mask on the door side by identifying the face mask from the face or open the door if the incoming person is wearing the mask. This system is based on the Internet of Things (IoT) and a Deep Learning algorithm called Convolutional Neural Network (CNN). For this purpose, images with and without masks were collected as samples from the university. The CNN algorithm is used to detect the mask and classify it as with or without masks. The IoT module controls the door operation based on the classification response sent to the IoT module by the CNN algorithm. The system was tested lively with the dummy door system in order to ensure the functionality of the face mask detection system and developed software applications for the system model are working as defined objectives. Our model had 99.36% accuracy with the training dataset and 99.29% accuracy with the validation set. Hence, the proposed system could be used for the automatic identification and classification of masks on the face and to operate the door to allow the person who is wearing the mask to pass through while keeping it closed when no mask is found on the face.