Face Mask Detection using Deep Learning CNN Architecture

Chinmay Arora, Ritvik Gupta, Dr. S. Sridhar

Abstract: The COVID-19 pandemic redefined many aspects of our definition of normal life in which social distancing and wearing of face masks plays a vital role in controlling the spread of the virus. But many people are not adhering to the social precautions that were once essential of wearing face masks in public places which increases the spread of viruses. Hence to avoid such situations we must scrutinize and make people aware of wearing face masks. Manual real-time monitoring of face mask wearing for a large group of people is becoming difficult. Humans cannot be overly involved in this process, due to the chance of getting affected by the coronavirus. This can be combated using arti?cial intelligence and deep learning, which is the main theme of our study. The proposed approach is based on two steps. One step aims to create a DL model that detects facemasks and whether they are correctly worn. An online step that deploys the DL model at edge computing in order to detect masks in real-time. We have used a pre-trained CNN model of the VGG19 class for image classification and identification. A dataset from Kaggle will be used for both training and testing purposes. Overall training accuracy of over 95% is targeted.

Keywords: Facemask detection, Real-time, COVID-19, Deep Learning, Convolutional Neural Network