Real-Time Face Mask Detection to Prevent COVID-19 in Confined Spaces

Aviva Munshi, Anoushka Mehra, Ashna Choudhury, Yogeswari Sahu, Asmit Gupta

Neural network, Image Data generator, Face mask detection, MobileNetV2.

In this pandemic, it is getting more and more challenging to keep track of people wearing masks regularly. We cannot solely depend on human efforts to take care of this task. Therefore we need to develop software that can automatically detect whether any given person is wearing a mask or not. In this time of Covid-19 pandemic, it is of utmost importance that we all use masks regularly. But there are always exceptions, and only manual human efforts are meaningless to track who are not wearing masks. Therefore there is a need to develop software that can detect if a person is wearing a mask or not automatically. Currently, face recognition field has seen many challenges and has evolved much. New and more accurate algorithms are being derived by using convolutional architectures. Convolutional architecture also has the benefit of its ability to extract pixel details. Fully convolutional architecture is used for training so that faces are segmented out of the image semantically. We can find whether a person is using a mask not or adequately by feature detection and feature extraction. For accurate model creation, a dataset of morphed images with veneers can be used by the face mask detector. Thus, the created model will be computationally efficient. This will also be easily deployable because it will use MobileNetV2 architecture (Google coral, Raspberry Pi, etc.). Due to Covid-19 pandemic, face mask detection is required, and this framework can be used for that in real-time applications. This project can be used at public places like airports, schools, train stations, workplaces and bus stops by merging with embedded applications. This will lead to a safer environment in pandemic times. This identification process will help us classify individuals by reducing the number of manpower required.

https://doi.org/10.31871/IJNTR.7.11.18