An Automated Approach for Screening COVID-19 from Thermal Images Using Convolutional Neural Network

Abstract

The world has seen the disastrous effect caused by COVID-19 on humankind. The rapidity of COVID-19 transmission, re-infections, post-COVID-19 symptoms, and the emergence of new COVID-19 strands have disrupted the global healthcare systems. Consequently, screening for COVID-19 cases has become of the utmost importance. As temperature and mask checks help significantly to prevent the rapid spread of COVID-19, automating this process in public places has become indispensable. In this work, we propose an end-to-end approach for mask detection followed by temperature for efficient screening. The proposed model achieved 93.5%, 96.7%, and 97.7% precision, recall, and mAP when trained on the thermal surveillance dataset and tested on a lightning dataset consisting of images with varying intensities. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.