Evaluation of AIoT performance in Cloud and Edge computational models for mask detection

Abstract

COVID-19 has caused serious health damage, infect-ing millions of people and unfortunately causing the several deaths around the world. The vaccination programs of each government have reduced those rates. Nevertheless, new coronavirus mutations have emerged in different countries, which are highly conta-gious, causing concern with vaccination effectiveness. So far, wearing facemasks in public continues being the most effective protocol to avoid and prevent COVID-19 spread. In this context, there is a demand of automatic facemask detection services to remind people the importance of wearing them appropriately. In this work, a performance evaluation of an AIoT system to detect correct, inappropriate, and non-face-mask wearing, based on two computational models: Cloud and Edge, was conducted. Having as objective to determine which model better suits a real environment (indoor and outdoor), based on: reliability of the detector algorithm, use of computational re-sources, and response time. Experimental results show that Edge-implementation got better performance in comparison to Cloud-implementation.