Real-time mask and social distance detection by unmanned aerial vehicle

The COVID-19 corona virus first appeared in Wuhan, China, in late December 2019. Usually, this virus is transmitted through small particles when speaking, coughing, sneezing, and in confined and unventilated spaces, mostly between people in close contact. The World Health Organization has identified the most effective ways to prevent the spread of the virus as maintaining physical distance and wearing a mask. State institutions and authorities have made it mandatory to keep the social distance of approximately 2 meters and to wear masks in public and closed areas such as schools, shopping centers and transportation facilities during the pandemic process. However, automatic control of whether people are wearing masks and the distance between them has become an important problem. In this study, MobilNetV2 for mask detection and Yolov3 for social distance detection, which are important in breaking the transmission chain of epidemics such as COVID-19, which are effective today, and slowing the speed of the epidemic, were carried out using deep learning algorithm and image processing methods. At the same time, an Unmanned Aerial Vehicle (UAV) was used to obtain images and to save time and provide easier control. As a result of the application, an accuracy of 99.87% to 100% was obtained in the proposed architectures. Keywords: Mask detection, Social distancing, MobileNetV2, Yolov3, COVID-19, UA