A Low-Cost Embedded Security System for UAV-Based Face Mask Detector Using IoT and Deep Learning to Reduce COVID-19

Abstract

Nowadays, the most effective method against the virus is wearing a mask. Hence, it is fundamental to wear a mask appropriately at open places like general stores and shopping malls. This paper proposes a novel human face mask detection method from UAV-captured frame sequences to solve the aforementioned problem. The proposed approach involves an offline stage and an inference stage. The offline stage generates the mask or no-mask by utilizing a convolutional neural network. We trained our model on a face mask dataset, and this enhancement allows the suggested system to obtain high accuracy in detecting unmasked people. The inference stage uses the already generated model to detect no mask humans and sends the alert to the smartphone-based Internet of Things. At this stage, Jetson nano was used to implement an embedded powerful real-time application for UAV-based face mask detection that runs at high frames per second. The proposed system monitors and detects people who have not worn a mask. Also, we used IoT techniques to send the pictures and notifications to the nearest police station to apply forfeit when it detects unmasked people. The main contributions in this paper lie in adjusting the deep learning, embedded platforms, IoT techniques, and Tello drone, generally dedicated to detecting unmasked people at a low cost. On average, detection accuracy is 99% based on the experimental evaluation of the proposed deep learning model for UAV-based face mask detection on the provided dataset. Overall, the proposed method can help decrease the spread of COVID-19 and other transmissible diseases.