Automatic medical face mask recognition for COVID-19 Mitigation: Utilizing YOLO V5 object detection

Abstract

The ongoing COVID-19 pandemic has significantly affected global public health, necessitating protective measures such as wearing face masks to reduce the spread of the disease. Recent advances in deep learning-based object detection have shown promise in accurately recognizing objects within images and videos. In this study, the state-of-The-Art You Only Look Once (YOLO) V5 object detection model was employed to classify individuals based on their mask-wearing status into three categories: none, poor, and adequate. YOLO V5 is known for its high efficiency and precision in object recognition tasks. Two datasets, the Face Mask Dataset (FMD) and the Medical Mask Dataset (MMD), were combined for simultaneous evaluation. The performance of the models was assessed based on crucial metrics such as Giga-Floating Point Operations (GFLOPS), workspace area, detection time, and mean average precision (mAP). Results indicated that the YOLO V5m model achieved the highest mAP (97.2%) for the "adequate" class, demonstrating its effectiveness in detecting proper mask usage for COVID-19 mitigation.