Attention-Guided Neural Network for Face Mask Detection

With the outbreak of COVID-19 and various influenza diseases, it is necessary to wear masks properly in crowded public places to prevent the spread of the virus. Therefore, detecting mask-wearing efficiently and accurately is essential for people’s physical health and safety. In this paper, we present a novel one-stage mask detection method, named attention-guided neural network (AGNN) that can efficiently detect non-mask-wearing faces in public. Specifically, we started with YOLOv5 as a baseline and integrated the coordinate attention mechanism module into YOLOv5 to guide the holistic model for improving the ability of feature extraction. Furthermore, we explored utilizing the focal loss to solve the problem of class imbalance. The experiment is conducted on the face mask detection dataset of real-life scenes with twenty different categories. Experimental results demonstrate that the proposed AGNN method achieves higher precision and recall than the original YOLOv5 in multi-classification mask detection.