YOLOv8-AS: Masked Face Detection and Tracking Based on YOLOv8 with Attention Mechanism Model

The development of intelligent surveillance systems relies significantly on the effectiveness of face detection. To identify suspicious individuals, this technology uses a face detection model that quickly and efficiently analyzes every frame generated by the system. This presents challenges for some public safety surveillance systems that rely on face detection and tracking. Hence, there is a growing need to develop effective algorithms for face detection and tracking, even when wearing masks. In this paper, we present a YOLOv8-AS model with an attention mechanism for real-time tracking designed for individuals with and without masks. This model was trained using publicly available face (ChokePoint and NRC-IIT) datasets that mostly feature individuals without masks. Although not directly trained on masked faces, our model demonstrated its strength by performing adeptly at accurately tracking individuals wearing masks.