Advanced Mask-Wearing Detection for Safety Compliance in Public and High-Risk Environments

Ensuring compliance with face mask regulations is critical in high-risk environments such as hospitals, laboratories, and industries handling hazardous materials. The COVID-19 pandemic and recent outbreaks like monkeypox (2022), H5N1 (2021-present), and Nipah virus (2023) have highlighted the importance of masks in preventing airborne and close-contact infections. However, manual enforcement is impractical, and existing automated systems struggle with variations in mask types and detecting improper usage. To address these challenges, we propose a robust mask-wearing detection framework using a fine-tuned YOLOv11 model combined with a novel module for refining low-confidence predictions. We also introduce the Balanced and Diversified Face Mask Dataset (BDFMD), a 92,000- image dataset covering diverse lighting conditions, facial fea-tures, angles, and mask types. Our approach achieves AP@50 of 89.23% on BDFMD and 95.33% in real-world scenarios, demonstrating high accuracy and adaptability.