Real-time face mask position recognition system based on MobileNet model.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Public Health

Background:

• COVID-19 remains a significant global health threat, necessitating effective preventative measures.
• Improper face mask usage, including covering the nose and mouth, compromises protection.
• Automatic recognition systems can enforce proper mask-wearing in public spaces.

Purpose of the Study:

• To develop and evaluate a deep learning model for automatic face mask position recognition.
• To create and release a diverse dataset for training and testing face mask detection models.
• To compare the proposed model's performance against existing methods.

Main Methods:

• A new dataset of 391 individuals' face mask images was collected and released.
• Six pre-trained deep learning architectures were studied.
• The state-of-the-art MobileNet model was fine-tuned for face mask recognition.

Main Results:

• The fine-tuned MobileNet model achieved 99.23% accuracy, 99.22% F1-score, and 99.19% Cohen's Kappa.
• The proposed model outperformed existing methods by approximately 2% in accuracy.
• The system demonstrated robust performance on both real and synthetic datasets without accuracy degradation.

Conclusions:

• An automatic face mask position recognition system was successfully developed.
• The system accurately identifies correct and incorrect face mask usage.
• This technology can significantly contribute to public health by ensuring proper mask adherence.