FMDNet: An Efficient System for Face Mask Detection Based on Lightweight Model during COVID-19 Pandemic in Public...

Area of Science:

• Artificial Intelligence
• Computer Vision
• Deep Learning

Background:

• Public health protocols, such as face mask mandates, require effective monitoring systems.
• Traditional surveillance methods struggle with real-time, accurate identification of non-compliance.
• The need for automated solutions to enforce public health guidelines is critical.

Purpose of the Study:

• To develop and evaluate a deep learning (DL) model for automated face mask detection in public spaces.
• To assess the model's accuracy, efficiency, and performance against existing DL models.
• To provide a tool for governments to monitor and enforce face mask protocols.

Main Methods:

• Creation of a private dataset comprising diverse face images with and without masks.
• Development of a deep learning model named FMDNet for face mask detection.
• Training and real-time testing of FMDNet on surveillance video feeds.

Main Results:

• FMDNet achieved 99.0% accuracy in detecting face mask violations (individuals not wearing masks).
• The model demonstrated superior performance compared to FSA-Net (24.03% better), MobileNet V2 (5.0% better), and ResNet (24.10% better).
• FMDNet is lightweight, operates with 99.0% confidence in constrained environments, and processes at 41.72 frames per second.

Conclusions:

• The developed FMDNet offers a highly accurate and efficient AI-based solution for real-time face mask detection.
• Its lightweight nature and high performance make it suitable for deployment in resource-constrained environments.
• FMDNet can significantly assist governments in enforcing public health Standard Operating Procedures (SOPs).