DTLMV2-A real-time deep transfer learning mask classifier for overcrowded spaces.

Area of Science:

• Computer Vision
• Machine Learning
• Public Health

Background:

• The COVID-19 pandemic necessitates effective strategies to curb viral transmission.
• Widespread mask-wearing is a key preventative measure, yet compliance varies.
• Monitoring mask usage in public spaces is crucial for public health surveillance.

Purpose of the Study:

• To develop and implement a reliable face mask identification model.
• To create a low-cost, efficient system for detecting mask usage in real-time.
• To leverage deep learning for public health monitoring during the pandemic.

Main Methods:

• Proposed DTLMV2 (Deep Transfer Learning MobileNetV2) model for classification.
• Utilized MobileNetV2, a lightweight Convolutional Neural Network (CNN).
• Integrated computer vision techniques for real-time image and video analysis.

Main Results:

• Achieved 97.01% accuracy on validation data.
• Attained 98% accuracy on training data.
• Demonstrated 97.45% accuracy on testing data.

Conclusions:

• The DTLMV2 model provides a highly accurate method for face mask detection.
• The system is suitable for deployment in open public spaces and surveillance.
• This technology can support public health initiatives by monitoring mask compliance.