An Efficient and Effective Deep Learning-Based Model for Real-Time Face Mask Detection.

Area of Science:

• Computer Science
• Artificial Intelligence
• Public Health

Background:

• The COVID-19 pandemic, originating in December 2019, has caused significant global health and economic crises.
• Transmission of the COVID-19 virus primarily occurs via respiratory droplets from infected individuals.
• The World Health Organization recommends face mask usage as a key strategy to mitigate COVID-19 spread.

Purpose of the Study:

• To investigate deep learning architectures for effective face mask detection.
• To propose an efficient and deployable model for real-time face mask detection.
• To enhance model robustness through data augmentation techniques.

Main Methods:

• Evaluation of various deep learning models including VGG16, VGG19, InceptionV3, ResNet-101, ResNet-50, EfficientNet, MobileNetV1, and MobileNetV2.
• Development of a novel model utilizing MobileNetV2 architecture combined with an autoencoder for feature abstraction.
• Implementation of extensive data augmentation techniques such as rotation, flipping, and various image transformations.

Main Results:

• The proposed MobileNetV2-based model demonstrated superior performance compared to other state-of-the-art models.
• The model effectively extracts salient features and forms abstract representations for accurate classification.
• Extensive data augmentation significantly improved the model's training effectiveness.

Conclusions:

• The developed model offers an efficient and effective solution for face mask detection, crucial for controlling COVID-19.
• The proposed architecture is suitable for deployment on edge devices, enabling widespread application.
• This research contributes to public health efforts by providing an advanced tool for monitoring mask compliance.