AI Based Monitoring of Different Risk Levels in COVID-19 Context.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Public Health Technology

Background:

• The COVID-19 pandemic caused significant global disruption, necessitating continuous monitoring of public behavior to control virus transmission.
• Effective public health strategies require tools to monitor compliance with safety measures and assess health indicators in real-time.

Purpose of the Study:

• To apply deep learning algorithms for detecting face masks in public spaces using RGB cameras.
• To develop a method for accurate body temperature measurement by detecting the eye caruncle using thermal cameras.

Main Methods:

• Utilized synthetic data generation to create hybrid datasets from public sources.
• Trained state-of-the-art algorithms including YOLOv5 object detector and a Resnet-50 based keypoint detector.
• Employed both RGB and thermal imaging for distinct detection tasks.

Main Results:

• YOLOv5 achieved 82.4% average precision for RGB mask detection.
• YOLOv5 and the keypoint detector achieved 96.65% and 78.7% average precision respectively for thermal mask, glasses, and caruncle detection.
• Publicly released RGB and thermal datasets to facilitate further research.

Conclusions:

• Deep learning models demonstrate high efficacy in detecting face masks and enabling non-contact temperature measurement.
• The developed methods and available datasets can support ongoing public health monitoring efforts.
• Integration of computer vision in public spaces offers valuable tools for pandemic response and management.