A hybrid tiny YOLO v4-SPP module based improved face mask detection vision system.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Surveillance Technology

Background:

• Face masks are increasingly common, complicating offender identification in surveillance.
• Existing object detection methods struggle with detecting masked individuals.

Purpose of the Study:

• To develop an accurate and efficient face mask detection system for real-time surveillance.
• To improve upon existing object detection models for masked face identification.

Main Methods:

• An improved tiny YOLOv4 object detector was developed by integrating a spatial pyramid pooling (SPP) module and an additional YOLO detection layer.
• The network was trained and validated on a custom dataset of over 50,000 images.
• Performance was evaluated using mean average precision (mAP) and average precision (AP) metrics.

Main Results:

• The proposed tiny YOLOv4-SPP network achieved a mAP of 64.31%, a 6.6% improvement over the original tiny YOLOv4.
• For detecting small objects like face masks, the system achieved an AP of 84.42%, a 14.05% increase compared to tiny YOLOv4.
• The system demonstrated capability for accurate, real-time detection of masks in surveillance.

Conclusions:

• The novel tiny YOLOv4-SPP network offers superior performance for face mask detection in surveillance.
• This advancement is crucial for law enforcement and security applications in the current climate.
• The system provides a reliable solution for identifying masked individuals in real-time scenarios.