COVID-Masks is and experiment based on cooperation and knowledge transfer between Poznan University of Technology and SME ABZero to develop the affordable computer system for automatically detecting people wearing correctly and incorrectly anti-virus masks. The developed system, comprising a camera module and microcomputer with face detection software and a trained convolutional neural network, offers advantages in mobility, compact size, reliability, and accuracy. The software accepts video streams from various sources and supports multiple platforms. The performance of the elaborated system was validated independently by both partners in various environments such as crowded corridors, entrances to the buildings and lecture rooms. The experimental results show a large potential to implementation of the proposed solution.

Objectives

COVID-Masks project presents an automatic system for detection of correct wearing anti-virus masks. The knowledge transfer between the experiment partners aimed at creating a low-cost and accurate implementation of the system that can be used to monitor crowded places to find people wearing anti-virus masks correctly as well as incorrectly.

Challenges

Reliable system for anti-virus mask wearing correctness detection Need of build dataset from scratch containing different environments, people, lighting conditions etc. Software possibility to enlarge dataset and train neural networks to achieve better results of proper detection Support of low-cost system hardware

Technology

The prepared system contains software for recognition of people wearing face masks on images and video sequences/streems as well as in live video streams. The system is based on neural network models: MTCNN for face detection (ROI) and the authors’ model for the mask detection. The package also includes a big database and a program, which can be used for training other models.

Market

Due to the recommendations of COVID-19 it was and still is important to check whether people wear face masks correctly or not. We offer a novel technology for prevention and surveillance targeted for both the public at large and also specific communities such: as refugee camps, shopping centres, small companies etc. The use of algorithms to recognize people wearing face masks can find further, constant applications even after the pandemics. In particular, wearing face masks is always mandatory in some environments, such as medical and industrial centres.

Consortium