A Correct Face Mask Usage Detection By Aiot

Abstract

The COVID-19 pandemic, which affected over 400 million people worldwide and caused nearly 6 million deaths, has caused a dramatic turbulence to the world’s health system. Along with vaccination, self-testing, and physical distancing, wearing a well-ftted mask can help protect people by reducing the chance of spreading the virus. Unfortunately, researchers indicate that most people do not wear masks correctly, with their nose, mouth, or chin uncovered. This issue makes masks, once believed as one of the most cheap and effective protections against the virus, completely counterproductive. Recent studies have attempted to use deep learning technology to recognize wrong mask usage behavior. However, current solutions either cover only the mask/non-mask classifcation problem or require heavy computational resources that cannot be deployed in a computational limited system. To fll the gap in current research, we focus on constructing a deep learning model that achieves high-performance results with low processing time. Then, a correct face mask usage detection framework was proposed. The Raspberry Pi, a low-cost, credit-card-sized computer, and Raspberry Pi camera were deployed to implement our proposed framework. By leveraging transfer learning, with only 4-6 hours of the training session on approximately 20,000 images, the author achieves a model with accuracy up to 99% and FPS up to 7 FPS on a Raspberry Pi 4 model B. Our proposed framework enables organizations and schools to implement cost-effective correct face mask usage detection on constrained devices. The source code for this thesis is available in https: //github.com/minhrongcon2000/mask-recognition.