Detection of Face Mask Wearing Conditions with Lightweight CNN Models on Raspberry Pi 4 and Jetson Nano
Mark Austin L. Pagarigan, Jarold Rendell F. Reyes, Dionis A. Padilla · 2022 7th International Conference on Image, Vision and Computing (ICIVC) · 2022
There are still instances where an individual fails to correctly use a face mask, giving them a false sense of security. In a pandemic, automated systems for monitoring public health safety are critical. This research aims to develop an embedded system that uses lightweight CNN models to detect the condition of people’s face masks. The method works by employing a camera to capture the faces. It will then determine if the recognized faces in the photos are wearing their face masks appropriately or incorrectly, displayed on an LCD monitor. The trained models performed admirably, with MobileNetV2 achieving 97 percent accuracy. The NasNetMobile model fared the lowest, with an overall accuracy of 92 percent. However, it is still higher than 90%.