Detector Face Mask using UAV-based CNN Transfer Learning of YOLOv5

Abstract

Detection of the use of masks on someone is helpful in health protocols during the COVID-19 pandemic. All public services or places require people to wear masks during the pandemic. There are about three types of masks commonly used by the public today: surgical/medical masks, cloth masks, and scuba masks. This research aims to detect masks by monitoring a user using a mask through a camera. also detects the type of mask used by the community. So that it can provide convenience in implementing discipline in carrying out the COVID-19 health protocol using masks. In addition, this research proposes the detection of masks on the face by monitoring using a drone. The detection method used in this research is Transfer Learning CNN. This algorithm is a deep learning method that can classify and detect in digital image processing. The initial step of the research is to collect the types of masks on the market in the form of digital images, followed by the application before being modeled into mathematical calculations, which will later be processed using the Convolutional Neural Network method. This research compares two architectural transfer learning methods in deep learning, namely mobile net V2 with YOLOv5. The system testing process will be carried out by analyzing the recall value, precision, and accuracy. The testing process on drone camera-based devices uses the python programming language. Based on the results of the transfer learning method using YOLOv5, the results of the data training accuracy are 97% in detecting masks.