Mask detection system at the entry of a room

Abstract

This study focuses on automatic mask detection tools that can open doors in a room to minimize violations of health protocols, one of which is the use of masks during the pandemic. The method used in this study is the CNN classification method. Where the CNN calcification method has several stages in it, including pre-processing, training, and testing. In the pre-processing, all image data used will be labeled using Labeling.axe. The training process at CNN uses TensorFlow framework version 1.15. In the testing process, the test and data testing will be carried out in real-time by entering new images and models that are made and then a classification process is carried out on objects caught by the camera, classified images are marked with boxes and names of data classes. This data class is divided into two, namely data on wearing masks and without masks. The results of the test were carried out by entering 200 facial image data. The system can correctly detect as much as 190 times from 200 data tested with an Accuracy rate of 95%. Based on the test results, it shows that the resulting model is good and suitable for the classification process of recognizing mask detection images. However, to produce a better model requires data with more variety and a larger amount of data.