A Simple Mask Detection Model Based On A Multi-Layer Perception Neural Network

A global health crisis is appeared due to the rapid transmission of the COVID-19 pandemic. According to the World Health Organization (WHO), one of the effective ways to decrease this transmission is wearing masks in crowded places. However, monitoring people by police is a weary and difficult process. Thanks to the improvement in technology and artificial intelligence that make task became easier. In this paper, a simple mask recognition model based on texture and color moments features is proposed. This model is deployed in two stages: first, texture and color moments features from the face image (31 features) are extracted using a hybridization between texture features and color moments features techniques. In order to extract the texture features, the image transformed into Gray Level Co-Occurrence Matrix (GLCM) then 22 statistical metrics were calculated. So as to extract the color moments features, the first, second and third moments have been calculated from each layer of the RGB image. Second, based on the extracted features, the images are classified using a Multi-Layer Perceptron model (MLP). The dataset used in this research consists of 1787 real images with masks and 1918 without masks. The obtained results showed that the accuracy achieved by the proposed model is 90.58% and the time complexity is 6.7379 seconds for training and 0.0023 seconds for prediction.