Face Mask Detection Under Low Light Condition Using Convolutional Neural Network (Cnn)

Abstract

The COVID-19 pandemic has been around for 3 years, and the virus is still spreading until now and using mask is an alternative for people to not get infected, but some people tend to let go of the mask for inconvenience reasons, especially under low light conditions which is difficult for humans to identify. Thus, this paper proposed and implemented a face mask detection model which can accurately detect a person that using a mask or not in such a condition as low light by using Convolutional Neural Network (CNN) architecture with OpenCV, TensorFlow and Keras. To achieve this, the first step is to transform the data by using Python Imaging Library (PIL) to create a low light image, then we process the data by using Contrast Limited Adaptive Histogram Equalization and with Gamma Correction. The second step is to augment the data by using TensorFlow ImageDataGenerator and define the CNN model. The final step is to create the face mask prediction by using Haar Cascade Algorithm to detect the face mask. The results of this research shows that CNN model can be trained with a recreational low light images to detect face mask under low light conditions. The result of the model produced an accuracy of 98%.

Using a One-Dimensional Convolutional Neural Network with Taguchi Parametric Optimization for a Permanent-Magnet Synchronous Motor Fault-Diagnosis System

“…The 1D CNN parameter design adopted in this study has seven control factors, one of which has two levels, and six control factors have three levels, as shown in Table 1. To reduce the number of experiments and improve the reliability of the experiments, according to the control factors, the number of levels, and the orthogonal array rules of the Taguchi method, L18(2 1 ,3 6 ) was selected for DOE, and the arrangement is shown in Table 2. The 1D CNN parameter design adopted in this study has seven control factors, one of which has two levels, and six control factors have three levels, as shown in Table 1.…”

“…[5] used motor vibration signals as analytic data and an extension neural network (ENN) to diagnose induction motor faults. Convolutional neural networks (CNNs) have been extensively used due to their outstanding characteristics and strong ability to extract features from complex information, such as face recognition [6], target tracking [7], target diagnosis [8], and timefrequency analysis [9]. Ref.…”

An Instance Segmentation Method for Nesting Green Sea Turtle’s Carapace using Mask R-CNN

“…Moreover, selecting the backbone of Mask R-CNN between ResNet-101 and ResNet50 is crucial because it will influence the trade-off between training time and accuracy. Thus, while ResNet-101 excels in detection and segmentation accuracy, it has a slower training time due to its numerous layers [13]. Since…”

“…However, the network sometimes misidentifies non-crystal spots outside the droplet as crystals, highlighting the need for enhance segmentation accuracy. Naufal et al [13] addressed the problem of face mask detection under low-light conditions during the Coronavirus disease (COVID-19) pandemic. The low-light condition can make image detection more difficult due to the presence of high-noise images, poor illumination, and reflectance.…”

RFID and Facemask Detector Attendance Monitoring System

“…However, accuracy can be affected by environment brightness. Low-light conditions make it challenging to distinguish faces from surroundings, while glare in bright light can hinder mask recognition [10]. Table 1 shows the system performs well across various lighting conditions, except in near darkness.…”