Real-Time face mask Surveillance System for the Pandemic of Covid-19

Real-Time face mask Surveillance System for the Pandemic of Covid-19 | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Real-Time face mask Surveillance System for the Pandemic of Covid-19 Momal Iqbal This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4637920/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The disease was first discovered in Wuhan City, Hubei Province, the People’s Republic of China in late 2019, and rapidly grow to many countries around the world in early 2020, steadily turning into a global extensive pandemic. More than 222 million confirmed cases have been reported in different countries and regions around the world, and more than 4.6 million have died, which is one of the large-scale epidemics in human history . The coronavirus spreads through small droplets during the discussion, coughing, sneezing, etc. In poorly and closed ventilated locations a higher risk of transmission rate However, wearing a face mask that prevents the transmission of droplets in the air. But the continuous inspection of preventive measures both inside and outside the building/offices to prevent the growth of COVID-19 is a major challenging task. Therefore, in this research work, we focused on implementing a Face Mask Detection model that is relying on the related technologies of machine vision, we adopted three different well-known and the most advanced end-to-end target detection algorithm named CNN, VGG16, and -YOLOv5 to realize the detection and recognition of whether the face is wearing a mask. In terms of data set collection, we use the face mask opensource data set. After the actual effect test, we found the accuracy, error rate, recall rate, precision rate, and F1 of the Yolov5 algorithm model have reached a high level. This solution tracks the people with or without masks in a real-time scenario and highlighted the person with a red rectangle box in the case of violation. With the help of this 24/7, either inside or outside the organization continuously monitoring is possible and it has a great impact to identify the violator and ensure the safety of every individual. Numerical Analysis Software Engineering Face Mask Detection Deep Learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 1 Introduction The Severe acute respiratory syndrome (SARS) is the novel corona virus found in 2019 is known as Coronavirus2(SARS-Cov2). SARS-CoV-2 is a novel corona virus strain that has never been seen in humans. Some corona viruses can be passed from person to person, usually through close contact with an infected patient, such as among family members or in a health care[ 26 ]. The novel corona virus that causes COVID-19 respiratory disease can be passed from person to person by intimate contact with a potential or confirmed case. These secretions are released from the mouth or nose when a sick person cough. Sneezes, talks, or signs. This COVID-19 health disaster is now affecting the entire global world. In the serious condition, it may have influenza which causes death after the failure of the respiratory system. Millions of people are affected by it daily in the entire world and hence it becomes a pandemic. It has its different shapes and variants with damaging of repository system like Alpha variant, beta variant and delta variant etc[ 19 ]. Because there is currently no medication for COVID-19, every governments goal is to lower the death rate and the number of affected people. Every country is facing enormous economic obstacles, particularly developing countries whose wealth expansion is based on trade-offs between countries, resulting in increased global poverty. People are not taking this COVID-19 seriously since, according to them, if they are protected from coronavirus, they will die of hunger. Until now it has been observed that covid-19 is spreading when people are near to each other or with an object or surface transmission. Covid-19 is transmitted to another person with mouth, nose or eye when an infected person breathes, cough, sneeze or speak[ 21 ]. Coronavirus is now a day spreading everywhere, it has 614K confirmed deaths; the actual death rate is likely higher than total COVID-19 confirmed cases. It shows how highly the community is affected by this pandemic. As a result, a big part of the community is sacred and afraid of it. Every country is not so dangerous enough. The government of affected countries took immediate action to control this disease and some countries safe now or they have got somewhat control over this disease. Everyone is trying to be as careful as possible but they didn’t want to stop their life activities especially their business activities. They didn’t want to stop the supply chain of their business or necessary traveling. After several months of stagnation, all public and private sectors/offices are slowly resuming, and two major measures have been adopted: social distancing and wearing Face masks. (2021) social distancing (physical distancing) is one of the primary prevention strategies from the COVID-19 by minimizing close contact between people. The World Health Organization (WHO) recommends wearing face masks so they can protect people from getting infections[ 9 ]. All public and private sectors/offices hire active administrative members for continuous inspection of preventive measures both inside and outside the building/offices to prevent the growth of COVID-19 [ 23 ]. But still, we faced different challenges in manually monitoring the system as follow, 1. Time-consuming process for the large number of individuals to check masks before entering into the office/factory. 2. After inspections individual may remove their masks By thinking about above referred to challenges it is now no longer monitor to reveal all individual manually in any sectors, but the with the collaboration of computer vision and AI design the AI-based COVID SOPs monitoring systems that monitor SOPs 24/7. 2 LITERATURE REVIEW A.Face Mask Facial masks have become almost universal since the SARS CoV-2 epidemic. Fear of infection has prompted everyone to wear a face mask, leading to a shortage of products. Face mask policies vary from state to state. According to the World Health Organization, facial masks are not recommended for healthy people unless they are caring for a suspect with SARS or respiratory symptoms. However, it is always recommended to wear a face mask to prevent the transmission of the disease from asymptomatic people [12]. People with low or moderate risk of infection are encouraged to wear face masks in China, but those with very low risk of infection do not. For the sake of transition in the stage where it does not appear in it, it is necessary to be satisfied with the submission of the righteous conviction for the cause if the betrayal seems to be the cause. It is obligatory on those who are exposed to danger, such as those who are older than those who have basic problems, to be satisfied. In order to reduce the number of ferrous fluids such as influenza and ferruginous corona, it is necessary to reduce the amount of ferrous leukemia. Based on these latest discoveries, several EU governments have made it mandatory for people to wear masks in an effort to limit the spread of SARS-CoV-2 in the so-called second stage of the epidemic [16]. Combine the Viola-Jones technique with a simple HOG feature extractor. Necklace feature selection, integral image construction, adobe training, and cascade classifiers are used. The mouth, nose and ears are in the first place. It is believed that if a person’s face cannot be identified, he is wearing a mask. To detect masked faces, use CNN based techniques. Its CNN design consists of three layers, each of which eliminates false identification windows and further strengthens the system. On its unique MASKED FACE dataset, its model achieves an accuracy of 86.6 percent. Function Extractor for ipal Component Analysis (PCA). they used a feature extractor to perform principal component analysis (PCA). His method detects faces using the Viola-Jones algorithm, then processes images using PCA to extract features. The model’s best accuracy on masked faces was 73.75 percent [8]. B.Avoid public gatherings When you speak, you produce 50 times more virus-laden aerosols than when you don’t. These particles can infect everyone within a 16-foot radius in minutes. Aerosols are far smaller than dust particles and can last for hours in the air. Because aerosols get more concentrated over time, the longer you stay at a home party, the more probable you are to become ill if someone else in the vicinity becomes contaminated [6]. Remember that roughly 20 percent of persons infected with the corona virus have no symptoms, so determining who is safe is impossible. The virus can also spread in enclosed and/or poorly ventilated spaces where people choose to spend more time. This is because aerosols can linger in the air or travel far beyond talking distance (often called long-range aerosols or long range airborne transmissions) [10]. Its emphasized the importance of standardizing standards for smart city communications. Noah with his companions. CDC provides information about Middle East Respiratory Syndrome (MERS), Severe Acute Respiratory Syndrome (SARS), and COVID-19 medical signs and symptoms, treatment, diagnosis, transit routes, prevention techniques, and risk factors. Submitted. Review of the CDC’s (Centers for Disease Control and Prevention) website and PubMed’s Comprehensive Literature. C. Wash hands with soap for 20 seconds regularly According to existing evidence, the virus spreads largely among persons who are in close proximity to one another, such as within communication distances. The virus is transferred through the mouth or nose into little liquid particles when an infected person coughs, sneezes, speaks, sings, or breathes. When infectious particles from the air are breathed close together (also known as short-distance aerosols or short distance air transmissions) or come into direct contact with the eyes, nose, or mouth, another person can get infected (Drop dispersion) [15]. Use a hand sanitizer with at least 60 Percent alcohol if soap and water aren’t accessible. Cover your hands with a towel and rub them together until dry. D. Refrain from touching nose, eyes, and mouth with un washed hands Hands should be washed often with soap and water or rubbed with alcohol. If at all possible, carry an alcohol-based hand sanitizer with you and use it frequently. Coughing and sneezing should be covered with a bent elbow or tissue as quickly as possible and placed into a closed box. After that, use an alcohol-based hand sanitizer to wash or rub your hands. COVID 19 outbreak was ranked sixth vice president of the Public Health Emergencies (SPHEC) by the World Health Organization (WHO) on January 30, 2020. The corona virus has been known to spread before. SARS-COV outbreaks and Middle Eastern respiratory corona virus (MERS-COV) outbreaks were two prior corona virus epidemics.The COVID-19 pandemic was the third corona virus outbreak, in evolving more than 209 nations, including Pakistan. According to the World Health Organization (WHO), there have been 1,093,349 confirmed cases, with 58,620 deaths. To date, the United States has had the most positive cases, followed by Italy and Spain [18]. E. Stay at six feet distance from other people If at all possible, stay away from sick people. Keep a 6- foot space between the sick person and other family members if at all possible. If you’re caring for someone who’s sick, make sure they’re wearing the right masks and taking other measures. Stay at least 6 feet away from other people if you are not vaccinated with COVID-19 shots, especially if you are at high risk of serious disease from COVID-19 [6]. COVID-19 is most commonly spread among persons who have been close for a long time (within 6 feet). Drops from an infected person’s mouth or nose fly through the air and into the mouths or noses of others when he or she coughs, sneezes, or speaks. Inhaling the droplets is also an option. Infected but asymptomatic people are more likely to help transmit COVID 19, according to new research. Because people can spread the virus before they even realize they’re sick, it’s crucial to keep at least 6 feet away from others, regardless of whether you’re sick or not. People who are at a higher risk of contracting COVID-19 should avoid contact with others [5]. Here are a few basic social distancing tips: Consider social distancing methods to travel safely Limit contact when running errands Choose safe social activities Maintain distance at events and gatherings Maintain distance while being active Most countries have enacted lockdown and social distance measures in response to the COVID-19 epidemic, closing schools, training institutes, and higher education facilities. Teachers deliver high-quality instruction across a variety of online platforms, which is a welcome shift. Despite the difficulties faced by both teachers and students, continuing to learn online, through remote learning, and through education has shown to be the treatment for this unparalleled global disease. For students and professors alike, transitioning from traditional face-to-face learning to online learning can be a profoundly different experience in which they have little or no choice but to change. The educational system and instructors have accepted” emergency education” and are being pressured to adopt systems for which they are unprepared [7]. Table 1: Comparison of Different Face Mask Detecting Technique. Technique Type Accuracy Dataset CNN ] [25] Face Mask 86.6Percent Wider Face HSV Color Channel+CNN [2] Face Mask 90Percent MAFA HOG with ViolaJones [20] Face Mask 46.6Percent Custom YOLOv4 + R-CNN [22] Person 41.2Percent Oxford Town 3 RESEARCH METHODOLOGY Reserach methodology of the proposed framework to achieve the objective of the research as the COVID-19 pandemic is covering the entire global community. It immediately became an increasing factor, and no one knows who is suspected or infected by COVID-19, as symptoms may arise after 14 days. In common it may have signs of cough, fever, shortness of breath, and tiredness. In the serious condition, it may have influenza which causes death after the failure of the repository system . Millions of people are affected by it daily in the whole world and The World Health Organization (WHO) recommends wearing face masks so they can protect people from getting infections [11]. This preventive measures both inside and outside the building/offices to prevent the growth of COVID-19. While the proposed interactive framework as shown in Figure 1 helps to monitor the face mask detection automatically via using the AI approach. In this research work, we applied three well-known algorithms and select the best one for the inference engine on the bases of the performance evaluation metric [13].In the above mentioned proposed framework, we take the advantage of the Face Mask dataset contains thousands of face with different facial positions and illuminations, as well as people in indoor and outdoor settings, individual faces, partially occluded faces, and crowded pictures with blurred faces, all of which contribute to the effectiveness of AI-based face mask recognition. To achieve the goal of this study, the pre-process the dataset. A. Pre-processing on Dataset In pre-processing operation eliminate the irrelevant and dismissed images through an effective operation which might cut back the working out time, increase the accuracy of learning, and simplify higher empathetic for accurate learning model or data. The pre-processing needs to perform a series of operations [1]. 1. Image Resizing All images of the dataset are resized to the 48*48 size by utilizing image processing. In resizing images, the main problem raised is that the Images lose quality when you apply to zoom. When a limited number of pixels are placed on a screen with a greater resolution, new pixels must be created in order to fill the gaps that would otherwise appear. As a result, resize the photographs to the same size but with a greater resolution. Resize the photographs to the same scale but with a higher resolution as a consequence [17]. 2. Histogram Calculation After resizing, proceed with brightness and contrast adjustments. Even related papers prove that the histogram normalization effect is the most stable. Therefore, we normalize the images and histogram and remove the noise from images because too much noisy data will lead to poor model generalization. 3. Augmentation Image augmentation is a method of improving current pictures in direction to produce extra data for the model testing process. To put it in a different way, it is the method of vastly boosting the dataset useable for learning a deep learning model and, with the use of augmentation, emphasizing the as a whole or municipal features of the image, making the original uncertain image perfect, and increasing the size of a dataset for better training of the model to predict more results [24]. B. Train-Test split In the Train-Test split, split the whole dataset into 70/30. The 70 Percent portion is used for the training of an algorithm while the remaining 30 Percent is for testing to ensure the validity of the models (Yolo 5, CNN, and VGG19) C.Yolov5 Yolo is an advanced algorithm for a real-time object recognition system that you only look once. For real-time use, Yolo approaches outperform all other methods on all parameters. Rather of go for the most essential area of the Picture, in a single run of the algorithm, the Yolo technique predicts classes and bounding boxes for the entire frame [14]. YOLO v5 is the latest neural network structure from the Yolo family that is consists of 3 main modules, they are: (1) Backbone A convolutional neural network that gathers and creates picture feature at finegrained levels across many images. (2) Neck A set of network layers that mix and blend picture characteristics before them to the prediction layer. (3) Head Image features may be predicted, bounding boxes can be generated, and classification can be predicted. D. CNN CNNs are Deep Neural Networks that can recognize and classify certain picture features. They are usually applied in computer vision. Feature Extraction is the technique by which a convolution tool extracts and identifies the various characteristics of a picture for analysis. Fully linked layer that predicts the image class using the feature retrieved in the previous step and the output of the convolution procedure [3]. 1. Convolution layer In the field of math, a convolutional operation is described as the mixture of two functions. As usual, this operation is practiced as a filter. A kernel filters all that isn’t significant for the feature map, just focusing on some particular data. So, two components are required to execute this operation: Input data Convolution filter (kernel) A feature map is obtained as the result of this operation. The quantity of feature maps (output channels) furnishes the neural network with the ability to acquire attributes. Each channel has meant to take in each new element from the picture that is convoluted. That’s why each channel is independent [4]. There is an amazing case on the edges of input. One kind of padding will dispose of the border of the input; later there is no more involvement close to it that can be filtered. Then again, the other padding will finish the contribution with an estimation of 0. It involves decreasing parameters while convoluting. For an all-encompassing clarification on the process behind the convolution action, explained. 2. Spatial sub-sampling (pooling) Sub-sampling is an activity as well-known as pooling. The function of the pooling is to progressively reduce the spatial size of data representation to decrease parameters and computation in the model. In this way, this one decreases the impact of the feature position on the feature map by reducing its three-dimensional goal. After the convolutional operations, pooling chooses the most receptive pixel from the data. Each pooling layer works independently. Pooling has two common types: average pooling and max pooling. Average pooling processes average of defined area; while, the maxpooling just chooses the most elevated incentive on the region. The region size can prompt a decrease in the prediction execution if the worth is excessively enormous. The pooling activity stored a higher value of pixel using the given filter region [3]. By using the pooling layer, the dimension of the feature map is decreased. Reduction in feature map tells that the framework learns features from the position. At that point, it summarizes the feature of a new model. 3. Dropout It is the method that randomly chooses neurons that are ignored in training. It is a simple way to avoid overfitting the model. Giving more independence to every unit decreases solid unit bias promoting solid regularization and better generalization. 4. Fully Connected layer By integrating weights and biases with neurons, the fully connects the neurons among two separate layers. The final several layers of CNN architecture are often positioned before the output layer. E. VGG19 The VGG-19 Neural Network comprises 19 layers of deep neural network and is heavier. The VGG-19 network is 574MB in terms of completely connected nodes. The correctness of face mask identification improves as the number of layers’ increases. The Vgg-19 model consists of 19 deep trainable convolutional layers that are fully connected by max pooling and dropout layers. A dropout layer was utilized to normalize a convolutional layer that had been trained to execute a specialized classification role using a fully convolutional classifier. VGG-19 is really useful, and it just employs 3X3 convent arranged as above to extend the depth as a handler, max –pooling layers are used to reduce the size. FCN layers are two in number, each with 4096 neurons. To limit the frequency of false positives, VGG was trained on individual lesions and tested on all sorts of lesions[33]. Convolution layers put on the convolution process to image at apiece pixel, allowing the result to pass through the next layer. Filters are utilized in a 3X3 dimensional convolution layer that is learned for feature extraction. Each stacked convolution layer is then followed by a Rectified Linear Unit (ReLU) layer and a maxpooling layer [2]. ReLU is the most well-known nonlinear initiation function that accepts only the positive portion of the input. ReLU is very effective in determining the optimal convergence behavior that eliminates the gradient issue. Following The ReLU activation function, a down sampling max-pooling layer is utilized. IN general, a filter with two dimensions has the same step size. IN each subregion the output will be of maximum value. F. Inference System Inference Engine is a core component that implements knowledge-based reasoning in an expert system. It is the realization of knowledge-based reasoning in a computer. It mainly includes two aspects of reasoning and control. It is an indispensable and important part of the knowledge system. It is based on the best predictive model and performs prediction on new data and makes it more and more generalized that detects the wearing state of the face mask [5]. G. Tools To complete various tasks, we employ two distinct instruments. For this study, we use Google Collab and Python. Python is used in our research since it executes code line by line. Python is a dynamic, free, and open programming language. It is famous for rich selection of libraries, especially for machine learning. It is an open-source data analysis tool. The key component of IDE includes advance editing, code analysis tool, the Python console, charts, debugger etc [12]. 4 RESULTS AND DISCUSSION In this section, we examined the results obtained by applying three well-known learning models to the given dataset, including CNN, Yolov5, and VGG 19, and analyzing the outcomes for face mask identification. We used Performance Evaluation Metric to compare the prediction results of different models. 4.0.1 Dataset Distribution As discussed earlier data distribution in the train test split. Divide the data into two sections: training and testing. 70 percent of the records are devoted to model training. 30 Percent of the data is for testing. This distribution is pretty well described in the table [9]. In this data, there are two types of different classes which are as follows. Mask No- Mask Table 2: Data Distribution Dataset Number of images Training images 3507 Testing images 1503 In dataset, a certain number of images or features labelled with every mask and non-mask state as shown in Table 3. Table 3: Dataset Distribution No Class Number of images 1 Mask 2937 2 No Mask 2073 Class distribution is shown in the form of the bar graph below Figure 5 4.0.2 Performance Evaluation Matric PEM (Performance Evaluation Matric) is a key analytical tool used to evaluate the predictive ability of the model on the bases of Overall Accuracy, Error Rate, Recall Rate Precision Rate, F1 Score. All these are calculated by confusion matrix. 4.1 Overall Accuracy Accuracy refers to the proportion of the number of correctly classified samples to the total number of samples. It is commonly used methods to evaluate the generalization ability of a learning model. It can be measured by the equation no. 1 Model Accuracy= ( n ( True − Positives ) + m ( True − Negatives )) /N n ( True − Positives ) andm ( True − Negatives ) are number of correctly classified samples. N is total number of samples. Table 4: Accuracy table Algorithm Name Accuracy 1 CNN 0.946 1 VGG19 0.943 2 Yolov5 0.98 4.2 Error Rate Error rate refers to the proportion of misclassified samples to the total number of samples. It can be calculated using the equation no. 2, Error Rate= ( n ( True − Negatives +) m ( False − Positives )) /N Eq no. 2. n ( True − Negatives ) andm ( False − Positives ) are total number of wrongly classified samples. N is total number of samples. Table 5: Error Rate Algorithm Name Error Rate 1 CNN 0.057 1 VGG19 0.054 2 Yolov5 0.02 4.3 Precision Rate Precision Rate refers to the proportion of the true positive class among all the results predicted as the positive class. It can be defined as, Precision Rate=P= n ( True − Positives ) / ( n ( True − Positives )+ m ( False − Positives )) Eq no. 3. n ( True − Positives ) is total number of actual classified samples. m ( True − Positives ) is total number of wrong samples that are mistakenly classified to be correct. Table 6: Precision Rate Algorithm Name Precision 1 CNN 0.90 1 VGG19 0.89 2 Yolov5 0.97 4.4 Recall Rate Recall rate refers to the proportion of the number of correctly classified sample by the model to the number of all positive samples. Recall Rate=R= n ( True − Positives ) / ( n ( True − Positives ) + m ( False − Negatives )) Eq no. 4. n ( True − Positives ) is number of correctly classified samples. m ( False -Negatives ) is number of samples that are classified to be false. Table 7: Recall Rate Algorithm Name Recall Rate 1 CNN 0.93 1 VGG19 0.94 2 Yolov5 0.98 4.5 F1 Measurement F1-Measurement is the evaluation standard which is the harmonic average of the precision and recall and it can be measured by the given equation, F1 Measurement=R= 2. PR/(P+R ) Eq no. 5. P = Precision Rate and R = Recall Rate. Evaluate the model’s predictive Table 8: F1 Measurement Algorithm Name F1 Measurement 1 CNN 0.916 1 VGG19 0.913 2 Yolov5 0.97 ability on the test set according to the previously determined evaluation criteria and reveal that the Yolo v5 model is found to be best as compared to all other models as shown in Figure 9 5 Conclusion The Covid-19 Pandemic has significantly altered our routines. While waiting for the virus’s spread to decreased, it is essential to follow a set of guidelines, such as wearing face masks in public places. The World Health Organization (WHO) provides the standard recommendations on preventing the spread of COVID-19 and the importance of face masks for protection from the virus [ 24 ]. The goal of the following work is to develop a solution for automatically identifying infractions committed by people who are not wearing face masks. In this study, we conduct a comparative study for face mask detection We employed three deep learning models: Yolov5, CNN, and VGG-19. We train these models on face mask detection dataset. Yolov5 has a 98 Percent accuracy rate, CNN achieved 94.6 accuracy and VGG-19 accomplished the 94.3 accuracy. YoloV5 outperforms other competitor algorithms in terms of accuracy. This algorithm works well with various performance parameters. Our proposed methodology facilitates in the detection of face masks in public places and the prevention of Corona virus disease. 6 Future Work We have the ability to create a mobile application. In which we keep track of which employees are frequently breaking the regulations and wearing face masks or not in the university/office. If someone breaks the rules, a message will be sent to their mobile phone. References U Rajendra Acharya, Shu Lih Oh, Yuki Hagiwara, Jen Hong Tan, and Hojjat Adeli. Deep convolutional neural network for the automated detection and diagnosis of seizure using eeg signals. Computers in biology and medicine , 100:270–278, 2018. Nadeem Akhtar and U Ragavendran. Interpretation of intelligence in cnn-pooling processes: a methodological survey. Neural computing and applications , 32(3):879–898, 2020. Mohammed A Al-Masni, Mugahed A Al-Antari, Mun-Taek Choi, SeungMoo Han, and Tae-Seong Kim. Skin lesion segmentation in dermoscopy images via deep full resolution convolutional networks. Computer methods and programs in biomedicine , 162:221–231, 2018. Shaojie Bai. The Effect of Pre-ReLU Input Distribution on DNN’s Performance . PhD thesis, Carnegie Mellon University, 2017. Satilmis Bilgin, Ozge Kurtkulagi, Gizem Bakir Kahveci, Tuba Taslamacioglu Duman, and Burcin Meryem Atak Tel. Millennium pandemic: a review of coronavirus disease (covid-19). Experimental Biomedical Research , 3(2):117–125, 2020. Wei Bu, Jiangjian Xiao, Chuanhong Zhou, Minmin Yang, and Chengbin Peng. A cascade framework for masked face detection. In 2017 IEEE international conference on cybernetics and intelligent systems (CIS) and IEEE conference on robotics, automation and mechatronics (RAM) , pages 458–462. IEEE, 2017. Jane Chiodini. Maps, masks and media–traveller and practitioner resources for 2019 novel coronavirus (2019-ncov) acute respiratory virus. Travel medicine and infectious disease , 33:101574, 2020. Gayatri Deore, Ramakrishna Bodhula, Vishwas Udpikar, and Vidya More. Study of masked face detection approach in video analytics. In 2016 Conference on Advances in Signal Processing (CASP) , pages 196–200. IEEE, 2016. Anwesha Dutta and Harry W Fischer. The local governance of covid-19: Disease prevention and social security in rural india. World Development , 138:105234, 2021. Md Sabbir Ejaz, Md Rabiul Islam, Md Sifatullah, and Ananya Sarker. Implementation of principal component analysis on masked and nonmasked face recognition. In 2019 1st international conference on advances in science, engineering and robotics technology (ICASERT) , pages 1–5. IEEE, 2019. Ali Mamdouh Elkahky, Yang Song, and Xiaodong He. A multi-view deep learning approach for cross domain user modeling in recommendation systems. In Proceedings of the 24th international conference on world wide web , pages 278–288, 2015. Shuo Feng, Chen Shen, Nan Xia, Wei Song, Mengzhen Fan, and Benjamin J Cowling. Rational use of face masks in the covid-19 pandemic. The Lancet Respiratory Medicine , 8(5):434–436, 2020. Leon Gatys, Alexander S Ecker, and Matthias Bethge. Texture synthesis using convolutional neural networks. Advances in neural information processing systems , 28, 2015. A Victor Ikechukwu, S Murali, R Deepu, and RC Shivamurthy. Resnet50 vs vgg-19 vs training from scratch: A comparative analysis of the segmentation and classification of pneumonia from chest x-ray images. Global Transitions Proceedings , 2(2):375–381, 2021. Talha Ikram, Abdullah Saeed, NoorUl Ayn, Muhammad Ali Tahir, and Rafia Mumtaz. A review of the prevalent ict techniques used for covid19 sop violation detection. In 2020 IEEE 17th International Conference on Smart Communities: Improving Quality of Life Using ICT, IoT and AI (HONET) , pages 194–198. IEEE, 2020. Nancy HL Leung, Daniel KW Chu, Eunice YC Shiu, Kwok-Hung Chan, James J McDevitt, Benien JP Hau, Hui-Ling Yen, Yuguo Li, Dennis KM Ip, JS Peiris, et al. Respiratory virus shedding in exhaled breath and efficacy of face masks. Nature medicine , 26(5):676–680, 2020. Muhammad Mateen, Junhao Wen, Nasrullah, Sun Song, and Zhouping Huang. Fundus image classification using vgg-19 architecture with pca and svd. Symmetry , 11(1):1, 2018. Noah C Peeri, Nistha Shrestha, Md Siddikur Rahman, Rafdzah Zaki, Zhengqi Tan, Saana Bibi, Mahdi Baghbanzadeh, Nasrin Aghamohammadi, Wenyi Zhang, and Ubydul Haque. The sars, mers and novel coronavirus (covid-19) epidemics, the newest and biggest global health threats: what lessons have we learned? International journal of epidemiology , 49(3):717–726, 2020. Jitendra Singh Rathore and Chaitali Ghosh. Severe acute respiratory syndrome coronavirus-2 (sars-cov-2), a newly emerged pathogen: an overview. Pathogens and disease , 78(6):ftaa042, 2020. Connor Shorten and Taghi M Khoshgoftaar. A survey on image data augmentation for deep learning. Journal of big data , 6(1):1–48, 2019. Stanislaw P Stawicki, Rebecca Jeanmonod, Andrew C Miller, Lorenzo Paladino, David F Gaieski, Anna Q Yaffee, Annelies De Wulf, Joydeep Grover, Thomas J Papadimos, Christina Bloem, et al. The 2019–2020 novel coronavirus (severe acute respiratory syndrome coronavirus 2) pandemic: A joint american college of academic international medicineworld academic council of emergency medicine multidisciplinary covid19 working group consensus paper. Journal of global infectious diseases , 12(2):47–93, 2020. Yanan Sun, Bing Xue, Mengjie Zhang, Gary G Yen, and Jiancheng Lv. Automatically designing cnn architectures using the genetic algorithm for image classification. IEEE transactions on cybernetics , 50(9):3840– 3854, 2020. Yu Wang, Huaiyu Tian, Li Zhang, Man Zhang, Dandan Guo, Wenting Wu, Xingxing Zhang, Ge Lin Kan, Lei Jia, Da Huo, et al. Reduction of secondary transmission of sars-cov-2 in households by face mask use, disinfection and social distancing: a cohort study in beijing, china. BMJ global health , 5(5):e002794, 2020. Jian Xiao, Jia Wang, Shaozhong Cao, and Bilong Li. Application of a novel and improved vgg-19 network in the detection of workers wearing masks. In Journal of Physics: Conference Series , volume 1518, page 012041. IOP Publishing, 2020. Metin Yıldız, Yakup Sarpdag˘ı, Mehmet Okuyar, Mehmet Yildiz, Necmettin C¸iftci, Ay¸se Elkoca, Mehmet Salih Yildirim, Muhammet Ali Aydin, Mehmet Parlak, and Bu¨nyamin Bingo¨l. Segmentation and classification of skin burn images with artificial intelligence: Development of a mobile application. Burns , 2024. Xinkai Zhou, Zhigui Wu, Ranran Yu, Shanni Cao, Wen Fang, Zhen Jiang, Fang Yuan, Chao Yan, and Dijun Chen. Modelling-based evaluation of the effect of quarantine control by the chinese government in the coronavirus disease 2019 outbreak. MedRxiv , pages 2020–03, 2020. Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4637920","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":318957210,"identity":"9909d17a-6175-4371-8588-454111ac4fc0","order_by":0,"name":"Momal Iqbal","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3klEQVRIiWNgGAWjYLCCBwYMCQzsjI0PgGwePqK0JIC0MDM3G4C0sBGnBYSY2dskQByCWnTbuxM/JBRY5/E3M7ZVfs2xk2FjYH746AYeLWZnzm6WSDBIL5Y4zNh2W3ZbMtBhbMbGOfi03MjdANRyOLEBpEVyGzNQCw+bNAEtm3+AtMwHaimW3FZPlJZtYFs2ALUwftx2mAgtZ85uswD5xfAwY7M047bjPGzMhPxyvHfzjQ9/rPPkjrc//PhzW7U9P3vzw8f4tEABM4TkQbCJ1ML4gzjVo2AUjIJRMMIAACZ9Sh3vmmeIAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0009-0004-8770-0241","institution":"Air University","correspondingAuthor":true,"prefix":"","firstName":"Momal","middleName":"","lastName":"Iqbal","suffix":""}],"badges":[],"createdAt":"2024-06-25 16:41:50","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-4637920/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4637920/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":59121451,"identity":"753b699f-b946-439b-ba17-da9334cebca9","added_by":"auto","created_at":"2024-06-26 14:55:26","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":137530,"visible":true,"origin":"","legend":"\u003cp\u003eBlock Diagram\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-4637920/v1/36b14b11a4681d6016cfb56d.png"},{"id":59121452,"identity":"3ba65ff2-fca8-4452-b00d-2f3cebf3d386","added_by":"auto","created_at":"2024-06-26 14:55:26","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":145855,"visible":true,"origin":"","legend":"\u003cp\u003e2 Yolov5 architecture\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-4637920/v1/3eaef434770180cb484b285d.png"},{"id":59121449,"identity":"fd241216-8e31-4a71-9fce-22678550bdc8","added_by":"auto","created_at":"2024-06-26 14:55:26","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":53750,"visible":true,"origin":"","legend":"\u003cp\u003eCNN architecture\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-4637920/v1/0c38e90aa15c3f2ba2d61320.png"},{"id":59120550,"identity":"a47088bb-975c-4b19-92d2-720bcf18712c","added_by":"auto","created_at":"2024-06-26 14:47:26","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":74866,"visible":true,"origin":"","legend":"\u003cp\u003eVGG19 Architecture\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-4637920/v1/3f8b6d3a8e8bb149ab999242.png"},{"id":59120544,"identity":"54c92ee5-c9f3-48f7-8242-f2877978ffca","added_by":"auto","created_at":"2024-06-26 14:47:26","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":32172,"visible":true,"origin":"","legend":"\u003cp\u003eData set Class distribution\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-4637920/v1/de143717bf8f1ff6b44e90e9.png"},{"id":59122230,"identity":"698d6e4b-7e03-4dbf-a13b-7d0aa63660b2","added_by":"auto","created_at":"2024-06-26 15:03:27","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":27814,"visible":true,"origin":"","legend":"\u003cp\u003eCNN Confusion Matric\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-4637920/v1/1257a8818730d817113ea801.png"},{"id":59122228,"identity":"63243f84-eb04-4a5f-98a1-9c3dfd26c2bc","added_by":"auto","created_at":"2024-06-26 15:03:26","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":27933,"visible":true,"origin":"","legend":"\u003cp\u003eVGG19 Confusion Matric\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-4637920/v1/4c80b5e1e9473f2e5e4b38ff.png"},{"id":59123107,"identity":"d638d201-01a3-4ef3-841e-a97fb06bad81","added_by":"auto","created_at":"2024-06-26 15:11:26","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":27908,"visible":true,"origin":"","legend":"\u003cp\u003eyolov5 Confusion Matric\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-4637920/v1/5d91919f686b029da2f349e7.png"},{"id":59120545,"identity":"2f12ce39-63b1-4484-abe6-54c0293dc715","added_by":"auto","created_at":"2024-06-26 14:47:26","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":51328,"visible":true,"origin":"","legend":"\u003cp\u003eModel Performance Evaluation Graph\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-4637920/v1/632202f69f2d863e6e2181b3.png"},{"id":59123996,"identity":"1421e7fe-38d8-46d9-8723-116a54130752","added_by":"auto","created_at":"2024-06-26 15:19:27","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1115950,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4637920/v1/6bffeeb5-d46a-4aa5-a343-89a1d65a8579.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eReal-Time face mask Surveillance System for the Pandemic of Covid-19\u003c/p\u003e","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eThe Severe acute respiratory syndrome (SARS) is the novel corona virus found in 2019 is known as Coronavirus2(SARS-Cov2). SARS-CoV-2 is a novel corona virus strain that has never been seen in humans. Some corona viruses can be passed from person to person, usually through close contact with an infected patient, such as among family members or in a health care[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. The novel corona virus that causes COVID-19 respiratory disease can be passed from person to person by intimate contact with a potential or confirmed case. These secretions are released from the mouth or nose when a sick person cough. Sneezes, talks, or signs. This COVID-19 health disaster is now affecting the entire global world. In the serious condition, it may have influenza which causes death after the failure of the respiratory system. Millions of people are affected by it daily in the entire world and hence it becomes a pandemic. It has its different shapes and variants with damaging of repository system like Alpha variant, beta variant and delta variant etc[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Because there is currently no medication for COVID-19, every governments goal is to lower the death rate and the number of affected people. Every country is facing enormous economic obstacles, particularly developing countries whose wealth expansion is based on trade-offs between countries, resulting in increased global poverty. People are not taking this COVID-19 seriously since, according to them, if they are protected from coronavirus, they will die of hunger. Until now it has been observed that covid-19 is spreading when people are near to each other or with an object or surface transmission. Covid-19 is transmitted to another person with mouth, nose or eye when an infected person breathes, cough, sneeze or speak[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Coronavirus is now a day spreading everywhere, it has 614K confirmed deaths; the actual death rate is likely higher than total COVID-19 confirmed cases. It shows how highly the community is affected by this pandemic. As a result, a big part of the community is sacred and afraid of it. Every country is not so dangerous enough. The government of affected countries took immediate action to control this disease and some countries safe now or they have got somewhat control over this disease. Everyone is trying to be as careful as possible but they didn\u0026rsquo;t want to stop their life activities especially their business activities. They didn\u0026rsquo;t want to stop the supply chain of their business or necessary traveling. After several months of stagnation, all public and private sectors/offices are slowly resuming, and two major measures have been adopted: social distancing and wearing Face masks. (2021) social distancing (physical distancing) is one of the primary prevention strategies from the COVID-19 by minimizing close contact between people. The World Health Organization (WHO) recommends wearing face masks so they can protect people from getting infections[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. All public and private sectors/offices hire active administrative members for continuous inspection of preventive measures both inside and outside the building/offices to prevent the growth of COVID-19 [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. But still, we faced different challenges in manually monitoring the system as follow, 1. Time-consuming process for the large number of individuals to check masks before entering into the office/factory. 2. After inspections individual may remove their masks By thinking about above referred to challenges it is now no longer monitor to reveal all individual manually in any sectors, but the with the collaboration of computer vision and AI design the AI-based COVID SOPs monitoring systems that monitor SOPs 24/7.\u003c/p\u003e"},{"header":"2 LITERATURE REVIEW","content":"\u003ch2\u003eA.Face Mask\u003c/h2\u003e\n\u003cp\u003eFacial masks have become almost universal since the SARS CoV-2 epidemic. Fear of infection has prompted everyone to wear a face mask, leading to a shortage of products. Face mask policies vary from state to state. According to the World Health Organization, facial masks are not recommended for healthy people unless they are caring for a suspect with SARS or respiratory symptoms. However, it is always recommended to wear a face mask to prevent the transmission of the disease from asymptomatic people [12]. People with low or moderate risk of infection are encouraged to wear face masks in China, but those with very low risk of infection do not. For the sake of transition in the stage where it does not appear in it, it is necessary to be satisfied with the submission of the righteous conviction for the cause if the betrayal seems to be the cause. It is obligatory on those who are exposed to danger, such as those who are older than those who have basic problems, to be satisfied. In order to reduce the number of ferrous fluids such as influenza and ferruginous corona, it is necessary to reduce the amount of ferrous\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eleukemia. Based on these latest discoveries, several EU governments have made it mandatory for people to wear masks in an effort to limit the spread of SARS-CoV-2 in the so-called second stage of the epidemic [16]. Combine the Viola-Jones technique with a simple HOG feature extractor. Necklace feature selection, integral image construction, adobe training, and cascade classifiers are used. The mouth, nose and ears are in the first place. It is believed that if a person\u0026rsquo;s face cannot be identified, he is wearing a mask. To detect masked faces, use CNN based techniques. Its CNN design consists of three layers, each of which eliminates false identification windows and further strengthens the system. On its unique MASKED FACE dataset, its model achieves an accuracy of 86.6 percent. Function Extractor for ipal Component Analysis (PCA). they used a feature extractor to perform principal component analysis (PCA). His method detects faces using the Viola-Jones algorithm, then processes images using PCA to extract features. The model\u0026rsquo;s best accuracy on masked faces was 73.75 percent [8].\u003c/p\u003e\n\u003ch2\u003eB.Avoid public gatherings\u003c/h2\u003e\n\u003cp\u003eWhen you speak, you produce 50 times more virus-laden aerosols than when you don\u0026rsquo;t. These particles can infect everyone within a 16-foot radius in minutes. Aerosols are far smaller than dust particles and can last for hours in the air. Because aerosols get more concentrated over time, the longer you stay at a home party, the more probable you are to become ill if someone else in the vicinity becomes contaminated [6]. Remember that roughly 20 percent of persons infected with the corona virus have no symptoms, so determining who is safe is impossible. The virus can also spread in enclosed and/or poorly ventilated spaces where people choose to spend more time. This is because aerosols can linger in the air or travel far beyond talking distance (often called long-range aerosols or long range airborne transmissions) [10]. Its emphasized the importance of standardizing standards for smart city communications. Noah with his companions. CDC provides information about Middle East Respiratory Syndrome (MERS), Severe Acute Respiratory Syndrome (SARS), and COVID-19 medical signs and symptoms, treatment, diagnosis, transit routes, prevention techniques, and risk factors. Submitted. Review of the CDC\u0026rsquo;s (Centers for Disease Control and Prevention) website and PubMed\u0026rsquo;s Comprehensive Literature.\u003c/p\u003e\n\u003ch2\u003eC. Wash hands with soap for 20 seconds regularly\u003c/h2\u003e\n\u003cp\u003eAccording to existing evidence, the virus spreads largely among persons who are in close proximity to one another, such as within communication distances. The virus is transferred through the mouth or nose into little liquid particles when an infected person coughs, sneezes, speaks, sings, or breathes. When infectious particles from the air are breathed close together (also known as short-distance aerosols or short distance air transmissions) or come into direct contact with the eyes, nose, or mouth, another person can get infected (Drop dispersion) [15]. Use a hand sanitizer with at least 60 Percent alcohol if soap and water aren\u0026rsquo;t accessible. Cover your hands with a towel and rub them together until dry.\u003c/p\u003e\n\u003ch2\u003eD. Refrain from touching nose, eyes, and mouth with un washed hands\u003c/h2\u003e\n\u003cp\u003eHands should be washed often with soap and water or rubbed with alcohol. If at all possible, carry an alcohol-based hand sanitizer with you and use it frequently. Coughing and sneezing should be covered with a bent elbow or tissue as quickly as possible and placed into a closed box. After that, use an alcohol-based hand sanitizer to wash or rub your hands. COVID 19 outbreak was ranked sixth vice president of the Public Health Emergencies (SPHEC) by the World Health Organization (WHO) on January 30, 2020. The corona virus has been known to spread before. SARS-COV outbreaks and Middle Eastern respiratory corona virus (MERS-COV) outbreaks were two prior corona virus epidemics.The COVID-19 pandemic was the third corona virus outbreak, in evolving more than 209 nations, including Pakistan. According to the World Health Organization (WHO), there have been 1,093,349 confirmed cases, with 58,620 deaths. To date, the United States has had the most positive cases, followed by Italy and Spain [18].\u003c/p\u003e\n\u003ch2\u003eE. Stay at six feet distance from other people\u003c/h2\u003e\n\u003cp\u003eIf at all possible, stay away from sick people. Keep a 6- foot space between the sick person and other family members if at all possible. If you\u0026rsquo;re caring for someone who\u0026rsquo;s sick, make sure they\u0026rsquo;re wearing the right masks and taking other measures. Stay at least 6 feet away from other people if you are not vaccinated with COVID-19 shots, especially if you are at high risk of serious disease from COVID-19 [6]. COVID-19 is most commonly spread among persons who have been close for a long time (within 6 feet). Drops from an infected person\u0026rsquo;s mouth or nose fly through the air and into the mouths or noses of others when he or she coughs, sneezes, or speaks. Inhaling the droplets is also an option. Infected but asymptomatic people are more likely to help transmit COVID 19, according to new research. Because people can spread the virus before they even realize they\u0026rsquo;re sick, it\u0026rsquo;s crucial to keep at least 6 feet away from others, regardless of whether you\u0026rsquo;re sick or not. People who are at a higher risk of contracting COVID-19 should avoid contact with others [5].\u003c/p\u003e\n\u003cp\u003eHere are a few basic social distancing tips:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eConsider social distancing methods to travel safely\u003c/li\u003e\n \u003cli\u003eLimit contact when running errands\u003c/li\u003e\n \u003cli\u003eChoose safe social activities\u003c/li\u003e\n \u003cli\u003eMaintain distance at events and gatherings\u003c/li\u003e\n \u003cli\u003eMaintain distance while being active\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eMost countries have enacted lockdown and social distance measures in response to the COVID-19 epidemic, closing schools, training institutes, and higher education facilities. Teachers deliver high-quality instruction across a variety of online platforms, which is a welcome shift. Despite the difficulties faced by both teachers and students, continuing to learn online, through remote learning, and through education has shown to be the treatment for this unparalleled global disease. For students and professors alike, transitioning from traditional face-to-face learning to online learning can be a profoundly different experience in which they have little or no choice but to change. The educational system and instructors have accepted\u0026rdquo; emergency education\u0026rdquo; and are being pressured to adopt systems for which they are unprepared [7].\u003c/p\u003e\n\u003cp\u003eTable 1: Comparison of Different Face Mask Detecting Technique.\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"389\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"44.473007712082264%\" valign=\"top\"\u003e\n \u003cp\u003eTechnique\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"17.480719794344473%\" valign=\"top\"\u003e\n \u003cp\u003eType\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.76606683804627%\" valign=\"top\"\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.28020565552699%\" valign=\"top\"\u003e\n \u003cp\u003eDataset\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"44.473007712082264%\" valign=\"top\"\u003e\n \u003cp\u003eCNN ] [25]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"17.480719794344473%\" valign=\"top\"\u003e\n \u003cp\u003eFace Mask\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.76606683804627%\" valign=\"top\"\u003e\n \u003cp\u003e86.6Percent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.28020565552699%\" valign=\"top\"\u003e\n \u003cp\u003eWider Face\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"44.473007712082264%\" valign=\"top\"\u003e\n \u003cp\u003eHSV Color Channel+CNN [2]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"17.480719794344473%\" valign=\"top\"\u003e\n \u003cp\u003eFace Mask\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.76606683804627%\" valign=\"top\"\u003e\n \u003cp\u003e90Percent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.28020565552699%\" valign=\"top\"\u003e\n \u003cp\u003eMAFA\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"44.473007712082264%\" valign=\"top\"\u003e\n \u003cp\u003eHOG with ViolaJones [20]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"17.480719794344473%\" valign=\"top\"\u003e\n \u003cp\u003eFace Mask\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.76606683804627%\" valign=\"top\"\u003e\n \u003cp\u003e46.6Percent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.28020565552699%\" valign=\"top\"\u003e\n \u003cp\u003eCustom\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"44.473007712082264%\" valign=\"top\"\u003e\n \u003cp\u003eYOLOv4 + R-CNN [22]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"17.480719794344473%\" valign=\"top\"\u003e\n \u003cp\u003ePerson\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.76606683804627%\" valign=\"top\"\u003e\n \u003cp\u003e41.2Percent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.28020565552699%\" valign=\"top\"\u003e\n \u003cp\u003eOxford Town\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"3\tRESEARCH METHODOLOGY","content":"\u003cp\u003eReserach methodology of the proposed framework to achieve the objective of the research as the COVID-19 pandemic is covering the entire global community. It immediately became an increasing factor, and no one knows who is suspected or infected by COVID-19, as symptoms may arise after 14 days. In common it may have signs of cough, fever, shortness of breath, and tiredness. In the serious condition, it may have influenza which causes death after the failure of the repository system . Millions of people are affected by it daily in the whole world and The World Health Organization (WHO) recommends wearing face masks so they can protect people from getting infections [11]. This preventive measures both inside and outside the building/offices to prevent the growth of COVID-19. While the proposed interactive framework as shown in Figure 1 helps to monitor the face mask detection automatically via using the AI approach. In this research work, we applied three well-known algorithms and select the best one for the inference engine on the bases of the performance evaluation metric [13].In the above mentioned proposed framework, we take the advantage of the Face Mask dataset contains thousands of face with different facial positions and illuminations, as well as people in indoor and outdoor settings, individual faces, partially occluded faces, and crowded pictures with blurred faces, all of which contribute to the effectiveness of AI-based face mask recognition. To achieve the goal of this study, the pre-process the dataset.\u003c/p\u003e\n\u003ch2\u003eA. Pre-processing on Dataset\u003c/h2\u003e\n\u003cp\u003eIn pre-processing operation eliminate the irrelevant and dismissed images through an effective operation which might cut back the working out time, increase the accuracy of learning, and simplify higher empathetic for accurate learning model or data. The pre-processing needs to perform a series of operations [1].\u003c/p\u003e\n\u003ch3\u003e1. Image Resizing\u003c/h3\u003e\n\u003cp\u003eAll images of the dataset are resized to the 48*48 size by utilizing image processing. In resizing images, the main problem raised is that the Images lose quality when you apply to zoom. When a limited number of pixels are placed on a screen with a greater resolution, new pixels must be created in order to fill the gaps that would otherwise appear. As a result, resize the photographs to the same size but with a greater resolution. Resize the photographs to the same scale but with a higher resolution as a consequence [17].\u003c/p\u003e\n\u003ch3\u003e2. Histogram Calculation\u003c/h3\u003e\n\u003cp\u003eAfter resizing, proceed with brightness and contrast adjustments. Even related papers prove that the histogram normalization effect is the most stable. Therefore, we normalize the images and histogram and remove the noise from images because too much noisy data will lead to poor model generalization.\u003c/p\u003e\n\u003ch3\u003e3. Augmentation\u003c/h3\u003e\n\u003cp\u003eImage augmentation is a method of improving current pictures in direction to produce extra data for the model testing process. To put it in a different way, it is the method of vastly boosting the dataset useable for learning a deep learning model and, with the use of augmentation, emphasizing the as a whole or municipal features of the image, making the original uncertain image perfect, and increasing the size of a dataset for better training of the model to predict more results [24].\u003c/p\u003e\n\u003ch2\u003eB. Train-Test split\u003c/h2\u003e\n\u003cp\u003eIn the Train-Test split, split the whole dataset into 70/30. The 70 Percent portion is used for the training of an algorithm while the remaining 30 Percent is for testing to ensure the validity of the models (Yolo 5, CNN, and VGG19)\u003c/p\u003e\n\u003ch2\u003eC.Yolov5\u003c/h2\u003e\n\u003cp\u003eYolo is an advanced algorithm for a real-time object recognition system that you only look once. For real-time use, Yolo approaches outperform all other methods on all parameters. Rather of go for the most essential area of the Picture, in a single run of the algorithm, the Yolo technique predicts classes and bounding boxes for the entire frame [14]. YOLO v5 is the latest neural network structure from the Yolo family that is consists of 3 main modules, they are:\u003c/p\u003e\n\u003ch3\u003e(1) Backbone\u003c/h3\u003e\n\u003cp\u003eA convolutional neural network that gathers and creates picture feature at finegrained levels across many images.\u003c/p\u003e\n\u003ch3\u003e(2) Neck\u003c/h3\u003e\n\u003cp\u003eA set of network layers that mix and blend picture characteristics before them to the prediction layer.\u003c/p\u003e\n\u003ch3\u003e(3) Head\u003c/h3\u003e\n\u003cp\u003eImage features may be predicted, bounding boxes can be generated, and classification can be predicted.\u003c/p\u003e\n\u003ch2\u003eD. CNN\u003c/h2\u003e\n\u003cp\u003eCNNs are Deep Neural Networks that can recognize and classify certain picture features. They are usually applied in computer vision. Feature Extraction is the technique by which a convolution tool extracts and identifies the various characteristics of a picture for analysis. Fully linked layer that predicts the image class using the feature retrieved in the previous step and the output of the convolution procedure [3].\u003c/p\u003e\n\u003ch3\u003e1. Convolution layer\u003c/h3\u003e\n\u003cp\u003eIn the field of math, a convolutional operation is described as the mixture of two functions. As usual, this operation is practiced as a filter. A kernel filters all that isn\u0026rsquo;t significant for the feature map, just focusing on some particular data. So, two components are required to execute this operation:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eInput data\u003c/li\u003e\n \u003cli\u003eConvolution filter (kernel)\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eA feature map is obtained as the result of this operation. The quantity of feature maps (output channels) furnishes the neural network with the ability to acquire attributes. Each channel has meant to take in each new element from the picture that is convoluted. That\u0026rsquo;s why each channel is independent [4]. There is an amazing case on the edges of input. One kind of padding will dispose of the border of the input; later there is no more involvement close to it that can be filtered. Then again, the other padding will finish the contribution with an estimation of 0. It involves decreasing parameters while convoluting. For an all-encompassing clarification on the process behind the convolution action, explained.\u003c/p\u003e\n\u003ch3\u003e2. Spatial sub-sampling (pooling)\u003c/h3\u003e\n\u003cp\u003eSub-sampling is an activity as well-known as pooling. The function of the pooling is to progressively reduce the spatial size of data representation to decrease parameters and computation in the model. In this way, this one decreases the impact of the feature position on the feature map by reducing its three-dimensional goal. After the convolutional operations, pooling chooses the most receptive pixel from the data. Each pooling layer works independently. Pooling has two common types: average pooling and max pooling. Average pooling processes average of defined area; while, the maxpooling just chooses the most elevated incentive on the region. The region size can prompt a decrease in the prediction execution if the worth is excessively enormous. The pooling activity stored a higher value of pixel using the given filter region [3]. By using the pooling layer, the dimension of the feature map is decreased. Reduction in feature map tells that the framework learns features from the position. At that point, it summarizes the feature of a new model.\u003c/p\u003e\n\u003ch3\u003e3. Dropout\u003c/h3\u003e\n\u003cp\u003eIt is the method that randomly chooses neurons that are ignored in training. It is a simple way to avoid overfitting the model. Giving more independence to every unit decreases solid unit bias promoting solid regularization and better generalization.\u003c/p\u003e\n\u003ch3\u003e4. Fully Connected layer\u003c/h3\u003e\n\u003cp\u003eBy integrating weights and biases with neurons, the fully connects the neurons among two separate layers. The final several layers of CNN architecture are often positioned before the output layer.\u003c/p\u003e\n\u003ch3\u003eE. VGG19\u003c/h3\u003e\n\u003cp\u003eThe VGG-19 Neural Network comprises 19 layers of deep neural network and is heavier. The VGG-19 network is 574MB in terms of completely connected nodes. The correctness of face mask identification improves as the number of layers\u0026rsquo; increases. The Vgg-19 model consists of 19 deep trainable convolutional layers that are fully connected by max pooling and dropout layers. A dropout layer was utilized to normalize a convolutional layer that had been trained to execute a specialized classification role using a fully convolutional classifier. VGG-19 is really useful, and it just employs 3X3 convent arranged as above to extend the depth as a handler, max \u0026ndash;pooling layers are used to reduce the size. FCN layers are two in number, each with 4096 neurons. To limit the frequency of false positives, VGG was trained on individual lesions and tested on all sorts of lesions[33]. Convolution layers put on the convolution process to image at apiece pixel, allowing the result to pass through the next layer. Filters are utilized in a 3X3 dimensional convolution layer that is learned for feature extraction. Each stacked convolution layer is then followed by a Rectified Linear Unit (ReLU) layer and a maxpooling layer [2]. ReLU is the most well-known nonlinear initiation function that accepts only the positive portion of the input. ReLU is very effective in determining the optimal convergence behavior that eliminates the gradient issue. Following The ReLU activation function, a down sampling max-pooling layer is utilized. IN general, a filter with two dimensions has the same step size. IN each subregion the output will be of maximum value.\u003c/p\u003e\n\u003ch2\u003eF. Inference System\u003c/h2\u003e\n\u003cp\u003eInference Engine is a core component that implements knowledge-based reasoning in an expert system. It is the realization of knowledge-based reasoning in a computer. It mainly includes two aspects of reasoning and control. It is an indispensable and important part of the knowledge system. It is based on the best predictive model and performs prediction on new data and makes it more and more generalized that detects the wearing state of the face mask\u003c/p\u003e\n\u003cp\u003e[5].\u003c/p\u003e\n\u003ch2\u003eG. Tools\u003c/h2\u003e\n\u003cp\u003eTo complete various tasks, we employ two distinct instruments. For this study, we use Google Collab and Python. Python is used in our research since it executes code line by line. Python is a dynamic, free, and open programming language. It is famous for rich selection of libraries, especially for machine learning. It is an open-source data analysis tool. The key component of IDE includes advance editing, code analysis tool, the Python console, charts, debugger etc [12].\u003c/p\u003e"},{"header":"4 RESULTS AND DISCUSSION","content":"\u003cp\u003eIn this section, we examined the results obtained by applying three well-known learning models to the given dataset, including CNN, Yolov5, and VGG 19, and analyzing the outcomes for face mask identification. We used Performance Evaluation Metric to compare the prediction results of different models.\u003c/p\u003e\n\u003ch2\u003e4.0.1 \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Dataset Distribution\u003c/h2\u003e\n\u003cp\u003eAs discussed earlier data distribution in the train test split. Divide the data into two sections: training and testing. 70 percent of the records are devoted to model training. 30 Percent of the data is for testing. This distribution is pretty well described in the table [9]. In this data, there are two types of different classes which are as follows.\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eMask\u003c/li\u003e\n \u003cli\u003eNo- Mask\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eTable 2: Data Distribution\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"203\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"51.23152709359606%\" valign=\"top\"\u003e\n \u003cp\u003eDataset\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"48.76847290640394%\" valign=\"top\"\u003e\n \u003cp\u003eNumber of images\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"51.23152709359606%\" valign=\"top\"\u003e\n \u003cp\u003eTraining images\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"48.76847290640394%\" valign=\"top\"\u003e\n \u003cp\u003e3507\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"51.23152709359606%\" valign=\"top\"\u003e\n \u003cp\u003eTesting images\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"48.76847290640394%\" valign=\"top\"\u003e\n \u003cp\u003e1503\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eIn dataset, a certain number of images or features labelled with every mask and non-mask state as shown in Table 3.\u003c/p\u003e\n\u003cp\u003eTable 3: Dataset Distribution\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"199\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"19.696969696969695%\" valign=\"top\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.303030303030305%\" valign=\"top\"\u003e\n \u003cp\u003eClass\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003eNumber of images\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"19.696969696969695%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.303030303030305%\" valign=\"top\"\u003e\n \u003cp\u003eMask\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003e2937\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"19.696969696969695%\" valign=\"top\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.303030303030305%\" valign=\"top\"\u003e\n \u003cp\u003eNo Mask\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003e2073\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eClass distribution is shown in the form of the bar graph below Figure 5\u003c/p\u003e\n\u003ch2\u003e4.0.2 \u0026nbsp; \u0026nbsp; \u0026nbsp; Performance Evaluation Matric\u003c/h2\u003e\n\u003cp\u003ePEM (Performance Evaluation Matric) is a key analytical tool used to evaluate the predictive ability of the model on the bases of Overall Accuracy, Error Rate, Recall Rate Precision Rate, F1 Score. All these are calculated by confusion matrix.\u003c/p\u003e\n\u003ch2\u003e4.1 Overall Accuracy\u003c/h2\u003e\n\u003cp\u003eAccuracy refers to the proportion of the number of correctly classified samples to the total number of samples. It is commonly used methods to evaluate the generalization ability of a learning model. It can be measured by the equation no. 1\u003c/p\u003e\n\u003cp\u003eModel Accuracy= (\u003cem\u003en\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003ePositives\u003c/em\u003e) + \u003cem\u003em\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003eNegatives\u003c/em\u003e))\u003cem\u003e/N n\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003ePositives\u003c/em\u003e)\u003cem\u003eandm\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003eNegatives\u003c/em\u003e) are number of correctly classified samples. N is total number of samples.\u003c/p\u003e\n\u003cp\u003eTable 4: Accuracy table\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"185\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"70.27027027027027%\" colspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eAlgorithm Name\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.72972972972973%\" valign=\"top\"\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"16.756756756756758%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"53.513513513513516%\" valign=\"top\"\u003e\n \u003cp\u003eCNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.72972972972973%\" valign=\"top\"\u003e\n \u003cp\u003e0.946\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"16.756756756756758%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"53.513513513513516%\" valign=\"top\"\u003e\n \u003cp\u003eVGG19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.72972972972973%\" valign=\"top\"\u003e\n \u003cp\u003e0.943\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"16.756756756756758%\" valign=\"top\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"53.513513513513516%\" valign=\"top\"\u003e\n \u003cp\u003eYolov5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.72972972972973%\" valign=\"top\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch2\u003e4.2 \u0026nbsp;Error Rate\u003c/h2\u003e\n\u003cp\u003eError rate refers to the proportion of misclassified samples to the total number of samples. It can be calculated using the equation no. 2, Error Rate=\u003c/p\u003e\n\u003cp\u003e(\u003cem\u003en\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003eNegatives\u003c/em\u003e+)\u003cem\u003em\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eFalse\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003ePositives\u003c/em\u003e))\u003cem\u003e/N\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eEq no. 2. \u003cem\u003en\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003eNegatives\u003c/em\u003e)\u003cem\u003eandm\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eFalse\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003ePositives\u003c/em\u003e) are total number of wrongly classified samples. N is total number of samples.\u003c/p\u003e\n\u003cp\u003eTable 5: Error Rate\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"193\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"67.35751295336787%\" colspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eAlgorithm Name\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"32.64248704663213%\" valign=\"top\"\u003e\n \u003cp\u003eError Rate\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"27.979274611398964%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"39.37823834196891%\" valign=\"top\"\u003e\n \u003cp\u003eCNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"32.64248704663213%\" valign=\"top\"\u003e\n \u003cp\u003e0.057\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"27.979274611398964%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"39.37823834196891%\" valign=\"top\"\u003e\n \u003cp\u003eVGG19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"32.64248704663213%\" valign=\"top\"\u003e\n \u003cp\u003e0.054\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"27.979274611398964%\" valign=\"top\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"39.37823834196891%\" valign=\"top\"\u003e\n \u003cp\u003eYolov5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"32.64248704663213%\" valign=\"top\"\u003e\n \u003cp\u003e0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch2\u003e4.3 \u0026nbsp; \u0026nbsp; \u0026nbsp; Precision Rate\u003c/h2\u003e\n\u003cp\u003ePrecision Rate refers to the proportion of the true positive class among all the results predicted as the positive class. It can be defined as,\u003c/p\u003e\n\u003cp\u003ePrecision Rate=P= \u003cem\u003en\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u003c/em\u003e\u0026minus;\u003cem\u003ePositives\u003c/em\u003e)\u003cem\u003e/\u003c/em\u003e(\u003cem\u003en\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u003c/em\u003e\u0026minus;\u003cem\u003ePositives\u003c/em\u003e)+\u003cem\u003em\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eFalse\u003c/em\u003e\u0026minus; \u003cem\u003ePositives\u003c/em\u003e)) Eq no. 3.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003en\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u003c/em\u003e\u0026minus;\u003cem\u003ePositives\u003c/em\u003e) is total number of actual classified samples. \u003cem\u003em\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u003c/em\u003e\u0026minus; \u003cem\u003ePositives\u003c/em\u003e) is total number of wrong samples that are mistakenly classified to be correct.\u003c/p\u003e\n\u003cp\u003eTable 6: Precision Rate\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"184\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"70.65217391304348%\" colspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eAlgorithm Name\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.347826086956523%\" valign=\"top\"\u003e\n \u003cp\u003ePrecision\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"16.847826086956523%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"53.80434782608695%\" valign=\"top\"\u003e\n \u003cp\u003eCNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.347826086956523%\" valign=\"top\"\u003e\n \u003cp\u003e0.90\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"16.847826086956523%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"53.80434782608695%\" valign=\"top\"\u003e\n \u003cp\u003eVGG19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.347826086956523%\" valign=\"top\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"16.847826086956523%\" valign=\"top\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"53.80434782608695%\" valign=\"top\"\u003e\n \u003cp\u003eYolov5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.347826086956523%\" valign=\"top\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch2\u003e4.4 \u0026nbsp; \u0026nbsp; \u0026nbsp; Recall Rate\u003c/h2\u003e\n\u003cp\u003eRecall rate refers to the proportion of the number of correctly classified sample by the model to the number of all positive samples.\u003c/p\u003e\n\u003cp\u003eRecall Rate=R= \u003cem\u003en\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003ePositives\u003c/em\u003e)\u003cem\u003e/\u003c/em\u003e(\u003cem\u003en\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003ePositives\u003c/em\u003e) + \u003cem\u003em\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eFalse\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003eNegatives\u003c/em\u003e)) Eq no. 4.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003en\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eTrue\u0026nbsp;\u003c/em\u003e\u0026minus; \u003cem\u003ePositives\u003c/em\u003e) is number of correctly classified samples.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003em\u003c/em\u003e\u003csub\u003e(\u003c/sub\u003e\u003cem\u003eFalse -Negatives\u003c/em\u003e) is number of samples that are classified to be false.\u003c/p\u003e\n\u003cp\u003eTable 7: Recall Rate\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"196\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"65.98984771573605%\" colspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eAlgorithm Name\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.01015228426396%\" valign=\"top\"\u003e\n \u003cp\u003eRecall Rate\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"13.775510204081632%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"52.04081632653061%\" valign=\"top\"\u003e\n \u003cp\u003eCNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.183673469387756%\" valign=\"top\"\u003e\n \u003cp\u003e0.93\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"13.775510204081632%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"52.04081632653061%\" valign=\"top\"\u003e\n \u003cp\u003eVGG19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.183673469387756%\" valign=\"top\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"13.775510204081632%\" valign=\"top\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"52.04081632653061%\" valign=\"top\"\u003e\n \u003cp\u003eYolov5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.183673469387756%\" valign=\"top\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch2\u003e4.5 \u0026nbsp; \u0026nbsp; \u0026nbsp; F1 Measurement\u003c/h2\u003e\n\u003cp\u003eF1-Measurement is the evaluation standard which is the harmonic average of the precision and recall and it can be measured by the given equation,\u003c/p\u003e\n\u003cp\u003eF1 Measurement=R= 2. PR/(P+R )\u003c/p\u003e\n\u003cp\u003eEq no. 5.\u003c/p\u003e\n\u003cp\u003eP = Precision Rate and R = Recall Rate. Evaluate the model\u0026rsquo;s predictive\u003c/p\u003e\n\u003cp\u003eTable 8: F1 Measurement\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"222\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"58.55855855855856%\" colspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eAlgorithm Name\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"41.44144144144144%\" valign=\"top\"\u003e\n \u003cp\u003eF1 Measurement\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"13.963963963963964%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"44.5945945945946%\" valign=\"top\"\u003e\n \u003cp\u003eCNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"41.44144144144144%\" valign=\"top\"\u003e\n \u003cp\u003e0.916\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"13.963963963963964%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"44.5945945945946%\" valign=\"top\"\u003e\n \u003cp\u003eVGG19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"41.44144144144144%\" valign=\"top\"\u003e\n \u003cp\u003e0.913\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"13.963963963963964%\" valign=\"top\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"44.5945945945946%\" valign=\"top\"\u003e\n \u003cp\u003eYolov5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"41.44144144144144%\" valign=\"top\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eability on the test set according to the previously determined evaluation criteria and reveal that the Yolo v5 model is found to be best as compared to all other models as shown in Figure 9\u003c/p\u003e"},{"header":"5 Conclusion","content":"\u003cp\u003eThe Covid-19 Pandemic has significantly altered our routines. While waiting for the virus\u0026rsquo;s spread to decreased, it is essential to follow a set of guidelines, such as wearing face masks in public places. The World Health Organization (WHO) provides the standard recommendations on preventing the spread of COVID-19 and the importance of face masks for protection from the virus [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. The goal of the following work is to develop a solution for automatically identifying infractions committed by people who are not wearing face masks. In this study, we conduct a comparative study for face mask detection We employed three deep learning models: Yolov5, CNN, and VGG-19. We train these models on face mask detection dataset. Yolov5 has a 98 Percent accuracy rate, CNN achieved 94.6 accuracy and VGG-19 accomplished the 94.3 accuracy. YoloV5 outperforms other competitor algorithms in terms of accuracy. This algorithm works well with various performance parameters. Our proposed methodology facilitates in the detection of face masks in public places and the prevention of Corona virus disease.\u003c/p\u003e"},{"header":"6 Future Work","content":"\u003cp\u003eWe have the ability to create a mobile application. In which we keep track of which employees are frequently breaking the regulations and wearing face masks or not in the university/office. If someone breaks the rules, a message will be sent to their mobile phone.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eU Rajendra Acharya, Shu Lih Oh, Yuki Hagiwara, Jen Hong Tan, and Hojjat Adeli. Deep convolutional neural network for the automated detection and diagnosis of seizure using eeg signals. \u003cem\u003eComputers in biology and medicine\u003c/em\u003e, 100:270\u0026ndash;278, 2018.\u003c/li\u003e\n \u003cli\u003eNadeem Akhtar and U Ragavendran. Interpretation of intelligence in cnn-pooling processes: a methodological survey. \u003cem\u003eNeural computing and applications\u003c/em\u003e, 32(3):879\u0026ndash;898, 2020.\u003c/li\u003e\n \u003cli\u003eMohammed A Al-Masni, Mugahed A Al-Antari, Mun-Taek Choi, SeungMoo Han, and Tae-Seong Kim. Skin lesion segmentation in dermoscopy images via deep full resolution convolutional networks. \u003cem\u003eComputer methods and programs in biomedicine\u003c/em\u003e, 162:221\u0026ndash;231, 2018.\u003c/li\u003e\n \u003cli\u003eShaojie Bai. \u003cem\u003eThe Effect of Pre-ReLU Input Distribution on DNN\u0026rsquo;s Performance\u003c/em\u003e. PhD thesis, Carnegie Mellon University, 2017.\u003c/li\u003e\n \u003cli\u003eSatilmis Bilgin, Ozge Kurtkulagi, Gizem Bakir Kahveci, Tuba Taslamacioglu Duman, and Burcin Meryem Atak Tel. Millennium pandemic: a review of coronavirus disease (covid-19). \u003cem\u003eExperimental Biomedical Research\u003c/em\u003e, 3(2):117\u0026ndash;125, 2020.\u003c/li\u003e\n \u003cli\u003eWei Bu, Jiangjian Xiao, Chuanhong Zhou, Minmin Yang, and Chengbin Peng. A cascade framework for masked face detection. In \u003cem\u003e2017 IEEE international conference on cybernetics and intelligent systems (CIS) and IEEE conference on robotics, automation and mechatronics (RAM)\u003c/em\u003e, pages 458\u0026ndash;462. IEEE, 2017.\u003c/li\u003e\n \u003cli\u003eJane Chiodini. Maps, masks and media\u0026ndash;traveller and practitioner resources for 2019 novel coronavirus (2019-ncov) acute respiratory virus. \u003cem\u003eTravel medicine and infectious disease\u003c/em\u003e, 33:101574, 2020.\u003c/li\u003e\n \u003cli\u003eGayatri Deore, Ramakrishna Bodhula, Vishwas Udpikar, and Vidya More. Study of masked face detection approach in video analytics. In \u003cem\u003e2016 Conference on Advances in Signal Processing (CASP)\u003c/em\u003e, pages 196\u0026ndash;200. IEEE, 2016.\u003c/li\u003e\n \u003cli\u003eAnwesha Dutta and Harry W Fischer. The local governance of covid-19: Disease prevention and social security in rural india. \u003cem\u003eWorld Development\u003c/em\u003e, 138:105234, 2021.\u003c/li\u003e\n \u003cli\u003eMd Sabbir Ejaz, Md Rabiul Islam, Md Sifatullah, and Ananya Sarker. Implementation of principal component analysis on masked and nonmasked face recognition. In \u003cem\u003e2019 1st international conference on advances in science, engineering and robotics technology (ICASERT)\u003c/em\u003e, pages 1\u0026ndash;5. IEEE, 2019.\u003c/li\u003e\n \u003cli\u003eAli Mamdouh Elkahky, Yang Song, and Xiaodong He. A multi-view deep learning approach for cross domain user modeling in recommendation systems. In \u003cem\u003eProceedings of the 24th international conference on world wide web\u003c/em\u003e, pages 278\u0026ndash;288, 2015.\u003c/li\u003e\n \u003cli\u003eShuo Feng, Chen Shen, Nan Xia, Wei Song, Mengzhen Fan, and Benjamin J Cowling. Rational use of face masks in the covid-19 pandemic.\u0026nbsp;\u003cem\u003eThe Lancet Respiratory Medicine\u003c/em\u003e, 8(5):434\u0026ndash;436, 2020.\u003c/li\u003e\n \u003cli\u003eLeon Gatys, Alexander S Ecker, and Matthias Bethge. Texture synthesis using convolutional neural networks. \u003cem\u003eAdvances in neural information processing systems\u003c/em\u003e, 28, 2015.\u003c/li\u003e\n \u003cli\u003eA Victor Ikechukwu, S Murali, R Deepu, and RC Shivamurthy. Resnet50 vs vgg-19 vs training from scratch: A comparative analysis of the segmentation and classification of pneumonia from chest x-ray images. \u003cem\u003eGlobal Transitions Proceedings\u003c/em\u003e, 2(2):375\u0026ndash;381, 2021.\u003c/li\u003e\n \u003cli\u003eTalha Ikram, Abdullah Saeed, NoorUl Ayn, Muhammad Ali Tahir, and Rafia Mumtaz. A review of the prevalent ict techniques used for covid19 sop violation detection. In \u003cem\u003e2020 IEEE 17th International Conference on Smart Communities: Improving Quality of Life Using ICT, IoT and AI (HONET)\u003c/em\u003e, pages 194\u0026ndash;198. IEEE, 2020.\u003c/li\u003e\n \u003cli\u003eNancy HL Leung, Daniel KW Chu, Eunice YC Shiu, Kwok-Hung Chan, James J McDevitt, Benien JP Hau, Hui-Ling Yen, Yuguo Li, Dennis KM Ip, JS Peiris, et al. Respiratory virus shedding in exhaled breath and efficacy of face masks. \u003cem\u003eNature medicine\u003c/em\u003e, 26(5):676\u0026ndash;680, 2020.\u003c/li\u003e\n \u003cli\u003eMuhammad Mateen, Junhao Wen, Nasrullah, Sun Song, and Zhouping Huang. Fundus image classification using vgg-19 architecture with pca and svd. \u003cem\u003eSymmetry\u003c/em\u003e, 11(1):1, 2018.\u003c/li\u003e\n \u003cli\u003eNoah C Peeri, Nistha Shrestha, Md Siddikur Rahman, Rafdzah Zaki, Zhengqi Tan, Saana Bibi, Mahdi Baghbanzadeh, Nasrin Aghamohammadi, Wenyi Zhang, and Ubydul Haque. The sars, mers and novel coronavirus (covid-19) epidemics, the newest and biggest global health threats: what lessons have we learned? \u003cem\u003eInternational journal of epidemiology\u003c/em\u003e, 49(3):717\u0026ndash;726, 2020.\u003c/li\u003e\n \u003cli\u003eJitendra Singh Rathore and Chaitali Ghosh. Severe acute respiratory syndrome coronavirus-2 (sars-cov-2), a newly emerged pathogen: an overview. \u003cem\u003ePathogens and disease\u003c/em\u003e, 78(6):ftaa042, 2020.\u003c/li\u003e\n \u003cli\u003eConnor Shorten and Taghi M Khoshgoftaar. A survey on image data augmentation for deep learning. \u003cem\u003eJournal of big data\u003c/em\u003e, 6(1):1\u0026ndash;48, 2019.\u003c/li\u003e\n \u003cli\u003eStanislaw P Stawicki, Rebecca Jeanmonod, Andrew C Miller, Lorenzo Paladino, David F Gaieski, Anna Q Yaffee, Annelies De Wulf, Joydeep Grover, Thomas J Papadimos, Christina Bloem, et al. The 2019\u0026ndash;2020 novel coronavirus (severe acute respiratory syndrome coronavirus 2) pandemic: A joint american college of academic international medicineworld academic council of emergency medicine multidisciplinary covid19 working group consensus paper. \u003cem\u003eJournal of global infectious diseases\u003c/em\u003e, 12(2):47\u0026ndash;93, 2020.\u003c/li\u003e\n \u003cli\u003eYanan Sun, Bing Xue, Mengjie Zhang, Gary G Yen, and Jiancheng Lv. Automatically designing cnn architectures using the genetic algorithm for image classification. \u003cem\u003eIEEE transactions on cybernetics\u003c/em\u003e, 50(9):3840\u0026ndash; 3854, 2020.\u003c/li\u003e\n \u003cli\u003eYu Wang, Huaiyu Tian, Li Zhang, Man Zhang, Dandan Guo, Wenting Wu, Xingxing Zhang, Ge Lin Kan, Lei Jia, Da Huo, et al. Reduction of secondary transmission of sars-cov-2 in households by face mask use, disinfection and social distancing: a cohort study in beijing, china. \u003cem\u003eBMJ global health\u003c/em\u003e, 5(5):e002794, 2020.\u003c/li\u003e\n \u003cli\u003eJian Xiao, Jia Wang, Shaozhong Cao, and Bilong Li. Application of a novel and improved vgg-19 network in the detection of workers wearing masks. In \u003cem\u003eJournal of Physics: Conference Series\u003c/em\u003e, volume 1518, page 012041. IOP Publishing, 2020.\u003c/li\u003e\n \u003cli\u003eMetin Yıldız, Yakup Sarpdag˘ı, Mehmet Okuyar, Mehmet Yildiz, Necmettin C\u0026cedil;iftci, Ay\u0026cedil;se Elkoca, Mehmet Salih Yildirim, Muhammet Ali Aydin, Mehmet Parlak, and Bu\u0026uml;nyamin Bingo\u0026uml;l. Segmentation and classification of skin burn images with artificial intelligence: Development of a mobile application. \u003cem\u003eBurns\u003c/em\u003e, 2024.\u003c/li\u003e\n \u003cli\u003eXinkai Zhou, Zhigui Wu, Ranran Yu, Shanni Cao, Wen Fang, Zhen Jiang, Fang Yuan, Chao Yan, and Dijun Chen. Modelling-based evaluation of the effect of quarantine control by the chinese government in the coronavirus disease 2019 outbreak. \u003cem\u003eMedRxiv\u003c/em\u003e, pages 2020\u0026ndash;03, 2020.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Air University","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Face Mask Detection, Deep Learning,","lastPublishedDoi":"10.21203/rs.3.rs-4637920/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4637920/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe disease was first discovered in Wuhan City, Hubei Province, the People’s Republic of China in late 2019, and rapidly grow to many countries around the world in early 2020, steadily turning into a global extensive pandemic. More than 222 million confirmed cases have been reported in different countries and regions around the world, and more than 4.6 million have died, which is one of the large-scale epidemics in human history . The coronavirus spreads through small droplets during the discussion, coughing, sneezing, etc. In poorly and closed ventilated locations a higher risk of transmission rate However, wearing a face mask that prevents the transmission of droplets in the air. But the continuous inspection of preventive measures both inside and outside the building/offices to prevent the growth of COVID-19 is a major challenging task. Therefore, in this research work, we focused on implementing a Face Mask Detection model that is relying on the related technologies of machine vision, we adopted three different well-known and the most advanced end-to-end target detection algorithm named CNN, VGG16, and -YOLOv5 to realize the detection and recognition of whether the face is wearing a mask. In terms of data set collection, we use the face mask opensource data set. After the actual effect test, we found the accuracy, error rate, recall rate, precision rate, and F1 of the Yolov5 algorithm model have reached a high level. This solution tracks the people with or without masks in a real-time scenario and highlighted the person with a red rectangle box in the case of violation. With the help of this 24/7, either inside or outside the organization continuously monitoring is possible and it has a great impact to identify the violator and ensure the safety of every individual.\u003c/p\u003e","manuscriptTitle":"Real-Time face mask Surveillance System for the Pandemic of Covid-19","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-06-26 14:47:22","doi":"10.21203/rs.3.rs-4637920/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"4426e416-0bec-4f9e-bf5b-6850cf0c3f86","owner":[],"postedDate":"June 26th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":33727949,"name":"Numerical Analysis"},{"id":33727950,"name":"Software Engineering"}],"tags":[],"updatedAt":"2024-06-26T14:47:22+00:00","versionOfRecord":[],"versionCreatedAt":"2024-06-26 14:47:22","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4637920","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4637920","identity":"rs-4637920","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}