Mask Detection and Social Distance Identification Using Internet of Things and Faster R-CNN Algorithm

Abstract

The drones can be used to detect a group of people who are unmasked and do not maintain social distance. In this paper, a deep learning-enabled drone is designed for mask detection and social distance monitoring. A drone is one of the unmanned systems that can be automated. This system mainly focuses on Industrial Internet of Things (IIoT) monitoring using Raspberry Pi 4. This drone automation system sends alerts to the people via speaker for maintaining the social distance. This system captures images and detects unmasked persons using faster regions with convolutional neural network (faster R-CNN) model. When the system detects unmasked persons, it sends their details to respective authorities and the nearest police station. The built model covers the majority of face detection using different benchmark datasets. OpenCV camera utilizes 24/7 service reports on a daily basis using Raspberry Pi 4 and a faster R-CNN algorithm.

This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction

The coronavirus disease 2019 (COVID-19) epidemic has arisen as a major menace all around the world. As the number of cases is gradually increasing day by day, the government has several difficulties in controlling the pandemic situation. The communication of this disease can only be lessened with the proper collaboration of people. Physical distancing, repeated hand sanitizing, and face masking have proven to be quite efficient to control the spreading of the virus, but everyone is not obeying the guidelines. Various technologies like machine learning (ML) algorithms [1], artificial intelligence (AI) approaches [1], Internet of things (IoT) [2–5], and unmanned aerial vehicles (UAV) [6] give a real-time scenario at any given point about (i) the number of people following physical distancing and (ii) whether people are wearing masks or not.

In today’s scenario, COVID-19 [3, 5] has come to be one of the most important topics that ought to be confronted properly. Therefore, there is a requirement to develop an approach that can tackle the issues like unmasked persons and nonmaintenance of physical distancing. With the major developments in the area of industrial IoT (IIoT), remote monitoring can be done with much more ease. The key concept at the back of IoT is to interconnect the different automated components throughout a network. In order to allow the data communication between them, each object is assigned with a different identity and the approach is to design a facial mask detection system based on IIoT.

As of January 20, 2021, the deadly COVID-19 illness, which has affected more than 200 countries and territories, including two international means of transport, has so far caused 96.1 million infections and 2.06 million fatalities globally. Even while there was a shortage of active pharmaceutical experts, there was also a lack of public opposition to COVID-19, making the populace even more vulnerable. The World Health Organization has labeled it a pandemic [7]. Because this is an epidemic with no cure, the only strategy left is to wear a mask. The fact that a face mask may prevent the spread of COVID-19 has led to an increase in its use among the general population. In order to stop the virus from spreading further, the global society must consider quarantine, as well as increasing the social barrier between infected and uninfected individuals. This face mask test is used to make sure that the individual is protected against infection with the airborne virus. Whenever someone coughs, talks, or sneezes, viruses will fly into the air, increasing the risk of spreading disease to their surroundings. To reduce the transmission of disease, infection control specialists use a number of measures, including surgical masks, to ward against contamination.

The main ways in which this paper makes its mark are as follows. A face mask recognition method is used based on quicker R-CNN and YOLO models [8]. To get a full understanding of the main issues involved in face mask detection, something may help in the future when it comes to developing the new face mask detectors. This transfer learning model includes the new information for developing algorithms. It is a machine learning strategy in which the computer learns the skills from a single task and then may apply these skills to other situations. The use of pretrained models as the base point in AI activities, especially those with computational complexity and time concerns, is an emerging trend. The authors used two algorithms, YOLOv3 and accelerated R-CNN, for forecasting, and their results have been compared. Since the R-CNN [8] uses two networks, the first of which is the region proposal network (RPN) [8] which makes recommendations to identify objects, it is more efficient. Each area box is assigned a value by RPN. YOLO is a sophisticated, real-time CNN that identifies objects. The procedure slices the image into regions and guesses where to train these models but it is difficult because of the camera angles and mask types in pictures. Another difficulty that the authors found was the absence of a large dataset with both masked and unmasked categories. The writers had to generate a fresh dataset and use transfer learning to finish the research because of this issue. The social distancing and mask detection are performed on drone footage using artificial intelligence and the faster R-CNN algorithm. The captured images are analyzed using datasets and trained models.

The proposed system identifies only the stored faces in the database. Upon detection of an unmasked face, it sends an e-mail containing the image of the person. In order to design the facemask recognition system, a setup containing Raspberry Pi 4 and an OpenCV camera is used. Raspberry Pi uses Raspbian Stretch as an operating system. In order to train the database of images, a huge number of images at different angles and lighting environments would be collected. A faster R-CNN algorithm is used here to detect facial mask recognition. The current research aims to provide the information to the user using open source technology that comprises faster R-CNN, IIoT, Raspberry Pi 4, OpenCV camera, and UAV. Most of the existing models are not tested in real time. However, the proposed model is verified in real time by embedding the built model on a drone. The faster R-CNN model is used for detecting faces and people’s activities. Live social distancing is measured using YOLOv3.

The paper is organized as follows: Section 2 discusses the survey of the existing work related to the proposed work. Section 3 discusses the proposed system model and the proposed methodology. Section 4 discusses the results and analysis section in terms of various parameters. At last, the conclusion is drawn.

2. Related Work

People are finding many new methods to secure themselves from the COVID-19 pandemic. The researchers working in the fields such as Internet of Things, machine learning, computer vision, and Blockchain are working on the various techniques that can secure and treat the people against the spreading of this virus. In this paper, research is made using the Industrial Internet of Things, deep learning, and a faster R-CNN algorithm to detect masked people and social distancing. The related existing research is discussed in this section.

In [9], an innovative intelligent technique based on a deep convolutional model is used to protect the people from COVID-19. The proposed system can detect automatically whether people are following the safety guidelines or not. In [8], a detailed comparison is done between the various deep learning approaches to monitor the disease from medical imaging. In another study [10], an IoT system based on temperature sensing, mask detection, and social distancing is suggested for the protection against COVID-19. Arduino Uno is used for the infrared sensing of temperature while Raspberry Pi is used for the mask detection and social distancing using computer vision techniques. In [11], an innovative mask identification technique is proposed by preprocessing of the image followed by face detection and image superresolution. The system is found to be very much accurate as far as identification of mask wearing is concerned. In [7], a model based on deep neural network is proposed for monitoring people and social distancing even in poor light conditions. The technique is found to be better than many other past techniques in terms of speed and accuracy. In [12], a computer vision based deep learning approach is used to examine the mask detection and social distancing automatically in real time. The model is developed on Raspberry Pi to monitor the different activities. It is found to save time and reduce the spread of COVID-19. In [13], an IoT-based deep learning foundation is provided for the detection of COVID-19. The provided model is used for the detection of a pandemic by applying this model on the X-ray of the chest. It is proved to be very accurate and hence quite helpful for the medical experts for the prompt detection of COVID-19. In [14], a technique based on open source computer vision is proposed to detect masked persons. The technique is found to be efficient especially in industrial applications.

In [15], the contribution of IoT and the concerned sensors for tracking and mitigating the virus is discussed. The study provides deep insight into e-health services based upon sensors for managing COVID-19 and discusses the subsequent IoT networks for the postpandemic era. In [16], a review of the technologies that may be used to detect COVID-19 is discussed. The review also discusses the future challenges in implementing these technologies. The technologies discussed are deep learning associated with X-ray, in vitro diagnostics (IVDs), and wearable sensors based on IoT for monitoring the COVID-19 patients. In [17], detection of the face wearing mask in the fine state has been done with context attention R-CNN technique with the help of special features for the purpose of region proposal and by dissociation of localization and classification fields. The context approach R-CNN has been found to be highly accurate. The authors in [18] have proposed a cellphone system-based detection system to extract four different kinds of features using the K-Nearest Neighbour algorithm. The system has been found to be excellent in terms of accuracy, precision, and recall. A system based on real time has been analyzed in [19] that uses the concept of generic detection. The accuracy of this system has been found to be high that may further be improved by including more parameters despite having a large computation time. Further in [20], a high-performance face mask detector based on deep learning has been proposed that is computationally less complex. The feature extraction has been improved by residual and regression modules.

Recently, many deep learning-based models have been developed for mask detection. The convolutional model is based on automatic recognition while the deep learning-based face mask detection technique is used for the high speed and accurate detection and feature extraction in a real-time environment even in bad light conditions. IoT-based models can be used for high accuracy detection while some systems based on superresolution image processing give extremely high accuracy. To overcome the gaps, the proposed system is based on IoT, deep learning, convolution, and automatic face recognition and computer vision to combine the advantages of all the techniques. Further, the proposed system uses the faster R-CNN technique to mitigate the effects of COVID-19.

3. The Proposed System Model

In this system, a model is suggested that uses the combination of OpenCV library with Raspberry Pi to build an Industrial Internet of Things (IIoT) application for mask detection and UAV application for social distance monitoring. The proposed system can identify or verify a person from a video frame. To see the masked face in a frame, first, we need to identify whether the facemask is present or not. If it is present, then it is marked as the region of interest (ROI) followed by its removal and processing for facial mask detection. The faster R-CNN algorithm for facemask detection works very well if the database contains clear images of persons. The employment of the OpenCV library tool proves to be very effective for mask detection and recognition.

3.1. Components of the Proposed Model

The various components of the proposed model are shown in Figure 1. A brief description of all the components is mentioned as follows.

[figure omitted; refer to PDF]

The pseudocode for detection of bounding box using OpenCV camera is shown in Figure 4.

[figure omitted; refer to PDF]

Figure 10 shows the overview of the drone system. The speaker alerting system is programmed with a condition referred to as “when the distances between the objects are less than 1 meter” in the IoT controller. After satisfying the condition, the recorded voice sounds are activated to be disposed of the crowd. This system is used only to alert the people to maintain social distancing and inform them to wear masks. The outcome result of this method is based on the recorded voice output and monitoring output through IoT clouding.

[figure omitted; refer to PDF]

The drone mechanism includes an IoT controller and a high-fidelity camera. Automatically, the speaker announces whenever the condition is solved.

5. Conclusion

The drones can be used to detect a group of people who are unmasked and do not maintain social distance. In this paper, a drone was used for detecting a number of objects using a faster R-CNN algorithm and YOLOv3. Initially, faster R-CNN was used for mask detection. Raspberry Pi 4 interfaced OpenCV camera in a real test case was implemented to capture the images. The built face classification was stored on the cloud; thus the built drone always remains connected with the cloud server using the Internet. YOLOv3 model was used for measuring the social distance. Faster R-CNN and YOLOv3 were utilized since they can easily calculate the captured images with better performance in a few milliseconds. Extensive results were drawn for social distancing, unmask/mask faces, and count of persons with real-time video streaming. PyCharm software was utilized using Python for reducing the cost of the system. Extensive comparative analyses revealed that the proposed model outperforms the competitive models in terms of various performance metrics.

In this paper, no novel deep learning model was proposed. Therefore, in near future, we will design novel deep learning models to achieve better results. Also, the proposed model can be extended for indoor application areas too.

Ethical Approval

This article does not contain any studies with human participants performed by any of the authors.

Acknowledgments

The authors extend their appreciation to the Researchers Supporting Project (number RSP-2021/314) King Saud University, Riyadh, Saudi Arabia.

References

[1] F. Rustam, A. A. Reshi, A. Mehmood, S. Ullah, B. On, W. Aslam, G. S. Choi, "COVID-19 Future Forecasting Using Supervised Machine Learning Models," IEEE Access, vol. 8,DOI: 10.1109/ACCESS.2020.2997311, 2020.

[2] N. Sindhwani, V. P. Maurya, A. Patel, R. K. Yadav, S. Krishna, R. Anand, "Implementation of intelligent plantation system using virtual IoT," Internet of Things and its Applications, pp. 305-322, DOI: 10.1007/978-3-030-77528-5_16, 2022.

[3] D. Singh, V. Kumar, M. Kaur, M. Y. Jabarulla, H.-N. Lee, "Screening of COVID-19 suspected subjects using multi-crossover genetic algorithm based dense convolutional neural network," IEEE Access, vol. 9,DOI: 10.1109/ACCESS.2021.3120717, 2021.

[4] R. Anand, A. Sinha, A. Bhardwaj, A. Sreeraj, "Flawed security of social network of things," Handbook of Research on Network Forensics and Analysis Techniques, pp. 65-86, DOI: 10.4018/978-1-5225-4100-4.ch005, 2018.

[5] R. Anand, N. Sindhwani, A. Saini, "Emerging Technologies for COVID‐19," Enabling Healthcare 4.0 for Pandemics: A Roadmap Using AI, Machine Learning, IoT and Cognitive Technologies, pp. 163-188, DOI: 10.1002/9781119769088.ch9, 2021.

[6] L. Kohli, M. Saurabh, I. Bhatia, N. Sindhwani, M. Vijh, "Design and Development of Modular and Multifunctional UAV with Amphibious Landing, Processing and Surround Sense Module," Unmanned Aerial Vehicles For Internet Of Things (IoT): Concepts, Techniques, and Applications,DOI: 10.1002/9781119769170.ch12, 2021.

[7] M. I. Uddin, S. A. A. Shah, M. A. Al-Khasawneh, "A novel deep convolutional neural network model to monitor people following guidelines to avoid COVID-19," Journal of Sensors, vol. 2020,DOI: 10.1155/2020/8856801, 2020.

[8] X. Liu, L. Faes, A. U. Kale, S. K. Wagner, D. J. Fu, A. Bruynseels, T. Mahendiran, G. Moraes, M. Shamdas, C. Kern, J. R. Ledsam, M. K. Schmid, K. Balaskas, E. J. Topol, L. M. Bachmann, P. A. Keane, A. K. Denniston, "A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis," The lancet digital health, vol. 1 no. 6, pp. e271-e297, DOI: 10.1016/s2589-7500(19)30123-2, 2019.

[9] N. Petrović, Đ. Kocić, "IoT-based system for COVID-19 indoor safety monitoring," Preprint), IcETRAN, 2020. In press

[10] B. Qin, D. Li, "Identifying facemask-wearing condition using image super-resolution with classification network to prevent COVID-19," Sensors, vol. 20 no. 18,DOI: 10.3390/s20185236, 2020.

[11] M. Rezaei, M. Azarmi, "DeepSOCIAL: social distancing monitoring and infection risk assessment in COVID-19 pandemic," Applied Sciences, vol. 10 no. 21,DOI: 10.3390/app10217514, 2020.

[12] S. Yadav, "Deep learning based safe social distancing and face mask detection in public areas for COVID-19 safety guidelines adherence," International Journal for Research in Applied Science and Engineering Technology, vol. 8 no. 7, pp. 1368-1375, DOI: 10.22214/ijraset.2020.30560, 2020.

[13] I. Ahmed, A. Ahmad, G. Jeon, "An IoT based deep learning framework for early assessment of Covid-19," IEEE Internet of Things Journal, vol. 8 no. 1,DOI: 10.1109/JIOT.2020.3034074, 2020.

[14] T. V. N. Rao, G. V. Sai, M. Chenna, M. S. Reddy, "An application for the detection of COVID-19 face mask using OpenCV," High Technology Letters, vol. 26 no. 7, pp. 965-975, 2020.

[15] M. Ndiaye, S. S. Oyewobi, A. M. Abu-Mahfouz, G. P. Hancke, A. M. Kurien, K. Djouani, "IoT in the wake of COVID-19: a survey on contributions, challenges and evolution," IEEE Access, vol. 8,DOI: 10.1109/access.2020.3030090, 2020.

[16] S. S. Kumar, "Emerging technologies and sensors that can Be used during the COVID-19 pandemic," Proceedings of the 2020 International Conference on UK-China Emerging Technologies (UCET),DOI: 10.1109/UCET51115.2020.9205424, .

[17] J. Zhang, F. Han, Y. Chun, W. Chen, "A novel detection framework about conditions of wearing face mask for helping control the spread of COVID-19," IEEE Access, vol. 9,DOI: 10.1109/access.2021.3066538, 2021.

[18] Y. Chen, M. Hu, C. Hua, G. Zhai, J. Zhang, Q. Li, S. X. Yang, "Face mask assistant: detection of face mask service stage based on mobile phone," IEEE Sensors Journal, vol. 21 no. 9,DOI: 10.1109/jsen.2021.3061178, 2021.

[19] X. Fan, M. Jiang, H. Yan, "A deep learning based light-weight face mask detector with residual context attention and Gaussian heatmap to fight against COVID-19," Ieee Access, vol. 9,DOI: 10.1109/access.2021.3095191, 2021.

[20] D. G. Dondo, J. A. Redolfi, R. G. Araguás, D. Garcia, "Application of deep-learning methods to real time face mask detection," IEEE Latin America Transactions, vol. 19 no. 6, pp. 994-1001, DOI: 10.1109/tla.2021.9451245, 2021.

[21] G. Hu, S.-H. Kay Chen, N. Mazur, "Deep neural network-based speaker-aware information logging for augmentative and alternative communication," Journal of Artificial Intelligence and Technology, vol. 1 no. 2, pp. 138-143, DOI: 10.37965/jait.2021.0017, 2021.

[22] B. Gupta, M. Tiwari, S. Lamba, "Visibility improvement and mass segmentation of mammogram images using quantile separated histogram equalisation with local contrast enhancement," CAAI Transactions on Intelligence Technology, vol. 4 no. 2, pp. 73-79, DOI: 10.1049/trit.2018.1006, 2019.

[23] S. Han, J. Pool, J. Tran, W. Dally, "Learning both weights and connections for efficient neural network," Advances in Neural Information Processing Systems, pp. 1135-1143, DOI: 10.5555/2969239.2969366, 2015.

[24] H. Singh, T. B. Rehman, C. Gangadhar, R. Anand, N. Sindhwani, M. Babu, "Accuracy detection of coronary artery disease using machine learning algorithms," Applied Nanoscience,DOI: 10.1007/s13204-021-02036-7, 2021.

[25] R. Kamalraj, S. Neelakandan, M. R. Kumar, V. C. S. Rao, R. Anand, H. Singh, "Interpretable filter based convolutional neural network (IF-CNN) for glucose prediction and classification using PD-SS algorithm," Measurement, vol. 183,DOI: 10.1016/j.measurement.2021.109804, 2021.

[26] D. Singh, V. Kumar, M. Kaur, "Densely connected convolutional networks-based COVID-19 screening model," Applied Intelligence, vol. 51 no. 5, pp. 3044-3051, DOI: 10.1007/s10489-020-02149-6, 2021.

[27] J. F. Henriques, R. Caseiro, P. Martins, J. Batista, "High-speed tracking with kernelized correlation filters," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37 no. 3, pp. 583-596, DOI: 10.1109/TPAMI.2014.2345390, 2014.

[28] F. N. Iandola, S. Han, M. W. Moskewicz, K. Ashraf, W. J. Dally, K. Keutzer, "SqueezeNet: AlexNet-Level Accuracy with 50x Fewer Parameters And," 2016. https://arxiv.org/abs/1602.07360

[29] M. Kaur, V. Kumar, V. Yadav, D. Singh, N. Kumar, N. N. Das, "Metaheuristic-based deep COVID-19 screening model from chest X-ray images," Journal of Healthcare Engineering, vol. 2021,DOI: 10.1155/2021/8829829, 2021.

[30] N. Sindhwani, M. S. Bhamrah, "An optimal scheduling and routing under adaptive spectrum-matching framework for MIMO systems," International Journal of Electronics, vol. 104 no. 7, pp. 1238-1253, DOI: 10.1080/00207217.2017.1293173, 2017.

[31] A. Srivastava, A. Gupta, R. Anand, "Optimized smart system for transportation using RFID technology," Mathematics in Engineering, Science and Aerospace, vol. 12 no. 4, pp. 953-965, 2021.

[32] N. Sindhwani, M. Singh, "Performance analysis of ant colony based optimization algorithm in MIMO systems," Proceedings of the 2017 International Conference on Wireless Communications, Signal Processing and Networking (WiSPNET), pp. 1587-1593, DOI: 10.1109/WiSPNET.2017.8300029, .

[33] R. Anand, P. Chawla, "A review on the optimization techniques for bio-inspired antenna design," 2016 Third International Conference on Computing for Sustainable Global Development (INDIACom), pp. 2228-2233, .

[34] M. Kristan, J. Matas, A. Leonardis, M. Felsberg, L. Cehovin, G. Fernandez, R. Pflugfelder, "The visual object tracking vot2015 challenge results," Proceedings of the IEEE International Conference on Computer Vision Workshops,DOI: 10.1109/ICCVW.2015.79, .

[35] H. Nam, B. Han, "Learning multi-domain convolutional neural networks for visual tracking," Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4293-4302, DOI: 10.1109/cvpr.2016.465, .

[36] J. Redmon, S. Divvala, R. Girshick, A. Farhadi, "You only look once: unified, real-time object detection," Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 779-788, DOI: 10.1109/cvpr.2016.91, .

[37] J. Redmon, S. K. Divvala, R. B. Girshick, A. Farhadi, "You only look once: unified, real-time object detection," 2015.

[38] S. Ren, K. He, R. Girshick, J. Sun, "Faster r-cnn: towards real-time object detection with region proposal networks," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39 no. 6, pp. 1137-1149, DOI: 10.1109/TPAMI.2016.2577031, 2017.

[39] K. Simonyan, A. Zisserman, "Deep convolutional networks for large-scale image recognition," 2014. https://arxiv.org/abs/1409.1556

[40] H. S. Basavegowda, G. Dagnew, "Deep learning approach for microarray cancer data classification," CAAI Transactions on Intelligence Technology, vol. 5 no. 1, pp. 22-33, DOI: 10.1049/trit.2019.0028, 2020.

[41] Y. Xu, T. T. Qiu, "Human activity recognition and embedded application based on convolutional neural network," Journal of Artificial Intelligence and Technology, vol. 1 no. 1, pp. 51-60, DOI: 10.37965/jait.2020.0051, 2021.

[42] A. Nag, J. N. Nikhilendra, M. Kalmath, "IOT based door access control using face recognition," Proceedings of the 2018 Third International Conference for Convergence in Technology (I2CT),DOI: 10.1109/I2CT.2018.8529749, .

[43] S. Liao, A. K. Jain, S. Z. Li, "A fast and accurate unconstrained face detector," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 38 no. 2, pp. 211-223, DOI: 10.1109/TPAMI.2015.2448075, 2015.

[44] M. Najibi, P. Samangouei, R. Chellappa, L. S. Davis, "Ssh: single stage headless face detector," Proceedings of the IEEE International Conference on Computer Vision, pp. 4875-4884, DOI: 10.1109/iccv.2017.522, .

[45] S. Ghosh, P. Shivakumara, P. Roy, U. Pal, T. Lu, "Graphology based handwritten character analysis for human behaviour identification," CAAI Transactions on Intelligence Technology, vol. 5 no. 1, pp. 55-65, DOI: 10.1049/trit.2019.0051, 2020.

[46] N. Takahashi, M. Gygli, B. Pfister, L. Van Gool, "Deep convolutional neural networks and data augmentation for acoustic event detection," 2016. https://arxiv.org/abs/1604.07160

[47] X. Xu, I. A. Kakadiaris, "Joint head pose estimation and face alignment framework using global and local CNN features," In Proceedings of the 2017 Tweleveth IEEE International Conference on Automatic Face & Gesture Recognition (FG 2017), pp. 642-649, DOI: 10.1109/FG.2017.81, .

[48] R. Ranjan, V. M. Patel, R. Chellappa, "Hyperface: a deep multi-task learning framework for face detection, landmark localization, pose estimation, and gender recognition," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 41 no. 1, pp. 121-135, DOI: 10.1109/TPAMI.2017.2781233, 2017.

[49] R. Anand, B. Singh, N. Sindhwani, "Speech perception & analysis of fluent digits’ strings using level-by-level time alignment," International Journal of Information Technology and Knowledge Management, vol. 2 no. 1, pp. 65-68, 2009.

[50] D. Jiang, G. Hu, G. Qi, N. Mazur, "A fully convolutional neural network-based regression approach for effective chemical composition analysis using near-infrared spectroscopy in cloud," Journal of Artificial Intelligence and Technology, vol. 1 no. 1, pp. 74-82, DOI: 10.37965/jait.2020.0037, 2021.

[51] S. K. Ryu, T. Kim, Y. R. Kim, "Image-based pothole detection system for ITS service and road management system," Mathematical Problems in Engineering, vol. 2015,DOI: 10.1155/2015/968361, 2015.

[52] S. R. Qasim, J. Kieseler, Y. Iiyama, M. Pierini, "Learning representations of irregular particle-detector geometry with distance-weighted graph networks," The European Physical Journal C, vol. 79 no. 7,DOI: 10.1140/epjc/s10052-019-7113-9, 2019.

[53] YOLOv3-Face-Mask-Detection, "Face mask detection using YOLOv3 on Google colab," 2021. https://github.com/VictorLin000/YOLOv3_mask_detect

[54] D. Patel, "Applying perspective transformations to 2d images," , 2021. https://github.com/deepampatel/BirdsEye