Student Attendance using Face Recognition

Technology
Jack Febrian Rusdi Frans Richard Kodong Richardus Eko Indrajit
Department of Informatics Department of Informatics Engineering Pradita University
Sekolah Tinggi Teknologi Bandung UPN “Veteran” Yogyakarta Department of Information Technology
Bandung, Indonesia Yogyakarta, Indonesia Tangerang, Indonesia
2020 2nd International Conference on Cybernetics and Intelligent System (ICORIS) | 978-1-7281-7257-6/20/$31.00 ©2020 IEEE | DOI: 10.1109/ICORIS50180.2020.9320819

[email protected] [email protected] [email protected]

Herry Sofyan Abdurrohman Robert Marco

Department of Informatics Engineering Department of Informatics Fakultas Ilmu Komputer
UPN “Veteran” Yogyakarta Sekolah Tinggi Teknologi Bandung Universitas Amikom Yogyakarta
Yogyakarta, Indonesia Bandung, Indonesia Yogyakarta, Indonesia
[email protected] [email protected] [email protected]

Abstract—In the world of education, recording student Communication (NFC) [10]. Various researchers and
attendance is one crucial element. Student presence record is application developers have implemented this attendance
one of the determinants of student graduation. In almost every system application. With various sensor technologies,
education field, attendance records that accurately recorded. attendance systems can be developed in various forms.
The presence of facial recognition technology is one of the
solutions used to record student attendance during class The presence of facial recognition technology is one of
meetings. This article discusses the model of student the solutions used to record student attendance during class
attendance recording that can become part of the campus meetings [2], [11]. This technology uses face capture
academic information system. The technologies used in through cameras and object protection systems that make up
recording student attendance include face recognition human faces [12]. This system then discusses each face's
technology through Computer Vision and Matlab. This study object, then each item is predicted with a list of verified-
discusses the system module and the factors that affect the faces in the library of student's faces.
accuracy and performance of the system. This model can be
the basis for developing student attendance systems at This study discusses the model of face recognition
universities or other educational institutions. This study applied to student participation systems in tertiary
proves that the process of recording through face recognition institutions.
has the potential to be applied to educational systems in
institutions, both as the primary system or as a support system
The facial recognition technology model applied to the
of an existing system. college attendance system can be one solution for
universities to use attendance systems on each campus.
Keywords—face recognition, attendance system, computer
vision, student attendance, machine learning.
II. RELATED STUDY
Computer Vision is Digital Image Processing [13]. So
that computers can see an object like humans know an object
I. INTRODUCTION [14]. Thus, the machine can recognize an item, make
decisions, take action, and count the number of objects [15].
In the world of education, recording student attendance
is one crucial element. Student presence record is one of the Through Computer Vision [16], computers learn
determinants of student graduation. In almost every existing data and learn by themselves, so machines can see
education field, attendance records that accurately recorded like humans see things [2]. Through this ability, humans use
[1], [2]. computers in various activities that observe and pay
attention to objects until finally, the machine use to the next
With the presence of various information and level, including student attendance systems in the learning
communication technology capabilities, this has resulted in process with their lecturers.
the development of processes from manual to semi-
automatic or automatic. The diversity of existing sensor The presence and development of an accelerated
technologies provides various combinations of processes Graphics Processing Unit (GPU) [17], machine learning has
that have the opportunity to be developed by researchers and become one of the emerging technologies lately [18],
developers in various methods, including the method of especially the use of fixed and moving images for various
using the camera [3]. processes needed by humans and their developers.
Various methods of recording student attendance are One of the technologies used is deep systematic learning
applied by higher education, both manual, semi-automatic, [19], [20]. The wide scope of deep learning is growing,
and even by using technology [4]. Different ways of making this technology popular in information processing
recording student attendance are available and developed at through natural language processing, including capability
this time; of course, each method has advantages and extraction and visualization processes.
disadvantages [5]. Deep Learning technology has an excellent ability to
Examples of the use of technology in this attendance handle processes related to face recognition, including its
system include RFID [6], social media [7], barcodes, application to Computer Vision. One of the capabilities used
Bluetooth [8], fingerprints[9], and Near Field

Authorized licensed use limited to: K K Wagh Inst of Engg Education and Research. Downloaded on September 18,2023 at 05:11:09 UTC from IEEE Xplore. Restrictions apply.
 is in terms of classifying objects in an image, known as The programming language used is to detect image-
Convolutional Neural Networks (CNN) [21], [22]. based objects, using R-CNN as a detector. R-CNN or
Regions with Convolutional Neural Networks is one of the
The detection of objects in the form of images continues object detectors used in deep learning, mainly related to
to grow, including the presence of various alternatives such convolutional neural network areas.
as the success of the regional proposal method and the R-
CNN [23]. IV. RESULTS AND DISCUSSION
III. METHOD The concept applied in this study show in Fig. 2. This
study provides a solution to the process of recording the
This research conduct at the Universitas Pembangunan presence of lectures with face recognition methods using
Nasional “Veteran”, Yogyakarta. The basis of information Computer Vision, both with deep learning and the use of
and knowledge base on the study of various kinds of Convolutional Neural Networks and Raster R-CNN Matlab.
literature sourced from multiple journals.
The study stages are based on the concept of the System
Development Life Cycle (SDLC) (see Fig. 1). These stages
include Communication, Planning, Modeling, Construction,
and Deployment [24].
At the communication stage, a process is carried out to
understand the needs of the attendance system. The planning
stage is a step related to preparing various related steps so Fig. 2. The design of attendance recording systems using Computer
that the system can be implemented. The modelling stage is Vision
the stage for creating a model based on the things that have
been planned. The Construction Phase is the technology In developing a student attendance recording system
phase built, including hardware, software, and supporting during lectures by utilizing face recognition technology, at
infrastructure. Meanwhile, the deployment stage is the least four modules prepared, starting from the face taking
application of the system so that it can be used properly. the module, training module, camera module, and
attendance module. This stage present Fig. 3.
The system developed using Computer Vision to utilize
capabilities to automate tasks related to the human Implementation of the system is carried out by recording
perspective, namely capturing faces, analysing, searching, each student's data, then conducting training data and testing
and recording assigned tasks. with functions in the CNN library. When the accuracy is
above 90%, the system is used in the student attendance
process, as in prototype experiments conducted in class.

Face taking Training Camera Attendance

Modul Modul Modul Modul

Fig. 3. Modules for face recognition on student attendance systems

Fig. 1. Life Cycle of Waterfall Software Development in the student In the system of recording the presence of students
attendance recording system using face recognition through the introduction of human faces as objects, there is
an initial stage process that works to record human faces that
Thus, the system built through this model has the are used as objects, namely the faces of students—this
potential to replace the existing manual system. The manual collection of face images stored in the library (Fig. 4).
process becomes an automatic attendance recording system,
including enriching research and studies from a
technological perspective in applying student attendance
recording systems.
In the processed system through this study, the CNN
detector used is the face pattern recognition method. The
CNN method library used comes from Matlab 2019.
Matlab is a numerical computational environment and
includes in the fourth-generation computer programming
language.
Matlab capabilities used in this method are mainly in the
object detection framework, including image manipulation,
data function plotting, algorithm implementation, and user Fig. 4. Library of face images that will be recognized by the system
interfaces.
This image library that used as a source during the face
initiation process sought by the system. If found, their

Authorized licensed use limited to: K K Wagh Inst of Engg Education and Research. Downloaded on September 18,2023 at 05:11:09 UTC from IEEE Xplore. Restrictions apply.
 attendance recorded in the student attendance system. The If the face found matches the predicted face trained, then the
more images taken and stored in the library will increase the attendance process is carried out—an example of the
level of accuracy of the system in human face recognition process in this module, shown in Fig. 5 below.
from the tests carried out.
The process in this camera module is to recognize
The gallery image requires sufficient space. The storage objects in the shape of faces, then capture faces to be
capacity is an issue to consider here. The capacity used is compared with face predictions in the face library that
directly proportional to the number of images stored in the previously initiated.
library. The higher the resolution used will help with more
precise prediction results. Likewise, with the number of Face detection in real-time which done during lectures
images used as a library, the more images that become in the classroom. For this, an interface that needed that
references will help the system have higher accuracy in connects the program Matlab with the equipment. In this
predicting students' faces. case, detection using a camera where scattered in class.
Placement of several cameras which done to enhance the
The second stage is the training module, which is the accuracy of facial recognition. Besides that, the installation
algorithm and completeness module related to the system's of some angle taking positions also increases the precision
process to train itself to recognize the faces of students based of facial pattern recognition.
on data stored in the library.
Accuracy information can also be monitored, especially
The code compiled in the training module provides steps as a consideration in building and developing systems. This
about the processes carried out in the system. Programs that level of accuracy present in Fig. 6.
are made through effective and efficient algorithms will help
the program to run correctly. Before this system was run, the attendance system was
manual. Manual data obtained is used for training on the
One part of the facial swallowing logic is as follows: system until the system produces facial recognition. From
direct experiments in class comparisons between manual
real attendance and those carried out through this system,
img = imread(student_face_file); testing on CNN showed the results of the system training
[img, face] = cropface(img); with high accuracy, reaching 99% of the actual data against
the predicted data from the students' faces.
%face value is 1 when it detects face in image or 0
if face == 1
img = imresize (img, [227 227]);
predict = classify (newnet, img);
end

In the above program logic, the image is first stored in

memory. Furthermore, the images in this memory are
identified only by the face, namely by cropping the human
face if it is found. If a face is found that face is predicted.
Next is the Camera Module. This module recognizes
human faces and looks for faces in the library of faces that
previously trained.
The camera is placed in a strategic location, which is a Fig. 6. The level of accuracy of the results of training with the CNN
method
location that is thought to be able to capture the faces of the
students clearly. In a variety of conditions, this position
The performance of facial recognition technology
helps to obtain accurate image capture by the system. Also,
applied to the attendance system which influenced by
the lighting system can influence the capture and clarity of
various criteria, such as algorithms, completeness of the
images obtained by the camera. Another thing to note about
library, facial recognition aids, camera resolution, processor
this camera module is related to the resolution of the image
and memory strength, number of object image libraries that
capture and the camera's autofocus ability.
identified.
Of the various cases found at the time of this study, facial
recognition technology's accuracy and performance applied
to the present system influenced by at least the following
three factors, the devices used, the system and logic, and the
situation on the ground (Fig. 7).
Accuracy and performance factors for devices are
affected by the processor, memory, camera resolution, and
network. The system and logic influence by several things.
This logic is influenced by how effective the algorithm is
used, how the system is built, the quantity, and the quality
of object image libraries' availability. As for the field
situation, accuracy and performance are influenced by,
among others, the available lighting power.
Fig. 5. Facial correction and capture of student attendance pictures

Authorized licensed use limited to: K K Wagh Inst of Engg Education and Research. Downloaded on September 18,2023 at 05:11:09 UTC from IEEE Xplore. Restrictions apply.
 Conference on Recent Trends in Electronics, Information &
Communication Technology (RTEICT), 2016.
[4] D. Sunaryono, J. Siswantoro, and R. Anggoro, “An android based
Accuracy course attendance system using face recognition,” J. King Saud
and Univ. - Comput. Inf. Sci., 2019.
Performance
[5] R. Taufiq, M. Baharun, B. Sunaryo, B. Pudjoatmodjo, and W. M.
Utomo, “Indonesia: Covid-19 and E-Learning in Student Attendance
Method,” SciTech Framew., vol. 2, no. 1, pp. 12–22, 2020.
Space [6] K. R. Pireva, J. Siqeca, and S. Berisha, “RFID: Management system
Equipment System Logic
Situation for students’ attendance,” in IFAC Proceedings Volumes (IFAC-
PapersOnline), 2013, vol. 15, no. PART 1, pp. 137–140.
[7] M. Syaiful Romadhon, A. Rahmah, and Y. Wirani, “Blended
learning system using social media for college student: A case of
tahsin education,” in Procedia Computer Science, 2019, vol. 161, pp.
Fig. 7. Factors that affect the accuracy and performance of the face 160–167.
recognition process [8] R. Lodha, S. Gupta, H. Jain, and H. Narula, “Bluetooth Smart based
attendance management system,” in Procedia Computer Science,
The system that tested through the student attendance 2015, vol. 45, no. C, pp. 524–527.
system during lectures using Matlab, R-CNN, and vission is [9] L. K. Almajmaie, O. N. Ucan, and O. Bayat, “Fingerprint
paperless electronic presence. This system can be developed recognition system based on modified multi-connect architecture
in determining participation in exams and can also (MMCA),” Cogn. Syst. Res., vol. 58, pp. 107–113, Dec. 2019.
overcome fraud in the presence that has been using the [10] I. L. Ruiz and M. Á. Gómez-Nieto, “Combining of NFC, BLE and
Physical Web Technologies for Objects Authentication on IoT
manual method (paper). Likewise, lecturers can determine Scenarios,” in Procedia Computer Science, 2017, vol. 109, pp. 265–
the value of student attendance in determining the 272.
evaluation value automatically for lecturers. Besides [11] U. Jayaraman, P. Gupta, S. Gupta, G. Arora, and K. Tiwari, “Recent
reducing the use of time wasted during teaching caused by development in face recognition,” Neurocomputing, 2020.
the attendance process, for meetings that have little face-to- [12] S. Afra and R. Alhajj, “Early warning system: From face recognition
face time, of course, this time issue becomes essential. by surveillance cameras to social media analysis to detecting
suspicious people,” Phys. A Stat. Mech. its Appl., vol. 540, p.
The benefits of developing a system through facial 123151, Feb. 2020.
recognition sensor technology can increase attendance [13] J. Qin, X. J. Shen, M. Zou, and S. P. Qin, “An Automotive Needle
accuracy and reduce the level of cheating that sometimes Meter Dynamic Test Method Based on Computer Vision and
occurs through manual absences manipulation. HILTechnology,” in Procedia Computer Science, 2018, vol. 154, pp.
588–595.
In a transparent and real-time manner, this system can [14] I. Jegham, A. Ben Khalifa, I. Alouani, and M. A. Mahjoub, “Vision-
monitor and control the attendance of lecture activities, and based human action recognition: An overview and real world
student examinations carried out through this system as the challenges,” Forensic Sci. Int. Digit. Investig., vol. 32, p. 200901,
Mar. 2020.
basis for the management of teaching and learning activities.
[15] A. Elmahmudi and H. Ugail, “Deep face recognition using imperfect
This study shows that the application of attendance by facial data,” Futur. Gener. Comput. Syst., vol. 99, pp. 213–225, Oct.
using a Computer Vision-based application can simplify the 2019.
student attendance process more easily and is done [16] S. V. Shavetov, I. I. Merkulova, A. A. Ekimenko, O. I. Borisov, and
V. S. Gromov, “Computer Vision in Control and Robotics for
automatically through a camera. Educational Purposes,” in IFAC-PapersOnLine, 2019, vol. 52, no. 9,
pp. 144–146.
V. CONCLUSIONS
[17] X. Liu, H. A. Ounifi, A. Gherbi, Y. Lemieux, and W. Li, “A hybrid
The use of face recognition technology supported by GPU-FPGA-based computing platform for machine learning,” in
Computer Vision can provide an alternative to record Procedia Computer Science, 2018, vol. 141, pp. 104–111.
student attendance. The process was carried out by forming [18] A. Winkler-Schwartz et al., “Artificial Intelligence in Medical
four machine learning modules specifically for recording Education: Best Practices Using Machine Learning to Assess
Surgical Expertise in Virtual Reality Simulation,” J. Surg. Educ.,
student attendance through student face recognition. The vol. 76, no. 6, pp. 1681–1690, Nov. 2019.
four process modules face taking the module, training [19] M. You, X. Han, Y. Xu, and L. Li, “Systematic evaluation of deep
module, camera module, and attendance module. The face recognition methods,” Neurocomputing, vol. 388, pp. 144–156,
system's accuracy and performance are influenced by May 2020.
various aspects, both from the computer equipment used, the [20] G. Guo and N. Zhang, “A survey on deep learning based face
system built, and the space situation. recognition,” Comput. Vis. Image Underst., vol. 189, Dec. 2019.
[21] A. Khatami, A. Nazari, A. Khosravi, C. P. Lim, and S. Nahavandi,
This study proves that the process of recording through “A weight perturbation-based regularisation technique for
face recognition has the potential to be applied to convolutional neural networks and the application in medical
educational systems in institutions, both as the primary imaging,” Expert Syst. Appl., vol. 149, Jul. 2020.
system or as a support system of an existing system. [22] D. X. Zhou, “Theory of deep convolutional neural networks:
Downsampling,” Neural Networks, vol. 124, pp. 319–327, Apr.
REFERENCES 2020.
[23] J. Zhang, L. He, M. Karkee, Q. Zhang, X. Zhang, and Z. Gao,
[1] K. Mohammed, A. S. Tolba, and M. Elmogy, “Multimodal student
“Branch detection for apple trees trained in fruiting wall architecture
attendance management system (MSAMS),” Ain Shams Eng. J., vol.
9, no. 4, pp. 2917–2929, Dec. 2018. using depth features and Regions-Convolutional Neural Network
(R-CNN),” Comput. Electron. Agric., vol. 155, pp. 386–393, Dec.
[2] S. M. Bah and F. Ming, “An improved face recognition algorithm 2018.
and its application in attendance management system,” Array, vol.
5, p. 100014, Mar. 2020. [24] de Vicente Mohino, Bermejo Higuera, Bermejo Higuera, and Sicilia
Montalvo, “The Application of a New Secure Software
[3] N. D. Veer and B. F. Momin, “An automated attendance system Development Life Cycle (S-SDLC) with Agile Methodologies,”
using video surveillance camera,” in 2016 IEEE International Electronics, vol. 8, no. 11, p. 1218, Oct. 2019.

Authorized licensed use limited to: K K Wagh Inst of Engg Education and Research. Downloaded on September 18,2023 at 05:11:09 UTC from IEEE Xplore. Restrictions apply.