International Journal of Engineering and Advanced Technology (IJEAT)

ISSN: 2249-8958 (Online), Volume-9 Issue-3, February 2020

Automatic Student Attendance System using

Face Recognition
Partha Chakraborty, Chowdhury Shahriar Muzammel, Mahmuda Khatun, Sk. Fahmida Islam,
Saifur Rahman

when they enter the classroom. Because of its non-intrusive

Abstract: The most common difficulty that every teacher faces nature and familiarity, face recognition is chosen because
in class room is to take the attendance of the students one by one people recognize other people primarily based on their facial
in each and every class. For the time being many automated characteristics. The biometric (facial) system consists of an
systems has been proposed for taking student attendance. In this
paper, I introduced an automated student attendance system
enrollment process in which the specific features of a
based on the use of unique techniques for face detection and person's face are registered in a database and then the
recognition. This system automatically detects the student when he recognition and authentication processes. The detected face
or she enters the classroom and recognizes that specific student (acquired from the webcam) in a picture is matched with the
and marks the student's attendance. This method also focuses on previously stored faces captured in this process at the time of
the specific features of different attributes such as the face, eye attendance.
and nose of humans. In order to evaluate the performance of
different face recognition system, different real-time situations are
Many universities still use conventional methods for
considered. This paper also suggests the technique for handling student attendance. Because this approach is used, many
the technique such as spoofing and avoiding student proxy. This students support their peers by signing in their attendance if
system helps track students compared to traditional or current they are absent from the school. So while using this process,
systems and thereby saves time. the faculty analyzes and manually manages attendance
records to know the number of present and absent students. If
Keywords: Face detection, Feature Extraction, Face
the attendance sheet is lost, the faculty will have to attend
recognition, Eigenface, Haar Cascade Classifier, Principal
Component Analysis (PCA) again and in this case students who are absent will have the
chance to make their present in a new sheet. The practice will
I. INTRODUCTION also affect students as time is spent on signing, checking and
sending the attendance sheet manually, in addition to being
Student attendance is taken manually through the use of inconvenient for the lecturer. Therefore, it is necessary to
attendance sheets issued by department heads as part of the develop a computerized system that can handle and allow the
legislation in most learning institutions. In these sheets, lecturers to easily attend and sustain the attendance. This
students sign up for future study, which is then filled in or program can be easily accessed by the faculty. The program
signed on to a computer manually. This approach is slow, must take careful monitoring and control of student
time-consuming and inconsistent as some students often sign attendance data in order to minimize the faculty's manual
for their missing colleagues. It also makes it difficult to track review of student attendance. The program analyzes all the
the attendance of individual students in a large classroom data automatically as it is transmitted by the faculty. Instead
setting. In our work, we propose the design and of traditional methods, by using our face detection and
implementation of a face detection and recognition system to recognition system, you can get a fast and effective method to
automatically detect students attending a lecture in the collect student attendance reliably while providing safe,
classroom and recognise their attendance by recognizing stable and robust system record storage where allowed.
their faces.
While other methods of biometric authentication may be A. Contribution of Our Work
more reliable, students usually have to queue for a long time We can get rid of common, manual and time-loss methods
for marking student attendance in classrooms using this
automatic method. Using face detection and recognition
Revised Manuscript Received on February 05, 2020. automated attendance can boost attendance control and
* Correspondence Author
Partha Chakraborty*, Department of CSE, Comilla University, management effectiveness. A software application module
Comilla - 3506, Bangladesh. Email: [email protected] that faces the recognition of the images faces recorded in the
Chowdhury Shahriar Muzammel, Department of CSE, Comilla video, marks the students ' register and then stores the results
University, Comilla - 3506, Bangladesh. Email: [email protected]
Mahmuda Khatun, Department of CSE, Comilla University, Comilla - in a database for future study. The system uses the Eigenface
3506, Bangladesh. Email: [email protected] recognition technique to recognize faces. The system
Sk. Fahmida Islam, Department of CSE, Jahangirnagar University, evaluates and measures Eigenfaces which are faces made of
Savar, Dhaka, Bangladesh. Email: [email protected]
Saifur Rahman, Department of CSE, Comilla University, Comilla - eigenvectors. This approach also estimates the Eigenfaces to
3506, Bangladesh. Email: [email protected] determine the presence of a person (face) and its identity.
This method involves the following steps:
© The Authors. Published by Blue Eyes Intelligence Engineering and
Sciences Publication (BEIESP). This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Published By:
Retrieval Number: B4207129219/2020©BEIESP
Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B4207.029320
Journal Website: www.ijeat.org
93 & Sciences Publication
 Automatic Student Attendance System using Face Recognition

Train face dataset. Next when we sense a face, we collection consists of M images. Where each image size is
calculate Eigenface for that face. The system then compares N/N
the current face's eigenvectors and the face picture stored Step 1: Convert the images to face vectors in the training
decides whether or not the face is detected. The final step set. The pictures are translated to vectors of the board. Let a
(optional) is that the system learns to recognize faces if the face image be an 8-bit intensity two-dimensional N by N
unknown face is found repeatedly. array. An image can also be regarded as a squared dimension
N vector. So a standard 50 by 50 size images becomes a 2500
II. RELATED WORK dimensional space vector. A set of images, then, maps this
huge space to a selection of points.
The main reason behind the system is to analyze and find
Step 2: Standardize face vectors, i.e. remove all common
solutions from other writers and recognize their suggested
features. This is done by measuring and subtracting from all
system imperfection and provide the best solutions.
the facial vectors the average facial vector.
Kawaguchi developed a classroom attendance system with a
Step 3: Calculate ownership of the covariance matrix. As
new method called continual monitoring, and student
the main components are obtained from the covariance
attendance reported dynamically by the camera capturing the
matrix in the PCA-based Eigenfaces method, we now
image of a student in the classroom [1]. The design of the
calculate the own vectors from the covariance matrix. It
network is simple as the class wall is fitted with two cameras.
would take a lot of time to find K own vectors from M=2500
The first is a camera that captures the image of students in the
own vectors with K<M due to the many calculations that
classroom; the other is a sensor camera to capture the chair of
need to be done and therefore the need for Step 3
a student. The program compares the image taken from a
Dimensionality reduction.
camera that records photographs and faces in the database
Step 4: Increasing the training set's dimensionality. To
that made a lot of time to make the attendance great. Another
each these calculations on the required individual faces, we
paper proposed by [2] implemented in the automated
calculate them from a reduced dimensionality covariance
attendance management system a real-time computer vision
matrix.
algorithm. The system installed the non-intrusive camera
Step 5: Choose K's best individual vectors so that K < M
capable of snapping images in the classroom and comparing
can represent the entire training set. The yellow bars show the
the captured face with faces within the system from the
selected vectors of K Eigen which are adequate to represent
camera's image. The approach often uses techniques that are
the entire training set.
often used in computer vision for machine learning. Haar
Step 6: Transform lower-dimensional K own vectors to the
classifiers were used in camera capture images to train. The
dimensionality of the original face.
camera's face snap converts to grayscale and retracts the files,
Step 7: Each image is represented as a linear combination
which is then forwarded for later saving and processing on
of all K own vectors plus the mean / average face.
the server in 2012Used two libraries including OpenCV is a
Step 8: Calculate and store the corresponding weight
computer vision library and FLTK to implement the program.
vectors for each image in the training set.
Both of these libraries contributed to the development of the
interface, such as the OpenCV support algorithm [4] and B. General Overview
FLTK. An integrated attendance management system was In this model, many similar components have been
implemented by Kar [3] using face recognition techniques. clustered to form sub-systems which then combine to form
The system provides matching requests and adding new the whole system. Breaking the system into modules and
information to the database. Request Matching's first move is sub-systems reveals the practical essence of the class
to open the shutter and snap the picture after removing the attendance system. Fig. 1 flow diagram shows the operation
front face. The next step is to classify the face with the of the processes.
training data and to paper the extracted face onto the main
feature visualization. A smart attendance marking system
that involves two differentiating algorithms such as Principal
Component Analysis and Artificial Neural Network ase
suggested by Jyotshana Kanti [5]. Zacharia proposed using
face recognition [6] an advanced attendance management
system in which only PCA was used. J.G. in their inquiry.
Roshan Tharanga et al proposed an intelligent way of
marking the attendance [7]. K. Senthamil Selvi et al
discussed recent developments in the topic in their paper[8].
Shireesha Chintalapati et al, in their report, discussed face
recognition uisng PCA, LDA, and LBPH [9].

III. METHODLOGY

A. Principle Face Recognition (PCA) Algorithm

Fig. 1. Sequence of events in the class attendance system
In this system, we used PCA algorithm for face recognition.
This is step by step algorithm overview.
Step 0: Creating and loading a training set. The training

Retrieval Number: B4207129219/2020©BEIESP Published By:

DOI: 10.35940/ijeat.B4207.029320 Blue Eyes Intelligence Engineering
Journal Website: www.ijeat.org 94 & Sciences Publication
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249-8958 (Online), Volume-9 Issue-3, February 2020

From Fig. 1, we can see that most of the components used are F. Functions of the two Sub –Systems
identical only to be used at different stages of the face The functionality of the components is shown in Figure 3
recognition process. block diagrams. The face recognizer system consists of two
main components, i.e. the manager of the training set and the
C. Training Set Manager Sub System face recognizer.
Those two modules must share the Faces database, the Face
The training set management sub-system's conceptual
Detector system image acquisition; as they are similar in their
architecture should consist of a component of image features. Therefore, we must partition the machine into two
acquisition, a face detection feature, and a component of subsystems and have to enforce their comprehensive
training set management. Together, to control the training conceptual designs.
collection, these elements work with the faces database.
These will be developed in an application form for Windows.
D. Face Recognizer Sub System
The Face Recognizer's conceptual architecture would
consist of the component of image processing, face
recognizer, and face detection system all operating with the
database of faces. Here, the portion of image acquisition and
face detection is the same as those in the Training Set
Manager sub-system as the same features. The only
difference is the face recognizer component and its user
interface controls. This will reload the training set on the
added faces to train the recognizer and show the calculated
Eigen-faces and average face. It should then display the Fig. 3. A block diagram showing functions of the
recognized picture in a photo box components
E. System Architecture G. Full Systems Logical Design
The system consists of a camera recording the student’s The logical design of the full system are composed several
photographs and submitting them to the server for image components. The system first takes an image sample of
enhancement. The picture will appear in the Face Detection student from input devices such as camera. Then we enhance
and Recognition modules after enhancement and then the the image using some image enhancement process for better
presence will be identified on the server of the database. This face detection and recognition. After successfully identifying
is shown in Figure 2 of the experimental setup. Templates of the face by comparing with face database the system can give
individual student’s face photos are stored in the Face attendance to the identified students and save it to attendance
database at the time of enrollment. The input image identifies sheet.
all the faces here, and the algorithm compares them one by
one with the face database. It saves a lot of time and this is a
highly secure procedure that no one can mark someone else's
attendance. Face database is the collection of face images and
extracted features during the registration process and the
second attendance inventory includes student information as
well as uses to identify attendance. The figure below
illustrates the three application subsystems ' conceptual
design and implementation.

Fig. 4. Whole system logical design

H. Tools Used
We used the EmguCV Library to process images. EmguCV
is a wrapper for the OpenCV image processing library.
Visual Studio, Xamarin Studio and Unity can compile the
wrapper, it can run on Windows, Linux, Mac OS X. EmguCV
uses a form of face detector called a hair cascade. The Haar
Cascade is a human face-trained classifier (detector). The
attendance result sheet is stored on this paper folder in a text
file.

Fig. 2. The logical design of the Desktop Module

Subsystems

Published By:
Retrieval Number: B4207129219/2020©BEIESP
Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B4207.029320
Journal Website: www.ijeat.org
95 & Sciences Publication
 Automatic Student Attendance System using Face Recognition

IV. RESULT compared to the database. After a good match, the name
associated with the face will be shown on the top edge of the
A. User Interface of the System rectangular box. Now shown in the highlighted box number 2
The attendance system adds faces to the training set for are the numbers of faces in the scene with their names. To
extracting feature. The image is taken from the marked red connect faces to the database, use face press.
box as can be seen in Fig. 5 and displayed in the yellow box
on a picture box. The Areas of Interest (AOI) i.e. face (es) in
the drawing would be automatically identified by drawing a
light yellow rectangular box. Then we assign a face mark to
the blue rectal to the extracted gray scale.

Fig. 8. Live camera feed window

To enhance the system's recognition performance, nine
photos were selected for training for each person from the
regular Trained Faces database; the remaining two photos
were selected for the test collection. Of the fifteen subjects
Fig. 5. Training set editor from the database of trained faces, twelve faces were
recognized correctly. This was in proportion to the precision
Then we can adjust the pairs of face marks in the purple box
of 90%. Time consuming faces where there is a large number
if they are mistakenly identified or even eliminate the faces if
of subjects involved.
they do not meet the criteria.
B. The Face Recognizer
The system match the observed image's input face to the
faces captured during the recording process. If it is a match
then it will retrieve the name associated with the input image.

Fig. 9. Images from class

C. Attendance Sheet Creating and Marking
When we click the CLICK HERE button a text file create
in Face recognizer folder. In a new day create a new text file
for attendance. In figure 10 shown the attendance register as a
Fig. 6. Face recognizer with single student database.

Fig. 7. Face recognizer with multiple students Fig. 10. Attendance register
Step 1 is for the marker to be qualified to recognize a face as The returned input face I d number was then used to fill the
known or unknown. Step 2 selects the recognizable face of records in the database of the attendance registry. Clicking
the image source. This could be complemented by a live the Attendance sheet button will reveal the sheet as shown in
camera recorded images. The face input image will then be figure 11. The images are identified, monitored and
shown in the recognizer picture box 2 as shown in Figure 6. recognized automatically in the video feed. Photos can also
The input face ID of the image will then be shown as shown be transferred from the live to the database.
in Figure 7, Step 5.
The outlined box 1 shows the camera's present view / scene
from Figure 8. The faces and eyes in the pictures are
automatically marked as shown by the rectangular boxes
around them. The face detected is extracted from and

Retrieval Number: B4207129219/2020©BEIESP Published By:

DOI: 10.35940/ijeat.B4207.029320 Blue Eyes Intelligence Engineering
Journal Website: www.ijeat.org 96 & Sciences Publication
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249-8958 (Online), Volume-9 Issue-3, February 2020

Veiled
10 10 7 70
Faces

Bearded
10 10 8 80
Faces

Unveiled
20 20 17 85
A

Unveiled
30 30 25 83.3
B

With a different combination of methods and algorithms this

Fig. 11. The Attendance Sheet system helps us to achieve the desired results more precisely.
D. Analysis The minimum variable detection scale eliminates the
This work takes part in five stages: identification and recognition distance issue for both close
Step 1: Training Data generation: The system that is the and group images. It improves the exactness of the face
cropped images is initially trained and saved to the database detection for upright faces to 100% and increases the
and is detected and recognized. This data will also be used to precision of the face recognition by usually 70%. Similarly,
compare the detected images in all the files uploaded and to the setting of minimum neighbors enhanced the precise face
mark the attendance. detection significantly. The rectangular section of the
Step 2: Capture: Use a tool to capture the video or image and observed face rather than the whole image captures and
upload the captured file to the server using the web transform decreases the influence of background noise on the
application. visual accuracy of the device.Histogram equalization is
Step 3: Face Detection: Using the Haar-cascade classifier, the performed to the sample images, achieving uniform intensity
file uploaded to the server is detected face. The obtained distribution of the 46 output images through the reassignment
frames are tested for the faces and for further identification of intensity pixels. All input images with different lighting
they are cropped. are therefore enhanced in detail, leading to better facial
Step 4: Face Recognition: The images detected are correlated recognition performance. Figure 12 shows the first 32 Eigen
with the database's trained images. This now recognizes the faces from a set of 50 faces created by each of the five men.
detected images. The first few Eigen-faces display dominant facial features
Step 5: Marking attendance: The recognized students are and the last 196 to 247 Eigen-faces, as shown in 13, are
searched in the database after the recognition process and mostly picture noise and are therefore discarded. The average
their attendance is marked. face from Figure 14 shows the smooth face structure of a
Tests were carried out for several subsets specifically of the generic human being.
Yale dataset out of 60 faces in the sample. Tabulated the
results obtained as in Table-I. The percentage identification
rate for the various subsets was determined as the sum of the
percentages. Recognition levels were not influenced by faces
with or without glasses. The average percentage
identification rate was 80.22%. The middle light faces had
the best overall ranking of 90%. The main issues in the
majority of the currently in use facial detection and
recognition systems are rotation, orientation, distance of
identification and lighting. Such issues reduce the system's
performance unless it is carried out under the required
constraints. Such restrictions would include putting the Fig. 12. The first 32 Eigen faces
subjects in specific positions, which would be very The most prominent facial characteristics of the photos to be
challenging in a real world classroom environment and not to tested are found by the first Eigenfaces. The successor
mention time consuming when there are a large number of Eigen-faces (main components) shows the next highly likely
subjects involved. facial features and more noise in consequence. Of the 247
Table- I: Experimental results training images, 195 main components and the average face
are adequate to recreate the entire training kit entirely. We
Successfully Successfully were therefore able to convert a set of correlated face
No of % Correct variables (M) into a set of values of uncorrelated variables
Datasets detected Recognized
Faces recognition
Faces Faces called own vectors. It was noted that the number of
Eigen-faces is less than the original images of the face i.e. K
< M, PCA-based.
Center
10 10 9 90
light

Left
15 15 11 73.3
light

Right
15 15 12 80
light

Published By:
Retrieval Number: B4207129219/2020©BEIESP
Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B4207.029320
Journal Website: www.ijeat.org
97 & Sciences Publication
 Automatic Student Attendance System using Face Recognition

system has been developed to popularize errors. The efficient

and reliable system of the attendance that can replace the old
manual methods in the office environment. This method is
sufficiently stable, accurate and ready for use. There is no
need for specialized equipment in the office to mount the
device. It can be designed with a camera and a computer.
Authors intend to improve the effectiveness of face
recognition by using the interaction between our system,
users, and administrators in further work. On the other hand,
our system can be used in a completely new face recognition
application dimension, mobile face recognition, which can
Fig. 13. The last Eigen-faces in the training set help ordinary people to know about any person being
photographed by cell phone camera, including proper
authorization to access a centralized database.

REFERENCES
1. Y. Kawaguchi and T. Shoji, “Face recognition-based lecture
attendance system,” Academic Center for Computing and Media
Studie, 2005.
2. K. S. Kumar, S. Prasad, V. BhaskarSemwal, and R. C. Tripathi, “Real
time face recognition using adaboost improved fast pca algorithm,” Int.
J. Artif. Intell.Appl, p. 45–58, 2011.
3. N. Kar and D. M. K. D. Barma, “Study of implementing automated
at-tendance system using face recognition technique,” International
Journal of Computer and Communication Engineering, 2012.
4. T. Bradski and A. Kaehle, “Opencv ml algorithms,” in Learning
OpenCV. O’Reilly Media, Inc, 2008, p. 580.
5. J. Kanti and J. Papola, “Smart attendance using face recognition with
percentage analyzer,” International Journal of Advanced Research in
Computer and Communication Engineering, vol. 3, 2014. [Online].
Fig. 14. The average face Available:
Interestingly, the key component analysis can be used for https://ijarcce.com/wp-content/uploads/2012/03/IJARCCE10A-a-Jyot
compression of images from the Eigenfaces obtained in the shana-smart-attendance.pdf.
face recognition stage, as shown by the dominant numbers of
6. J. Joseph and K. Zacharia, “Automated attendance management system
Eigen faces that can comfortably represent all images in the using face recognition,” International Journal of Science and research,
trainings package. Just 195 faces are required for the 247 vol. 2, 2013.
pictures of the training set to completely re-create all the 247 7. R. Tharanga, Samarakoon, Karunarathne, Liyanage, and D. Parer,
faces. “Smart Attendance using real time face recognition (smart-fr),”
Semantic Scholar, 2013.
8. K. Selvi, P.Chitrakala, and A. Jenitha, “Face recognition based
V. CONCLUSION attendance marking system,” IJCSMC, no. 3, p. 337 –342, 2014.
9. S. Chintalapati and M. Raghunad, “Automated attendance
A program has been suggested to ensure attendance. This management system based on face recognition algorithms,”
replaces the manual system with a simple, reliable, International Conference on Computational Intelligence and
cost-effective, time-saving automated system as it eliminates Computing Research, 2013.
the stationary material and paper work. It is therefore
predicted that this method would produce desired results and AUTHORS PROFILE
could be applied for logout in the future. In the near future,
other strategies could also improve efficiency. We can say Partha Chakraborty is a member of the faculty at the
that a safe, stable, rapid and efficient class attendance Department of Computer Science and Engineering,
Comilla University, Bangladesh is currently working as
management system was developed to replace a manual and an assistant professor. He received a M.Sc. from the
inefficient system. The face recognition program saves time, Jahangirnagar University, a renowned Bangladesh Public
reduces administration work and replaces obsolete electronic University, and a B.Sc. in the field of computer science
equipment for the currently in use stationery content. The and engineering. He also has an interest in computer vision, robotics, image
processing and artificial intelligence for his work. To various international
device must not be mounted by specialized equipment, journals, he presented his research articles. He is actively involved in the
because it requires only a computer and a camera. The picture fields of education and learing.
quality and performance must be tested in the real time
scenario, particularly when the systems are operated from a Chowdhury Shahriar Muzammel a faculty member
of Comilla University, Bangladesh, Faculty of
live camera supply, and thus the camera plays a decisive role Engineering, is currently working as a lecturer in the
in the system's work. The mechanism can also be used for Department of Computer Science and Engineering. He
permission-based access control systems and for safe access was awarded M.Sc. (Engineering) Bachelor's degree
authentication (restricted installations). The greatest threat to and B.Sc. (Technology) Comilla University degree in
Computer Science and Engineering, a renowned Bangladesh public
the system is spoofing. Anti-spotting techniques like eye university. Signal processing, Bangla Natural Language Processing, Image
twitch detection can be used in the event that the face Processing, and Artificial Neural Network are his current research interests.
recognition is provided by grabbed images for possible In various international journals, he presented his research articles.
improvements to separate live from static images. A feature
can therefore be added that lists all unidentified faces and can
be manually checked by the user. Automatic attendance

Retrieval Number: B4207129219/2020©BEIESP Published By:

DOI: 10.35940/ijeat.B4207.029320 Blue Eyes Intelligence Engineering
Journal Website: www.ijeat.org 98 & Sciences Publication
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249-8958 (Online), Volume-9 Issue-3, February 2020

Mahmuda Khatun, a faculty member of the Faculty

of Engineering, Comilla University, Bangladesh, is
currently working as a Lecturer at the department of
Computer Science and Engineering. She has completed
her B.Sc. (Hons) in Computer Science and Engineering
from Jahangirnagar University, Savar, Dhaka in 2016 and
M.Sc. from the same university in 2018. Her teaching experience includes
different graduate (M.Sc.) and under graduate courses. Her current research
interest includes Wireless Communication, Machine Learning, Artificial
Intelligence and Image Processing.

Sk. Fahmida Islam received the B.Sc. and M.Sc.

degrees in computer science and engineering from
Jahangirnagar University, Savar, Dhaka,
Bangladesh.and currently a Ph.D student in the
department of Computer Science and Engineering of
Jahangirnagar University. Her current research interests
include image processing, Internet of Things, vehicle tracking and mobile
cloud computing.

Saifur Rahman is currently a student in the Department

of Computer Science and Engineering of Comilla
University, Bangladesh. His research interests are in
Data Science, Machine Learning and Natural Language
Processing.

Published By:
Retrieval Number: B4207129219/2020©BEIESP
Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B4207.029320
Journal Website: www.ijeat.org
99 & Sciences Publication