International Journal of Scientific Research in Engineering and Management (IJSREM)

Volume: 07 Issue: 06 | June - 2023 SJIF Rating: 8.176 ISSN: 2582-3930

Attendance System using Face Recognition

is 77% and its false-positive rate is 28%. This

Abstract—In the current educational system, system is recognizing students even when
student’s frequent attendance in class is very students are wearing glasses or grown a beard.
important for performance evaluation and the Face Recognition of unknown persons is nearly
quality control. Most institutions still use 60% for both with and without applying
insecure, time-consuming traditional methods threshold value. Its false-positive rate is 14% and
like calling people by the name or having them 30% with and without applying threshold
sign documents. Face recognition technology respectively.
based on high-definition monitor video and other
information technologies is used to solve the Keywords—Face recognition, face detection,
problem of the recognizing face for taking machine learning, Haar Cascade Algorithm, Local
attendance. A facial recognition attendance Binary Pattens Histograms (LBPH).
system uses facial technology to automatically
record attendance while identifying and verifying
a person based on their facial features. This
research focusing on a face recognition-based I. INTRODUCTION
attendance system with getting a less false- Control of machines and their process with
positive rate using a threshold to confidence i.e., various technologies based on computer software
Euclidean distance value while detecting is called automation. In this modern age, these
unknown persons and save their images. advancements have proven to increase accuracy
Compare to other Euclidean distance-based and also helping us to improve our livelihood.
algorithms like Eigenfaces and Fisher faces, Innovations such as these save lots of labor work.
Local Binary Pattern Histogram (LBPH) One advancement in the field of automation is the
algorithm is better. We used Haar cascade for
Automated.
face detection because of their robustness and
The proposed system aims to create a face
LBPH algorithm for face recognition. It is robust
recognition-based attendance system with a less
against monotonic grayscale transformations.
Scenarios such as face recognition rate, false- false-positive rate. It uses Haar cascade and LBPH
positive rate for that and false-positive rate with algorithm, which are robust against monotonic
and without using a threshold in detecting grayscale transformations. It also detects and saves
unknown persons are considered to evaluate our the images of any unknown person in the class
system. We got face recognition rate of students whose Information is not in the database.

© 2023, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM23966 | Page 1

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 06 | June - 2023 SJIF Rating: 8.176 ISSN: 2582-3930

In [2] the authors projected an attendance This system provides functionalities such as taking
System utilizing RFID cards. In this approach, the images of students, training the images in the
projected RFID tag uses strength from the tag database and on the camera, and tracking people
reciter. The question at this moment approach is entering the classroom. When students enter the
that an person who is unfamiliar can use a genuine classroom, the system detects their faces from the
ID card and come the University. There are more camera and pre-processes them for further
documents including utilizing biometrics for
processing. The implementation of each stage is
attendance. In [3] mark is secondhand for
detailed in the next section.
designating the attendance of scholars. There is a
biometric sensor that will take the mark, feature
distillation approved at which point dossier. If it
is for admission therefore that dossier is stocked
in the table else if it is for confirmation therefore
that dossier is begun corresponding
accompanying the dossier in the table. The
question at this moment procedure is that for
attendance graduates concede possibility make
use of the place this fittings design is situated or
pass the fittings device about the juniors all the
while class that maybe a confusion to the pupils.
And in [4] and [12] iris located attendance Fig. 1. Overview of Proposed System
structure is secondhand. The question at this .
moment approach is that it is alert incidental III. Experiments
determinants. In [1] authors projected a face
acknowledgment located attendance system We created a database of 18+ individuals with 60
established Eigenface acknowledgment. Images images of each for this project, and 10 individuals
are con verted into eigenfaces, Recognition is are considered for testing recognition of unknown
acted by equating eigenface took from the persons. There are 18 individuals with 60 images
recommendation countenance and eigenfaces in and 10 individuals are considered for testing
the table. The question at this moment approach recognition.
is that this plan is very awake face backdrop, head We tested our system using a live real-time
orientations and it doesn’t acknowledge the face video in which students and unknown persons
of one if the customer is tiring pince-nez or a come and stand in front of the camera. Fig. 3
mature beard, etc. But in the approach proposed shows a few images after the pre-processing stage.
in this place paper, our structure is not alert face
education, head orientations and it perceives
one’s face even though he evolves a beard or
wears cheaters, etc.

II. PROPOSED SYSTEM

The proposed automated attendance
management system is based on haar cascade for
face detection and the LBPH algorithm for face
recognition. Fig 1

© 2023, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM23966 | Page 2

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 06 | June - 2023 SJIF Rating: 8.176 ISSN: 2582-3930

the best out of many as shown in Fig.4.[9]

Then Region of Interest (ROI) i.e
containing faces is extracted and sent to
next stage.

B. Face Recognition
The LBPH algorithm is used for face
recognition due to its robustness,
capability to recognize both front and side
faces, and better compared to Eigenfaces
and Fisherfaces[6]. It is easier to use as it
characterizes the image locally and
compares the result to each of the pictures
within the dataset. It works better in
different environments and light
conditions than other algorithms.
Local Binary Pattern (LBP) [5]
operation creates an image which
highlights the characteristics of a image in
a better way. It uses the concept of the
sliding window and the parameters, radius
and neighbors [7]. It is showed in Fig.
4.[10]
The most important details in this
text are the steps involved in converting a
frame into matrices of 3X3 pixels. If a
neighbor pixel in a matrix is greater than
the median pixel of that matrix, a binary
number is set and replaced with the median
pixel value. As the image is now converted
into LBP form, extract histograms from
each grid and concatenate to form a new
and larger histogram. These histograms
A. Preprocessing and Face Detection indicate the characteristics of the original
image and represent the facial image from
First, we convert the frame from color
the database. For the new image, it
to grayscale. To detect the faces, we used a
performs the same steps and gets a new
haar cascade classifier which is proposed
histogram for the image.
in [8] where a cascade function is trained
and detect features in other images. For C. Post processing
this, we use haar features like edge, line,
The most important details in this text
and four-rectangle.
are that the system compares the new
histogram with the histograms from the
For a large image or variable size of an training dataset and chooses the histogram
image, it takes a lot of computations and with the lowest confidence. If confidence is
features and most of them will be less than 50, details belong to the extracted ID
irrelevant. But AdaBoost manages to select are shown on the frame [11] and the names are
© 2023, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM23966 | Page 3
 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 06 | June - 2023 SJIF Rating: 8.176 ISSN: 2582-3930

updated into an excel sheet if the student’s list of defaulters is added to the attendance
name is not in the excel sheet. If confidence is report by creating a new column called
greater than 95, the person’s image is saved in "Defaulters" and marking each student
a separate folder. This helps in identifying any who has missed more than the threshold
intruders in the class and reduces the wrong number of classes as a defaulter.
classification of students to an unknown
person. IV. RESULT AND ANALYSIS
The Face Recognition rate of
students is 77% and its false-positive rate is
28%. Face Recognition of unknown
persons is 60% and its false-positive rate is
14% and 30% respectively refer to Table 1.
The threshold value only affects the false
positive rate of an unknown person. In the
existing system, when the person in the
video turns his head greater slightly then
confidence value for that frame may get
Table 1 greater than favorable filter value then the
person in the frame is considered as an
unknown person. In the proposed system,
if confidence is greater than 50 and 95 then
only a person is considered as an unknown
person and that person’s image is saved as
an unknown person.

VI. CONCLUSION
LBPH is one of the prominent techniques
for face recognition. Our system successfully
recognizes a student with unintentional changes
like wearing glasses or growing beard. Here the
problem is the dataset is small. In future, An effort
could be made to build a better dataset, that might
D. Calculation of Defaulters
practically give a more accurate result. We can
The defaulters' part of the attendance
report is responsible for identifying Improve haar cascade
students who have missed more than a classifiers through the synthesis of new training
certain number of classes and marking examples which can improve the recognition rate
them as "defaulters" in the attendance of unknown persons. A system alert (voice and
report. A threshold value is set and the visual) can be included if an intruder is detected in
attendance percentage for each week is the class.
calculated using the formula (number of
classes attended in the week / total number
of classes in the week) * 100. If the
attendance percentage for any week is
below the threshold value, that student is
marked as a defaulter for that week. The

© 2023, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM23966 | Page 4

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 06 | June - 2023 SJIF Rating: 8.176 ISSN: 2582-3930

© 2023, IJSREM | www.ijsrem.com DOI: 10.55041/IJSREM23966 | Page 5