See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/366005392

Face Recognition Attendance System based on Real- Time Video

Article in International Journal of Computer Applications · April 2022

DOI: 10.5120/ijca2022921983

CITATION READS

1 178

2 authors, including:

Eiti Jain
SAM College of Engineering Bhopal
4 PUBLICATIONS 3 CITATIONS

SEE PROFILE

All content following this page was uploaded by Eiti Jain on 05 December 2022.

The user has requested enhancement of the downloaded file.

 International Journal of Computer Applications (0975 – 8887)
Volume 184 – No.6, April 2022

Face Recognition Attendance System based on Real-

Time Video

Eiti Jain Santosh Mishra

M.Tech Scholar, Assistent Prof.,
RGPV Bhopal RGPV Bhopal

ABSTRACT Needless mistakes frequently occur whenever these attending

A person's face becomes their identification. Since the systems are designed. Using the present attendance tracking
invention of computer vision applications, the strategies for system for example, the study discovered it has a failure rate
exploiting this physical property have changed dramatically. c of around 5%, and that there'll be a phenomenon where palm
Every school, and collage library keeps track of attendance. prints could be hit, which has a significant impact on
Traditional approach for attendance is professor calls student attendance effectiveness, especially for large participation
name & record attendance. It takes some time to record sites where congestion is more likely. Workers swiped card
attendance. Suppose duration of class of one subject is about with someone else is a problem with the cards attendance
50 minutes & to record attendance takes 5 to 10 minutes. It is monitoring system, making it impossible to meet the goal of
a waste of time for every lecture. To eliminate these losses, real-time attending. Face recognition provides greater
we're going to apply an automated picture processing accuracy and reliability than other two attendance methods as
procedure. Face detection and recognition systems are used in there are more locations for facial recognition software, which
this unique technique. Face detection distinguishes face from would be more exact than other methods. It is tough to
non-faces, which is necessary for correct attendance. The congeal now that the situation has much improved. Despite
alternative method is to use face recognition to track a the fact that China's studies into biometric technology was
student's attendance. By the proposed method ( HAAR, delayed, our scientific experts have caught up, and several key
AdaBoost and Deep Neural Network) we are getting an players in the area have created their own industrial positions.
accuracy of 99.41 percent on the automatic face detection With the introduction of the big data age of today's world as
system. well as the economic value of biometric technology, the future
of this research and technology is quite bright, and there is a
Keywords lot of market demand for it. Faces in surveillance recordings
Video Processing, Face Recognition Technology, Face are frequently subjected to severe picture blur, posture shifts,
Recognition Attendance, Attendance System, Video and obstruction. Ding C has suggested a complete framework
Recognition. based on deep learning to tackle the issues of video-based
facial recognition software (VFR) (CNN). To compensate for
1. INTRODUCTION the lack of genuine instructional videos data, Ding C
A technique based just on Viola Jones technique with face deliberately blur the training data made of clear still photos in
detection, following by the equalisation histogram for order to build a fuzzy and strong feature representation. CNN
extracting features, as well as the (SVM) classifiers for facial is encouraged to automatically discover fuzzy insensitive data
recognition software was presented for developing an utilizing data for training made up of still photos and
automation process for human identification and attending generated fuzzy data. Secondly, CNN has presented a trunks
accountancy. developed an automated systems for face branching CNN model (TBE-CNN), that extracts
detection using the same method, as well as a classification complementary from the entire face picture and patched all
for facial recognition software. These approaches, on the other around face sections Data, in order to improve the resilience
hand, were based on traditional machines algorithms. A of CNN characteristics to posture shifts and occlusion. The
current FaceTime technique employs a cascaded deep neural beginning of the 20th century face identification problem has
network with face detection as well as a convolutional again been investigated by researchers such as V B Nemirovskiy.
for production of facial characteristics, which are The description is based on grouping the closeness of the
subsequently utilised for recognition. Face identification using chromatic distributions of the split image's brightness clusters.
multi-task coalesced convolutions (MTCNN) and a Siamese Nemirovskiy V B employs three sorts of distances to
neural network to capture facial features into a 128-element determine closeness: There are three types of distances:
vector. Most commercially automation software utilizes close- cosine, Euclidean,and Leibler. For picture segmentation and
up photographs of faces, with an open, well-lit face being the grouping of proximity measures, a recursive neural network
most important criterion for effective operation. Image programming model is utilised. Face recognition attendance
process is done in steps, which is a time-consuming management system with actual video computation is
procedure. Face detection and recognition software favourable to business development and will have a good
application have rapidly grown as an information security impact on future business development. The goal of this paper
technology throughout the world . a recent, specifically now, is to create a real-time video processing-based face
when acts of terrorism are on the rise, this technique has recognition timesheet system. Four inquiry experiments were
gotten a lot of attention. In the fields of public safety, civic performed as part of this project: the prediction accuracy of a
economy, and home media, face identification system has a face detection system during exact check-in; the consistency
wide range of applications. The availability of staff must be of the face recognition timesheet structure utilising actual
recorded in the pipelines of normal businesses, that has video preparation; study of a skip rate of the face detection
become a fundamental obligation of the organisation. and recognition attendance system using actual video

12
 International Journal of Computer Applications (0975 – 8887)
Volume 184 – No.6, April 2022

production; face detection and recognition attendance tracking to a know one based on the similarity of recognised attributes.
interface setup using real-time video sorting The experimental Recognition software has a wide range of uses, including
findings show that attendance management system delivers intelligent safety and security systems, biometrics,
the predicted timing & attended outcomes using facial safeguarding, verification, and attendee tracking, Machine
recognition and a computer, demonstrating that the entire vision, for example. Biometric identity methods are being
methodology is feasible. Students who completed the created dynamically because traditional methods of
attending sign-in system accomplished their responsibilities identification, such like keys or passwords, do not give a high
swiftly, got rid of the confusing roll call sign, and rapidly level of dependability. The face has a lot of benefits over
grasped the sign of operations and function. Future other approaches for determining a personality:
technology timing and form of attendance tracking
conversions have made significant advancements, • The individual does not have to be physically approached,
dramatically enhancing the rates of attending and the and it's the most common method for bulk applications.
dependability of face recognition technology. It is worth of • There is no need for intricate or spend at least.
your researchers' further investigation and realisation.
Any organisation that wants to analyze its members' success
2. METHODOLOGY
Nowadays days, technology attempts to convey a large
in part dependent on attendees must keep track of their
amount of knowledge-based technical advancements. Deep
attendance. Automated tracking that is exact and time and cost
Learning is an intriguing subject that allows a machine to
efficient is thus a pressing demand for today's large
educate itself using data as inputs and then deliver an
companies and enterprises. The process of face recognition
acceptable output throughout testing using various learning
may be characterised as the comparison of a still digital
techniques. Nowadays, attendance is seen as a critical aspect
picture still or frames of a video produced from a video file to
for both students and teachers in educational institutions. With
face images or attributes of existing facial images with the
the progress of deep learning technology, the computer can
aim of verifying or identifying an individual. This is
now automatically recognise the kids' attending performance
accomplished by comparing facial traits from a picture to
and keep records of it. In generally, a student's attendance
faces in a collection. Because of its common use in security
management system can be kept in two different ways,
devices, the method may be likened to other finger print or
namely,
eye iris recognizer. It is preferred because it is contactless and
non-invasive, despite the fact that it is less accurate than iris
 Manual Attendance System (MAS)
and fingerprints identification. This innovative, yet divisive
technology offers a wide range of applications, including  Automated Attendance System (AAS).
improved human-computer interaction, video surveillance,
automated picture indexing, and video databases, among Manual Student Attendance Management is a method in
others. Figure 1 depicts the entire process in a diagram to help which a teacher in charge of a specific topic must manually
visualise the series of events. called the students' names and record their attendance.
Manually attention may be viewed as a time-consuming
process, and it is possible that the instructor will miss
someone, or that pupils will respond to the absences of their
friends many times. As a result, the issue emerges when we
consider the usual method of collecting class attendance. We
use an Automated Attendance System to address all of these
difficulties (AAS).
Nowadays days, technology attempts to convey a large
Fig.1 Systematic flow of generic face recognition
amount of knowledge-based technical advancements. Deep
Monitoring attendance of students in each session is a Learning is an intriguing subject that allows a machine to
moment task for university professors, especially when educate itself using datasets as input and then deliver an
courses are big. Facial recognition is a method of recognising acceptable output during testing using various learning
or verifying a person's identity in photographs, films, or in techniques. Nowadays, attendance is seen as a critical aspect
live time. The attending system can track employee for both students and teachers in educational institutions. With
whereabouts and date of checking in/out, as well as the results the progress of artificial intelligence technology, the computer
will be recorded down. The attendance system will next can now automatically recognise the kids' punctuality record
process the information and provide timesheet records. and keep a record of it. In generally, a student's attendance
Students face a tremendous hardship if punctuality is tracked management system can be kept in one of two ways: Manual
by hand. To solve this issue, a smart and automatic attendance Attendance System (MAS) or Automatic Attendance
system is being built. One of the technologies for detecting Management (AAS).
people and monitoring their attendance is facial images. To
use this technique, the problem of proxy and student been Input is passed through numerous "layer" of neural network
labelled current even if they're not physically there may be models in deep learning models, each of which provides a
readily solved. The gadget would therefore save not just time, reduced version of the data to the next layer.
but also the dedication that professors were supposed to bring
to each lecture.
The majority of algorithms perform effectively on datasets of
Face identification system, which is used to autonomously just few thousand characteristics or columns. Unfortunately,
locate faces on video or identify people, is gaining popularity. an unorganized dataset, including one derived from a picture,
The definition of "recognition" is the attribution of the has so many characteristics that this approach becomes
examined item to one of recognised classes. "Identifying" is inefficient or impossible. Machine learning algorithms can't
the process of determining the identification of a known thing handle 2.4 million parameters in a single 800-by-1000-pixel
RGB colour picture.

13
 International Journal of Computer Applications (0975 – 8887)
Volume 184 – No.6, April 2022

Fig.2 Face Recognition Attendance System Based

Even as image passes thru every neural network layer, deep
learning methods learn more about it. Early layers learn to
recognise low-level elements like edges, while later layers
integrate those features towards a more holistic picture. A
intermediate layer, for example, may image is determined to
detect sections of an item in a shot, including a leg or even a
branch, but a deep layer would detect the whole thing,
including a dog or even a tree.

Principal Component Analysis (PCA)-Principal Fig. .3. Proposed Methodology

Components, or PCA, is a dimensional space approach for Face detection is currently utilised in a variety of areas,
reducing the dimension of big data sets by converting a large particularly on image-hosting services such as Picasa,
collection of factors into a small one that retains the majority Facebook , and Photobucket. The automatically tagging tool
of the data in the huge array. provides a new level to sharing photos among the individuals
in the photo, as well as giving other people an indication of
Naturally, lowering the amount of factors in a data set reduces who the person in the photo is.
accurate; nevertheless, the answer to dimension reduction is to
exchange some accuracy for simplicity. Since small sets of In the AdaBoost approach, wavelet features are used to
data are simpler to study and display, and also because characterise both face and non-facial pictures. In the Eigen
machine learning techniques can analyse data more easily and Face technique, the feature vector of a face and nonface
quickly without having to deal with superfluous factors. picture is generated using (PCA). PCA is also used to
compress the provided data vector. The position of a face or a
To summarise, PCA's goal is to decrease the number of face region with facial characteristics is determined using a
variables in a data collection while keeping as much face detection algorithm (eye, nose, lips etc). In social
information as possible. situations, the face is our primary centre of attention, and it
plays a crucial role in transmitting identity.
3. PROPOSED METHOD
Just on PYTHON Graphical Interface, the proposed technique
is used to detect the face from a video cameras recording
frame (GUI). The Viola-Jones approach is credited with being
the first object detection method to provide real-time tracking.
Furthermore, because it is quick and adaptable in establishing
a platform with a high detection, this approach is suitable for
real-time application.

Attentional Cascade, Image integral, feature extraction,

classifier with Image integral, Structure, and AdaBoost are
three fundamental principles in the Viola-Jones Face
Detection system that enable for the construction of excellent
facial recognition that can be used in real-time applications.

Fig 4: Face Detection System

The supplied information vector is compressed using PCA. In
DNN, the kernel function is used to characterise face and non-
face pictures. The pattern of the human face provides
important information concerning human face recognition.

14
 International Journal of Computer Applications (0975 – 8887)
Volume 184 – No.6, April 2022

Windows scanners or a divided region can be used to apply 1 N

template matching. c
n
   T  AAT , ( MN  MN )
i 1 i i

Step 1: Prepare the training faces. .......................(5)

A  [1 2..... n],( MN  n) .....................................(6)

Obtain the face images I1, I2, I3, I4, ......, In (training faces).
The representations of the face must be centered and of the
same proportions.
where A is the matrix of the concatenation of the column
Step 2: Set up a data set. vectors after removing the mean face.

Each face image in the database is converted into a vector and Step 6: Obtain the eigenvectors and eigenvalues.
located in a training set S.
𝑢𝑖 = 𝐴𝑣𝑖 , 𝑖 = 1,2,3, … . 𝑛 – 1.............................(7)
S = {Γ1, Γ2, Γ3, … … Γ𝑛 }.............(1)

Dimension Reduced Matrix

𝑢𝑖 is the eigenvector of AAT whereas 𝑣𝑖 is eigenvector of
AT A . Eigen values
T
of A A , are calculated and sorted.
Eigen values less than 1 are eliminated so the number of non-
zero eigenvectors may be less than (n-1). Eigen face is the
principal component distribution of facial image.

Step 7: Projection of facial image to Eigen face.

𝛺𝑖 = 𝑈𝑇(𝛤𝑖 − 𝜑) , 𝑖 = 1,2, … , 𝑛 − 1 (9) In order to produce

the projected image Ωi, the facial expression was projected on
the Eigen Face using the equation. Γ𝑖−𝜑 as the centered
Step 3: Calculate the average vector of the face.
vector which the mean face is removed. Steps 1 to 7 are used
to train the training image set and only step 1,2,3,4 and 7 is
The average face vector (𝜑) is calculated based on the
required for the test image. Step 5 and step 6 are not required
formula:
for the test image, as the Eigen face is only required to be
computed once during training. The Mahalanobis Distance is
1 N
  i
n i 1
used as distance classifier to calculate the shortest distance
between the projected image and projected test image for
...................(2) recognition.

3.1 HAAR
Haar characteristics are made up of two or three rectangle.
Candidates' faces are scanned and examined for Haar
characteristics relevant to the current stage. The weight are
constants that the learning method generates. As seen in Fig.
............(3)
4.4, there are many different types of characteristics.
Step 4: Subtract the mean/average face vector.

The original faces are subtracted with the average face vector
and the output result is stored in the variable 𝛷𝑖

𝛷𝑖 = 𝛤𝑖 − 𝜑𝑖 = 1,2, … , 𝑛......................(4)

Fig. 5. Examples of Haar features

Areas of white and black regions are multiplied by their
respective weights and then summed in order to get the Haar
feature value.Each Haar element has a values that is
determined by multiplying the size of each rectangle by their
corresponding weights, then adding the results together. To
use the integral picture, you can quickly calculate the area of
Step 5: Calculate the covariance matrix. each rectangle. Using the integral picture, you can find the
total of all the pixels above and to the left of any corners of a
The covariance matrix is used to calculate the eigenvectors
rectangular. Because L1 has been deducted twice, it must be
and eigenvalues. However, 𝐴𝐴𝑇 have an 𝑀𝑁 × 𝑀𝑁 put back in to acquire the right rectangle area. Using the
dimension that is extremely large to calculated. AAT and positions of the integral picture, the surface of the rectangles
T R, denoted as the rectangle integral, may be determined as
A A have the same eigenvalues and their eigenvectors can follows: L4-L3-L2+L1.
be related as 𝑢𝑖 = 𝐴𝑣𝑖.

15
 International Journal of Computer Applications (0975 – 8887)
Volume 184 – No.6, April 2022

3.2 AdaBoost compares the picture of the recorded face to the photographs
in the face database. Finally, the database was updated with
Face detection is a computer algorithm that detects the the attendance information of identified attendees as well as
location, size, and posture of a person's face in a photo or the time collected during detection. At the exit door, the same
video sequences. Face detection is a hot topic in the computer procedure is followed, with two separate timings being
vision world; finding a human face in an image is critical for recorded for the final computation of engaged time for each
applications including face identification, video surveillance, attendance.
HCI, database administration, and searching picture
databases. Viola and Jones created the most successful face
identification method based on the Adaboost classifiers.
Although their system proved successful in detecting faces, it
has issues with false alarms, which may become more
difficult in the existence of a complex backdrop. False
positives can cause issues in an applications and require
further postprocessing to eliminate. We offer a probabilistic
strategy to adjust the scoring system of the Adaboost
algorithm as our contribution to reducing the amount of false
positives, which involves the extension of important concepts
and supported experimental findings over the previous
versions.

The Adaboost method is adaptive in the sense that it boosts

misclassified data from prior classifiers during training by Fig6: Face Recognition initialize
giving it a higher weight than correctly classified data. The
incoming data collection and accompanying categorization
labels make up the training database. Over the training data
set, Adaboost continually calls a weaker learning ability. At
each level, the most optimum parameters of weak learning
algorithms are determined, resulting in the lowest
classification error. A best classifier is a weak training
classification with optimum parameter for a specific training
stage. The intake data set is originally weighted evenly; but,
throughout the training process, the weaker training algorithm
emphasises the misclassified data more than the perfectly
classified material. This is performed by increasing the
weights of the miscategorized data in relation to the properly
classified data at each iteration. The following are the major
phases of the Adaboost method for efficiently classifying
data. In transmembrane identification, a neural network is a Fig. 7 : Initialize Recognition Window
widely utilised approach. Its concept is to construct a
hierarchical system out of a huge number of simple When Rashmi clicked the photo for the first time, the photo of
calculating units. Each basic unit can only perform quick rashmi was not being mentioned in her camera, after that
arithmetic, but a system made up of components in complex when she took the second photo, her photo appeared in it.
configurations might provide a difficult problem. The neural
network technique has also obtained good results in facial
recognition. BP network, self-organizing connections,
convolution systems, and other regularly utilised networks are
examples. There's enough training examples to potentially
recognise all faces if the networks is large enough. While
neural networks offer certain benefits when it comes to
recognize faces, they have also significant flaws.

4. RESULTS
Captions The overall real-time human detection and face
recognition framework consist of video input then captured,
face detection, face value detection, face recognition and
attendance record. The system's framework is depicted in
Figure 6. The system begins with a picture obtained from the
entry video camera. Two cameras were employed, one for Fig. 8 Face Recognition
entry records and the other for exit records, one at the This photo of Rashmi is visible in the figure 8 picture and
entrance door and the other at the departure door. Attendees' through that photo we can count the presence of Rashmi and
faces were photographed and sent into the system for face no one can cheat in it because first the camera will recognize
detection. The technology will detect the face once it has been the face, only then the presence of Rashmi will be there. , if
spotted. First, take an image from a camera and detect a face, the face camera does not recognize the face, then at that time
then use a har and deep learning algorithm to categorise the there will be absence of Rashmi.
object of the face and gather face value features, then compare
with the existing data base. The face recognition module

16
 International Journal of Computer Applications (0975 – 8887)
Volume 184 – No.6, April 2022

2. Mohd Suhairi Real-time mask 73.3 %

Md Suhaimin detection
Video Image
3. Hao Yang Recognition System 96%
Shubhobrata An automatic 85%
4. Bhattacharya attendance
management system
5. HAAR,AdaBoost
Proposed and Deep Neural 100%
Network
In the above given comparison table, author Qiuyan Li
searched on Face Recognition API, whose accuracy is 94
Fig. 9 Face Mark Attendance percent. Mohd Suhairi Md Suhaimin did a search on real-time
mask detection, which has an accuracy of 73.3 percent. Hao
The present work consists of photographs of four different Yang did research on Video Image Recognition System,
photographs in a collection of faces for the evaluation process. whose accuracy is 96 percent. Shubhobrata Bhattacharya did
Model transitions such as random crop, random transition, and research on an automatic attendance management system,
face orientation are added to create different images of the whose accuracy 85%. By propoposed method we get 99.41%
same face to obtain more facial features. Some facial features accuracy.
can be obtained from a small dataset using the above image
transformations. "In this part, the task is proposed in two
classroom sessions and is referred to as "Testing Session 1"
and "Testing Session-2". Test session -2 consists of a 100% Accuracy
component of the test dataset, with different faces, visibility
and lighting. The detected and detected face is shown in
Figure (1).
0%
This is exactly the situation where the participants for such
lectures are accordingly identified and the CSV file is Accuracy
generated and sent to the faculty concerned. Attendees are
labeled with the date and time of lecture for eti found in a
column next to their name. In the last tab, average attendance
is estimated, as photographs are revised in additional lecture
sessions. The Rashmi names identified from the proposed
model are then saved in a csv file and sent to the respective
professors, as was done in test Session-1.
Fig. 10 Accuracy Graph
Accuracy : You should read about True positive and True
negative, false positive and negatives. With this formula of
5. CONCLUSION
Dissertations presented come with some great features. As the
your
system used Haar cascade model for feature extraction. This
model works well for feature extraction as it uses integral
Accuracy=(TP+TN)/(Total).
pictures and a boosting technique that speeds up the detection
method and aids in the selection of the best features. So that,
face recognition accuracy cab be measured according to the
when the machine is being trained no other extra things come
percentage of the recognized faces per the total number of
into the frame. With LBPH the Haar Cascade classifier can
tested faces of the same person.
recognize faces from any angle, even a student is not looking
Accuracy= (Recognized faces(TP)/ Total number of tested at the camera and if the head is tilted. The system can
faces(TF))x100 recognize faces when the student has glasses. It can also take
attendance of multiple students at the same time. The system
Total Faces(TF) Proposed Method is quite complex. After using the proposed model, we get
100% percent accuracy, for an inexperienced person taking
True Positive (TP) 50 50 attendance will not be so easy. The main disadvantage is that,
because it is a face recognition-based attendance system, the
False Positive 50 0 dataset must be updated frequently, and the system will go
blind if there is no power supply. Attendance may be
False Negative 50 3 interrupted due to technical fault.

Detection Accuracy Rate 100% 6. REFERENCES

[1] Vignesh S, Sriram A, Venkatesan G, Usha A “Face Mask
Table No-5.1 Comparison of methods and accuracy Attendance System Based On Image Recognition”
[2021].
S.No Author Method Accuracy
[2] Shubham Kadam, Sagar Khedkar “Review Paper on
1. Qiuyan Li Face Recognition 94 % Contactless Attendance System based on Face
API Recognition”[2021].

17
 International Journal of Computer Applications (0975 – 8887)
Volume 184 – No.6, April 2022

[3] Qiuyan Li “Research on the Application of English Monitoring System (SAMS): A Face Recognition based
Course Online Examination System Based on Face Attendance System for Classroom Environment” [2018].
Recognition Technology” [2021].
[12] Yuslinda Wati Mohamad Yusof, Yuslinda Wati
[4] Mohd Suhairi Md Suhaimin , Mohd Hanafi Ahmad Mohamad Yusof, Muhammad Asyraf Mohd Nasir, Kama
Hijazi , Chung Seng Kheau , Chin Kim On “Real-time Azura Othman, Saiful Izwan Suliman, Shahrani
mask detection and face recognition using eigenfaces and Shahbudin, Roslina Mohamad” Real-Time Internet
local binary pattern histogram for attendance system” Based Attendance Using Face Recognition System”
[2021]. [2018].
[5] Hao Yang And Xiaofeng Han “Face Recognition [13] Sakshi Patel, Prateek Kumar , Shelesh Garg, Ravi Kumar
Attendance System Based on Real-Time Video “Face Recognition based smart attendance system using
Processing” [2020]. IOT” [2018].
[6] Lin Zhi-heng, Li Yong-zhen “Design and [14] Smit Hapani, Nikhil Parakhiya,Nandana Prabhu, Mayur
Implementation of Classroom Attendance System Based Paghdal “Automated Attendance System using Image
on Video Face Recognition” [2019]. Processing” [2018].
[7] Edy Winarno, Imam Husni Al Amin, Herny [15] Huda Mady, Shadi M. S. Hilles “Face recognition and
Februariyanti “Attendance System Based on Face detection using Random forest and combination of LBP
Recognition System Using CNN-PCA Method and Real- and HOG features” [2018].
time Camera” [2019].
[16] Borra Surekha, Kanchan Jayant Nazare, S. Viswanadha
[8] Shreyak Sawhney, Karan Kacker ,Samyak Jain, Raju and Nilanjan Dey “Attendance Recording System
Shailendra Narayan Singh , Rakesh Garg “Real-Time Using Partial Face Recognition Algorithm” [2017].
Smart Attendance System using Face Recognition
Techniques” [2019]. [17] Akshara Jadhav, Akshay Jadhav Tushar Ladhe, Krishna
Yeolekar “Automated Attendance System Using Face
[9] Mayank Kumar Rusia, Dushyant Kumar Singh, Mohd. Recognition” [2017].
Aquib Ansari “Human Face Identification using LBP and
Haarlike Features for Real Time Attendance Monitoring” [18] Marko Arsenovic, Srdjan Sladojevic, Andras Anderla,
[2019]. Darko Stefanovic “FaceTime – Deep Learning Based
Face Recognition Attendance System” [2017].
[10] L. Deepshikha1 , Dr.Ch.Venkateswara Rao2 , G. Uday
Kumar “Automated Attendance Management System [19] Prof. Arun Katara, Mr. Sudesh V. Kolhe, Mr. Amar P.
Based on Face Recognition Algorithms” [2018]. Zilpe, Mr. Nikhil D. Bhele , Mr. Chetan J. Bele
“Attendance System Using Face Recognition and Class
[11] Shubhobrata Bhattacharya, Gowtham Sandeep Nainala, Monitoring System” [2017].
Prosenjit Das and Aurobinda Routray “Smart Attendance

IJCATM : www.ijcaonline.org 18

View publication stats