REVIEW PAPER ON

FACE RECOGNITION ATTENDANCE

SYSTEM
Manas Juneja1 Snehal Moghe2
1
UG student, Department of Computer Science and Engineering, Medi-Caps University Rau, Indore,
manasjunejaa@gmail.com
2
Faculty, Department of Computer Science and Engineering, Medi-Caps University Rau, Indore,
snehalm.ind2022@gmail.com

Abstract - The use of face recognition software is Face recognition technology identifies an
increasingly common in organizations and individual by comparing a live capture to a
colleges for attendance management. This database of registered individuals. It is a biometric
technology enables automated and accurate method that is becoming a popular choice because
attendance marking, reducing the need for it is relatively easy to use when compared to other
manual intervention. The system captures an biometric options.
image of an employee or student's face and Machine learning is a powerful domain that
matches it to a database of registered individuals, provides desirable outcomes by applying various
marking the attendance of the individual as algorithms to a given dataset. Previously,
present. Face recognition technology offers educational institutions were very cautious when
benefits such as increased accuracy, speed, and tracking student attendance. However,
efficiency, as well as the elimination of fraudulent computerized attendance management systems that
attendance practices. However, there are privacy use face recognition can handle time, safety, and
and security concerns associated with this delegation issues more efficiently.
technology, such as data breaches, inaccurate or Haar Cascade is a complementary technology that
biased results, and lack of consent. To address is often used to analyse images in these systems.
these concerns, appropriate data protection and Attendance Management Systems that use Face
security measures must be implemented, and Recognition technology can save time, prevent
individuals' consent must be obtained. fraudulent attendance, and improve security.
Additionally, regular reviews and updates of However, it is essential to consider privacy and
technology and policies are necessary to ensure data security concerns when implementing this
compliance with relevant regulations and ethical technology. Appropriate measures should be taken
standards. to protect the collected data and ensure compliance
with relevant regulations and ethical standards.
Keywords - Time consumption, Automatic
The traditional technique of marking student
process, Image processing, Face detection, Face
attendance is often troublesome and disruptive to
recognition
the teaching process. The face recognition student
attendance system provides a simpler solution by
eliminating the need for calling out names or
I. INTRODUCTION checking identification cards. This system also
helps to prevent distractions for students during
Attendance tracking using traditional methods is exam sessions. In addition, attendance sheets
not an attractive option for institutes, and it often passed around in large classes can be difficult to
leads to inconsistencies. As a result, many manage. The proposed face recognition system is
institutions are turning to Attendance Management aimed at replacing the manual signing of
Systems that rely on Face Recognition technology. attendance, which can be burdensome and
These systems replace manual attendance taking by distracting for students. Moreover, this automated
teachers with an automated process. They are system can also help to prevent fraudulent
widely used by various administrations and attendance practices, and lecturers do not have to
educational institutions for tracking attendance. manually count the number of students present.
However, facial identification presents difficulties
such as distinguishing between known and In 2018, Kritika Shrivastava, Shweta Manda, and
unknown images, training process being slow and Prof. P.S. Chavan introduced an Automated
time-consuming, and the impact of different Attendance System based on Face Recognition and
lighting and head poses on the accuracy of the Gender Classification. The system employs the
system. Therefore, a real-time operating student Haar Cascade and LBPH algorithms along with the
attendance system that can identify students within LDA Model [6].
a defined time frame and maintain consistency in
facial features despite changing backgrounds, In 2019, Jenif D Souza proposed an Automated
illumination, pose, and expression is needed. The Attendance Marking and Management System
system's performance will be evaluated based on based on Facial Recognition technology. The
high accuracy and fast computation time. system automatically marks students' attendance
using a camera that captures their photo in the
classroom. To identify faces, the system uses the
Histogram algorithm, which converts the face
II. LITERATURE SURVEY image to matrix form and employs histograms for
In 2016, E Vardharajan, R Dharani, S Jeevitha, and recognition. This approach avoids the time-
SHemalata proposed an Automatic Attendance consuming process of manual attendance marking
[7]
Management System that incorporates Face .
Recognition technology. The system implements In 2019, Shreyak Sawhney, Karan Kicker, and
the Eigen Faces and Eigen Weight methods for Samyak Jain proposed a Real-Time Smart
detecting faces. When a student appears, the Attendance Management System that utilizes Face
camera captures an image and the system crops the Recognition techniques. However, the system uses
face and compares it to the student database to find two cameras - one for face detection and
a match. Once the student is identified, the system recognition at the entrance of the classroom, while
marks their attendance as present [1]. the other camera is used inside the classroom to
In 2017, Samuel John has designed an attendance check for proxy attendance [8].
management system that incorporates face In 2019, Nandhini R. proposed an Attendance
recognition technology and features GSM System that uses Face Recognition technology. The
Notification. The system employs the Viola Jones system captures a video of students, which is then
algorithm to detect faces and the Fisher faces divided into frames, and these frames are stored in
algorithm to create facial patterns. These patterns a database. The system uses Convolutional Neural
are then stored as templates in a database. To create Network (CNN) algorithm to detect faces, which
the graphical user interface, an SDK is used in leads to better accuracy and faster processing [9].
combination with the OpenCV library [2].
In 2017, Poornima S and Sripriya N proposed an
Attendance Management System that incorporates III. METHODOLOGY
Facial Recognition technology, audio output, and
gender classification. The system uses Principal Detecting faces in images is a complex task in
Component Analysis for face detection and computer vision as it involves identifying whether
recognition. Then, Microsoft Speech API is used to an image window contains a face or not. Faces
announce the names of absent students, providing a exhibit a wide range of variations in terms of age,
cross-check [3]. skin colour, facial expressions, and other factors.
Furthermore, factors such as lighting conditions,
In 2017, Prof. Arun Katara and Mr. Sudesh V. image quality, and occlusion can make the task
Kolhe presented an Attendance System that even more challenging. A robust face detection
incorporates Face Recognition technology and system should be able to detect faces accurately in
Class Monitoring. They introduced the utilization any lighting condition and under any background.
of Raspberry Pi with the OpenCV library installed
for both. The system employs a web camera To address this problem, researchers have
connected to the Raspberry Pi, which in turn is developed different approaches to face detection,
connected to a database [4]. including traditional computer vision techniques
and deep learning-based methods. Traditional
In 2018, Omkar Abdul Rhmansa Lim developed a approaches, such as Haar cascades, use hand-
Class Attendance Management System that uses crafted features like edges, corners, and lines to
Face Recognition technology. The system is based detect faces, while deep learning-based methods,
on the Raspberry Pi and captures images of such as Convolutional Neural Networks (CNNs),
students' faces when they face the camera [5]. can automatically learn features from raw image
data. Although traditional approaches are
computationally efficient and can work in real-
time, deep learning-based methods have When the user requests for recognition, the frontal
demonstrated superior performance and can detect face is extracted from the captured video frame,
faces with high accuracy even under challenging and the eigenvalues are recalculated for the test
conditions. face. The system then matches the test face with the
stored data to determine the closest neighbour.
However, even the most advanced face detection
systems may not be able to detect all faces under all We utilized several tools to develop the HFR
conditions, and there is still room for improvement. system, and their contributions were essential to the
Therefore, researchers continue to work on successful completion of the project. One of the
developing more robust and accurate face detection most crucial tools we used was the OpenCV
methods that can handle challenging scenarios. library, which is an open-source computer vision
library with the primary aim of providing an easy-
The face detection task can be divided into two to-use infrastructure for building sophisticated
steps: classification and face localization. In the vision applications quickly. The library comprises
first step, the system determines whether there are more than 500 functions covering numerous areas
any faces in the image, outputting a binary value of in vision. Our HFR system relied heavily on
"yes" or "no." In the second step, the system OpenCV's face recognition technology. To
localizes the faces within the image and outputs recognize a user's face, they stood in front of the
their location as a bounding box with coordinates camera, ensuring a minimum distance of 50cm and
(x, y, width, height). After capturing an image, the allowed their image to be captured. The frontal face
system will compare it with the images in its was then extracted from the image, converted to
database and provide the most relevant result. grayscale, and stored. We used the Principal
To implement this system, we will use the Component Analysis (PCA) algorithm on the
Raspbian operating system and the OpenCV images and saved the eigenvalues in an XML file.
platform, and we will write the code in Python When a user requested recognition, the frontal face
language. was extracted from the captured video frame via
the camera. The system then recalculated the
eigenvalue for the test face and matched it with the
stored data for the closest match.
OpenCV is a crucial tool we used to build our HFR
system. It is a powerful library that provides a wide
range of image processing functions. This cross-
platform library is available for free under the
open-source BSD license. OpenCV is particularly
useful because it provides several image processing
functions that produce the expected results without
requiring any coding. Some of the supported
functions include gradient and Laplacian
computing, contours delimitation, Hough
transforms for lines, segments, circles, and
geometrical shapes detection, histogram computing
and equalization, segmentation, filtering,
The proposed approach for implementing Face morphological operations, cascade detectors for
recognition involves the use of the open-source detection of face, eye, and car plates, interest points
computer vision library, OpenCV. OpenCV is detection and matching, video processing like
designed to provide a user-friendly infrastructure optical flow, background subtraction, and camshaft
for building advanced vision applications quickly. for object tracking, and photography like
With over 500 functions spanning various areas in panoramas realization, high definition imaging
vision, it is the primary technology behind Face (HDR), and image inpainting.
recognition.
There are several Integrated Development
To use the system, the user is required to stand in Environments (IDEs) available for Python,
front of the camera at a distance of at least 50cm, including PyCharm, Thonny, Ninja, and Spyder.
and their image is captured. The frontal face is then Among these, we chose to use Spyder, as it is more
extracted from the image, converted to grayscale, feature-rich than Ninja and is also free. While
and stored. Principal component analysis (PCA) Spyder is slightly heavier than Ninja, it is still
algorithm is applied to the images, and the resulting much lighter than PyCharm. Additionally, it is
eigenvalues are stored in an XML file. possible to run Spyder on the Raspberry Pi and
access the GUI on your PC through SSH-Y. We
installed Spyder using the following command line:
 sudo apt-get install spyder  Suitable for use in various machines such
as ATMs, medical machines, automatic
1.4GHz 64-bit quad-core processor, dual-band vending machines and industrial machines
wireless LAN, Bluetooth 4.2/BLE, faster Ethernet,
and Power-over-Ethernet support (with separate  Parameters can be adjusted including
PoE HAT) brightness, contrast, saturation, white
balance, gamma, definition and exposure.
Specifications:
a. Power Source
The Raspberry Pi 3 Model B+ is the latest revision
in the Raspberry Pi 3 line-up, with the following We used Mi 20000 mAH Power Bank for our
specifications: power source.
• Broadcom BCM2837B0, Cortex-A53 (ARMv8) IV. COMPARISON OF FACE RECOGNITION
64-bit SoC @ 1.4GHz TECHNIQUES
• 1GB LPDDR2 SDRAM There are various technical papers available on
Attendance Management Systems that use Face
• 2.4GHz and 5GHz IEEE 802.11.b/g/n/ac wireless Recognition technology. After studying some of
LAN, Bluetooth 4.2, BLE these papers, it was found that different methods
• Gigabit Ethernet over USB 2.0 (maximum and techniques were employed. The following are
throughput 300 Mbps) some of the techniques used for Face Recognition:

• Extended 40-pin GPIO header 1. Haar Cascade Classifier: This technique

uses Haar-like features to detect objects,
• Full-size HDMI including faces. Although effective in
detecting faces, it can be slow.
• 4 USB 2.0 ports
2. Local Binary Patterns Histograms
• CSI camera port for connecting a Raspberry Pi (LBPH): This technique is a texture-based
camera approach that compares the local image
• DSI display port for connecting a Raspberry Pi properties to recognize faces. It is
touchscreen display computationally efficient but can be
sensitive to lighting conditions.
• 4-pole stereo output and composite video port
3. Convolutional Neural Network (CNN):
• Micro SD port for loading your operating system This technique utilizes a deep learning
and storing data model that can be trained to accurately
recognize faces. It requires a large amount
• 5V/2.5A DC power input. of training data, but it can perform well
The ELP HD 8-megapixel USB CMOS board even in varying lighting conditions.
camera module is equipped with a Sony sensor 4. Principal Component Analysis (PCA):
(1/3.2") IMX179 and is compatible with various This technique transforms the face images
operating systems such as Linux, Windows, and into a lower-dimensional space where the
Android. This makes it a versatile option for use in variations between images are captured in
different types of equipment. principal components. It is fast and
Specifications (1/3.2-inch Sony IMX179 USB efficient, but it can suffer from lighting
webcam): and pose variations.

 8-megapixel resolution The selection of a specific technique depends on

the specific requirements and constraints of the
 Mjpeg USB camera that supports UVC for Attendance Management System. Each method has
Windows, Linux, Mac and Android its own strengths and weaknesses.
systems
 Compatible with Raspberry Pi, Ubuntu,
OpenCV, AMCap and other USB web V. RESULT ANALYSIS AND FUTURE SCOPE
camera software and hardware a. Face Detection:
 Equipped with a 2.8mm lens To start capturing images through the client-side
 38x38/32x32mm mini micro-USB board web camera:
camera
 1. Pre-process the captured image and extract Throughout the development process, the following
the facial image. results were achieved:
2. Calculate the eigen value of the captured • The system is user-friendly and can be
facial image. administered by a technician without an IT
background.
3. Compare the calculated eigen value with
the eigen values of existing faces in the • The system is fully developed and ready for
database. commercial use.
4. If the eigen value does not match with the • The system has the ability to recognize up to
existing ones, store the new facial image 1,000 faces.
information in the face database (xml file).
• The system can be scaled to serve any number of
5. If the eigen value matches with the people within an organization.
existing one, move on to the recognition
step. Class Participation can be one aspect where this
whole idea of face recognition can be applied.
The following steps were taken for face recognition Class Participation (CP) means that every student
using the PCA algorithm: who asks questions to Faculty in class or answers
questions of the Faculty, should be given marks out
1. Find the face information of the matched of 10/ A to E grade or 0 or negative if found using
face image from the database. mobile phone. Other conditions can also be
2. Update the log table with the applied, etc. So digitization of education basically
corresponding face image and system means that CP will be done via desktop / tablet
time, completing attendance for individual application, where Teaching Associate or Faculty
students. himself will grade students sitting in a particular
specific seating chartwise, location in the class. No
In this section, the experiment conducted to capture changes are allowed, unless given permission by
the face into a grayscale image of 50x50 pixels is PGP (Post Graduate Programme) Office. Hence,
presented along with its results. for every session CP will be marked and at the end,
tabular format of output will be generated in form
of MS Excel format or SQL or MS Access. This
mode of CP Marking is helpful in attendance as
Here are some potential areas for future
well. In the last run, grades of CP Marking will be
development of this project:
visible along with session-wise marking of every
• Enhancing security: This could involve adding subject subscribed for.
additional layers of authentication, such as multi-
Financials of this Project can be worked on per
factor authentication or biometric data.
lecture/per faculty wise or per semester wise flat
• Using neural networks for improved accuracy: rate across IITs/ IIMs.
Neural networks are a type of machine learning
algorithm that can be trained to recognize patterns
in data, including facial recognition. Implementing VI. REFERENCES
neural networks could potentially enhance the
accuracy of the system. [1] E Varadharajan, R Dharani, S. Jeevitha, B
Kavinmathi, S. Hemalatha “Automatic Attendance
• Scaling up for larger operations: This system Management system using face detection” at
could be used in larger factories or for employee ICGET 2016. 2020 Department of Information
attendance in corporate settings, with more Technology.
complex integration and management
requirements. [2] Kennedy O., EtinosaNomaOsaghae, Samuel J.,
Kalu-Anyah Grace, Imhade Okokpujie “A face
• Developing a fully web-based system: Currently, Recognition Attendance system with GSM
this system requires physical hardware and Notification” in IEEE NIGERICON 2017.
software installation. In the future, the system
could be moved entirely online, accessible through [3] Poornima, Sripriya, Vijayalakshmi B,
a web browser. This would necessitate additional Vishnupriya P “Attendance monitoring system
development to ensure security and accessibility. using facial recognition with audio output gender
classification” in ICCCSP 2017.
[4] Arun Katara, Sudesh V. Kolhe “Attendance
System Using Face Recognition and Class
Monitoring System” in IJRITCC 2017.
[5] Omar Abdul Rhman Salim, Rashidha
Olanrewaju, Wasiu Balogun “Class attendance
management system usingface recognition” in
ICCCE 2018
[6] Kritika Shrivastava, Shweta Manda, P.S
Chavan, “Conceptual model for proficient
automated attendance system based on face
recognition and gender classification using Haar-
cascade” in IJEAT 2018.
[7] Jenif D Souza, Jothi S, Chandrasekar,
“Automated Attendance Marking and Management
System by Facial Recognition using Histogram” in
ICACCS 2019.
[8] Shreyak Sawhney, Karan Kacker, Samayak
Jain, Shailendra Narayan, Rakesh Garg “Real Time
Smart Attendance system using face recognition
techniques” in international conference on cloud
computing data science and engineering 2019.
[9] Nandhini R, Duraimurugan N, S.P
Chollalingam “Face Recognition Attendance
System”in IJEAT in 2019.