Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)

IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

Smart Attendance Tracking System using Various

Machine Learning Algorithms
Saravanan.M.S Faiyaz Ahmad R.T.Thivya Lakshmi,
Professor, Department of Computer Assistant Professor, Asst.Professor, Department of CSE, Vel
Science and Engineering, Saveetha Department of Computer Engineering, Tech Rangarajan Dr.Sagunthala R&D
2024 Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI) | 979-8-3315-4066-1/24/$31.00 ©2024 IEEE | DOI: 10.1109/ICOICI62503.2024.10696638

School of Engineering, Jamia Millia Islamia, Institute of Science and Technology,

Saveetha Institute of Medical and New Delhi, Pin-110025, India Chennai - 600062.
Technical Sciences, [email protected] [email protected]
Chennai, Tamilnadu, India.
[email protected]

Abstract - The major goal of this study is to predict the verification purpose with or without minor changes [4] [5]. The
effectiveness of maintaining day-to-day activities of attendant data suggested method has an average recognition accuracy of 92
by using a unique Support Vector Machines (SVM) method to percent. Daily attendance is frequently tracked using this
monitor the faces. Materials and Methods: Novel SVM algorithm approach, reducing the chance of human mistake [6] [7]. Using
with sample size = 42 and Convolutional Neural Networks (CNN)
with sample size = 42 were validated many times to forecast
real-time facial recognition with the help of the Convolution
efficiency percentage. Novel SVM algorithm makes use of face Neural Network, this method is frequently used to take
vectors to recognize countenance (face) data. Results and attendance at institutions or offices. [6]. In this approach, two
Discussions: The recognition of countenance data by monitoring different methods are proposed as Support Vector Machine and
faces performs more efficiently using SVM (Support Vector Convolutional Neural Networks.
Machine) algorithm with mean efficiency 92.65% and comparing
Convolutional Neural Networks with mean efficiency 79.02%. The CNN is used for face detection and SVM is used for
Support Vector Machine algorithm performed significantly better extracting features [8]. This research focused on introducing a
than the Convolutional Neural Networks algorithm is 0.001 smart attendance system supported for face recognition which
(p<0.005). Conclusion: Novel SVM algorithm performs more
efficiently than the CNN algorithm to predict correct attendance
will be used to take attendance of all students in a classroom [9]
tracking using countenance images. at the same time. The suggested technique differs from similar
systems by proposing a novel approach that use image
Keywords - Novel Support Vector Machine, Convolutional Neural processing to take attendance and a face recognition algorithm
Network, Camera, Countenance Recognition, Attendance. to identify facial landmarks for efficient outcomes [10] [11].
We have collaborated with various authors across our
I. INTRODUCTION institutions that has made us complete the project with ease and
accuracy [12] [13]. The drawbacks of the existing systems are
The major goal of this study is to increase the rate of lack of accuracy while implementing and monitoring the faces,
efficiency in order to improve the speed of attendance register without predefined data it won't do anything, then few problems
automatically by applying various machine learning algorithms arises and this was giving more false positives, errors and
such as support vector machine and convolutional neural wrongly recognizing the faces as it is very sensitive to the light.
network to recognize faces. As of now many methods are lack of efficiency while implementing and experimenting the
introduced like biometric attendance and traditional ways of method [14] [15]. The goal of this study is to boost the
marking attendance by calling names one by one [1] which efficiency of personage attendant surveillance.
consumes a long period of time. This proposed research helps
to reduce the manual work, time and stress of faculties. It II. METHODOLOGY
monitors the personage attendant system based on countenance
data that also reduces the false positive images and proxies. The This research work carried out in our machine learning and
countenance achieved through the proposed system is to mark cyber security laboratory at Saveetha School of Engineering,
attendance by recognizing the faces and storing the data into Saveetha Institute of Medical and Technical Sciences
excel sheets [2]. Then it allows new users to register their faces (SIMATS). Implemented personage attendant monitoring
on the spot while marking the attendance using SVM algorithm. system [16] based on countenance data with sample size 84
Novel face recognition personage attendant system carried using IBM SPSS analysis. It is divided into 2 groups. Group 1
out by the researchers to promote the business are around 116 was taken as SVM algorithm and group 2 was taken as CNN
articles are published google scholar and IEEE Xplore. The face algorithm [17].The novel SVM algorithm is applied on training
recognition-based attendance system using cameras is data or live capture of input images from the camera are divided
extremely effective and efficient for further improvement of into frames at the rate of 30 frames per second by recognizing
accuracy and user data. It has high accurate processing in order both front face and side face then mark the attendance in real
to produce a system that is reliable and powerful to spot human time. Compared the efficiency of SVM algorithm with CNN
faces in real-time [3]. The technique to picture augmentation algorithm and achieved lesser mean value.
for face recognition tasks is presented as a model; it might be A. CNN Algorithm
included into another system as a supporting element for

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1108

Authorized licensed use limited to: Saveetha Engineering College. Downloaded on January 08,2025 at 04:18:38 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)
IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

The Convolutional Neural Networks (CNN) technique is

used to successfully locate normalized human faces in a
fraction of the time required by earlier approaches. The
technique begins by detecting heads on colour images utilising
colour and structural abnormalities in the background's face.
The figure 1 has the architecture diagram to understand the
facial detection process using various machine learning
algorithms for attendant systems All faces should be turned into
a comparable size and position by normalizing the space and
position of the reference points. Tests on a common database
indicate that the method is dependable and works at any time.
The CNN algorithm is given below.
Fig. 1. Architecture of face recognition and detection using facial data using
Step 1: Start the program machine learning algorithms
Step 2: Take argument as video dataset using OpenCV library
▪ It works well with a visible separation margin; it is
Step 3: Start while loop for frames one by one
effective in high-dimensional spaces; and it is
a) Extract region of interest for each frame and
effective when the number of dimensions exceeds the
declare var limit
number of samples.
b) Compute CNN features
▪ It is memory economical since it employs a subset of
c) Classify regions (persons)
coaching points (called support vectors) within the
d) Calculate distance between coordinates.
judgement function.
e) If distance > limit display warning
f) Output if there are any frames let Computer hardware specs are detailed descriptions of the
g) Else break loop components and capabilities of a computer. Hard disk drive of
Step 4: Release the video 1TB, main memory-4GB RAM, processing speed of 600 MHZ,
Step 5: Destroy all windows opened Processor Pentium IV, RAM of capacity 4 GB, Processor with
Step 6: End the process minimum space of 2.5GHZ. Computer software specifications
used requirements are Operating System – Windows (7/8/9)
B. SVM Algorithm
tool - PyCharm and language - Python
Support Vector Machine (SVM) algorithms are a set of
supervised learning methods used for classification, regression, C. Statistical Analysis
and outliers. These are the data points that are closest to the
IBM SPSS used for statistical analysis. The independent
decision boundary. They are critical in defining the position and
variable is pixel length and the dependent variable is color of
orientation of the hyperplane. In SVM, the decision boundary
the bounding box. The independent T-Test analysis is
between classes is a hyperplane. For two-dimensional data, this
performed.
is a line; for three-dimensional data, it’s a plane, and for higher
dimensions, it’s called a hyperplane. It is trained with a set of
III. RESULTS & DISCUSSION
knowledge that has already been categorized and is developing
Table 1 presents the dataset detailing various groups of
the model. As a result, SVM is classified as a non-binary linear
individuals, along with the detection outcomes and locations
classifier. SVM is one of the finest algorithms for employing a
using a CNN algorithm. Table 2 illustrates the simulated
face recognition technology to track attendance. It will use a
efficiency analysis of a novel SVM algorithm. Table 3 provides
line to divide data that is linearly ordered, but it will not be able
a group statistical analysis, showing mean values of 92.65%
to draw one line for non-linear data. The following algorithms
and 79.02%, with standard deviations of 2.407 and 3.083 for
has a step-by-step explanation of the novel SVM method. The
the novel algorithms, respectively. Table 4 displays the results
SVM algorithms is given below.
of an independent T-test analysis for both groups, revealing
Step 1: Start
two-tailed significance values of p<0.001.
Step 2: Adding a new image in the program for face
recognition TABLE I. INPUT DATASET SAMPLES WITH VARIOUS SIZES AND RELATED
Step 3: The SVM algorithm used to recognizes the new ACCURACIES FOR CNN MACHINE LEARNING ALGORITHM
images
Step 4: Then it compares this histogram with the existing
ones.
Step 5: Finally, it finds the best match and returns the label
person associated with this best match
Step 6: Then, mark attendance and store data
Step 7: Go to step 2 and Repeat up to step 6
Step 8: End the process

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1109

Authorized licensed use limited to: Saveetha Engineering College. Downloaded on January 08,2025 at 04:18:38 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)
IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

S.No Data Set size Accuracy Levene’s Test for

Equality of T-test for Equality of Means
Variance
1 1652 74.96
Std. 95% Conf.
2 1765 75.36 Mean Error Diff.
F Sig t df Sig(2- Diff. Diff.
tailed)
3 1874 76.18
Low Upp
er er
4 1986 77.36
Accur 18.44 <.001 2.892 18 .001 4.572 4.4 2.84 9.38
acy 60 3 6
5 2048 78.48 2.904 16.7 .000 7.738
6 4.4 3.72 8.44
71 1 3
6 2153 79.64

7 2249 80.24

8 2358 81.41 Figure 2 shows a bar graph comparing the SVM and CNN
algorithms' mean accuracy. SVM algorithm is 12.2% more
9 2485 82.65 accurate than CNN algorithm. After observing all the results it
is evident that the SVM algorithm when used for detecting faces
10 2578 83.90 gives more efficient results than the detection model developed
using the CNN algorithm.
TABLE II. INPUT DATASET SAMPLES WITH VARIOUS SIZES AND RELATED
ACCURACIES FOR SVM MACHINE LEARNING ALGORITHM

S.No Data Set size Accuracy

1 1652 89.11

2 1765 89.71

3 1874 90.46

4 1986 91.74

5 2048 92.49

6 2153 93.98
FIG. 2. “BAR GRAPH ANALYSIS OF NOVEL SVM ALGORITHM AND CNN
7 2249 93.05 ALGORITHM. GRAPHICAL REPRESENTATION SHOWS THE MEAN EFFICIENCY OF
92.65% AND 79.02% FOR THE PROPOSED ALGORITHM (SVM) AND (CNN)
8 2358 94.15 RESPECTIVELY. X-AXIS: SVM VS CNN, Y-AXIS: MEAN ACCURACY ± 2 SD.”

9 2485 95.68
The proposed approach in this particular research intends
10 2578 96.11 to observe that the SVM (Support Vector Machine) algorithm
seems to possess better accuracy than CNN (Convolutional
TABLE III. GROUP STATISTICS OF SVM AND CNN ALGORITHM WITH MEAN Neural Networks) based on the significance value. The mean
VALUE OF 92.65% AND 79.02%. efficiency of SVM algorithm (96.11%) and CNN algorithm
(83.9%) from table 2 and table 3, It is clear that SVM has better
Group N Mean Std. Deviation Std. Error efficiency [18].The countenance detection system was
Mean developed for universities and organizations during this
research, it mainly consists of developing the attendance
SVM 10 92.65 2.407 0.761
management system [19]. This challenge supplied a framework
CNN 10 79.02 3.083 0.975 which attendance control turned into automatic with 89.62%
efficiency. In this research [20], substitute the deep learning
TABLE IV. GROUP STATISTICS OF SVM AND CNN ALGORITHM WITH MEAN knowledge of primarily based visual attendance model in the
VALUE OF 92.65% AND 79.02% WITH STANDARD ERROR MEAN WITH entire method of growing face element through combining
DIFFERENCE modern techniques, it determines the reduced number of face
photos in the entire system, as well as the recommended
approach of augmentation with high accuracy [21]. There are
no such opposite findings with regarding of existing object
detection for predicting community seclusion violations. The
proposed novel SVM model has accuracy of 92.65% and

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1110

Authorized licensed use limited to: Saveetha Engineering College. Downloaded on January 08,2025 at 04:18:38 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)
IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

provides an easy face recognition instruction tool, which is Banu, “A novel face recognition technology to enhance health and safety
measures in hospitals using SBC in pandemic prone areas,” Materials
better than the remaining algorithms because it needs just one Today: Proceedings, vol. 45, pp. 2584–2588, Jan. 2021.
image to figure and not necessary to convert it into grayscale [3] S. J. L. Chenxin, “In-pit coal mine personnel uniqueness detection
image [22]. The above article concludes that the attendance technology based on personnel positioning and face recognition,”
system [23] was made with a combination of many different International Journal of Mining Science and Technology, vol. 23, no. 3,
algorithms and it works with good accuracy and efficiency as pp. 357–361, May 2013.
[4] H. A. M. Bishnu Deo Kumar, “Exam form automation using facial
shown in table 5. recognition,” Materials Today: Proceedings, Jul. 2021, doi:
10.1016/j.matpr.2021.06.190.
TABLE V. INPUT DATASET SAMPLES WITH VARIOUS SIZES AND RELATED [5] S. C. Hoo and H. Ibrahim, “Biometric-Based Attendance Tracking System
ACCURACIES FOR SVM AND CNN MACHINE LEARNING ALGORITHMS for Education Sectors: A Literature Survey on Hardware Requirements,”
Journal of Sensors, vol. 2019, Sep. 2019, doi: 10.1155/2019/7410478.
Iterations SVM CNN [6] S. Arya, H. Mesariya, and V. Parekh, “Smart Attendance System Usign
CNN,” Apr. 22, 2020. Accessed: Jun. 19, 2021. [Online]. Available:
http://arxiv.org/abs/2004.14289
1 89.11 74.96 [7] M. Narendra, G. S. Veena, D. Harekal, and J. S. Kallimani, “Biometric
Based Attendance Management System using Cloud,” International
2 89.71 75.36 Journal of Advanced Science and Technology, vol. 29, no. 3s, pp. 130–
138, Mar. 2020.
3 90.46 76.18 [8] A. A. Raj, A. Arjun Raj, M. Shoheb, K. Arvind, and K. S. Chethan, “Face
Recognition Based Smart Attendance System,” 2020 International
4 91.74 77.36 Conference on Intelligent Engineering and Management (ICIEM). 2020.
doi: 10.1109/iciem48762.2020.9160184.
[9] P. V. Sanivarapu, “Multi-Face Recognition Using CNN for Attendance
5 92.49 78.48 System,” in Machine Learning for Predictive Analysis, Springer,
Singapore, 2021, pp. 313–320.
6 93.98 79.64 [10] Saravanan.M.S, S. S. Teja, "Enhancing the Accuracy of Emoji Conversion
from Text using Novel Lagrangian Support Vector Machine Against Long
7 93.05 80.24 Short-Term Memory Algorithm," 2023 Global Conference on Information
Technologies and Communications (GCITC), Bangalore, India, 2023, pp.
1-5, doi: 10.1109/GCITC60406.2023.10426512.
8 94.15 81.41
[11] P. Neelakantan, C. Subbarao, C. V. Subbarao, G. De-Deus, and M.
Zehnder, “The impact of root dentine conditioning on sealing ability and
9 95.68 82.65 push-out bond strength of an epoxy resin root canal sealer,” Int. Endod. J.,
vol. 44, no. 6, pp. 491–498, Jun. 2011.
10 96.11 83.90 [12] V. Krishnan and T. Lakshmi, “Bioglass: A novel biocompatible
innovation,” J. Adv. Pharm. Technol. Res., vol. 4, no. 2, pp. 78–83, Apr.
2013.
[13] A. Mootha, S. Malaiappan, N. D. Jayakumar, S. S. Varghese, and J. Toby
The system proposed during this research triggers to solve the Thomas, “The Effect of Periodontitis on Expression of Interleukin-21: A
above problems by combining face recognition technologies, Systematic Review,” Int. J. Inflam., vol. 2016, p. 3507503, Feb. 2016.
with excellent results and efficiency. The testing results reveal [14] T. Lakshmi, V. Krishnan, R. Rajendran, and N. Madhusudhanan,
“Azadirachta indica: A herbal panacea in dentistry - An update,”
that the suggested attendance system performed well in the Pharmacogn. Rev., vol. 9, no. 17, pp. 41–44, Jan. 2015.
server when it came to face recognition. Although the results of [15] D. Wang, R. Fu, and Z. Luo, “Classroom Attendance Auto-management
the research are better in both experimental and statistical Based on Deep Learning,” in 2017 2nd International Conference on
analysis, there are certain limitations within the work. Education, Sports, Arts and Management Engineering (ICESAME 2017),
Jun. 2017, pp. 1523–1528.
Algorithm is sensitive to the lightning shadows and scale faces [16] S. A. S. W. Goroa, “An android based course attendance system using face
occurs within the images. The future scope of this research is to recognition,” Journal of King Saud University - Computer and
be further developed to use in other avenues or other sensitive Information Sciences, vol. 33, no. 3, pp. 304–312, Mar. 2021.
places. [17] Khaled Mohammeda, A.S.Tolbaa, Mohammed Elmogy, “Multimodal
student attendance management system (MSAMS),” Ain Shams
Engineering Journal, vol. 9, no. 4, pp. 2917–2929, Dec. 2018.
V. CONCLUSION [18] P. D. D. M. Chandrakala, “Two-stage classifier for face recognition using
HOG features,” Materials Today: Proceedings, Apr. 2021, doi:
The proposed system successfully achieves the goal of 10.1016/j.matpr.2021.04.114.
[19] M. S. Saravanan, D. Suresh, I. Sudha, T. J. Nandhini, S. Sivashankar and
high precision while maintaining low computational K. Kalyani, "Quality Enhancement for Uninvited Content of Social Media
complexity. This method is cost-effective and minimizes Using Support Vector Machine and Alexnet," 2023 International
manual labor. For face recognition, two algorithms were Conference on Data Science, Agents & Artificial Intelligence (ICDSAAI),
employed: convolutional neural networks (CNNs) and support Chennai, India, 2023, pp. 1-5, doi:
10.1109/ICDSAAI59313.2023.10452586.
vector machines (SVMs). Among these, the SVM algorithm [20] Murat Taskirana, Nihan Kahramana, Cigdem Eroglu Erdemb, “Face
demonstrated the highest accuracy, reaching 96.11%. In recognition: Past, present and future (a review),” Digit. Signal Process.,
contrast, convolutional neural networks, although characterized vol. 106, p. 102809, Nov. 2020.
by low computational complexity, were found to be less [21] Salah Elaskari, Muhammad Imran, Abdurrazag Elaskric, Abdullah
Almasoudi, “Using Barcode to Track Student Attendance and Assets in
efficient. Higher Education Institutions,” Procedia Comput. Sci., vol. 184, pp. 226–
233, Jan. 2021.
REFERENCES [22] M. D. Iffandi, R. N. Adiningrat, J. R. Pandapota, J. F. Ramadhan, B.
Kanigoro, and E. Irwansyah, “Attendance System with Face Recognition,”
[1] M. K. Nancy Singla, “Automated latent fingerprint identification system: in Software Engineering Perspectives in Intelligent Systems, Oct. 2020,
A review,” Forensic Sci. Int., vol. 309, p. 110187, Apr. 2020. pp. 751–757.
[2] S. Sulochanan Karthick Ramanathan, Rani Fathima Kamal Basha, Asan [23] S. Reddy Boyapally, “Facial Recognition and Attendance System Using

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1111

Authorized licensed use limited to: Saveetha Engineering College. Downloaded on January 08,2025 at 04:18:38 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Second International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI 2024)
IEEE Xplore Part Number : CFP24VG0-ART ; ISBN : 979-8-3315-4066-1

Dlib and Face_Recognition Libraries,” Mar. 2021, Accessed: Apr. 26,

2021. [Online]. Available: https://papers.ssrn.com/abstract=3804334

979-8-3315-4066-1/24/$31.00 ©2024 IEEE 1112

Authorized licensed use limited to: Saveetha Engineering College. Downloaded on January 08,2025 at 04:18:38 UTC from IEEE Xplore. Restrictions apply.